Variations of baseflow across multi-temporal scales in China’s Loess Plateau: Impacts of catchment attributes and climate change

Baseflow is a key component of river discharge for maintaining sustainable water resources globally. However, the differences and attributions of baseflow across multi-temporal scales, particularly under extreme conditions (floods/droughts), remain inadequately understood. This study analyzed baseflow variations across five temporal scales (annual, rainy season, non-rainy season, flood period, and drought period) during 1956-2022 in fifteen catchments within the Loess Plateau. A machine learning-based stepwise regression model was developed with catchment attributes (normalized difference vegetation index [NDVI], index of hydrological connectivity [IC], soil water content [SWC]) and climatic factors (precipitation [P], temperature [T], potential evapotranspiration [PET], aridity index [AI]), to quantify their contributions to baseflow variations. The multi-year average baseflow index (BFI, ratio of baseflow to runoff) was 0.57, with 0.50 and 0.72 during the rainy and non-rainy seasons, respectively. Notably, BFI reached its minimum (0.25) during flood period and maximum (0.80) during drought period. At the five temporal scales, both runoff and baseflow decreased significantly ( p < 0.05), while the BFI increased significantly ( p < 0.05) in most catchments. The IC, P, and PET were the primary influencing factors for annual BFI variations (total relative contribution: 0.65, hereafter). Moisture conditions (SWC and P) controlled BFI variations during the rainy season (0.32) and flood period (0.50), whereas catchment attributes (IC and NDVI) played a dominant role during the non-rainy season (0.49) and drought period (0.41). The above findings enhance the understanding of hydrological processes and contribute to development of water resources utilization strategies and restoration of ecological restoration in semi-arid catchments.