It is generally acknowledged in the environmental sciences that the choice of a computational model impacts the research results. In this study of a flood and drought event in the Swiss Thur basin, we show that modeling decisions during the model configuration, beyond the model choice, also impact the model results. In our carefully designed experiment we investigated four modeling decisions in ten nested basins: the spatial resolution of the model, the spatial representation of the forcing data, the calibration period, and the performance metric. The flood characteristics were mainly affected by the performance metric, whereas the drought characteristics were mainly affected by the calibration period. The results could be related to the processes that triggered the particular events studied. The impact of the modeling decisions on the simulations did, however, vary among the investigated sub-basins. In spite of the limitations of this study, our findings have important implications for the understanding and quantification of uncertainty in any hydrological or even environmental model. Modeling decisions during model configuration introduce subjectivity from the modeler. Multiple working hypotheses during model configuration can provide insights on the impact of such subjective modeling decisions.

@en

Hydrologic projections are of vital socio-economic importance. However, they are also prone to uncertainty. In order to establish a meaningful range of storylines to support water managers in decision making, we need to reveal the relevant sources of uncertainty. Here, we systematically and extensively investigate uncertainty in hydrologic projections for 605 basins throughout the contiguous US. We show that in the majority of the basins, the sign of change in average annual runoff and discharge timing for the period 2070-2100 compared to 1985-2008 differs among combinations of climate models, hydrologic models, and parameters. Mapping the results revealed that different sources of uncertainty dominate in different regions. Hydrologic model induced uncertainty in the sign of change in mean runoff was related to snow processes and aridity, whereas uncertainty in both mean runoff and discharge timing induced by the climate models was related to disagreement among the models regarding the change in precipitation. Overall, disagreement on the sign of change was more widespread for the mean runoff than for the discharge timing. The results demonstrate the need to define a wide range of quantitative hydrologic storylines, including parameter, hydrologic model, and climate model forcing uncertainty, to support water resource planning.

@en

We discuss two definitions of reproducibility, and question if both definitions are required to be met in computational hydrological studies.@en

The transfer of parameter sets over different temporal and spatial resolutions is common practice in many large-domain hydrological modelling studies. The degree to which parameters are transferable across temporal and spatial resolutions is an indicator of how well spatial and temporal variability is represented in the models. A large degree of transferability may well indicate a poor representation of such variability in the employed models. To investigate parameter transferability over resolution in time and space we have set up a study in which the Variable Infiltration Capacity (VIC) model for the Thur basin in Switzerland was run with four different spatial resolutions (1 km × 1 km, 5 km × 5 km, 10 km × 10 km, lumped) and evaluated for three relevant temporal resolutions (hour, day, month), both applied with uniform and distributed forcing. The model was run 3150 times using the Hierarchical Latin Hypercube Sample and the best 1 % of the runs was selected as behavioural. The overlap in behavioural sets for different spatial and temporal resolutions was used as an indicator of parameter transferability. A key result from this study is that the overlap in parameter sets for different spatial resolutions was much larger than for different temporal resolutions, also when the forcing was applied in a distributed fashion. This result suggests that it is easier to transfer parameters across different spatial resolutions than across different temporal resolutions. However, the result also indicates a substantial underestimation in the spatial variability represented in the hydrological simulations, suggesting that the high spatial transferability may occur because the current generation of large-domain models has an inadequate representation of spatial variability and hydrologic connectivity. The results presented in this paper provide a strong motivation to further investigate and substantially improve the representation of spatial and temporal variability in large-domain hydrological models.

@en

A meta-analysis on 192 peer-reviewed articles reporting on applications of the variable infiltration capacity (VIC) model in a distributed way reveals that the spatial resolution at which the model is applied has increased over the years, while the calibration and validation time interval has remained unchanged. We argue that the calibration and validation time interval should keep pace with the increase in spatial resolution in order to resolve the processes that are relevant at the applied spatial resolution. We identified six time concepts in hydrological models, which all impact the model results and conclusions. Process-based model evaluation is particularly relevant when models are applied at hyper-resolution, where stakeholders expect credible results both at a high spatial and temporal resolution.

@en

A meta-analysis on 192 peer-reviewed articles reporting applications of the Variable Infiltration Capacity (VIC) model in a distributed way reveals that the spatial resolution at which the model is applied has increased over the years, while the calibration and validation time interval has remained unchanged. We argue that the calibration and validation time interval should keep pace with the increase in spatial resolution in order to resolve the processes that are relevant at the applied spatial resolution. We identified six time concepts in hydrological models, which all impact the model results and conclusions. Process-based model evaluation is particularly relevant when models are applied at hyper-resolution, where stakeholders expect credible results both at a high spatial and temporal resolution.@en

Flash flooding is a potentially destructive natural hazard known to occur in the Ce´vennes-Vivarais region in southern France. HyMeX (Hydrological Cycle in the Mediterranean Experiment) is an international program focused on understanding the hydrological cycle in the Mediterranean basin. Soil moisture is known to be a useful indicator of catchment response, however, establishing a meaningful estimation of soil moisture at the catchment level can be difficult due to its high variability in space and time.In a small gauged catchment in the Ce´vennes-Vivarais region in southern France, a series of manual soil moisture measurements was taken from September to December 2012 at both the field and catchment scale during the Special Observation Period 1 (SOP1) as part of the HyMeX program. Six plots were selected along a trajectory of a microwave link installed in the catchment and were chosen to represent different elevations in the catchment. Within each field plot, surface soil moisture was measured along a 50. m transect at 2. m intervals. This allowed the study of changes in within-field variability as well as between-field variability in response to precipitation events and during the drying out phase.Several precipitation events occurred over this autumn 2012 period which caused a significant wetting-up of the catchment, allowing the study of soil moisture processes over a wide range of wetness conditions. The influence of antecedent catchment conditions (soil moisture) on rainfall-runoff dynamics is demonstrated through the comparison of storm hydrographs for the various events. Dry catchment conditions result in minimal response in event flow, whereas large precipitation events occurring during wetter conditions produce much stronger responses in event flow. This further confirms the importance of quantifying catchment initial conditions to enhance the prediction of flash flood occurrences.

@en

In many rainfall-runoff models, at least some calibration of model parameters has to take place. Especially for ungauged or poorly gauged basins this can be problematic, because there is little or no data available for calibration. A possible solution to overcome the problems caused by data scarcity is to set up a measurement campaign for a limited time period. In this study, we determine the minimum amount of data required to determine robust parameter values for a simple model with two parameters. The model is constructed such that the parameters can be determined not only with automatic calibration, but also by recession analysis and a priori from Boussinesq theory. The model has been applied to a research catchment in Switzerland. For automatic calibration and recession analysis, one season (5 months) is found to be sufficient to give robust parameters for simulation of high flows over the full observation period. For automatic calibration, this should be the season with the highest precipitation, for recession analysis the season with least evapotranspiration. The Boussinesq equation is able to give good parameter estimates for modeling high flows, but detailed in situ knowledge of the catchment is required. Automatic calibration outperforms recession analysis and Boussinesq theory by far when it comes to parameter estimation with a focus on prediction of low flows. It was shown that a single set of parameters cannot simultaneously describe high and low flows with a reasonable accuracy, suggesting that more than two parameters are needed to characterize subsurface properties. Key Points Automatic calibration, recession analysis, and Boussinesq theory are compared Automatic calibration leads to the highest model efficiencies One season of data is needed for robust parameter estimation

@en