Machine Learning‐Based Drought Stress Mapping Using Landsat and Sentinel‐2 Imagery: A Remote Sensing Approach

ABSTRACT Drought poses a significant threat to agricultural productivity, particularly in rice‐growing regions. The use of multi‐source remote sensing satellites to monitor drought stress during the peak rice growing season remains largely unexplored. Therefore, this study employed integrating approaches such as Landsat‐8 and Sentinel‐2 imagery to monitor drought stress during the peak rice growing season (July–September) from 2021 to 2024 in Hunan province. Results of Landsat‐8 satellite indicate the Vegetation Health Index (VHI) identified extreme drought conditions (37.9%) in August during critical rice growth stages in 2022. The Normalized Difference Vegetation Index (NDVI) values decreased in September 2022. Similarly, the results of Actual Evapotranspiration (ETa) analysis showed elevated values in August 2022. Together, these indicators reveal a coupled response of vegetation stress and increased water demand during the peak rice growing period under severe drought conditions. Moreover, NDVI anomaly mapping demonstrated pronounced negative deviations (−40% to −20%) in July and August 2022, with vegetation stress persisting into September, confirming the lagged response of vegetation to drought conditions. While land use land cover (LULC) analysis revealed cropland area highly decreased from 2022 to 2024 by 8.68%, built‐up areas expanded continuously from 2021 to 2024, increasing by 13.1%. Tree covered area remained relatively stable; thus, a negligible change was noted by 0.87% during the study period. In conclusion, the integrated multi‐indicator approach effectively determined the spatiotemporal drought impacts on total cropland and effects on the peak rice growing season in 2022 during the entire studied period. These findings provide valuable understanding for developing drought early warning systems and informing water management strategies in rice production systems vulnerable to climate variability.