Lightning is a phenomenon that can cause death or injury to humans and animals, damage to infrastructures, and can be a hazard to various sectors like the aviation and forestry industries. There is a need for prediction techniques to ensure the protection of people and property. In this dissertation, a new lightning threat index (LTI) is proposed for southern Africa. The aim of the LTI is to identify the areas where lightning is likely to occur during the day. Before the LTI could be developed, it was necessary to identify candidate model predictors capable of predicting the occurrence of lightning. In total 25 predictors were selected from literature that showed promising results to forecast the occurrence of lightning. The selected predictors are different variations from the following six groups of parameters; convective available potential energy, lifted index, precipitable water, equivalent potential temperature, relative humidity and air temperature. This study identifies the parameter from each of the six groups capable of predicting the occurrence of lightning over southern Africa the best during spring and summer by means of stepwise logistic regression techniques. The six parameters identified in this study for spring are; the most unstable convective available potential energy in the 1 - 6 km above ground level range, surface lifted index, mean precipitable water in the 850 to 300 hPa layer, minimum relative humidity in the 3-6 km above ground level layer, equivalent potential temperature lapse rate between 700 and 500 hPa and mean temperature in the 850 700 hPa layer. During summer, the same parameters were identified, except that the average relative humidity in the 3-6 km above ground level layer and equivalent potential temperature lapse rate between 850 and 400 hPa were identified. After the most appropriate parameters, capable of predicting the occurrence of lightning, were identified, the development of the new LTI could commence. Since the goal was to develop a single index that utilises the different model predictors to forecast the binary outcome of lightning occurrence (yes or no), attention was given to binary logistic regression techniques. In this study a rare-event binary logistic regression technique is used to develop equations for the LTI that utilise NWP model output early in the morning to provide a probability forecast of where lightning is expected to occur during the day between 07:00 and 21:00 UTC. The new LTI is evaluated over an entire independent spring and summer season. Results show that the LTI forecasts have a high sensitivity and specificity for both the spring and summer seasons. The LTI is not so reliable during the spring season, since it over-forecasts the occurrence of lightning, but during the summer season, the LTI forecast is reliable, only slightly over-forecasting the lightning activity. The LTI produces sharp forecasts during both the spring and summer seasons. The LTI will be a useful tool to operational weather forecasters or sectors interested in lightning forecasts, to provide guidance early in the morning on the areas of interest where lightning can be expected during the day, and can ultimately contribute to society by aiding with timely warnings of lightning or thunderstorms to protect humans, animals and property.