Use of geostatistical methods for the estimation of spatial distribution of rainfall for Pangani basin

The ultimate goal of this study was to model the spatial variation of rainfall and to estimate the rainfall surface for Pangani basin by using geo-statistical techniques of characterising the spatial continuity. The research was carried out in two steps. The first step is the application of experimental measures of spatial variability and variogram models. Whilst the second step is the application of interpolation techniques (Kriging). In the first step, the variogram model was used to model spatial fluctuation of the quantity under study (in this case rainfall), software called VARIOWIN (Panntier, 1996) is used for this purpose. The kriging technique was used to weight irregular space point data to estimate rainfall of regularly spaced prediction grid with certain resolution. In this analysis various geostatistical techniques are implemented, Ordinary Kriging (OK), Kriging with an external drift (KT), Co-Kriging (COK) and Ratio Kriging. The Ordinary Kriging method has given a reasonably good R2 of 59.9%. An attempt was made to improve the rainfall estimation by incorporating related secondary variable, in this case Cold Cloud Cover (CCD) and Digital Elevation Model (DEM). The results have shown good improvement for the kriging with an external drift (KT), R2 was improved from 59.9 to 65.1 when DEM is taken as secondary variable. An encouraging improvement is also observed in including the secondary variables as ratio to the primary variable (rainfall). The results of R2 were improved from 59.9 to 76.7 when DEM is the secondary variable and 59.9 to 63.31 when CCD is the secondary variable. Encouraging result was also obtained from the Co-Kriging Method. The R2 was not improved when CCD is the secondary variable but improvement of R2 from 59.9 to 61.0 was obtained when DEM is used as secondary variable. Generally, in all techniques of spatial rainfall, use of secondary variable has improved the estimated rainfall and thus they are essential components of spatial rainfall characteristics.