• English 
  • Spanish 

Skill of raw and downscaled seasonal predictions of monsoonal precipitation from the ENSEMBLES multimodel over Senegal and Ghana

The aim of this work is to compare the skill of raw and downscaled seasonal predictions of precipitation during the monsoon season (July-August-September) over Senegal and Ghana. For such purpose, the state-of-the-art ENSEMBLES seasonal multimodel, which covers the period of study 1961-2000, is used.
In a first step, the model’s ability to reproduce monsoonal rain two months in advance is evaluated by using a robust probabilistic validation based on terciles. The poor performance found (especially for Senegal) might be related to the GCM coarse resolution, which is not able to capture small-scale processes triggering monsoonal rains. Therefore, downscaling approaches need to be undertaken.
In this context, a meticulous search for identifying the most adequate downscaling configurations (predictors-domain) in Perfect Prognosis is conducted. Then, various methodologies are followed to statistically downscale the GCM outputs to the local stations where observational data of good quality are available.
Apart from probabilistic scores, comparison between raw and downscaled predictions is given in terms of other indicators such as the frequency of wet days or the mean length of wet spells.
This study is part of the work done in the QWeCI project, funded by the European Commission's 7th Framework Program through contract 243964.