6B.5 Multi seasonal rainfall assessments for southren Africa using the CSIRO9 AGCM

Monday, 5 April 1999: 9:30 AM
Hannes (C.J.deW.) Rautenbach, University of Pretoria, Pretoria, South Africa

The CSIRO-9 atmospheric general circulation model (AGCM) has an R21 spectral resolution (64 grids from east to west and 28 Gaussian latitudes from the pole to equator) as well as 9-levels in the vertical. Observed SSTA for the period 1985 to 1996 (12 years) was prescribed as model thermal ocean forcing to evaluate model rainfall. Five-year ensemble means were calculated over seven model grid-boxes over South Africa. The resultant model rainfall for each month, and at each one of the seven model grid boxes, were ranked and classified in three categories (terciles) namely tercile 1 (below-normal), tercile 2 (normal) and tercile 3 (above-normal). The model rainfall time series for each month comprised 12 values (1985 to 1996) which implies that only four values were available per category for rainfall validation. Monthly area averaged observed rainfall (1985 to 1996) was subsequently also calculated for each one of the seven model grid-boxes, and similar to the model output, was categorized in terms of terciles. These results were used to determine (forecast) the percentage probability for below-normal, normal or above-normal rainfall to appear, provided that the model simulated rainfall lies in the below-normal, normal or above-normal category. Three month average canonical correlation analyses (CCA) predicted SSTA, issued by the Research Group for Seasonal Climate Studies (RGSCS) at the South African Weather Bureau, were prescribed as model ocean boundary forcing over the Pacific-, Atlantic- and Indian Oceans to provide nine-month model rainfall predictions. Three model predictions, with a difference in time lag, were completed to predict rainfall for the October, November, December (OND) 1998 and January, February, March (JFM) 1999 seasons. Weights were allocated to rainfall according to three-month time lags. Categorical probabilistic predictions (based on the observed climate) for below-normal, normal or above-normal rainfall were constructed and are compared with observed rainfall.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner