Let evolution do the work: A new way for hydrological model construction Florian Ulrich Jehn1, Philipp Kraft1 ,Lutz Breuer1,2, Tobias Houska1, 1Institute
for Landscape Ecology and Resources Management (ILR), Research Centre for BioSystems, Land Use and Nutrition (iFZ), Justus Liebig University Giessen, Heinrich-Buff-Ring 26, 35390 Giessen, Germany 2Centre for International Development and Environmental Research (ZEU), Justus Liebig University Giessen, Senckenbergstrasse 3, 35392 Giessen, Germany
Why?
How?
Model building in hydrology is hard work. One has to consider many different processes and thus, there is a plethora of possible model implementations. The more complex a model, the more difficult it gets to find suitable structure and parametrizations, up to a point where it is almost impossible to overlook all the possibilities. This problem has become even clearer through the introduction of modelling frameworks such as the Catchment Modelling Framework (e.g. Kraft et al., 2011) in the wake of the method of multiple hypotheses (Clark et al., 2011), because they allow an easier and quicker construction of models. Therefore, new ways of exploring the space of potential model structures are needed.
Possible structural elements To address the differences in catchments, ACME needs to incorporate many processes. To keep things manageable and allow for an acceptable runtime, a first version will use a complex lumped model as a basic layout from which all other models are derived. Thus, ACME can construct models from this elements: - Several storages in series
- Linear or Power Law flow
Contact Florian Ulrich Jehn M.Sc., PhD-Student
[email protected]
Find me at GitHub (https://github.com/zutn) and Research Gate
- Several storages in parallel
- Usage of a canopy and snow
With this algorithm an Automated Construction of Models by Evolution (ACMEMethod) is possible. When using ACME, a human bias in model structure building can be avoided, while allowing a much more thorough sampling of the space of possible model structures. ACME is not meant to find the “best” model structure in general, but to help identify what processes are most important for a catchment and allow an easy construction of models for catchments whose properties are not well known. To avoid “stupid curve fitting” the model structures are evaluated using the Kling-Gupta Efficiency and hydrological signatures.
Precipitation
Example for a possible resulting model structure
Observed (black) and simulated discharge (blue) of the final model structure. Ready for further analysis.
References Clark, M. P., Kavetski, D. and Fenicia, F.: Pursuing the method of multiple working hypotheses for hydrological modeling: Hypothesis testing in hydrology, Water Resour. Res., 47(9), doi:10.1029/2010WR009827, 2011. Kraft, P., Vaché, K. B., Frede, H.-G. and Breuer, L.: CMF: A Hydrological Programming Language Extension For Integrated Catchment Models, Environ. Model. Softw., 26(6), 828–830, doi:10.1016/j.envsoft.2010.12.009, 2011. For a similar approach, see: Jehn, F. U., Breuer, L., Houska, T., Bestian, K. and Kraft, P.: Incremental model breakdown to assess the multi-hypotheses problem, Hydrol. Earth Syst. Sci. Discuss., 1–22, doi:10.5194/hess-2017-691, 2017.
