Daily climatology of minimum and maximum temperature anomalies ( DTmin, DTmax) and also accumulated precipitation were studied with data from the Instituto Agronômico do Paraná (IAPAR)'s meteorological stations. It was shown that frequency distribution of DTmin occurrences for all stations is almost normal, whereas frequency distribution of DTmax occurrences is almost uniform and frequency distribution of rainfall occurrences has positive symmetry. A Model Output Statistics (MOS) was adapted using data from National Center for Environmental Prediction (NCEP) numerical model outputs and from 26 IAPAR stations and with information of minimum and maximum temperatures and accumulated precipitation. In the MOS implementation it was utilized an analogs group method. MOS was implemented for periods of april-september (colder period) and october-march (warmer period). Forecasts were calculated and evaluated using a cross validation method for up to 4 days for the period of April 1997 to March 2000. The rainfall occurrences were predicted correctly 78%, 76%, 75%, 71% for 1 to 4 days, respectively, during the winter, and 72%, 71%, 68%, 67% for 1 to 4 days, respectively, during the summer. Mean absolute errors for minimum temperature are 1.8, 1.9, 2.1 and 2.2 for 1 to 4 days, respectively, during the winter, and 1.2, 1.3, 1.4, 1.4 for 1 to 4 days, respectively, during the summer. Maximum temperature is predicted with mean absolute errors of 2.1, 2.4, 2.5, 2.7 for 1 to 4 days, respectively, during the winter, and 2.0, 2.2, 2.2, 2.4 for 1 to 4 days, respectively, during the summer. Comparisons of MOS forecasts with persistence and climatology for the same period showed the advantage of using MOS forecasts.
Mos; Climatology; Statistical forecast; Minimum and maximum temperature; Accumulated precipitation
