Forecasting Method minimum temperatures for weather stations in the western region

Abstract

A method based on “Perfect Prog” to predict the minimum temperature to 24 hours in the western region of Cuba during the winter season is developed. The dependent sample conforms the months of November through April of the years 1990 to 2005. Predictors of major contribution to increasing the variance explained by the equations are the isobaric surface temperature of 1 000 hPa and 925 hPa, and the persistence. The equations, except one, are able to explain the variance between 71 % and 82 % and the mean absolute errors vary between 0.9 ºC and 1.6 ºC. In the independent sample validation (winter season 2010-2011), the mean absolute errors vary between 0.9 ºC and 2.2 ºC. These errors are mainly due to the Global Forecast System forecasts tend to overestimate the surface wind speed at 12 UTC and underestimate the air temperature at the level of 1 000 hPa at 06 and 12 UTC. Suggested corrections for wind and temperature forecast by the Global Forecast System improve substantially the predictions, approaching the mean errors of these to those obtained with the inclusion of the actual data of the predictor variables. In general, the predictions by the method developed improve the persistence, showing in this way a satisfactory skill.