H103-07
Hydrological Model Experiment Setups Using the eWaterCycle Platform

Thursday, 10 December 2020: 19:30
Virtual
Jerom P.M. Aerts1, Rolf Hut1, Nick Van De Giesen1, Niels Drost2, Peter Kalverla2, Bouwe Andela2, Fakhereh Alidoost2, Ben van Werkhoven2, Jaro Camphuijsen2, Inti Pelupessy2, Stefan Verhoeven2, Willem van Verseveld3 and Albrecht Weerts4, (1)Delft University of Technology, Faculty of Civil Engineering and Geosciences, Delft, Netherlands, (2)Netherlands eScience Center, Amsterdam, Netherlands, (3)Deltares, Inland Water Systems, Delft, Netherlands, (4)Deltares, Operational Water Management, Delft, Netherlands
Recorded Presentation
Abstract:
Setting up hydrological models for hypotheses testing is a time consuming and human error prone process. At its core the setup for validating a hydrological model consists of three general parts; pre-processing external forcing data, model setup (including parameterization), and model validation.

New insights in model user behaviour based on a quantitative analyses conclude that users choose models based on experience rather than model adequacy for testing hypotheses (Addor and Melsen., 2019). This trend in behaviour can delay progress in the field of computational hydrology due to conservative model selection. To overturn this trend we provide model users with the tools that allow them to experiment with different model setups without re-inventing the model setup process.

Based on these tools we showcase model experiment setups that differ in complexity. The complexity varies in methods used for pre-processing of external forcing data (from simple linear resampling to stochastic downscaling of precipitation fields using climatology reference data), types of hydrological model (simple conceptual, process-based), and model validation methods.

The model experiments are build up-on the eWaterCycle platform (https://www.ewatercycle.org/), which is an Open Source community driven system that is designed to uphold the Open and FAIR data principles in hydrological modelling. The hydrological models are run in a containerized environment to ensure reproducibility. The pre-processing of external forcing is done by creating a common input processing pipeline based on an existing climate model analysis tool: ESMValTool (https://www.esmvaltool.org).

In this showcase we show setups including those that assess the relevant spatial modelling scales while simulating streamflow for multiple basins in the Contiguous United States using the Wflow SBM hydrological model. A stepwise modelling approach is presented in which spatial resolutions vary between 3km and 200m.

<< Previous Abstract | Next Abstract >>