Mauritius has recently experienced drastic weather conditions such as flashfloods that caused major collateral damage and life loss. Predicting such weather conditions with conventional forecasting systems is not possible because these systems provide predictions for large regions over hours. However, there are several weather forecasting systems based on localised sensors connected to cloud computing facilities that can provide short-term forecasts for small regions. These forecasting systems make use of techniques such as neural networks, Fuzzy logic, time series and regression analysis. The main aim of this research is to set up a short-term weather forecasting system for Mauritius that will collect weather data such as temperature, humidity, water level etc. The sensors will relay the data to a cloud that will apply different forecasting algorithms and mobile alerts can even be generated when required. Investigations will be performed in different regions and time periods to compare the performances of different predictions algorithms. Subsequently, an adaptive algorithm that will provide optimum prediction based for a given region and time interval will be developed. This forecasting system will be of relatively low-cost and will have good potential for commercialisation. For example educational and other institutions might consider setting up their own weather stations. Additionally, the system could be used in the tourism sector, especially by hotels for their guests to anticipate weather conditions in regions they want to visit and also in the agriculture sector where chain crop producers can particularly plan their daily schedules so as to increase efficiency and productivity.