Flood modelling in the Ammer watershed using coupled meteorological and hydrological models

Numerous applications of hydrologic models have shown their
capability to simulate hydrologic processes up to a reasonable
degree of certainty. In terms of flood modeling, the hereby
required quality of the key input parameter precipitation is of
vast importance but often remains questionable. This Ph.D. thesis
presents a critical review of experiences made in the EU-funded
RAPHAEL project. Different meteorological data sources are
evaluated to assess their applicability for flood modeling and
forecasting in the Bavarian pre-alpine watershed of the Ammer river
(709 km²), for which the manifold hydrologic aspects of runoff
production as well as the complex nature of floods are described.
Apart from conventional rain gauge data, forecasts of several
Numerical Weather Prediction Models (NWP) as well as rain radar
data and precipitation derived from METEOSAT are examined, scaled
and applied within the framework of a GIS-structured and physically
based hydrologic model. Multi-scenario results are quantitatively
compared and analysed. The synergetic approach leads to promising
results under certain meteorological conditions, but also
emphasizes a variety of drawbacks. At present, NWPs are the only
data source to provide placed rainfall forecasts (up to 96 hours)
with large spatial coverage and high temporal resolution. On the
other hand, the coarse spatial resolution of NWP grids cannot as
yet image the heterogeneous structures of orographic rainfields in
complex convective situations and hence introduces a major
downscaling problem for mountain catchment applications. As shown
for two selected Ammer flood events, a high variability in
prediction accuracy still has to be accepted at present.
Sensitivity analysis of both meteo-data input and hydrological
model performance in terms of process description are discussed,
drawing positive conclusions for future applications of an advanced
meteo-hydro model synergy.