Efficient Evaluation Method for Paddy Field Quality at Plot Scale Based on Multi‐Source Data and Random Forest Model

ABSTRACT China is rapidly constructing well‐facilitated farmland projects, with paddy fields as a key component, making efficient monitoring of paddy field cultivated land quality a core issue in agricultural resource management. However, the existing cultivated land quality evaluation methods still have some problems, such as complex evaluation index, low data acquisition efficiency and poor timeliness. Therefore, in this study, the paddy field of Jiaxing City, which is a good representative of well‐facilitated farmland construction, was selected as the research area, and a simplified paddy field quality index system including seven core indicators was screened through literature review analysis, which could be quickly calculated by combining remote sensing and associated data. The Random Forest (RF) algorithm was used to construct a multi‐index fusion evaluation model to achieve rapid and accurate evaluation of paddy field quality at plot scale, which solved the subjective problems of traditional methods. The results showed that: (1) The accuracy of the paddy field evaluation model based on random forest reached 89.5%, showing a high level of consistency with the quality grading of the paddy fields; (2) Indicators exhibited low redundancy (max correlation 0.36 between normalized difference vegetation index (NDVI) and net primary productivity (NPP)); light‐temperature production potential (LTPP) contributed most (0.47), followed by NPP (0.19) and distance to fields (0.10), key indicators for differences in paddy field production potential and quality. (3) The quality grade of paddy fields in Jiaxing showed obvious aggregation characteristics, and 55.83% of paddy fields were rated as high grade. This study validates the effectiveness of the simplified indicator system and multi‐source data fusion, providing an efficient method for monitoring the construction effect of well‐facilitated farmland projects.