Temporal and spatial scaling impacts on extreme precipitation
- 1Climate Service Center 2.0, Hamburg, Germany
- 2Hydrology Research unit, SMHI, Norrköping, Sweden
- 3Niels Bohr Institute, Copenhagen, Denmark
- 4Max Planck Institute for Meteorology, Hamburg, Germany
Abstract. Convective and stratiform precipitation events have fundamentally different physical causes. Using a radar composite over Germany, this study separates these precipitation types and compares extremes at different spatial and temporal scales, ranging from 1 to 50 km and 5 min to 6 h, respectively. Four main objectives are addressed. First, we investigate extreme precipitation intensities for convective and stratiform precipitation events at different spatial and temporal resolutions to identify type-dependent space and time reduction factors and to analyze regional and seasonal differences over Germany. We find strong differences between the types, with up to 30% higher reduction factors for convective compared to stratiform extremes, exceeding all other observed seasonal and regional differences within one type. Second, we investigate how the differences in reduction factors affect the contribution of each type to extreme events as a whole, again dependent on the scale and the threshold chosen. A clear shift occurs towards more convective extremes at higher resolution or higher percentiles. For horizontal resolutions of current climate model simulations, i.e., ~10 km, the temporal resolution of the data as well as the chosen threshold have profound influence on which type of extreme will be statistically dominant. Third, we compare the ratio of area to duration reduction factor for convective and stratiform events and find that convective events have lower effective advection velocities than stratiform events and are therefore more strongly affected by spatial than by temporal aggregation. Finally, we discuss the entire precipitation distribution regarding data aggregation and identify matching pairs of temporal and spatial resolutions where similar distributions are observed. The information is useful for planning observational networks or storing model data at different temporal and spatial scales.