Tropical cyclones claim a huge number of lives and cause substantial damage to property and crops in many regions each year. Southern Africa is no exception. This makes the process of forecasting tropical cyclones of great importance to the region’s economy and to public safety. Skillful seasonal forecasts of tropical cyclone activity could be used to warn the communities affected by tropical cyclones of the likely occurrence of such systems ahead of the cyclone season. This could result in reduced damage and fatalities associated with such systems. Both statistical and dynamical techniques have been employed in an attempt to predict tropical cyclone activity on a seasonal time scale over a number of ocean basins. The skills of such techniques vary from one technique to another and from one basin to another. This study investigates the predictability of tropical cyclone activity on a seasonal time scale over the southwestern Indian Ocean (SWIO) by nesting a regional climate model (RCM), the RegCM3 within a coarse-resolution atmospheric general circulation model (AGCM), the ECHAM4.5. The national meteorological centres of most southern African countries do not have the required dedicated computational resources to run the high-resolution GCMs that are suitable to predict these systems operationally. However, these systems can be very devastating on the southern African region and need to be predicted on various time scales, including the seasonal time scale. Therefore, it is instructive that research be done to better our understanding of these systems and their predictability using physical models. This study examines the simulations of the genesis locations and the number of tropical cyclones produced in RCM integrations nested within an AGCM forced by observed sea-surface temperatures (SSTs). The season of interest is the mid-summer period of December to February. Four members of the AGCM generated at the International Research Institute for Climate and Society (IRI) are used to force the RCM. Four-month integrations over a 10-year period (1991/92-2000/01) are performed. An objective procedure for detecting model-generated tropical cyclones is applied to this ensemble. Some characteristics of the simulated cyclones are compared with the observations. In addition, some statistical techniques are employed to evaluate the capability of the RCM to reproduce some aspects of the observed tropical cyclones during the aforementioned period. The results show that there is a good agreement between two of the simulated and observed environmental variables that influence tropical cyclone formation, viz. vertical wind shear and relative vorticity. In particular, the simulated and observed vertical wind shear show a similar pattern in most parts of the model domain. With regards to the relative vorticity, the highest agreement is found in the Mozambique Channel and in the region east of Madagascar. In addition, there is an appreciable agreement between the simulated and observed tropical cyclone characteristics such as tropical cyclone genesis locations and frequency. The model also simulated the interannual variability in the tropical cyclone frequency skillfully. AFRIKAANS : Tropiese siklone is verantwoordelik vir ‘n goot aantal sterftes en veroorsaak beduidende skade aan eindom asook oeste oor etlike areas elke jaar. Suidelike Afrika is nie ‘n uitsondering nie. Hierdie verliese maak die voorspelling van tropiese siklone van groot belang vir die gebied se ekonomie asook vir publieke veiligheid. Vaardige seisoenale voorspelling van tropiese sikloon aktiwiteit kan gebruik word om gemeenskappe wat onderhewig is aan die invloed van tropiese siklone te waarsku oor die kans vir sulke sisteme om voor te kom voordat die tropiese sikloon seisoen ‘n aanvang neem. Vroegtydige waarskuwings kan tot gevolg hê dat daar minder verwant skade en laer sterftes is. Beide statistiese en dinamiese tegnieke is al in die verlede gebruik om tropiese sikloon aktiwiteit oor verskeie oseaankomme op ‘n seisoenale tydskaal te probeer voorspel. Die vaardigheid van hierdie tegnieke hang af van die tipe tegniek wat gebruik word asook watter oseaankom beskou word. Hierdie studie ondersoek die voorspelbaarheid van tropies sikloon aktiwiteit op ‘n seisoenale tydskaal oor die suid-westelike Indiese Oseaan deur gebruik te maak van ‘n streeksmodel, die RegCM3, genes in ‘n growwe-resolusie algemene sirkulasie model van die atmosfeer, die ECHAM4.5. Die nasionale weerdienste van die meerderheid Suider-Afrikaanse lande beskik nie oor die nodige rekenaars om geskikte hoë-resolusie algemene sirkulasie modelle te loop om sodanige sisteme mee operasioneel te voorspel nie. Desnieteenstaande kan hierdie tropiese sisteme verwoestend wees en daarom behoort hulle voorspel te word op verskeie tydskale, insluitende seisoenale tydskale. Dit sal dus insiggewend wees om navorsing te doen om sodoende ons begrip oor hierdie sisteme en hul voorspelbaarheid te verbeter deur gebruik te maak van fisiese modelle. Hierdie studie gaan ondersoek instel oor die simulasie van tropiese siklone oor hul ontwikkelingsgebiede en die aantal tropiese siklone wat ‘n streeksmodel, genes in ‘n algemen sirkulasie model van die atmosfeer wat geforseer word deur waargeneemde see-oppervlak temperature, kan produseer. Die seisoen van belang is die mid-somer periode van Desember tot Februarie. Vier ensemble lede afkomstig vanaf die algemene sirkulasie model wat geloop is by die International Research Institute for Climate and Society word gebruik om die streeksmodel mee te forseer. Model integrasies word oor ‘n 4-maand periode gedoen en vir ‘n 10-jaar tydperk (1991/92-2000/01). ‘n Objektiewe vorteks opsporingsprosedure word dan toegepas op die 4-lid ensemble om model-geskepte tropiese siklone te identifiseer. Sommige van die karakteristieke van die gesimuleerde siklone word dan vergelyk met die waargeneemde tropiese stelsels. Hiermee saam word statisiese tegnieke ingespan vir die genoemde tydperk om die vermoë van die streeksmodel te ondersoek om sekere aspekte van waargeneemde storms te herproduseer. Die resultate wys dat daar ‘n goeie ooreenkoms is tussen twee van die gesimuleerde en waargeneemde omgewingsveranderlikes wat tropiese sikloon ontwikkeling beinvloed, nl, vertikale windskuiwing en relatiewe vortisiteit. In besonder het die gesimuleerde en waargeneemde vertikale windskuiwing ooreenstemmende patrone gelewer oor die grootste gedeelte van die streeks model-area. Wat relatiewe vortisiteit betref, is die beste ooreenkoms oor die Mosambiek kanaal en in die gebied oos van Madagaskar gevind. Verder is daar ‘n sterk ooreenkoms tussen die gesimuleerde en waargeneemde tropiese sikloon karakteristieke soos by die tropiese siklone se ontwikkelingsgebiede asook hul frekwensie. Die model he took daarin geslaag om die inter-jaarlikse veranderlikheid van tropiese sikloon frekwensie suksesvol te simuleer. / Dissertation (MSc)--University of Pretoria, 2010. / Geography, Geoinformatics and Meteorology / MSc / Unrestricted