Drought and the predictability of the October-November rains over Sri Lanka

Drought occurrence in Sri Lanka during the main cultivation season - the Maha - that lasts from October to March was analyzed using the 6-monthly standardized precipitation index (SPI) from 1961-2001. Two severe droughts (-1.99>SPI>-1.5), 10 moderate droughts (-1.49<SPI<-1.0) and 7 mild droughts (-1.0<SPI<-0.5) took place between 1961 and 2001. One severe drought (1995), 9 moderate droughts (1964, 1968, 1971, 1974, 1980, 1981, 1991, 2000 and 2001) and 3 mild droughts (1984, 1989 and 1996) coincide with suppressed October-November (ON) inter-monsoonal rainfall. Diagnostic analyses of large-scale atmospheric features affecting ON rainfall over Sri Lanka from 1961-2005, using Canonical Correlation Analysis, revealed that the strength of the contemporaneous zonal (westerly) wind at 850hPa (U850) over the central Indian Ocean (5°S to 15°N and 40°E to 105°E) has a profound impact on the strength of ON rainfall. Temporal scores of the leading U850 mode were analyzed to identify salient characteristics of the U850 field in years when the ON rains failed. ON rain failure takes place under two scenarios: first, when the U850 is anomalously strong (anomaly = 0.32ms-1); and second, when U850 is relatively weak (anomaly = -0.44ms-1). Droughts associated with the first scenario occurred in 1964, 1971, 1980, 1981, 1988, 1995, 1996 and 2001 while droughts of the second scenario occurred in 1968, 1984, 1989 and 1991. The anomalously strong U850 field - associated with droughts of the first scenario - occurs most often (but not always) in conjunction with La Niña events. In all drought years, a combination of factors - including a strong vertical shear of the mean zonal wind; weak low-level vorticity; and anomalously low mid-tropospheric vertical velocity - suppress convection over Sri Lanka. This study demonstrates that ON rain failure can be predicted with good skill using predicted fields of contemporaneous U850 issued with a lag of one month in September. It uses retrospective forecasts of U850, for 1982-2002, from three GCM ensemble runs - NCEP’s fully-coupled Climate Forecast System (CFS); the ECHAM4.5 forced with persisted sea surface temperature anomalies (SSTAs); and the ECHAM4.5 forced with constructed analogues of SSTAs - to test whether such predictions could be generated on an operational basis. All three models perform well at capturing droughts of the first scenario. The only exception is the ambiguous signal represented in the NCEP_CFS’s tercile probability forecast for the severe drought of 1995. Droughts of the second scenario that occurred in 1984 and 1989 are clearly captured by the NCEP_CFS and the ECHAM4.5 forced with persisted SSTAs. The second-scenario drought of 1991 is not captured by any model. Causes of the 1991 drought are being investigated. Some factors under consideration include whether there was a role for internal atmospheric variability or the radiative influence of the 1991 Pinatubo eruption.