A decision support system for Lake Victoria: hydrologic module and lake behavior

A critical review of the hydrological module of the Lake Victoria decision support system has been done in this study. The hydrologic models i.e. Sacramento-lumped and Sacramento-distributed have been applied on all gauged catchments contributing to lake Victoria as delineated in Lake Victoria decision support system. All system models i.e. simple linear system models, non-linear models and simple non-linear models were applied on the same catchments and results compared. A conceptual model (Xinanjiang model) was also applied to the same catchments for the same period. Linear perturbation model gave the best results among the system models applied. The conceptual Xinanjiang model gave results far better than both Sacramento-lumped and Sacramento-distributed and better than all system models based on Nash & Suchtlife efficiency criteria. Average monthly means of observed and estimated series of Xinanjiang and Sacramento models were determined and graphs drawn to further show the superiority of Xinanjiang model on the catchments. Each catchment was analyzed separately for better comparison. Hydrological diagrams were also obtained to give more evidence on the performance. This implies that the hydrological model used in the hydrological module of Lake Victoria is inadequate to be used in forecasting inflows into the lake. The Sacramento-lumped gave better results than Sacramento-distrubuted which demonstrates the inability of distributed model especially when the grid used is big (e.g. grid > 500 km2). The annual water balance of the lake was done for both observed periods and simulated periods. For the observed periods, the precipitation data, rainfall data, evaporation data, inflows and outflows were prepared as described in Part 11 of the study. For the simulation period, the all data except evaporation were simulated. Rainfall over the lake was simulated using Markov chain of order one model. The magnitudes for wet day series were modeled using lognormal distribution with method of moments as method of estimating parameters. The inflows and outflows were generated using the famous Thomas-Feiring model for the next 100 years. The Evaporation data used was similar to the one used in the observation period i.e. estimated using modified complementary relationship evaporation model (Morton's model). The annual balances were obtained from monthly model balances, described in part II of the study. The lake levels for both periods were plotted against their respective years to get the general trend of the lake levels. The net annual basin supplies for both periods were obtained and plotted against their respective years. The variation of the response time for the lake due to constant net basin supply was determined as shown in the graphs of constant net basin supply versus years from the start period by assuming constant annual net basin supply of ranging from 0-1.0. The periods of constant annual net basin supply were identified from the graph and data for both periods as shown. Equilibrium levels of the constant net basin supply were obtained from the rating curve and the behavior of the lake predicted for the past and present century. The annual water balance showed that the variations in the lake levels are primarily related to variations of rainfall over the lake and the surrounding basin. It was further confirmed that the response time of the lake is between 10-20 years and present and future trends of the lake are governed by small climatologically homogeneous periods with abrupt and predictable shifts between each period.