Spatiotemporal variation of actual evapotranspiration in a transboundary river basin based on MGWR downscaling

Study region

Tumen River Basin.

Study focus

Achieving high spatial resolution of actual evapotranspiration (ETa) is crucial for effective management of regional soil and water resources. However, current satellite technology limits the simultaneous acquisition of high temporal and spatial resolution data. Thus, downscaling ETa from lower to higher resolution is a practical solution. In this paper, the multi-scale geographically weighted regression (MGWR) model is used to downscale ETa data in the Tumen River Basin (TRB) from 1 km to 30 m resolution, using normalized difference vegetation index (NDVI), land surface temperature (LST), and digital elevation model (DEM) as independent variables. Furthermore, this study analyzed the spatiotemporal evolution of ETa in the TRB from 1990 to 2020 using high-resolution ETa data obtained from downscaling across seven periods.

New hydrological insights for the region

The results show the MGWR model's effectiveness, achieving an R² of 0.912, and accurately capturing the spatially heterogeneous relationships between ETa and the independent variables. The downscaled ETa data enhances spatial detail while preserving the original low-resolution patterns. From 1990–2020, ETa in the TRB showed a generally non-significant decreasing trend, with the basin-wide average decreasing from 102.37 mm to 101.06 mm. On the Chinese side, it declined from 107.12 mm to 105.93 mm, and on the side of the Democratic People's Republic of Korea (DPRK), from 91.78 mm to 90.22 mm.