The impact of small-scale surface representation in WRF on hydrological modeling in a glaciated catchment

Abstract. High-elevation alpine catchments are particularly affected by the global rise in temperature. Understanding the drivers of climate-induced changes in the hydrological response of these catchments in the past is relevant for developing future adaptation strategies for water resources and risk management. However, the study of long-term changes since the last Little Ice Age (around 1850) is strongly limited by the availability of hydrometeorological observation data. Regional climate models (RCMs) can bridge this limitation and provide comprehensive meteorological forcing data for hydrological models (HM). We used the Weather Research Forecasting Model (WRF) to dynamically downscale a global reanalysis product (20CRv3) to a 2 km x 2 km spatial and 1 h temporal resolution from 1850 to 2015 as forcing for an HM (WaSiM). The main challenge is transferring the forcing data to the much finer grid resolution (i.e., 25 m) of the HM, considering the complex topography and plausible sub-daily precipitation and temperature lapse rates (TLRs). Thus, we developed a workflow for extracting and transferring hourly TLRs from the WRF atmosphere to the small-scale topography of the HM domain. In addition, we corrected WRF precipitation frequencies with observation data and re-distributed the precipitation according to the small-scale topography. Our study demonstrates the impact of TLRs computed from different WRF layers (i.e., 2 m and free atmosphere) on the HM results of a highly glaciated Alpine catchment in the European Alps. In a multi-data evaluation procedure, we found that the TLRs and the HM results are significantly dependent on the coarse surface properties of WRF. Temperature-sensitive processes such as snow and glacier evolution, as well as the streamflow response, are more realistically simulated when the HM is forced by TLRs originating from the WRF free atmosphere rather than with simulated near-surface temperature. The HM results are also consistent with observation data over a simulation period beginning in 1969, suggesting the corrected WRF temperature can reliably reproduce the non-stationarity in local temperature observations. Our study addresses several aspects, limitations, and potential solutions in applying a standard modeling chain of an RCM and a physics-based HM for climate sensitivity studies in high-elevation alpine regions.