Pub Date : 2026-06-01Epub Date: 2026-05-26DOI: 10.1016/j.aiig.2026.100226
Ihcene Djouama, Nabil Kadache, Rachid Seghir
High-resolution near-surface temperature data are essential in mountainous regions, where complex topography induces strong spatial and temporal variability. However, coarse-resolution reanalysis products such as ERA5-Land (9 km) fail to represent fine-scale thermal patterns. This study presents a UNet based deep learning framework to downscale ERA5-Land 2-m air temperature to 250 m resolution at a 6-hourly temporal frequency over the Trentino–South Tyrol Alpine region. The model is trained and validated using the observation-based ALPINE-TST-250 gridded dataset, derived from more than 300 weather stations, and integrates high-resolution elevation data along with physically meaningful auxiliary predictors. Including 2-m dew point temperature significantly improves performance, reducing the RMSE from 2.32 to 2.05 and the MAE from 1.76 to 1.54 . The resulting 2011–2021 downscaled dataset successfully reproduces sub-kilometer spatial gradients and temporal variability absent in the original ERA5-Land fields. These results highlight the potential of deep learning approaches to enhance temperature representation in complex alpine terrain for climate and environmental applications.
{"title":"Deep learning-based downscaling of ERA5-Land temperature to 250 m resolution over the Trentino–South Tyrol Alpine region","authors":"Ihcene Djouama, Nabil Kadache, Rachid Seghir","doi":"10.1016/j.aiig.2026.100226","DOIUrl":"10.1016/j.aiig.2026.100226","url":null,"abstract":"<div><div>High-resolution near-surface temperature data are essential in mountainous regions, where complex topography induces strong spatial and temporal variability. However, coarse-resolution reanalysis products such as ERA5-Land (9 km) fail to represent fine-scale thermal patterns. This study presents a UNet based deep learning framework to downscale ERA5-Land 2-m air temperature to 250 m resolution at a 6-hourly temporal frequency over the Trentino–South Tyrol Alpine region. The model is trained and validated using the observation-based ALPINE-TST-250 gridded dataset, derived from more than 300 weather stations, and integrates high-resolution elevation data along with physically meaningful auxiliary predictors. Including 2-m dew point temperature significantly improves performance, reducing the RMSE from 2.32 <span><math><mrow><mo>°</mo><mi>C</mi></mrow></math></span> to 2.05 <span><math><mrow><mo>°</mo><mi>C</mi></mrow></math></span> and the MAE from 1.76 <span><math><mrow><mo>°</mo><mi>C</mi></mrow></math></span> to 1.54 <span><math><mrow><mo>°</mo><mi>C</mi></mrow></math></span>. The resulting 2011–2021 downscaled dataset successfully reproduces sub-kilometer spatial gradients and temporal variability absent in the original ERA5-Land fields. These results highlight the potential of deep learning approaches to enhance temperature representation in complex alpine terrain for climate and environmental applications.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"7 2","pages":"Article 100226"},"PeriodicalIF":4.2,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148177940","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-06-01Epub Date: 2026-05-28DOI: 10.1016/j.aiig.2026.100231
Anouar Taibou, Halima Jounaid, Nachchach Badr, Said Ou Moua
Groundwater potential mapping (GWP) is essential for sustainable resource managmment, particlarly in semi-arid and data-scarce regions. This study aims to delineate groundwater potential zones in the Azrou-Khenifra basin by integrating topographical, hydrological, geological, cilmatological and Land use/Land cover factors using machine learning and ensemble learning models. A total of twenty groundwater influencing factors were considered, including elevation, slope, slope aspect, profile and plan curvature, TPI, TRI, SPI, TWI, drainage density, distance to drainage, lithology, distance to faults, fault density, distance to lineaments, lineament density, rainfall, LST, as well LULC, NDVI, most of which were derived from satellite data. Frequency Ratio (FR) and Weight of Evidence (WoE) were aplied to analyze the spatial relationship between water points and individual factor classes and to evaluate their contribution. Groundwater potential mpas were generated using three machine learning algorithms: Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbors, and four ensemble learning models (RF-SVM, RF-KNN, SVM-KNN, and RF-SVM-KNN). Models performance was evaluated using precision, recall, F1-score, overall accuracy, and ROC-AUC. The results indicate that the RF model achieved the highest predictive perfomance among individual models, with an accuracy of 0.74 and ROC-AUC of 0.8112, followed by KNN (AUC = 0.787) and SVM (AUC = 0.778). Ensemble learning models competitiveand stable performance, particularly RF-based ensembles and the final RF-SVM-KNN voting model, which achieved balanced metrics with and overall accuracy of about 0.68-0.72 and AUC values reaching 0.8112. Feature importance analysis highlights rainfall, Fault density, and distance to faults as the most influential factors controlling groundwater occurrence, followed by topograohic indices and hydrological factors. Additional validation based on the spatial distribution of water points productivity across groundwater potential classes confirmed the reliability of the generated maps, as higher flow-rate points were predominantly located within high and very high potential zones. The findings demonstrate that integrating machine learning and ensemble learning techniques provides reliable groundwater potential predictions and offers a useful framework for groundwater exploration and sustainable management in hetergeneous and data-limited regions.
{"title":"Machine learning and ensemble learning models for groundwater potential mapping in a fractured basin: Case of the Azrou-Khenifra basin, central massif, Morocco","authors":"Anouar Taibou, Halima Jounaid, Nachchach Badr, Said Ou Moua","doi":"10.1016/j.aiig.2026.100231","DOIUrl":"10.1016/j.aiig.2026.100231","url":null,"abstract":"<div><div>Groundwater potential mapping (GWP) is essential for sustainable resource managmment, particlarly in semi-arid and data-scarce regions. This study aims to delineate groundwater potential zones in the Azrou-Khenifra basin by integrating topographical, hydrological, geological, cilmatological and Land use/Land cover factors using machine learning and ensemble learning models. A total of twenty groundwater influencing factors were considered, including elevation, slope, slope aspect, profile and plan curvature, TPI, TRI, SPI, TWI, drainage density, distance to drainage, lithology, distance to faults, fault density, distance to lineaments, lineament density, rainfall, LST, as well LULC, NDVI, most of which were derived from satellite data. Frequency Ratio (FR) and Weight of Evidence (WoE) were aplied to analyze the spatial relationship between water points and individual factor classes and to evaluate their contribution. Groundwater potential mpas were generated using three machine learning algorithms: Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbors, and four ensemble learning models (RF-SVM, RF-KNN, SVM-KNN, and RF-SVM-KNN). Models performance was evaluated using precision, recall, F1-score, overall accuracy, and ROC-AUC. The results indicate that the RF model achieved the highest predictive perfomance among individual models, with an accuracy of 0.74 and ROC-AUC of 0.8112, followed by KNN (AUC = 0.787) and SVM (AUC = 0.778). Ensemble learning models competitiveand stable performance, particularly RF-based ensembles and the final RF-SVM-KNN voting model, which achieved balanced metrics with and overall accuracy of about 0.68-0.72 and AUC values reaching 0.8112. Feature importance analysis highlights rainfall, Fault density, and distance to faults as the most influential factors controlling groundwater occurrence, followed by topograohic indices and hydrological factors. Additional validation based on the spatial distribution of water points productivity across groundwater potential classes confirmed the reliability of the generated maps, as higher flow-rate points were predominantly located within high and very high potential zones. The findings demonstrate that integrating machine learning and ensemble learning techniques provides reliable groundwater potential predictions and offers a useful framework for groundwater exploration and sustainable management in hetergeneous and data-limited regions.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"7 2","pages":"Article 100231"},"PeriodicalIF":4.2,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148177944","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-06-01Epub Date: 2026-05-20DOI: 10.1016/j.aiig.2026.100225
Felinto J. da Costa, Mariana R. Urbano
Rapid urban expansion poses significant challenges for land use planning, spanning infrastructure provision to environmental monitoring. Accurate and detailed classification of urban land cover (ULC) is essential to support evidence-based decision-making, particularly in complex and densely built environments. Although deep learning (DL) models dominate image-based classification tasks, most are pre-trained on general-purpose datasets and require adaptation to the specific characteristics of remote sensing (RS) data. However, full fine-tuning requires substantial computational resources. Parameter-Efficient Fine-Tuning (PEFT) methods, such as Low-Rank Adaptation (LoRA), offer a scalable alternative by updating only a small subset of parameters. This study investigates integrating LoRA into state-of-the-art (SOTA) vision models for multicategory ULC classification using high-resolution remote sensing (HRRS) imagery. We conducted a comprehensive comparative evaluation of 20 DL models across three benchmark RS image-classification datasets; when LoRA was applied, performance matched or exceeded baseline methods. However, while existing RS datasets provide ULC classes suitable for many purposes, urban morphological patterns vary substantially across world regions, ranging from compact to sprawling and hybrid forms with distinct built-environment characteristics. Therefore, we developed a new dataset (LDB10) with 2687 HRRS images of Londrina, a Brazilian mid-sized city, and a localized taxonomy covering 10 categories, including areas with variable building density, industrial zones, and green spaces of diverse types. We further evaluated the models on LDB10 and applied the best-performing model (DINOv2 ViT-L/14, 96.65% accuracy) to map ULC in Londrina and two neighboring municipalities, enabling detailed spatial and cross-domain analyses of urban, economic, and environmental patterns. The results validate the effectiveness of PEFT for adapting DL models to specific urban mapping tasks, facilitating the deployment of sophisticated DL solutions in municipal contexts with constrained computational resources.
快速的城市扩张给土地利用规划带来了巨大的挑战，从基础设施供应到环境监测。城市土地覆盖（ULC）的准确和详细分类对于支持基于证据的决策至关重要，特别是在复杂和密集建筑环境中。尽管深度学习（DL）模型主导着基于图像的分类任务，但大多数模型都是在通用数据集上进行预训练的，并且需要适应遥感（RS）数据的特定特征。然而，全面的微调需要大量的计算资源。参数高效微调（PEFT）方法，如低秩自适应（LoRA），提供了一种可扩展的替代方案，只需更新一小部分参数。本研究探讨了利用高分辨率遥感（HRRS）图像将LoRA集成到最先进的（SOTA）视觉模型中，用于多类别ULC分类。我们在三个基准RS图像分类数据集上对20个DL模型进行了全面的比较评估；当应用LoRA时，性能匹配或超过基线方法。然而，尽管现有的RS数据集提供了适合多种用途的ULC类别，但世界各地的城市形态模式差异很大，从紧凑型到蔓延型，以及具有不同建筑环境特征的混合形式。为此，本文利用巴西中等城市Londrina的2687幅HRRS影像，构建了一个新的数据集（LDB10），并建立了包含可变建筑密度区域、工业区和不同类型绿地等10个类别的局部分类系统。我们进一步在LDB10上对模型进行了评估，并将表现最佳的模型（DINOv2 viti - l /14，准确率为96.65%）应用于Londrina和两个邻近城市的ULC地图，从而对城市、经济和环境模式进行了详细的空间和跨域分析。研究结果验证了PEFT在使深度学习模型适应特定城市地图任务方面的有效性，促进了在计算资源受限的城市环境中部署复杂的深度学习解决方案。
{"title":"Efficient specialization of foundation vision models for urban land cover classification","authors":"Felinto J. da Costa, Mariana R. Urbano","doi":"10.1016/j.aiig.2026.100225","DOIUrl":"10.1016/j.aiig.2026.100225","url":null,"abstract":"<div><div>Rapid urban expansion poses significant challenges for land use planning, spanning infrastructure provision to environmental monitoring. Accurate and detailed classification of urban land cover (ULC) is essential to support evidence-based decision-making, particularly in complex and densely built environments. Although deep learning (DL) models dominate image-based classification tasks, most are pre-trained on general-purpose datasets and require adaptation to the specific characteristics of remote sensing (RS) data. However, full fine-tuning requires substantial computational resources. Parameter-Efficient Fine-Tuning (PEFT) methods, such as Low-Rank Adaptation (LoRA), offer a scalable alternative by updating only a small subset of parameters. This study investigates integrating LoRA into state-of-the-art (SOTA) vision models for multicategory ULC classification using high-resolution remote sensing (HRRS) imagery. We conducted a comprehensive comparative evaluation of 20 DL models across three benchmark RS image-classification datasets; when LoRA was applied, performance matched or exceeded baseline methods. However, while existing RS datasets provide ULC classes suitable for many purposes, urban morphological patterns vary substantially across world regions, ranging from compact to sprawling and hybrid forms with distinct built-environment characteristics. Therefore, we developed a new dataset (LDB10) with 2687 HRRS images of Londrina, a Brazilian mid-sized city, and a localized taxonomy covering 10 categories, including areas with variable building density, industrial zones, and green spaces of diverse types. We further evaluated the models on LDB10 and applied the best-performing model (DINOv2 ViT-L/14, 96.65% accuracy) to map ULC in Londrina and two neighboring municipalities, enabling detailed spatial and cross-domain analyses of urban, economic, and environmental patterns. The results validate the effectiveness of PEFT for adapting DL models to specific urban mapping tasks, facilitating the deployment of sophisticated DL solutions in municipal contexts with constrained computational resources.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"7 2","pages":"Article 100225"},"PeriodicalIF":4.2,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148177942","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-06-01Epub Date: 2026-05-22DOI: 10.1016/j.aiig.2026.100229
Ammar J. Abdlmutalib , Korhan Ayranci , Umair Bin Waheed , James A. MacEachern , Alessandro Traversa
Sedimentary structure classification is fundamental to facies interpretation, depositional environment reconstruction, and reservoir characterization. While Convolutional Neural Networks (CNNs) have demonstrated success in geological image analysis, their spatial-domain operations may limit their ability to capture repetitive and globally organized stratification patterns. In this study, we introduce a Fourier Neural Operator (FNO)-based framework for automated classification of eleven sedimentary structure classes from core images and benchmark it against EfficientNet-B2, ResNet-50, MobileNet-V3, and a transformer-based Vision Transformer (ViT), specifically the Data-efficient Image Transformer (DeiT-B16), under a harmonized 5-fold cross-validation protocol. The FNO achieved the highest overall accuracy (0.986) and macro-averaged F1-score (0.983), outperforming EfficientNet-B2 (0.975), DeiT-B16 (0.972), MobileNet-V3 (0.967), and ResNet-50 (0.960). Paired statistical tests across folds confirmed that performance differences between FNO and all benchmark models are statistically significant (p < 0.05). Precision-recall curves further demonstrate balanced class-wise performance despite dataset imbalance, with average precision (AP) values exceeding 0.98 across all classes. Ablation analysis revealed that performance gains are attributable to learned frequency-domain representations rather than increased model capacity. Removal of Fourier blocks leads to systematic degradation and confusion patterns. Additional frequency-domain visualizations provided an intuitive illustration of how representative bedding geometries are preserved in the retained low-frequency content and refined through the spectral layers of the FNO. An additional resolution-robustness analysis showed that FNO retained the highest prediction stability under progressive image degradation, indicating stronger tolerance to reduced image detail than the benchmark models. Despite containing substantially fewer parameters (0.01 M) and lower computational cost (0.21 GFLOPs), FNO achieved both the highest predictive accuracy and the lowest inference latency (0.0023 s per image). These results indicate that frequency-domain modeling provides measurable and statistically supported improvements for sedimentary structure classification, particularly for rhythmically layered structure facies.
{"title":"Deep learning with Fourier Neural Operators for sedimentary structure recognition","authors":"Ammar J. Abdlmutalib , Korhan Ayranci , Umair Bin Waheed , James A. MacEachern , Alessandro Traversa","doi":"10.1016/j.aiig.2026.100229","DOIUrl":"10.1016/j.aiig.2026.100229","url":null,"abstract":"<div><div>Sedimentary structure classification is fundamental to facies interpretation, depositional environment reconstruction, and reservoir characterization. While Convolutional Neural Networks (CNNs) have demonstrated success in geological image analysis, their spatial-domain operations may limit their ability to capture repetitive and globally organized stratification patterns. In this study, we introduce a Fourier Neural Operator (FNO)-based framework for automated classification of eleven sedimentary structure classes from core images and benchmark it against EfficientNet-B2, ResNet-50, MobileNet-V3, and a transformer-based Vision Transformer (ViT), specifically the Data-efficient Image Transformer (DeiT-B16), under a harmonized 5-fold cross-validation protocol. The FNO achieved the highest overall accuracy (0.986) and macro-averaged F1-score (0.983), outperforming EfficientNet-B2 (0.975), DeiT-B16 (0.972), MobileNet-V3 (0.967), and ResNet-50 (0.960). Paired statistical tests across folds confirmed that performance differences between FNO and all benchmark models are statistically significant (p < 0.05). Precision-recall curves further demonstrate balanced class-wise performance despite dataset imbalance, with average precision (AP) values exceeding 0.98 across all classes. Ablation analysis revealed that performance gains are attributable to learned frequency-domain representations rather than increased model capacity. Removal of Fourier blocks leads to systematic degradation and confusion patterns. Additional frequency-domain visualizations provided an intuitive illustration of how representative bedding geometries are preserved in the retained low-frequency content and refined through the spectral layers of the FNO. An additional resolution-robustness analysis showed that FNO retained the highest prediction stability under progressive image degradation, indicating stronger tolerance to reduced image detail than the benchmark models. Despite containing substantially fewer parameters (0.01 M) and lower computational cost (0.21 GFLOPs), FNO achieved both the highest predictive accuracy and the lowest inference latency (0.0023 s per image). These results indicate that frequency-domain modeling provides measurable and statistically supported improvements for sedimentary structure classification, particularly for rhythmically layered structure facies.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"7 2","pages":"Article 100229"},"PeriodicalIF":4.2,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148177943","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-06-01Epub Date: 2025-11-29DOI: 10.1016/j.aiig.2025.100170
Hua Wang , Meng Li , Qiang Wang , Shaopeng Shi , Gengxiao Yang , Zhilong Fang , Aihua Tao , Meng Wang
Cement bond quality evaluations are essential for assessing zonal isolation between formation strata, providing crucial information for ensuring environmental and ecological safety in oil and gas exploitation, geothermal energy injection and geological carbon dioxide sequestration. In the past decade, the ultrasonic pulse-echo and pitch-catch logging techniques have emerged as effective and non-destructive methods for quantitatively evaluating bond quality at both the casing-cement and cement-formation interfaces. This review presents a comprehensive overview of recent advancements in cement bond quality assessment based on ultrasonic measurements. Key developments include automatic waveform quality assessment, inversion techniques for mud and cement impedance, tool trajectory corrections, separation of flexural and extensional mode waves, machine learning-based extraction and enhancement of TIE waveforms, and imaging of the cement-formation interface using the reverse time migration approach. The review thoroughly explores the methodological principles and applications of these techniques, supported by synthetic datasets, full-scale physical well experiments, and field well data. Considering the recent progress in machine learning and the growing availability of advanced computational resources, we highlight the most significant achievements and ongoing challenges in data processing, while discussing the potential advancements these techniques could offer in the near future.
{"title":"Recent advances and challenges of cement bond evaluation based on ultrasonic measurements in cased holes","authors":"Hua Wang , Meng Li , Qiang Wang , Shaopeng Shi , Gengxiao Yang , Zhilong Fang , Aihua Tao , Meng Wang","doi":"10.1016/j.aiig.2025.100170","DOIUrl":"10.1016/j.aiig.2025.100170","url":null,"abstract":"<div><div>Cement bond quality evaluations are essential for assessing zonal isolation between formation strata, providing crucial information for ensuring environmental and ecological safety in oil and gas exploitation, geothermal energy injection and geological carbon dioxide sequestration. In the past decade, the ultrasonic pulse-echo and pitch-catch logging techniques have emerged as effective and non-destructive methods for quantitatively evaluating bond quality at both the casing-cement and cement-formation interfaces. This review presents a comprehensive overview of recent advancements in cement bond quality assessment based on ultrasonic measurements. Key developments include automatic waveform quality assessment, inversion techniques for mud and cement impedance, tool trajectory corrections, separation of flexural and extensional mode waves, machine learning-based extraction and enhancement of TIE waveforms, and imaging of the cement-formation interface using the reverse time migration approach. The review thoroughly explores the methodological principles and applications of these techniques, supported by synthetic datasets, full-scale physical well experiments, and field well data. Considering the recent progress in machine learning and the growing availability of advanced computational resources, we highlight the most significant achievements and ongoing challenges in data processing, while discussing the potential advancements these techniques could offer in the near future.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"7 1","pages":"Article 100170"},"PeriodicalIF":4.2,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145698039","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-06-01Epub Date: 2026-01-13DOI: 10.1016/j.aiig.2026.100187
{"title":"Thank you reviewers!","authors":"","doi":"10.1016/j.aiig.2026.100187","DOIUrl":"10.1016/j.aiig.2026.100187","url":null,"abstract":"","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"7 2","pages":"Article 100187"},"PeriodicalIF":5.4,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148553777","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In Indonesia, wildfires have become an annual disaster that results in significant losses across various aspects of life, including ecological, social, and economic conditions. To minimize these losses, accurate wildfire predictions are urgently needed for prevention, early detection, and wildfire management decision support. This study employs an ensemble learning approach to develop a prediction model for wildfire occurrences and create a susceptibility map of fire-prone areas on a national scale in Indonesia. The proposed model is Stacking Ensemble Learning (SEL), which integrates K-Nearest Neighbor (KNN), Adaptive Boosting (AdaBoost), and Extreme Gradient Boosting (XGBoost) as the base learners, with Random Forest (RF) serving as the meta-learner. The final performance results indicate an accuracy 0.985, balanced accuracy 0.97, precision 0.96, average precision 0.896, recall 0.92, F1-Score 0.954, Matthew's Correlation Coefficient (MCC) 0.944, and AUC Score 0.97, with no signs of overfitting and optimal computational efficiency. SHAP Explainable AI analysis is employed to identify the most influential factors, revealing that proximity to residential areas and climate factors are the most significant contributors to wildfire occurrences in Indonesia. The susceptibility mapping results highlight the provinces with the largest and most vulnerable areas: West Kalimantan, South Sumatra, and South Sulawesi. The outcome of this study can assist stakeholders in mitigating wildfires and protecting the environment to achieve sustainable development goals.
{"title":"GIS-based wildfire prediction model in Indonesia using stacking ensemble learning","authors":"Clarisha Hanandya Puspitadewi , Arian Dhini , Enrico Laoh , Dodi Sudiana","doi":"10.1016/j.aiig.2026.100228","DOIUrl":"10.1016/j.aiig.2026.100228","url":null,"abstract":"<div><div>In Indonesia, wildfires have become an annual disaster that results in significant losses across various aspects of life, including ecological, social, and economic conditions. To minimize these losses, accurate wildfire predictions are urgently needed for prevention, early detection, and wildfire management decision support. This study employs an ensemble learning approach to develop a prediction model for wildfire occurrences and create a susceptibility map of fire-prone areas on a national scale in Indonesia. The proposed model is Stacking Ensemble Learning (SEL), which integrates K-Nearest Neighbor (KNN), Adaptive Boosting (AdaBoost), and Extreme Gradient Boosting (XGBoost) as the base learners, with Random Forest (RF) serving as the meta-learner. The final performance results indicate an accuracy 0.985, balanced accuracy 0.97, precision 0.96, average precision 0.896, recall 0.92, F1-Score 0.954, Matthew's Correlation Coefficient (MCC) 0.944, and AUC Score 0.97, with no signs of overfitting and optimal computational efficiency. SHAP Explainable AI analysis is employed to identify the most influential factors, revealing that proximity to residential areas and climate factors are the most significant contributors to wildfire occurrences in Indonesia. The susceptibility mapping results highlight the provinces with the largest and most vulnerable areas: West Kalimantan, South Sumatra, and South Sulawesi. The outcome of this study can assist stakeholders in mitigating wildfires and protecting the environment to achieve sustainable development goals.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"7 2","pages":"Article 100228"},"PeriodicalIF":4.2,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148177941","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-06-01Epub Date: 2026-06-06DOI: 10.1016/j.aiig.2026.100232
Abdullah Ali Ali Hussein , Chunhua Qiu , Ibrahim Althamary , Peng Xiao , Jiangbo Wang , Lu Li , Jie Ren
Undrained shear strength () is a key parameter for evaluating slope stability, offshore foundation design, and submarine geohazards in marine environments. Conventional methods for predicting often fall short in accuracy because they fail to account for the complex and nonlinear relationships among sediment properties. Here, we propose a physics-informed neural network (PINN) framework to predict the three-dimensional structure of using various observed physical parameters, including bulk density, porosity, P-wave velocity, gamma ray attenuation, and natural gamma ray. The framework embeds governing physical laws—total vertical stress , pore water pressure , and effective stress σ′(z)—as constraints within the loss function to ensure physically consistent and accurate predictions. The results show that our physics-informed model significantly improves prediction accuracy and stability, achieving R2 up to 0.91 (mean ≈ 0.85) and reducing prediction error by more than 18% for marine sediments compared to purely data-driven models. Sensitivity analysis highlights that bulk density and porosity are the most influential inputs for predicting , consistent with their fundamental role in sediment consolidation and strength. The predicted exhibits a northwest–southeast gradient, with the strongest increase at depths of approximately 100–200 m below the seafloor. The proposed model provides spatially continuous maps and demonstrates that embedding first-order physical laws in network training greatly improves the reliability and interpretability of predictions, offering a practical tool for marine geotechnical applications in areas with complex stratigraphy.
{"title":"Predicting undrained shear strength in marine sediments using a physics-informed neural network (PINN)","authors":"Abdullah Ali Ali Hussein , Chunhua Qiu , Ibrahim Althamary , Peng Xiao , Jiangbo Wang , Lu Li , Jie Ren","doi":"10.1016/j.aiig.2026.100232","DOIUrl":"10.1016/j.aiig.2026.100232","url":null,"abstract":"<div><div>Undrained shear strength (<span><math><mrow><msub><mi>S</mi><mi>U</mi></msub></mrow></math></span>) is a key parameter for evaluating slope stability, offshore foundation design, and submarine geohazards in marine environments. Conventional methods for predicting <span><math><mrow><msub><mi>S</mi><mi>U</mi></msub></mrow></math></span> often fall short in accuracy because they fail to account for the complex and nonlinear relationships among sediment properties. Here, we propose a physics-informed neural network (PINN) framework to predict the three-dimensional structure of <span><math><mrow><msub><mi>S</mi><mi>U</mi></msub></mrow></math></span> using various observed physical parameters, including bulk density, porosity, P-wave velocity, gamma ray attenuation, and natural gamma ray. The framework embeds governing physical laws—total vertical stress <span><math><mrow><mi>σ</mi><mrow><mo>(</mo><mi>z</mi><mo>)</mo></mrow></mrow></math></span>, pore water pressure <span><math><mrow><mi>u</mi><mrow><mo>(</mo><mi>z</mi><mo>)</mo></mrow></mrow></math></span>, and effective stress σ′(z)—as constraints within the loss function to ensure physically consistent and accurate predictions. The results show that our physics-informed model significantly improves prediction accuracy and stability, achieving R<sup>2</sup> up to 0.91 (mean ≈ 0.85) and reducing prediction error by more than 18% for marine sediments compared to purely data-driven models. Sensitivity analysis highlights that bulk density and porosity are the most influential inputs for predicting <span><math><mrow><msub><mi>S</mi><mi>U</mi></msub></mrow></math></span>, consistent with their fundamental role in sediment consolidation and strength. The predicted <span><math><mrow><msub><mi>S</mi><mi>U</mi></msub></mrow></math></span> exhibits a northwest–southeast gradient, with the strongest increase at depths of approximately 100–200 m below the seafloor. The proposed model provides spatially continuous <span><math><mrow><msub><mi>S</mi><mi>U</mi></msub></mrow></math></span> maps and demonstrates that embedding first-order physical laws in network training greatly improves the reliability and interpretability of <span><math><mrow><msub><mi>S</mi><mi>U</mi></msub></mrow></math></span> predictions, offering a practical tool for marine geotechnical applications in areas with complex stratigraphy.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"7 2","pages":"Article 100232"},"PeriodicalIF":4.2,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148241065","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-06-01Epub Date: 2025-12-18DOI: 10.1016/j.aiig.2025.100182
A. Sanches, B. Ağbulut, L. Castro, M. Vieira
{"title":"Corrigendum to “Machine learning applied to recognition of dinoflagellate cysts: Type study with the species Batioladinium longicornutum” [Artif. Intell. Geosciences, (6), Issue 2, 2025, 100150]","authors":"A. Sanches, B. Ağbulut, L. Castro, M. Vieira","doi":"10.1016/j.aiig.2025.100182","DOIUrl":"10.1016/j.aiig.2025.100182","url":null,"abstract":"","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"7 2","pages":"Article 100182"},"PeriodicalIF":5.4,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148553617","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2026-02-25DOI: 10.1016/j.aiig.2026.100197
Zhaoxing Xu, Peng Liu, Qinqin Wu
Seismic impedance inversion is a key technique for extracting reservoir information from seismic data. Traditional model-driven inversion methods often prove inadequate when dealing with complex reservoirs, which has led to the growing adoption of data-driven sparse representation constrained inversion approaches. These methods typically employ redundant dictionary learning to adaptively extract feature information from logging data for effective inversion constraints. Although they excel in enhancing the vertical resolution and accuracy of inversion results, they still suffer from limitations such as high computational complexity and a lack of horizontal feature constraints, resulting in insufficient horizontal continuity, overall accuracy, and computational efficiency. To address these issues, this paper proposes a fast sparse representation-based impedance inversion method using online adaptive reservoir features. Based on logging and seismic data, the method employs an online dictionary learning strategy to adaptively extract both vertical and horizontal reservoir characteristics for sparse representation inversion constraints. To further improve computational efficiency, orthogonal dictionary learning is introduced to reduce computational costs. Ultimately, an impedance inversion method is developed based on online orthogonal dictionary learning that simultaneously imposes adaptive joint constraints on both vertical and horizontal features. Experimental results demonstrate that the proposed method not only achieves high accuracy and high-resolution inversion results but also offers significant advantages in computational efficiency.
{"title":"Fast sparse representation impedance inversion method based on online adaptive reservoir characterization","authors":"Zhaoxing Xu, Peng Liu, Qinqin Wu","doi":"10.1016/j.aiig.2026.100197","DOIUrl":"10.1016/j.aiig.2026.100197","url":null,"abstract":"<div><div>Seismic impedance inversion is a key technique for extracting reservoir information from seismic data. Traditional model-driven inversion methods often prove inadequate when dealing with complex reservoirs, which has led to the growing adoption of data-driven sparse representation constrained inversion approaches. These methods typically employ redundant dictionary learning to adaptively extract feature information from logging data for effective inversion constraints. Although they excel in enhancing the vertical resolution and accuracy of inversion results, they still suffer from limitations such as high computational complexity and a lack of horizontal feature constraints, resulting in insufficient horizontal continuity, overall accuracy, and computational efficiency. To address these issues, this paper proposes a fast sparse representation-based impedance inversion method using online adaptive reservoir features. Based on logging and seismic data, the method employs an online dictionary learning strategy to adaptively extract both vertical and horizontal reservoir characteristics for sparse representation inversion constraints. To further improve computational efficiency, orthogonal dictionary learning is introduced to reduce computational costs. Ultimately, an impedance inversion method is developed based on online orthogonal dictionary learning that simultaneously imposes adaptive joint constraints on both vertical and horizontal features. Experimental results demonstrate that the proposed method not only achieves high accuracy and high-resolution inversion results but also offers significant advantages in computational efficiency.</div></div>","PeriodicalId":100124,"journal":{"name":"Artificial Intelligence in Geosciences","volume":"7 1","pages":"Article 100197"},"PeriodicalIF":4.2,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147396653","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号