Real-time precipitation forecasts over tropical South America for September 1999 and
March 2000 performed using PSU/NCAR MM5 with Grell cumulus parameterization scheme were
compared with simulations performed under the same model conditions but using Kuo scheme. The area
of study was divided into regions with an area of 5ox5o and monthly precipitation averages
were calculated for each of these. The same averaging was done on satellite-based precipitation
estimates. A precipitation score index was defined using the model precipitation and the satellite
estimates and used to evaluate the precipitation skill of MM5 for each region. The model precipitation
data was also correlated with the satellite estimates in different regions of South America in order
to evaluate if the precipitation patterns were correctly reproduced by the model using the two schemes.
Our results show that, in general, Grell scheme represented the precipitation patterns correctly,
specially in the equatorial region (10oS-10oN), although it over-estimated rainfall in this
region during March 2000. Kuo scheme tends to under-estimate over the tropical region and does not
represent adequately the precipitation patterns.