This work analysed heavy rainfall events and their predictability on Rio de Janeiro, Brazil, using rain gauge data from 2000 to 2010, atmospheric model outputs, and an artificial neural network based on adaptive resonance theory. The latter was applied on top of atmospheric simulations for 2009 and 2010, and we were able to predict 55% of the heavy rainfall events using a combination of relative humidity at 900 hPa and meridional winds at 10 m for a domain covering central and southern Brazil, which represents a relative gain of 67% on predictability when compared to the model predicted rainfall.
Redes Sociais