Seasonal Predictability in Northwestern Africa: Considering the Mix of Mid-Latitude and Tropical Influences

A diagnostic predictability study is reported for precipitation and surface temperature across the land domain defined by approximately: 20N northward to the Mediterranean Sea, 15E westward to the Atlantic Ocean. Much of the region is known to have strong connections to mid-latitude atmospheric modes such as the North Atlantic Oscillation, but the availability of relatively new global datasets allows for some new analyses across the whole target predictand domain defined above. We compare and contrast linkages to the north and south, and in this context, assess the potential predictability through sea-surface temperature (SST) influences on the controlling atmospheric components. Various combinations of station, satellite and reanalysis products are drawn upon, focusing mostly on the period 1979 to 2010. For the target predictands, 2m temperature (ECMWF Interim reanalysis) is analyzed for the whole domain, while for precipitation (GPCP merged station/satellite product), the domain is reduced, with the region of very low climatological rainfall in the center of the Sahara desert masked out. Supporting and validating analyses are made with station data. Diagnostic analysis tools applied include a rotated principal component analysis of the predictand gridded sets (to reveal the leading modes of spatial variation) and canonical correlation analyses (CCA) to assess large-scale climate linkages with the predictands, and explore predictability achievable from SST. The main focus is the boreal winter half of the year (October to April) which covers the primary rainfall season, although important aspects of predictability for the boreal summer half of the year (primarily related to surface temperature as a predictand) are also introduced. Analyses are initially for two-month overlapping seasons, and aspects of evolution in variability patterns and teleconnections through the annual cycle are assessed.

First, with a view to internal atmospheric mechanisms, the extent of control emanating from mid-latitude atmospheric modes versus conditions to the south (Sahara desert to the inter-tropical convergence zone) is assessed. The previously identified influence of primarily mid-latitude modes such as the North Atlantic Oscillation, the East Atlantic pattern and the Scandinavian pattern are all quantified. This is compared with the strength of teleconnections to the south, as revealed in reanalysis products and station observations.

Second, building on the above insights, the degree of SST control on the variations of the regional circulation system is evaluated leading to a diagnostic predictability assessment, focusing primarily on linear relationships with SST (as revealed by regression/CCA analysis). Similarities and differences across the predictand domain, especially from west to east, are discussed including synoptic interpretations. Some aspects considered include direct influence of tropical North Atlantic and Mediterranean SST on nearby rainfall and temperature over land, and the remote influence of tropical Pacific SST building on the known but modest teleconnections into the eastern Atlantic/northwestern Africa sector. Especially for surface temperature, seasonal predictability assessments using detrended data are consulted, so that results are not dominated by measures/relationships that simply track trends in the data. Nonetheless, the low-frequency component is of interest and relevance of itself, and must be treated in any seasonal prediction assessment, despite the reduced degrees of freedom for establishing relationships. The low-frequency component is discussed especially in the context of the expression of the Atlantic Multidecadal Oscillation.