Diagnostics and predictability of east African rainfall with tropospheric circulation parameters

Abstract

Rainfall is one of the weather and climate parameters that is highly variable in time and space with extremes sometimes manifested as excessive rainfall or floods and severe rainfall deficits or droughts. The national economies of East Africa are heavily dependent on rainfall which mainly occurs during the two major rainfall seasons, namely March-May (long rains) and October-December (short rains) making the predictability of seasonal rainfall very crucial. Any major change in the onset, duration and intensity in seasonal rainfall has severe adverse impacts on several socio-economic activities in the region. There is to some extent some prediction potential in the predictability of seasonal rainfall in the region. However, most of the current prediction tools are based on distant global and regional climate systems associated with rainfall anomalies in the region. The main goal of the present study is to develop new reliable local predictors to improve the skill in the prediction of seasonal rainfall and enhance provision of timely climate early warning products and services for application in the planning and management of various climate sensitive socio-economic activities in the East African region. The specific objectives of the study is to determine the best set of upper tropospheric variables that could provide new predictors of the two major rainfall seasons and improve the seasonal rainfall predictability in the region. The upper tropospheric variables investigated included 300 hPa temperature anomaly (thermal index) in relation to rainfall and linkages with strong Indian Ocean Dipole (I0D) and EI Nino/Southern Oscillation (ENSO) climate systems; 300 hPa. geopotential anomaly (geopotential Index) and the geopotential derived Energy Index expressed as the variation of geopotential anomalies at 500 hPa and 300 hPa surfaces. The data sets used in the study included monthly and seasonal rainfall; upper-air monthly temperature and geopotential heights; sea surface temperatures (SSTs) and Southern Oscillation Index (SOl) covering the period 1956-2000. The rainfall stations used in the study were selected from homogenous rainfall zones derived from Principal Component Analysis (PCA) by IGAD Climate Prediction and Applications Centre (lCPAC) for the two major rainfall seasons in the East African region. The suitability of the estimated missing rainfall data were however examined first by fitting straight lines on rainfall mass curves to ensure quality and homogeneity of the data before inclusion in the analyses. The methods adopted included diagnostic analysis to determine the best set of upper tropospheric variables that could provide new and reliable local predictors of seasonal rainfall. The diagnostic tools used included composite and correlation analyses. Existence of any unique linkages between upper tropospheric thermal anomalies during the years of strong IOD and ENSO associated with some regional seasonal rainfall anomalies in the region were also investigated. The prediction potentials of the upper-tropospheric predictors in the timely seasonal rainfall outlooks were also examined using both correlation and regression analyses. The regression analysis was used to provide the best empirical equation that can describe the relationship between the predict and and the new predictors. The analysis results of monthly rainfall from homogeneous rainfall zones displayed three types of rainfall patterns namely unimodal, bimodal and trimodal where the majority of the stations in the region were dominated by the bimodal rainfall pattern. The analysis of the fitted straight lines to cumulative rainfall curves were indicative of good quality and homogeneity of the rainfall data that were used in the study. Results from normalised rainfall indices indicated common existence of extreme droughts and floods in the rainfall seasons of March-May and October-December. It was discernible from the results that it is not common for both the March-May and October-December rainfall seasons to have similar rainfall anomaly signals during the same year. It was also noted that many severe floods/droughts observed in a given season were often followed by droughts/floods respectively in the following season/year. This may be due to strong linkages between regional rainfall and ENSO events. However, some of the extreme rainfall events also occurred during the strong lOD years. Results from the study indicated evidence of significant relationships between seasonal rainfall anomalies and 300 hPa temperature anomaly (thermal index) also linked to strong lOD and ENSO events; 300 hPa. geopotential anomaly (geopotential index) and the Energy Index. The spatial correlation patterns indicated that there were some reasonable time lags between seasonal rainfall anomalies and the various upper-troposphere indices with significant predictive values. For example onset of warm/cold upper-troposphere thermal index regime starts appearing in January that could be useful for the prediction of the March- May seasonal rains while the July and August values could also be used as predictors of the October-December seasonal rains. The thermal index displayed the strongest signal in the prediction of March-May rainfall season. The regression analysis results indicated that prediction of seasonal rainfall using upper tropospheric temperature and geopotential indices had higher skill during the October-December season compared to the March-May season. The skill of the seasonal forecast however varied significantly from one location to another with percentages variance ranging from 30 to 70%. The study has for the first time found new local and reliable prediction tools to provide improved seasonal rainfall predictions and enhance the early warning and disaster preparedness systems for timely planning, management and mitigation from adverse impacts of the weather and climate related catastrophes on the socio-economic activities in the East African region.