Evaluation of the Climate Forecast System Reanalysis weather data for the hydrological model in the Arctic watershed Malselv

The high-resolution Climate Forecast System Reanalysis (CFSR) data have recently become an alternative input for hydrological models in data-sparse regions. However, the quality of CFSR data for running hydrological models in the Arctic is not well studied yet. This paper aims to compare the quality of CFSR data with ground-based data for hydrological modeling in an Arctic watershed, Malselv. The QSWAT model, a coupling of the hydrological model SWAT (soil and water assessment tool) and the QGIS, was applied in this study. The model ran from 1995 to 2012 with a 3-year warm-up period (1995-1997). Calibration (1998-2007), validation (2008-2012), and uncertainty analyses were conducted by the model for each dataset at five hydro-gauging stations within the watershed. The objective function Nash-Sutcliffe coefficient of efficiency for calibration is 0.65-0.82 with CFSR data and 0.55-0.74 with ground-based data, which indicate higher performance of the high-resolution CFSR data than the existing scattered ground-based data. The CFSR weather grid points showed higher variation in precipitation than the ground-based weather stations across the whole watershed. The calculated average annual rainfall by CFSR data for the whole watershed is approximately 24% higher than that by ground-based data, which results in some higher water balance components. The CFSR data also demonstrate its high capacities to replicate the streamflow hydrograph, in terms of timing and magnitude of peak and low flow. Through examination of the uncertainty coefficients P-factors (>= 0.7) and R-factors (<= 1.5), this study concludes that CFSR data are a reliable source for running hydrological models in the Arctic watershed Malselv.

Automated summary

The study compares the quality of CFSR data with ground-based data for hydrological modeling in the Arctic watershed Malselv. The results indicate that the high-resolution CFSR data perform better, accurately simulating precipitation in the watershed and replicating the streamflow hydrograph.