The study analyses 19 years of hourly quantitative precipitation estimates (RADKLIM-RW) concerning the spatial scaling behaviour of extreme rainfall for Germany as well as general heavy rainfall patterns. Four regions with the same size of 256 km × 256 km were selected with varying characteristics. For each region, spatial rainfall maxima for different aggregation steps with box length of 2, 4, 8, 16, 32, 64, and 128 km were retrieved using sliding and fixed window methods. Rainfall was classified with two parameters into patterns dominated by convection or advection, respectively. Based on the calculated spatial maxima, potential maximum rain rates were extrapolated for sub-pixel scale. Though the power laws of the resulting scaling relationship have a very good fit, no mono-scale power law could be derived from the data due to different rainfall mechanisms on different spatial scales. Typical power laws exist for more convective characteristics up to a scale of 8 km box length and for more advective characteristics with larger box lengths. A regional power law for maxima of convective rainfall could be used to statically refine the radar resolution to 250 m or even 125 m box length. However, when scaled to smaller box lengths or to the size of a Hellmann gauge (200 cm2) the values become increasingly unrealistic. The results indicate that radar derived areal precipitation values are limited to the radar scale (here 1 km2), despite the fact that they are scaled to fit observed values from rain gauges.