Mondal et al. (2026) Advancements in Spatio-temporal agricultural drought monitoring and modeling: a comprehensive review on multi-source remote sensing and machine learning techniques

Identification

• Journal: Theoretical and Applied Climatology
• Year: 2026
• Date: 2026-03-18
• Authors: Suresh Mondal, Kumar Arun Prasad, A. L. Achu, S. Kaliraj, K. Balasubramani
• DOI: 10.1007/s00704-026-06136-8

Research Groups

• Department of Geography, School of Earth Sciences, Central University of Tamil Nadu, Thiruvarur, India
• Department of Climate Variability and Aquatic Ecosystems, Kerala University of Fisheries and Ocean Studies (KUFOS), Kochi, Kerala, India
• National Centre for Earth Science Studies (NCESS), Ministry of Earth Sciences, Thiruvananthapuram, Kerala, India

Short Summary

This comprehensive review synthesizes advancements in spatio-temporal agricultural drought monitoring and modeling, focusing on the integration of multi-source remote sensing data with machine learning (ML) and deep learning (DL) techniques. It highlights the effectiveness, cost-efficiency, and transferability of these advanced geospatial methods for assessing and predicting agricultural drought conditions across various scales.

Objective

• To comprehensively synthesize the contributions of multi-source remote sensing data and artificial intelligence (AI) approaches (including machine learning and deep learning) for monitoring, evaluating, and modeling agricultural drought.
• To review the latest developments in time-series satellite data analysis, remote-sensing spectral indices, and numerical models used globally for rapid and accurate drought monitoring and modeling.
• To critically examine case studies of drought assessment using ML/DL coupled with satellite, multi-parametric, and ground-based data for local and regional scale monitoring and prediction.

Study Configuration

• Spatial Scale: Global, regional, and local scales, covering diverse agroclimatic regions worldwide.
• Temporal Scale: Review of literature spanning from 1974 to 2025, focusing on long-term time series observations and near-real-time monitoring capabilities.

Methodology and Data

• Models used:
  • Machine Learning (ML): Decision Tree models (CART, Conditional Inference Trees, Multivariate Regression Trees), Support Vector Machine (SVM) (Linear SVM, Kernel SVMs, Support Vector Regression, One-Class SVM), Boosting models (AdaBoost, Gradient Boosting Machines, XGBoost, CatBoost), Ensemble models (Random Forest, Bagging, Extremely Random Trees, Stacking), Artificial Neural Networks (ANN), Multilayer Perceptron (MLP) NNs.
  • Deep Learning (DL): Convolutional Neural Networks (CNN), Autoencoders, Long Short-Term Memory (LSTM), Graph Neural Networks (GNN).
  • Traditional Models: Stochastic time-series models (ARIMA, SARIMA), Dynamic models (Global Circulation Model), Index-based approaches.
  • Emerging: Hybrid and Physics-Informed Neural Networks (PINN), Process-Guided Neural Networks (PGNNs).
• Data sources:
  • Multi-source Remote Sensing Data: Satellite data products (Landsat series, Sentinel series, MODIS, SMAP, METEOSAT, NOAA/AVHRR, SPOT, IKONOS, WorldView, ASTER, AMSR-E, SMOS, AWiFS, LISS-III, ETM+, ALOS-PALSAR, Hyperion EO-1, PRISMA, EnMAP, HyspIRI).
  • Remote Sensing-Derived Indices: Vegetation indices (NDVI, EVI, VCI, MSI, WDI, NDWI, NDMI, NMDI, VSDI, etc.), Temperature-based indices (LST, TCI, VTCI, TVDI, VHI), Soil moisture indices (SMDI, MPDI, MWSI, AMWSI, PDI), Combined indices (NDDI, VegDRI, agCDI, ASIS, IADI).
  • Meteorological Data: Precipitation, temperature, humidity, evapotranspiration.
  • Ground-based Data: In-situ observations, agricultural surveys.
  • Multi-parametric Datasets: Climate indices (e.g., El Niño-Southern Oscillation - ENSO).
  • Emerging Technologies: UAV-based high-resolution data, high-precision field-based sensors, Internet of Things (IoT).

Main Results

• Multi-source remote sensing data, combined with advanced AI (ML/DL) techniques, provide accurate, cost-effective, and transferable solutions for spatio-temporal agricultural drought monitoring and prediction.
• Widely used remote sensing spectral indices for agricultural drought characterization include NDVI, SMA, VCI, and VHI, while precipitation-based indices like SPI and SPEI are common for assessment and prediction.
• MODIS and Landsat datasets are frequently utilized due to their long-term continuity and accessibility.
• Multi-index integration consistently offers more robust drought mapping compared to single-index approaches.
• The performance of ML/DL models for drought prediction is highly dependent on input feature selection, lag design, validation strategy, and spatial transferability.
• Future research should prioritize high-resolution root-zone soil moisture estimation, develop advanced Physics-guided and hybrid ML approaches for improved generalization and reliability under climate non-stationarity, and integrate drought monitoring with impact assessment (e.g., crop yield-loss modeling and vulnerability/risk mapping).
• Advancements in this field directly contribute to achieving United Nations Sustainable Development Goals (SDGs), particularly SDG 2 (Zero Hunger), SDG 6 (Clean Water and Sanitation), SDG 13 (Climate Action), and SDG 15 (Life on Land).

Contributions

• Provides a comprehensive theoretical framework and summary of advanced geospatial and AI/ML/DL approaches for agricultural drought monitoring and modeling.
• Offers a critical examination of the strengths, limitations, and suitability of various remote sensing indices, sensors, and AI frameworks across different scales.
• Highlights the increasing importance and potential of big data analytics, ensemble, hybrid, and physics-based machine learning and deep learning approaches in drought research.
• Identifies key research gaps and future directions, emphasizing the need for high-resolution soil moisture data, advanced hybrid models, and integrated impact assessment.
• Serves as a valuable resource for decision-makers, policy planners, and research organizations involved in sustainable and climate-smart agriculture.

Funding

• The first author acknowledges the University Grants Commission, Government of India, for providing a fellowship to pursue doctoral research.
• No direct funds, grants, or other support were received for the preparation of this specific manuscript.

Citation

@article{Mondal2026Advancements,
  author = {Mondal, Suresh and Prasad, Kumar Arun and Achu, A. L. and Kaliraj, S. and Balasubramani, K.},
  title = {Advancements in Spatio-temporal agricultural drought monitoring and modeling: a comprehensive review on multi-source remote sensing and machine learning techniques},
  journal = {Theoretical and Applied Climatology},
  year = {2026},
  doi = {10.1007/s00704-026-06136-8},
  url = {https://doi.org/10.1007/s00704-026-06136-8}
}