温室气体监测与AI模型应用（CO₂、CH₄浓度预测）

No abstract available

CO2 is one of the most important greenhouse gases. Its concentration and distribution in the atmosphere have always been important in studying the carbon cycle and the greenhouse effect. This study is the first to validate the XCO2 of satellite observations with total carbon column observing network (TCCON) data and to compare the global XCO2 distribution for the passive satellites Orbiting Carbon Observatory-2 (OCO-2) and Greenhouse Gases Observing Satellite (GOSAT), which are on-orbit greenhouse gas satellites. Results show that since GOSAT was launched in 2009, its mean measurement accuracy was −0.4107 ppm with an error standard deviation of 2.216 ppm since 2009, and has since decreased to −0.62 ppm with an error standard deviation of 2.3 ppm during the past two more years (2014–2016), while the mean measurement accuracy of the OCO-2 was 0.2671 ppm with an error standard deviation of 1.56 ppm from September 2014 to December 2016. GOSAT observations have recently decreased and lagged behind OCO-2 on the ability to monitor the global distribution and monthly detection of XCO2. Furthermore, the XCO2 values gathered by OCO-2 are higher by an average of 1.765 ppm than those by GOSAT. Comparison of the latitude gradient characteristics, seasonal fluctuation amplitude, and annual growth trend of the monthly mean XCO2 distribution also showed differences in values but similar line shapes between OCO-2 and GOSAT. When compared with the NOAA statistics, both satellites’ measurements reflect the growth trend of the global XCO2 at a low and smooth level, and reflect the seasonal fluctuation with an absolutely different line shape.

A deep learning network is introduced to predict concentrations of gases in the underground coal mine enclosed region using various IoT-enabled gas sensors installed in a metallic gas chamber. The air is sucked automatically at specific intervals from the sealed-off site utilizing a solenoid valve, suction pump, and programmed microprocessor. The gas sensors monitor the gas content in the underground coal mine and communicate gas concentration to the surface server room through a wireless network and cloud storage media. The t-SNE_VAE_bi-LSTM model is proposed in this study as a prediction model that combines the t-SNE, VAE, and bi-LSTM networks. The proposed model's t-SNE method aims to minimize the dimensionality of the recorded gas concentration; and VAE layer intends to retrieve the inner characteristics of low-dimensional gas concentration. Finally, the given model's Bi-LSTM layer tries to forecast the concentrations of CH4, CO2, CO, O2, and H2 gases. The proposed model's prediction accuracy is compared with the existing two models, namely auto-regressive integrated average moving (ARIMA) and chaos time series (CHAOS). The experiment findings demonstrate that the t-SNE_VAE_bi-LSTM model forecasted mean square error (MSE) is more accurate, and it has lesser MSE value of 0.029 and 0.069 for CH4; 0.037 and 0.019 for CO2; 0.092 and 0.92 for CO; 1.881 and 1.892 for O2; and 1.235 and 1.200 for H2 than the ARIMA and CHAOS models, respectively.

A multi-gas sensing system was developed based on the detection principle of the non-dispersive infrared (NDIR) method, which used a broad-spectra light source, a tunable Fabry–Pérot (FP) filter detector, and a flexible low-loss infrared waveguide as an absorption cell. CH4, C2H6, and CO2 gases were detected by the system. The concentration of CO2 could be detected directly, and the concentrations of CH4 and C2H6 were detected using a PCA-BP neural network algorithm because of the interference of CH4 and C2H6. The detection limits were achieved to be 2.59 ppm, 926 ppb, and 114 ppb for CH4, C2H6, and CO2 with an averaging time of 429 s, 462 s, and 297 s, respectively. The root mean square error of prediction (RMSEP) of CH4 and C2H6 were 10.97 ppm and 2.00 ppm, respectively. The proposed system and method take full advantage of the multi-component gas measurement capability of the mid-infrared broadband source and achieve a compromise between performance and system cost.

The detection of hazardous gases and the localization of gas leaks heavily rely on electronic nose (E-nose) systems. However, most existing robotic detection and localization approaches rely on a single gas sensor and assume a single-gas leakage scenario, neglecting the potential interference from other gas components in complex environments. Moreover, conventional machine learning algorithms, constrained by the sequential process of data preprocessing, feature extraction, and training-validation, fail to meet the real-time concentration prediction. To address these limitations, this paper proposes a novel multi-task learning convolutional neural network (MTL-CNN) algorithm capable of simultaneous gas classification and concentration prediction. Compared to traditional methods, the proposed MTL-CNN achieves significant improvements in both classification and regression tasks, while enabling real-time prediction by eliminating manual feature extraction. Experiments were conducted using methane (CH4), carbon dioxide (CO2), and their mixtures. The results demonstrate a classification accuracy of 97.7778%, with mean absolute errors (MAE) for concentration regression of 2.2825 ppm (CH4), 26.242 ppm (CO2), and 2.8191 ppm (CH4 in mixtures), respectively.

Very limited studies have forecast the gas concentration and faults of high voltage direct current (HVDC) converter transformers. In this study, a hybrid algorithm consisting seasonal autoregressive integrated moving average-convolutional neural network-gated recurrent unit (SARIMA-CNN-GRU) was proposed to forecast the seven gas concentrations, i.e., hydrogen (H2), methane (CH4), ethane (C2H6), ethylene (C2H4), acetylene (C2H2), carbon-dioxide (CO2), and carbon monoxide (CO). Moreover, extreme gradient boost (XGBoost), adaptive boosting (AdaBoost), and light gradient boosting machine (LightGBM) were utilized to predict the fault inside the HVDC converter transformer. Two types of datasets collected from different sources were used to train the regression and classification models. The performance of the regression model was evaluated by root mean squared error and mean absolute error, whereas the performance of the classification model was evaluated by accuracy, precision, and recall. Finally, a comparative analysis was conducted to showcase the superiority of the proposed methodology.

Carbon dioxide is one of the major component of landfill gas being emitted by sanitary landfills. High concentration of this gas may cause several health condition. It is also one of the greenhouse gas that consistently contributes to climate change. Monitoring and assessing the carbon dioxide concentration in landfills is vital to ensure better living conditions. This study presents the development of carbon dioxide concentration model based on machine learning algorithms. A prototype was developed using Arduino Uno, Wi-Fi module, DHT11 temperature and humidity sensor, MQ4 and MQ135 gas sensors. This prototype was used to gather CO2 and CH4 concentrations, humidity and air temperature of the sanitary landfill. Five machine learning model based on linear regression, support vector machine, regression trees, boosted regression trees and neural network was trained and evaluated. Matlab software was used in this study for the development of each model. The R-square and MSE of each model was calculated and compared which results to an almost identical r-square value of 0.75 and 0.76. An MSE of 6.90857e-05 for the neural network model followed by SVM, Boosted Regression Trees, Regression Trees and Linear Regression with an MSE of 8.8168e-05, 9.0085e-05, 9.4227e-05 and 9.4652e-05 respectively was also obtained. Based on these results, it was concluded that the machine learning model based on neural network is the best algorithm for the carbon dioxide concentration modelling in sanitary landfills since it obtained the lowest MSE among the five models.

No abstract available

No abstract available

This paper takes a fresh look at the influence of CO2 and CH4 for the identification of “El Niño” episodes. We first introduce the statistical methodology for early detection based on the analysis of the historical information Temperature Anomaly produced in the Oceans by the “El Niño” episode. Second, we introduce a statistical analysis to study a possible relation between CO2 and CH4 with the Temperature Anomaly. We apply three Neural Network models and we analyse the correlation between the obtained series and the real one as well as the respective error metrics. The outcome shows that a Convolutional Encoder-Decoder model is the most suitable structure to carry out this purpose because it shows a correlation coefficient of 0.991 and error metrics of MSE = 0.096, RMSE = 0.309, MAE = 0.249. However, the amount of historical information on CH4 becomes a limitation to detect a further relationship.

No abstract available

No abstract available

To achieve the goal of sustainable development, China has put forward a “dual-carbon” strategy aimed at comprehensively reducing carbon consumption and emissions and providing strong support for building an ecological civilization and a green, low-carbon society. Responding to the national green call and practicing the “dual-carbon” strategy, CO2 emissions from vehicle exhausts at urban transportation intersections are being researched. Firstly, a distributed multi-intelligence system model was established for urban road intersections in a city to optimize CO2 capture at urban road intersections by improving the efficiency of pedestrian and vehicle traffic at the intersections. Secondly, the CO2 concentration detection methods were analyzed for the CO2 viscosity, and the CO2 concentration of vehicle exhaust was detected using appropriate methods. Then, a BP neural network prediction model was established by collecting data on CO2 emissions from automobile exhaust at a city intersection in a winter alpine region, and the prediction model was used to accurately predict the CO2 emission of automobile exhaust from an intersection in the city. Finally, a series of CO2 capture strategies were proposed for the CO2 emission from automobile exhaust at an urban intersection. In this study, beneficial experience is provided for studying carbon emission reduction in other similar regions, contributing an important part to realizing Chinese sustainable development goals.

ABSTRACT The sector with the largest share (32%) of total GHG emissions is now electricity and heat production. This means that more fossil fuels have to be burned to meet the ever-increasing demand for electricity and heat. This study aims to estimate the pollutants such as CO2, CH4, N2O, F-gases, and total GHGs produced during electricity generation in Türkiye using LSTM and 1-DCNN. In the training and prediction processes of these algorithms, the 1990-2020 values from the TURKSTAT were used as the data set. Seven of the data were used in the prediction process, while the remaining data were used in the training process of the algorithms. In 1990, a total of 57,543 GWh of electricity was produced in Türkiye, compared to 306,703 GWh in 2020. In 2020, 34.5% of the electricity was produced from coal, 0.1% from liquid fuels, 23.1% from natural gas, 25.5% from hydropower, and 16.8% from renewables and waste. While CO2 emissions comprised 69.05% of Türkiye’s total GHG emissions in 1990, this percentage rose to 78.92% in 2020. It was concluded that the algorithms preferred in this study gave satisfactory results separately for the estimation of GHG emissions released to the atmosphere during electricity production in Türkiye.

Uncertainty is the state of all operation, components, and objective environment that makes impossible to describe the existing state. Forecasting techniques are essential in the field of knowledge development to overcome the uncertainty to increase the efficiency of all systems. In this paper, artificial neural network algorithm is applied to forecast the CO2 concentration in an office building. The algorithm is implemented in Rstudio software using neural net package. The case study of the paper presents two scenarios with different input data to propose the impacts of train data on forecasting algorithms results. The used dataset in the case study is real data that have been monitored for 2 years. The obtained results of algorithms show the predicted values of CO2 concentration in one office for 600 minutes of a working day. The mean percentage error means absolute percentage error, and standard deviation of predicted data for both scenarios are presented in results section.

Atmospheric methane (CH4) concentrations have increased to 2.5 times their pre-industrial levels, with a marked acceleration in recent decades. CH4 is responsible for approximately 30% of the global temperature rise since the Industrial Revolution. This growing concentration contributes to environmental degradation, including ocean acidification, accelerated climate change, and a rise in natural disasters. The column-averaged dry-air mole fraction of methane (XCH4) is a crucial indicator for assessing atmospheric CH4 levels. In this study, the Sentinel-5P TROPOMI instrument was employed to monitor, map, and estimate CH4 concentrations on both regional and global scales. However, TROPOMI data exhibits limitations such as spatial gaps and relatively coarse resolution, particularly at regional scales or over small areas. To mitigate these limitations, a novel Convolutional Neural Network Autoencoder (CNN-AE) model was developed. Validation was performed using the Total Carbon Column Observing Network (TCCON), providing a benchmark for evaluating the accuracy of various interpolation and prediction models. The CNN-AE model demonstrated the highest accuracy in regional-scale analysis, achieving a Mean Absolute Error (MAE) of 28.48 ppb and a Root Mean Square Error (RMSE) of 30.07 ppb. This was followed by the Random Forest (RF) regressor (MAE: 29.07 ppb; RMSE: 36.89 ppb), GridData Nearest Neighbor Interpolator (NNI) (MAE: 30.06 ppb; RMSE: 32.14 ppb), and the Radial Basis Function (RBF) Interpolator (MAE: 80.23 ppb; RMSE: 90.54 ppb). On a global scale, the CNN-AE again outperformed other methods, yielding the lowest MAE and RMSE (19.78 and 24.7 ppb, respectively), followed by RF (21.46 and 27.23 ppb), GridData NNI (25.3 and 32.62 ppb), and RBF (43.08 and 54.93 ppb).

Accurate estimation of anthropogenic CO2 emissions is crucial for effective climate change mitigation policies. This study aims to improve CO2 emission estimates in China using remote sensing measurements of column-averaged dry air mole fractions of CO2 (XCO2) and a neural network approach. We evaluated XCO2 anomalies derived from three background XCO2 concentration approaches: CHN (national median), LAT (10-degree latitudinal median), and NE (N-nearest non-emission grids average). We then applied the Generalized Regression Neural Network model, combined with a partition modeling strategy using the K-means clustering algorithm, to estimate CO2 emissions based on XCO2 anomalies, net primary productivity, and population data. The results indicate that the NE method either outperformed or was at least comparable to the LAT method, while the CHN method performed the worst. The partition modeling strategy and inclusion of population data effectively improved CO2 emission estimates. Specifically, increasing the number of partitions from 1 to 30 using the NE method resulted in mean absolute error (MAE) values decreasing from 0.254 to 0.122 gC/m2/day, while incorporating population data led to a decrease in MAE values between 0.036 and 0.269 gC/m2/day for different partitions. The present methods and findings offer critical insights for supporting government policy-making and target-setting.

Methane (CH4) is one of the main greenhouse gases, whose retrieval is easily affected by atmospheric water (H2O) and surface albedo. In this letter, based on a radiative transfer model, the Airborne Visible/Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) radiance with different H2O and surface albedo is simulated as training data. Back propagation (BP) feed-forward neural network algorithm in machine learning is used to train the CH4 retrieval model, which is applied to quantify the atmospheric CH4 concentration. This method can effectively decrease the impact of atmospheric H2O and surface albedo on CH4 retrieval. Moreover, this machine learning-based approach separates the processes of model training and prediction. This enables rapid characterization of CH4 emission point sources in images as the matched filter (MF) method, while also obtaining the column-averaged concentration of CH4, similar to the optimal estimation (OE) method. The research results indicate that the mean absolute percentage error (MAPE) of the optimal BP model is as low as 0.33%. If necessary, further increases in training data can improve the resolution and applicability of the model.

This paper aims to study the evolution of CO2 concentrations and emissions on a conventional farm with weaned piglets between 6.9 and 17.0 kg live weight based on setpoint temperature, outdoor temperature, and ventilation flow. The experimental trial was conducted during one transition cycle. Generally, the ventilation flow increased with the reduction in setpoint temperature throughout the cycle, which caused a reduction in CO2 concentration and an increase in emissions. The mean CO2 concentration was 3.12 g m–3. Emissions of CO2 had a mean value of 2.21 mg s−1 per animal, which is equivalent to 0.195 mg s−1 kg−1. A potential function was used to describe the interaction between 10 min values of ventilation flow and CO2 concentrations, whereas a linear function was used to describe the interaction between 10 min values of ventilation flow and CO2 emissions, with r values of 0.82 and 0.85, respectively. Using such equations allowed for simple and direct quantification of emissions. Furthermore, two prediction models for CO2 emissions were developed using two neural networks (for 10 min and 60 min predictions), which reached r values of 0.63 and 0.56. These results are limited mainly by the size of the training period, as well as by the differences between the behavior of the series in the training stage and the testing stage.

This study presents a hybrid neural network architecture incorporating LSTM, CNN and MLP layers for predicting CO? concentrations in marine environments. The proposed architecture uses 96 input variables and generates 48 output variables, enabling it to efficiently capture temporal and spatial patterns in the analysed data. Cross-validation with five folds was implemented to assess the generalisability of the previously validated model. Additionally, new time series data from four different locations on three Navy training vessels operating in the same area were incorporated to test the model's performance in real conditions. The statistical techniques applied proved effective and accurate. In training conditions, the model achieved a MAE of 8.5-12.5 ppm and a MAPE of 2-3%, demonstrating its effectiveness in environmental monitoring in training boats.

No abstract available

In this study, four models were developed and assessed, including Autoregressive Integrated Moving Average (ARIMA), Feedforward Neural Network (FNN), Gated Recurrent Unit (GRU), and Long Short-Term Memory (LSTM) to evaluate current status and future forecast of global CO2 concentration. A total of 554 global monthly datasets were employed to train and test the developed models, aimed at estimating future CO₂ concentrations. Then, each developed model was also utilized to estimate CO₂ concentrations for the future 110 months, from March 2025 to February 2035. Among all generated techniques, the LSTM model showed the highest estimation accuracy with an MAPE of 0.05%, an MAE of 0.2028 ppm, and an RMSE of 0.3216 ppm. Whereas GRU and FNN techniques also obtained good results with the same MAPE of 0.05%, their MAE and RMSE values were slightly higher. The four developed models (ARIMA, FNN, GRU, and LSTM) agree on a continuous rise in atmospheric CO2 level within the range between March 2025 and early 2035, and they typically show CO2 concentrations starting from approximately 425 ppm in early 2025 to 442-443 ppm by the end of 2034.

Methane is a potent greenhouse gas, and its accurate detection is critical for addressing global climate change. Although remote sensing has been a crucial technique for understanding the spatial distribution and temporal dynamics of methane emissions, it is still urgently needed to automate the identification of methane emission plumes and effectively deconvolve the signal from background noise. In this study, we propose the methane plume segmentation UNet (MPSUNet) to achieve precise segmentation of methane plumes from remote sensing imagery. The MPSUNet incorporates the pyramid squeeze attention (PSA) module to enhance feature representation and employs a joint loss function combining Dice Loss and Focal Loss to address challenges such as class imbalance and noisy data. A novel dataset, MPDataset, was constructed using Earth surface mineral dust source investigation (EMIT) methane enhancement and RGB radiance data, providing 4172 high-quality samples for model training and evaluation. Our results show that MPSUNet achieves a mean intersection over union (MIoU) of 78.20%, mean precision of 80.78%, recall of 71.11%, and mean pixel accuracy (MPA) of 85.41% on the complete four-channel MPDataset. Compared with seven classical segmentation models, the most improvements of MPSUNet in MIoU, MPrecision, recall, and MPA reach up to 5.33%, 12.28%, 18.04%, and 8.94%, respectively. Notably, the integration of RGB channels enhances the segmentation of small and intricate plume structures. Cross-dataset evaluation using Sentinel-2 data further validates the model’s robustness, achieving an MIoU of 77.65% and an MPA of 83.61%. Generally, the proposed MPSUNet model marks a substantial performance in methane detection, which provides a robust technical framework for global-scale methane emission monitoring as emphasized by global climate agreements.

Methane is a powerful greenhouse gas that contributes significantly to global warming. Accurate detection of methane emissions is the key to taking timely action and minimizing their impact on climate change. We present AttMetNet, a novel attention-enhanced deep learning framework for methane plume detection with Sentinel-2 satellite imagery. The major challenge in developing a methane detection model is to accurately identify methane plumes from Sentinel-2's B11 and B12 bands while suppressing false positives caused by background variability and diverse land cover types. Traditional detection methods typically depend on the differences or ratios between these bands when comparing the scenes with and without plumes. However, these methods often require verification by a domain expert because they generate numerous false positives. Recent deep learning methods make some improvements using CNN-based architectures, but lack mechanisms to prioritize methane-specific features. AttMetNet introduces a methane-aware architecture that fuses the Normalized Difference Methane Index (NDMI) with an attention-enhanced U-Net. By jointly exploiting NDMI's plume-sensitive cues and attention-driven feature selection, AttMetNet selectively amplifies methane absorption features while suppressing background noise. This integration establishes a first-of-its-kind architecture tailored for robust methane plume detection in real satellite imagery. Additionally, we employ focal loss to address the severe class imbalance arising from both limited positive plume samples and sparse plume pixels within imagery. Furthermore, AttMetNet is trained on the real methane plume dataset, making it more robust to practical scenarios. Extensive experiments show that AttMetNet surpasses recent methods in methane plume detection with a lower false positive rate, better precision recall balance, and higher IoU.

Abstract. Efficiently detecting large methane point sources (super-emitters) in oil and gas fields is crucial for informing stakeholder decisions about mitigation actions. Satellite measurements by multispectral instruments, such as Sentinel-2, offer global and frequent coverage. However, methane signals retrieved from satellite multispectral images are prone to surface and atmospheric artifacts that vary spatially and temporally, making it challenging to build a detection algorithm that applies everywhere. Hence, laborious manual inspection is often necessary, hindering widespread deployment of the technology. Here, we propose a novel deep-transfer-learning-based methane plume detection framework. It consists of two components: an adaptive artifact removal algorithm (low-reflectance artifact detection, LRAD) to reduce artifacts in methane retrievals and a deep subdomain adaptation network (DSAN) to detect methane plumes. To train the algorithm, we compile a dataset comprising 1627 Sentinel-2 images from six known methane super-emitters reported in the literature. We evaluate the ability of the algorithm to discover new methane sources with a suite of transfer tasks, in which training and evaluation data come from different regions. Results show that DSAN (average macro F1 score 0.86) outperforms four convolutional neural networks (CNNs), MethaNet (average macro F1 score 0.70), ResNet-50 (average macro F1 score 0.77), VGG16 (average macro F1 score 0.73), and EfficientNet-V2L (average macro F1 score 0.78), in transfer tasks. The transfer learning algorithm overcomes the issue of conventional CNNs, which is their performance degrades substantially in regions outside regions with training data. We apply the algorithm trained with known sources to an unannotated region in the Algerian Hassi Messaoud oil field and reveal 34 anomalous emission events during a 1-year period, which are attributed to three methane super-emitters associated with production and transmission infrastructure. These results demonstrate the potential of our deep-transfer-learning-based method in contributing towards efficient methane super-emitter discovery using Sentinel-2 across different oil and gas fields worldwide.

Satellite‐based detection of methane super‐emitters in oil and gas fields is critical to inform methane mitigation actions. Multispectral satellite instruments such as Sentinel‐2 offer frequent global coverage, making them suitable for monitoring methane super‐emitters worldwide. However, automatically detecting methane emissions from the vast amount of noisy multispectral satellite data remains challenging. Recent studies have shown that deep learning is promising for this task, but it requires a large set of representative training samples, which are still limited. Hyperspectral data, particularly from airborne sources, are relatively mature and have accumulated some data sets, for example, from Carbon Mapper. Here, we develop PlumeBed, which consists of a synthetic image generation module and a domain adversarial neural network (DANN) module. The synthetic image generation module synthesizes training data by combining Carbon Mapper methane plumes and Sentinel‐2 background noises. The DANN module is then trained to detect methane plumes from Sentinel‐2 images. Evaluation against testing data sets compiled from previously reported super‐emitters shows that the PlumeBed detector achieves an average macro‐F1 score of 0.86, outperforming the conventional deep learning frameworks such as ResNet‐50. We further apply PlumeBed to a previously unseen region in the Dauletabad gas field of Turkmenistan. This application unveils 14 methane super‐emitters based on 1‐year of Sentinel‐2 data. Our study demonstrates that utilizing airborne hyperspectral data through transfer learning is promising to efficiently detect methane super‐emitters in the global‐coverage multispectral satellite data.

Methane emissions are a major contributor to global warming, making the detection and monitoring of anthropogenic methane sources, such as oil and gas operations and landfills, a critical task. Remote sensing using hyperspectral sensors has proven to be a valuable tool for this purpose. However, traditional methane detection algorithms based on matched filters often produce spurious results and require laborious postprocessing to accurately identify actual plumes. In this letter, we propose a novel approach for estimating the target signature of a matched filter as a learnable layer integrated into a user-selected segmentation model, trainable in an end-to-end manner. By considering the signature as a parameter of the segmentation model, our methodology enables the signature to adapt both to the task and the user-selected segmentation approach. We use hyperspectral images from the EMIT instrument and their associated publicly available methane plumes to train deep learning segmentation models. Our results show consistent improvements in both convolutional neural networks (CNNs) and transformer architectures, as demonstrated by the $F1$ segmentation score increasing by approximately 18% compared to the baseline values. This allows for more accurate and automated plume segmentation, ultimately aiding in the identification of methane leaks from point source emitters.

The new generation of hyperspectral imagers, such as PRISMA, has improved significantly our detection capability of methane (CH4) plumes from space at high spatial resolution (30m). We present here a complete framework to identify CH4 plumes using images from the PRISMA satellite mission and a deep learning model able to detect plumes over large areas. To compensate for the relative scarcity of PRISMA images, we trained our model by transposing high resolution plumes from Sentinel-2 to PRISMA. Our methodology thus avoids computationally expensive synthetic plume generation from Large Eddy Simulations by generating a broad and realistic training database, and paves the way for large-scale detection of methane plumes using future hyperspectral sensors (EnMAP, EMIT, CarbonMapper).

Methane emissions from oil and gas infrastructure, wetlands, and livestock contribute to the greenhouse gas inventory. The analysis of satellite short-wave infrared imagery offers opportunities for screening large areas to detect methane leaks. Deep learning algorithms excel at analyzing these data, however, they require large annotated datasets for model calibration that are difficult to get. To overcome this limitation, we explore a methodology to spot methane plumes using deep binary classifiers trained on a large dataset of synthetically created methane plumes, customized for this specific task, using publicly available images of the Sentine1-2 satellites. To build the database, we simulate plume patterns using the Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT) and use a simple stochastic model to account for reflectance attenuation due to methane in band 12 centered at 2190 nm. To help distinguish methane plumes from the image background, we compute a methane signature image based on a background subtraction technique. Once calibrated, the classification model is applied to image patches centered in the local minima of the methane signature within the satellite image, scoring a value ranging from 0 to 1 associated with the presence of a methane plume. We compare experimentally the general-purpose ResNet architecture and MethaNet, a domain-specific convolutional neural network, using simulated data. Then, we evaluate the feasibility of our approach in detecting large methane leaks at two study sites located in the Hassi Messaoud oil field in Algeria and the Permian Basin in the US, each covering an area of 0.25$\times$ 0.25 degrees. We found that ResNet is effective in identifying large, known methane plumes that were set aside for testing purposes. This method could be considered as a component of a solution for planning mitigation activities.

Ruminal acidosis is a prevalent metabolic disorder in dairy cattle causing significant economic losses and animal welfare concerns. Current diagnostic methods rely on invasive pH measurement, limiting scalability for continuous monitoring. We present FUME (Fused Unified Multi-gas Emission Network), the first deep learning approach for rumen acidosis detection from dual-gas optical imaging under in vitro conditions. Our method leverages complementary carbon dioxide (CO2) and methane (CH4) emission patterns captured by infrared cameras to classify rumen health into Healthy, Transitional, and Acidotic states. FUME employs a lightweight dual-stream architecture with weight-shared encoders, modality-specific self-attention, and channel attention fusion, jointly optimizing gas plume segmentation and classification of dairy cattle health. We introduce the first dual-gas OGI dataset comprising 8,967 annotated frames across six pH levels with pixel-level segmentation masks. Experiments demonstrate that FUME achieves 80.99% mIoU and 98.82% classification accuracy while using only 1.28M parameters and 1.97G MACs--outperforming state-of-the-art methods in segmentation quality with 10x lower computational cost. Ablation studies reveal that CO2 provides the primary discriminative signal and dual-task learning is essential for optimal performance. Our work establishes the feasibility of gas emission-based livestock health monitoring, paving the way for practical, in vitro acidosis detection systems. Codes are available at https://github.com/taminulislam/fume.

Prioritizing methane for near-term climate action is crucial due to its significant impact on global warming. Previous work used columnwise matched-filter (CMF) products from the airborne AVIRIS-NG imaging spectrometer to detect methane plume sources; convolutional neural networks (CNNs) discerned anthropogenic methane plumes from false positive enhancements. However, as an increasing number of remote sensing platforms are used for methane plume detection, there is a growing need to address cross-platform alignment. In this work, we describe model- and data-driven machine learning approaches that leverage airborne observations to improve spaceborne methane plume detection, reconciling the distributional shifts inherent with performing the same task across platforms. We develop a spaceborne methane plume classifier using data from the EMIT imaging spectroscopy mission. We refine classifiers trained on airborne imagery from AVIRIS-NG campaigns using transfer learning, outperforming the standalone spaceborne model. Finally, we use CycleGAN, an unsupervised image-to-image translation technique, to align the data distributions between airborne and spaceborne contexts. Translating spaceborne EMIT data to the airborne AVIRIS-NG domain using CycleGAN and applying airborne classifiers directly yields the best plume detection results. This methodology is useful not only for data simulation but also for direct data alignment. Though demonstrated on the task of methane plume detection, our work more broadly demonstrates a data-driven approach to align related products obtained from distinct remote sensing instruments.

ABSTRACT Methane is the second most significant greenhouse gas after carbon dioxide. As global warming intensifies, the quantification of methane point sources is becoming increasingly crucial. However, retrieving high-quality methane signals from remote sensing data remains challenging due to various factors, including surface reflectance, atmospheric interference, sensor noise, wind speed, and sensor sensitivity. In real-world scenarios, methane remote sensing quantification often encounters unfavourable conditions that lead to low signal-to-noise ratio (SNR) signals, resulting in reduced quantification accuracy. To address these challenges, we introduce a novel multi-task learning-based approach. Specifically, we incorporate a denoising auxiliary task into the quantification network by introducing an additional denoising branch that recovers clean plume column concentration maps. The network is trained with supervision using both denoising and quantification losses, which enables it to acquire robust feature representations to noise and benefits the quantification of low SNR images. During the inference phase, the denoising branch is removed, resulting in an efficient and robust single quantification network with reduced inferring time, supporting onboard computation. We construct a low SNR database based on the AVIRIS-NG sensor and evaluate the generalization ability of our method. DQNet achieves the highest RMSE, MAPE, and R of 16.478 kg/h, 15.296%, and 95.167% respectively on the synthetic dataset. Across the entire range of SNR, DQNet exhibits a greater relative advantage as SNR decreases. However, our approach is not limited to AVIRIS-NG and can be easily extended to various multispectral and hyperspectral satellites.

No abstract available

No abstract available

Abstract. We present a deep learning model, CH4Net, for automated monitoring of methane super-emitters from Sentinel-2 data. When trained on images of 23 methane super-emitter locations from 2017–2020 and evaluated on images from 2021, this model detects 84 % of methane plumes compared with 24 % of plumes for a state-of-the-art baseline while maintaining a similar false positive rate. We present an in-depth analysis of CH4Net over the complete dataset and at each individual super-emitter site. In addition to the CH4Net model, we compile and make open source a hand-annotated training dataset consisting of 925 methane plume masks as a machine learning baseline to drive further research in this field.

No abstract available

No abstract available

This study investigated the possibility of using a laser methane detector (LMD) and optical gas imaging (OGI) to detect and quantify enteric methane () produced by ruminants in vitro. Four single‐flow continuous fermenters were used for rumen culture incubation with four different treatment diets: Control (50:50 forage to concentrate [F:C] ratio), Control + Bromoform (CBR), Low Forage (LF; 20:80), and High Forage (HF; 80:20). After 10 days of incubation, all fermenter contents were transferred and used in a 24 h ANKOM batch culture to measure gas production with LMD and OGI. The authors introduce the Controlled Diet (CD) dataset, a large‐scale collection of 4,885 plume images captured using an FLIR GF77 OGI camera under varying dietary conditions. The performance of six semantic segmentation models (FCN, U‐Net, Vision Transformer, Swin Transformer, DeepLabv3+, and Gasformer) on the CD dataset is compared. Results showed that LMD data for followed a similar pattern to the gas chromatography (GC) instrument results. The in vitro results showed that different diets and F:C ratios had an impact on gas production and rumen fermentation characteristics. Adding bromoform to the control diet fully inhibited emission. The HF diet produced more compared to all treatments () when measured with GC and LMD. CBR produced the lowest values when measured with GC and LMD. The Gasformer architecture achieved the highest performance with mean IoU of 85.1% and mean F‐score of 91.72%. These findings demonstrate that OGI technology combined with advanced semantic segmentation models offers a promising solution for predicting and quantifying emissions in the livestock sector, potentially aiding in the development of mitigation strategies to combat climate change.

Methane is one of the most potent greenhouse gases, and its short atmospheric half-life makes it a prime target to rapidly curb global warming. However, current methane emission monitoring techniques primarily rely on approximate emission factors or self-reporting, which have been shown to often dramatically underestimate emissions. Although initially designed to monitor surface properties, satellite multispectral data has recently emerged as a powerful method to analyze atmospheric content. However, the spectral resolution of multispectral instruments is poor, and methane measurements are typically very noisy. Methane data products are also sensitive to absorption by the surface and other atmospheric gases (water vapor in particular) and therefore provide noisy maps of potential methane plumes, that typically require extensive human analysis. Here, we show that the image recognition capabilities of deep learning methods can be leveraged to automatize the detection of methane leaks in Sentinel-2 satellite multispectral data, with dramatically reduced false positive rates compared with state-of-the-art multispectral methane data products, and without the need for a priori knowledge of potential leak sites. Our proposed approach paves the way for the automated, high-definition and high-frequency monitoring of point-source methane emissions across the world.

The need for safe and healthy air quality has become critical as urbanization and industrialization increase, leading to health risks and environmental concerns. Gas leaks, particularly of gases like carbon monoxide, methane, and liquefied petroleum gas (LPG), pose significant dangers due to their flammability and toxicity. LPG, widely used in residential and industrial settings, is especially hazardous because it is colorless, odorless, and highly flammable, making undetected leaks an explosion risk. To mitigate these dangers, modern gas detection systems employ sensors, microcontrollers, and real-time monitoring to quickly identify dangerous gas levels. This study introduces an IoT-based system designed for comprehensive environmental monitoring, with a focus on detecting LPG and butane leaks. Using sensors like the MQ6 for gas detection, MQ135 for air quality, and DHT11 for temperature and humidity, the system, managed by an Arduino Mega, collects data and sends these to the ThingSpeak platform for analysis and visualization. In cases of elevated gas levels, it triggers an alarm and notifies the user through IFTTT. Additionally, the system includes a microphone and a CNN model for analyzing audio data, enabling a thorough environmental assessment by identifying specific sounds related to ongoing activities, reaching an accuracy of 96%.

Aerial imagery plays a critical role in identifying oil and gas (O&G) infrastructure for environmental monitoring and methane emission control. This study evaluates ten deep learning models and a multimodal large language model (GPT-4o-mini) for detecting O&G facilities using high-resolution images from the OGNet dataset. InceptionV3 achieved the highest F1 score (0.875), while GPT-4o-mini reached 0.857 through iterative prompt tuning. Though CNNs like ResNet50 and VGG16 were computationally efficient, GPT-4omini offered competitive accuracy with higher inference costs. These findings highlight the promise of combining CNN and vision-LLM approaches to support scalable and sustainable facility detection.

CO2 injection is a promising technology for enhancing gas recovery (CO2-EGR) that concomitantly reduces carbon emissions and aids the energy transition, although it has not yet been applied commercially at the field scale. We develop an innovative workflow using raw data to provide an effective approach in evaluating CH4 recovery during CO2-EGR. A well-calibrated three-dimensional geological model is generated and validated using actual field data—achieving a robust alignment between history and simulation. We visualize the spread of the CO2 plume and quantitatively evaluate the dynamic productivity to the single gas well. We use three deep learning algorithms to predict the time histories of CO2 rate and CH4 recovery and provide feedback on production wells across various injection systems. The results indicate that CO2 injection can enhance CH4 recovery in water-bearing gas reservoirs—CH4 recovery increases with injection rate escalating. Specifically, the increased injection rate diminishes CO2 breakthrough time while concurrently expanding the swept area. The increased injection rate reduces CO2 breakthrough time and increases the swept area. Deep learning algorithms exhibit superior predictive performance, with the gated recurrent unit model being the most reliable and fastest among the three algorithms, particularly when accommodating injection and production time series, as evidenced by its smallest values for evaluation metrics. This study provides an efficient method for predicting the dynamic productivity before and after CO2 injection, which exhibits a speedup that is 3–4 orders of magnitudes higher than traditional numerical simulation. Such models show promise in advancing the practical application of CO2-EGR technology in gas reservoir development.

As global warming intensifies, increased attention is being paid to monitoring fugitive methane emissions and detecting gas plumes from landfills. We have divided methane emission monitoring into three subtasks: methane concentration inversion, plume segmentation, and emission rate estimation. Traditional algorithms face certain limitations: methane concentration inversion typically employs the matched filter, which is sensitive to the global spectrum distribution and prone to significant noise. There is scant research on plume segmentation, with many studies depending on manual segmentation, which can be subjective. The estimation of methane emission rate frequently uses the IME algorithm, which necessitates meteorological measurement data. Utilizing the WENT landfill site in Hong Kong along with PRISMA hyperspectral satellite imagery, we introduce a novel deep learning-based framework for quantitative methane emission monitoring from remote sensing images that is grounded in physical simulation. We create simulated methane plumes using large eddy simulation (LES) and various concentration maps of fugitive emissions using the radiative transfer equation (RTE), while applying augmentation techniques to construct a simulated PRISMA dataset. We train a U-Net network for methane concentration inversion, a Mask R-CNN network for methane plume segmentation, and a ResNet-50 network for methane emission rate estimation. All three deep networks yield higher validation accuracy compared to traditional algorithms. Furthermore, we combine the first two subtasks and the last two subtasks to design multi-task learning models, MTL-01 and MTL-02, both of which outperform single-task models in terms of accuracy. Our research exemplifies the application of multi-task deep learning to quantitative methane monitoring and can be generalized to a wide array of methane monitoring tasks.

Effective cloud and cloud shadow detection is a critical prerequisite for accurate retrieval of concentrations of atmospheric methane (CH4) or other trace gases in hyperspectral remote sensing. This challenge is especially pertinent for MethaneSAT, a satellite mission launched in March 2024, to fill a significant data gap in terms of resolution, precision and swath between coarse-resolution global mappers and fine-scale point-source imagers of methane, and for its airborne companion mission, MethaneAIR. MethaneSAT delivers hyperspectral data at an intermediate spatial resolution (approx. 100 x 400, m), whereas MethaneAIR provides even finer resolution (approx. 25 m), enabling the development of highly detailed maps of concentrations that enable quantification of both the sources and rates of emissions. In this study, we use machine learning methods to address the cloud and cloud shadow detection problem for sensors with these high spatial resolutions. Cloud and cloud shadows in remote sensing data need to be effectively screened out as they bias methane retrievals in remote sensing imagery and impact the quantification of emissions. We deploy and evaluate conventional techniques-including Iterative Logistic Regression (ILR) and Multilayer Perceptron (MLP)-with advanced deep learning architectures, namely U-Net and a Spectral Channel Attention Network (SCAN) method. Our results show that conventional methods struggle with spatial coherence and boundary definition, affecting the detection of clouds and cloud shadows. Deep learning models substantially improve detection quality: U-Net performs best in preserving spatial structure, while SCAN excels at capturing fine boundary details... Our data and code is publicly available at: https://doi.org/10.7910/DVN/IKLZOJ

Seabed gas and oil emissions appear as bubble plumes ascending through the water column in various environments. Understanding bubble characteristics—size, rise speed—is important for estimating escape rates of fluids like methane, oil, and carbon dioxide. However, measuring underwater gas bubbles is challenging, often requiring expensive specialized equipment. This study presents a novel methodology using two calibrated consumer‐grade cameras to estimate bubble size distribution, rise velocities, and corresponding gas or oil flow rates. Our approach, named BURST (Bubble Rise and Size Tracking), uses a trained neural network for automated bubble detection in diverse camera footage, effectively analyzing under varying lighting conditions and visibility, without requiring a uniform backlit background for bubble identification. Post‐detection, bubbles are tracked and synchronized between the cameras, with three‐dimensional triangulation used to deduce sizes and rise speeds, enabling flow rate calculations. We demonstrate the efficacy of our methodology through basin experiments capturing methane bubble plumes with controlled flow rates. Additionally, we successfully apply this methodology to existing footage from natural methane emission sites in the Hopendjupet seeps within the central Barents Sea, measuring methane flow rates of approximately 46 and 24 mmol CH4 min−1 at water depths of 327 and 341 m, respectively. These results underscore the practical applicability of BURST in complex underwater environments without disrupting natural bubble flow. By utilizing readily available equipment, BURST enables reliable bubble measurements in challenging real‐world conditions, including the analysis of legacy footage not initially intended for bubble flow rate quantification. The BURST python script is available at https://github.com/BUbbleRST/BURST/.

No abstract available

Seafloor gas seeps, primarily methane, play a crucial role in the global carbon cycle, influencing ocean chemistry, biodiversity, and geohazards while signaling potential natural gas reserves. Detecting and monitoring these seeps over time is crucial to understanding underwater resources and their environmental impacts. However, current detection methods rely heavily on human interpretation of water column imagery, limiting efficiency and scalability of the detection process. This study proposes a supervised machine learning (ML) framework that mimics human visual interpretation by analyzing sequences of multibeam echosounder (MBES) water column data, enabling automated seep detection by capturing temporal patterns across frames. Instead of processing only single frames, ML-based detection is developed on a sequence of MBES frames at a time. The model is evaluated on MBES sequences of varying lengths and confidence levels and its performance is compared with single-frame-based detection. Although both approaches perform similarly on low-confidence seep sequences, the sequential model demonstrates a clear advantage in high-confidence cases. Sequential analysis reduces the high false positive rate of the single-frame method by improving precision from 66% to 78%, maintains a more balanced recall of 71%, and achieves a higher overall accuracy of 87%. This approach offers a more robust and reliable seafloor gas seep detection system compared to single-frame-based analysis, highlighting the benefits of leveraging temporal context in water column data.

No abstract available

A novel method is introduced to improve the detection performance of photoacoustic spectroscopy for trace gas detection. For effectively suppressing various types of noise, this method integrates photoacoustic spectroscopy with residual networks model which encompasses a total of 40 weighted layers. Firstly, this approach was employed to accurately retrieve methane concentrations at various levels. Secondly, the analysis of the signal-to-noise ratio (SNR) of multiple sets of photoacoustic spectroscopy signals revealed significant enhancement. The SNR was improved from 21 to 805, 52–962, 98–944, 188–933, 310–941, and 587–936 across the different concentrations, respectively, as a result of the application of the residual networks. Finally, further exploration for the measurement precision and stability of photoacoustic spectroscopy system utilizing residual networks was carried out. The measurement precision of 0.0626 ppm was obtained and the minimum detectable limit was found to be 1.47 ppb. Compared to traditional photoacoustic spectroscopy method, an approximately 46-fold improvement in detection limit and 69-fold enhancement in measurement precision were achieved, respectively. This method not only advances the measurement precision and stability of trace gas detection but also highlights the potential of deep learning algorithms in spectroscopy detection.

Safe and economical infrastructure facilitates gas and oil transportation via pipeline networks. Leaks in natural gas pipelines, caused by ageing and other reasons, are common yet difficult to identify. Nevertheless, the most prevalent issue that impacts pipeline operation is leakage failure. To a large extent, the flow properties of the particular pipeline determine the accuracy with which a leaking location may be located. One helpful method for analyzing the many assumptions used to explain the fluid flow process and parameters is numerical modelling based on Artificial Intelligence (AI). Hence, this r4search suggests the Deep Learningassisted Gas Pipeline Leakage Detection System (DLGPLDS) using infrared cameras to accurately detect the leak location. A system of infrared cameras is suggested for use in gas plants, transportation, and manufacturing to track the undetectable leaks of methane gas in real-time. This research recommends using a histogram equalization procedure to remove the background noise effectively. The leaking of invisible gas may be successfully identified by this knowledge-based infrared camera system. This novel approach uses CNNs to categorize infrared camera images into groups based on whether or not they contain natural gas leaks. The research findings demonstrate that the suggested DLGPLDS model increases the gas leakage prediction ratio by 98.7%, classification accuracy ratio by $\mathbf{9 7. 4 \%}$, PSNR ratio by 96.5%, and error rate by 10.3% compared to other state-of-art models.

The detection of methane emissions has drawn significant attention from researchers, as methane is one of the main contributors to global warming. Due to their high accessibility and extensive coverage, satellite-based remote sensing data have become a widely used tool for studying methane emissions. The multispectral satellite Sentinel-2 is particularly prominent for studies related to methane, as it provides shortwave infrared (SWIR) spectral band data that align with the methane absorption line. Previous studies on methane emission detection using Sentinel-2 data rely mainly on synthetic data, which involves simulated methane emission plumes overlaid on real Sentinel-2 backgrounds. However, our analysis reveals that Sentinel-2 images with simulated plumes exhibit different data distributions and visual characteristics from those of real-world data. Moreover, these studies typically use only a single reference image as a clean background along with the plume image, which we find to be insufficient. Finally, the lack of publicly available benchmarks hinders research and development in this domain. To address these challenges, we introduce the MethaneS2CM dataset for methane emission detection, collected from Sentinel-2 multispectral data and the CarbonMapper platform. Our dataset consists of two main products (L1C and L2A) of Sentinel-2, including over 4000 plumes from 43 countries, ranging from 2016 to 2024. The dataset incorporates both long-term and short-term variations in Sentinel-2 images, with each sample consisting of images captured at different time periods. Furthermore, we conduct extensive experiments to evaluate a wide range of existing methane emission detection methods and models based on this dataset. We subsequently propose a deep methane emission classification model, MEECNet. Experimental results demonstrate that MEECNet outperforms all baseline methods. Our dataset is publicly available at https://huggingface.co/datasets/H1deaki/MethaneS2CM.

Abstract. Reducing methane emissions from human activities is essential to tackle climate change. To monitor these emissions, we rely on satellite observations, which enable regular, global-scale tracking. Methane emissions are typically quantified by their source rate – the mass of gas emitted per unit of time. Our goal here is to estimate the emission source rate of methane plumes detected by hyperspectral imagers such as PRISMA or EnMAP. For this task, we generated a large synthetic dataset using large eddy simulation (LES) to train a deep learning model. This dataset was specifically designed to avoid network overfitting with careful plume temporal sampling and plume scaling. Our deep learning network, MetFluxNet, does not require any wind information or a plume mask. Moreover, it accurately predicts the source rate even in the presence of false positives. MetFluxNet performs well on our dataset with a mean absolute percentage error (MAPE) of 8.3 % over a wide range of source rates from 500 to 25 000 kg h−1. Notably, it remains effective at lower source rates, where background noise is typically high. To validate its real-world applicability, we tested MetFluxNet on real plumes with known ground-truth fluxes. The predicted source rates fell systematically within the 95 % confidence intervals, demonstrating its reliability for real-world plume estimation. Finally, in a comparison with recent state-of-the-art methods, MetFluxNet outperformed the deep learning-based S2MetNet and the physics-based integrated mass enhancement (IME) method.

Scientists from leading institutions have warned that global warming is the next potential world ending catastrophe, which will result in increasing natural disasters and social disruptions. Methane gas (CH4) is up to 80 times more effective in trapping heat than carbon dioxide, and it is considered to be the largest contributor to global climate change. Methane gas is invisible, odorless and colorless, which makes it impossible to detect, much less combat. Previous machine learning methods require expert intervention and are often largely inaccurate. Hyperspectral mask-RCNN (H-MCRNN) is a recently developed method among the deep learning-based approaches, which allows for accurate automatic detection due to its incorporation of hyperspectral imagery in conjunction with deep learning neural networks. While powerful, the original H-MRCNN model lacks optimization in terms of the hyperparameters, which directly affects the efficiency and accuracy of the model, resulting in multiple unfeasible detections. This project focuses on the optimization of hyperparameters to reach the full potential of the H-MRCNN model to realize true automatic methane detection without expert intervention. Results have shown that higher epoch numbers and lower training rate can produce clearer and more accurate masks, with a minimum of 24 epochs and a training rate of 1e-6. By using the optimized hyperparameters, 87% success rate can be achieved in the detection of methane leaks.

This paper tackles the challenging problem of detecting methane plumes, a potent greenhouse gas, using Sentinel-2 imagery. This contributes to the mitigation of rapid climate change. We propose a novel deep learning solution based on U-Net with a ResNet34 encoder, integrating dual spectral enhancement techniques (Varon ratio and Sanchez regression) to optimise input features for heightened sensitivity. A key achievement is the ability to detect small plumes down to $400 ~\mathrm{m}^{2}$ (i.e., for a single pixel at 20 m resolution), surpassing traditional methods limited to larger plumes. Experiments show our approach achieves a 78.39 % F1-score on the validation set, demonstrating superior performance in sensitivity and precision over existing remote sensing techniques for automated methane monitoring, especially for small plumes.

Abstract. Current methods for detecting atmospheric plumes and inferring point-source rates from high-resolution satellite imagery are labor-intensive and not scalable with regard to the growing satellite dataset available for methane point sources. Here, we present a two-step algorithm called U-Plume for automated detection and quantification of point sources from satellite imagery. The first step delivers plume detection and delineation (masking) with a U-Net machine learning architecture for image segmentation. The second step quantifies the point-source rate from the masked plume using wind speed information and either a convolutional neural network (CNN) or a physics-based integrated mass enhancement (IME) method. The algorithm can process 62 images (each measuring 128 pixels × 128 pixels) per second on a single 2.6 GHz Intel Core i7-9750H CPU. We train the algorithm using large-eddy simulations of methane plumes superimposed on noisy and variable methane background scenes from the GHGSat-C1 satellite instrument. We introduce the concept of point-source observability, Ops=Q/(UWΔB), as a single dimensionless number to predict plume detectability and source rate quantification error from an instrument as a function of source rate Q, wind speed U, instrument pixel size W, and instrument-dependent background noise ΔB. We show that Ops can powerfully diagnose the ability of an imaging instrument to observe point sources of a certain magnitude under given conditions. U-Plume successfully detects and masks plumes from sources as small as 100 kg h−1 in GHGSat-C1 images over surfaces with low background noise and successfully handles larger point sources over surfaces with substantial background noise. We find that the IME method for source quantification is unbiased over the full range of source rates, while the CNN method is biased towards the mean of its training range. The total error in source rate quantification is dominated by wind speed at low wind speeds and by the masking algorithm at high wind speeds. A wind speed of 2–4 m s−1 is optimal for detection and quantification of point sources from satellite data.

Operational deployment of a fully automated greenhouse gas (GHG) plume detection system remains an elusive goal for imaging spectroscopy missions, despite recent advances in deep learning approaches. With the dramatic increase in data availability, however, automation continues to increase in importance for natural and anthropogenic emissions monitoring. This work reviews and addresses several key obstacles in the field: data and label quality control, prevention of spatiotemporal biases, and correctly aligned modeling objectives. We demonstrate through rigorous experiments using multicampaign data from airborne and spaceborne instruments that convolutional neural networks (CNNs) are able to achieve operational detection performance when these obstacles are alleviated. We demonstrate that a multitask model that learns both instance detection and pixelwise segmentation simultaneously can successfully lead towards an operational pathway. We evaluate the model's plume detectability across emission source types and regions, identifying thresholds for operational deployment. Finally, we provide analysis-ready data, models, and source code for reproducibility, and work to define a set of best practices and validation standards to facilitate future contributions to the field.

Real-time identification and quantification of greenhouse-gas emissions under transient atmospheric conditions is a critical challenge in environmental monitoring. We introduce a spatio-temporal inversion framework that embeds a deep-learning surrogate of computational fluid dynamics (CFD) within a sequential Monte Carlo algorithm to perform Bayesian inference of both emission rate and source location in dynamic flow fields. By substituting costly numerical solvers with a multilayer perceptron trained on high-fidelity CFD outputs, our surrogate captures spatial heterogeneity and temporal evolution of gas dispersion, while delivering near-real-time predictions. Validation on the Chilbolton methane release dataset demonstrates comparable accuracy to full CFD solvers and Gaussian plume models, yet achieves orders-of-magnitude faster runtimes. Further experiments under simulated obstructed-flow scenarios confirm robustness in complex environments. This work reconciles physical fidelity with computational feasibility, offering a scalable solution for industrial emissions monitoring and other time-sensitive spatio-temporal inversion tasks in environmental and scientific modeling.

The rapid proliferation of industrial activities and urban development has markedly intensified anthropogenic climate change, underscoring the critical need for effective mitigation strategies targeting greenhouse gas emissions. While most existing carbon emission forecasts focus on quarterly or annual scales, accurate daily-level prediction is essential. This study utilizes daily atmospheric CO concentration data from January 25, 2015, to February 13, 2025, to forecast future CO levels. Two machine learning-based time series modelsProphet and Long Short-Term Memory (LSTM)are employed to perform daily-scale carbon emission forecasting. Descriptive statistical analysis reveals a clear upward trend and seasonal fluctuations in the data. Results indicate that while the Prophet model performs well on the training set, it exhibits limited generalization ability on the test set. In contrast, the LSTM model performs better on the test set, showing stronger adaptability to unseen data. This study provides insights into CO2 concentration trends, aiding climate change assessment, informing emission mitigation strategies, and guiding the selection of time series forecasting models.

This study explores the application of machine learning methods, specifically XGBoost and Random Forest models, for predicting carbon dioxide (CO₂) emissions from gas turbine equipment. Accurate forecasting of harmful emissions is a critical task in the context of ensuring energy security and achieving environmental sustainability. The research utilizes a real-world dataset collected in Turkey between 2011 and 2015, containing over 36,000 hourly monitoring records of technical and atmospheric parameters from a gas turbine installation.To enhance model performance, data preprocessing was conducted, including cleaning, feature correlation analysis, and standardization of numerical values. Based on the selected parameters, two regression models were developed: XGBRegressor and RandomForestRegressor, both demonstrating high accuracy in evaluation. The highest coefficient of determination (R²) achieved was 0.742 for the Random Forest model, indicating its effectiveness in capturing the relationships between technical parameters and CO₂ emission levels. Feature importance analysis revealed that turbine temperature-related parameters had the greatest influence on emission levels, whereas external atmospheric conditions played a secondary role. The obtained results are practically significant for the energy sector, as they can be used for emission monitoring, planning the modernization of energy facilities, informing environmental policy, and selecting optimal operational modes for equipment. Implementing such models in industrial practice will contribute to reducing greenhouse gas emissions, increasing energy efficiency, and supporting the realization of national sustainable development strategies. The study confirms that the use of modern machine learning-based analytical tools can make a substantial contribution to ensuring a country's environmental and energy security.

Accurate forecasting of carbon dioxide (CO2) emissions is crucial for developing effective environmental policies and mitigating climate change. In this study, we apply machine learning models, including Random Forest, XGBoost, LightGBM, and CatBoost, to predict CO2 emissions based on a dataset covering 107 countries from 2000 to 2020. We investigate the influence of key economic, social, environmental, and energy-related factors on CO2 emissions and assess the predictive performance of each model. To enhance interpretability, we employ feature importance analysis to identify the most significant drivers of CO2 emissions. By leveraging Permutation Importance, we quantify the contribution of various features across different models. Our methodology integrates a time-window-based feature engineering approach, allowing us to capture temporal patterns in CO2 emissions trends. Experimental results show that CatBoost delivers the highest overall predictive performance, benefiting from its Ordered Boosting mechanism and superior handling of categorical data. LightGBM and XGBoost also achieve strong results, with XGBoost demonstrating notable advantages in controlling prediction bias. The feature importance analysis highlights the dominant role of energy-related factors, particularly electricity consumption from fossil fuels and renewables, in shaping CO2 emissions. Additionally, social and economic indicators, such as land area and GDP growth, exhibit varying levels of impact across models. This study underscores the efficacy of machine learning techniques in CO2 emissions forecasting and provides valuable insights into the underlying drivers of emissions. The findings contribute to advancing data-driven environmental policy-making.

No abstract available

Air quality monitoring and forecasting have become increasingly critical in urban environments due to rising pollution levels and their impact on public health. Recent advances in Internet of Things (IoT) technology and machine learning offer promising alternatives to traditional monitoring stations, which are limited by high costs and sparse deployment. This paper presents the development of a real-time, low-cost air quality forecasting system that integrates IoT-based sensing units with predictive machine learning algorithms. The proposed system employs low-cost gas sensors and microcontroller-based hardware to monitor pollutants such as particulate matter, carbon monoxide, carbon dioxide and volatile organic compounds. A fully functional prototype device was designed and manufactured using Fused Deposition Modeling (FDM) with modular and scalable features. The data acquisition pipeline includes on-device adjustment, local smoothing, and cloud transfer for real-time storage and visualization. Advanced feature engineering and a multi-model training strategy were used to generate accurate short-term forecasts. Among the models tested, the GRU-based deep learning model yielded the highest performance, achieving R2 values above 0.93 and maintaining latency below 130 ms, suitable for real-time use. The system also achieved over 91% accuracy in health-based AQI category predictions and demonstrated stable performance without sensor saturation under high-pollution conditions. This study demonstrates that combining embedded hardware, real-time analytics, and ML-driven forecasting enables robust and scalable air quality management solutions, contributing directly to sustainable development goals through enhanced environmental monitoring and public health responsiveness.

With the growing severity of global climate change, forecasting and managing carbon dioxide (CO2) emissions has become one of the critical tasks in addressing climate change. To improve the accuracy of CO2 emission forecasting, an innovative framework based on variational mode decomposition (VMD), improved black-winged kite algorithm (IBKA), and BiLSTM networks is proposed. This framework aims to address the challenges associated with predicting non-stationary data and optimizing model hyperparameters. Initially, experiments were conducted on 29 benchmark functions using the IBKA algorithm, demonstrating its superior performance in highly nonlinear and complex environments. Subsequently, the BiLSTM model optimized by IBKA was employed to predict CO2 emission trends across four major industries in China, confirming its enhanced prediction accuracy. Finally, a comparative analysis with other mainstream machine learning and deep learning models revealed that the BiLSTM model consistently achieved the best predictive performance across all industries. This research proposes an efficient and practical technical pathway for intelligent carbon emission prediction under the “dual-carbon” strategic goals, offering scientific support for policy formulation and the low-carbon transition.

Carbon dioxide (CO2) emissions are among the most significant contributing factors to climate change. The increase in industrial activities, transportation, and forest exploitation are direct causes of this ongoing crisis. Due to its rapid development, Taiwan faces the dual challenge of sustaining economic growth while mitigating environmental impacts. Building on previous studies, this paper presents a novel approach that considers the interconnectedness of global CO2 patterns. This study trains seven machine learning (ML) models using the optimal training ratio to forecast Taiwan’s CO2 trends. The three top-performing models were identified as Gradient Boosting Regressor, FeedForward Neural Network (FFNN), and Random Forest Regressor. After optimization, Gradient Boosting achieved an R² score of 0.997, FFNN scored 0.996, and Random Forest reached 0.995. These models demonstrated high accuracy with no signs of overfitting. This paper aims to provide policymakers in Taiwan with insights to effectively address CO2 emissions.

Soil respiration, the natural process by which carbon dioxide ($\text{CO}_{2}$) is released from the soil into the atmosphere, is the largest flux of $\text{CO}_{2}$ from the Earth's surface. Despite its significance, accurately predicting $\text{CO}_{2}$ emissions caused by soil respiration remains challenging, and the influence of environmental variables are not yet fully understood. Early studies using artificial intelligence (AI) to predict soil respirations often relied on linear models or overly simplified datasets which lacked real-world applications. This study introduces a novel integration of XGBoost and Shapley Additive exPlanations (SHAP) to achieve both high predictive accuracy and detailed feature interpretability. In this study, an innovative ensemble model, XGBoost, is developed to estimate annual soil respiration (Rs) using spatial, temporal, climatic, and other independent variables. Additionally, a deep neural network (DNN) model is explored. Among the machine learning models evaluated, XGBoost shows the best performance, with a 4% reduction in Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) and a 3.7 % increase in $\mathbf{R}^{\mathbf{2}}$ compared with the Random Forest model, which was so far believed to be the state-of-the-art method. To aid in model explainability, SHAP is used to give feature importance at both the global and individual levels of analysis. Results indicate that Mean Annual Temperature (MAT) and Mean Annual Precipitation (MAP) are the most dominant environmental factors, with geographic variables such as latitude and longitude also significant. Other strong predictors include Study Year, Soil Type, soil composition-related features, Elevation, and Basal Area (BA). This study fills the gap between predictive modeling and interpretability, providing an explainable AI framework that is scalable for realworld environmental research.

This research aims to model and predict greenhouse gas (GHG) emissions in Saudi Arabia by examining their association with crucial socio-economic and environmental factors. Utilizing annual data from 1980 to 2023, the study focuses on three emission variables as dependent variables: carbon dioxide (CO₂) emissions from the power sector, methane (CH₄) emissions from the power sector, and nitrous oxide (N₂O) emissions from industrial activities. The independent variables include agricultural land area, urban population, GDP growth, exports, trade openness, foreign direct investment, and manufacturing output. A comparative assessment of various modeling approaches Ordinary Least Squares (OLS), Ridge Regression, Least Absolute Shrinkage and Selection Operator (LASSO), Elastic Net (Enet), Random Forest (RF), and a new hybrid method that merges Elastic Net and Random Forest (ENRF) was performed. The performance of the models was evaluated based on Mean Squared Error (MSE) and Root Mean Squared Error (RMSE). The results indicated that the ENRF model consistently surpassed both traditional and machine learning techniques, achieving the lowest MSE and RMSE values. The outcomes underscore the efficacy of hybrid statistical and machine learning models in reliably predicting emissions and informing environmental policy in complex, big data contexts.

This The tenacity of this study is to improve a Machine Learning (ML) model to enhance the precision of Carbon Dioxide (CO2) emission predictions. This study utilizes the cutting-edge forecasting techniques for a more accurate understanding of environmental impact. This study will harness the power of smart forecasting to inform strategic decision-making in carbon mitigation efforts. In this research, the performance metrics of four commonly used ML classifiers, namely LR, Gaussian Process, MLP and SMOreg have been evaluated to foretell CO2 emissions using the dataset collected from Kaggle. The dataset was pre-processed, and all the algorithms were trained and tested. The number of instances used in this study is 935. The results of this investigation show that machine learning algorithms are capable of producing accurate CO2 emission forecasts. The findings suggest that the SMOreg Classifier is more accurate than the LR (LR), Gaussian Process Regression (GPR) and Multilayer Perceptron (MLP) classifiers for predicting CO2 emissions. This study emphasizes the possibility of using ML algorithms to predict CO2 emissions. The error values such as MSE, RMSE, MAE, Correlation Coefficient and Root relative squared error indicates the performance of SMOreg is a superior classifier for the forecasting, these results have an important implication in climate change for improving prediction models, which could assist in early detection of climate change.

Climate change is a critical problem that causes global environmental and social issues due to increased greenhouse gas emissions caused by human activities. Carbon dioxide (CO₂) emissions, in particular, are one of the main elements of global warming and have devastating effects on ecosystems. Keeping track of carbon dioxide emissions resulting from human activities like burning fossil fuels, clearing forests, and farming, as well as forecasting their future patterns, is essential for creating effective sustainable environmental strategies. The study utilized machine-learning models to evaluate CO₂ emissions per individual, using the dataset from the Global Carbon Atlas that has released in 2023. In the study, the traditional ARIMA model and deep learning-based LSTM networks were comparatively discussed. The models were trained with the aim of predicting Türkiye's future CO₂ emission levels by learning from past data, and their performances were evaluated with MAE, MSE, RMSE, and R² metrics. The LSTM model achieved an R² score of 90.4%, while the ARIMA model achieved an R² score of 94.3%. The findings show that machine learning techniques are a powerful tool in the fight against climate change and provide valuable insights for policymakers. The findings of the study guide more effective monitoring of CO₂ emissions and determination of strategies for sustainable development goals.

In response to escalating climate concerns, precise industrial Carbon Dioxide (CO2) emissions prediction is paramount. Employing advanced Machine Learning (ML) techniques, this study focuses on forecasting industrial CO2 emissions using global data from the Our World In Data Dataset (containing information on annual emissions from cement, coal, flaring, gas, and oil industries). Various regression models including Support Vector Regression (SVR), Linear Regression, and XGBoost were explored, with a primary emphasis on time series forecasting models for yearly CO2 emissions. Leveraging time series forecasting, intricate temporal trends in emissions data are discerned, offering enhanced predictive insights. CO2 prediction literature was reviewed, data collected and preprocessed, and various ML algorithms implemented, followed by hyperparameter tuning. The models, rigorously trained and evaluated, yield accurate emission predictions. Results highlight the superior performances of the Transformer model and the Neural Prophet Library developed by Stanford University in collaboration with Facebook Inc., with RMSE scores of 416.58 and 470.30, impressively low MAPE scores of both 0.01, and relatively lower MAE of 349.07 and 380.40 compared to other tested models. DeepTCN also demonstrates competitive predictive capabilities but falls short of Transformer model and Neural Prophet model accuracy. Traditional models including ARIMA, Naive Forecasting, Auto Regression (AR), Exponential Smoothing, and SARIMA lag in performance compared to both Neural Prophet and Transformer. These findings underscore the promising role of ML in advancing sustainable environmental management and pave the way for subsequent research endeavors.

This paper presents a novel two-part pipeline for monitoring progress towards the UN Sustainable Development Goals (SDG's) related to Climate Action and Sustainable Cities and Communities. The pipeline consists of two main parts: the first part takes a raw satellite image of a motorway section and produces traffic count predictions for count sites within the image; the second part takes these predicted traffic counts and other variables to produce estimates of Local Authority (LA) motorway Average Annual Daily Traffic (AADT) and Greenhouse Gas (GHG) emissions on a per vehicle type basis. We also provide flexibility to the pipeline by implementing a novel method for estimating emissions when data on AADT per vehicle type or/and live vehicle speeds are not available. Finally, we extend the pipeline to also estimate LA A-Roads and minor roads AADT and GHG emissions. We treat the 2017 year as training and 2018 as the test year. Results show that it is possible to predict AADT and GHG emissions from satellite imagery, with motorway test year $R^2$ values of 0.92 and 0.78 respectively, and for A-roads' $R^2$ values of 0.94 and 0.98. This end-to-end two-part pipeline builds upon and combines previous research in road transportation traffic flows, speed estimation from satellite imagery, and emissions estimation, providing new contributions and insights into these areas.

Climate change poses significant challenges with potentially far-reaching consequences. Accurate prediction of climate change patterns is crucial for informed decision-making and proactive mitigation. This research presents a novel approach to climate change prediction using deep learning techniques. The study integrates deep learning algorithms with extensive climate datasets, including satellite imagery, historical weather records, oceanic data, and greenhouse gas concentrations, to model and forecast variables such as temperature, precipitation, and sea levels. The methodology leverages convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to capture complex spatial and temporal patterns. The deep learning models are trained on pre-processed datasets, employing advanced optimization techniques, transfer learning, and ensembling methods to enhance accuracy and robustness. The research also investigates model interpretability to gain insights into the driving factors behind climate change, aiding policymakers and scientists in developing targeted strategies. Results show that deep learning models outperform traditional statistical methods, demonstrating strong generalization across geographic regions and effectively capturing long-term climate trends and abrupt changes. The findings hold significant implications for climate science and policy formulation, offering new opportunities for informed decision-making and effective mitigation efforts. As the world faces urgent climate challenges, this study provides a valuable contribution to the scientific community and policymakers.

Onshore seeps are recognized as strong sources of methane (CH4), the second most important greenhouse gas. Seeps actively emitting CH4 were recently found in floodplains of West Siberian rivers. Despite the origin of CH4 in these seeps is not fully understood, they can make substantial contribution in regional greenhouse gas emission. We used high-resolution satellite Sentinel-2 imagery to estimate seep areas at a regional scale. Convolutional neural network based on U-Net architecture was implemented to overcome difficulties with seep recognition. Ground-based field investigations and unmanned aerial vehicle footage were coupled to provide reliable training dataset. The seep areas were estimated at 2885 km2 or 1.5% of the studied region; most seep areas were found within the Ob’ river floodplain. The overall accuracy of the final map reached 86.1%. Our study demonstrates that seeps are widespread throughout the region and provides a basis to estimate seep CH4 flux in entire Western Siberia.

Abstract Measuring and attributing greenhouse gas (GHG) emissions remains a challenging problem as the world strives toward meeting emissions reductions targets. As a significant portion of total global emissions, the road transportation sector represents an enormous challenge for estimating and tracking emissions at a global scale. To meet this challenge, we have developed a hybrid approach for estimating road transportation emissions that combines the strengths of machine learning and satellite imagery with localized emissions factors data to create an accurate, globally scalable, and easily configurable GHG monitoring framework.

Greenhouse gas emissions heavily contribute to global warming, causing devastating impacts on society. Climate change has increased the number and size of wildfires along with extending wildfire seasons, which only further damages the environment. A study of the 196 countries in the Paris Agreement found most underreporting their emissions; for example, Malaysia emitted 422 million tons of emissions in 2016 but only reported 81 million. Since tracking is done at a country level, each country uses different tracking methods and can falsify its data. With accurate and unbiased emissions tracking, setting climate goals and determining the effectiveness of solutions are possible.There is a need for an unbiased, quantitative, real-time, and consistent compliance solution for emissions tracking. 30% of emissions come from facilities, mostly coal power plants emitting smoke plumes. Satellite imagery covers all facilities in the world at high frequency. Meanwhile, the US Environmental Protection Agency (EPA) accurately gathers emissions data from over 8,000 US facilities. By taking facilities from this data, collecting satellite imagery of the facilities using SentinelHub, segmenting the smoke plumes from the images using my own YoloV8 segmentation model, and extracting features from the smoke plumes, I have created a deep-learning regression model correlating the smoke plumes to the actual emissions, achieving an 8% Mean Absolute Percentage Error (MAPE) on my testing dataset, and proving the feasibility of accurately tracking facility emissions worldwide in real-time using satellite imagery and deep-learning AI to keep countries accountable.The goal of my project is to accurately estimate greenhouse gas emissions from different facilities (power plants, waste facilities, etc.) using high frequency, publicly available 12-channel satellite imagery, which enables unbiased, accurate tracking of emissions worldwide without relying on local agencies.

Climate change triggered by greenhouse gas (GHG) emissions from the agricultural sector is a critical challenge in sustainable development. This study aims to develop a spatial prediction and classification system of GHG emissions in tropical agricultural land based on artificial intelligence to support low-carbon agricultural policies in Indonesia. The system is built through the integration of multisource data, including field measurements, environmental sensors, external sources, drone imagery, and spatial data. The methodology used includes data preprocessing, dimension reduction using Principal Component Analysis, spatial grouping with K-Means, and emission classification through three machine learning models: Logistic Regression, Support Vector Machine, and Naive Bayes. The results showed that Logistic Regression resulted in the highest accuracy in emission classification $(99.8 \%)$, with superior performance in detecting high emission categories. Environmental variables such as humidity, soil temperature, and pH play a significant role in predicting emissions. This research makes a significant contribution to the artificial intelligence literature for predicting tropical agricultural emissions and strengthens the scientific foundation in the development of data-driven sustainable food systems. These findings open up opportunities for further integration in the national emissions mitigation framework through a digital technology approach.

Atmospheric methane, a powerful greenhouse gas (GHG), plays a significant role in accelerating global warming. This study leverages machine learning to analyze methane emissions and forecast their patterns using comprehensive datasets, including TROPOMI Sentinel-5P satellite methane data, MIDAS UK soil temperature records, and the NERC EDS 2021 land cover dataset. By constructing predictive models through a machine learning pipeline, this research identifies key drivers of methane emissions and addresses data gaps caused by incomplete satellite measurements. Experimental results demonstrate the superior performance of the Random Forest model compared to other machine learning models, achieving the highest accuracy with an RMSE of 29.49 and an MAE of 20.84 for methane (CH4) prediction. Shapley analysis is used to enhance model explainability by evaluating how different attributes (e.g., time, land usage, soil temperature) influence methane production. This study highlights a notable increase in methane levels over recent years, and underscores the critical role of sustainable land management, particularly agricultural practices, in strategies to reduce methane emissions and combat global warming. This work is part of a broader initiative to develop data-driven, AI-powered Digital Twins for analyzing the interplay between human activities and natural processes in advancing climate action.

The stability of the global climate is seriously threatened by the growing buildup of Greenhouse Gases (GHGs), especially carbon dioxide (CO₂) and methane (CH₄). Because it is colorless and odorless, methane is particularly hard to detect, even though it has a global warming potential that is more than 30 times that of CO₂ over a 100-year period. Traditional detection techniques, like satellite remote sensing, point sensors, and manual inspection, frequently have poor sensitivity, high latency, and limited spatial resolution when it comes to identifying diffuse or low-concentration leaks. This study suggests an AI-powered, multimodal detection framework that combines real-time data processing, nanophotonic sensor improvements, and advanced infrared (IR) spectroscopy in order to overcome these constraints. To increase sensitivity, selectivity, and energy efficiency, the system makes use of the near-infrared (NIR), shortwave infrared (SWIR), and Longwave Infrared (LWIR) bands, which are supported by materials like Metal-Organic Frameworks (MOFs), Carbon Nanotubes (CNTs), and quantum dots. For low-latency performance, embedded FPGA hardware is used for signal processing. Principal Component Analysis (PCA) in conjunction with deep neural networks facilitates the quantification of CO₂ from spectral data, while AI models such as Vision Transformers (e.g., Gasformer) and temporal networks (e.g., GLRNet) are used to segment methane plumes in thermal imagery. Further advancements in system miniaturization and deployability include CMOS-compatible imagers and micro-LED-based illumination sources, enabling compact, low-power, and scalable integration for real-world applications. By combining interpretable, AI-driven analytics with nanophotonic sensor design, this work offers a unified and scalable solution for future GHG monitoring.

This study investigated the impact of grassland and cropland expansion on carbon (C) and nitrous oxide (N2O) emissions using remote sensing data and machine learning models. The research focused on agricultural land-use changes in South Sumatra from 1992 to 2018, utilizing Landsat satellite imagery and Google Earth Engine (GEE) for spatial and temporal analysis. Machine learning algorithms, including gradient boosting trees (GBT), random forest (RF), support vector machines (SVM), and classification and regression trees (CART), were employed to estimate greenhouse gas emissions based on multiple environmental parameters. These parameters include enhanced vegetation index (EVI), land surface temperature (LST), normalized difference vegetation index (NDVI), albedo, elevation, humidity, population density, precipitation, soil moisture, and wind speed. The results revealed a strong correlation between agricultural expansion and increased C and N2O emissions, with RF and GBT models demonstrating superior predictive accuracy. Specifically, GBT and RF achieved the highest R2 value (0.71, 0.59) and the lowest error metrics in modeling emissions, whereas SVM performed poorly across all cases. The study highlights the effectiveness of machine learning in quantifying emission dynamics and underscores the necessity of sustainable land management strategies to mitigate greenhouse gas emissions. By integrating remote sensing and data-driven methodologies, this research contributes to climate change mitigation policies and precision agriculture strategies aimed at balancing food security and environmental sustainability.

Air pollution represents a pivotal environmental challenge globally, playing a major role in climate change via greenhouse gas emissions and negatively affecting the health of billions. However predicting the spatial and temporal patterns of pollutants remains challenging. The scarcity of ground-based monitoring facilities and the dependency of air pollution modeling on comprehensive datasets, often inaccessible for numerous areas, complicate this issue. In this work, we introduce GeoFormer, a compact model that combines a vision transformer module with a highly efficient time-series transformer module to predict surface-level nitrogen dioxide (NO2) concentrations from Sentinel-5P satellite imagery. We train the proposed model to predict surface-level NO2 measurements using a dataset we constructed with Sentinel-5P images of ground-level monitoring stations, and their corresponding NO2 concentration readings. The proposed model attains high accuracy (MAE 5.65), demonstrating the efficacy of combining vision and time-series transformer architectures to harness satellite-derived data for enhanced GHG emission insights, proving instrumental in advancing climate change monitoring and emission regulation efforts globally.

Fossil fuel combustion produces large quantities of carbon dioxide (CO2), a major greenhouse gas (GHG), which is one of the main drivers of climate change. A quantitative assessment of GHG emissions is fundamental to predicting climate change effects, enforcing emission regulations, and monitoring pollution trading schemes. Unfortunately, the reporting of GHG emissions is only required in some countries, resulting in insufficient global coverage. At the same time, the transition from fossil fuels to zero carbon to limit climate change is at the heart of several ecological movements, hence the need for quantifying energy production, as well. In this work, we propose an end-to-end method to estimate power generation rates for fossil fuel power plants from satellite images, based on which we approximate GHG (CO2) emission rates. We present a physics-guided multitask deep-learning approach able to simultaneously predict from a single-satellite image of a power plant: 1) the pixel-area covered by plumes; 2) the type of fired fuel; and 3) the power generation rate. To ensure physically realistic predictions from our model we account for environmental conditions and empirical physical constraints. We then convert the predicted power generation rate into estimates for the rate at which CO2 is being emitted, using a fuel-dependent conversion factor. Experimental results show that our multitask learning approach improves the power generation estimation mean absolute error (MAE) by 23% compared to a single-task network trained on the same dataset.

This research introduces a novel deep learning framework that integrates a Convolutional Neural Network with Long Short-Term Memory (CNN-LSTM) for predicting gas concentration from Photoacoustic Spectroscopy (PAS) signals, and it evaluates its performance against the Partial Least Squares (PLS) method. PAS signals are highly sensitive to noise, especially in low-concentration scenarios. The CNN-LSTM architecture extracts local characteristics via convolutional layers and then captures the temporal dependencies by leveraging Long Short-Term Memory (LSTM) layers, enhancing feature learning. In contrast, PLS relies on dimensionality reduction and linear regression. Experimental results show that the CNN-LSTM model achieves more accurate concentration prediction under complex noise conditions, outperforming PLS in terms of noise resistance and generalization.

Predicting CO2 concentration in post‐combustion carbon capture (PCC) systems is challenging due to complex operating conditions and multivariate interactions. This study proposes an enhanced RIME algorithm (ERIME) optimization‐based convolutional neural network (CNN)‐long short‐term memory (LSTM)‐multi‐head‐attention (ECLMA) model to improve prediction accuracy. The local outlier factor (LOF) algorithm was used to remove noise from the data, while mutual information (MI) determined time lags, and the smoothed clipped absolute deviation (SCAD) method optimized feature selection. The CNN‐LSTM‐multi‐head‐attention model extracts meaningful features from time series data, and parameters are optimized using the ERIME algorithm. Using a simulated dataset from a 600 MW supercritical coal‐fired power plant, the results showed that after LOF outlier removal, root mean square error (RMSE) and mean absolute error (MAE) improved by 10%–13%. Post‐MI delay reconstruction reduced RMSE to 0.00999 and MAE to 11.6937, with R2 rising to 0.9929. After variable selection, RMSE and MAE further reduced to 0.00907 and 9.9697, with R2 increasing to 0.9983. After ERIME optimization, the ECLMA model outperformed traditional models, reducing RMSE and MAE by up to 91.55% and 84.94%, respectively, compared to CNN, and by 85.91% and 69.47%, respectively, compared to LSTM. These results confirm the model's superior accuracy and stability.

As particulate organic carbon (POC) from lakes plays an important role in lake ecosystem sustainability and carbon cycle, the estimation of its concentration using satellite remote sensing is of great interest. However, the high complexity and variability of lake water composition pose major challenges to the estimation algorithm of POC concentration in Class II water. This study aimed to formulate a machine-learning algorithm to predict POC concentration and compare their modeling performance. A Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) algorithm based on spectral and time sequences was proposed to construct an estimation model using the Sentinel 2 satellite images and water surface sample data of Chaohu Lake in China. As a comparison, the performances of the Backpropagation Neural Network (BP), Generalized Regression Neural Network (GRNN), and Convolutional Neural Network (CNN) models were evaluated for remote sensing inversion of POC concentration. The results show that the CNN–LSTM model obtained higher prediction precision than the BP, GRNN, and CNN models, with a coefficient of determination (R2) of 0.88, a root mean square error (RMSE) of 3.66, and residual prediction deviation (RPD) of 3.03, which are 6.02%, 22.13%, and 28.4% better than the CNN model, respectively. This indicates that CNN–LSTM effectively combines spatial and temporal information, quickly captures time-series features, strengthens the learning ability of multi-scale features, is conducive to improving estimation precision of remote sensing models, and offers good support for carbon source monitoring and assessment in lakes.

Estimating the concentration levels of gas components in a mixture is vital for taking preventive actions and ensuring public safety in various environments, including industrial settings, urban areas, and confined spaces where gas leaks or accumulation can pose significant health and safety risks. This is because general gas sensors for target pollutants are easily affected by other gases. In this study, we assessed the effectiveness of combining data segmentation, feature extraction, and normalization on the performance of three deep learning (DL) models: 1-D convolutional neural network (CNN-1D), long short-term memory (LSTM), and gated recurrent unit (GRU) network for predicting gas levels in two combinations: ethylene (Ety)–methane (Met) and Ety-carbon monoxide (CO). We evaluated performance using mean absolute error (MAE), root mean squared error (RMSE), and R-squared (<inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula>). Among the models evaluated, the GRU network outperformed the CNN-1D and LSTM in terms of averaged MAE (0.1578), RMSE (0.0481), and <inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula> (0.96375). Comparative analysis with existing studies confirmed that our proposed GRU model outperformed others so that we propose its use in gas concentration estimation tasks.

Curbing methane emissions is among the most effective actions that can be taken to slow down global warming. However, monitoring emissions remains challenging, as detection methods have a limited quantification completeness due to trade-offs that have to be made between coverage, resolution, and detection accuracy. Here we show that deep learning can overcome the trade-off in terms of spectral resolution that comes with multi-spectral satellite data, resulting in a methane detection tool with global coverage and high temporal and spatial resolution. We compare our detections with airborne methane measurement campaigns, which suggests that our method can detect methane point sources in Sentinel-2 data down to plumes of 0.01 km2, corresponding to 200 to 300 kg CH4 h−1 sources. Our model shows an order of magnitude improvement over the state-of-the-art, providing a significant step towards the automated, high resolution detection of methane emissions at a global scale, every few days. Accurate monitoring of methane emissions is essential to understand its contribution to global warming. The authors here employ multi-spectral satellite data to create a methane detection tool with global coverage and high temporal and spatial resolution.

Methane (CH4) is the chief contributor to global climate change. Recent Airborne Visible-Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) has been very useful in quantitative mapping of methane emissions. Existing methods for analyzing this data are sensitive to local terrain conditions, often require manual inspection from domain experts, prone to significant error and hence are not scalable. To address these challenges, we propose a novel end-to-end spectral absorption wavelength aware transformer network, MethaneMapper, to detect and quantify the emissions. MethaneMapper introduces two novel modules that help to locate the most relevant methane plume regions in the spectral domain and uses them to localize these accurately. Thorough evaluation shows that MethaneMapper achieves 0.63 mAP in detection and reduces the model size (by 5×) compared to the current state of the art. In addition, we also introduce a large-scale dataset of methane plume segmentation mask for over 1200 AVIRIS-NG flight lines from 2015–2022. It contains over 4000 methane plume sites. Our dataset will provide researchers the opportunity to develop and advance new methods for tackling this challenging green-house gas detection problem with significant broader social impact. Dataset and source code link11https://github.com/UCSB-Vrl/methaneMapper-Spectral-Absorption-aware-Hyperspectral-Transformer-for-Methane-Detection.

Livestock methane emissions represent 32% of human-caused methane production, making automated monitoring critical for climate mitigation strategies. We introduce GasTwinFormer, a hybrid vision transformer for real-time methane emission segmentation and dietary classification in optical gas imaging through a novel Mix Twin encoder alternating between spatially-reduced global attention and locally-grouped attention mechanisms. Our architecture incorporates a lightweight LR-ASPP decoder for multi-scale feature aggregation and enables simultaneous methane segmentation and dietary classification in a unified framework. We contribute the first comprehensive beef cattle methane emission dataset using OGI, containing 11,694 annotated frames across three dietary treatments. GasTwinFormer achieves 74.47% mIoU and 83.63% mF1 for segmentation while maintaining exceptional efficiency with only 3.348M parameters, 3.428G FLOPs, and 114.9 FPS inference speed. Additionally, our method achieves perfect dietary classification accuracy (100%), demonstrating the effectiveness of leveraging diet-emission correlations. Extensive ablation studies validate each architectural component, establishing GasTwinFormer as a practical solution for real-time livestock emission monitoring. Please see our project page at gastwinformer.github.io.

No abstract available

No abstract available

The spatial prediction of soil CO2 flux is of great significance for assessing regional climate change and high-quality agricultural development. Using a single satellite to predict soil CO2 flux is limited by climatic conditions and land cover, resulting in low prediction accuracy. To this end, this study proposed a strategy of multi-source spectral satellite coordination and selected seven optical satellite remote sensing data sources (i.e., GF1-WFV, GF6-WFV, GF4-PMI, CB04-MUX, HJ2A-CCD, Sentinel 2-L2A, and Landsat 8-OLI) to extract auxiliary variables (i.e., vegetation indices and soil texture features). We developed a tree-structured Parzen estimator (TPE)-optimized extreme gradient boosting (XGBoost) model for the prediction and spatial mapping of soil CO2 flux. SHapley additive explanation (SHAP) was used to analyze the driving effects of auxiliary variables on soil CO2 flux. A scatter matrix correlation analysis showed that the distributions of auxiliary variables and soil CO2 flux were skewed, and the linear correlations between them (r < 0.2) were generally weak. Compared with single-satellite variables, the TPE-XGBoost model based on multiple-satellite variables significantly improved the prediction accuracy (RMSE = 3.23 kg C ha−1 d−1, R2 = 0.73), showing a stronger fitting ability for the spatial variability of soil CO2 flux. The spatial mapping results of soil CO2 flux based on the TPE-XGBoost model revealed that the high-flux areas were mainly concentrated in eastern and northern farmlands. The SHAP analysis revealed that PC2 and the TCARI of Sentinel 2-L2A and the TVI of HJ2A-CCD had significant positive driving effects on the prediction accuracy of soil CO2 flux. The above results indicate that the integration of multiple-satellite data can enhance the reliability and accuracy of spatial predictions of soil CO2 flux, thereby supporting regional agricultural sustainable development and climate change response strategies.

Peatlands play a key role in the circulation of the main greenhouse gases (GHG) – methane (CH4), carbon dioxide (CO2), and nitrous oxide (N2O). Therefore, detecting the spatial pattern of GHG sinks and sources in peatlands is pivotal for guiding effective climate change mitigation in the land use sector. While geospatial environmental data, which provide detailed spatial information on ecosystems and land use, offer valuable insights into GHG sinks and sources, the potential of directly using remote sensing data from satellites remains largely unexplored. We predicted the spatial distribution of three major GHGs (CH4, CO2, and N2O) sinks and sources across Finland. Utilizing 143 field measurements, we compared the predictive capacity of three different data sets with MaxEnt machine-learning modeling: (1) geospatial environmental data including climate, topography and habitat variables, (2) remote sensing data (Sentinel-1 and Sentinel-2), and (3) a combination of both. The combined dataset yielded the highest accuracy with an average test area under the receiver operating characteristic curve (AUC) of 0.845 and AUC stability of 0.928. A slightly lower accuracy was achieved using only geospatial environmental data (test AUC 0.810, stability AUC 0.924). In contrast, using only remote sensing data resulted in reduced predictive accuracy (test AUC 0.763, stability AUC 0.927). Our results suggest that (1) reliable estimates of GHG sinks and sources cannot be produced with remote sensing data only and (2) integrating multiple data sources is recommended to achieve accurate and realistic predictions of GHG spatial patterns. We compared remote sensing and geospatial data to predict peatland greenhouse gas sinks and sources across Finland. Remote sensing data perform less effectively than habitat and climate-related variables. We recommend integrating various data sources for modeling greenhouse gas sinks and sources. We compared remote sensing and geospatial data to predict peatland greenhouse gas sinks and sources across Finland. Remote sensing data perform less effectively than habitat and climate-related variables. We recommend integrating various data sources for modeling greenhouse gas sinks and sources.

No abstract available

The CO2 efflux from forest soil (FCO2) is one of the largest components of the global carbon cycle. Accurate estimation of FCO2 can help us better understand the carbon cycle in forested areas and precisely predict future climate change. However, the scarcity of field-measured FCO2 data in the subtropical forested area greatly limits our understanding of FCO2 dynamics at regional and global scales. This study used an automatic cavity ring-down spectrophotometer (CRDS) analyzer to measure FCO2 in a typical subtropical forest of southern China in the dry season. We found that the measured FCO2 at two experimental areas experienced similar temporal trends in the dry season and reached the minima around December, whereas the mean FCO2 differed apparently across the two areas (9.05 vs. 5.03 g C m−2 day−1) during the dry season. Moreover, we found that both abiotic (soil temperature and moisture) and biotic (vegetation productivity) factors are significantly and positively correlated, respectively, with the FCO2 variation during the study period. Furthermore, a machine-learning random forest model (RF model) that incorporates remote sensing data is developed and used to predict the FCO2 pattern in the subtropical forest, and the topographic effects on spatiotemporal patterns of FCO2 were further investigated. The model evaluation indicated that the proposed model illustrated high prediction accuracy for the training and testing dataset. Based on the proposed model, the spatiotemporal patterns of FCO2 in the forested watershed that encloses the two monitoring sites were mapped. Results showed that the spatial distribution of FCO2 is obviously affected by topography: the high FCO2 values mainly occur in relatively high altitudinal areas, in slopes of 10–25°, and in sunny slopes. The results emphasized that future studies should consider topographical effects when simulating FCO2 in subtropical forests. Overall, our study unraveled the spatiotemporal variations of FCO2 and their driving factors in a subtropical forest of southern China in the dry season, and demonstrated that the proposed RF model in combination with remote sensing data can be a useful tool for predicting FCO2 in forested areas, particularly in subtropical and tropical forest ecosystems.

Abstract. Arctic methane (CH4) budgets are uncertain because field measurements often capture only fragments of the wet-to-dry gradient that control tundra CH4 fluxes. Wet hotspots are over-represented, while dry, net-sink sites are under-sampled. We paired over 13 000 chamber flux measurements during peak growing season in July (2019–2024) from Trail Valley Creek in the western Canadian Arctic with co-registered remotely sensed predictor variables to test how spatial resolution (1 m vs. 10 m) and choice of machine-learning algorithm shape upscaled CH4 flux maps over our 3.1 km2 study domain. Four algorithms for CH4 flux scaling (Random Forest (RF), Gradient Boosting Machine (GBM), Generalised Additive Model (GAM), and Support Vector Regression (SVR)) were tuned using the same stack of multispectral indices, terrain derivatives and a six-class landscape classification. Tree-based models such as RF and GBM offered the best balance of 10-fold cross-validated R2 (≤0.75) and errors, so RF and GBM were used in a subsequent step for upscaling to the study area. With 1 m resolution, GBM captured the full range of microtopographic extremes and predicted a mean July flux of 99 mg CH4 m−2 per month. In contrast, RF, which smoothed local extremes, yielded an average flux of 519 mg CH4 m−2 per month. The disagreement between flux estimates using GBM and RF correlated mainly with the Normalized Difference Water Index (NDWI), a moisture proxy, and was most pronounced in waterlogged, low-lying areas. Aggregating predictors to 10 m averaged the sharp metre-scale flux highs in hollows and lows on ridges, narrowing the GBM-RF difference to ∼75 mg CH4 m−2 per month while broadening the overall flux distribution with more intermediate values. At 1 m, microtopography was the main driver. At 10 m, moisture proxies explained about half of the variance. Our results demonstrate that: (i) metre predictors are indispensable for capturing the wet-dry microtopography and its CH4 signals, (ii) upscaling algorithm selection strongly controls prediction spread and uncertainty once that microrelief is resolved, and (iii) coarser grids smooth local microtopographic details, resulting in flattened CH4 flux peaks and wider distribution. At 10 m, however, flux estimates became more consistent between models and better represented broad moisture-driven patterns, suggesting improved generalisability despite some loss of detail. This is supported by findings for remote sensing derived seasonal subsidence which reflects moisture gradients. All factors combined lead to potentially large differences in scaled CH4 flux budgets, calling for a careful selection of scaling approaches, spatial predictor layers (e.g., vegetation, moisture, topography), and grid resolution. Future work should couple ultra-high-resolution imagery with temporally dynamic indices to reduce upscaling bias along Arctic wetness gradients.

Global warming potential (GWP) has been widely used in the life cycle assessment (LCA) to quantify the climate impacts of energy technologies. Most LCAs are static analyses without considering the dynamics of greenhouse gas (GHG) emissions and changes in background GHG concentrations. This study presents a dynamic approach to analyze the life-cycle GWP of energy technologies in different timeframes and representative GHG concentration pathways. Results show that higher atmospheric GHG concentrations lead to higher life-cycle GWP for long-term analysis. The impacts of background GHG concentrations are more significant for technologies with large operational emissions or CH4 emissions than technologies with low operational emissions. The case study for the U.S. electricity sector in 2020-2050 shows the impacts of background GHG concentrations and different LCA methods on estimating national climate impacts of different energy technology scenarios. Based on the results, it is recommended for future LCAs to incorporate temporal effects of GHG emissions when (1) the technology has large operational GHG emissions or CH4 emissions; (2) the analysis time frame is longer than 50 years; (3) when LCA results are used for policymaking or technology comparisons for mitigating climate change.

The need for new technology towards the innovation of a hybrid system for the production of electric energy is very necessary for today’s growing population. Because of an increasing population there is a change in weather condition which is affecting the renewable energy sources. Nowadays solar PV technology is a great source of energy production. But due to the effect of different environmental factors, there are some changes in the efficiency of the solar PV panel. A Photovoltaic module efficiency mainly depends on the ambient temperature, module temperature, incoming solar radiation intensity, and photovoltaic material composition. So these terms are affected by some environmental factors. So to know how the solar panel efficiency is affected by greenhouse gas I have conducted an experiment by taking two panels. One in normal atmosphere and another in GHG (CH4) chamber. It shows that with an increase in temperature of that chamber, there is a decrease in efficiency of solar PV.

The terrestrial net biome production (NBP) is considered as one of the major drivers of interannual variation in atmospheric CO2 levels. However, the determinants of variability in NBP under the background climate (i.e., preindustrial conditions) remain poorly understood, especially on decadal‐to‐centennial timescales. We analyzed 1,000‐year simulations spanning 850‐1,849 from the Community Earth System Model (CESM) and found that the variability in NBP and heterotrophic respiration (RH) were largely driven by fluctuations in the net primary production (NPP) and carbon turnover rates in response to climate variability. On interannual to multidecadal timescales, variability in NBP was dominated by variation in NPP, while variability in RH was driven by variation in turnover rates. However, on centennial timescales (100‐1,000 years), the RH variability became more tightly coupled to that of NPP. The NBP variability on centennial timescales was low, due to the near cancellation of NPP and NPP‐driven RH changes arising from climate internal variability and external forcings: preindustrial greenhouse gases, volcanic eruptions, land use changes, orbital change, and solar activity. Factorial experiments showed that globally on centennial timescales, the forcing of changes in greenhouse gas concentrations were the largest contributor (51%) to variations in both NPP and RH, followed by volcanic eruptions impacting NPP (25%) and RH (31%). Our analysis of the carbon‐cycle suggests that geoengineering solutions by injection of stratospheric aerosols might be ineffective on longer timescales.

The objective of this experiment was to investigate the effects of an essential oil (EO) blend on lactational performance, rumen fermentation, nutrient utilization, blood variables, enteric methane emissions and manure greenhouse gas-emitting potential in dairy cows. A randomized complete block design experiment was conducted with 26 primiparous and 22 multiparous Holstein cows. A 2-wk covariate and a 2-wk adaptation periods preceded a 10-wk experimental period used for data and sample collection. Treatments were: (1) basal diet supplemented with placebo - CON; and (2) basal diet supplemented with a blend of EO containing eugenol and geranyl acetate as main compounds. Supplementation with EO did not affect dry matter intake, milk and energy-corrected milk yields, and feed efficiency of cows, compared with CON. Milk fat and lactose concentrations were increased, and milk TS concentration and milk fat yield tended to be increased by EO. Multiparous cows supplemented with EO tended to have slightly decreased dry matter and crude protein digestibility compared with CON multiparous cows. There was a tendency for increased ruminal pH by EO, whereas other rumen fermentation variables did not differ between treatments. Daily methane emission was not affected by EO supplementation, but methane emission intensity per kg of milk fat was decreased by 8.5% by EO. Methane emission intensity per kg of milk lactose and milk TS were decreased and methane emission intensity per kg of milk yield tended to be decreased by up to 10% in EO multiparous cows, but not in primiparous cows. The greenhouse gas-emitting potential of manure was not affected by EO supplementation. Compared with CON, fecal nitrogen excretion was increased by EO supplementation in multiparous, but not in primiparous cows, and milk nitrogen secretion (as a % of nitrogen intake) tended to be increased in EO supplemented cows. Blood variables were not affected by EO supplementation in the current study. Overall, dietary supplementation of EO did not affect lactational performance of the cows, although milk fat and lactose concentrations were increased. Most enteric methane emission metrics were not affected, but EO decreased methane intensity per kg of milk fat by 8.5%, compared with the control.

As greenhouse gas emissions from dairy farms are on the rise, effective monitoring of these emissions has emerged as a crucial tool for assessing their environmental impacts and promoting sustainable development. Most of the existing studies on GHGs from dairy farms involve stationary detections with long response times and high costs. In this study, a greenhouse gas detection system was constructed based on NDIR technology using a single broadband light source and a four-channel thermopile detector for the detection of CH4, N2O, and CO2; the detection range of CH4 was 0~100 ppm; that of N2O was 0~500 ppm; and that of CO2 was 0~20%. After the concentration calibration, the cross-interference between the gas measurement channels was studied, and the least-squares method was used to correct the interference between the three gases. The experimental results showed that the full-range deviation of the detection device was lower than 0.81%, the repeatability was lower than 0.39%, the stability was lower than 0.61%, and the response time was lower than 10 s. This study also carried out on-site testing in Luoyang Shengsheng Ranch (Luoyang, China), and the results show that the error between this device and the PTM600 portable gas analyzer is within 9.78%, and the dynamic response time of this device is within 16 s, at which point the content of greenhouse gases in dairy farms can be measured quickly and accurately. The objective of this study is to enhance the precision and effectiveness of greenhouse gas (GHG) emissions monitoring from dairy farms, thereby contributing to environmental protection and sustainable development goals. By achieving this, we aim to facilitate societal progress towards a greener and low-carbon future.

Making a cost-effective governance of greenhouse gas (GHG) emissions and air pollution is of great importance for megacities to pursue a sustainable future. To achieve this, the present study advocates megacities to implement a two-tier synergic governance system consisting of both synergic governance between GHG and air pollutant emission reductions and between megacities and their surrounding regions. Based on the LEAP model and WRF-SMOKE-CMAQ simulation platform, this study found that climate governance of China's megacity, Shenzhen, could synergistically contribute to decreasing urban annual PM2.5 concentration by 5.6% in 2030. Using synergic governance with surrounding regions could further help cap and then quickly decrease the megacity's GHG emissions and lower its PM2.5 concentrations by an additional 11.8%. The results demonstrated the substantial effects of transdepartment and transregional synergic governance on Shenzhen's GHG emission reduction and air quality improvement. Furthermore, this study suggested road transportation and power generation and supply as the two priority fields for wide-ranging megacities to promote two-tier synergic governance, highlighting an integration of improved urban electrification with high-efficiency electricity use and a renewable-based power supply.

The Yedoma layer, a permafrost layer containing a massive amount of underground ice in the Arctic regions, is reported to be rapidly thawing. In this study, we develop the Permafrost Degradation and Greenhouse gasses Emission Model (PDGEM), which describes the thawing of the Arctic permafrost including the Yedoma layer due to climate change and the greenhouse gas (GHG) emissions. The PDGEM includes the processes by which high-concentration GHGs (CO 2 and CH 4 ) contained in the pores of the Yedoma layer are released directly by dynamic degradation, as well as the processes by which GHGs are released by the decomposition of organic matter in the Yedoma layer and other permafrost. Our model simulations show that the total GHG emissions from permafrost degradation in the RCP8.5 scenario was estimated to be 31-63 PgC for CO 2 and 1261-2821 TgCH 4 for CH 4 (68 th percentile of the perturbed model simulations, corresponding to a global average surface air temperature change of 0.05–0.11 °C), and 14-28 PgC for CO 2 and 618-1341 TgCH 4 for CH 4 (0.03–0.07 °C) in the RCP2.6 scenario. GHG emissions resulting from the dynamic degradation of the Yedoma layer were estimated to be less than 1% of the total emissions from the permafrost in both scenarios, possibly because of the small area ratio of the Yedoma layer. An advantage of PDGEM is that geographical distributions of GHG emissions can be estimated by combining a state-of-the-art land surface model featuring detailed physical processes with a GHG release model using a simple scheme, enabling us to consider a broad range of uncertainty regarding model parameters. In regions with large GHG emissions due to permafrost thawing, it may be possible to help reduce GHG emissions by taking measures such as restraining land development.

No abstract available

The static chamber approach is often used for greenhouse gas (GHG) flux measurements, whereby the flux is deduced from the increase of species concentration after closing the chamber. Since this increase changes diffusion gradients between chamber air and soil air, a nonlinear increase is expected. Lateral gas flow and leakages also contribute to non linearity. Several models have been suggested to account for this non linearity, the most recent being the Hutchinson–Mosier regression model (hmr). However, the practical application of these models is challenging because the researcher needs to decide for each flux whether a nonlinear fit is appropriate or exaggerates flux estimates due to measurement artifacts. In the latter case, a flux estimate from the linear model is a more robust solution and introduces less arbitrary uncertainty to the data. We present the new, dynamic and reproducible flux calculation scheme, kappa.max, for an improved trade-off between bias and uncertainty (i.e. accuracy and precision). We develop a tool to simulate, visualise and optimise the flux calculation scheme for any specific static N2O chamber measurement system. The decision procedure and visualisation tools are implemented in a package for the R software. Finally, we demonstrate with this approach the performance of the applied flux calculation scheme for a measured flux dataset to estimate the actual bias and uncertainty. The kappa.max method effectively improved the decision between linear and nonlinear flux estimates reducing the bias at a minimal cost of uncertainty.

No abstract available

BackgroundFlow sheet options for integrating ethanol production from spent sulfite liquor (SSL) into the acid-based sulfite pulping process at the Sappi Saiccor mill (Umkomaas, South Africa) were investigated, including options for generation of thermal and electrical energy from onsite bio-wastes, such as bark. Processes were simulated with Aspen Plus® for mass- and energy-balances, followed by an estimation of the economic viability and environmental impacts. Various concentration levels of the total dissolved solids in magnesium oxide-based SSL, which currently fuels a recovery boiler, prior to fermentation was considered, together with return of the fermentation residues (distillation bottoms) to the recovery boiler after ethanol separation. The generation of renewable thermal and electrical energy from onsite bio-wastes were also included in the energy balance of the combined pulping-ethanol process, in order to partially replace coal consumption. The bio-energy supplementations included the combustion of bark for heat and electricity generation and the bio-digestion of the calcium oxide SSL to produce methane as additional energy source.ResultsEthanol production from SSL at the highest substrate concentration was the most economically feasible when coal was used for process energy. However this solution did not provide any savings in greenhouse gas (GHG) emissions for the concentration-fermentation-distillation process. Maximizing the use of renewable energy sources to partially replace coal consumption yielded a satisfactory economic performance, with a minimum ethanol selling price of 0.83 US$/l , and a drastic reduction in the overall greenhouse gas emissions for the entire facility.ConclusionHigh substrate concentrations and conventional distillation should be used when considering integrating ethanol production at sulfite pulping mills. Bio-wastes generated onsite should be utilized at their maximum potential for energy generation in order to maximize the GHG emissions reduction.

No abstract available

No abstract available

Background Little is known about the combined impacts of global environmental changes and ecological disturbances on ecosystem functioning, even though such combined impacts might play critical roles in shaping ecosystem processes that can in turn feed back to climate change, such as soil emissions of greenhouse gases. Methodology/Principal Findings We took advantage of an accidental, low-severity wildfire that burned part of a long-term global change experiment to investigate the interactive effects of a fire disturbance and increases in CO2 concentration, precipitation and nitrogen supply on soil nitrous oxide (N2O) emissions in a grassland ecosystem. We examined the responses of soil N2O emissions, as well as the responses of the two main microbial processes contributing to soil N2O production – nitrification and denitrification – and of their main drivers. We show that the fire disturbance greatly increased soil N2O emissions over a three-year period, and that elevated CO2 and enhanced nitrogen supply amplified fire effects on soil N2O emissions: emissions increased by a factor of two with fire alone and by a factor of six under the combined influence of fire, elevated CO2 and nitrogen. We also provide evidence that this response was caused by increased microbial denitrification, resulting from increased soil moisture and soil carbon and nitrogen availability in the burned and fertilized plots. Conclusions/Significance Our results indicate that the combined effects of fire and global environmental changes can exceed their effects in isolation, thereby creating unexpected feedbacks to soil greenhouse gas emissions. These findings highlight the need to further explore the impacts of ecological disturbances on ecosystem functioning in the context of global change if we wish to be able to model future soil greenhouse gas emissions with greater confidence.

No abstract available

No abstract available

Abstract Methane (CH4) is a potent greenhouse gas and the second highest contributor to global warming. CH4 emissions are still growing at an alarmingly high pace. To limit global warming to 1.5 °C, one of the most effective strategies is to reduce rapidly the CH4 emissions by developing large‐scale methane removal methods. The purpose of this perspective paper is threefold. (1) To highlight the technology gap dealing with low concentration CH4 (at many emission sources and in the atmosphere). (2) To analyze the challenges and prospects of solar‐driven gas phase advanced oxidation processes for CH4 removal. And (3) to propose some ideas, which may help to develop solar‐driven gas phase advanced oxidation processes and make them deployable at a climate significant scale.

No abstract available

No abstract available

No abstract available

Tuna are globally distributed species of major commercial importance and some tuna species are a major source of protein in many countries. Tuna are characterized by dynamic distribution patterns that respond to climate variability and long‐term change. Here, we investigated the effect of environmental conditions on the worldwide distribution and relative abundance of six tuna species between 1958 and 2004 and estimated the expected end‐of‐the‐century changes based on a high‐greenhouse gas concentration scenario (RCP8.5). We created species distribution models using a long‐term Japanese longline fishery dataset and two‐step generalized additive models. Over the historical period, suitable habitats shifted poleward for 20 out of 22 tuna stocks, based on their gravity centre (GC) and/or one of their distribution limits. On average, tuna habitat distribution limits have shifted poleward 6.5 km per decade in the northern hemisphere and 5.5 km per decade in the southern hemisphere. Larger tuna distribution shifts and changes in abundance are expected in the future, especially by the end‐of‐the‐century (2080–2099). Temperate tunas (albacore, Atlantic bluefin, and southern bluefin) and the tropical bigeye tuna are expected to decline in the tropics and shift poleward. In contrast, skipjack and yellowfin tunas are projected to become more abundant in tropical areas as well as in most coastal countries' exclusive economic zones (EEZ). These results provide global information on the potential effects of climate change in tuna populations and can assist countries seeking to minimize these effects via adaptive management.

No abstract available

No abstract available

Nitrous oxide (N2O) is an environmentally important atmospheric trace gas because it is an effective greenhouse gas and it leads to ozone depletion through photo-chemical nitric oxide (NO) production in the stratosphere. Mitigating its steady increase in atmospheric concentration requires an understanding of the mechanisms that lead to its formation in natural and engineered microbial communities. N2O is formed biologically from the oxidation of hydroxylamine (NH2OH) or the reduction of nitrite (NO−2) to NO and further to N2O. Our review of the biological pathways for N2O production shows that apparently all organisms and pathways known to be involved in the catabolic branch of microbial N-cycle have the potential to catalyze the reduction of NO−2 to NO and the further reduction of NO to N2O, while N2O formation from NH2OH is only performed by ammonia oxidizing bacteria (AOB). In addition to biological pathways, we review important chemical reactions that can lead to NO and N2O formation due to the reactivity of NO−2, NH2OH, and nitroxyl (HNO). Moreover, biological N2O formation is highly dynamic in response to N-imbalance imposed on a system. Thus, understanding NO formation and capturing the dynamics of NO and N2O build-up are key to understand mechanisms of N2O release. Here, we discuss novel technologies that allow experiments on NO and N2O formation at high temporal resolution, namely NO and N2O microelectrodes and the dynamic analysis of the isotopic signature of N2O with quantum cascade laser absorption spectroscopy (QCLAS). In addition, we introduce other techniques that use the isotopic composition of N2O to distinguish production pathways and findings that were made with emerging molecular techniques in complex environments. Finally, we discuss how a combination of the presented tools might help to address important open questions on pathways and controls of nitrogen flow through complex microbial communities that eventually lead to N2O build-up.

Water temperature is critical for the ecology of lakes. However, the ability to predict its spatial and seasonal variation is constrained by the lack of a thermal classification system. Here we define lake thermal regions using objective analysis of seasonal surface temperature dynamics from satellite observations. Nine lake thermal regions are identified that mapped robustly and largely contiguously globally, even for small lakes. The regions differed from other global patterns, and so provide unique information. Using a lake model forced by 21st century climate projections, we found that 12%, 27% and 66% of lakes will change to a lower latitude thermal region by 2080–2099 for low, medium and high greenhouse gas concentration trajectories (Representative Concentration Pathways 2.6, 6.0 and 8.5) respectively. Under the worst-case scenario, a 79% reduction in the number of lakes in the northernmost thermal region is projected. This thermal region framework can facilitate the global scaling of lake-research. Water temperature is a critical variable for lakes, but its spatial and temporal patterns are not well characterised globally. Here, the authors use surface temperature dynamics to define lake thermal regions that group lakes with similar patterns, and show how these regions shift under climate change.

No abstract available

The effects of nitrogen (N) deposition on soil organic carbon (C) and greenhouse gas (GHG) emissions in terrestrial ecosystems are the main drivers affecting GHG budgets under global climate change. Although many studies have been conducted on this topic, we still have little understanding of how N deposition affects soil C pools and GHG budgets at the global scale. We synthesized a comprehensive dataset of 275 sites from multiple terrestrial ecosystems around the world and quantified the responses of the global soil C pool and GHG fluxes induced by N enrichment. The results showed that the soil organic C concentration and the soil CO2, CH4 and N2O emissions increased by an average of 3.7%, 0.3%, 24.3% and 91.3% under N enrichment, respectively, and that the soil CH4 uptake decreased by 6.0%. Furthermore, the percentage increase in N2O emissions (91.3%) was two times lower than that (215%) reported by Liu and Greaver (Ecology Letters, 2009, 12:1103–1117). There was also greater stimulation of soil C pools (15.70 kg C ha−1 year−1 per kg N ha−1 year−1) than previously reported under N deposition globally. The global N deposition results showed that croplands were the largest GHG sources (calculated as CO2 equivalents), followed by wetlands. However, forests and grasslands were two important GHG sinks. Globally, N deposition increased the terrestrial soil C sink by 6.34 Pg CO2/year. It also increased net soil GHG emissions by 10.20 Pg CO2‐Geq (CO2 equivalents)/year. Therefore, N deposition not only increased the size of the soil C pool but also increased global GHG emissions, as calculated by the global warming potential approach.

Sustainable nitrogen cycle is an essential biogeochemical process that ensures ecosystem safety and byproduct greenhouse gas nitrous oxide reduction. Antimicrobials are always co-occurring with anthropogenic reactive nitrogen sources. However, their impacts on the ecological safety of microbial nitrogen cycle remain poorly understood. Here, a denitrifying bacterial strain Paracoccus denitrificans PD1222 was exposed to a widespread broad-spectrum antimicrobial triclocarban (TCC) at environmental concentrations. The denitrification was hindered by TCC at 25 μg L-1 and was completely inhibited once the TCC concentration exceeded 50 μg L-1. Importantly, the accumulation of N2O at 25 μg L-1 of TCC was 813 times as much as the control group without TCC, which attributed to the significantly downregulated expression of nitrous oxide reductase and the genes related to electron transfer, iron, and sulfur metabolism under TCC stress. Interestingly, combining TCC-degrading denitrifying Ochrobactrum sp. TCC-2 with strain PD1222 promoted the denitrification process and mitigated N2O emission by 2 orders of magnitude. We further consolidated the importance of complementary detoxification by introducing a TCC-hydrolyzing amidase gene tccA from strain TCC-2 into strain PD1222, which successfully protected strain PD1222 against the TCC stress. This study highlights an important link between TCC detoxification and sustainable denitrification and suggests a necessity to assess the ecological risks of antimicrobials in the context of climate change and ecosystem safety.

The concentration of dissolved oxygen (DO) is an important attribute of aquatic ecosystems, influencing habitat, drinking water quality, biodiversity, nutrient biogeochemistry, and greenhouse gas emissions. While average summer DO concentrations are declining in lakes across the temperate zone, much remains unknown about seasonal factors contributing to deepwater DO losses. It is unclear whether declines are related to increasing rates of seasonal DO depletion or changes in seasonal stratification that limit re‐oxygenation of deep waters. Furthermore, despite the presence of important biological and ecological DO thresholds, there has been no large‐scale assessment of changes in the amount of habitat crossing these thresholds, limiting the ability to understand the consequences of observed DO losses. We used a dataset from >400 widely distributed lakes to identify the drivers of DO losses and quantify the frequency and volume of lake water crossing biologically and ecologically important threshold concentrations ranging from 5 to 0.5 mg/L. Our results show that while there were no consistent changes over time in seasonal DO depletion rates, over three‐quarters of lakes exhibited an increase in the duration of stratification, providing more time for seasonal deepwater DO depletion to occur. As a result, most lakes have experienced summertime increases in the amount of water below all examined thresholds in deepwater DO concentration, with increases in the proportion of the water column below thresholds ranging between 0.9% and 1.7% per decade. In the 30‐day period preceding the end of stratification, increases were greater at >2.2% per decade and >70% of analyzed lakes experienced increases in the amount of oxygen‐depleted water. These results indicate ongoing climate‐induced increases in the duration of stratification have already contributed to reduction of habitat for many species, likely increased internal nutrient loading, and otherwise altered lake chemistry. Future warming is likely to exacerbate these trends.

No abstract available

Nitrous oxide (N2O) is a powerful greenhouse gas and the main driver of stratospheric ozone depletion. Since soils are the largest source of N2O, predicting soil response to changes in climate or land use is central to understanding and managing N2O. Here we find that N2O flux can be predicted by models incorporating soil nitrate concentration (NO3−), water content and temperature using a global field survey of N2O emissions and potential driving factors across a wide range of organic soils. N2O emissions increase with NO3− and follow a bell-shaped distribution with water content. Combining the two functions explains 72% of N2O emission from all organic soils. Above 5 mg NO3−-N kg−1, either draining wet soils or irrigating well-drained soils increases N2O emission by orders of magnitude. As soil temperature together with NO3− explains 69% of N2O emission, tropical wetlands should be a priority for N2O management.In a global field survey across a wide range of organic soils, the authors find that N2O flux can be predicted by models incorporating soil nitrate concentration (NO3–), water content and temperature. N2O emission increases with NO3– and temperature and follows a bell-shaped distribution with water content.

No abstract available

The cost of monitoring greenhouse gas emissions from landfill sites is of major concern for regulatory authorities. The current monitoring procedure is recognised as labour intensive, requiring agency inspectors to physically travel to perimeter borehole wells in rough terrain and manually measure gas concentration levels with expensive hand-held instrumentation. In this article we present a cost-effective and efficient system for remotely monitoring landfill subsurface migration of methane and carbon dioxide concentration levels. Based purely on an autonomous sensing architecture, the proposed sensing platform was capable of performing complex analytical measurements in situ and successfully communicating the data remotely to a cloud database. A web tool was developed to present the sensed data to relevant stakeholders. We report our experiences in deploying such an approach in the field over a period of approximately 16 months.

Urine diversion has been proposed as an approach for producing renewable fertilizers and reducing nutrient loads to wastewater treatment plants. Life cycle assessment was used to compare environmental impacts of the operations phase of urine diversion and fertilizer processing systems [via (1) a urine concentration alternative and (2) a struvite precipitation and ion exchange alternative] at a city scale to conventional systems. Scenarios in Vermont, Michigan, and Virginia were modeled, along with additional sensitivity analyses to understand the importance of key parameters, such as the electricity grid and wastewater treatment method. Both urine diversion technologies had better environmental performance than the conventional system and led to reductions of 29-47% in greenhouse gas emissions, 26-41% in energy consumption, approximately half the freshwater use, and 25-64% in eutrophication potential, while acidification potential ranged between a 24% decrease to a 90% increase. In some situations, wastewater treatment chemical requirements were eliminated. The environmental performance improvement was usually dependent on offsetting the production of synthetic fertilizers. This study suggests that urine diversion could be applied broadly as a strategy for both improving wastewater management and decarbonization.

Ventilation air methane (VAM) is one of the main greenhouse gas sources. Owing to the characteristics of low concentration of ventilation air methane and high moisture content, we build an experimental platform and take the oxidative combustion temperature and methane conversion rate as the research indexes, and the systematic research finds that the inhibitory effect of moisture on the oxidative combustion of ultra‐low concentration of methane (<1%) is a nonlinear polynomial law. In the meantime, we constructed OH(H2O)n+CH4 and studied its reaction potential energy surface using quantum chemical calculations, which used the most significant primitive reaction of methane combustion, OH+CH4→H2O+CH3, as the theoretical basis. We found that as moisture content increased, so did its reaction energy barrier, making the reactants more stable, strengthening the three‐body collision effect, and reducing the number of free radicals, all of which hindered the methane chain reaction. The study aimed to validate the experimental finding that moisture inhibits the oxidative combustion of ventilation air methane by examining the internal mechanism of methane oxidation. © 2024 Society of Chemical Industry and John Wiley & Sons, Ltd.

Dactylorhiza hatagirea is a terrestrial orchid listed in Appendix II of the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) and classified as threatened by International Union for Conservation of Nature (IUCN). It is endemic to the Hindu-Kush Himalayan region, distributed from Pakistan to China. The main threat to its existence is climate change and the associated change in the distribution of its suitable habitats to higher altitudes due to increasing temperature. It is therefore necessary to determine the habitats that are suitable for its survival and their expected distribution after the predicted changes in climate. To do this, we use Maxent modelling of the data for its 208 locations. We predict its distribution in 2050 and 2070 using four climate change models and two greenhouse gas concentration trajectories. This revealed severe losses of suitable habitat in Nepal, in which, under the worst scenario, there will be a 71–81% reduction the number of suitable locations for D. hatagirea by 2050 and 95–98% by 2070. Under the most favorable scenario, this reduction will be 65–85% by 2070. The intermediate greenhouse gas concentration trajectory surprisingly would result in a greater reduction by 2070 than the worst-case scenario. Our results provide important guidelines that local authorities interested in conserving this species could use to select areas that need to be protected now and in the future.

Underground mines are responsible for a large number of the released methane worldwide. Our study paper can contribute to mitigating methane emissions, hence reducing the concentration about greenhouse emissions within the environment and mitigating the associated hazardous risks. In this study, more than 45 recent journals on Methane emission in Coal Mines were gathered from Web of Science, IEEE Xplore, ScienceDirect, and ResearchGate. A systematic review is accomplished of the past four years of various parts of Methane gas emission such as anthropogenic emission sources, gas emissions detection, prediction of methane using different technology and the Artificial intelligence projection model for methane emission. The outcomes reveal that since the methane emission management has obtained increasing attention over the past four years. This study also shows that big countries are using technology to control and utilize the methane emission to reduce the energy crisis. To decrease the coal mine injuries, academic understanding of underground methane management has increased, different technologies are integrated and support from various IT departments has amplified for the forecasting. In the future, the most critical task for coal mines risk assessment is to restore the worker's trust in mine safety, and the primary solution is to give more awareness to the underground management and workers through utilizing Artificial Intelligence (AI) mainly Artificial Neural Networks (ANN).

No abstract available

No abstract available

No abstract available

Supercritical water gasification (SCWG) of lignocellulosic biomass is a promising pathway for the production of hydrogen. However, SCWG is a complex thermochemical process, the modeling of which is challenging via conventional methodologies. Therefore, eight machine learning models (linear regression (LR), Gaussian process regression (GPR), artificial neural network (ANN), support vector machine (SVM), decision tree (DT), random forest (RF), extreme gradient boosting (XGB), and categorical boosting regressor (CatBoost)) with particle swarm optimization (PSO) and a genetic algorithm (GA) optimizer were developed and evaluated for prediction of H2, CO, CO2, and CH4 gas yields from SCWG of lignocellulosic biomass. A total of 12 input features of SCWG process conditions (temperature, time, concentration, pressure) and biomass properties (C, H, N, S, VM, moisture, ash, real feed) were utilized for the prediction of gas yields using 166 data points. Among machine learning models, boosting ensemble tree models such as XGB and CatBoost demonstrated the highest power for the prediction of gas yields. PSO-optimized XGB was the best performing model for H2 yield with a test R2 of 0.84 and PSO-optimized CatBoost was best for prediction of yields of CH4, CO, and CO2, with test R2 values of 0.83, 0.94, and 0.92, respectively. The effectiveness of the PSO optimizer in improving the prediction ability of the unoptimized machine learning model was higher compared to the GA optimizer for all gas yields. Feature analysis using Shapley additive explanation (SHAP) based on best performing models showed that (21.93%) temperature, (24.85%) C, (16.93%) ash, and (29.73%) C were the most dominant features for the prediction of H2, CH4, CO, and CO2 gas yields, respectively. Even though temperature was the most dominant feature, the cumulative feature importance of biomass characteristics variables (C, H, N, S, VM, moisture, ash, real feed) as a group was higher than that of the SCWG process condition variables (temperature, time, concentration, pressure) for the prediction of all gas yields. SHAP two-way analysis confirmed the strong interactive behavior of input features on the prediction of gas yields.

No abstract available

No abstract available

Separating carbon dioxide (CO2) from gaseous streams released into the atmosphere is becoming critical due to its greenhouse effect. Membrane technology is one of the promising technologies for CO2 capture. SAPO-34 filler was incorporated in polymeric media to synthesize mixed matrix membrane (MMM) and enhance the CO2 separation performance of this process. Despite relatively extensive experimental studies, there are limited studies that cover the modeling aspects of CO2 capture by MMMs. This research applies a special type of machine learning modeling scenario, namely, cascade neural networks (CNN), to simulate as well as compare the CO2/CH4 selectivity of a wide range of MMMs containing SAPO-34 zeolite. A combination of trial-and-error analysis and statistical accuracy monitoring has been applied to fine-tune the CNN topology. It was found that the CNN with a 4-11-1 topology has the highest accuracy for the modeling of the considered task. The designed CNN model is able to precisely predict the CO2/CH4 selectivity of seven different MMMs in a broad range of filler concentrations, pressures, and temperatures. The model predicts 118 actual measurements of CO2/CH4 selectivity with an outstanding accuracy (i.e., AARD = 2.92%, MSE = 1.55, R = 0.9964).

Interfacial tension (IFTC-B) between CO2 and brine depends on chemical components in multiphase systems, intricately evolving with a change in temperature. In this study, we developed a convolutional neural network with a multibranch structure (MBCNN), which, in combination with a compiled data set containing measurement data of 1716 samples from 13 available literature sources at wide temperature and pressure ranges (273.15-473.15 K and 0-70 MPa), was used to quantitatively explore the correlation of various chemical components with IFTC-B at varying temperature, aiming to achieve accurate predictions of IFTC-B under complex conditions. Our multibranch neural network analysis yielded some important insights: (1) Leveraging the convolutional and multibranch structure, MBCNN effectively mitigates the adverse effects of sparse matrices resulting from the absence of certain basic components, exhibiting higher prediction accuracy particularly for low IFTC-B scenarios (MAE = 0.47, and R2 = 0.9921) than other AI models. (2) The multibranch structure allows MBCNN to additionally capture the interattribute relationship between temperature and each chemical component. Such interattribute relationships are quantitatively correlated with IFTC-B, demonstrating that varying temperature significantly influences the dependence of IFTC-B on chemical components in gas and brine by causing the variation in their solubility. Specifically, the ratio of IFTC-B to the molality of monovalent cations (Na+ and K+) and bivalent cations (Ca2+ and Mg2+) in brine, as well as to the mole fraction of non-CO2 components (CH4 and N2) in the gas phase, varies with increasing temperature, approximately following a quadratic function. (3) By formulating the effect of each attribute on IFTC-B and quantifying their respective weight, we derived a new piecewise function for predicting IFTC-B at three temperature intervals (T ≤ 293.15 K, 293.15 K < T ≤ 324.4 K, and T > 324.4 K), with high prediction performance (MAE = 2.3672, R2 = 0.9263) across a wide temperature range in saline aquifers.

The biological methanation process has emerged as a promising alternative to thermo-catalytic methods due to its ability to operate under milder conditions. However, challenges such as low hydrogen solubility and the need for precise trace element supplementation (Fe(II), Ni(II), Co(II)) constrain methane production yield. This study investigates the combined effects of trace element concentrations and applied pressure on biological methanation, addressing their synergistic interactions. Using a face-centered composite design, batch mode experiments were conducted to optimize methane production. Response Surface Methodology (RSM) and Artificial Neural Network (ANN)—Genetic Algorithm (GA) approaches were employed to model and optimize the process. RSM identified optimal ranges for trace elements and pressure, while ANN-GA demonstrated superior predictive accuracy, capturing nonlinear relationships with a high R² (>0.99) and minimal prediction errors. ANN-GA optimization indicated 97.9% methane production efficiency with a reduced conversion time of 15.9 h under conditions of 1.5 bar pressure and trace metal concentrations of 25.0 mg/L Fe(II), 0.20 mg/L Ni(II), and 0.02 mg/L Co(II). Validation experiments confirmed these predictions with deviations below 5%, underscoring the robustness of the models. The results highlight the synergistic effects of pressure and trace metals in enhancing gas–liquid mass transfer and enzymatic pathways, demonstrating the potential of computational modeling and experimental validation to optimize biological methanation systems, contributing to sustainable methane production.

Predicting carbon dioxide (CO2) solubility in water and brine is crucial for understanding carbon capture and storage (CCS) processes. Accurate solubility predictions inform the feasibility and effectiveness of CO2 dissolution trapping, a key mechanism in carbon sequestration in saline aquifers. In this work, a comprehensive data set comprising 1278 experimental solubility data points for CO2–brine systems was assembled, encompassing diverse operating conditions. These data encompassed brines containing six different salts: NaCl, KCl, NaHCO3, CaCl2, MgCl2, and Na2SO4. Also, this databank encompassed temperature spanning from 273.15 to 453.15 K and a pressure range spanning 0.06–100 MPa. To model this solubility databank, cascade forward neural network (CFNN) and generalized regression neural network (GRNN) were employed. Furthermore, three optimization algorithms, namely, Bayesian Regularization (BR), Broyden–Fletcher–Goldfarb–Shanno (BFGS) quasi-Newton, and Levenberg–Marquardt (LM), were applied to enhance the performance of the CFNN models. The CFNN-LM model showcased average absolute percent relative error (AAPRE) values of 5.37% for the overall data set, 5.26% for the training subset, and 5.85% for the testing subset. Overall, the CFNN-LM model stands out as the most accurate among the models crafted in this study, boasting the highest overall R2 value of 0.9949 among the other models. Based on sensitivity analysis, pressure exerts the most significant influence and stands as the sole parameter with a positive impact on CO2 solubility in brine. Conversely, temperature and the concentration of all six salts considered in the model exhibited a negative impact. All salts exert a negative impact on CO2 solubility due to their salting-out effect, with varying degrees of influence. The salting-out effects of the salts can be ranked as follows: from the most pronounced to the least: MgCl2 > CaCl2 > NaCl > KCl > NaHCO3 > Na2SO4. By employing the leverage approach, only a few instances of potential suspected and out-of-leverage data were found. The relatively low count of identified potential suspected and out-of-leverage data, given the expansive solubility database, underscores the reliability and accuracy of both the data set and the CFNN-LM model’s performance in this survey.

The effective capturing of carbon dioxide (CO₂) from flue gases represents an important environmental and economic challenge, where such emissions are considered a major contributor to global climate issue. Traditional capturing techniques such as amine scrubbing are energy-intensive with high cost, due to the necessity of cooling high temperature flue gases before the separation process. In this study, we investigated the utilization of ceramic membranes fabricated from Saudi red clay which considered an available, cost-effective local material as a sustainable solution for high-temperature CO₂ capture. The research evaluates separation efficiency under high temperatures and different pressure values, which enable direct CO₂ capturing from hot flue streams without precooling processes. Experimental results demonstrate the membranes efficacy in separating CO₂ from flue gas, in which the presence of iron oxide (Fe₂O₃) constituents in the clay enhancing capture efficiency through weak chemisorption. In addition, the membranes showed robust structural integrity and consistent performance under high temperature conditions, compared to polymeric membranes that degrade thermally and offer advantages over metal-organic framework-enhanced ceramics, which incur higher costs and lower thermal tolerance. An ANN model is constructed to estimate the membrane performance (CO2 concentration (%) in permeate) using results obtained from the present experimental results and utilizing pressure and temperature as ANN input parameters. The process of training incorporates the analysis of the loss function on training and validation data for controlling the weights and biases using backpropagation while feed forward propagate the selected input parameters. A total of 8 hidden layers consisting of 12 neurons each has been used in constructing the ANN, and training process is optimized using the ADAM algorithm to minimize the loss function. The Final layer uses the linear activation function while all the hidden layers use the rectified Linear Units Activation function (ReLU). The ANN model demonstrates excellent predictive performance, yielding values close to 1 for R2 and r, along with extremely low values for MSE, MAPE, MSLE, and log-cosh loss (0.00033, 0.146%, 4.1×10⁻6, 0.00016 respectively), demonstrating the ANN model's high predictive accuracy.

Carbon evasion from urban river networks becomes increasingly significant as urbanization accelerates. However, there remains a limited understanding of the overall carbon emission impact integrating CO2 and CH4 dynamics, particularly in response to ecological restoration efforts. In this study, we investigated patterns of fluvial CO2 and CH4 diffusive fluxes across an urban river network in Wuxi, China. Our results reveal that water quality variables, especially dissolved oxygen (DO) and phosphorus content, predominantly influence the variability of carbon emissions. These factors exhibit a stronger correlation with CO2 emissions compared to CH4, indicating a net increase in carbon emissions as water quality deteriorates. Seasonally, higher water temperatures, phosphate levels, and lower DO concentrations lead to increased carbon emissions during summer months. Spatially, areas with lower carbon emissions (averaged 86 mmol m−2 d−1 CO2 and 0.13 mmol m−2 d−1 CH4) are primarily situated near the lake and in river sections where significant water quality improvements have been achieved through ecological restoration efforts. Cluster analysis shows that over 60% of high‐carbon emission (averaged 162 mmol m−2 d−1 CO2 and 1.21 mmol m−2 d−1 CH4) sites in the study area have undergone ecological restoration, suggesting potential for further carbon emission reduction through enhanced restoration practices. Our findings underscore the importance of implementing carbon reduction strategies such as nutrient removal and aeration for oxygenation within water ecological restoration initiatives. Effective matching of restoration strategies holds further potential for mitigating carbon emissions from urban river networks.

Metal-organic cages (MOCs) have been considered as emerging zero-dimensional (0D) porous fillers to generate molecularly homogenous MOC-based membrane materials. However, the discontinuous pore connectivity and low filler concentrations limit the improvement of membrane separation performance. Herein, we propose the dimension augmentation of MOCs in membranes using three-dimensional (3D) supramolecular MOC networks as filler materials in mixed matrix membranes (MMMs). We further explore the packing engineering of MOC networks to produce distinct polymorphs (α and β phases) for tailoring membrane performance. Synchrotron X-ray absorption and positron annihilation lifetime spectroscopy were employed to differentiate distinct MOC polymorphous networks within membranes. Gas permeation tests revealed that the corresponding MMMs showed superior CO2/CH4 separation performance, exceeding the Robeson upper bound. Our proposed approach is expected to enrich the repertoire of reticular chemistry pertaining to molecular-based networks.

No abstract available

This study investigated the simulation and optimization of synthetic methane production over Ni/MgAl2O4 in a multi-tubular fixed-bed reactor. The study comprises process simulation conducted using Aspen Hysys software, modeling and optimization using response surface methodology (RSM) and artificial neural network (ANN) performed using Design Experts and MATLAB software respectively. In the process simulation for the CO2methanation, sensitivity analyses were performed to determine the effects of temperature, pressure, H2/CO2 ratio, and CO fraction in the feedstock on CO2 conversion,CH4 yield, and CH4 selectivity. RSM and ANN models were built using datapoints provided by the process simulation results to model the relationship between input variables and output responses and perform optimisation for RSM model and ANN model coupled with genetic algorithm (GA). The process simulation results profoundly highlighted theimpact of temperature in enhancing CO2 conversion and CH4 yield. Higher temperatures favoured the endothermic reversed water-gas shift (RWGS) reaction, leading to increased CO2 conversion and CH4 yield. Both CO2 conversion, CH4 selectivity, and yield were found to be minimally affected by pressure. CO fraction in the feed was found to exert a delicate influence on the CO2 conversion and CH4 yield. Excessive CO fractions hindered the methanation process, reducing both CO2 conversion and CH4 yield. Additionally, the H2/CO2 ratio proved critical as higher ratios facilitated higher CO2 conversion, CH4 selectivity, and yield, emphasizing the significance of optimal hydrogen to CO2 ratio for efficient methanation which was proposed to be at values higher than the stoichiometric value of 4:1. Furthermore, the ANN-GA model outperformed RSM in terms of prediction accuracy and optimization. The ANN model demonstrated superior capabilities in capturing the complex relationships between the input variables and output responses demonstrated by the performance metrics including R2 values, MSE, RMSE etc. The optimisation results of the ANN-GA model provided more precise and efficient predictions when compared with RSM, offering a deeper understanding of the intricate interactions within the methanation process.

Pressure swing adsorption (PSA) technology is among the most efficient techniques for purifying and separating hydrogen. A layered adsorption bed composed of activated carbon and zeolite 5A for a gas mixture (H2: 56.4 mol%, CH4: 26.6 mol%, CO: 8.4 mol%, N2: 5.5 mol%, CO2: 3.1 mol%) PSA model was built. The simulation model was validated using breakthrough curves. Then, a six-step PSA cycle model was built, and the purification performance was studied. The Box–Behnken design (BBD) method was utilized in Design Expert software (version 10) to optimize the PSA purification performance. The independent optimization parameters included the adsorption time, the pressure equalization time, and the feed flow rate. Quadratic regression models can be derived to represent the responses of purity and productivity. To explore a better optimization solution, a novel optimization method using machine learning with a back propagation neural network (BPNN) was then proposed, and a kind of heuristic algorithm–genetic algorithm (GA) was introduced to enhance the architecture of the BPNN. The predicted outputs of hydrogen production using two kinds of models based on the BPNN–GA and the BBD method integrated with the BPNN–GA (BBD–BPNN–GA). The findings revealed that the BBD–BPNN–GA model exhibited a mean square error (MSE) of 0.0005, with its R–value correlation coefficient being much closer to 1, while the BPNN–GA model exhibited an MSE of 0.0035. This suggests that the BBD–BPNN–GA model has a better performance, as evidenced by the lower MSE and higher correlation coefficient compared to the BPNN–GA model.

Geopolymer Concrete (GPC) is maintainable construction material which is applied for conventional cement concrete. Various factors impact effectiveness of GPCS like types of binder material utilized, molar concentration of an activator solution as well as preserving conditions. Recently, the recycling of brickwork destruction has attracted growing attention. In this research, development of eco-friendly geopolymers feasibility is identified by the utilization of Ground Mixed Recycled Aggregates (GMRA). In this determination process, the flexural tests as well as compressive according to geopolymers are performing. The slag modification, alkali concentration as well as SiO2/Na2O ratio effects were determined. The relationship among the flexural asset as well as compressive for geopolymer based on GMRA was developed by test data fitting. In that approach, the intensity of the CO2 as well as charge effectiveness of GMRA with geopolymer was minimized as well as enhanced the slag replacement ratio. However, insignificant consequence of alkali concentration on CO2 concentration as well as charge effectiveness is experimented. The data are analyzed using Dual Quaternion Feed Forward Neural Network (DQFFNN) which obtains a less Mean Square Error (MSE) of 0.5432 compared to existing methods like Recurrent Neural Network (RNN).

This study investigates CO2 biofixation and pyrolytic kinetics of microalga G. pectorale using model-fitting and model-free methods. Microalga was grown in two different media. The highest rate of CO2 fixation (0.130 g/L/day) was observed at a CO2 concentration of 2%. The pyrokinetics of the biomass was performed by a thermogravimetric analyzer (TGA). Thermogravimetric (TG) and derivative thermogravimetric (DTG) curves at 5, 10 and 20°C/min indicated the presence of multiple peaks in the active pyrolysis zones. The activation energy was calculated by different model-free methods such as Friedman, Flynn-Wall-Ozawa (FWO), Kissinger-Akahira-Sunose (KAS), and Popescu. The obtained activation energy which are 61.7–287 kJ/mol using Friedman, 40.6–262 kJ/mol using FWO, 35–262 kJ/mol using KAS, and 66.4–255 kJ/mol using Popescu showed good agreement with the experimental values with higher than 0.96 determination coefficient (R2). Moreover, it was found that the most probable reaction mechanism for G. pectorale pyrolysis was a third-order function. Furthermore, the multilayer perceptron-based artificial neural network (MLP-ANN) regression model of the 4-10-1 architecture demonstrated excellent agreement with the experimental values of the thermal decomposition of the G. pectoral. Therefore, the study suggests that the MLP-ANN regression model could be utilized to predict thermogravimetric parameters.

No abstract available

No abstract available

Transformer is an important part of power system, its life mainly depends on the mechanical strength and electrical integrity of its insulation. CO and CO2 are one of the main fault characteristic gases dissolved in transformer oil. Thus, monitoring the CO and CO2 dissolved in transformer oil is considered by the power industry as an important means to ensure the safe operation of power systems. This work proposes the measurement of the CO and CO2 dissolved in transformer oil using Raman spectroscopy and generalized regression neural network (GRNN). Firstly, 200 samples were divided into training sets and test sets. After that, Raman spectroscopy as the input of GRNN, and The concentration of CO and CO2 dissolved in transformer oil as the output of GRNN. Afterwards, the experimental method was incorporated to get the best smoothing factor σ for GRNN. Then, the GRNN with the best smoothing factor σ=24 was trained with the training sets, and the model was verified with the test sets. The experimental results show that the prediction accuracy of the prediction model of the CO and CO2 dissolved in transformer oil based on GRNN is 90%. It also provides a new method for evaluating the health of transformer.

Because of their rich mineral resources and extreme ecological environments, deep-sea hydrothermal vent areas are significantly valuable both scientifically and economically. Optical imaging is the most direct method for verifying deep-sea hydrothermal activity. Autonomous detection and recognition of hydrothermal activity using optical imaging can significantly improve underwater robot detection efficiency. However, because of the difficulty in obtaining optical images of deep-sea hydrothermal activity and the limited accumulation of data, to date, there have been few studies on the autonomous detection of various characteristic targets of hydrothermal activity. To achieve autonomous detection and recognition of deep-sea hydrothermal polymetallic sulfides, plumes, and their biological communities, this study developed an optical image dataset of deep-sea hydrothermal activity, proposed an improved YOLOv8-based multi-target real-time detection algorithm, introduced the CBAM attention mechanism, and modified the network structure. Experimental results show that the improved algorithm achieves better target detection performance, and the response time reaches 6.8ms, which satisfies the real-time performance for underwater robots to autonomously identify hydrothermal vents.

Detecting and locating emitted fluids in the water column is necessary for studying margins, identifying natural resources, and preventing geohazards. Fluids can be detected in the water column using multibeam echosounder data. However, manually analyzing the huge volume of this data for geoscientists is a very time-consuming task. Our study investigated the use of a YOLO-based deep learning supervised approach to automate the detection of fluids emitted from cold seeps (gaseous methane) and volcanic sites (liquid carbon dioxide). Several thousand annotated echograms collected from three different seas and oceans during distinct surveys were used to train and test the deep learning model. The results demonstrate first that this method surpasses current machine learning techniques, such as Haar-Local Binary Pattern Cascade. Additionally, we thoroughly analyzed the composition of the training dataset and evaluated the detection performance based on various training configurations. The tests were conducted on a dataset comprising hundreds of thousands of echograms i) acquired with three different multibeam echosounders (Kongsberg EM302 and EM122 and Reson Seabat 7150) and ii) characterized by variable water column noise conditions related to sounder artefacts and the presence of biomass (fishes, dolphins). Incorporating untargeted echoes (acoustic artefacts) in the training set (through hard negative mining) along with adding images without fluid-related echoes are the most efficient way to improve the performance of the model and reduce the false positives. Our fluid detector opens the door for near-real time acquisition and post-acquisition detection with efficiency, reliability and rapidity.

No abstract available

Since the Industrial Revolution, methane has become the second most important greenhouse gas component after CO2 and the second most important culprit of global warming, leading to serious climate change problems such as droughts, fires, floods, and glacial melting. While most of the methane in the atmosphere comes from emissions from energy activities such as petroleum refining, storage tanks are an important source of methane emissions during the extraction and processing of crude oil and natural gas. Therefore, the use of high-resolution remote sensing image data for oil and gas production sites to achieve efficient and accurate statistics for storage tanks is important to promote the strategic goals of “carbon neutrality and carbon peaking”. Compared with traditional statistical methods for studying oil storage tanks, deep learning-based target detection algorithms are more powerful for multi-scale targets and complex background conditions. In this paper, five deep learning detection algorithms, Faster RCNN, YOLOv5, YOLOv7, RetinaNet and SSD, were selected to conduct experiments on 3568 remote sensing images from five different datasets. The results show that the average accuracy of the Faster RCNN, YOLOv5, YOLOv7 and SSD algorithms is above 0.84, and the F1 scores of YOLOv5, YOLOv7 and SSD algorithms are above 0.80, among which the highest detection accuracy is shown by the SSD algorithm at 0.897 with a high F1 score, while the lowest average accuracy is shown by RetinaNet at only 0.639. The training results of the five algorithms were validated on three images containing differently sized oil storage tanks in complex backgrounds, and the validation results obtained were better, providing more accurate references for practical detection applications in remote sensing of oil storage tank targets in the future.

Methane (CH_4) is a potent anthropogenic greenhouse gas, contributing 86 times more to global warming than Carbon Dioxide (CO_2) over 20 years, and it also acts as an air pollutant. Given its high radiative forcing potential and relatively short atmospheric lifetime (9\textpm1 years), methane has important implications for climate change, therefore, cutting methane emissions is crucial for effective climate change mitigation. This work expands existing information on operational methane point source detection sensors in the Short-Wave Infrared (SWIR) bands. It reviews the state-of-the-art for traditional as well as Machine Learning (ML) approaches. The architecture and data used in such ML models will be discussed separately for methane plume segmentation and emission rate estimation. Traditionally, experts rely on labor-intensive manually adjusted methods for methane detection. However, ML approaches offer greater scalability. Our analysis reveals that ML models outperform traditional methods, particularly those based on convolutional neural networks (CNN), which are based on the U-net and transformer architectures. These ML models extract valuable information from methane-sensitive spectral data, enabling a more accurate detection. Challenges arise when comparing these methods due to variations in data, sensor specifications, and evaluation metrics. To address this, we discuss existing datasets and metrics, providing an overview of available resources and identifying open research problems. Finally, we explore potential future advances in ML, emphasizing approaches for model comparability, large dataset creation, and the European Union's forthcoming methane strategy.

The forecasting of carbon dioxide (CO2) emissions plays a critical role in the formulation of effective climate change mitigation strategies. In this study, a comprehensive comparative analysis of hybrid statistical models is conducted by integrating Complementary Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and a range of supervised machine learning algorithms, including Support Vector Machine (SVM), Artificial Neural Network (ANN), Decision Tree (DT), Random Forest (RF), K-Nearest Neighbors (KNN), and Naïve Bayes (NB). The proposed methodology follows a two-stage procedure. Firstly, the daily carbon emission data of Russia, covering the period from January 1, 2022, to December 31, 2023, are decomposed into several Intrinsic Mode Functions (IMFs) and a single monotonic residue. Secondly, the actual carbon emission as well as the decomposed data is sliced into 80% training and 20% testing. All the hybrid models are trained on 80% data and prediction were made for the reaming 20% test data. The performance of the six hybrid models is evaluated using widely adopted performance metrics, including Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Coefficient of Determination (R2), Nash–Sutcliffe Efficiency (NSE), and Kling–Gupta Efficiency (KGE). The empirical findings demonstrate that the CEEMDAN-ANN hybrid model outperforms all other models, achieving the lowest RMSE (0.00041), MAE (0.00034), MAPE (0.0350), alongside the highest values of R2 (0.9999), NSE (0.9996) and KGE (0.9950). These results highlight the superior predictive capability of the CEEMDAN-ANN approach, underscoring its potential as an effective tool for CO2 emissions forecasting in support of climate policy and decision-making.

No abstract available

No abstract available

Climate change is considered a global disaster that has wreaked havoc worldwide. Climate change conditions are primarily driven due to emission of carbon dioxide and other greenhouse gases. Around the globe, several countries are facing serious challenges related to global warming and climate change. The adverse effects of global warming and climate change are recently observed in the form heat waves, floods, extreme rainfall events and drought conditions. The escalating CO2 emission level causes threats of food security and water scarcity. To overcome this, timely and accurate CO2 emissions forecasting is essential for developing effective mitigation strategies in such a manner that CO2 emission impacts can be reduced. The aim of this study is to forecast the total CO2 emissions of upcoming years using various statistical, machine learning models such as ARIMA (Auto-Regressive Integrated Moving Average), SARIMA (Seasonal ARIMA), Linear Regression, and Decision tree regressor. We had forecasted CO2 emission on the basis of influencing factors, which highly contributes in CO2 emission. The performance of the models is validated using multiple evaluation metrics such as MAE, MSE RMSE,MAPE and R2 square. The results show that CO2 emission is highly affected by fossil fuel consumption. In particular, Linear regression performs better among all applied models, with R2 value 0.93 and MAE, RMSE and MAPE values 5.52, 6.47 and 0.03, respectively. Our research delineates the results and insights that will assist concerned authorities, policymakers and environmental agencies. The present study also aim to inform authorities, helping them to better understand the future of CO2 emission trajectories and make timely efforts and interventions to tackle the climate change and global warming challenges.

Accurate estimation of fuel consumption and emissions is crucial for assessing the impact of materials and stringent emission control techniques on climate change, particularly in the transportation industry, which accounts for a significant portion of global greenhouse gases and hazardous pollutants emissions. To address these concerns, the government of Canada has collected a large sensor-based dataset containing detailed information on 7384 light-duty vehicles from 2017 to 2021, with the goal of reducing CO2 emissions by 40–45% by 2030. To this end, various researchers worldwide have developed vehicle emissions and consumption models to comply with these targets and achieve the Canadian government’s ambitious objectives. In this work, we propose the development of boosting and other regression models to predict carbon dioxide emissions for light-duty vehicle designs, with the aim of creating ensemble learning models that leverage vehicle specifications to forecast emissions. Our proposed boosting model is capable of accurately predicting CO2 emissions, even with only one car attribute as input. Moreover, our regression models, in conjunction with the boosting algorithm, can effectively make predictions from various vehicle inputs. Our proposed technique, categorical boosting (Catboost), provides critical insights into transportation-generated air pollution, offering valuable recommendations for both vehicle users and manufacturers. Importantly, Catboost performs data processing in less time and with less memory than other algorithms proposed in the literature. Future research efforts should focus on developing higher performance models and expanding datasets to further improve the accuracy of predictions.

No abstract available

No abstract available

This study provides a comprehensive time series analysis of daily industry-specific, country-wise CO$_2$ emissions from January 2019 to February 2023. The research focuses on the Power, Industry, Ground Transport, Domestic Aviation, and International Aviation sectors in European countries (EU27 & UK, Italy, Germany, Spain) and India, utilizing near-real-time activity data from the Carbon Monitor research initiative. To identify regular emission patterns, the data from the year 2020 is excluded due to the disruptive effects caused by the COVID-19 pandemic. The study then performs a principal component analysis (PCA) to determine the key contributors to CO$_2$ emissions. The analysis reveals that the Power, Industry, and Ground Transport sectors account for a significant portion of the variance in the dataset. A 7-day moving averaged dataset is employed for further analysis to facilitate robust predictions. This dataset captures both short-term and long-term trends and enhances the quality of the data for prediction purposes. The study utilizes Long Short-Term Memory (LSTM) models on the 7-day moving averaged dataset to effectively predict emissions and provide insights for policy decisions, mitigation strategies, and climate change efforts. During the training phase, the stability and convergence of the LSTM models are ensured, which guarantees their reliability in the testing phase. The evaluation of the loss function indicates this reliability. The model achieves high efficiency, as demonstrated by $R^2$ values ranging from 0.8242 to 0.995 for various countries and sectors. Furthermore, there is a proposal for utilizing scandium and boron/aluminium-based thin films as exceptionally efficient materials for capturing CO$_2$ (with a binding energy range from -3.0 to -3.5 eV). These materials are shown to surpass the affinity of graphene and boron nitride sheets in this regard.

No abstract available

Sustainable development and climate change are central agendas in global policy and research. This study examines and compares three ensemble learning models using Gradient Boosting Machine, Categorical Boosting, and Extreme Gradient Boosting for forecasting vehicle carbon dioxide (CO2) emissions. Data preprocessing with Interquartile Range (IQR) and median imputation is among the methods used to address missing values in CO₂ rating and smog rating variables. SHAP and PDP were employed for feature importance analysis and model interpretability. The findings from the third experiment demonstrate that Extreme Gradient Boosting (XGBoost) outperformed other models achieving a Coefficient Determination of 0.9988, Root-Mean-Square Error of 2.1696, Mean-Absolute Error of 0.4977, and Mean-Absolute-Percentage Error of 0.0019. The primary predictive features included combined fuel consumption (liters/100 km), city and highway fuel consumption, ethanol fuel consumption, model year, engine size and diesel consumption. The findings suggest the potential of boosting-based models for supporting sustainable transport planning, policy for emission reduction, and evidence-based policy making.

Air is one of the most significant elements of the environment. The increasing global air pollution crisis poses an unavoidable threat to human health, environmental sustainability, ecosystems, and the earth's climate. Air pollution has been referred to as a silent killer due to its insidious nature. Its indirect impact on human health further underscores its dangerous effects. Early detection of air quality can potentially save millions of lives globally. A unique and transformative approach can harness the power of machine learning to combat air pollution. This research presents a manual and web-based automatic prediction system that provides real-time alerts on air quality status and can help prevent premature deaths, chronic diseases, and other health problems. Air pollutants, including carbon monoxide (CO), ozone (O3), nitrogen dioxide (NO2), and particulate matter (PM 2.5), are used in this study for feature analysis and extraction. The system utilizes publicly available data from 23,463 different cities worldwide. Data preprocessing was performed before feeding the data into the machine learning models for feature correlation and evaluation. The proposed research uses various machine learning models to predict air quality, including Random Forest (100%), Logistic Regression (79%), Decision Tree (100%), Support Vector Machine (93%), Linear SVC (98%), K-Nearest Neighbor (99%), and Multinomial Naïve Bayes (52%). A user-friendly Django-based web interface offers an accessible platform for users to monitor air quality in real-time, based on the two best-performing models: Random Forest and Decision Tree techniques.

No abstract available

No abstract available

No abstract available

Background The continuous increase in carbon dioxide (CO2) emissions from fuel vehicles generates a greenhouse effect in the atmosphere, which has a negative impact on global warming and climate change and raises serious concerns about environmental sustainability. Therefore, research on estimating and reducing vehicle CO2 emissions is crucial in promoting environmental sustainability and reducing greenhouse gas emissions in the atmosphere. Methods This study performed a comparative regression analysis using 18 different regression algorithms based on machine learning, ensemble learning, and deep learning paradigms to evaluate and predict CO2 emissions from fuel vehicles. The performance of each algorithm was evaluated using metrics including R2, Adjusted R2, root mean square error (RMSE), and runtime. Results The findings revealed that ensemble learning methods have higher prediction accuracy and lower error rates. Ensemble learning algorithms that included Extreme Gradient Boosting (XGB), Random Forest, and Light Gradient-Boosting Machine (LGBM) demonstrated high R2 and low RMSE values. As a result, these ensemble learning-based algorithms were discovered to be the most effective methods of predicting CO2 emissions. Although deep learning models with complex structures, such as the convolutional neural network (CNN), deep neural network (DNN) and gated recurrent unit (GRU), achieved high R2 values, it was discovered that they take longer to train and require more computational resources. The methodology and findings of our research provide a number of important implications for the different stakeholders striving for environmental sustainability and an ecological world.

The urgent imperative to mitigate carbon dioxide (CO2) emissions from power generation poses a pressing challenge for contemporary society. In response, there is a critical need to intensify efforts to improve the efficiency of clean energy sources and expand their use, including wind energy. Within this field, it is necessary to address the variability inherent to the wind resource with the application of prediction methodologies that allow production to be managed. At the same time, to extend its use, this clean energy should be made accessible to everyone, including on a small scale, boosting devices that are affordable for individuals, such as Raspberry and other low-cost hardware platforms. This study is designed to evaluate the effectiveness of various machine learning (ML) algorithms, with special emphasis on deep learning models, in accurately forecasting the power output of wind turbines. Specifically, this research deals with convolutional neural networks (CNN), fully connected networks (FC), gated recurrent unit cells (GRU), and transformer-based models. However, the main objective of this work is to analyze the feasibility of deploying these architectures on various computing platforms, comparing their performance both on conventional computing systems and on other lower-cost alternatives, such as Raspberry Pi 3, in order to make them more accessible for the management of this energy generation. Through training and a rigorous benchmarking process, considering accuracy, real-time performance, and energy consumption, this study identifies the optimal technique to accurately model such real-time series data related to wind energy production, and evaluates the hardware implementation of the studied models. Importantly, our findings demonstrate that effective wind power forecasting can be achieved on low-cost hardware platforms, highlighting the potential for widespread adoption and the personal management of wind power generation, thus representing a fundamental step towards the democratization of clean energy technologies.

Residential energy consumption is rapidly increasing every year due to demographic and behavioral changes, such as the rising population and the adoption of work-from-home post-COVID-19. High energy consumption emits a substantial amount of carbon dioxide and other Greenhouse Gases, contributing to global warming. It becomes crucial to accurately predict residential load. To enable smart home electricity consumption control, as well as efficient generation, planning, and usage, we predict household energy consumption at very short-term, short-term, and medium-term forecast levels using univariate and multivariate time series data. This study assesses the impact of different household units (water heater and air conditioning), areas (kitchen, laundry, office, living room, bathroom, ironing room, teenager room, and parents’ room), and time (i.e., hour, day, and month) on energy consumption. Comparative analysis and numerical experimental results between the most used approaches, Support Vector Regression and Long Short-Term Memory, reveal that the former outperforms the latter across all forecast levels using different datasets. The findings of this paper will be useful to energy companies and household owners in enhancing energy efficiency and earning carbon credits by reducing the emission of carbon dioxide and other Greenhouse Gases.

India and the rest of the world are growing more and more worried about polluted atmosphere on a daily basis. A comprehensive prevision and prognostication of air quality parameters is vital due to the major harm that air pollution causes to both the environment and public health, causing concern on a global scale. In-depth analyses of the methods for predicting ambient air pollutants, like carbon monoxide (CO), sulfur dioxide (SO2), nitrogen dioxide (NO2), particulate matter with diameters less than 10μ (PM10) and less than 2.5μ (PM2.5), and ozone (O3), are provided in this work in tandem with the modeling of the Air Quality Index (AQI).To further enhance the anticipated precision and applicability of these models, the assessment additionally employs trend analysis to determine precedents and new trends in air quality. This paper offers insights into recent advances in algorithms using deep learning and machine learning for anticipating AQI and forecasting pollutant concentrations by combining current research in this topic. In order to inform policy decisions and measures aimed at reducing air pollution and its adverse effects on public health, trend analysis integration affords a more thorough comprehension of the dynamics of air quality.

As global concerns about climate change intensify, accurately predicting and managing carbon dioxide (CO2) emissions becomes paramount for sustainable environmental practices. This project employs a machine learning-driven approach to forecast CO2 emissions, utilizing historical data and advanced predictive modeling techniques. The study integrates diverse datasets, including environmental parameters and temporal information, to enhance the accuracy of predictions. The predictive models, primarily based on linear regression, are trained and evaluated using a comprehensive dataset. The approach not only explores the performance of different machine learning models but also delves into feature engineering and model interpretability. Through this research, we aim to contribute to the growing body of knowledge in environmental science and provide valuable insights for policymakers and stakeholders. The results of this study are anticipated to facilitate informed decision-making for mitigating environmental impact and promoting sustainable practices.

The transportation sector is a major contributor to carbon dioxide (CO2) emissions in Canada, making the accurate forecasting of CO2 emissions critical as part of the global push toward carbon neutrality. This study employs interpretable machine learning techniques to predict vehicle CO2 emissions in Canada from 1995 to 2022. Algorithms including K-Nearest Neighbors, Support Vector Regression, Gradient Boosting Machine, Decision Tree, Random Forest, and Lasso Regression were utilized. The Gradient Boosting Machine delivered the best performance, achieving the highest R-squared value (0.9973) and the lowest Root Mean Squared Error (3.3633). To enhance the model interpretability, the SHapley Additive exPlanations (SHAP) and Accumulated Local Effects methods were used to identify key contributing factors, including fuel consumption (city/highway), ethanol (E85), and diesel. These findings provide critical insights for policymakers, underscoring the need for promoting renewable energy, tightening fuel emission standards, and decoupling carbon emissions from economic growth to foster sustainable development. This study contributes to broader discussions on achieving carbon neutrality and the necessary transformations within the transportation sector.

The exponential escalation of carbon dioxide (CO2) emissions in the U.S. presents a pressing environmental challenge with substantial implications for climate change and public health. The principal objective of this study was to devise robust machine learning algorithms particularly designed for forecasting CO2 emissions in the United States. This focused exclusively on CO2 emission data pertinent to America, reflecting the economic, unique environmental, and regulatory context of the nation. The dataset for analysis consisted of a broad-based set of information focused on the main contributors of CO2 emissions in the United States, ranging from energy consumption and industrial activity to transportation and historical CO2 emission data. The energy consumption data included facts on electricity generated, fuel consumed, and absolute energy consumption among different sectors of the economy, and industrial activities information provides data on specific outputs from such processes and their emissions. It also included transportation facts on vehicle trends, fuel intensity, and energy-related emissions associated with the sector. These three datasets have been garnered from reliable resources, including the US. These range from detailed EPA emissions inventories and energy reports from the U.S. The analyst deployed credible algorithms such as Random Forest, Logistic Regression, and Support Vector Classifier which had different strengths that can be leveraged based on characteristics of the dataset. According to their accuracy scores, the Random Forest model led the race compared to the other two models, with a higher accuracy rate. With such large integrations of machine learning predictions into climate policy, great opportunities might develop vis-à-vis sustainable development goals in the USA. Advanced analytics will let the policy analyst capture emission and resource trends with greater insight than ever before into the effectiveness of existing regulations; this will let it plug into the SDGs on Climate Action, Sustainable Cities, and Responsible Consumption. For enhancing environmental monitoring systems' efficiency, environmental planning should be incorporated with machine learning models.

Cities and buildings represent the core of human life, the nexus of economic activity, culture, and growth. Although cities cover less than 10% of the global land area, they are notorious for their substantial energy consumption and consequential carbon dioxide (CO2) emissions. These emissions significantly contribute to reducing the carbon budget available to mitigate the adverse impacts of climate change. In this context, the designers’ role is crucial to the technical and social response to climate change, and providing a new generation of tools and instruments is paramount to guide their decisions towards sustainable buildings and cities. In this regard, data-informed digital tools are a viable solution. These tools efficiently utilise available resources to estimate the energy consumption in buildings, thereby facilitating the formulation of effective urban policies and design optimisation. Furthermore, these data-driven digital tools enhance the application of algorithms across the building industry, empowering designers to make informed decisions, particularly in the early stages of design. This paper presents a comprehensive literature review on artificial intelligence-based tools that support performance-driven design. An exhaustive keyword-driven exploration across diverse bibliographic databases yielded a consolidated dataset used for automated analysis for discerning the prevalent themes, correlations, and structural nuances within the body of literature. The primary findings indicate an increasing emphasis on master plans and neighbourhood-scale simulations. However, it is observed that there is a lack of a streamlined framework integrating these data-driven tools into the design process.

In the face of global climate change, accurately predicting carbon dioxide emissions has become an urgent requirement for environmental science and policy-making. This article provides a systematic review of the literature on carbon dioxide emission forecasting, categorizing existing research into four key aspects. Firstly, regarding model input variables, a thorough discussion is conducted on the pros and cons of univariate models versus multivariable models, balancing operational simplicity with high accuracy. Secondly, concerning model types, a detailed comparison is made between statistical methods and machine learning methods, with a particular emphasis on the outstanding performance of deep learning models in capturing complex relationships in carbon emissions. Thirdly, regarding model data, the discussion explores annual emissions and daily emissions, highlighting the practicality of annual predictions in policy-making and the importance of daily predictions in providing real-time support for environmental policies. Finally, regarding model quantity, the differences between single models and ensemble models are examined, emphasizing the potential advantages of considering multiple models in model selection. Based on the existing literature, future research will focus on the integration of multiscale data, optimizing the application of deep learning models, in-depth analysis of factors influencing carbon emissions, and real-time prediction, providing scientific support for a more comprehensive, real-time, and adaptive response to the challenges of climate change. This comprehensive research outlook aims to provide scientists and policymakers with reliable information on carbon emissions, promoting the achievement of environmental protection and sustainable development goals.

Contemporarily, Asia is the world's largest carbon emitter among other continents according to the recorded data, which plays an essential role in climate change. This paper introduces Asia's carbon dioxide emission effect and mainly focuses on the emission prediction for the future five years based on the time series data from 1950 to 2021. Some statistical models are implemented in forecasting, including ARIMA, SARIMA, and SARIMAX models. Besides, two general machine learning models, linear regression and random forest regression would also be applicable in this paper. According to the analysis, based on the model performance matrix, the study found that the SARIMA model is the optimal model to explain and forecast the data. Specifically, it has the lowest MSE, MRSE, and MAE values and the highest AIC and BIC scores. These results shed light on guiding further exploration of carbon dioxide emission prediction in the next five years in Asia.

This study aims to assess the current state and progress of academic research on using machine learning techniques to simulate the problems associated with air pollution. Air is the composition of Nitrogen, oxygen and various other gases. The excessive presence of few gases namely NO2, CO, SO2 etc., causes air pollution which has a significant influence on the atmosphere . This leads to global warming , acid rain causing serious health effects. An efficient Air Pollution Monitoring System is essential to avoid imbalance in nature. Therefore, there is a need to put forward a Real -Time Air Pollution Monitoring and Prediction of the Air Quality Index (AQI) of Carbon Monoxide (CO), Sulphur Dioxide (SO2) and Nitrogen Dioxide (NO2) by introducing a Forecast BySsa , a machine learning forecasting uncertainties may arise in the results of air pollution models due to gaps in the data used for their development. Singular Spectrum Analysis (SSA) is a non-parametric method that offers several advantages over the other machine learning algorithms , which eliminates the need for assumptions about the underlying data distribution, making it more flexible than other parametric methods such as linear regression , which can be constrained by data assumptions . It can be used to filter out noise and other unwanted components from time series data, allowing researchers to focus on the underlying signals of interest . Overall , SSA is a powerful data analysis technique that offers several advantages , particularly for analyzing time series data. algorithm for environmental protection which is cost effective, reliable, scalable and accurate. An IOT device is utilized as the prototype to collect data from gas sensors such as MQ2, MQ9, MQ 135. A web application that displays the air pollution statistics for the next 10 days is created using . NET Framework , for further analysis and prediction. This paper emphasizes on the usage of Singular Spectrum Analysis for training and evaluation of the machine learning model . In addition, SSA has an advantage of working with long range non-linear datasets.

The amount of carbon dioxide in the atmosphere has risen over recent years, with a growth of over 40%. This study examines transportation-related carbon dioxide (CO 2 ) emissions in Canada, which contribute significantly to the country’s overall emissions. The study investigates the rise of carbon dioxide (CO 2 ) due to various reasons such as economic development, transportation, as well as population growth, but the study focuses on transportation-related CO 2 emission in Canada. Various machine learning algorithms, such as Deep Neural Networks, Support Vector Machines, and Random Forests, are utilized to forecast CO 2 emissions. The results show promising outcomes, with R2 values ranging from 0.9532 to 0.9996, RMSE values ranging from 1.0974 to 13.6561, MAPE scores from 0.0088 to 0.0010, MBE scores ranging from -0.0594 to 1.0366, rRMSE score ranging from 0.4259 to 5.3002, and MABE score ranging from 0.2643 to 5.6582 for all six (6) algorithms. To meet greenhouse gas reduction targets, this paper recommends further efforts to reduce CO 2 emissions from transportation sources and suggests the adoption of Vehicle Alternative Fuel Types and low-carbon fuels.

The broad use of wind power plants is a result of the rising need for renewable energy. However, it is difficult to effectively harness wind energy due to the inconsistent and erratic behavior of the wind. Improved wind energy system effectiveness depends on reliable wind speed forecasting. We suggest a unique marine predator-optimized convolutional deep belief network (MP-CDBN) in this study for predicting wind speed. The MPO technique is employed for optimizing the MP-CDBN framework once it has been trained using prior wind data. An evaluation and comparison of the suggested model with other wind speed prediction techniques are conducted. The suggested MP-CDBN model's precise wind speed predictions have the potential to increase the effectiveness of wind energy installations. The suggested approach can aid in lowering carbon dioxide emissions and encouraging the production of renewable energy by increasing the effectiveness of wind power plants.

In this article, an experimental investigation is done to study the impact of carbon dioxide (CO2) vehicle emissions on global climatic changes. The dataset for the analysis is taken from the Canadian Ministry of Transportation. Forecasting carbon emissions from automobiles is studied using five prominent ML models: regression models, decision trees, random forests, and neural networks. Variables like vehicle characteristics, fuel type, mileage, driving habits, and traffic conditions are considered for forecasting CO2 emissions. The study also discusses the possible advantages of including more variables in the predictive models, such as weather, road conditions, and vehicle fleet composition, which can improve forecast accuracy and make it possible for decision-makers to pinpoint specific emission-reduction measures. This research study highlights how machine learning could help stakeholders and policymakers make informed choices to develop low-carbon transportation networks.

Abstract Background Northern Australia’s tropical savannas (~118 M ha), dominated by high biomass grassy Eucalypt woodlands, are amongst the world’s most frequently burned landscapes. Fire severity strongly influences biodiversity and greenhouse gas emissions but remains difficult to map consistently at biome scales. Aims To develop and validate a biome-scale fire severity mapping framework that generates annual fire severity mapping across north Australia, using MODIS 250 m reflectance data, moderate scale burnt area mapping, active fire data, and a large field dataset of fire impacts. Methods We derived the relative difference in the near infrared (RdNIR) for 10-day median end-of-month image composites, stratified by season, in 3 distinct pyro-geographical regions, and undertook a Supervised classification from extensive field data. Results The calibrated approach achieved 93% overall accuracy. Adjusted thresholds reduced overall over-classification of severe fires in low-canopy environments, markedly improving correspondence with field data. Implications This framework delivers a biome-scale fire severity map for Australia’s tropical savannas, critical for biodiversity conservation, emissions accounting, and Indigenous-led savanna burning projects. Limitations include reliance on NIR alone. Incorporation of fractional cover and SWIR data represents a priority for future refinement. Extensive field data is key to accuracy.

Cattle farming is responsible for 8.8\% of greenhouse gas emissions worldwide. In addition to the methane emitted due to their digestive process, the growing need for grazing areas is an important driver of deforestation. While some regulations are in place for preserving the Amazon against deforestation, these are being flouted in various ways, hence the need to scale and automate the monitoring of cattle ranching activities. Through a partnership with \textit{Global Witness}, we explore the feasibility of tracking and counting cattle at the continental scale from satellite imagery. With a license from Maxar Technologies, we obtained satellite imagery of the Amazon at 40cm resolution, and compiled a dataset of 903 images containing a total of 28498 cattle. Our experiments show promising results and highlight important directions for the next steps on both counting algorithms and the data collection process for solving such challenges. The code is available at \url{this https URL}.

Given that soil carbon sequestration is now widely recognized as an important tool for reducing greenhouse gas emissions and improving soil health, encouragement to adopt such strategies is necessary as the acceleration of climate change. Due to their vast and complex nature, agroforestry systems that are known for their potential. The remote sensing approach for monitoring the soil carbon sequestration in agroforestry systems presented in this study is based on a satellite. The satellite imagery using which the proposed framework is integrated is combined with ground truth data to estimate radiometric calibration, atmospheric correction, vegetation indices, soil carbon. NDVI (Normalized Difference Vegetation Index) values are thoroughly analyzed by correlating this data to soil carbon in order to understand in detail carbon sequestration on various agroforestry practices. Overall, the results show that mixed species and intercropping agroforestry systems have much higher soil carbon storage capabili3ty than single species or non agroforestry practices. Satellite based remote sensing offers both advantages and disadvantages when it comes to informing soil carbon dynamics of this study shows. With these findings, it becomes evident that diversified agroforestry are absolutely crucial in the mitigation of climate change and enhancement of soil quality. This approach enables one to base decisions on sustainable land management on satellite data and reinforces the role that agroforestry can play as a transformative intervention aimed at enhancing environmental sustainability.

No abstract available

Land use change in agriculture and forestry is a key sector in the IPCC reports on climate change, as it significantly impacts greenhouse gas emissions and carbon sequestration. Advances in remote sensing technology and new satellite generations provide opportunities for more frequent and detailed monitoring of land cover. This study focuses on analyzing and evaluating the effectiveness of machine learning models, including Random Forest (RF), Boosted Tree (BT), and Support Vector Machine (SVM), for land cover classification in Dak Nong. Sentinel-2 MSI and ALOS World 3D data were used in machine learning classification models. Results indicate that RF is the most effective classification model, achieving the highest overall accuracy of 78.6% with strong stability, and Kappa value of 0.75. BT classifier also performed well on dry-season imagery, while SVM exhibited lower accuracy due to its sensitivity to parameter settings. The findings highlight the importance of integrating multi-seasonal imagery with spectral indices and topographic data to improve classification accuracy in complex landscapes as well as their potential application to support AFOLU-related emissions inventory.

Monitoring and understanding Land Use/Land Cover (LU/LC) is critical for sustainable development, as it can impact various environmental, social, and economic systems. For example, deforestation and land degradation can lead to soil erosion, loss of biodiversity, and greenhouse gas emissions, affecting the quality of soil, air, and water resources. The present research examined changes in (LU/LC) within the underdeveloped regions of Balochistan and Sindh provinces, which are situated in Pakistan. In order to monitor temporal variations of LU/LC, we employed Geographic Information System (GIS) technique, to conduct an analysis of satellite imagery obtained from the Landsat 8 Operational Land Imager (OLI) during the time period spanning from 2013 to 2023. In order to obtain an accurate LU/LC classification, we used principal component analysis (PCA) and a supervised classification approach using the maximum likelihood algorithm (MLC). According to the results of our study, there was a decrease in the extent of water bodies (− 593.24 km2) and vegetation (− 68.50 km2) by − 3.43% and − 0.40% respectively. In contrast, the area occupied by settlements in the investigated region had a 2.23% rise, reaching a total of 385.66 square kilometers. Similarly, the extent of barren land also expanded by 1.60%, encompassing a total area of 276.04 square kilometers, during the course of the last decade. The overall accuracy (94.25% and 95.75%) and K value (91.75% and 93.50%) were achieved during the year 2013 and 2023 respectively. The enhancement of agricultural output in Pakistan is of utmost importance in order to improve the income of farmers, mitigate food scarcity, stimulate economic growth, and facilitate the expansion of exports. To enhance agricultural productivity, it is recommended that the government undertake targeted initiatives that aimed at enhancing water infrastructure and optimizing land use to foster a sustainable ecological framework. Integrating the sustainable ecological framework provides a foundation for informed decision-making and effective resource management. By identifying areas of urban expansion, agricultural intensification, or alterations in natural LU/LC, stakeholders can design targeted conservation strategies, mitigating potential environmental degradation and promoting biodiversity conservation. In conclusion, the integration of GIS and Remote Sensing (RS) may effectively facilitate the monitoring of land use patterns over a period of time. This combined approach offers valuable insights and recommendations for the judicious and optimal management of land resources, as well as informing policy decisions.

Solar energy plays a crucial role in mitigating greenhouse gas emissions in the context of global climate change. However, its deployment for green electricity generation can significantly influence regional climate and vegetation dynamics. While prior studies have examined the impacts of solar power plants on vegetation, the accuracy of these assessments has often been constrained by the availability of publicly accessible multispectral, high-resolution remotely sensed imagery. Given the abundant solar energy resources and the ecological significance of the Tibetan Plateau, a thorough evaluation of the vegetation effects associated with solar power installations is warranted. In this study, we utilize sub-meter resolution imagery from the GF-2 satellite to reconstruct the fractional vegetation cover (FVC) at the Gonghe solar thermal power plant through image classification, in situ sampling, and sliding window techniques. We then quantify the plant’s impact on FVC by comparing data from the pre-installation and post-installation periods. Our findings indicate that the Gonghe solar thermal power plant is associated with a 0.02 increase in FVC compared to a surrounding control region (p < 0.05), representing a 12.5% increase relative to the pre-installation period. Notably, the enhancement in FVC is more pronounced in the outer ring areas than near the central tower. The observed enhancement in vegetation growth at the Gonghe plant suggests potential ecological and carbon storage benefits resulting from solar power plant establishment on the Tibetan Plateau. These findings underscore the necessity of evaluating the climate and ecological impacts of renewable energy facilities during the planning and design phases to ensure a harmonious balance between clean energy development and local ecological integrity.

A significant amount of global greenhouse gas (GHG) emissions comes from Indonesia, largely driven by deforestation and land degradation. As a developing nation, it is also dealing with the growing pressures of urban expansion. This study assesses the distribution of carbon stock in Parepare City, South Sulawesi, Indonesia. Notably, Parepare City has not yet experienced extensive land-use transformations, retaining substantial carbon stock, which positions it as a proactive case study for preventing future carbon loss amidst ongoing urbanization. Using the InVEST Carbon Storage and Sequestration model with SPOT 7 satellite imagery (2016) and global carbon density data, the research quantifies carbon storage across various land use/land cover (LULC) types. Analysis reveals natural ecosystems, particularly mixed forests and fields, hold the highest carbon storage potential. The total estimated carbon stock in Parepare City is 1,456,909.41 Mg C. These findings emphasize the urgent need for climate-responsive land management, including forest conservation, and urban greening, to enhance local carbon sinks and support Indonesia's climate change mitigation goals. This assessment provides crucial insights for urban planners and policymakers to balance growth with ecosystem conservation for a susta00inable future.

Loss of reactive nitrogen (N) from agricultural fields in the U.S. Midwest is a principal cause of the persistent hypoxic zone in the Gulf of Mexico. We used eight years of high resolution satellite imagery, field boundaries, crop data layers, and yield stability classes to estimate the proportion of N fertilizer removed in harvest (NUE) versus left as surplus N in 8 million corn (Zea mays) fields at subfield resolutions of 30 × 30 m (0.09 ha) across 30 million ha of 10 Midwest states. On average, 26% of subfields in the region could be classified as stable low yield, 28% as unstable (low yield some years, high others), and 46% as stable high yield. NUE varied from 48% in stable low yield areas to 88% in stable high yield areas. We estimate regional average N losses of 1.12 (0.64–1.67) Tg N y−1 from stable and unstable low yield areas, corresponding to USD 485 (267–702) million dollars of fertilizer value, 79 (45–113) TJ of energy, and greenhouse gas emissions of 6.8 (3.4–10.1) MMT CO2 equivalents. Matching N fertilizer rates to crop yield stability classes could reduce regional reactive N losses substantially with no impact on crop yields, thereby enhancing the sustainability of corn-based cropping systems.

Alfalfa is the most widely grown forage crop worldwide and is thought to be a significant carbon sink due to high productivity, extensive root systems, and nitrogen-fixation. However, these conditions may increase nitrous oxide (N_2O) emissions thus lowering the climate change mitigation potential. We used a suite of long-term automated instrumentation and satellite imagery to quantify patterns and drivers of greenhouse gas fluxes in a continuous alfalfa agroecosystem in California. We show that this continuous alfalfa system was a large N_2O source (624 ± 28 mg N_2O m^2 y^−1), offsetting the ecosystem carbon (carbon dioxide (CO_2) and methane (CH_4)) sink by up to 14% annually. Short-term N_2O emissions events (i.e., hot moments) accounted for ≤1% of measurements but up to 57% of annual emissions. Seasonal and daily trends in rainfall and irrigation were the primary drivers of hot moments of N_2O emissions. Significant coherence between satellite-derived photosynthetic activity and N_2O fluxes suggested plant activity was an important driver of background emissions. Combined data show annual N_2O emissions can significantly lower the carbon-sink potential of continuous alfalfa agriculture. Long-term continuous greenhouse gas measurements in alfalfa cropland showed that the magnitude of the carbon sink was significantly offset by large nitrous oxide (N_2O) emission events following irrigation and rainfall.

The issue of global warming continues to be a concern for the international world. One of the causes of global warming is the greenhouse gas (GHG) effect caused by an increase in the amount of emissions in the atmosphere. Indonesia has set targets for reducing GHG emissions in the Nationally Determined Contribution (NDC) by 31.89% on its own capability and 43.20% with international assistance in 2030. Therefore, to support the program, it is necessary to conduct research to inventory the emissions produced and emissions absorbed by each region in Indonesia. This study calculated the net-zero emission index in an area by comparing the value of carbon emission with the value of carbon sequestration based on land cover in an area. Regional emission data was obtained from Aksara Bappenas, while regional sequestration data was obtained from RAD-GRK and satellite imagery interpretation based on land cover. Meanwhile, to calculate emissions at the city/regency level, individual carbon footprint data was used and converted into city/regency emission data. The calculation of the carbon index was carried out nationally in all provinces of Indonesia. Several provinces were further tested as examples to determine sequestration based on land cover from imagery interpretation. The results show that the Indonesian net-zero emission index is lower than 1, but some provinces with dense populations have a net-zero emission index >1, namely: DKI Jakarta, DI Yogyakarta, North Sumatra, and Riau.

Land use change in the Bangsri sub-catchment of the Brantas basin in East Java during the past three decades has increased both local and global climate vulnerability, by degrading soils and contributing to net greenhouse gas emissions, respectively. Our study aimed to estimate land use change and its impact on net carbon emissions, as well as to formulate strategies for adaptation to and mitigation of climate change downstream of a National park. We analyzed land cover changes from satellite imagery and measured carbon stocks in biomass, necromass and soil (0 to 30 cm) pools. Satellite imagery of land cover in 1994, 2001, 2011 and 2017 showed a decrease in natural forest area and an increase in the area of shrubs, agroforestry, production forests and annual crop land. Net CO2 emission increased from 2.4 to 6.4 Mg ha−1 year−1 in the periods 2001 to 2011 and 2011 to 2017, respectively. Sand mining is the most destructive land use pattern in the area, as it leaves soil profiles stripped of their topsoil. Vulnerability to less reliable rainfall has been addressed by the common creation of small reservoirs and the abundant use of irrigation for vegetables growing under partial shade in the agroforestry zone.

The forestry and Other Land Use (FOLU) sector in Indonesia is expected to contribute 60% of the greenhouse gas emission reduction. The priority location for enhancing carbon stock as the mitigation action is natural forests, such as Mount Halimun Salak National Park (MHSNP). The objectives of this research were to estimate vegetation cover changes in 2016, 2019, 2022, and to analyze the forestry programs affecting forest cover in MHSNP. This research used Landsat 8 satellite imagery. NDVI was categorized into five classes, specifically class 1 (the clouds/non-vegetation), class 2 (very low dense vegetation), class 3 (low dense vegetation), class 4 (moderately dense vegetation), and class 5 (highly dense vegetation). MHSNP vegetation cover consecutive in 2016, 2019 and 2022 is dominated by class 1 (35,94% or 31.508,45 ha), class 2 (30,86% or 27.053,73 ha), and class 5 (58,76% or 51.543,18%), respectively. In general, the large vegetation cover from 2016 to 2022 is increasing from 87.662,06 ha to 87.716,88 ha and is significantly denser. It might be caused by the success of the forestry program in MHSNP, such as increasing the rehabilitation area, tree adoptions, and restoration in collaboration with communities and companies.

Gas flaring is an environmental problem of local, regional and global concerns. Gas flares emit pollutants and greenhouse gases, yet knowledge about the source strength is limited due to disparate reporting approaches in different geographies, whenever and wherever those are considered. Remote sensing has bridged the gap but uncertainties remain. There are numerous sensors which provide measurements over flaring-active regions in wavelengths that are suitable for the observation of gas flares and the retrieval of flaring activity. However, their use for operational monitoring has been limited. Besides several potential sensors, there are also different approaches to conduct the retrievals. In the current paper, we compare two retrieval approaches over an offshore flaring area during an extended period of time. Our results show that retrieved activities are consistent between methods although discrepancies may originate for individual flares at the highly temporal scale, which are traced back to the variable nature of flaring. The presented results are helpful for the estimation of flaring activity from different sources and will be useful in a future integration of diverse sensors and methodologies into a single monitoring scheme.

To enhance our understanding of the sources and sinks of greenhouse gases (GHGs), particularly carbon dioxide and methane, it is vital to monitor their atmospheric concentrations in near real time and at high spatial resolution. This study proposes a novel approach involving a constellation of miniature satellites equipped with compact spectrometers possessing nanometer-scale spectral resolution. Climate change, one of the most pressing global challenges, exerts profound impacts across environmental, social, and economic domains. Effective monitoring and assessment are therefore essential to inform policy decisions and mitigate its consequences. Remote sensing technology has emerged as an indispensable tool in climate change research, offering the ability to observe, evaluate, and predict environmental changes on a global scale. By utilizing satellite imagery, aerial surveys, and other sensing methods, scientists and policymakers can collect robust datasets, monitor long-term climate trends, and make evidence-based decisions. The integration of miniaturized satellite spectrometers represents a significant advancement in the effort to improve the timeliness and accuracy of GHG monitoring.

No abstract available

ABSTRACT Clark, A.; Moorman, B.; Whalen, D., and Vieira, G., 2025. Segmentation and classification of Pléiades Satellite imagery for complex shoreline proxy delineation in the Western Canadian Arctic. Journal of Coastal Research, 41(3), 391–408. Charlotte (North Carolina), ISSN 0749-0208. Permafrost coasts are vulnerable to the effects of climate change, including above average warming in the Arctic, sea-level rise, and changes to sea-ice extent and duration. As a result, coastal erosion represents a prominent hazard that impacts communities and habitats locally, but it also potentially releases significant amounts of organic carbon that are consequential to understanding global greenhouse gas emissions, which requires appropriate quantification and monitoring. The extent and complexity of Arctic coasts represent a challenge for effective broad-scale monitoring using traditional methods of manual coastline delineation. In this study, an alternative to manual coastline delineation is presented using object-based image analysis to classify very high resolution Pléiades satellite imagery (0.5 m/pixel) scenes and subsequently extract two common coastline proxies, the tundra line and waterline.This study focused on three large, varied, coastal stretches in the Western Canadian Arctic: the Yukon North Slope, Tuktoyaktuk Peninsula, and Darnley Bay coasts. Twenty-five (25) images were classified with high accuracy (92% average), while the majority of extracted coastline proxies were within 3.0 m of the reference features and in many scenarios had accuracies better than 1.5 m, which is comparable to expected digitizing error, or loss of accuracy, associated with the variance of repeated shoreline digitization of the same coastal extent. This work presents an important step towards broad-scale Arctic coastal change monitoring and quantification through semi-automated classification and feature extraction techniques, which will also enhance contextual information useful in conducting richer Arctic coastal erosion studies.

Abstract. Observation-based monitoring of the status of greenhouse gas emissions goals set at the 2015 Paris Climate Summit is critical to provide timely, accurate and precise information on the progress towards these goals. Observations also permit the identification of potential deviations from the adopted policies that could compromise the efforts to reduce the future impact of pollutants on the climate. Current remote sensing capabilities of atmospheric CO2 have demonstrated the ability to estimate emissions from the strongest sources of CO2 based on imagery of individual plumes in conjunction with wind speed estimates. However, a realistic evaluation of the accuracy of the obtained estimates is essential. Here, we examine how the stochastic nature of daytime atmospheric turbulence affects the estimation of CO2 emissions from a lignite coal power plant in Bełchatów, Poland. For this investigation, we use a high-resolution (400 m × 400 m × 85 levels) atmospheric model set up in a realistic configuration. We demonstrate that persistent structures in the downwind concentration fields of emitted plumes can cause significant uncertainties in the retrieved fluxes, on the order of 10 % of the total source strength, when the commonly used cross-sectional mass-flux (CSF) method is applied with short distances between individual estimates. These form a significant contribution to the overall uncertainty which remains unavoidable in the presence of atmospheric turbulence. Furthermore, we applied temporally tagged tracers for the decomposition of the plume variability into its constituent parts. These tracers helped us to explain why spatial scales of variability in plume intensity are far larger than the size of turbulent eddies – a finding that challenges previous assumptions.

The construction of the sea dike and Semarang-Demak toll road has severed the mangrove ecosystem inside the dike, as well as increased greenhouse gas impacts due to transportation activities and the growth of built-up areas around the dike and toll road. The aim of this research is to formulate a regression model based on spatial data that can be used to measure the impact of transportation activities and building intensity on LST. The data used in this study are the number of motorized vehicles crossing the main roads in Semarang City and LST obtained from the Landsat 8 thermal infrared sensor band in 2013 and 2019. This research utilizes Geographic Information System, Remote Sensing, and statistical methods to model the environmental impact of the sea dike and toll road development. This model used to predict the environmental impact of the sea dike and Semarang-Demak Toll Road in the future. The result shows that the increase in the number of motorized vehicles and building intensity has a high contribution to LST. Every additional 1,000 passenger cars on a road will make LST increase from 0.0150C to 0.0380C, whereas every 10% increase in land intensity will make LST increase by 0.030C. In addition, there is an increase in the LST value of 300C from 260C previously. This model is expected to provide input for each stakeholder to mitigate the potential environmental impacts of the Semarang-Demak Sea dike and toll road in the future, and hope that the Semarang-Demak Sea dike.

The mangrove vegetation within the Niger Delta region of Nigeria is ravaged by anthropogenic practices including but not limited to rapid urbanization, aquaculture expansion and oil exploration which penultimately distorts the biodiversity of both the mangrove and marine environments, culminating in the loss of structural and functional integrity of these ecosystems, specifically their role in climate change regulation. The study aimed at assessing the changes in mangrove covers from 1987 to 2022 in the study area as well as examining the changes in GHGs emissions resulting from the mangrove changes. The methodology adopted a remote sensing-based research design utilizing satellite imagery to analyze temporal changes in mangrove cover and evaluated their association with climate variables such as CO2 emissions and LST of the study area. Each satellite image geo-referenced in ArcGIS 10.8 & LULC changes calculated using geometry module of ArcGIS 10.8. NDIR spectroscopy was used in examining the variation in GHGs emissions. The data obtained revealed mangrove reduction from 12,991 km2 in 1987 to 9,089km2 in 2022 resulting in the loss of 3,904.00 km2 of mangrove forest. The reduction resulted in increased CO₂ emissions from 370.70 ppm to 403.29 ppm between 1987 and 2022. These results illustrate a clear link between mangrove cover change and CO₂ emissions, highlighting the critical role mangroves play in regulating climate change. The study was able to show that significant losses in mangrove cover have been closely associated with increased CO₂ emissions, thus reflecting the vital role these ecosystems play in carbon sequestration which underscores the importance of preserving these vital ecosystems to mitigate local and global climate impacts.

Agricultural ponds are a significant source of greenhouse gases, contributing to the ongoing challenge of anthropogenic climate change. Nations are encouraged to account for these emissions in their national greenhouse gas inventory reports. We present a remote sensing approach using open-access satellite imagery to estimate total methane emissions from agricultural ponds that account for (1) monthly fluctuations in the surface area of individual ponds, (2) rates of historical accumulation of agricultural ponds, and (3) the temperature dependence of methane emissions. As a case study, we used this method to inform the 2024 National Greenhouse Gas Inventory reports submitted by the Australian government, in compliance with the Paris Agreement. Total annual methane emissions increased by 58% from 1990 (26 kilotons CH4 year–1) to 2022 (41 kilotons CH4 year–1). This increase is linked to the water surface of agricultural ponds growing by 51% between 1990 (115 kilo hectares; 1,150 km2) and 2022 (173 kilo hectares; 1,730 km2). In Australia, 16,000 new agricultural ponds are built annually, expanding methane-emitting water surfaces by 1,230 ha yearly (12.3 km2 year–1). On average, the methane flux of agricultural ponds in Australia is 0.238 t CH4 ha–1 year–1. These results offer policymakers insights into developing targeted mitigation strategies to curb these specific forms of anthropogenic emissions. For instance, financial incentives, such as carbon or biodiversity credits, can mobilize widespread investments toward reducing greenhouse gas emissions and enhancing the ecological and environmental values of agricultural ponds. Our data and modeling tools are available on a free cloud-based platform for other countries to adopt this approach.

Solar Photovoltaic (PV) technology is increasingly recognized as a pivotal solution in the global pursuit of clean and renewable energy. This technology addresses the urgent need for sustainable energy alternatives by converting solar power into electricity without greenhouse gas emissions. It not only curtails global carbon emissions but also reduces reliance on finite, non-renewable energy sources. In this context, monitoring solar panel farms becomes essential for understanding and facilitating the worldwide shift toward clean energy. This study contributes to this effort by developing the first comprehensive global dataset of multispectral satellite imagery of solar panel farms. This dataset is intended to form the basis for training robust machine learning models, which can accurately map and analyze the expansion and distribution of solar panel farms globally. The insights gained from this endeavor will be instrumental in guiding informed decision-making for a sustainable energy future. https://github.com/yzyly1992/GloSoFarID

Assessing the aboveground biomass and carbon stock of rubber plantation areas accurately and precisely is crucial for utilizing them as tools to drive low-carbon societies in line with the Sustainable Development Goals (SDGs). Goal 13 specifically targets reducing greenhouse gas emissions. This study classified rubber plantation areas using satellite image data from Planet NICFI through Random Forest (RF) classification. Additionally, it evaluated the above-ground biomass and soil carbon assessment of rubber plantation areas in Bueng Kan Province using GEDI data combined with random forest regression (RFR). The study found that the overall accuracy and kappa of rubber plantation area classification in Bueng Kan Province exceeded 85%. Meanwhile, for above-ground biomass assessment and soil carbon sequestration in rubber plantation areas, the model’s correlation coefficient (R2) was 0.77, indicating a relatively high level of accuracy with a root mean square error (RMSE) of 65.98 megagrams per hectare. The spatial distribution of carbon sequestration ranged from 15.16 to 212.28 megagrams per hectare. These results can support biomass and carbon stock education technology for use in formulating spatial management policies and plans from regional to local levels and serve as tools to drive low-carbon societies alongside sustainable environmental management.

Climate change is mainly anthropogenic mostly caused by urbanization, human activities in economics, industry, and transportation. The expansion of built-up land, deforestation and the loss of farmland are closely linked to land use and land use change. Greenhouse gas emissions produced by the land use sector can significantly affect global carbon budgets by changing the carbon storage level in terrestrial ecosystem vegetation and soil. In 2005, Indonesia was responsible for approximately 85% of carbon emissions. The Indonesian government is combating environmental issues by mandating local governments, including Palembang City, to conduct greenhouse gas inventories. Changes in land use and the amount of carbon stock in Palembang City can be taken into consideration by the Palembang City Government in dealing with climate change. Data analysis was carried out by interpreting satellite imagery SPOT-7 and classification of land use data into six classes based on AFOLU guidelines. The area derived from land use transition matrix of the period 2012-2018 is used as a basis to calculate greenhouse gas emissions. The greenhouse gas emissions were then calculated using the Gain-Loss method based on the IPCC journal as a reference. Due to land use and land use change from 2012 to 2018, Palembang City emits greenhouse gas as much as -149098.5827 Tonnes C/Year in total. Forest Land Category -26557.22425 Tonnes C/Year, Crop Land Category -112739.8894 Tonnes C/Year, Grass Land Category -32257.56413 Tonnes C/Year, Wetland Category -20721.68315 Tonnes C/Year, Settlement Category 43273.249 Tonnes C/Year and Other Land Category -95.4708 Tonnes C/Year. Inventories on greenhouse gas (GHG) emissions and absorption trends are crucial for climate change mitigation strategies in Palembang. One important strategy towards achieving net zero emissions by 2060, as initiated by the Government of Indonesia, is to curb carbon release associated with land use changes.

Tracking and measuring national carbon footprints is key to achieving the ambitious goals set by the Paris Agreement on carbon emissions. According to statistics, more than 10% of global transportation carbon emissions result from shipping. However, accurate tracking of the emissions of the small boat segment is not well established. Past research looked into the role played by small boat fleets in terms of greenhouse gases, but this has relied either on high-level technological and operational assumptions or the installation of global navigation satellite system sensors to understand how this vessel class behaves. This research is undertaken mainly in relation to fishing and recreational boats. With the advent of open-access satellite imagery and its ever-increasing resolution, it can support innovative methodologies that could eventually lead to the quantification of greenhouse gas emissions. Our work used deep learning algorithms to detect small boats in three cities in the Gulf of California in Mexico. The work produced a methodology named BoatNet that can detect, measure and classify small boats with leisure boats and fishing boats even under low-resolution and blurry satellite images, achieving an accuracy of 93.9% with a precision of 74.0%. Future work should focus on attributing a boat activity to fuel consumption and operational profile to estimate small boat greenhouse gas emissions in any given region.

No abstract available

Gas flaring is a major source of greenhouse gas and a waste of nonrenewable resources. Remote sensing is suggested to be a capable proxy to provide independent, reliable, and regular observations on global gas flaring. Multiple satellite sensors, especially nighttime imagers, have been used to detect and characterize gas flares. However, the hundred-meter level coarse resolution of current nighttime sensors impacts the accuracy of detection and estimation with small fires undetected. In this study, we explore the potential of high resolution multispectral SDGSAT-1 glimmer imager for urbanization (GIU) data and we propose a fine-grained gas flaring detection and estimation method based on the visible optical structure zones of gas flares. The study area of this work locates mainly in eight Persian Gulf countries. Existences, locations, volumes, and extents of gas flares within the study area are captured and analyzed. The proposed method achieves a detection accuracy of 90.46% with 891 active fire sites correctly recognized. Comparisons with daytime approach for gas flaring investigation V2 data show that our method generates less miss-detection especially when detecting small-sized gas flares, which demonstrates the utility of high resolution nighttime imagery and the proposed optical-based method.

To accurately predict the concentration ratio of wood dye solutions, a hybrid neural network model incorporating Principal Component Analysis (PCA), Adaptive Step-size Dung Beetle Optimization (ASDBO), and Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) was proposed. The model first applied PCA to perform principal component analysis on the absorbance data, reducing redundancy and retaining key features. It then used CNN to extract spatial features from the data, which were subsequently input into the LSTM for training. Finally, the ASDBO algorithm was used to optimize the model's hyperparameters. The model's performance was evaluated using multiple metrics, including Root Mean Square Error (RMSE) and the CIEDE2000 Color Difference Formula (CIEDE2000). The results demonstrated that the model significantly outperforms other models (such as SVR, RF, etc.), with RMSE (0.0483), MAE (0.0257), and R² (0.95). The RMSE was reduced by 29.5% compared to the CNN-GRU model. In the CIEDE2000 color difference evaluation, the proportion of no perceptible difference reached 81.92%, which is an improvement of 18.8% over CNN-GRU, meeting the requirements for industrial-grade color matching.

In order to solve the problem that the existing PM2.5 concentration prediction methods ignore the spatial and temporal influencing factors of PM2.5 concentration, this paper constructs a spatial characteristic factor of PM2.5 concentration based on the maximum information coefficient, and proposes a CNN‐LSTM combined prediction model based on multi‐feature fusion, which transforms the abstract spatial and temporal influencing factors into quantifiable features. The model has good feature extraction ability and strong ability to capture short‐term transient information and long‐range dependent information in time series data, which improves the prediction performance of the model. The experimental results show that the prediction accuracy of CNN‐LSTM model based on multi‐feature fusion is 87.21%, and MAPE is 6.25, 4.84, and 1.29 less than BP, SVR, and LightGBM, and 1.91 and 7.04 less than CNN and LSTM.

The rapid and accurate analysis of ion concentrations in mixed salt solutions is a critical aspect of utilizing salt lake chemical resources. To explore an efficient and non-destructive detection method, this study proposes a deep learning model that fuses a Convolutional Neural Network (CNN), a Long Short-Term Memory (LSTM) network, and an Attention mechanism for the prediction of salt solution concentrations using Near-Infrared Spectroscopy (NIRS). First, single-component and two-component mixed salt solution samples of NaCl, KCl, and MgCl₂ were prepared, and their near-infrared spectral data were collected. After applying Savitzky-Golay smoothing and derivative preprocessing to the spectra, a CNN-LSTM-Attention prediction model was constructed. A comparative analysis was conducted against common models such as Partial Least Squares Regression (PLSR), Support Vector Regression (SVR), and Random Forest (RF), and an ablation study was performed to analyze the contribution of each deep learning module. The results show that for single-component salt solutions, the proposed model's performance is comparable to that of the high-performing SVR and RF models. In complex mixed solutions with severe spectral overlap, the CNN-LSTM-Attention model demonstrated significant superiority, with its prediction accuracy surpassing all traditional baseline models across all mixed datasets, achieving a coefficient of determination (R²) as high as 0.973. The study concludes that the proposed CNN-LSTM-Attention model can effectively address the challenge of spectral overlap, demonstrating the potential of using deep learning for the quantitative analysis of complex mixture systems via near-infrared spectroscopy.

Considering that ozone is essential to understanding air quality and climate change, this study presents a deep learning method for predicting atmospheric ozone concentrations. The method combines an attention mechanism with a convolutional neural network (CNN) and long short-term memory (LSTM) network to address the nonlinear nature of multivariate time-series data. It employs CNN and LSTM to extract features from short time series, enhanced by the attention mechanism to improved short-term prediction accuracy. It takes eight meteorological and environmental parameters from 16,806 records (2018–2019) as input, which are selected principal component analysis (PCA). It features an attention-based CNN-LSTM hybrid deep learning model with specific settings: a time step of 5, a batch size of 25, 15 units in the LSTM layer, the ReLU activation function, 25 epochs, and an overfitting avoidance strategy with a dropout rate of 0.15. Experimental results demonstrate that this hybrid model outperforms individual models and the CNN-LSTM model, especially in forward prediction with a multi-hour time lag. The model exhibits a high coefficient of determination (R2 = 0.971) and a root mean square error of 3.59 for a 1-hour time lag. It also exhibits consistent accuracy across different seasons, highlighting its robustness and superior time-series prediction capabilities for ozone concentrations.

The concentrations of atmospheric particulate matter (PM10 and PM2.5) significantly impact global environment, human health, and climate change. This study developed a particulate matter concentration retrieval method based on multi-source data, proposing a dual-branch retrieval network architecture named CSLTNet that integrates Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The CNN branch is designed to extract spatial features, while the LSTM branch captures temporal characteristics, with attention modules incorporated into both the CNN and LSTM branches to enhance feature extraction capabilities. Notably, the model demonstrates robust spatial generalization capability across different geographical regions.Comprehensive experimental evaluations demonstrate the outstanding performance of the CSLTNet model. For the Beijing–Tianjin–Hebei region in China: in PM10 retrieval, sample-based 10-fold cross-validation achieved R2 = 0.9427 (RMSE = 16.47μg/m3), while station-based validation yielded R2 = 0.9213 (RMSE = 19.50μg/m3); for PM2.5 retrieval, sample-based 10-fold cross-validation resulted in R2 = 0.9579 (RMSE = 6.49μg/m3), with station-based validation reaching R2 = 0.9296 (RMSE = 8.32μg/m3). For Northwest China: in PM10 retrieval, sample-based 10-fold cross-validation achieved R2 = 0.9236 (RMSE = 34.52μg/m3), while station-based validation yielded R2 = 0.9046 (RMSE = 37.24μg/m3); for PM2.5 retrieval, sample-based 10-fold cross-validation resulted in R2 = 0.9279 (RMSE = 10.56μg/m3), with station-based validation reaching R2 = 0.8787 (RMSE = 13.71μg/m3).

Fine particulate matter (PM2.5) is a major air pollutant affecting human survival, development and health. By predicting the spatial distribution concentration of PM2.5, pollutant sources can be better traced, allowing measures to protect human health to be implemented. Thus, the purpose of this study is to predict and analyze the PM2.5 concentration of stations based on the integrated deep learning of a convolutional neural network long short-term memory (CNN-LSTM) model. To solve the complexity and nonlinear characteristics of PM2.5 time series data problems, we adopted the CNN-LSTM deep learning model. We collected the PM2.5data of Qingdao in 2020 as well as meteorological factors such as temperature, wind speed and air pressure for pre-processing and characteristic analysis. Then, the CNN-LSTM deep learning model was integrated to capture the temporal and spatial features and trends in the data. The CNN layer was used to extract spatial features, while the LSTM layer was used to learn time dependencies. Through comparative experiments and model evaluation, we found that the CNN-LSTM model can achieve excellent PM2.5 prediction performance. The results show that the coefficient of determination (R2) is 0.91, and the root mean square error (RMSE) is 8.216 µg/m3. The CNN-LSTM model achieves better prediction accuracy and generalizability compared with those of the CNN and LSTM models (R2 values of 0.85 and 0.83, respectively, and RMSE values of 11.356 and 14.367, respectively). Finally, we analyzed and explained the predicted results. We also found that some meteorological factors (such as air temperature, pressure, and wind speed) have significant effects on the PM2.5 concentration at ground stations in Qingdao. In summary, by using deep learning methods, we obtained better prediction performance and revealed the association between PM2.5 concentration and meteorological factors. These findings are of great significance for improving the quality of the atmospheric environment and protecting public health.

In order to solve the problem of operators of SCR flue gas denitrification equipment in thermal power plants relying on experience to adjust the opening of the ammonia injection valve and reduce the concentration of NO x in the outlet flue gas of the SCR denitrification system, a method for predicting the NO x concentration at the SCR outlet is proposed. This method, based on a one-dimensional

Prolonged exposure to high concentrations of suspended particulate matter (SPM), especially aerodynamic fine particulate matter that is ≤2.5 μm in diameter (PM2.5), can cause serious harm to human health and life via the induction of respiratory diseases and lung cancer. Therefore, accurate prediction of PM2.5 concentrations is important for human health management and governmental environmental management decisions. However, the time-series processing of PM2.5 concentration based only on a single region and a special time period is less explanatory, and thus, the spatial-temporal applicability of the model is more restricted. To address this problem, this paper constructs a PM2.5 concentration prediction optimization model based on Convolutional Neural Networks-Long Short-Term Memory (CNN-LSTM). Hourly data of atmospheric pollutants, meteorological parameters, and Precipitable Water Vapor (PWV) of 10 cities in the Beijing-Tianjin-Hebei metropolitan area during the period of 1–30 September 2021/2022 were used as the training set, and the PM2.5 data of 1–7 October 2021/2022 were used for validation. The experimental results show that the CNN-LSTM model optimizes the average root mean square error (RMSE) by 25.52% and 14.30%, the average mean absolute error (MAE) by 26.23% and 15.01%, and the average mean absolute percentage error (MAPE) by 35.64% and 16.98%, as compared to the widely used Back Propagation Neural Network (BPNN) and Long Short-Term Memory (LSTM) models. In summary, the CNN-LSTM model is superior in terms of applicability and has the highest prediction accuracy in the Beijing-Tianjin-Hebei metropolitan area. The results of this study can provide a reference for the relevant departments in the Beijing-Tianjin-Hebei metropolitan area to predict PM2.5 concentration and its trend in specific time periods.

Proper use of disinfectants in the cold chain environment is critical to prevent the cold chain transmission of the new coronavirus and other viruses. To address this issue, this paper presents a model for predicting the disinfectant concentrations based on 1D CNN-LSTM to quantify the concentration of four types of disinfectants in cold chain. An electronic nose platform developed in the laboratory is used and combined with the algorithm. The electronic nose signal incorporates information such as disinfectant concentration, and combining 1D CNN and LSTM models can achieve multiple advantages. 1D CNN is able to capture local patterns and features in sequences, while LSTM is able to model long-term dependencies, enabling the model to efficiently process sequence data and reduce the computational complexity and memory consumption of the model, effectively preventing overfitting. Experimental results demonstrate that the algorithm can effectively quantify the concentration of disinfectants in the cold chain environment, with a reduction in error.

Accurately predicting PM2.5 concentration can effectively avoid the harm caused by heavy pollution weather to human health. In view of the non-linearity, time series characteristics, and the problem of large multi-step prediction errors in PM2.5 concentration data, a method combining Long Short-term Memory Network and Convolutional Neural Network with Time Pattern Attention mechanism (TPA-CNN-LSTM) is proposed. The method uses historical PM2.5 concentration data, historical meteorological data, and surrounding station data to predict the future 6-hour PM2.5 concentration of air quality monitoring stations. Firstly, CNN is used to obtain the spatial characteristics between multiple stations, secondly, LSTM is added after CNN to extract the temporal changes of non-linear data, and finally, to capture the key features of temporal information, Temporal Pattern Attention mechanism (TPA) is added. TPA can automatically adjust weights based on the input of each time step, and select the most relevant time step for prediction, thereby improving the accuracy of the model. An example analysis is conducted on the measured data of Beijing's air quality stations in 2018, and compared with other mainstream algorithms. The results show that the proposed model has higher prediction accuracy and performance.

Particulate matter forecasting is fundamental for early warning and controlling air pollution, especially PM2.5. The increase in this level of concentration will lead to a negative impact on public health. This study develops a hybrid model of CNN-LSTM and CONV-LSTM by combining a convolutional neural network (CNN) with an LSTM network to forecast PM2.5 concentration for the next few hours in Kemayoran DKI Jakarta, which is known as a busy area. We discovered the advantages of CNN in effectively extracting features and LSTM in learning long-term historical data from PM2.5 concentration time series data. The predictive model of CNN-LSTM is carried out in a different architecture where the CNN process is carried out first to become the input of LSTM. For CONV-LSTM, it is carried out in one architecture where the multiplication in the LSTM architecture is coupled with the convolution process. This research will explain how the method of developing hybrid CNN-LSTM and CONV-LSTM in predicting PM2.5 concentrations. Based on metric evaluation, the two models are compared to find the best model. Both predictive models produce MAPE values that fall into the good enough category with values < 20%. Results were obtained for CONV-LSTM with MAE worth 6.52, RMSE 8.55, and MAPE 16.39%. As a result, the CONV-LSTM model performs better than CNN-LSTM in nowcasting PM2.5.

The water consumption of urban users is complex, and there are great differences between residential water and industrial wastewater. Real-time monitoring of water quality is an important task of urban water management. Among the water quality parameters, dissolved oxygen concentration (DO) is an important index to evaluate the quality of water. Therefore, the accuracy of dissolved oxygen prediction results is very important for the management of water bodies in sewage treatment. In this paper, a model based on DBSCAN-CNN-LSTM is proposed to solve the problems of inaccurate prediction accuracy and poor data convergence in the existing water quality prediction model in the sewage treatment process. Firstly, the obtained water quality data are preprocessed to obtain water quality factors with high correlation with DO.Secondly, the input data are predicted and output by establishing a hybrid model based on DBSCAN-CNN-GRU. The DBSCAN algorithm can cluster the sewage data and reduce the influence of local minima on the prediction results, so as to improve the prediction accuracy and data convergence. Compared with the traditional neural network models GRU, LSTM and CNN-LSTM, the predicted values is closer to the real values.

With the intensification of global climate change, accurate prediction of air quality indicators, especially PM2.5 concentration, has become increasingly important in fields such as environmental protection, public health, and urban management. To address this, we propose an air quality PM2.5 index prediction model based on a hybrid CNN-LSTM architecture. The model effectively combines Convolutional Neural Networks (CNN) for local spatial feature extraction and Long Short-Term Memory (LSTM) networks for modeling temporal dependencies in time series data. Using a multivariate dataset collected from an industrial area in Beijing between 2010 and 2015 -- which includes hourly records of PM2.5 concentration, temperature, dew point, pressure, wind direction, wind speed, and precipitation -- the model predicts the average PM2.5 concentration over 6-hour intervals. Experimental results show that the model achieves a root mean square error (RMSE) of 5.236, outperforming traditional time series models in both accuracy and generalization. This demonstrates its strong potential in real-world applications such as air pollution early warning systems. However, due to the complexity of multivariate inputs, the model demands high computational resources, and its ability to handle diverse atmospheric factors still requires optimization. Future work will focus on enhancing scalability and expanding support for more complex multivariate weather prediction tasks.

Dissolved oxygen (DO) affects both the health of aquatic animals and the quality of the ambient water. Traditional modelling approaches to predict changes in DO do not account for two aspects: the dependencies between locations and correlations between dates over longer time periods. This paper describes a CNN-LSTM approach that aims to improve the accuracy of DO predictions. Data collection occurred over a two-month period, and included temperature, pH, salinity, and ammonia concentration collected every 15 minutes. Sequential modelling is accomplished through the LSTM layers, while spatial feature extraction is accomplished through the convolutional layers. In the experiments, the CNN-LSTM outperformed regression models such as LSTM, SVR, and RF, with an R2 of 98.46%, RMSE of 0.089 mg/L, and MAE of 0.065 mg/L. These results indicate that the CNN-LSTM approach is suitable for making real-time predictions in support of sustainable aquaculture management.

Mine water inrush poses a severe threat to safe mine production. Rapid and accurate identification of water source types is crucial for effective prevention and control of such hazards. This study proposes an efficient water source identification method integrating Laser-Induced Fluorescence (LIF) technology. Compared to traditional physicochemical analysis methods, LIF enables faster, more accurate, and lower-cost identification. For the qualitative identification task, a CNN-LSTM fusion model optimized with the Starfish Optimization Algorithm (SFOA) was developed. For the quantitative prediction task, a prediction model integrating CNN-Attention-LSTM with Kolmogorov-Arnold Networks (KAN) was constructed. Experiments utilized mixed water samples of sandstone water and goaf water at varying concentration ratios. Adaptive Least Squares (ALS) and Variational Mode Decomposition (VMD) were employed for data preprocessing, while Linear Discriminant Analysis (LDA) and Uniform Manifold Approximation and Projection (UMAP) were applied for feature extraction. Results demonstrate that for qualitative identification, the CNN-LSTM model—optimized by SFOA and utilizing data preprocessed with ALS and subjected to feature extraction via UMAP—delivered the most exceptional performance. For quantitative prediction, the CNN-Attention-LSTM-KAN model utilizing VMD-preprocessed data achieved superior prediction effectiveness. The integrated scheme combining LIF technology with these advanced intelligent algorithms, developed in this research, provides a reliable technical foundation for high-precision and high-efficiency identification of mine water sources, holding significant importance for mine water hazard prevention and control.

No abstract available

—Predicting dissolved oxygen (DO) concentration in water is an important problem in environmental management and aquaculture, but it is often challenging due to the nonlinear, rapidly fluctuating and locally noisy characteristics of the observed time series. Although deep learning models such as Transformer are capable of exploiting long-term dependencies through self-attention mechanisms, they lack the ability to memorize temporal states and are ineffective in handling short-term patterns and local fluctuations. To overcome these limitations, in this paper we propose a hybrid CNN-LSTM-Transformer model, which combines the advantages of three architectures: (1) CNN − that extracts local features and denoises; (2) LSTM − that memorizes state-sequential information; and (3) Transformer − that learns nonlinear global dependencies in the data series. The model is evaluated on a real-world monitoring dataset (collected in Ria Formosa, Portugal), which includes water quality parameters such as temperature, conductivity, salinity, pH, and DO. Experimental results show that our proposed model outperforms the single and traditional models including LR, SVR, LSTM, and Transformer. As a consequence, the model not only accurately predicts long-term trends but also responds well to short-term fluctuations − a key factor in real-time water quality monitoring systems.

Accurate monitoring of chlorophyll-a (Chl-a) is essential for managing coastal aquaculture, as Chl-a indicates phytoplankton biomass and water quality. This study developed a hybrid deep learning model integrating convolutional neural networks (CNN), bidirectional long short-term memory (BiLSTM), and an attention mechanism (Attention) to classify Chl-a using hourly, water quality datasets collected from the GOT001 station in Si Racha Bay, Eastern Gulf of Thailand (2020–2024). A random forest (RF) identified sea surface temperature (SEATEMP), dew point temperature (DEWPOINT), and turbidity (TURB) as the most influential variables, accounting for over 90% of the accuracy. Chl-a concentrations were categorized into ecological groups (low, medium, and high) using quantile-based binning and K-means clustering to support operational classification. Model performance comparison showed that the CNN–BiLSTM model achieved the highest classification accuracy (81.3%), outperforming the CNN–LSTM model (59.7%). However, the addition of the Attention did not enhance predictive performance, likely due to the limited number of key predictive variables and their already high explanatory power. This study highlights the potential of CNN–BiLSTM as a near-real-time classification tool for Chl-a levels in highly variable coastal ecosystems, supporting aquaculture management, early warning of algal blooms or red tides, and water quality risk assessment in the Gulf of Thailand and comparable coastal regions.

Air pollution is a global issue affecting the health of people, the sustainability of the environment, and the planning of urban areas. The present work utilises a smart air quality data monitoring analysis system that uses machine learning algorithms in forecasting and studying atmospheric pollution concentration. Multi-pollutant forecasting in Ust- Kamenogorsk utilises the collaborative use of LSTM and CNN. The Gaussian approximation is used in detecting outliers and meteorological input is added in order to support predictive precision. The LSTM-CNN blended model was utilized in predicting the concentration of different contaminants, including PM2.5, PM10, NO2, SO2, CO, and O3. The predictive accuracy of the model was average, considering its Root Mean Squared Error (RMSE) of 0.3297. Mean absolute error (MAE) was 0.2741, indicating differences in prediction ability among contaminants. However, R² score at -0.3210 suggests that the model needs to be tuned for greater predictability. Identification of outliers was done through residual analysis, which provided a 1.0 recall but poor precision of 0.0676, indicating high false positive rate. Despite its limitations, the model has the capacity to anticipate air quality in real time and detect anomalies. Future enhancements will include hyperparameter optimization, the addition of new data sources, and the refining of the anomaly detection method for greater accuracy and dependability. This contribution goes toward the development of intelligent air quality monitoring technologies to support data-driven environmental management and policy.

In response to the problems of redundant highdimensional data features, insufficient local information capture, and difficulty in modeling long-term dependencies in predicting SO ${ }_{2}$ concentration in flue gas from coal-fired power plants, this study proposes a multimodal prediction framework that integrates local linear embedding (LLE), convolutional neural network (CNN), Transformer, and long short-term memory network (LSTM). Using LLE algorithm to perform nonlinear dimensionality reduction on high-dimensional running data and extract low dimensional manifold features to simplify model complexity; By integrating the local feature capture capability of CNN, the global attention mechanism of Transformer, and the time series memory characteristics of LSTM, a hybrid model is constructed to enhance prediction robustness. The experiment is verified based on the actual data of a 550 MW unit desulfurization system. The results show that the MAPE of the fusion model is $\mathbf{9 7. 3 2 \%}$, and the performance of the fusion model is significantly better than that of the single model.

Reliable and accurate prediction of SO2 concentration would be conducive to the effective manipulation and maintenance for wet flue gas desulfurization (WFGD) unit and is of great significance for saving resources and protecting the environment. The desulfurization system has a large time delay and strong nonlinearity. Aiming at the problems above, this paper proposed a novel hybrid model based on Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) for predicting the SO2 concentration at outlet. Firstly, the CNN method is used to extract features from the inputs, and decrease the dimension of data. Secondly, LSTM has unique advantages for processing time series which is applied to solve the problem of time delay in desulfurization system. Dataset is sampled from a practical wet limestone-gypsum desulfurization system to evaluate the performance of the proposed model, and experiments are carried out in which the proposed model is compared with the LSTM model. The results indicate that the proposed CNN-LSTM model has lower errors and outperforms on the predicted results than LSTM model.

Forecasting air quality is a crucial technical approach to effectively respond to severe pollution conditions. The evolution of pollutant concentration has spatial correlation Due to the challenge of identifying monitoring stations with significant spatial correlation, a method utilizing the K-means clustering algorithm is proposed for partitioning air quality monitoring stations. Taking Nantong as an example, based on the selection of historical pollutant data in the target area, combined with meteorological data, the hybrid CNN-LSTM model, which consists of the convolutional neural network(CNN) and the long short-term memory(LSTM) neural network, is used to predict the pollutants, and finally realize the extraction of the temporal and spatial evolution characteristics of the pollutant concentration to complete the high accuracy of air quality forecast. Experimental results show that, after adding historical pollutant concentration data from stations in the cluster, the CNN-LSTM model can forecast PM2.5 concentrations precisely.

No abstract available

This paper aims to assess the water quality of the Ismailia Canal, Egypt, in accordance with Article 49 of Law 92/2013. QUAL2K and Convolutional Neural Networks and Long Short-Term Memory (CNN-LSTM) are utilized to simulate the water quality parameters of dissolved oxygen (DO), pH, biological oxygen demand (BOD), chemical oxygen demand (COD), total phosphorus (TP), nitrate nitrogen (NO3-N), and ammonium (NH3-N) in winter and summer 2023. The parameters of the QUAL2K and CNN-LSTM models were calibrated and validated in both winter and summer through trial and error, until the simulated results agreed well with the observed data. Additionally, the model’s performance was measured using different statistical criteria such as mean absolute error (MAE), root mean square (RMS), and relative error (RE). The results showed that the simulated values were in good agreement with the observed values. The results show that all parameter concentrations follow and did not exceed the limit of Article 49 of Law 92/2013 in winter and summer, except for dissolved oxygen concentration (8.73–4.53 mg/L) in winter and summer, respectively, which exceeds the limit of 6 mg/L, and in June, biological oxygen demand exceeds the limit of 6 mg/L due to increased organic matter. It is imperative to compare QUAL2K and CNN-LSTM models because QUAL2K provides a physics-based simulation of water quality processes, whereas CNN-LSTM employs deep learning in modeling complex temporal patterns. The two models enhance prediction accuracy and credibility towards enabling enhanced decision-making for Ismailia Canal water management. This research can be part of a decision support system regarding maximizing the benefits of the Ismailia Canal.

Jakarta is among the most polluted cities globally, with PM 2.5 posing the greatest health risks. Accurate prediction of PM 2.5 levels is therefore essential to support early warning systems and public health decisions. This study develops a hybrid deep learning model, CNN-LSTM, to forecast daily PM 2.5 concentrations in Central Jakarta for the next seven days. The dataset combines air pollutant records from Satu Data Jakarta and meteorological data from BMKG, covering January 2023 to December 2024. Data preparation included cleaning, handling missing values, outlier analysis, feature selection using Pearson correlation and linear regression, and normalization. Model performance was evaluated using MSE, MAE, RMSE, and MAPE, and compared against standalone CNN and LSTM models under two data configurations: pollutant-only and pollutant & meteorology. Results show that CNN-LSTM achieved the lowest MSE and RMSE when using combined data, indicating high predictive accuracy. However, the LSTM model with pollutant-only input provided the most consistent and efficient results, with lower error values across all metrics. These findings suggest that historical pollutant patterns alone are sufficient for short-term PM 2.5 forecasting, while meteorological factors offer limited additional benefit. The study demonstrates the potential of deep learning in supporting air quality monitoring and public health protection.

When a hydropower unit operates in a sediment-laden river, the sediment accelerates hydro-turbine wear, leading to efficiency loss or even shutdown. Therefore, wear fault diagnosis is crucial for its safe and stable operation. A hydro-turbine wear fault diagnosis method based on improved WT (wavelet threshold algorithm) preprocessing combined with IWSO (improved white shark optimizer) optimized CNN-LSTM (convolutional neural network-long-short term memory) is proposed. The improved WT algorithm is utilized to denoise the preprocessing of the original signals. Chaotic mapping, bird flock search, and cosine elite variation strategies are introduced to enhance the WSO algorithm's robust performance, and the CNN-LSTM model's hyperparameters are optimized using the IWSO algorithm to improve the diagnostic performance. The experimental results show that the accuracy of the proposed method reaches 96.2%, which is 8.9% higher than that of the IWSO-CNN-LSTM model without denoising. The study also found that the diagnostic accuracy of hydro-turbine wear faults increased with increasing sediment concentration in the water. This study can supplement the existing hydro-turbine condition monitoring and fault diagnosis system. Meanwhile, diagnosing wear faults in hydro-turbines can improve power generation efficiency and quality and minimize resource consumption.

One of the most important environmental problems brought about by rapid population growth and industrialization is air pollution. Today, air pollution is generally caused by heating, industry and motor vehicles. In addition, factors such as unplanned urbanization, topographic structure of cities, atmospheric conditions and meteorological parameters, building and population density also cause pollution to increase. Pollutants with concentrations above limit values have negative effects on humans and the environment. In order to prevent people from being negatively affected by these pollutants, it is necessary to know the pollution level and take action as soon as possible. In this study, a hybrid ConvLSTM model was developed in order to quickly and effectively predict air pollution, which has such negative effects on humans and the environment. ConvLSTM was compared with LR, RF, SVM, MLP, CNN and LSTM using approximately 4 years of air quality index data from the city of Gurugram in India. Experimental results showed that ConvLSTM was significantly more successful than the base models, with 30.645 MAE and 0.891 R2.

No abstract available

Accurate aerosol optical depth (AOD) prediction remains challenging due to complex aerosol‐radiation interactions and highly variable spatio‐temporal patterns. Three critical scientific issues motivate this work: understanding whether and how physical principles can enhance deep learning predictions, identifying which aerosol properties most strongly govern AOD variations, and improving the prediction of extreme AOD events critical for air quality management. Herein, utilizing MERRA‐2 reanalysis data (1980–2024) over the Huaihe River Basin in eastern China, a Physics‐Guided deep learning framework is presented for Aerosol Optical Depth (AOD) prediction. The model proposed integrates Convolutional Neural Networks (CNN), Long Short‐TermMemory (LSTM) networks, and multi‐head attention mechanisms to capture both spatio‐temporal features and physical relationships of aerosol properties. Three key aspects are involved: First, a hybrid deep learning model is developed and evaluated, which combines CNNs for spatial correlation extraction, bidirectional LSTM for temporal dependency modeling, and multi‐head attention for feature interaction learning. Second, a comprehensive feature importance analysis is conducted by examining the relationships between different aerosol properties (mass concentration, scattering coefficient, and Ångström exponent) and AOD prediction, offering physical insights into the model's decision‐making process. Third, a specialized approach is proposed for extreme AOD event prediction, focusing on early detection and accurate forecasting of high‐AOD episodes. Overall, the results demonstrate the model's efficacy in capturing both regular AOD variations and extreme events, with the Physics‐Guided architecture showing superior performance compared to traditional methods. This integrated approach enhances AOD prediction accuracy and deepens insights into aerosol‐radiation interactions, thereby improving atmospheric monitoring and air quality forecasting. While MERRA‐2 has inherent temporal delays, this framework provides valuable capabilities for historical trend analysis, numerical model validation, and can be readily adapted for real‐time applications through transfer learning with satellite observations.

PM2.5 is one of the most important pollutants related to air quality, and the increase of its concentration will aggravate the threat to people’s health. Therefore, the prediction of surface PM2.5 concentration is of great significance to human health protection. In this study, A hybrid CNN-LSTM model is developed by combining the convolutional neural network (CNN) with the long short-term memory (LSTM) neural network for forecasting the next 24h PM2.5 concentration in Beijing, which makes full use of their advantages that CNN can effectively extract the features related to air quality and the LSTM can reflect the long term historical process of input time series data. The air quality data of the last 7days and the PM2.5 concentration of the next day are first set as the input and output of the model due to the periodicity, respectively. Subsequently four models namely univariate LSTM model, multivariate LSTM model, univariate CNN-LSTM model and multivariate CNN-LSTM model, are established for PM2.5 concentration prediction. Finally, mean absolute error (MAE) and root mean square error (RMSE) are employed to evaluate the performance of these models and results show that the proposed multivariate CNN-LSTM model performs the best results due to low error and short training time.

No abstract available

No abstract available

In modern society, air pollution is an important topic as this pollution exerts a critically bad influence on human health and the environment. Among air pollutants, Particulate Matter (PM2.5) consists of suspended particles with a diameter equal to or less than 2.5 μm. Sources of PM2.5 can be coal-fired power generation, smoke, or dusts. These suspended particles in the air can damage the respiratory and cardiovascular systems of the human body, which may further lead to other diseases such as asthma, lung cancer, or cardiovascular diseases. To monitor and estimate the PM2.5 concentration, Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) are combined and applied to the PM2.5 forecasting system. To compare the overall performance of each algorithm, four measurement indexes, Mean Absolute Error (MAE), Root Mean Square Error (RMSE) Pearson correlation coefficient and Index of Agreement (IA) are applied to the experiments in this paper. Compared with other machine learning methods, the experimental results showed that the forecasting accuracy of the proposed CNN-LSTM model (APNet) is verified to be the highest in this paper. For the CNN-LSTM model, its feasibility and practicability to forecast the PM2.5 concentration are also verified in this paper. The main contribution of this paper is to develop a deep neural network model that integrates the CNN and LSTM architectures, and through historical data such as cumulated hours of rain, cumulated wind speed and PM2.5 concentration. In the future, this study can also be applied to the prevention and control of PM2.5.

PM2.5 is particle with a diameter of less than or equal to 2.5 microns, which contains large amounts of toxic substances and has a big effect on human fitness and environmental quality, so Predicting PM2.5 concentration is very important. The change of PM2.5 concentration is the result of different factors, and the change process is non-linear, so it is very trouble to use conventional approach to predict. In this paper, PM10, SO2, CO are used as the prediction indexes of PM2.5, and the prediction model of PM2.5 based on CNN-LSTM hybrid neural network is constructed. Pearson correlation analysis method was used for correlation analysis of multiple pollutant indicators, missing data were processed and normalized. The multi-variable CNN-LSTM prediction model was built, and the hourly data of six pollutants in Nanjing in 2021 were used for experiments. The experimental results show that the hybrid model can accurately predict the change of PM2.5.

Industrial wastewater classification is the basic work of water pollution prevention and control and water resources management, water chemical oxygen demand is the core indicator to measure the quality of water, compared with domestic sewage detection, the research on industrial wastewater classification is relatively lagging behind. Aiming at the shortcomings of COD classification model of industrial wastewater using convolutional neural network or long shortterm memory network alone, this paper constructs a hybrid model based on CNN and LSTM. The COD data of industrial wastewater were measured by ultraviolet-visible spectroscopy, abstract features were extracted from the data using CNN, and finally entered into LSTM, and the results of classification of industrial wastewater according to COD concentration were obtained, and the classification accuracy of the model reached 98.66%, which was 3.36% and 4.7% higher than that of CNN and LSTM alone.

Increased concentrations of nitrogen dioxide (NO2) in urban air have become a serious issue due to their impact on respiratory health, especially for vulnerable groups. To support mitigation efforts, an accurate NO2 concentration prediction system is needed as a basis for developing an early warning system. This study compares the performance of four deep learning-based approaches, namely LSTM, GRU, CNN, and the hybrid models CNN-LSTM and CNN-GRU. The evaluation was conducted using the Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and coefficient of determination (R2) metrics. The results show that the pure CNN model produced the highest performance with an MAE of 0.896, an RMSE of 1.076, and an R2 of 0.996. Meanwhile, CNN-GRU also showed competitive results with an MAE of 1.718, an RMSE of 1.929, and an R2 of 0.989. In contrast, the CNN-LSTM combination produced the lowest performance. These findings confirm that the CNN architecture is capable of capturing spatial-temporal patterns in air pollution data very well and has the potential to be applied as the basis for an early warning system for NO2 exposure. Thus, this study contributes to the development of artificial intelligence-based predictive solutions for environmental and public health.

Despite significant advancements in environment perception capabilities for autonomous driving and intelligent robotics, cameras and LiDARs remain notoriously unreliable in low-light conditions and adverse weather, which limits their effectiveness. Radar serves as a reliable and low-cost sensor that can effectively complement these limitations. However, radar-based object detection has been underexplored due to the inherent weaknesses of radar data, such as low resolution, high noise, and lack of visual information. In this article, we present TransRAD, a novel 3-D radar object detection model designed to address these challenges by leveraging the retentive vision transformer (RMT) to more effectively learn features from information-dense radar range-Azimuth–Doppler (RAD) data. Our approach leverages the retentive Manhattan self-attention (MaSA) mechanism provided by RMT to incorporate explicit spatial priors, thereby enabling more accurate alignment with the spatial saliency characteristics of radar targets in RAD data and achieving precise 3-D radar detection across RAD dimensions. Furthermore, we propose location-aware nonmaximum suppression (LA-NMS) to effectively mitigate the common issue of duplicate bounding boxes in deep radar object detection. The experimental results demonstrate that TransRAD outperforms state-of-the-art (SOTA) methods in both 2-D and 3-D radar detection tasks, achieving higher accuracy, faster inference speed, and reduced computational complexity. Code is available at https://github.com/radar-lab/TransRAD.

Existing Vision Transformer (ViT)‐based object detection methods for remote sensing images (RSIs) face significant challenges due to the scarcity of RSI samples and the over‐reliance on enhancement strategies originally developed for natural images. This often leads to inconsistent data distributions between training and testing subsets, resulting in degraded model performance. In this study, we introduce an optimized data distribution learning (ODDL) strategy and develop an object detection framework based on the Faster R‐CNN architecture, named ODDL‐Net. The ODDL strategy begins with an optimized augmentation (OA) technique, overcoming the limitations of conventional data augmentation methods. Next, we propose an optimized mosaic algorithm (OMA), improving upon the shortcomings of traditional Mosaic augmentation techniques. Additionally, we introduce a feature fusion regularization (FFR) method, addressing the inherent limitations of classic feature pyramid networks. These innovations are integrated into three modular, plug‐and‐play components—namely, the OA, OMA, and FFR modules—ensuring that the ODDL strategy can be seamlessly incorporated into existing detection frameworks without requiring significant modifications. To evaluate the effectiveness of the proposed ODDL‐Net, we develop two variants based on different ViT architectures: the Next ViT (NViT) small model and the Swin Transformer (SwinT) tiny model, both used as detection backbones. Experimental results on the NWPU10, DIOR20, MAR20, and GLH‐Bridge datasets demonstrate that both variants of ODDL‐Net achieve impressive accuracy, surpassing 23 state‐of‐the‐art methods introduced since 2023. Specifically, ODDL‐Net‐NViT attained accuracies of 78.3% on the challenging DIOR20 dataset and 61.4% on the GLH‐Bridge dataset. Notably, this represents a substantial improvement of approximately 23% over the Faster R‐CNN‐ResNet50 baseline on the DIOR20 dataset. In conclusion, this study demonstrates that ViTs are well suited for high‐accuracy object detection in RSIs. Furthermore, it provides a straightforward solution for building ViT‐based detectors, offering a practical approach that requires little model modification.

ABSTRACT Natural and man-made disasters take place around the world and cause significant financial and human losses. An accurate and fast post-disaster building damage mapping could play a crucial role in rapid rescue planning and operations. Remote sensing satellite images are the main source of building damage map generation. Usually, both pre-disaster and post-disaster satellite images are used for the generation of building damage maps, which encounter some challenges such as registration errors, noise, and atmospheric conditions. This study proposed a new method for building damage detection based only on post images with the UNet architecture network and Global Context Vision Transformer blocks. The deep learning network proposed in this research is automatic without any further processing. The proposed method comprises four main steps: (1) pre-processing, (2) network training, (3) building damage map generation, and (4) accuracy assessment for the final damage map. This network is applied to three different natural and man-made disaster datasets. The first dataset is the post-satellite image of the 2023 Turkey earthquake, the second one is the post-satellite image of the 2021 Bata explosion and the last one is the post-satellite image of the 2011 haiti earthquake. Results of the final building damage map indicate that the proposed method is highly effective with OA above 96%, which is superior to the other deep learning methods.

Issues such as complex noise interference and the long-tail distribution of data present many challenges to the multicategory ship detection task in synthetic aperture radar (SAR) images. This article proposes an efficient multicategory-oriented SAR ship detector, which adopts a powerful lightweight feature enhancement vision transformer (FEViT) backbone for a comprehensive feature representation in SAR ship images and, hence, is referred to as FEVT-SAR. FEViT includes two innovative lightweight modules: localized feature interactive convolution block (LFICB) and dual-granularity attention transformer block (DGTB). LFICB fuses multireceptive field local features to suppress speckle noise, while DGTB employs a coarse-to-fine self-attention to capture the global dependencies and avoids enormous computational costs. Moreover, a selective CopyPaste augmentation paradigm is designed to rebalance ship data distribution through data sampling. Finally, the performance of the FEVT-SAR is evaluated on two typical SAR ship datasets, namely SRSDD and HRSID. Experimental results show that the mean average precision 50 of FEVT-SAR reaches 68.59% and 89.62%, respectively. The proposed FEVT-SAR outperforms several state-of-the-art-oriented bounding box detectors in the multicategory ship dataset SRSDD while demonstrating its robustness in the single-category ship dataset HRSID.

Forest health monitoring at scale requires high-spatial-resolution remote sensing images coupled with deep learning image analysis methods. However, high-quality large-scale datasets are costly to acquire. To address this challenge, we explored the potential of freely available National Agricultural Imagery Program (NAIP) imagery. By comparing the performance of traditional convolutional neural network (CNN) models (U-Net and DeepLabv3+) with a state-of-the-art Vision Transformer (SegFormer), we aimed to determine the optimal approach for detecting unhealthy tree crowns (UTC) using a publicly available data source. Additionally, we investigated the impact of different spectral band combinations on model performance to identify the most effective configuration without incurring additional data acquisition costs. We explored various band combinations, including RGB, color infrared (CIR), vegetation indices (VIs), principal components (PC) of texture features (PCA), and spectral band with PC (RGBPC). Furthermore, we analyzed the uncertainty associated with potential subjective crown annotation and its impact on model evaluation. Our results demonstrated that the Vision Transformer-based model, SegFormer, outperforms traditional CNN-based models, particularly when trained on RGB images yielding an F1-score of 0.85. In contrast, DeepLabv3+ achieved F1-score of 0.82. Notably, PCA-based inputs yield reduced performance across all models, with U-Net producing particularly poor results (F1-score as low as 0.03). The uncertainty analysis indicated that the Intersection over Union (IoU) could fluctuate between 14.81% and 57.41%, while F1-scores ranged from 8.57% to 47.14%, reflecting the significant sensitivity of model performance to inconsistencies in ground truth annotations. In summary, this study demonstrates the feasibility of using publicly available NAIP imagery and advanced deep learning techniques to accurately detect unhealthy tree canopies. These findings highlight SegFormer’s superior ability to capture complex spatial patterns, even in relatively low-resolution (60 cm) datasets. Our findings underline the considerable influence of human annotation errors on model performance, emphasizing the need for standardized annotation guidelines and quality control measures.

Invading pests and diseases always degrade the quality and quantity of plants. Early and accurate identification of plant diseases is critical for plant health and growth. This work proposes a smartphone-based solution using a Vision Transformer (ViT) model for identifying healthy plants and unhealthy plants with diseases. The collected dataset of tomato leaves was used to collectively train Vision Transformer and Inception V3-based deep learning (DL) models to differentiate healthy and diseased plants. These models detected 10 different tomato disease classes from the dataset containing 10,010 images. The performance of the two DL models was compared. This work also presents a smartphone-based application (Android App) using a ViT-based model, which works on the basis of the self-attention mechanism and yielded a better performance (90.99% testing) than Inception V3 in our experimentation. The proposed ViT-SmartAgri is promising and can be implemented on a colossal scale for smart agriculture, thus inspiring future work in this area.

: Mango farming significantly contributes to the economy, particularly in developing countries. However, mango trees are susceptible to various diseases caused by fungi, viruses, and bacteria, and diagnosing these diseases at an early stage is crucial to prevent their spread, which can lead to substantial losses. The development of deep learning models for detecting crop diseases is an active area of research in smart agriculture. This study focuses on mango plant diseases and employs the ConvNeXt and Vision Transformer (ViT) architectures. Two datasets were used. The first, MangoLeafBD, contains data for mango leaf diseases such as anthracnose, bacterial canker, gall midge, and powdery mildew. The second, SenMangoFruitDDS, includes data for mango fruit diseases such as Alternaria, Anthracnose, Black Mould Rot, Healthy, and Stem and Rot. Both datasets were obtained from publicly available sources. The proposed model achieved an accuracy of 99.87% on the MangoLeafBD dataset and 98.40% on the MangoFruitDDS dataset. The results demonstrate that ConvNeXt and ViT models can effectively diagnose mango diseases, enabling farmers to identify these conditions more efficiently. The system contributes to increased mango production and minimizes economic losses by reducing the time and effort needed for manual diagnostics. Additionally, the proposed system is integrated into a mobile application that utilizes the model as a backend to detect mango diseases instantly.