THE PREDICTABILITY OF SEASONAL MAXIMUM TEMPERATURE IN SOUTH-AFRICA

An analysis of the variability and predictability of seasonal mean max imum temperature over South Africa is conducted. Possible relationship s between global scale sea-surface temperature fields and seasonal mea n maximum temperature over South Africa are sought by means of a stati stical technique, namely 'singular value decomposition'. Associations between sea-surface temperature and seasonal mean maximum temperature over South Africa are evident. In particular, associations between El Nino/Southern Oscillation related signals and seasonal mean maximum te mperature over the eastern half of the country were found. Also, the l ong-term warming trend in the ocean temperatures has important influen ces on the temperature variability over South Africa. In this study ca nonical correlation analysis is used to construct a forecast scheme fo r the prediction of seasonal mean maximum temperature over South Afric a at different lead times. Evolutionary patterns in the ocean temperat ure field are used as predictors from which categorical seasonal mean maximum temperatures for 77 stations in South Africa are predicted. Ca tegorical seasonal mean maximum temperatures are predicted and evaluat ed for an independent test period to verify the skill and therefore us efulness of such predictions. (C) 1998 Royal Meteorological Society.