전국 유역 대상 InVEST Annual Water Yield 모형의 경험적 검증 및 매개변수 평가

초록

This study evaluated the applicability and performance of the InVEST Annual Water Yield (InVEST AWY) model by comparing its outputs with observed annual total discharge across watersheds in South Korea. We delineated DEM-based watersheds for 201 streamflow gauging stations nationwide and selected 75 watersheds for validation, ensuring their delineated areas differed by no more than 20% from the reported gauging station drainage areas. We conducted a sensitivity analysis on the seasonality constant Z and compared four precipitation datasets that included global gridded products and domestic station-based interpolated data. We also conducted stratified 5-fold cross-validation to assess the stability of parameter selection. The results showed that for the watersheds in this study, a seasonality constant of Z = 1 yielded the highest explanatory power and the lowest prediction error. Cross-validation consistently selected this value, confirming the robustness of parameter selection. Among the precipitation datasets, the domestic station-based interpolated data outperformed the global data, with the spline-interpolated dataset achieving the highest R2 of 0.829 and being selected as the final input. Using this final configuration, we estimated the total annual water yield in South Korea at approximately 59.3 billion m³ yr-1, with forests contributing the largest share at 55.4% of the estimated national water yield. These findings suggest that the InVEST AWY model can provide reliable estimates of water supply under Korean conditions when properly parameterized. This study offers a model configuration that can be applied to future assessments of water supply ecosystem services in South Korea.

키워드