| A fourteen year monthly data set of Normalized Difference Vegetation Index (NDVI) measurements for Africa is used as a proxy indicator for interannual climate variability. The data have been analyzed at two temporal scales: short time scale, 2-5 years and long time scale, 14 years. Using standardized Principal Components Analysis (PCA), spatial and temporal patterns are extracted. The first five, and sometimes six components extracted represent patterns related to the time integral of NDVI, various modes of seasonality and anomalies related sensor degradation. There is evidence of two or more interannual patterns immediately following the seasonal components that are related to El Nino Southern Oscillation (ENSO). One of the modes of interannual variation detected, affecting most particularly East Africa and the Sahel, does not exhibit a consistent ENSO relationship. In particular instances the interannual components related to ENSO are separated into two components, a deficiency of the analysis technique to resolve phenomena that propagate over space. In order to examine the geographic patterns of ENSO manifestation, growing season anomaly maps are derived and fitted using trend surface techniques. The derived surfaces illustrate that there are two types of alternating patterns of ENSO manifestation over Southern Africa. Type I, characteristic of the 1986-87 and 1994-95 ENSO warm events, are characterized by movement from the southwest to the northeast, Type II, 1982-83 and 1991-92 events are characterized by a more or less stationary drought cell that shows limited movement but with a tendency to grow out in a cellular manner during the growing season.; Analysis of the entire time series shows three dominant modes of interannual variability, the Sahel-East Africa component related to QBO, a southeastern Africa component centered over eastern Zimbabwe and Mozambique closely related to ENSO, and lastly an anomaly pattern centered over Botswana, South Africa and the West African coast that shows a high frequency monthly variability but with an underlying 7-10 year variability pattern. Over the Sahel-East Africa regions, the NDVI anomaly pattern is largely inversely related to ENSO, although this is not common to all events. The analyses indicates that ENSO patterns are best retrieved using 2-5 year time series data sets. It is not possible to clearly map and construct a single index of ENSO based on PCA long time series results. A composite NDVI anomaly index is therefore derived from the various NDVI time windows analyzed that shows the spatio-temporal land surface response to individual ENSO events over Africa between 1982 and 1995. The overall correlations results show that there are very strong and significant links between interannual variability of NDVI over Africa and ENSO. (Abstract shortened by UMI.)... |