New spatial distribution dataset of cropland in China from 1850 to 2020

Accurately reconstructing the historical spatial distribution of cropland is crucial for quantifying the environmental effects of agricultural land use change and understanding the interactions within human−environment systems. However, existing statistical inference methods, while covering long historical periods, often yield insufficient precision in the spatial distribution of reconstructed cropland. Remote sensing inversion methods offer high spatial resolution but are limited in temporal coverage and are prone to systematic biases in area estimation. To address these issues, this study proposed a novel method that integrates the advantages of both approaches, reconstructing a yearly cropland spatial distribution dataset for China from 1850 to 2020 at a 5 km resolution. First, multi-source data, including historical archives, statistical yearbooks, and national land surveys, were integrated. Through phased identification, correction, and interpolation, a temporally continuous and internally consistent provincial-level cropland area time series for China was constructed. Second, a dynamically weighted joint probability index was established, coupling spatial distribution probabilities derived from remote sensing inversion with potential distribution probabilities based on physical suitability. This approach effectively bridged the cropland patterns before and after remote sensing data became available, enabling the spatial allocation of cropland.