Validation of GCM precipitation in Lake Victoria basin: the case Study of Lake Victoria Basin Tanzanian Side

Since 2000, Lake Victoria levels have fallen dramatically to an extent that affects availability of water resources. This situation is related to both deteriorated weather and irrational water release from the Lake. The objective of the study is to validate GCM at various temporal and spatial scales. The study involved 21 GCM precipitation data of 30 years from 1960 and observed precipitation data for 136 stations over Tanzanian side of Lake Victoria. The analysis was carried out at Monthly, Annual, Short rain season (Vuli) and Long rain season (Masika) at observation point stations and basin wise scale; in order to determine a suitable GCM at given spatial and temporal resolution for water resources management. The attained best result of R2 (Coefficient of Determination) of the data at station scale on average are 0.414, 0.115, 0.728 and 0.181 for Monthly, Annual, Vuli and Masika respectively, and their corresponding values at Basin level were 0.544, 0.216, 0.339 and 0.339. Analysis on “Volumetric Fit Index” (VFI) has shown an average best fit of 0.965, 0.965, 0.949 and 1.034 for Monthly, Annual, Vuli and Masika respectively, and their corresponding values at Basin level were 0.962, 0.962, 0.948 and 1.027. The study has shown that the R2 is weak on independent stations and basin scale. The maximum R2 observed is around 41.41% for stations and 54.44% for basin scale. In general the VFI has indicated the best fit for 6 GCM of around and GCM scenario bcm2 has shown relatively best results at station and basin levels. Using the R2 and VFI ranks at point station scale for monthly and annual temporal scale, GCM scenario echam4.picntrl ranked the highest followed by echo_g.picntrl then by pcm1.picntrl and bcm2_0.picntrl. Overall results for this study indicate GCM scenario fgoals1.0_g.picntrl is potentially suitable for climate changes studies in the basin. Further studies such as statistical downscaling are recommended