Spatiotemporal Dynamics of Desertification in Rajasthan (1990–2020), India: A Machine Learning and Remote Sensing Approach Using Google Earth Engine

Desertification poses serious challenges to ecosystems and human livelihoods, requiring a comprehensive understanding of its dynamics. While previous studies in Rajasthan have explored desertification, most relied on coarse-resolution datasets, limited area and timeframe studies, or traditional classification methods. This study advances desertification analysis by integrating multi-index spectral modelling and machine learning (ML) classification in Google Earth Engine, enabling reproducible, multi-decadal (1990–2020) parallel processing of the Landsat archive for the entire state of Rajasthan. Seven cloud-free summer-season composite images were generated at five-year intervals using multi-year aggregation, from which six key spectral indicators - NDVI, MSAVI, TGSI, Albedo, NBSI, and NDSI were derived. These indicators were used as predictor variables in a Random Forest (RF) classifier to map five desertification levels: ex-severe, severe, moderate, light, and water. The RF model was trained and validated using a 70:30 split of labelled samples and benchmarked against XGBoost, CART, SVM, and Naive Bayes ML models, achieving 94.17% overall accuracy and a kappa coefficient of 0.917. To characterise desertification dynamics, the study applied gain-loss analysis, interchange analysis, change intensity assessment, and gravity-centre shift mapping across five classified desertification levels. The relationship between the spectral indicators and the meteorological factors (precipitation, wind speed, soil moisture, maximum temperature and evapotranspiration) was quantified using correlation analysis and residual trend analysis. Wind speed emerged as the dominant meteorological control, and annual precipitation explained 15–48% of the indicator variance, while statistically significant trends remaining in the residual series (NDVI τ = +0.510, Albedo τ = -0.553, both p < 0.001) indicate a substantial non-climatic contribution to the observed change. The results show that the share of ex-severe desertification in Rajasthan decreased from 41% to 31% from 1990 to 2020. Similarly, combined severe, moderate and light desertification levels increased from 59% to 69%, and water increased from around 0% to 1%, largely due to anthropogenic water infrastructure development over the years. This study calls for the development of effective land management strategies aimed at understanding and reducing desertification’s adverse impacts on ecosystems and human well-being.

Authors

Institutions

Publication Details

Journal
Journal of the Indian Society of Remote Sensing
Published
2026-11-18
DOI
https://doi.org/10.1007/s12524-026-02618-4
Primary Topic
Remote Sensing in Agriculture
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Spatiotemporal Dynamics of Desertification in Rajasthan (1990–2020), India: A Machine Learning and Remote Sensing Approach Using Google Earth Engine

Biswajeet Pradhan, Aali Pant, Mukesh Ray
3 links
Journal of the Indian Society of Remote Sensing
Remote Sensing in Agriculture
article

Spatiotemporal Dynamics of Desertification in Rajasthan (1990–2020), India: A Machine Learning and Remote Sensing Approach Using Google Earth Engine

Biswajeet Pradhan, Aali Pant, Mukesh Ray
article en

Abstract

Desertification poses serious challenges to ecosystems and human livelihoods, requiring a comprehensive understanding of its dynamics. While previous studies in Rajasthan have explored desertification, most relied on coarse-resolution datasets, limited area and timeframe studies, or traditional classification methods. This study advances desertification analysis by integrating multi-index spectral modelling and machine learning (ML) classification in Google Earth Engine, enabling reproducible, multi-decadal (1990–2020) parallel processing of the Landsat archive for the entire state of Rajasthan. Seven cloud-free summer-season composite images were generated at five-year intervals using multi-year aggregation, from which six key spectral indicators - NDVI, MSAVI, TGSI, Albedo, NBSI, and NDSI were derived. These indicators were used as predictor variables in a Random Forest (RF) classifier to map five desertification levels: ex-severe, severe, moderate, light, and water. The RF model was trained and validated using a 70:30 split of labelled samples and benchmarked against XGBoost, CART, SVM, and Naive Bayes ML models, achieving 94.17% overall accuracy and a kappa coefficient of 0.917. To characterise desertification dynamics, the study applied gain-loss analysis, interchange analysis, change intensity assessment, and gravity-centre shift mapping across five classified desertification levels. The relationship between the spectral indicators and the meteorological factors (precipitation, wind speed, soil moisture, maximum temperature and evapotranspiration) was quantified using correlation analysis and residual trend analysis. Wind speed emerged as the dominant meteorological control, and annual precipitation explained 15–48% of the indicator variance, while statistically significant trends remaining in the residual series (NDVI τ = +0.510, Albedo τ = -0.553, both p < 0.001) indicate a substantial non-climatic contribution to the observed change. The results show that the share of ex-severe desertification in Rajasthan decreased from 41% to 31% from 1990 to 2020. Similarly, combined severe, moderate and light desertification levels increased from 59% to 69%, and water increased from around 0% to 1%, largely due to anthropogenic water infrastructure development over the years. This study calls for the development of effective land management strategies aimed at understanding and reducing desertification’s adverse impacts on ecosystems and human well-being.

Journal of the Indian Society of Remote Sensing
University of Technology Sydney (AU), Indian Institute of Technology Jodhpur (IN)
National Aeronautics and Space Administration, University of Technology Sydney, U.S. Geological Survey
Climate action, Life on land
Remote Sensing in Agriculture
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon