Abstract
Regional scale distributed conceptual models are typically developed with a bottom-up approach, which is process-inclusive but prone to over-parameterization. Here we demonstrate a proof of concept top-down approach for distributed conceptual model development, intended to emphasize dominant streamflow generating processes and to fulfill the principle of model parsimony. A key challenge in applying the top-down approach to distributed model development is devising a model comparison experiment that is both informative and limited to a few model alternatives. Here, we show how such model comparisons can be informed by a perceptual model of key processes that control streamflow response variability at the regional scale. We demonstrate our approach for the 27,100 km2 Moselle catchment, using the perceptual model developed in Part 1 of this two-part paper. We develop 5 distributed model structures for simulating daily streamflow at 26 subcatchments, and validate them on subcatchments that are not used during the calibration process. Our model comparisons illustrate how the spatial distribution of precipitation, lithology and topography affect the simulation of key signatures of streamflow response variability in the Moselle catchment, providing a basis to justify model decisions. Our analyses show how a minimally parameterized distributed model, with 12 calibration parameters, matches signatures of streamflow average (r = 0.96), baseflow index (r = 0.86), and hydrograph lag time (correct at 22 out of 26 subcatchments). Our proposed top-down approach contributes to improving distributed model development strategies, and can be used to develop parsimonious process based regional models elsewhere.
Highlights
Distributed hydrological models have been used by hydrologists since the first flood predictions on the Durnace River in France by Imbeaux (1892)
While field work and process knowledge can help with the se lection of an existing model or even structural development of a new distributed model at the small catchment scale, such selection or development options are few at regional scales and beyond (e.g. Loritz et al, 2018; Ehret et al, 2020)
In FM2022 we developed a perceptual model of the 27,100 km2 Moselle catchment, which explains the spatial variability of streamflow signatures observed at 26 gauged subcatchments
Summary
Distributed hydrological models have been used by hydrologists since the first flood predictions on the Durnace River in France by Imbeaux (1892). Loritz et al, 2018; Ehret et al, 2020) In these large catchments, even simple questions and decisions are not straightforward: Should a model have a coarser spatial resolution to limit its complexity, or a finer one, to enable a more detailed process representation? Even simple questions and decisions are not straightforward: Should a model have a coarser spatial resolution to limit its complexity, or a finer one, to enable a more detailed process representation? The answers to these questions are rather ad hoc This exposes the regional scale distributed model to the many criticisms that have been levied against them: the danger of overparameterization when available data are not sufficient to constrain model parameters Despite > 130 years of work since Imbeaux’s first distributed catchment model on the Durnace, we are still wondering how to inform the many decisions that developing a distributed model requires
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
