Publication Date:
2023-04-20
Description:
The factors determining a flood can essentially be divided into two components: the meteorological drivers such as rain and snowmelt, and the catchment factors such as antecedent soil moisture, relief or topography, which determine runoff formation. In its simplest form, the flood volume can therefore be determined by a combination of precipitation and runoff coefficient. According to previous analyses, the flood peak can be specified from the flood volume if the event types and their timescales are categorised beforehand. Based on this basic assumption, a deterministic-stochastic event generator was developed, which generates simulated pairs of flood volumes and their corresponding peaks on the basis of statistical distributions of the input variables. An important aspect of the generator is the classification of flood events by flood-generating factors. This allows the consideration of different distributions of precipitation, depending on the flood type. For the runoff coefficient, however, such a differentiation according to flood types alone is not sufficient. Instead, the characteristics of the catchment area itself as well as the initial soil moisture conditions have to be considered, too. This is done by applying machine-learning based algorithms, which estimate the runoff coefficient based on precipitation and antecedent soil-moisture in the catchment for hydrologically similar regions. The type-specific pairs of flood peak and flood volume generated in this way can then be used for flood statistics, e.g. to determine confidence intervals in type-based flood statistics. The deterministic-stochastic structure of the event generator allows an application to ungauged catchments.
Language:
English
Type:
info:eu-repo/semantics/conferenceObject
Permalink