Skill In Forecasting Daily Temperature and Precipitation: Some Experimental Results

Abstract

Estimates of skill in prediction of daily temperature and precipitation are obtained from a six-year record of real-time forecasts for Boston made in the Department of Meteorology at Massachusetts Institute of Technology. No secular increase of skill is found during the period 1966?1972, despite continuous improvement of predicted synoptic-scale flow patterns at the surface and at 500 mb, which were used for guidance. Skill, defined as incremental accuracy of the forecasts over forecasts of the climatological mean, is near 50% for the first day. The decrease of skill with increasing range is more rapid for precipitation than for temperature, the 10% level being reached in two and one-half days and four days, respectively. A pronounced seasonal variation of skill, with a summer minimum, is attributed to a similar variation in the importance of synoptic-scale, as opposed to mesoscale, sources of weather variability. The latter sources are seen as primarily responsible for the present limit of first-day skill. Bias in the probability forecasts of precipitation is relatively small. For the first day, unbiased forecasts representing near certainty are often made, while for the fourth day probability statements representing departures of more than 10 or 15% from the climatological probability are not reliable. Comparison of our results with trends in skill of temperature and precipitation forecasts made at the National Meteorological Center confirms that an increase in the skill of synoptic circulation prognoses is no guarantee of increased skill in predicting the weather.