Forecasting droughts in East Africa EmmahMwangi1, Fredrik Wetterhall2, Emanuel Dutra2, Francesca Di Giuseppe2, and Florian Pappenberger2 1. Kenya Meteorological Agency 2. European Centre for Medium Range Weather Forecasts
Introduction – Climate of East Africa • East Africa: two rainy seasons (Mar-May & Oct-Dec) • Movement of ITCZ • IOD, ENSO, MJO, QBO • GDP - rainfed agriculture
Introduction – drought outlook • Increase in frequency and intensity of droughts: 2008-2009, 2010-2011 • Major economic and humanitarian impacts • Accurate drought predictions with adequate lead time is essential • Existing seasonal forecasting system; GHACOF (Greater Horn of Africa Climate Outlook Forum)
Sudan Eritrea Djibouti Ethiopia Somalia Uganda Kenya Rwanda Burundi Tanzania Great Horn of Africa region (GHACOF) • ICPAC (IGAD Climate Prediction and Application Centre) • GHACOFs – 1998 • GHACOFs – 3 times a year; March-May, July-August, October-December
Regionalization of the countries using PCA into homogeneous climatological zones • Correlation analysis with SSTs • QBO, IOD, Ocean gradients • Regression analysis • Analogue technique: find years with similar climate drivers as the current year • Dynamical models from several centres
Statement • Problems related to water scarcity are likely to occur in northwestern and northeastern Kenya ; monitoring and contingency measures are necessary in order to adequately cope with the situation. • Diseases associated with water scarcity • Food security is expected to deteriorate in the eastern sector October-December 2013
Research questions: • Does ECMWF seasonal forecast of precipitation have skill over eastern Africa? • If so, is this information useful for the decision makers?
Observational data and forecast • Monthly rainfall for the 34 homogeneous zones over the period 1961–2009 • Hindcasts of ECMWF System 4, 15 members from 1981-2010 Skill assessment: Quantitative skill in of precipitation forecast (ACC, CRPSS, ROC) Qualitative evaluation mimicking the outlook forecast • Seasonal forecasts of precipitation anomalies • Seasonal forecasts of standardised precipitation index
Continuous Rank Probability Skill Scores (CRPSS) • Prediction skill declines with increasing lead time • Skill is higher in the OND than in MAM • For both methods, there is higher skill in lead time 2 than lead time1 in the OND season • SYS-4’s negative drift in SSTs over the NINO 3.4 region which highly impacts precipitation over East Africa.
Use of system-4 in the consensus framework – OND 2000 SYS-4 September and October forecasts and the consensus forecast, then the outlook could have been adjusted for the Kenya coast, Ethiopia and Sudan.
Use of system-4 in the consensus framework – OND 2006 • If the consensus would have been updated in October using SYS-4 forecast, then the wet conditions observed on the Eastern part could have been captured.
Use of system-4 in the consensus framework – MAM 2009 • Combining the outlook and SYS-4’s March forecast would have helped adjust the wet forecast over Ethiopia and Sudan to dry • .
Conclusions • SYS-4 has significant skill in forecasting precipitation over the study area with in predicting the short rains for October-December • The subjective assessment showed that there is a potential added advantageusing SYS4, especially in terms of a late update of the forecast • Needs to be further evaluated • Use of SPI made the forecast more easy to interpret and showed the areas with anomalies in a more homogenous way
Thank you for you attention! Mwangi, E., Wetterhall, F., Dutra, E., Di Giuseppe F. and Pappenberger, F., (2014), Forecasting droughts in East Africa, Hydrology and Earth System Sciences, 18, 611-620