Exploratory Long-Range Models to Estimate Summer Climate Variability over Southern Africa

• Published in: Journal of climate Vol. 12; no. 7; pp. 1892 - 1899
• Main Authors: Jury, Mark R., Mulenga, Henry M., Mason, Simon J.
• Format: Journal Article
• Language: English
• Published: Boston, MA American Meteorological Society 01.07.1999
• Subjects: Atmosphere; Climate; Climate models; Climate variability; Corn; Earth, ocean, space; El Nino; Exact sciences and technology; External geophysics; Forecasting models; Juries; Meteorology; Modeling; Ocean currents; Oceans; Other topics in atmospheric geophysics; Rain; Summer; Tropical climates; Indian Ocean; Southern Africa; Atlantic Ocean; Africa; Teleconnection; Validation; Atmospheric circulation; Sea surface; Statistical analysis; Predictor; Theoretical study; Summer; Surface temperature; Multivariate analysis; Rain; El Nino; Regression model; Southern oscillation
• ISSN: 0894-8755; 1520-0442
• DOI: 10.1175/1520-0442(1999)012<1892:ELRMTE>2.0.CO;2

Teleconnection predictors are explored using multivariate regression models in an effort to estimate southern African summer rainfall and climate impacts one season in advance. The preliminary statistical formulations include many variables influenced by the El Niño–Southern Oscillation (ENSO) such as tropical sea surface temperatures (SST) in the Indian and Atlantic Oceans. Atmospheric circulation responses to ENSO include the alternation of tropical zonal winds over Africa and changes in convective activity within oceanic monsoon troughs. Numerous hemispheric-scale datasets are employed to extract predictors and include global indexes (Southern Oscillation index and quasi-biennial oscillation), SST principal component scores for the global oceans, indexes of tropical convection (outgoing longwave radiation), air pressure, and surface and upper winds over the Indian and Atlantic Oceans. Climatic targets include subseasonal, area-averaged rainfall over South Africa and the Zambezi river basin, and South Africa’s annual maize yield. Predictors and targets overlap in the years 1971–93, the defined training period. Each target time series is fitted by an optimum group of predictors from the preceding spring, in a linear multivariate formulation. To limit artificial skill, predictors are restricted to three, providing 17 degrees of freedom. Models with colinear predictors are screened out, and persistence of the target time series is considered. The late summer rainfall models achieve a meanr² fit of 72%, contributed largely through ENSO modulation. Early summer rainfall cross validation correlations are lower (61%). A conceptual understanding of the climate dynamics and ocean–atmosphere coupling processes inherent in the exploratory models is outlined. Seasonal outlooks based on the exploratory models could help mitigate the impacts of southern Africa’s fluctuating climate. It is believed that an advance warning of drought risk and seasonal rainfall prospects will improve the economic growth potential of southern Africa and provide additional security for food and water supplies.