Comparação de modelos de previsão de precipitação pluviométrica no Paraná na presença de um Gold Standard

Abstract

The forecasts of agrometeorological variables are applied to guide several agricultural activities. The rainfall forecasts are important due to the impacts, which water excess or deficit can cause on the development of agricultural crops. The forecasts are performed by several national centers with parameterizations that can produce different forecasts for a certain agrometeorological variable in the same site. Thus, this trial aims at carrying out comparisons in the presence of a gold standard for georeferenced rainfall forecasts in some regions of Paran´a State - Brazil. The first research studied the state of Paran´a, from October to March of cropping years 2011/2012 to 2015/2016. The forecasts were clustered in fiveday periods for step 24-hour, and ten-day for step 240-hour, in which the data are from the Canadian Meteorological Center (CMC), European Center for Medium-Range Weather Forecasts (ECMWF), National Centers for Environmental Prediction (NCEP) and Center for Weather Forecasting and Climate Studies (CPTEC) according to the base Interactive Grand Global Ensemble (TIGGE). The measures gold standard were obtained from meteorological stations of the SIMEPAR Technological Institute. The results showed some variability when selecting the forecasting centers for the studied sites, mainly in mesoregions of Paran´a state. The Western Paran´a mesoregion area was studied in the second trial, from October to March of the soybean cropping years 2010/2011 to 2015/2016. The predictions of CMC, ECMWF, NCEP and CPTEC centers at the 240-hour were compared with the gold standard, which is obtained by mean values of observations from National Water Agency - Brazil (ANA) meteorological stations within a 0.36◦ radius of the centroid of the predicted site (pixel) clustered in ten-day periods. The agreements were obtained with normal distribution, and normality was achieved by Yeo-Johnson transformation with data divided into four ranges. The results indicated variability based on centers which were chosen in the mesoregion, according to the precipitation range. The third trial was based on the second paper data and the agreements were obtained using the three-parameter gamma distribution and the gamma difference distribution. The agreements with gamma distribution modelling ratified the variability of selecting centers in the Western Paran´a mesoregion for each range.