Calibration of the Global Flood Awareness System (GloFAS) using daily streamflow data.

This paper presents the calibration and evaluation of the Global Flood Awareness System (GloFAS), an operational system that produces ensemble streamflow forecasts and threshold exceedance probabilities for large rivers worldwide. The system generates daily streamflow forecasts using a coupled H-TESSEL land surface scheme and the LISFLOOD model forced by ECMWF IFS meteorological forecasts. The hydrology model currently uses a priori parameter estimates with uniform values globally, which may limit the streamflow forecast skill. Here, the LISFLOOD routing and groundwater model parameters are calibrated with ECMWF reforecasts from 1995 to 2015 as forcing using daily streamflow data from 1287 stations worldwide. The calibration of LISFLOOD parameters is performed using an evolutionary optimization algorithm with the Kling-Gupta Efficiency (KGE) as objective function. The skill improvements are quantified by computing the skill scores as the change in KGE relative to the baseline simulation using a priori parameters. The results show that simulation skill has improved after calibration (KGE skill score > 0.08) for the large majority of stations during the calibration (67% globally and 77% outside of North America) and validation (60% globally and 69% outside of North America) periods compared to the baseline simulation. However, the skill gain was impacted by the bias in the baseline simulation (the lowest skill score was obtained in basins with negative bias) due to the limitation of the model in correcting the negative bias in streamflow. Hence, further skill improvements could be achieved by reducing the bias in the streamflow by improving the precipitation forecasts and the land surface model. The results of this work will have implications on improving the operational GloFAS flood forecasting (www.globalfloods.eu).

The impact of lake and reservoir parameterization on global streamflow simulation.

Lakes and reservoirs affect the timing and magnitude of streamflow, and are therefore essential hydrological model components, especially in the context of global flood forecasting. However, the parameterization of lake and reservoir routines on a global scale is subject to considerable uncertainty due to lack of information on lake hydrographic characteristics and reservoir operating rules. In this study we estimated the effect of lakes and reservoirs on global daily streamflow simulations of a spatially-distributed LISFLOOD hydrological model. We applied state-of-the-art global sensitivity and uncertainty analyses for selected catchments to examine the effect of uncertain lake and reservoir parameterization on model performance. Streamflow observations from 390 catchments around the globe and multiple performance measures were used to assess model performance. Results indicate a considerable geographical variability in the lake and reservoir effects on the streamflow simulation. Nash-Sutcliffe Efficiency (NSE) and Kling-Gupta Efficiency (KGE) metrics improved for 65% and 38% of catchments respectively, with median skill score values of 0.16 and 0.2 while scores deteriorated for 28% and 52% of the catchments, with median values -0.09 and -0.16, respectively. The effect of reservoirs on extreme high flows was substantial and widespread in the global domain, while the effect of lakes was spatially limited to a few catchments. As indicated by global sensitivity analysis, parameter uncertainty substantially affected uncertainty of model performance. Reservoir parameters often contributed to this uncertainty, although the effect varied widely among catchments. The effect of reservoir parameters on model performance diminished with distance downstream of reservoirs in favor of other parameters, notably groundwater-related parameters and channel Manning's roughness coefficient. This study underscores the importance of accounting for lakes and, especially, reservoirs and using appropriate parameterization in large-scale hydrological simulations.

A global streamflow reanalysis for 1980-2018.

Global and continental scale hydrological reanalysis datasets receive growing attention due to their increasing number of applications, ranging from water resources management, climate change studies, water related hazards and policy support. Until recently, their use was mostly limited to qualitative assessments, due to their coarse spatial and temporal resolution, large uncertainty and bias in the model output, and limited extent of the dataset in space and time. This research reports on the setup of a gridded hydrological model with quasi-global coverage, able to reproduce a seamless 39-year streamflow simulation in all world's medium to large river basins. The model was calibrated at 1226 river sections with a total drainage area of 51 million km2 within 66 countries, using ECMWF's latest atmospheric reanalysis ERA5. A performance assessment revealed large improvements in reproducing past discharge observations, in comparison to the calibration used in the current operational setup of the hydrological model as part of the Copernicus - Global Flood Awareness System (GloFAS, www.globalfloods.eu), with median scores of Kling-Gupta Efficiency KGE = 0.67 and correlation r = 0.8. The simulation bias was also dramatically reduced and narrowed around zero, with more than 60% of stations showing percent bias within ±20%. Pronounced regional differences in the simulation results remain, pointing out the need for detailed investigation of the hydrological processes in specific regions, including parts of Africa and South Asia. In addition, observed discharges with high data quality is key to achieving skillful model output. The new calibrated model will become part of the operational runs of GloFAS in the next system release foreseen for Spring 2020, together with a near real time extension of the streamflow reanalysis.