Runoff change controlled by combined effects of multiple environmental factors in a headwater catchment with cold and arid climate in northwest China.

The limited runoff in cold and arid regions is sensitive to environmental changes, and it is thus urgent to explore the change and controlling factors of runoff under the background of global warming and intensified human activities. However, previous studies have rarely considered the combined effects of multiple controlling factors at varying scales over time. With the headwater region of the Manas River in northwest China as the study area, we investigated the change in runoff for the period of 1954-2016 and its relationship with regional environmental factors (e.g. precipitation PCP, temperature TMP, potential evapotranspiration ET, snow cover extent SCE, land use, and normalized difference vegetation index NDVI) and/or global atmospheric circulation (e.g. North Atlantic Oscillation NAO, Arctic Oscillation AO, Pacific Interdecadal Oscillation PDO, and El Nino Southern Oscillation ENSO). In particular, the combined effects of multiple environmental factors were determined at different scales by the multiple wavelet coherence. The annual runoff significantly increased at a rate of 0.508 × 10 m/decade, and the climate tended to be warmer and wetter. Among the regional and global environmental factors, NDVI and ENSO were the single factor mostly correlated with runoff, while NDVI-TMP and ENSO-PDO were the combined factors with the stronger relations on runoff, respectively. The regional environmental factors had larger impacts on runoff than the global environmental factors, and the natural factors outperformed human activities in controlling runoff. The accelerated melting of snow/glacier induced by the increasing temperature dominated runoff change, and the increasing water inputs from wetter climate may play a second role in runoff. The runoff characteristics in cold and arid regions seem to be different from those regions with little snow/glacier, which should be paid more attention. The employed multiple wavelet coherence is helpful in determining the processes dominating runoff change.