No abstract

<p class="IOPText"><span lang="EN-GB">Global hydrological models (GHMs) supply key information for international stakeholders and policymakers, simulating the impacts of the water cycle associated with climate change. Uncertainty in simulation, e.g., linked to climate models, model structure and parameters, jeopardizes valuable decision support. Various scenario data sets have been used, and model‑intercomparison studies have been performed in climate change studies to account for uncertainty linked to climate models and model structure, respectively (Kundzewicz et al., 2018). However, uncertainty in baseline data, used (1) for parameter adjustment of GHMs, and (2) assessment of relative changes in future, has rarely been addressed. Here we show that neglecting the uncertainty related to baseline data can mislead decision-making when assessing the impacts of climate change. We found that three different calibrated versions of the GHM WaterGAP3 (using three different sources of baseline data, namely EWEMBI2b, E-OBS and German Weather Service) reveal contradicting results regarding future streamflow for the German part of the Danube basin. Whereas one data set shows a decreasing 90th percentile of streamflow, indicating less heavy flood occurrence, the other datasets show an increasing 90th percentile of streamflow, indicating the opposite. Although the impact of baseline data (and consecutive parameter estimation) is already present at the mesoscale (Remesan & Holman, 2015), it is often overlooked in climate change studies using GHMs. Our results demonstrate that the choice of baseline data must be considered a source of uncertainty for climate change studies using calibrated GHMs. We anticipate that our study will increase awareness of baseline data's importance and contribute to valuable decision support for international policy related to floods, drought, and human water management.</span></p>

No abstract

Global hydrological models (GHMs) supply key information for stakeholders and policymakers simulating past, present and future water cycles. Inaccuracy in GHM simulations, i.e., simulation results that poorly match observations, leads to uncertainty that hinders valuable decision support. Improved parameter estimation is one key to more accurate simulations of global models. Here, we introduce an efficient and transparent way to understand the parameter control of GHMs to advance parameter estimation using global sensitivity analysis (GSA). In our analysis, we use the GHM WaterGAP3 and find that the most influential parameters in 50% of 347 basins worldwide are model parameters that have traditionally not been included when calibrating this model. Parameter importance varies in space and between metrics. For example, a parameter that controls groundwater flow velocity is influential on signatures related to the flow duration curve but not on traditional statistical metrics. Parameters linked to evapotranspiration and high flows exhibit unexpected behaviour, i.e., a parameter defining potential evapotranspiration influences high flows more than other parameters we would have expected to be relevant. This unexpected behaviour suggests that the model structure could be improved. We also find that basin attributes explain the spatial variability of parameter importance better than Köppen-Geiger climate zones. Overall, our results demonstrate that GSA can effectively inform parameter estimation in GHMs and guide the improvement of the model structure. Thus, using GSA to advance parameter estimation supports more accurate simulations of the global water cycle and more robust information for stakeholders and policymakers.