Hyperspectral-Informed Sentinel-2-Based Monitoring of Paddy Residue Burning through Crop-State Discrimination

Abstract

Abstract Accurate mapping of agricultural residue burning using satellite data remains challenging due to the rapid temporal overlap and spectral similarity of mature crops, harvested fields, and burnt residues during peak harvest periods. This study presents a scalable, decision-rule–based methodology for the concurrent mapping of mature, harvested, and burnt paddy fields, integrating field-scale hyperspectral measurements with multi-temporal Sentinel-2 multispectral imagery. Hyperspectral observations captured systematic changes in crop reflectance associated with maturity, harvest intensity, and post-burn ash deposition, which were subsequently upscaled to Sentinel-2 spectral bands to evaluate a comprehensive set of vegetation and burn-sensitive indices. The analysis identified the Chlorophyll Absorption Ratio Index (CARI) as the most effective indicator for separating mature from harvested rice, while the delta Normalized Burn Ratio (dNBR) exhibited the highest sensitivity for distinguishing harvested fields from burnt residues. These indices were combined within a hierarchical decision-tree framework and applied to multi-date Sentinel-2 imagery to map rice burning dynamics across intensively cultivated districts in northern India. The approach achieved an overall classification accuracy of 92.57% with a kappa coefficient of 0.80, demonstrating strong spatial and temporal consistency with field observations. By explicitly addressing intra-seasonal spectral confusion in agricultural landscapes, the proposed framework advances burned-area mapping beyond single-index detection toward integrated crop-state discrimination. The methodology is computationally efficient, sensor-transferable, and suitable for operational implementation, offering significant potential for large-scale agricultural monitoring, emission assessment, and policy-driven residue management in rice-based cropping systems globally.

Plant phenotyping relevance

圃場規模のハイパースペクトル／Sentinel-2データからイネの成熟・収穫・焼却状態を識別する手法を開発・検証しており、植物の作物状態を抽出する方法が研究の中心である。

Code and data availability

The preprint describes hyperspectral field measurements, ground-truth GPS/KML data, and R ('hsdar')/ENVI analysis, but contains no data or code availability statement, no public deposit, and no author-provided URL for any paper-specific dataset, imagery, or scripts. The only URL present (Sentinel-2 MSI user guide) is a

No evidence-backed public reproduction asset is currently recorded.