The water phase of precipitation-a comparison between observed, estimated and predicted values

The results from two projects aimed at improving precipitation products for the Norwegian Road Authorities are presented and compared in this paper: A precipitation forecast with water phase type (rain, sleet, snow) from the HIRLAM weather prediction model, and a precipitation type estimate based on synoptic observations and weather radar data. Operational forecasts of water phase type from HIRLAM are biased towards predicting too many cases with snow. This is caused by the mainly positive difference between model terrain and real topography. It is shown that additional calculations of snow melting based on temperature and humidity in the lowest model level reduce this error. The second product presented here is a precipitation type estimate based on synoptic observations and weather radar data. Synoptic observations of 2 m-temperature and dew point temperature at 3 h temporal resolution are used to estimate the water phase of precipitation at ground level. The dew point temperature is converted to relative humidity. Temperature and relative humidity are interpolated in space using a digital elevation model after deriving local vertical gradients of the two variables. The water phase is estimated and merged with the actual radar precipitation estimate every 15 min. The result is a high resolution image showing the intensity and the water phase of precipitation. A comparison to observed water phase types at the synoptic stations for the winter season 2002/2003 shows that the percentage correct is 86%. The Probability of Detection is 0.84 for rain and 0.97 for snow. For January 2002, the observed, estimated, and forecasted water phase types are compared at an independent set of precipitation stations in Southern Norway. The results confirm the good quality of the precipitation type estimate and show that the adjustment of HIRLAM water phase type to real topography improves the forecast in cases with observed rain.