Remote sensing inversion and prediction of land use land cover in the middle reaches of the Yangtze River basin, China

Land use and land cover (LULC) changes are dynamic and have been extensively studied; the change in LULC has become a crucial factor in decision making for planners and conservationists owing to its impact on natural ecosystems. Deriving accurate LULC data and analyzing their changes are important for assessing the energy balance, carbon balance, and hydrological cycle in a region. Therefore, we investigated the best classification method from the four methods and analyzed the change in LULC in the middle Yangtze River basin (MYRB) from 2001 to 2020 using the Google Earth Engine (GEE). The results suggest that (1) GEE platform enables to rapidly acquire and process remote sensing images for deriving LULC, and the random forest (RF) algorithm was able to calculate the highest overall accuracy and kappa coefficient (KC) of 87.7% and 0.84, respectively; (2) forestland occupied the largest area from 2001 to 2020, followed by water bodies and buildings. During the study period, there was a significant change in area occupied by both water bodies (overall increase of 46.2%) and buildings (decrease of 14.3% from 2001 to 2005); and (3) the simulation of LULC in the MYRB area was based on the primary drivers in the area, of which elevation changes had the largest effect on LULC changes. The patch generated land use simulation model (PLUS) was used to produce the simulation, with an overall accuracy and KC of 89.6% and 0.82, respectively. This study not only was useful for understanding the spatial and temporal characteristics of LULC in the MYRB, but also offered the basis for the simulation of ecological quality in this region.