Entwicklung eines räumlich verteilten, konzeptionellen Niederschlag-Abfluss-Modells für den Einsatz in Schnee-dominierten datenarmen Gebieten

In many neighboring regions of continental high mountain ranges water from snow and glacier melt is the main water resource. The snowmelt-runoff process is usually modeled by rainfall-runoff models, which have special routines for the simulation of the snow and glacier melt. However, the determination of the spatial distribution of the required meteorological input data from ground stations is very difficult and is subject to great uncertainties due to the high morphological variability and the generally low station density in most high mountain regions.
The aim of this thesis is to evaluate the use of today’s global remote sensing and reanalysis data as meteorological input for the modeling of the rainfall-runoff process with a special focus on snow-dominated catchments with poor data availability.
To model the snowmelt-runoff process, the Snowmelt Runoff Model (SRM) developed by Martinec (1975) that has been established worldwide for use in snow-dominated areas was developed further into a spatially distributed, conceptual rainfall-runoff model (rSRM). The three main input variables of the model are: air temperature, snow covered area and precipitation. A new method for reconstructing MODIS land surface temperatures has been developed, which provides the required spatially distributed daily mean temperatures needing only one reference ground station. Accordingly, existing approaches have been refined in order to close the spatial gaps in the daily MODIS snow-cover data-sets. The precipitation input was derived from an ensemble of remote sensing and reanalysis data. For this purpose, an artificial neural network as a deterministic method, the Model Conditional Processor as a probabilistic method and a combination of both approaches were used to determine the optimal data-set.
To evaluate the methods, a comparison against data of ground stations was carried out in the Swiss Alps. Subsequently, the methods were transferred to a poor-data catchment in order to test the practical applicability. To validate the model results, the rainfall-runoff process was simulated for two Swiss and one snow-dominated poor-data catchment with rSRM, HEC-HMS and an artificial neural network.
In the framework of this thesis, it could be demonstrated that the available global remote sensing and reanalysis data is extremely suitable as meteorological input for the modeling of the rainfall-runoff process in snow-dominated catchments. Of particular importance is a complete reconstruction of the remote sensing data, which is often incomplete because of cloud coverage, for the air temperature and fractional snow cover distribution. The reconstruction methods developed in this thesis provide daily data-sets with high spatial resolution that contribute decisively to a more accurate simulation of the snowmelt process within rSRM.