Large-scale and high-resolution paddy rice intensity mapping using downscaling and phenology-based algorithms on Google Earth Engine

Abstract

Listen

Translate article

Accurate mapping of paddy rice cropping intensity (PRCI) affects precision agriculture, water use management, and informed decision-making. Many PRCI mapping approaches have been developed and achieved remarkable performance. However, large-scale and high-resolution PRCI mapping in southern China, especially in the Hunan Province, remains challenging. In this region, optical data suffer from serious data loss owing to frequent cloud coverage, whereas radar data are usually affected by mountainous terrain. Multisource data fusion may mitigate data loss; however, many data fusion approaches cannot be used in large-scale applications because of their significant computational complexity. To achieve accurate PRCI mapping in large-scale areas of southern China, we propose a novel large-scale downscaling method for multi-source data fusion. A new framework for PRCI mapping was developed using the Google Earth Engine. To test its performance, we mapped the 10 m PRCI of Hunan Province in 2020. The overall accuracy of the mapping was 90.70 % with a kappa coefficient of 0.81. Furthermore, our estimated sown area was highly correlated with the statistical yearbook (R2 = 0.84, RMSE = 11.47 kha). Our experimental results demonstrated the remarkable performance of the proposed framework for large-scale and high-resolution PRCI mapping in southern China.