A meta-analysis based review of quantifying the contributions of runoff components to streamflow in glacierized basins

Abstract

Quantifying the contributions of runoff components (CRCs) to streamflow is of significant importance for understanding the dynamics of water resources under changing climate in glacierized basins. This article presents a meta-analysis on different approaches for quantifying runoff components in glacierized basins, including the tracer-based end-member mixing method and the hydrological modeling approach. We collected estimated CRCs from 312 glacierized basin cases, as well as values of five characteristics in these basins including mean basin elevation (MBE), mean annual air temperature (MAT), mean annual precipitation (MAP), winter precipitation fraction (WPF) and glacierized area ratio (GAR). Relations between CRCs and the basin characteristics were assessed using a random forest (RF) algorithm. The review showed that CRCs were most often quantified by the hydrological modeling approach (73% of the basin cases). Compared to hydrological modeling, the tracer-based approach (applied in 19% of the basin cases) was more likely to be used in smaller basins < 50 km2, rather at the seasonal than annual time scale and within shorter study periods of <=5 years. Meta-analysis results indicate that: (1) At the annual time scale, the most important influencing basin characteristics were GAR and MBE for the ice melt contribution, WPF and MAP for the snow melt contribution, and GAR and WPF for the rainfall contribution. RF algorithm based on the five basin characteristics was able to explain 56%, 40%, and 40% of the variability of the reported annual contributions of ice melt, snowmelt and rainfall, respectively; the variability of seasonal CRCs and annual contribution of groundwater could be less well explained by the five basin characteristics. (2) Comparing different definitions of runoff components based on water-input or flow-pathway indicated that the ice melt contribution to total water input (sum of rainfall and melt water) based on the water-input definition was close to the contribution of ice melt-induced surface flow to total runoff based on the flow-pathway definition. In contrast, based on the reviewed studies, rainfall and snowmelt contributions based on the water-input definition were around 9%-14% higher than the contributions of rainfall and snowmelt induced surface flow to total runoff. (3) The tracer-based end-member mixing method tended to estimate larger uncertainties of CRCs than hydrological modeling, but uncertainties of modeled CRCs were likely underestimated as often only one or two of the three uncertainty sources of model parameter, model input and model structure were considered in the modeling studies. We propose that more efforts are required to cross validate CRCs estimated by the tracer-based and hydrological modeling methods, and to reduce uncertainties of CRCs by integrations of hydro-meteorological data and water tracer data.