Global Evaluation of Streamflow from Ten State-of-the-Art Hydrologic Models

Abstract:

Observed streamflow (Q) data from 1286 medium sized catchments (1000–5000 km²) around the globe were used to comprehensively evaluate the Q estimates (1979–2012) of six global hydrologic models (GHMs) and four land surface models (LSMs) run within the EartH2Observe project. The models were all driven by the WATCH Forcing Data methodology applied to ERA-Interim reanalysis (WFDEI) meteorological dataset, but used various datasets for non-meteorologic variables and were run at various spatial and temporal resolutions, although all data were re-sampled to a common 0.5° spatial and daily temporal resolution. For the evaluation we used a broad range of performance metrics related to various important aspects of the hydrograph. The GHMs were found to perform slightly better overall than the LSMs, and the calibrated models were found to generally outperform the uncalibrated models for the metrics incorporated in the objective functions used for calibration. We further found that the ensemble mean had only slightly worse performance than the best (calibrated) model, and that despite adjustments using in-situ observations and corrections for orography, the WFDEI precipitation data still contain substantial biases which are reflected in the simulated Q. We argue that model parameter regionalization deserves more attention, given that the large majority of the land surface is ungauged or poorly gauged. We suggest that a parameter ensemble for a single sufficiently flexible model is easier to realize yet provides the same benefits as a multi-model ensemble. The results of this study have important implications for studies assessing the hydrologic impacts of climate and land-use change.