Extracting statistical parameters of extreme precipitation from a NWP model

Abstract. Precipitation simulations on an 8×8 km grid using the PSU/NCAR Mesoscale Model MM5 are used to estimate the M5 and C_i statistical parameters in order to support the M5 map used for flood estimates by Icelandic engineers. It is known a priori that especially wind anomalies occur on a considerably smaller scale than 8 km. The simulation period used is 1962–2005 and 73 meteorological stations have records long enough in this period to provide a validation data set. Of these only one station is in the central highlands, so the highland values of the existing M5 map are estimates. A comparison between the simulated values and values based on station observations set shows an M5 average difference (observed-simulated) of −5 mm/24 h with a standard deviation of 17 mm, 3 outliers excluded. This is within expected limits, computational and observational errors considered. A suggested correction procedure brings these values down to 4 mm and 11 mm, respectively.