Optimizing Rice Field Mapping in the Northern Region of China: An Asynchronous Flooding Signal and Object-Based Method

Abstract

Accurate delineation of paddy fields holds importance in ensuring food security, efficient water resource management, and precise evaluation of greenhouse gas emissions. Here we propose an innovative approach, the asynchronous flooding and object-based (AF-OB) model, aimed at optimizing phenology-based paddy field mapping. The AF-OB model capitalizes on the asynchronous flooding phenomenon observed between paddy fields and nonpaddy fields, along with the seasonal variations in the normalized difference vegetation index. The simple noniterative clustering algorithm is integrated to mitigate the common issue of the “pretzel effect” encountered in paddy field mapping. Evaluation through independent samples yields compelling results, with the paddy field map generated by the AF-OB method achieving an overall accuracy of 94.28%. The paddy fields extracted using the AF-OB method exhibit alignment with statistical data, surpassing comparable algorithms relying on alternative land use products in terms of visual quality. Furthermore, the AF-OB model exhibits stability across time, space, and sensors, thus enhancing its applicability and robustness. The outputs of the AF-OB method offer reference data for informed agricultural production planning and the effective management of water resources.

Keywords

• Google earth engine
• paddy field mapping
• phenology
• simple noniterative clustering (SNIC) and object-based