Phenology-assisted supervised paddy rice mapping with the Landsat imagery on Google Earth Engine: Experiments in Heilongjiang Province of China from 1990 to 2020

Abstract

Accurate spatial distribution maps of paddy rice played crucial roles in food security and market stability. Decades-spanning Landsat images were useful for long-term paddy rice mapping. However, it still remained challenging to achieve consistent paddy rice mapping using the Landsat series images due to many factors such as sparse observations, frequent weather contamination, and shortage of training samples. To address these challenges, this study proposed a flexible Phenology-assisted Supervised Paddy Rice (PSPR) mapping framework on Google Earth Engine (GEE). This was achieved by utilizing the automation of the phenological methods, training data generation and purification, and the all-season classification capacity of the machine learning methods. We demonstrated the method by generating high-resolution 30-m paddy rice maps of Heilongjiang Province of China from 1990 to 2020. The derived rice maps were validated using abundant reference samples, four existing paddy rice products, and available agricultural statistics. The result showed that improved performance was verified in comparison to previous studies and a high linear relationship was observed with an average R2 of 0.993. Based on the spatiotemporal analysis, it was discovered that the rice planting in Heilongjiang has significantly shifted northward in the last three decades and this northward shift surprisingly appeared earlier than the previous studies, which was to our best knowledge first to be revealed in related studies. The multi-year dataset is useful for rice monitoring, water management, and policy making. All data and codes used in this study can be accessed on GitHub (https://github.com/MKGenesis/PSPR-rice-HLJ).