Crop Classification in a Semi-Arid Region Using Multi-Temporal Sentinel-2 Data and GIS: A Comprehensive Case Study of Beed District, Maharashtra, India

Vishal Shirsat, Sanjay Tupe, Balaji Yadav, Sandipan Sawant and Shafiyoddin Sayyad

International Journal of Scientific Research in Science and Technology, 2026, vol. 13, issue 2, 493-506

Abstract: Agriculture in semi-arid regions like Beed district in Maharashtra, India, is the backbone of the local economy but is perpetually vulnerable to climatic stressors, including erratic monsoons, frequent droughts, and rising temperatures. Accurate, timely, and spatially explicit crop inventory data is critical for informed decision-making regarding water resource management, drought relief, insurance assessment, and sustainable agricultural planning. Traditional survey methods are inadequate for capturing the dynamic and fragmented cropping patterns of this region. This study addresses this gap by developing and evaluating a robust, scalable, and automated framework for multi-crop classification and acreage estimation using open-source tools and data. We leveraged the multi-spectral and high-temporal capabilities of Sentinel-2 satellite imagery within the cloud-computing environment of Google Earth Engine (GEE). A dense time-series from 2020 to 2023 was processed to generate seasonal median composites for the Kharif (monsoon) and Rabi (winter/post-monsoon) seasons. Key spectral indices—Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Normalized Difference Moisture Index (NDMI)—were computed to encapsulate crop phenology and water stress. A high-confidence reference dataset comprising samples for eight classes—Sorghum (Jowar), Pearl Millet (Bajra), Pigeon Pea (Tur), Soybean, Cotton, Fallow/Land, Forest/Scrub, and Water Bodies—was created through extensive field campaigns and visual interpretation. Three state-of-the-art machine learning classifiers—Support Vector Machine (SVM), Random Forest (RF), and Classification and Regression Trees (CART)—were trained and evaluated. The SVM algorithm achieved the highest performance with an Overall Accuracy (OA) of 96.7% and a Kappa coefficient (κ) of 0.96, significantly outperforming RF (OA=94.5%, κ=0.93) and CART (OA=91.2%, κ=0.89). Detailed spatio-temporal analysis of the generated maps reveals significant inter-annual variability in crop acreage, strongly correlated with pre-monsoon rainfall and reservoir levels. Sorghum and Pearl Millet dominated the cropped area, demonstrating the region's adaptation to water scarcity. The study conclusively demonstrates that integrating multi-temporal Sentinel-2 data, phenological indices, and advanced ML models on the GEE platform provides a powerful, cost-effective solution for operational crop monitoring in semi-arid, smallholder-dominated agricultural systems, with direct applications in enhancing climate resilience and food security.

Keywords: Crop classification; Sentinel-2; Semi-arid agriculture; Support Vector Machine (SVM); Precision agriculture; Time-series analysis; Google Earth Engine; Random Forest (search for similar items in EconPapers)
Date: 2026
References: Add references at CitEc
Citations:

Downloads: (external link)
https://ijsrst.com/home/article/view/IJSRST2613319 Abstract page (text/html)
https://ijsrst.com/home/article/download/IJSRST2613319/IJSRST2613319 Full text (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:etm:ijsrst:v13:y2026:i2:id:1477

DOI: 10.32628/IJSRST2613319

Access Statistics for this article

More articles in International Journal of Scientific Research in Science and Technology from Technoscience Academy
Bibliographic data for series maintained by Pankaj Sharma ().