Abstract
A new system is developed which gives both conditional and unconditional probability of snow amount forecasts and categorical forecasts at stations for the 12–24 h period after both 0000 and 1200 GMT. To derive the new equations, the Model Output Statistics technique is used with output from the Limited-area Fine Mesh model (LFM).
First, experimental probability of snow amount forecast equations are developed for the 12–18, 18–24, 12–24 and 24–36 h projections. Verification of the experimental forecasts on independent data indicates that the scores generally deteriorated as compared to the scores for the developmental sample. The deterioration was worse for the 18–24 and 24–36 h projections than for the 12–18 and 12–24 h projections. It's concluded that the snow amount forecasts for the 24–36 h projection aren't skillful enough for operational implementation at this time.
Next, to improve the stability of the forecast equations for the 12–24 h projection, the developmental and independent samples are combined (nine winter seasons) and new equations are developed for ≥5,≥10 and ≥15 cm snow amount categories. In this development, the results of a statistical screening procedure indicate that, generally, the most important predictor is the LFM precipitation amount forecast followed by the mean relative humidity (surface–500 mb) forecast. Other predictors which are relatively important include LFM forecasts of circulation intensity such as 700 mb vertical velocity, 700 mb east–west wind component, 850 mb divergence, 850 mb relative vorticity, and 850 mb east–west wind component. Verification of the new equations on another independent sample indicates that the scores were generally stable except for some deterioration for the ≥15 cm category. The new equations were implemented in the fall of 1982 and should be more useful than the old operational system, which provided forecasts for the ≥10 cm category only and which had been operational within the National Weather Service since October 1977.
