| description abstract | The variability of the West African monsoon on the intraseasonal time scale is a major issue for agricultural strategy, as the occurrence of dry spells can strongly impact yields of rain-fed crops. This study investigates this intraseasonal variability of rainfall over West Africa and gives a first overview of its predictability at a medium lead time. A statistical method, the singular spectrum analysis, is applied to a ground-based rainfall index in West Africa to describe first temporal patterns of the main leading modes of intraseasonal variability. The results point out the existence of one oscillatory mode of 34 days, one of 20 days, and one of 14 days. The same methodology is applied to rainfall from two reanalysis datasets and to deep convection from satellite data in order to assess the accuracy of the representation of intraseasonal variability in these datasets. It is shown that although the day-to-day variability of rainfall is not well captured in these datasets, intraseasonal features and, in particular, the low-frequency mode are very well reproduced. The medium lead-time predictability (5?10 days) of the intraseasonal modes is investigated using both the dynamical forecast scheme of the ECMWF and a statistical method, the maximum entropy method. For the latter method, an operational application using unfiltered input data is also considered. The performance of these prediction schemes is compared using a simple reference technique in which forecasts are based entirely on persistence. It is found that statistical predictions are much more promising than the dynamical ones, though they encounter problems when applied operationally. In an operational application, the forecast skill for the 10?90-day intraseasonal band is low but the predictability of individual intraseasonal modes is higher. The stability of the forecast skill levels is influenced by the characteristics of the intraseasonal mode. When the characteristics (i.e., amplitude and period) of the considered intraseasonal mode are well defined, skillful forecasts can be obtained. However, when the characteristics change rapidly, the forecast fails. | |
