Item – Theses Canada

OCLC number
1007026568
Link(s) to full text
LAC copy
LAC copy
Author
Ntale, Henry Kayondo,1967-
Title
The analysis and prediction of droughts in East Africa.
Degree
Ph. D. -- University of Alberta, 2001
Publisher
Ottawa : National Library of Canada = Bibliothèque nationale du Canada, [2003]
Description
2 microfiches
Notes
Includes bibliographical references.
Abstract
In this investigation on East African meteorological droughts, drought properties and patterns have been identified, three drought indices tested and modified, and two statistical teleconnection models developed for predicting its seasonal rainfall totals. Using harmonic analysis, East Africa was delineated into 6 homogeneous rainfall zones, and the important rainfall seasons (in terms of % rainfall contribution to the annual rainfall) identified for each zone. Three drought indices (Palmer Drought Severity Index or PDSI, Bhalme Mooley Index or BMI, and Standardized Precipitation Index or SPI) were analyzed, modified where necessary, and compared in terms of their consistency in detecting the initiation, evolution, severity and termination of meteorological droughts in East Africa. It seems that SPI is more versatile and consistent than PDSI and BMI in tracking East African droughts. From 6-month and 12-month SPI data, East Africa was delineated into 7 drought homogeneous zones whose spatial boundaries bear a resemblance to the 6 homogeneous zones identified from harmonic analysis. From composites developed out of 22 El Niño and 13 La Niña events and 6-month SPI data, El Niño-Southern Oscillation (ENSO) has been found to exert an influence on the moisture regime of East Africa. The degree and temporal patterns of ENSO response vary between the above-identified droughts zones of East Africa. It seems northeastern Tanzania has the strongest response to ENSO. El Niño seems to exert a stronger influence on East Africa than La Niña. Two statistical models, combined Canonical Correlation Analysis-Simplex (CCA-Simplex) and projection pursuit regression (PPR) models were developed to predict East African seasonal rainfall. CCA-Simplex is a linear model while PPR can model nonlinear associations. PPR performed better than the stand-alone CCA. By adjusting the prediction fields with 24 weights optimally determined by the Simplex algorithm, we found CCA-Simplex to consistently produce better forecasts than using un-weighted predictor fields. PPR-Simplex did not yield better results than the stand-alone PPR.
ISBN
0612689778
9780612689770