Browsing by Author "Batungwanayo, Pacifique"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Spatio-temporal modelling of precipitation in Tanzania
    (University of Dar es Salaam, 2016) Batungwanayo, Pacifique
    Many environmental variables that are relevant to climate change and variabilities, such as precipitation, wind and temperature, may experience both temporal and spatial variations. Having accurate information about such variables at any particular point in space and time, can often be of interest to different users, and may assist in decision-making. In this work, the best linear unbiased prediction called kriging in geostatistics is proposed for estimating monthly precipitation values in Tanzania. Two main analyses were focused: First, we explain the basic elements of time series models with application to the case of mean monthly precipitation data from Tanzania Meteorological Agency (TMA) from 11 stations. Seasonal autoregressive integrated moving average (SARIMA) models were built for each station and can be used for a short time forecasting. Secondly, different elements of space-time geostatistics were discussed, particularly variogram models, which are extremely important in kriging. The weighted least squares method for parameter estimation was considered for fitting variogram models. A number of variogram models are fitted and only optimal parameters are chosen. Thereafter, simple kriging, ordinary kriging and universal kriging methods are tried out based on the best model selected and precipitation maps are produced for monthly data from Climate Forecast System Reanalysis (CFSR) dataset for a number of 806 stations. Spatial trends and seasonal differences in precipitation for each month of the year 2014 are observed and discussed. The results show that the universal kriging method improves the estimation and fits significantly