Satellite-based Rabi rice paddy field mapping in India: a case study on Telangana state

2025-07-07

AI Summary

Fragmented smallholder farmland impedes accurate remote sensing–based rice area mapping, undermining food security and agricultural policy formulation. Method: This study develops a phenology-driven, regionally tailored remote sensing classification framework for the winter rice season in Telangana, India. It introduces an ecologically informed phenological calibration strategy that integrates multi-temporal Sentinel-2 and Landsat imagery with field-scale phenological metrics to accommodate local spatiotemporal heterogeneity. Contribution/Results: The method significantly enhances small-field detection capability, producing a high-resolution rice area map covering 732,000 ha. It achieves an overall accuracy of 93.3%—an 8.0-percentage-point improvement over conventional approaches—and exhibits strong agreement with official statistics (R² = 0.981). This advances a fine-grained, interpretable remote sensing monitoring paradigm for complex agricultural landscapes.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics

Abstract

Accurate rice area monitoring is critical for food security and agricultural policy in smallholder farming regions, yet conventional remote sensing approaches struggle with the spatiotemporal heterogeneity characteristic of fragmented agricultural landscapes. This study developed a phenology-driven classification framework that systematically adapts to local agro-ecological variations across 32 districts in Telangana, India during the 2018-19 Rabi rice season. The research reveals significant spatiotemporal diversity, with phenological timing varying by up to 50 days between districts and field sizes ranging from 0.01 to 2.94 hectares. Our district-specific calibration approach achieved 93.3% overall accuracy, an 8.0 percentage point improvement over conventional regional clustering methods, with strong validation against official government statistics (R^2 = 0.981) demonstrating excellent agreement between remotely sensed and ground truth data. The framework successfully mapped 732,345 hectares by adapting to agro-climatic variations, with Northern districts requiring extended land preparation phases (up to 55 days) while Southern districts showed compressed cultivation cycles. Field size analysis revealed accuracy declining 6.8 percentage points from medium to tiny fields, providing insights for operational monitoring in fragmented landscapes. These findings demonstrate that remote sensing frameworks must embrace rather than simplify landscape complexity, advancing region-specific agricultural monitoring approaches that maintain scientific rigor while serving practical policy and food security applications.

Problem

• Mapping Rabi rice fields in fragmented landscapes accurately
• Adapting to local agro-ecological variations in Telangana
• Improving accuracy over conventional remote sensing methods

Innovation

• Phenology-driven classification adapts locally
• District-specific calibration boosts accuracy
• Adapts to agro-climatic field variations