A Model Based Attribution of Streamflow Trends in Headwater Catchments of the Tarim River, Central Asia

Abstract:

Over the last decades, increasing streamflow has been observed in glacierized headwater catchments of the Tarim River, Central Asia. We investigated to what extent these streamflow increases can be attributed to increasing precipitation, or increasing air temperature and resulting glacier mass loss, using a data-driven approach based on multiple linear regression, and a simulation-driven approach with a hydrological model.

The hydrological model was calibrated in a multi-objective way using criteria for discharge and glacier mass balances. This is crucial in such catchments with high uncertainties in precipitation, as a deficient streamflow simulation due to inadequate precipitation input may be compensated by erroneous simulation of glacier melt, if both variables are not jointly considered for model calibration. Temporal variations in glacier changes were calibrated against an observed glacier mass balance time series, whilst geodetic glacier mass balance estimates were used for constraining the absolute mass balance loss. The model considers the effects of changes in glacier area and in the elevation of the glacier surface over the course of the 48-years study period.

Simulations with detrended time series indicate that changes in the less glacierized Kokshaal catchment are caused by increases in precipitation and glacier melt to a comparable extent. In contrast, for the more highly glacierized Sary-Djaz catchment, increasing glacier melt plays a dominant role. Results from the data-based approach were generally comparable, but pointed to a higher influence of temperature also in the Kokshaal catchment.

Overall, the hydrological modeling approach was considered as advantageous because it allows to (1) directly investigate cause-effect relations, (2) analyze the impact of changes in variables that are not measured (e.g., changes in the fraction of glacier melt), and (3) integrate different kinds of data (like glacier mass balances or snow cover data). In this way, (in-)consistencies between different hypotheses for the causes of changes, the model results and the various data can be revealed.