Determining the hydrological responses to climate variability and land use/cover change in the Loess Plateau with the Budyko framework

Understanding and quantifying the impacts of land use/cover change and climate variability on hydrological responses are important to the design of water resources and land use management strategies for adaptation to climate change, especially in water-limited areas. The elasticity method was used to detect the responses of streamflow and runoff coefficient to various driving factors in 15 main catchments of the Loess Plateau, China between 1961 and 2009. The elasticity of streamflow (Q) and runoff coefficient (Rc) to precipitation (P), potential evapotranspiration (E0), and catchment characteristics (represented by the parameter m in Fu's equation) were derived based on the Budyko hypothesis. There were two critical values of m=2 and E0/P=1 for the elasticity of Q and Rc. The hydrological responses were mainly affected by catchment characteristics in water-limited regions (E0/P>1), and in humid areas (E0/P<1), climate conditions played a more important role for cases of m>2 whereas catchment characteristics had a greater impact for cases of m<2. The annual Q and Rc in 14 of the 15 catchments significantly decreased with average reduction of 0.87mmyr−1 and 0.18%yr−1, respectively. The mean elasticities of Q to P, E0 and m were 2.66, −1.66 and −3.17, respectively. The contributions of land use/cover change and P reduction to decreased Q were 64.75% and 41.55%, respectively, while those to decreased Rc were 75.68% and 32.06%, respectively. In contrast, the decreased E0 resulted in 6.30% and 7.73% increase of Q and Rc, respectively. The contribution of land use/cover changes was significantly and positively correlated with the increase in the percentage of the soil and water conservation measures area (p<0.05). The Rc significantly and linearly decreased with the vegetation coverage (p<0.01). Moreover, the Rc linearly decreased with the percentage of measures area in all catchments (eight of them were statistically significant with p<0.05).