Abstract
The Standardized Precipitation Index (SPI) is usually defined as the arithmetic mean of the normalized precipitation recorded at several stations over a region of interest where the standard deviation computed at each station over a period of reference is used as the normalizing factor. It is common to use this index in order to diagnose whether the Sahelian region can be considered as wet or dry for a given year. There are several key factors interfering with the relevance of the SPI as a measure of how rainy is a season over the Sahel. The strong spatial variability of the Sahelian rainfall at the annual scale, the uneven distribution of the raingauge network and the mean interannual climatological gradients across the region are the most important of these factors, and their influence is studied in detail here. Using an optimal interpolation algorithm to compute the SPI, the effects of various sampling schemes are first studied showing that the SPI computed as a single mean value over the whole Sahelian region is relatively robust with respect to these effects. However, the central key question remains that computing a single mean SPI over the Sahelian region hides the strong underlying spatial variability of this index. For instance, 2006 was a significantly dry season over the Sahel as a whole, but working at the 0.5Β° Γ 0.5Β° resolution shows that in fact only 28% of the area was significantly dry, while 15% of the Sahel was significantly wet. From conditional empirical distributions a distribution function is proposed to determine the spatial distributions of the 0.5Β° Γ 0.5Β° SPI values for a given mean regional SPI value. Studying in detail the spaceβtime pattern of the SPI over the 1950β2006 period, also shows that recent years are characterized by a greater interannual variability than the previous 40 years, and by a contrast between the western Sahel remaining dry and the eastern Sahel returning to wetter conditions. Copyright Β© 2008 Royal Meteorological Society