Variability of summer rainfall over tropical north Africa (1906–92): Observations and modelling

Abstract

The nature and causes of seasonal (July to September (JAS)) rainfall variability over tropical north Africa are investigated using a combination of empirical and modelling approaches. Concentration is focused on three regions: the Sahel, Soudan and Guinea Coast. The variability of seasonal rainfall through the twentieth century is analysed for each region. The well known dipole of anomalies between the Sahel and Guinea Coast is evident, and new analysis reveals that this fluctuates primarily on time‐scales of five years or less. Attention is then focussed on the causes of rainfall variability in each region, by examining the relationships with global sea‐surface‐temperature (SST) patterns; a partitioning of the data into low‐ and high‐frequency components is found to be particularly useful. So as to attribute more convincingly the primary cause of seasonal rainfall variability to global SST forcing, a suite of general circulation model experiments are performed, aimed at simulating JAS rainfall anomalies for ten past years between 1949 and 1990. Each is forced by the observed SST patterns for the appropriate year. In almost every case, the model quite skilfully simulates the magnitude and pattern of JAS rainfall anomalies across tropical north Africa, thus strengthening the idea that global SST variations are indeed responsible for most of the variability of seasonal rainfall. The relative impact of two additional sources of variability is also investigated. First the role of internal atmospheric variability is quantified (using an ‘analysis of variance’ technique), and found to be small in all three regions. Second, and perhaps more controversially, the possible effects of land‐surface‐moisture feedback are explored. This is done by replacing the normal interactive soil‐moisture scheme with a model‐derived climatology; results suggest that in some years moisture evaporated from the land surface may play a key role, but that in general SST forcing still dominates. Finally, an assessment of the model's skill at sub‐seasonal time‐scales reveals that fluctuations of monthly rainfall about each year's seasonal mean (intra seasonal variations) are insensitive to SST forcing, in part due to a larger influence of internal atmospheric variations.