Assessing the quality of quantitative precipitation forecasts from numerical weather prediction models has been accomplished by either comparing a precipitation forecast grid to observation stations or to an analysed gridded field of precipitation. If the aim for evaluating precipitation forecasts is to asses the accuracy of the forecast at certain location, then one needs to interpolate the forecast to the observation location and use the actual observations as verification data (grid-to-point approach). This verification approach is used operationally at the European Centre for Medium Range Weather Forecasts, Deutscher Wetterdienst, Turkish Meteorological Service and some other places. This approach does not smooth observations. Its disadvantage is that it smoothes the forecasted precipitation field, which increases the minima and reduces the maxima. This method does not conserve the total precipitation forecast by the model. If the aim is to examine multiple forecast models, then one should compare, on a common verification grid, an analysed field of precipitation to the gridded precipitation forecast (grid-to-grid approach). This verification technique is currently used operationally at the National Centers for Environmental Predictions (NCEP) and at the Australian Bureau of Meteorology. Some advantages of this technique are that the gridded (re-mapped) observations better represent the grid-scale quantities predicted by the model, and the sampling is spatially uniform. The technique preserves, for some degree of accuracy, the total forecast precipitation of the native grid, although it introduces some minimal smoothing due to interpolation. Its disadvantage is that the analysis tends to smooth the observations. This paper outlines the basic properties of the two QPF verification techniques that employ the two basic approaches introduced above and presents statistical results generated during the DTC Winter Forecast Experiment (DWFE), which was performed from 15th January to 31st March 2005 (For details of DWFE see Nance et al. 2005 and Bernardet et al. 2005a).
This publication was presented at the following: