Remote Sensing Applications-Society and Environment最新文献

{"title":"TabPFN for crop classification using multi-temporal ALOS-4/PALSAR-3 SAR data under imbalanced conditions","authors":"Rei Sonobe","doi":"10.1016/j.rsase.2026.102185","DOIUrl":"10.1016/j.rsase.2026.102185","url":null,"abstract":"<div><div>Accurate crop classification using synthetic aperture radar (SAR) data is essential for operational agricultural monitoring but remains challenging due to complex scattering mechanisms, correlated features, and class imbalance in reference datasets. This study evaluates the applicability of Tabular Prior-Data Fitted Networks (TabPFN) for SAR-based crop classification and compares its performance with Random Forest, Support Vector Machine, and Extreme Gradient Boosting. Field-level classification was conducted using multi-temporal SAR-derived features, and model performance was assessed through cross-validation with fold numbers ranging from 2 to 10. Overall Accuracy, Cohen's Kappa, class-wise User's Accuracy, Producer's Accuracy, and F1-score were examined under four sampling strategies: no oversampling, and three Synthetic Minority Over-sampling Technique (SMOTE)-based approaches (Borderline-SMOTE, DeepSMOTE, and KMeans-SMOTE). Permutation feature importance was used to ensure consistent interpretation across models. Results show that TabPFN consistently achieved the highest classification performance across all validation settings, reaching an Overall Accuracy of up to 91.9% and a Cohen's Kappa of up to 0.896, outperforming the best-performing conventional machine learning model. Results also demonstrated strong robustness to class imbalance, with the best performance obtained using the original imbalanced dataset without any SMOTE-based oversampling. These findings suggest that model selection may be more influential than synthetic oversampling in SAR-based crop mapping. Furthermore, the robustness of TabPFN may reduce the reliance on additional oversampling procedures, thereby simplifying the preprocessing workflow for operational applications. Taken together, the results highlight the potential of TabPFN as a practical and effective approach for agricultural monitoring and crop mapping.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"43 ","pages":"Article 102185"},"PeriodicalIF":4.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148731802","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}

{"title":"Evaluation of remote sensing spectral indices in monitoring gregarious desert locusts using Sentinel-2 data in Google Earth Engine","authors":"Gachie Eliud Baraka,&nbsp;Guido D'Urso,&nbsp;Oscar Rosario Belfiore,&nbsp;Vahid Akbari,&nbsp;Anthony O'Hare,&nbsp;Gavin Abernethy","doi":"10.1016/j.rsase.2026.102171","DOIUrl":"10.1016/j.rsase.2026.102171","url":null,"abstract":"<div><div>Desert locusts are some of the most destructive agricultural pests globally. However, besides Normalised Difference Vegetation Index, Enhanced Vegetation Index and Leaf Area Index, there are limited publications on the application of other remote sensing spectral indices in monitoring desert locusts. This study evaluated the application of 12 spectral indices in monitoring gregarious desert locusts in Loiyangalani ward, Kenya. The study used 442 desert locust presence reports collected in 2020 and 2021 to generate time series data of remote sensing spectral indices from Sentinel 2 data in Google Earth Engine. The study visualised time series data using a line graph, before analysing the correlation and regression. The findings revealed that the peak of mature swarm and hopper band incidents coincided with the highest spectral index values in May 2020. However, the peak of immature swarm incidetns coincided with low spectral index values in February 2021. The Modified Soil Adjusted Vegetation Index produced the best results (β=0.711; t=3.354; p=0.006) for monitoring mature swarms and hoppers because it accounts for soil background interference in Loiyangalani, which is an arid and semi-arid ward. The study concludes that remote sensing spectral indices can be used to map favourable seasons and identify suitable breeding areas for mature swarms of gregarious desert locust, but may not reliably identify when and where immature swarms might be present in invasion zones. The study recommends careful application of remote sensing spectral indices in monitoring gregarious desert locusts, based on site-specific vegetation and soil properties.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"43 ","pages":"Article 102171"},"PeriodicalIF":4.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148635661","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}

{"title":"Projecting and mitigating urban heat futures in arid cities: A machine learning framework applied to Jeddah, Saudi Arabia","authors":"Alshimaa Aboelmakarem Farag","doi":"10.1016/j.rsase.2026.102186","DOIUrl":"10.1016/j.rsase.2026.102186","url":null,"abstract":"<div><div>The rapid urbanization of arid cities intensifies the Urban Heat Island (UHI), affecting environmental sustainability and public health. Although links between urban expansion and rising Land Surface Temperature (LST) are documented, land-cover associations with LST change in arid environments and predictive frameworks for policy assessment remain underdeveloped. This study integrates remote sensing, GIS, and machine learning to examine associations between decadal LST change and land-cover index change in Jeddah, Saudi Arabia (2015–2025). Using Landsat-derived data for 196 districts, changes in LST and in built-up (NDBI), vegetation (SAVI), water (MNDWI), and bare soil (NDBaI) indices were analyzed. Ridge Regression, Random Forest, Gradient Boosting, and an artificial neural network (ANN) were compared; the selected ANN (test R² = 0.88; RMSE = 0.56 °C) was used to project mean ΔLST for 2030 and 2040 under Business-as-Usual, Accelerated Urbanization, and Sustainable Development scenarios. Mean LST rose by +2.39 °C over 10 years (2015-2025). Among the four indices examined, ΔNDBaI showed the strongest preliminary bivariate association with ΔLST (r = −0.47, p &lt; 0.05; r² ≈ 0.22). Exploratory projections diverged markedly by 2040: +8.33 °C under Accelerated Urbanization versus −0.78 °C under Sustainable Development (∼9.11 °C contrast under assumed index trajectories). Compared with prior Jeddah studies, this work links district-scale land-cover change, comparative machine learning evaluation, and multi-scenario thermal projection. Results highlight green–blue infrastructure as planning-relevant levers for mitigating urban heat in arid cities.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"43 ","pages":"Article 102186"},"PeriodicalIF":4.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148730274","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}

{"title":"A domain-informed optimization and machine learning framework for land subsidence susceptibility mapping and building exposure assessment: A case study of Tehran, Iran","authors":"Reza Saeed,&nbsp;Mahdis Rezapour,&nbsp;Armin Moghimi,&nbsp;Mahdi Motagh,&nbsp;Ali Mohammadzadeh,&nbsp;Mohammad Javad Valadan Zoej","doi":"10.1016/j.rsase.2026.102195","DOIUrl":"10.1016/j.rsase.2026.102195","url":null,"abstract":"<div><div>Land subsidence (LS) is a major geohazard causing irreversible damage to urban infrastructure and substantial socioeconomic and environmental impacts. This study presents a domain-informed optimization and machine-learning framework that combines Genetic Algorithm (GA), Bayesian-guided Genetic Algorithm (BayGA), and Marine Predators Algorithm (MPA) optimization with Support Vector Machine (SVM), Random Forest (RF), and eXtreme Gradient Boosting (XGB) models to map LS susceptibility (LSS) in the Tehran metropolitan area, Iran. A dual-resolution grid sampling was employed to automatically generate subsidence and non-subsidence samples from an open InSAR-derived inventory data. Twenty-six environmental and anthropogenic conditioning factors representing the principal controls on LS were incorporated. Rather than optimizing all predictors, the proposed framework retains essential process-related variables as baseline features and optimizes only the remaining variables using metaheuristic feature selection. This strategy preserves physically meaningful information while improving model interpretability and robustness over conventional data-driven feature-selection approaches. The results showed that all models achieved high predictive performance (ROC–AUC &gt; 0.96) under the adopted random point-level validation strategy. BayGA-XGB achieved the highest accuracy (96%) and F1-score (95%), although its advantage over the other leading XGB- and RF-based models was marginal (0.2 to 0.8 percentage points in F1-score). SHapley Additive exPlanations (SHAP) identified elevation, sediment properties, geology, fault proximity, and groundwater-level decline as the dominant predictors. Overlay with Microsoft Building Footprints showed concentrated exposure in the southern and central districts. Groundwater trends for 2012–2019 provided a hydrogeological consistency check for the physical plausibility of the mapped susceptibility patterns. The framework enables interpretable regional LSS and exposure analysis for urban planning in Tehran and comparable metropolitan settings.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"43 ","pages":"Article 102195"},"PeriodicalIF":4.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148730278","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}

{"title":"KDAQnet: A lightweight and accurate deep learning framework for satellite-based multi-pollutant air quality forecasting with knowledge distillation","authors":"Sudhir Kumar,&nbsp;Vaneet Kour,&nbsp;Nimisha Jain,&nbsp;Shivendu Mishra,&nbsp;Rajiv Misra,&nbsp;T.N. Singh","doi":"10.1016/j.rsase.2026.102203","DOIUrl":"10.1016/j.rsase.2026.102203","url":null,"abstract":"<div><div>Air pollution forecasting has become increasingly vital for public health and environmental management, particularly with the growing availability of satellite-based datasets. Several studies have proposed deep learning models to predict air quality using such datasets; however, these approaches often suffer from high computational costs due to large model parameters. To address this challenge, we introduce KDAQnet, a lightweight and efficient Knowledge Distillation (KD) based deep learning framework for multi-pollutant air quality prediction. KDAQnet is based on a ResNet+FCN architecture that effectively extract spatial and temporal features. It demonstrates superior performance across five major pollutants, CH₄, CO, NO₂, O₃, and SO₂, achieving an average <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> score of 0.8710, while reducing MSE and MAE to 544.66 and 13.20, respectively. The training time and model complexity are also reduced to 20,033.92 s and 84%, respectively, underscoring the model’s efficiency. Furthermore, comparisons against baseline architectures such as ResNet+Transformer, CNN+FCN, AlexNet+Transformer, and SQUnet+Transformer confirm its consistent superiority. Additionally, a KD-based ablation study using varying <span><math><mi>α</mi></math></span> values shows the student model achieves up to 9.38% higher <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> and approximately 50% faster convergence. These results establish KDAQnet as a scalable and effective solution for real-time, multi-pollutant air quality forecasting using satellite data and ground truth.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"43 ","pages":"Article 102203"},"PeriodicalIF":4.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148731700","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}

{"title":"Detection of potential burial-related sites through open-source geospatial investigation and remote sensing: evidence from Russian-occupied areas in Ukraine (2022–2025)","authors":"Giampiero D'Amico,&nbsp;Maria Petriccione,&nbsp;Aldino Bondesan","doi":"10.1016/j.rsase.2026.102208","DOIUrl":"10.1016/j.rsase.2026.102208","url":null,"abstract":"<div><div>Armed conflicts inscribe measurable signatures on the landscape, and these can support humanitarian documentation when field access is constrained. We develop and test an OSINT-led, multi-temporal remote-sensing workflow to identify and document potential burial-related ground disturbance in Russian-occupied areas of Ukraine, with a core analytical window spanning February 2022 to May 2023 and selected sites followed into 2024–2025. Candidate locations across nine sites in Donetsk and Luhansk were prioritised through the Eyes on Russia WebGIS using targeted keyword queries and then examined through repeated photointerpretation of very-high-resolution optical imagery (Google Earth Pro, ESRI World Imagery Wayback, Pleiades, Pleiades Neo) within QGIS. Manual digitisation, near-infrared false-colour inspection, and FIRMS/NASA thermal anomalies are integrated to characterise disturbance morphology, quantify change through time, and constrain interpretation through contextual indicators of hostilities. Results are reported as case-based, measurable landscape change instead of index-derived anomaly counts, with confidence in site attributions discussed through qualitative triangulation between OSINT audiovisual material, multi-temporal satellite observations, and conflict dynamics. The workflow is desk-based and intended as a preliminary screening and prioritisation step within a longer investigative chain, for settings where field access is precluded by ongoing conflict. It does not substitute field verification or provide definitive victim counts, and outputs are presented following a do no harm rationale. Its purpose is to provide an immediate, evidence-based identification of areas potentially affected by burial-related activity, so that investigative and humanitarian resources can be directed efficiently once access becomes possible.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"43 ","pages":"Article 102208"},"PeriodicalIF":4.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148731701","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}

{"title":"Habitat (dis)connectivity modulates spatio-temporal dynamics of waterbird community assemblages in humanized floodplain wetlandscapes: Insights from eastern India","authors":"Biyas Dey,&nbsp;Debajit Datta,&nbsp;Proshanta Kumar Ghosh,&nbsp;Aznarul Islam,&nbsp;Biplab Sarkar","doi":"10.1016/j.rsase.2026.102166","DOIUrl":"10.1016/j.rsase.2026.102166","url":null,"abstract":"<div><div>Waterbird communities of tropical floodplain wetlands, being sensitive to changes in habitat characteristics, are regarded as key bio-indicators of landscape health. This study investigates the spatio-temporal correspondence between waterbird habitability and landscape-level environmental conditions in the Ganga-Ichhamati Interfluve, India, integrating empirical observations from 2017 to 2023 with a simulated 2029 projection; i.e. how the composition, configuration, and connectivity of habitat patches affect waterbird habitability. While various landscape-level spatial metrics are parameterized as environmental covariates; several ecological indicators based on habitat-specific line-transect data are applied to construct a community ecological structure index of waterbirds (CESI_W). Multiple machine-learning techniques were applied to assess the relationship between CESI_W and applied covariates, among which the Extreme Gradient Boosting showed best performance (R² = 0.85, mean squared error = 0.35). Using this, the waterbird habitability models were developed and validated for 2017 and 2023, followed by near-term projection of the 2029 scenario. Among the covariates, area of waterbody emerged as the most influential predictor with highest Shapley value (0.70), followed by hydrological connectivity index (0.24) and landscape shape index (0.13). The findings denoted alarming contractions of 19.74% in core habitat patches and 43.85% in high-quality corridors from 2017 to 2029. Clustered wetlands with persistent hydrological connectivity exhibited better habitability, compared to the isolated ones. This was followed by the presence of vegetated corridors between wetlands amidst the densely populated floodplain. Accordingly, we argue that scalable tethering of these connectivity and modulating features into landscape-level planning would serve efficiently for site-specific interventions in sustainable floodplain management.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"43 ","pages":"Article 102166"},"PeriodicalIF":4.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148635658","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}

{"title":"Microwave-guided OPTRAM: A novel approach for mapping soil moisture by fusing SMAP and MODIS satellite observations","authors":"Neda Mohamadzadeh,&nbsp;Morteza Sadeghi,&nbsp;Jida Wang,&nbsp;Gaurav Jha,&nbsp;Lan Liang,&nbsp;Uditha Bandara,&nbsp;Craig Altare,&nbsp;Tyler Hatch,&nbsp;Marcellus M. Caldas","doi":"10.1016/j.rsase.2026.102179","DOIUrl":"10.1016/j.rsase.2026.102179","url":null,"abstract":"<div><div>This study presents Microwave-Guided OPTRAM (MG-OPTRAM), an enhanced approach that fuses microwave and optical satellite data for improved remote sensing of soil moisture. Building upon the Optical TRApezoid Model (OPTRAM), which estimates soil moisture based on the relationship between vegetation index and surface reflectance in optical imagery, MG-OPTRAM enhances soil moisture retrieval by incorporating Soil Moisture Active Passive (SMAP) observations into the calibration process. The method directly calibrates OPTRAM using SMAP data at its native 9-km resolution and applies the calibrated model at Moderate Resolution Imaging Spectroradiometer (MODIS) 500-m resolution. Unlike conventional downscaling methods that rely on land surface temperature, MG-OPTRAM leverages optical surface reflectance, enabling a one-time calibration using physical relationships and extending soil moisture mapping beyond SMAP's operational period (2015–present) back to 2000. Validation of MG-OPTRAM at coarse resolution against SMAP data across Central Valley, California, from 2020 to 2022 shows strong correlation with a Coefficient of Determination (R²) of 0.75 and a Root Mean Square Error (RMSE) of 0.04 m³m^-3. Comparisons with in-situ measurements over longer period (where available) yield lower correlations (R² ranging from 0.59 to 0.81) and site-level biases of 0.02-0.09 m³m^-3, likely due to spatial mismatches between point-scale observations and 9-km pixel-averaged estimates. At 500-m resolution, MG-OPTRAM reconstructs SMAP-like soil moisture with high temporal fidelity and improves spatial resolution, revealing detailed landscape features not captured by SMAP. The correlation with in-situ data at this finer scale slightly decreases (R² ranging from 0.53 to 0.75) and biases increases at most sites, due to model simplicity and subpixel heterogeneity. However, MG-OPTRAM consistently outperforms the original OPTRAM in accuracy. Overall, MG-OPTRAM offers a scalable framework for long-term, high-resolution soil moisture monitoring using widely available satellite data.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"43 ","pages":"Article 102179"},"PeriodicalIF":4.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148635660","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}

{"title":"Long-term satellite reconstruction of harmful algal bloom dynamics in Lake Cascade (1972–2024): Linking remote sensing with hydrochemical drivers","authors":"Saeed Poorasadollah,&nbsp;Omid Abdi,&nbsp;Sophia Hghani","doi":"10.1016/j.rsase.2026.102189","DOIUrl":"10.1016/j.rsase.2026.102189","url":null,"abstract":"<div><div>Harmful algal blooms (HABs) are increasingly recognized as manifestations of interacting ecohydrological processes rather than solely responses to nutrient enrichment. However, long-term investigations integrating satellite observations with hydrochemical monitoring remain limited for regulated reservoirs. This study developed an integrated remote sensing–ecohydrological framework to reconstruct HAB dynamics in Lake Cascade (Idaho, USA) using multi-decadal Landsat observations (1972–2024), Sentinel-2 imagery and long-term hydrochemical records. Satellite-derived indicators, including the Normalized Difference Vegetation Index (NDVI), Normalized Difference Turbidity Index (NDTI), bloom area, and lake area, were integrated with nutrient, dissolved oxygen, temperature, and stoichiometric datasets to evaluate bloom occurrence, environmental controls, ecological transitions, and spatial dynamics. Results demonstrate that HABs have been a persistent feature of Lake Cascade for more than five decades, with the most severe events occurring during 1999–2001 when extensive blooms coincided with elevated nutrient concentrations, hypolimnetic oxygen depletion, and seasonal reservoir contraction. Correlation analyses revealed stronger associations between bloom indicators and nutrient stoichiometry, oxygen dynamics, and ecohydrological variables than with individual nutrient concentrations alone. Pettitt change-point, Kruskal–Wallis, and principal component analyses identified significant ecological reorganization characterized by declining bloom intensity, reduced bloom extent, altered nutrient–bloom coupling, and increasing ecosystem variability after the early 2000s. Satellite observations further revealed a transition from spatially extensive, high-intensity blooms toward more persistent but generally lower-density blooms during recent years. These findings demonstrate that long-term satellite observations, when integrated with hydrochemical monitoring, provide a robust framework for evaluating HAB evolution, ecohydrological controls, and ecological regime shifts in regulated freshwater systems.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"43 ","pages":"Article 102189"},"PeriodicalIF":4.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148635729","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}

{"title":"Block-level drought vulnerability zonation using remote sensing and explainable machine learning in lateritic Rarh plateau, Eastern India","authors":"Moumita Kundu,&nbsp;Arnab Ghosh","doi":"10.1016/j.rsase.2026.102207","DOIUrl":"10.1016/j.rsase.2026.102207","url":null,"abstract":"<div><div>In heterogenous lateritic landscapes, drought assessments require integrated frameworks that combine expert and data-driven approaches, for reliable ecological zonation. This study introduces a new Fuzzy AHP–XGBoost–SHAP approach to achieve block-level zonation of the water stress in the lateritic plateau of Rarh (Paschim Medinipur and Jhargram districts, west Bengal, India) in 29 community development blocks. Multi-scale SPEI analysis (1980-2025) showed that there are decadal oscillations, with a notable drying episode (1996-2010) corresponding to the El Niño suppressed monsoon. The novelty is not only on methods, but rather the combination of expert-driven and data-driven evidence as mutually interpreting layers, with their agreement quantified, and the extraction of operational non-linear stress thresholds.The Water Stress Zonation Index (WSZI) was consequently developed by synthesizing nine satellite spectral–thermal indices and SPEI anomaly using triangular fuzzy number–weighted AHP, which clearly revealed a northwest to southeast gradient in the extent of vulnerability with ∼73.5% of the region falling under moderate-to-high stress. The XGBoost–SHAP analysis was used not as independent validation, but to break down the WSZI and reveal the dominant drivers (TVDI, LST, SPEI) and non-linear stress thresholds, and a strong inverse correlation with independent validation using a block-wise Yield Anomaly Index was found across the four years of assessment (ρ = −0.76 to −0.90, p &lt; 0.001). From the compound-factor prioritization, Chandrakona II, Garbeta III and Salboni were identified as the most vulnerable blocks and this could be replicated for spatial prioritization of drought mitigation in lateritic ecosystem.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"43 ","pages":"Article 102207"},"PeriodicalIF":4.9,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148731807","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}