Plant disease mapping in paddy growing stages using remotely sensed data

This study presents an advanced phenology-based approach to map rice fields and their growth stages using Sentinel-1 and Sentinel-2 time-series data combined with k-means clustering and fuzzy deep neural networks (FDNN). The method relies on two main observations: the detection of flooding during transplanting through Sentinel-1 VH backscatter analysis and the identification of growth phase fluctuations using normalized difference vegetation index (NDVI) time series. This approach was applied to Gilan, Iran (14,042 km²), and Niigata, Japan (12,584 km²). Data from January 2019 to December 2020 were processed on the Google Earth engine (GEE) platform, resulting in high-resolution maps at a 10-meter scale. The methodology integrated unsupervised classification and FDNN to effectively identify areas with similar phenological characteristics and detect plant diseases. The resulting maps detailed the extent, intensity, and schedules of rice cultivation, showcasing the seasonal rhythms of planting and harvesting. Validation against high-resolution Google Earth street viewpoints confirmed the accuracy and reliability of the generated maps, demonstrating the robustness and efficiency of the proposed approach. The findings underscore the cost-effectiveness of this method, which accurately delineates rice fields, identifies growth stages, and detects plant diseases over large areas. This capability is crucial for monitoring progress towards self-sufficiency in rice production and for calculating methane emissions from rice paddies. By providing detailed insights into the spatial and temporal dynamics of rice farming, the study offers a valuable tool for agricultural planning and environmental management. The approach’s success in diverse geographic regions highlights its potential for broader application in other rice-producing areas worldwide.