Spatial and temporal variability of rainfall on different time scales

Rainfall is highly variable in space and time. The knowledge of precipitation variability is very important for design or for uncertainty assessment of models. In this contribution two different aspects of variability are investigated – the treatment of zero observations for spatial interpolation and the problem of high order dependence. The finer the temporal resolution of precipitation observations the more zeros have to be considered. Should one include all zeros for the description of the spatial variability (for example variograms)? Examples corresponding to different time aggregations are show that zeros need a specific treatment. High order dependence is investigated using time series observed at multiple sites. Results are compared to a meta-Gaussian approach. A large high-resolution dataset from South-West Germany is used to demonstrate the problems and the different approaches.