Climatic Determinants and Statistical Prediction of Tropical Cyclone Days in the Southwest Indian Ocean

Climatic determinants of tropical cyclone (TC) days in the southwest Indian Ocean area (10°–25°S, 50°–70°E) are analyzed using statistical techniques. A TC days index is formulated from records of local meteorological services over the December–March season in the period 1961–91. The index is correlated with gridded fields of sea surface temperature (SST), outgoing longwave radiation (OLR), and tropospheric winds, using monthly standardized departures at various lags. SST relationships with TC days are positive over the entire southwest Indian Ocean from −4 to +2 months, as expected. Peak correlations of >+0.5 occur in the genesis region 0°–10°S, 50°–60°E to the northeast of Madagascar at lag −4 (September). The synoptic-scale response of monsoon convection is approximated by OLR correlations. Negative correlations (associated with increased convection) are found to the northeast of Madagascar at lag −4 and 0 months. At lags −4 and −2 (November) opposing positive OLR correlations are found over Africa, suggesting a convective sink region during the spring season transition. Wind correlation vectors at the 200-hPa level indicate the persistence of an anticyclonic gyre centered near 35°S, 70°E in the south Indian Ocean and upper easterly flow in the equatorial zone. Surface northwesterly flow is a prominent feature in the central Indian Ocean (Diego Garcia), while strengthened midlatitude westerlies are found at lag −4 (September). In November surface northwesterly flow anomalies dominate the entire tropical zone with respect to summers with increased TC days. At lag 0 and to a lesser extent +2 months, a distinct cyclonic anomaly is centered on 20°S, 55°E with enhanced monsoon westerlies to the north. The correlation patterns offer statistical guidance in long-range forecasts and insights to the climatic processes involved in the interannual variability of TC days in the southwest Indian Ocean. Using predictors selected from present analysis, a linear multivariate model is constructed. The model has three predictors from the preceding July to November period and accounts for 59% of the variance over the 1971–92 period. The model performs adequately, achieving a jackknife correlation of 70% and a Heidke tercile score of 52.5%. A conceptual framework is used to highlight relationships between the predictors, the Indian monsoon, and tropical cyclogenesis.