Hu, Xueqian; Zhang, Shuo; Li, Li; Huang, Jianxi; Zhao, Zhenyu; Liu, Kangyi; Zhang, Zejia; Yao, Xiaochuang
Abstract
Northeast China (NE) is a vital grain production base, playing a crucial role in ensuring food security. Maize, rice, and soybean, the three major grain crops (MGC), account for 98.7 % of NE's total grain output. However, high-resolution MGC maps for this region are lacking, which are essential for analyzing spatial and temporal changes in crop planting patterns and developing agricultural policies. This study aims to create a 10-meter resolution map of MGC in NE (NE-MGCM) from 2019 to 2023 with a method based on sample generation and ensemble learning strategies, combined with optical and synthetic aperture radar (SAR) images. NE-MGCM achieves high mapping accuracy, with a five-year average overall accuracy (OA) of 94.96 % and an average Kappa coefficient of 0.9264. The producer accuracy (PA) for maize, rice, and soybean ranged from 91.12 % to 94.55 %, 94.89-96.36 %, and 94.53-95.69 %, respectively. Compared to four public datasets, NE-MGCM demonstrates superior overall performance and high consistency with statistical data. NE-MGCM effectively tracks the spatial and temporal distribution changes of MGC in NE, providing robust data support for food security and agricultural management.
Major grain crop mapping in Northeast China using sample generation method and ensemble learning.pdf
