Evaluation of LFM-2 Quantitative Precipitation Forecasts

Abstract

The results of a near real time experiment designed to assess the state of the art of quantitative precipitation forecasting skill of the operational NMC LFM-2 are described. All available LFM-2 quantitative precipitation forecasts were verified on an area-averaged basis for southern and central New England for the period 0000 GMT 3 January through 0000 GMT 14 May 1979. Individual point verifications were also made for Albany, Boston, Concord, New York and Portland. On an area-averaged basis the LFM-2 beat (lost to) the climatological control by +18.5% (?7.4%) for the 12?24 h (24?36 h) forecast projection. On a point basis the overall LFM-2 forecasts lost to climatology by ?7.2 and ?21.9% for these forecast projections. Close examination of the results suggests that much of the loss of predictive skill in the model forecasts is the result of systematic overprediction of precipitation accompanying major cyclonic events. Possible reasons for this behavior are examined through a discussion of individual synoptic cases.