Rebscher, Angela ; Schmalz, Britta (2018)
How to generate input for small-scale modeling - Example of soil erosion.
17th Biennial Conference ERB2018 (Euromediterranean Network of Experimental and Representative Basins). Darmstadt (11.-14.09.2018)
Conference or Workshop Item, Bibliographie

Abstract

Soil erosion through precipitation and surface runoff from agricultural areas is a main cause of soil degradation in Europe. Modeling approaches are needed to estimate the magnitude of erosion processes and evaluate appropriate protection measures. In physically-based approaches, used for small scales, a general difference exits between continuous and event-based modeling. While the latter is driven by high spatial and temporal resolution and an emphasis on good estimation of initial conditions, the main challenge for continuous models is the parametrization of the enormous amount of input data. At the Chair of Engineering Hydrology and Water Management (ihwb) of Technische Universität Darmstadt (Germany) a systematical analysis was conducted with the example of the Water Erosion Prediction Project (WEPP) model. Eight years of measurement of discharge and sediment yield from several small watersheds in the German low mountain range area (Fiener and Auerswald 2007) were analyzed and compared with calibrated and uncalibrated modeling results in WEPP. Overparametrization proved to be more severe than in other hydrological tasks. On one hand this is a consequence of the much bigger amount of input data, including more detailed representation of land use, hydraulic input for routing of surface runoff and additional soil parameters for the erosion module. Yet on the other hand calibration events for sediment yield and even runoff are rarely available, because of the very small catchment scales. Therefor a calibration –validation approach is often not applicable.Without calibration yearly average was modeled accurately, yet the modeled and measured hydrograph and sedigraph fit poorly. Due to the combination of several modules (climate, land cover, hydrology, hydraulics, erosion) model structure and interactions of modules is extremely complex, which can lead to unexpected outcome. Uncertainty analysis showed that all parameter groups influenced the modeled discharge and sediment yield drastically. An emphasis was given on input parameters that would not be the subject of calibration in classic hydrological modeling i.e. the generation of climate input and the representation of topography. None the less these input parameters proved very important for the quality of results. Hence, measurement uncertainty has to be accounted for. Unfortunate estimation or generation of input can result in misleading modeling results even though still appearing reasonable. Additionally, high resolution input did not necessarily produce better outcome –again due to the complex model structure. Using broad bands of possible input within reasonable uncertainty proved to give the most reliable results for uncalibrated models.

Send an inquiry

Show editorial Details