The present work analyses the inter-seasonal predictability of precipitation during the austral summer in a subregion of Southeastern South America that includes Rio Grande do Sul and Northern Uruguay (RGS-NU), and proposes a methodology to produce probabilistic precipitation forecasts for this region, based on the use of NOAA CFS v2. It is found that the correlation between ENSO and the precipitation over RGS-NU during December-January-February is statistically significant after the late 70's, but not before. Considering that this relationship changes in different multidecadal periods, it is useful to explore a forecast system based on numerical models. We studied the hindcasts from NOAA CFS v2 initialized during October of the years 1983 to 2009, and found that the hindcasts of meridional wind at 850 hPa averaged over certain region of South America have statistically significant skill, in terms of correlation, to predict the observed precipitation over RGS-NU. The proposed forecasts are based on this relationship. The CFS v2 hindcasts also show realistic anomalous circulation patterns associated with the anomalous precipitation in the region and the season considered. This suggests that the forecasts may be further improved by using regional models combined with the CFS v2 outputs.
Seasonal forecast; Statistical downscaling; Southeastern South America; Hindcasts
