Teleconnections between decadal rainfall variability and global sea surface temparatures and simulation of future climate scenarios over East Africa
Aming'o, Philip O
MetadataShow full item record
The overall objective of this study is to investigate the dominant spatial and temporal decadal rainfall variability modes and their teleconnection with decadal variability modes of the specific global oceans. Knowledge derived from the teleconnection is used to examine the predictability potentials of the modes of decadal variability of East African rainfall. Specific objectives that were undertaken to achieve the overall objective of the study include delineation of the region into homogenous zones with similar decadal variability modes; investigation of the teleconnection of the regional decadal rainfall variability patterns with global Sea Surface Temperatures modes; examination of the predictability potentials of the regional decadal rainfall variability patterns together with probable future regional climate scenarios and compare near-term projections with predicted decadal rainfall using Regional Climate Model (RCM). The data sets used in the study include monthly observed rainfall over East Africa and global sea surface temperature covering the period 1950 to 2008. Other data sets used include gridded data of the University of East Anglia's Climate Research Unit (CRU) for the period 1961 to 1990; the European Centre for Medium-range Weather Forecasts (ECMWF) 40-year Re-Analysis (ERA40) and European Community Hamburg Model version 4 (ECHAM4) model output. The observed rainfall and sea surface temperature data used in the study were smoothed using a nine point binomial coefficient filter to remove all fluctuations equal to and less than 9 years. The methods used to investigate the specific objectives included mass curve analysis to assess the quality of data used, trend and spectral analyses to investigate the dominant patterns of the existing decadal rainfall variability. Principal Component Analysis (PCA) was used to delineate the region into homogenous decadal rainfall zones while Cannonical Correlation Analysis (CCA) and Singular Value Decomposition (SVD) techniques were used to investigate teleconnection amongst decadal rainfall with Sea Surface Temperature (SST) modes over various parts of the global oceans. The predictability potentials of the regional decadal rainfall variability patterns were assessed using correlation and Multiple Linear Regression (MLR) methods. A high resolution Regional Climate Model (RCM), PRECIS (Providing Regional Climates for Impacts Studies), was applied to generate regional climate scenarios. Besides, PRECIS model was applied to produce hindcasts, that were compared with observed rainfall, and nearterm future projections that were compared with predicted decadal rainfall. The patterns of decadal varaibility showed that although wet and dry decades were recurrent, and sometimes extend over large areas, there were very few decades when floods or drought covered the whole of East Africa region except for the wet decade of 1961-1970 during the short rains (October-December) season. Results from spectral density analysis of rainfall time series smoothed with a 9-point binomial coefficients filter showed dominance of ten years period that were significant at 95% confidence level when both white and red noise hypotheses were used. The Principal Component Analysis results for decadal rainfall records yielded seven and nine homogeneous decadal rainfall zones for October-December (OND) and March-May (MAM) seasons respectively. Thus for the first time this study has provided detailed analysis and characterization of decadal rainfall modes for East Africa and delineated the region into homogenous zones, based on modes of decadal variability. This can be of great use in the long-term planning and management of all rainfall dependent activities in the region. The results obtained from analyses of teleconnection between the regional decadal rainfall variability patterns and the global sea surface temperatures using Singular Value Decomposition (SVD) analysis showed high values of Square Covariance Fractions (SCF) explained by the first three modes of the global oceans. The results for the first SVD modes for Indian, Atlantic and Pacific oceans, respectively, contributed to 50%, 43% and 38% of the total square covariance for March - May, 65%, 48% and 40% for September - December rainfall seasons. During June- August season however, the first SVD modes accounted for 61 %, 39% and 42% respectively for the same oceans. It was very evident that the EI Nino modes were prominent over the Pacific Ocean, while Indian Ocean dipole was a key feature over the Indian Ocean basin. An inter-hemispheric dipole mode that is common during ENSO was a prominent feature in the Atlantic Ocean. In general, the results from the SVD highlighted the significant roles of all the global oceans in the observed decadal rainfall variability modes over the region. Based on previous studies conducted on the rainfall varaiblity over the region, this study is the first one to clearly demonstrate strong covariance between decadal rainfall variability and specific global oceans SSTs over parts of East Africa. Results from Multiple Linear Regression (MLR) method showed substantial variation of the model prediction skill of the decadal rainfall variability modes within various homogenous zones and seasons. The Heidke Skill Score (HSS) values derived from model simulations were ~ 0.30 at all locations indicating that the models could provide forecasts with useful skills. The percentage of correct forecast in all categories was found to be between 38% and 64% for all zones. Thus, the study has for the first time demonstrated that SST decadal variability modes can provide useful insight of decada rainfall variability over East Africa region. The regional model (PRECIS) simulations reproduced realistic annual cycles and inter annual variability patterns characteristic of the regional climate. PRECIS model also captured the general trend of the interannual rainfall anomalies for the training period of the model. The amplitudes of the individual extreme rainfall events were however not well resolved in most years and seasons. However, the downscaled global climate scenarios using PRECIS model were in general agreement with projected decadal variations using Multiple Linear Regression (MLR) models in most locations of the region. The results of this study, therefore, have provided some very useful insights, tools, methods and products that can be broadly used to inform short to long term planning and management of all rainfall dependent activities in the region. In particular, the outcomes of this study will be useful milestone in mainstreaming decadal climate variability information into the regional economic development strategies. The results can be integrated in the development of new climate risk management tools to aid in coping with current climate variability and adaptation to future climate changes in East Africa.