Glacio-hydrological modeling informed by observations from a research basin in the Central Andes of Chile

Water resources availability in the Central Andes of Chile largely depends on glacier melt, which is undergoing accelerated changes as a consequence of global warming. Studying the hydrological response of glacierized systems requires understanding their interaction with catchment-scale processes, which in turn demands detailed observations that are rarely available in glacierized basins. Physically based hydrological modeling offers the opportunity of representing processes at large spatial extents through the informed selection and transference of observable parameters. Here, the Cold Regions Hydrological Model (CRHM) was implemented to evaluate the feasibility of transferring parameters from the intensively monitored Glaciar Echaurren Research Basin (app. 4 km²) to two larger glacierized basins: the Yeso River basin at Termas del Plomo (RYTP, app. 60 km²) and the Mapocho River basin at los Almendros (RMLA, app. 640 km²).

Simulations were performed using both locally calibrated parameters and parameters transferred from the experimental basin, and model performance was evaluated in terms of streamflow, fractional snow-covered area (fSCA), and snow water equivalent (SWE). Full transfer of the calibrated parameters from the small intensive study catchment to the larger RMLA basin resulted in reductions of up to 71% in streamflow KGE and 97% in SWE NSE compared to the basin’s own calibration. In the intermediate RYTP basin, the transfer of snow-related parameters adequately reproduced the seasonal pattern of SWE, although with a −59.2% bias and a 70.6% decrease in streamflow KGE relative to the calibrated version.

Individual parameter transfer revealed that snow-related parameters, such as snow roughness length and active layer thickness, explain a large fraction of the loss in SWE performance. On the other hand, the degradation in streamflow performance was dominated by parameters associated with surface storage processes. Overall, the results indicate that parameter transferability is only partially viable: while some parameters can be generalized across basins with similar characteristics, highly sensitive and locally dependent parameters require site-specific calibration to preserve model representativeness.