UCL logo

UCL Discovery

UCL home » Library Services » Electronic resources » UCL Discovery

Seasonal prediction of African rainfall with a focus on Kenya

Rourke, J.M.A.; (2011) Seasonal prediction of African rainfall with a focus on Kenya. Doctoral thesis, UCL (University College London). Green open access

[img]
Preview
PDF - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
5Mb

Abstract

Africa's climate is prone to extended rainfall deficits. In extreme cases these may lead to droughts and humanitarian disasters. Skilful prediction of seasonal rainfall would bring sound humanitarian and economic benefit to the many African countries that depend on rain-fed agriculture. Seasonal rainfall hindcast skill from the DEMETER multi-model ensemble system is examined across Africa. Skill at 0-month lead is found to be weak over much of Africa, except for the August-October (ASO) season in the Sahel and the November-January (NJD) season in equatorial East Africa, Nigeria and South Africa. For the ASO season, correlation values of 0.3-0.8 (p-values < 0.1) are found across the sub-Sahara belt. For the NDJ season, correlation values of 0.5-0.6 (p-values < 0.1) occur in Kenya, Tanzania and Uganda. Innovative statistical seasonal rainfall hindcast models are developed for six homogeneous rainfall regions in Kenya, using linear regression techniques. Kenya has experienced seven severe droughts over the period 1991-2008 affecting over 35 million people. Lagged sea surface temperature and atmospheric wind predictors are selected based on having a significant and temporally stable correlation with regional rainfall indices, and a clear physical-linking mechanism. Moderate-to-high rainfall hindcast skill is found for most regions at 0- and 1-month leads for the October-December rainy season. In contrast, no robust predictors are found for the March-May rainy season. In 2009 an improved version of DEMETER, called EUROSIP, was released. This study is the first to assess the skill of the EUROSIP rainfall hindcasts for the Kenyan October-December rainy season and to compare this with the statistical model skill. For the most heavily populated and cultivated West and Southwest regions of Kenya, which are home to 68% of the Kenyan population, the statistical models outperform the EUROSIP model with correlation values ≥ 0.42 (p-values ≤ 0.06) over the common verification period 1987-2005.

Type:Thesis (Doctoral)
Title:Seasonal prediction of African rainfall with a focus on Kenya
Open access status:An open access version is available from UCL Discovery
Language:English
UCL classification:UCL > School of BEAMS > Faculty of Maths and Physical Sciences > Space and Climate Physics

View download statistics for this item

Archive Staff Only: edit this record