Abgeschlossene Arbeiten seit 2004
Nutzen Sie die Recherchemöglichkieten im Forschungsinformationssystem (FIS) zur weiteren Einschränkung der aufgelisteten Abschlussarbeiten.
Site classification according to risk of diffuse pollution - Adressing the issue of model complexity
Art der Abschlussarbeit
Diplomarbeit
Autoren
- David, Telse
Betreuer
- Prof. Dr. sc. techn. Peter Krebs
- Dr.-Ing. Jens Tränckner
Weitere Betreuer
Dr. Christian Stamm (eawag aquatic research), Martin Frey (eawag aquatic research)
Abstract
"Fast runoff from agricultural areas contributes to diffuse pollution of surface water bodies. Generally, the fast runoff originates only from a limited part of the catchment. The identification of this part relies on spatially distributed models which predict the hydrological processes. I address the question whether simple models are capable to represent the relevant runoff processes. I compare three spatially distributed models of different complexity: the FAL risk assessment of nitrogen and phosphorus losses, an approach to delineate the dominant hydrological runoff processes (DRP) and the Soil Moisture Distribution and Routing model (SMDR). The former two models are expert systems, whereas the latter is a mechanistic model. These three models were applied to a small agricultural catchment in the Swiss Plateau. The comparison reveals that the models agree
in the general spatial predictions of runoff patterns. A low FAL risk corresponds to a slow reacting runoff process based on DRP and a low vulnerability to runoff formation based on SMDR. But the SMDR predictions vary to a large extent within the FAL and DRP classes because SMDR takes interactions between adjacent areas into account. Such spatial interactions are neglected by the two expert systems. In a second part of my thesis, I attempt to improve the simple expert-type models by deriving a meta-model from the output of the mechanistic SMDR simulations. A regression tree was built which explains the surface runoff pattern in the study area as predicted by SMDR. The tree resembles the spatial distribution fairly well. Nevertheless, it only explains 52 % of the variability. The limited amount of explained variability suggests that a new measure needs to be developed to better represent the interactions between adjacent areas. Such a measure should consider the topography and the soil properties within the specific catchment of an area."
in the general spatial predictions of runoff patterns. A low FAL risk corresponds to a slow reacting runoff process based on DRP and a low vulnerability to runoff formation based on SMDR. But the SMDR predictions vary to a large extent within the FAL and DRP classes because SMDR takes interactions between adjacent areas into account. Such spatial interactions are neglected by the two expert systems. In a second part of my thesis, I attempt to improve the simple expert-type models by deriving a meta-model from the output of the mechanistic SMDR simulations. A regression tree was built which explains the surface runoff pattern in the study area as predicted by SMDR. The tree resembles the spatial distribution fairly well. Nevertheless, it only explains 52 % of the variability. The limited amount of explained variability suggests that a new measure needs to be developed to better represent the interactions between adjacent areas. Such a measure should consider the topography and the soil properties within the specific catchment of an area."
Schlagwörter
Diffuse losses, nutrients and agrochemicals, model complexity, regression tree
Berichtsjahr
2008