Over the last decade our answer to the question: “but what is eWaterCycle?” has changed considerably. In 2014 we presented the first iteration of eWaterCycle: we showed that it was feasible to build a real time hydrological forecasting system that ran an ensemble of global models, forced with weather forecasts, assimilating satellite observations at every timestep, all from pre-existing openly available components.

In the light of the discussion in the hydrological community on reproducible science [Hutton, 2016] we build the next iteration of eWaterCycle: a platform that allows everyone to use commonly available hydrological models. Years (and a pandemic) later this platform is now openly available [Hut, 2022]. The vision behind the platform is to take as much as possible the computer-related headaches of running other people’s models away to let hydrologists focus on the hydrology. Furthermore, eWaterCycle is ‘FAIR by Design’: it should be easy to make any analysis done by eWaterCycle adhere to the FAIR principals. Using eWaterCycle MSc students have been able to do the type of research that previously was done by a PhD and PhDs have done the type of research that previously would require a whole team of people. Large Sample hydrology studies, Model coupling and climate change impact studies have all been done using eWaterCycle.

Adding one’s own model to the platform, however, still required considerable effort which limited the uptake by the broader hydrological community. That’s why recently we released v2.0 of eWaterCycle which fixes this: it is now significantly easier to add models to eWaterCycle!

Looking forward, among other things we will be:

• Making teaching material on hydrological modelling available as Open Educational Resources through eWaterCycle [funded project]
• Adding data assimilation as a module to eWaterCycle [funded project]
• Add easy access to Large Sample Hydrology datasets (camels / caravan) [looking for students]
• Study the impact of climate change on all catchments of the world, using many different hydrological models [looking for students]
• Connect or host eWaterCycle on the infrastructure currently being developed for Destination Earth (DestinE) [looking for funds and collaborations]

In this presentation I will reflect on the achievements of the last decade, highlight the scientific results generated with eWaterCycle and look forward to the next decade.

Hutton, C., T. Wagener, J. Freer, D. Han, C. Duffy, and B. Arheimer (2016), Most computational hydrology is not reproducible, so is it really science?, Water Resour. Res., 52, 7548–7555, doi:10.1002/2016WR019285.

Hut, R., Drost, N., van de Giesen, N., van Werkhoven, B., Abdollahi, B., Aerts, J., Albers, T., Alidoost, F., Andela, B., Camphuijsen, J., Dzigan, Y., van Haren, R., Hutton, E., Kalverla, P., van Meersbergen, M., van den Oord, G., Pelupessy, I., Smeets, S., Verhoeven, S., de Vos, M., and Weel, B.: The eWaterCycle platform for open and FAIR hydrological collaboration, Geosci. Model Dev., 15, 5371–5390, https://doi.org/10.5194/gmd-15-5371-2022, 2022.

How to cite: Hut, R., Drost, N., van de Giesen, N., Kalverla, P., Verhoeven, S., Schilperoort, B., and Aerts, J.: 10 years of eWaterCycle: from prototype-forecast to platform for Open and FAIR hydrology, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-13157, https://doi.org/10.5194/egusphere-egu24-13157, 2024.

The use of machine learning (ML) methods in rainfall-runoff modelling has apparently led to better prediction, but there are some concerns about the interpretability of these models. The emergence of hybrid modelling, which couples the data driven approach with the classical physics-based conceptual approach, has shown promise in enhancing both interpretability and accuracy. ML models and conceptual models each come with their own modelling practices and habits. To develop a hybrid approach, it is necessary to consider them.

While some of the steps in these modelling chains are similar (for instance the selection of the right metric during the calibration or learning step), others are more specifics, such as the optimization of the hyper-parameters of ML models. Furthermore, the hybrid approach comes with specific methodological challenges that emerge when coupling the two different types of models. For instance, depending on the choice made by the modeller, the parameters of the conceptual model are either trained with the ML model parameters or calibrated separately by a non-ML method.

There is a need to better understand the variety of hybrid approaches and to estimate the impact of their methodological choices. This work is based on a literature review and on large-sample modelling experiments with hybridizations of two classical models running at different time steps: the monthly GR2M model and the daily GR4J model. 

How to cite: Degenne, A., Bourgin, F., Perrin, C., and Andréassian, V.: Towards a better understanding of the hybrid modelling methodology for streamflow prediction, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-10546, https://doi.org/10.5194/egusphere-egu24-10546, 2024.

Shifts in generation processes of streamflow events driven by advancing climate change are raising concerns about the adaptability of conceptual hydrological models to changing hydrological systems. Using 30-year streamflow data across 395 German catchments, we evaluate the performance of a conceptual rainfall-runoff model for three distinct streamflow event types: events associated with snow and icy conditions (Snow-or-Ice), rainfall on dry soils (Rain-on-Dry), or wet soils (Rain-on-Wet). We focus on a two-dimensional evaluation of the timing and magnitude of streamflow events using the Series-Distance approach (Seibert et al., 2016) while also diagnosing the impact of inherent process limitations on model performance using random forest. The results reveal that the modelled streamflow consistently exhibits time delays and underestimations of magnitude for all types of events. Specifically, the Rain-on-Dry are associated with the most considerable delays, while underestimation of streamflow is the largest for Snow-or-Ice events. Given the statistically significant increasing trends in the occurrence of Rain-on-Dry events across 78.8% of catchments (Mann-Kendall test, p < 0.05), it can be assumed that the timing errors might further deteriorate in the future, compromising the reliability of the model-based early-warning systems for future flood events. Additionally, the errors vary across different hydrograph components (rising limbs, peaks, and recessions) for each type of streamflow event. Peaks are the most underestimated component in all events. Further diagnostics of the links between errors and drivers identifies the pre-event errors are the most important factors of timing and magnitude errors during the events. The process limitation in the model (e.g., groundwater recharge and fast runoff process) and properties of the events themselves (e.g., duration and peak discharge of events) cause the error heterogeneity among the events and exacerbate the errors in peaks of the events. Therefore, our study highlights the critical need for further improvement of process representation in hydrological models and more accurate simulation of pre-event conditions in order to address emerging challenges posed by changing hydrological systems.

Seibert, S. P., Ehret, U., & Zehe, E. (2016). Disentangling timing and amplitude errors in streamflow simulations. Hydrology and Earth System Sciences, 20(9), 3745-3763.

How to cite: Wang, Z., Tarasova, L., and Merz, R.: Event-type-based Multi-dimensional Diagnostics of Process Limitations in Hydrological Models, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-3415, https://doi.org/10.5194/egusphere-egu24-3415, 2024.

Understanding the structural uncertainties within current conceptual hydrological models is crucial, as an appropriate model structure is essential for achieving accurate and reliable hydrological simulations. The development and evaluation of conceptual models have primarily focused on replicating streamflow dynamics, with less attention given to other important processes, such as the conversion from potential evapotranspiration (PET) to actual evapotranspiration (AET). This study assesses the performance of 33 existing conceptual model structures in simulating 8-day-scale AET across 671 catchments in the United States. These models are calibrated using both daily streamflow data and 8-day remote-sensing AET data. While most models demonstrate comparable performance in streamflow simulations, significant differences are observed in their performance in AET simulations. None of these models can consistently performs well in AET simulations across all 671 catchments, indicating that the “one-model-fits-all” assumption is not applicable. The performance of most models is found to be related to one or more catchment attributes. The most relevant catchment features are climatic, vegetation and topographical characteristics, including climatic aridity, precipitation seasonality, fraction of precipitation falling as snow, green vegetation fraction and catchment mean slope. In contrast to the “one-model-fits-all” assumption, catchments with distinct climatic, vegetation and/or topographical conditions require different ways to represent the AET process. Specifically, most models tend to underestimate AET in humid catchments where the majority of rainfall occurs in winter, except those account for interception evaporation. Additionally, models that explicitly include a vegetation transpiration component tend to perform better in catchments with denser vegetation cover. This work highlights the structure uncertainties related to AET simulations and may help model structure selections in a way to reasonably represent AET process.

How to cite: Wu, S., Yang, Y., and Zhao, J.: Effects of Conceptual Model Structure Uncertainties on Actual Evapotranspiration Simulation, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-3940, https://doi.org/10.5194/egusphere-egu24-3940, 2024.

Estimating design flood hydrographs in ungauged basins requires the determination of hydrological response parameters. These parameters are derived from relationships that often lack a solid foundation in the specific physical characteristics of the basin. Among the parameters subject to greater uncertainty, the characteristic time of the IUH function certainly stands out. Indirect methods for estimating this parameter involve the use of empirical or analytical formulas and, in the engineering practice, the use of one or more formulas is often justified on heuristic grounds, lacking solid scientific considerations to guide the choice towards the most appropriate formulation.

Here, we propose a methodological approach to provide support in choosing a robust formulation for estimating basin flood response time. We have selected 35 formulas from the literature, all containing parameters related to the basin's length and slope. After verifying the real meaning of the input parameters and units required by the formulations in the original articles where they were published, the structure of the formulas considered has been analyzed in dimensional terms, using a reasoning scheme consistent with the hydraulic relations of resistance formulas. In this way, 17 hydraulically consistent formulas have been identified.

At this stage, we point out the advantage of comparing the formulas in terms of equivalent average flow velocity rather than in terms of observed travel times. Starting from the celerities obtained as the ratio between the length of the basin's drainage path and the response times provided by each formula and using the morphology of the river network of 135 basins in northwestern Italy, we compared the variability of estimated mean travel velocities. In line with literature observations, which highlight a slight increase in mean velocities with basin size, some formulas are deemed physically inconsistent, while 5 of them were identified as hydraulically robust and consistent with empirical observations. These formulas are Chow (1962), NERC (1975), SCS (1954), McEnroe and Zhao (1999), and Watt and Chow (1985).

The results obtained analytically identify the relationships between the exponents of length and slope in each formula and those governing empirical relationships between lengths and slopes of main river reaches in the basins. These relationships allow us to identify the range of values for the exponents of length and slope in the formulas for the characteristic time for which velocity estimates increase with the basin area. Based on these relationships, it is also possible to provide a guideline for the calibration of new formulations.

How to cite: Evangelista, G., Woods, R., and Claps, P.: How to avoid unreliable formulas for time of concentration in ungauged basins, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-1038, https://doi.org/10.5194/egusphere-egu24-1038, 2024.

Extraordinary floods like the one in July 2021 have induced catastrophic consequences on both societal and economic domains. Robust model simulations are crucial for mitigating the adverse effects of such extreme events on human life. However, accurately reproducing and predicting exceptional floods remain a challenge in particular when only one such flood extreme is available in the reference record period. This single flood could be included either in calibration and evaluation period. In both cases, extreme events are missing in the other period. To analyze how to best handle a single extreme flood, we present a framework for calibrating and evaluating the mesoscale Hydrologic Model (mHM) using the July 2021 flood in western Germany as a case study. Hereby, we tested the effect of including the extreme 2021 flood in calibration or evaluation periods.

Our study shows that including the exceptional 2021 flood event in model calibration proves crucial for accurately reproducing high streamflow. Without including the 2021 flood in the calibration period, the model cannot learn how to reproduce extreme floods. Our findings reveal that employing the modified weighted Nash-Sutcliffe Efficiency (wNSE) as the objective function significantly improves mHM's performance in capturing flood peaks. This leads to a notable reduction from -35% to -7.8% in the difference between the simulated and observed/reconstructed peaks as demonstrated for the catchment outlet. The hydrological model performance was validated spatially for an independent set of gauges. Spatial validation is necessary for assessing model performance when only one exceptional historical event is available. In conclusion, our framework provides valuable insights into improving hydrologic modeling accuracy, emphasizing the importance of specific calibration strategies and spatial validation in capturing exceptional flood events.

How to cite: Han, L., Guse, B., Nguyen, V. D., Rakovec, O., Najafi, H., Guan, X., Vorogushyn, S., Samaniego, L., and Merz, B.: What are the best strategies for managing a single extreme flood event in hydrological model evaluation? – Insights from the extreme flood 2021 in Western Germany, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-16172, https://doi.org/10.5194/egusphere-egu24-16172, 2024.

The National Hydrological Model for Denmark (DK-model) is a distributed, integrated hydrological model coupling 3D groundwater flow to descriptions of root zone processes, overland flow and river routing, including anthropogenic interference with the hydrological cycle. It covers all of Denmark (~43,000km²) at 500m and 100m grid scale. Its constant development over the last three decades has both been driven by research projects and projects for public authorities. It is being used for various tasks such as water resource assessments, climate change impact assessments, hydrological real-time monitoring and nutrient transport studies.

Recently, we endeavored novel ways to calibrate and parameterize the DK-model. The model is placed on the edge between research interest and practical applications, with a demand for adequately representing various aspects of the hydrological cycle across the entirety of the model domain. In combination with its large-scale distributed nature and high computational demand, conventional (groundwater) model optimization techniques are challenged: The complex nature and versatile applications of the DK-model require suitable parametrization schemes and inclusion of diverse calibration and evaluation data, beyond conventional groundwater head observations and streamflow. This also leads to trade-offs between the multiple objective functions. Hence, we moved beyond previously used single solution, gradient-based optimization algorithms.

The Pareto Archived Dynamically Dimensioned Search (PADDS) algorithm allows us to use a global parameter optimization, effective even at a few hundred model runs. Another major advantage of PADDS is that it does not require the a-priori weighting of objective function groups – instead, it explores the tradeoffs (pareto front) between the different objective function groups, allowing weighting after gaining knowledge about tradeoffs during the optimization process. Also, all solutions explored during the optimization are stored and remain open to analysis after finished optimization. This not only sheds light on tradeoffs between different objective functions in a unique manner, but also supports understanding of parameter sensitivity and uncertainty in a manner which otherwise is hard to achieve due to computational constraints.

Moreover, we included evapotranspiration patterns from satellite products as well as a machine learning based estimate of artificial drain flow as novel spatial data in the model evaluation. This helps us constraining some of the model processes crucial for e.g. nutrient transport, but otherwise poorly constrained by conventional data such as streamflow (305 stations) and groundwater heads (24,000 wells) covering practically the entire model domain.

We explored the benefits of this optimization setup applied to the DK-model, advancing not only the calibration process itself, but also our understanding of model process representation and performance.

How to cite: Schneider, R., Troldborg, L., Højberg, A. L., Ondracek, M., Christiansen, D. T., and Stisen, S.: Innovating the calibration of a national-scale integrated hydrological model, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-14555, https://doi.org/10.5194/egusphere-egu24-14555, 2024.

Land Surface Models (LSM) grounded on physically-based mathematical models for energy and water balance can be characterized by various levels of complexity, especially when they integrate numerous processes. Diverse mathematical models (i.e., sub-models) can sometimes be formulated for some processes, due to different assumptions made during the system conceptualization stage. Therefore, running LSMs require (i) selection of a set of processes and related mathematical formulations that will be used and (ii) estimation of the corresponding parameters. A convenient way to guide model (and parameter) choice is to rely on global sensitivity analysis. In this work, we analyze sensitivity of 3 common hydrological outputs (evaporation, transpiration, and groundwater recharge fluxes) to models and parameters involved in typical LSMs. The global sensitivity analysis relies on random (Monte Carlo) sampling of values of parameters associated with each of the different formulations considered for the sub-models embedded in the LSM. This enables us to quantify the relative importance of process formulation and ensuing parameters. Three diverse indices based on (i) the whole (sample) probability density function (pdf) of the model output (Borgonovo et al., 2011) and (ii) the first and second moment of the pdf (corresponding to the moment-based sensitivity indices introduced by Dell’Oca et al. (2017)) are used. The joint use of these metrics is exemplified upon relying on realistic field conditions (in terms of, e.g., climate, vegetation, and soil type) associated with two watersheds in the Vosges region (France). Results show that sensitivity analysis plays a crucial role in identifying sub-models and parameters that contribute significantly to the uncertainty of model outputs. It is found that the main characteristics of the soil comprising the litter layer and root zone play an important role in the evaluation of the evaporation and groundwater recharge fluxes. As such, our results strengthen the need for targeted studies on the characterization of flow in these layers.

Borgonovo et al., https://doi.org/10.1111/j.1539-6924.2010.01519.x, 2011.

Dell’Oca et al., https://doi.org/10.5194/hess-21-6219-2017.

How to cite: Ackerer, P., Luttenauer, D., Dell'Oca, A., Guadagnini, A., and Weill, S.: Land surface hydrological modeling: do we really need complex model formulations?, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-14444, https://doi.org/10.5194/egusphere-egu24-14444, 2024.