Multi-model driven by diverse precipitation datasets increases confidence in identifying dominant factors for runoff change in a subbasin of the Qaidam Basin of China

Abstract

Quantifying the climatic and anthropogenic effects on hydrological processes has received considerable attention. However, diverse conclusions could be drawn when different models and forcing datasets are used. This is particularly uncertain and challenging in poorly gauged arid regions. Here we aim to tackle this issue in the poorly gauged Xiangride River Basin within the Qaidam Basin, one of the three prominent inland basins in China. We applied two distinct models (Budyko Mezentsev-Choudhurdy-Yang and process-based SWAT) to a poorly-gauged inland basin in West China. The model simulations were driven by four precipitation products including Tropical Rainfall Measuring Mission (TRMM) 3B42 V7, Global Precipitation Measurement (GPM) IMERG V6, Multi-Source Weighted-Ensemble Precipitation (MSWEP) and China Meteorological Assimilation Driving Datasets (CMADS). Our results indicate that MSWEP performed best (NSE = 0.64 vs. 0.36–0.59 for other datasets) in the baseline period (2009–2012), whereas CMADS was more accurate during the impacted period (2013–2016); CMADS and GPM might underestimate the precipitation in the baseline and impacted period, respectively. Hydrological processes during the impacted period are presumed to be influenced by climate variation and/or human activities, compared to the relatively natural status in the baseline period. We conclude that runoff decline between the two periods was mainly affected by human activities (−66 to 94%), whereas the contribution of climate variation was more likely positive. A literature survey reveals that major anthropogenic effects in the study area includes reservoir, road construction and cropland expansion that could lead to runoff decrease. We recommend the use of process-based model (e.g., SWAT) in studies like this, as process-based models driven by high-quality remote-sensed or reanalysis climate datasets, better represents the spatiotemporal hydrological change under altered conditions, whereas the steady-state assumption of soil water for the Budyko model may not be fully satisfied during a short period.