Accuracy assessment of approaches to spatially explicit reconstruction of historical cropland in Songnen Plain,Northeast China

To understand historical human-induced land use/cover change (LUCC) and its climatic effects, it is essential to reconstruct historical land use/cover changes with explicit spatial information. In this study, based on the historically documented cropland area at county level, we reconstructed the spatially explicit cropland distribution at a cell size of 1 km × 1 km for the Songnen Plain in the late Qing Dynasty (1908 AD). The reconstructions were carried out using two methods. One method (hereafter, referred to as method I) allocated the cropland to cells ordered from a high agricultural suitability index (ASI) to a low ASI, but they were all within the domain of potential cropland area. The potential cropland area was created by excluding natural woodland, swamp, water bodies, and mountains from the study area. The other method (hereafter, method II) allocated the cropland to cells in the order from high ASI to low ASI within the domain of cropland area in 1959. This method was based on the hypothesis that the cropland area domain in 1959 resulted from enlargement of the cropland area domain in 1908. We then compared these two reconstructions. We found that the cropland distributions reconstructed by the two methods exhibit a similar spatial distribution pattern. Both reconstructions show that the cropland was mostly found in the southern and eastern parts of the Songnen Plain. The two reconstructions matched each other for about 68% of the total cropland area. By spatially comparing the unmatched cropland cells of the two reconstructions with the settlements for each county, we found that unmatched cropland cells from method I are closer to settlements than those from method II. This finding suggests that reconstruction using method I may have less bias than reconstruction with method II.