Predictability of Zimbabwe summer rainfall

Abstract

Predictors of Zimbabwe summer rainfall are investigated with a view to improved long‒range forecasts. Teleconnectivity is assessed in respect of sea‒surface temperatures, the Southern Oscillation index, the Quasi‒biennial Oscillation (QBO), outgoing longwave radiation (OLR) and wind. Spectral analyses of historical rainfall gives an indication of cycles in the range 2·3, 18 and 3·8 years, possibly associated with the QBO, the luni‒solar tide and the El Nĩo–Southern Oscillation (ENSO), respectively. Pair‒wise correlations are found between Zimbabwe summer rainfall and SST in the central Indian Ocean (r<−0·5) in austral spring. Below normal OLR values in September over southern Africa corresponds with good rains in the following summer. Rainfall–upper‒wind correlations are optimum (r<−0·7) over the equatorial Atlantic in spring. Comparatively weak correlation with the QBO may also reflect biennial adjustment of monsoon and global ENSO teleconnections. Additional predictor variables are utilized and multivariate models are formulated for early and late summer rainfall and maize yield in Zimbabwe. The models use three to five predictors, are trained over a 22‒year period and perform well in jack‒knife skill tests. Summer rainfall forecasts with one season lead times are viable and could ameliorate hardship caused by drought. ©1997 Royal Meteorological Society. Int.J.Climatol., Vol.17, 1421-1432 (No. of Figures: 5 No. of Tables: 0 No. of