| description abstract | This study 1) presents a logistic regression equation of the snow ratio (SR) for use in a conversion of numerically predicted precipitation amounts into snowfall depths and 2) examines the quality of snowfall-depth forecasts using the proposed SR equation. A logistic regression equation of SR has been derived with surface air temperature as the predictor, using observed 3-h snow ratio and surface air temperature. It is obtained for each of several ranges of the precipitation rate to reduce the large variability of SR. The proposed scheme is found to reproduce the observed SRs better than other schemes, according to verification against an independent observation dataset. Predictions of precipitation and snowfall using the Weather Research and Forecasting (WRF) model and the proposed SR equation have shown some skill for a low threshold [1 mm (6 h)?1 and 1 cm (6 h)?1 for precipitation and snowfall depth, respectively]: the 10-case mean threat scores (TSs) are 0.47 and 0.43 for precipitation and snowfall forecasts, respectively. For higher thresholds [5 mm (6 h)?1 and 5 cm (6 h)?1 for precipitation and snowfall depth, respectively], however, TSs for snowfall forecasts tend to be significantly lower than those for the precipitation forecasts. Examination indicates that the poor predictions of relatively heavy snowfall are associated with incorrect prediction(s) of precipitation amount and/or surface air temperature, and the errors of the estimated SRs. The proposed SR equation can be especially useful for snowfall prediction for an area where the spatial variation of precipitation type (e.g., wet or dry snow) is significant. | |
