Computer modelling for flood propagation in the lower Niger

The Muskingum flood routing model is calibrated with the aid of computer for the lower Niger in Nigeria. The adequacy of the model for flood prediction in a typical Nigerian river system is here examined with the hope of transferring similar techniques to other river systems prone to flood occurences in the near future. The model herein developed routes hydrographs of 12 hourly mean flows from Lokoja, situated at the confluence between the Niger and Benue, through four other gauging stations and finally to Aboh, a little downstream of the Niger bifurcation. The period studied is from May, when the floods start to rise, to November, when flows are both low and fairly steady. The Muskingum flood routing model is fitted by a nonlinear least squares estimation technique and optimal, model parameters determined using the steepest descent optimisation procedure. This seeks the minimum value of the sum of squared deviations between observed and simulated outflow discharges. In addition to the usual muskingum parameters, a third parameter is included in the model to account for ungauged lateral inflows. This was necessary because no measurement of lateral inflows were available nor could unit hydrographs be used to transform rainfall to runoff. Estimates of lateral inflows are assumed uniform over each month, but vary according to the local monthly rainfall distribution patterns. The adequacy of the model is evaluated in a split sample test by comparing the simulated and observed discharge hydrographs and their monthly statistics. Results show that high floods are predicted with satisfactory accuracy and fairly well for low flows. A constained optimisation was introduced to obtain reasonable parameter values using the penalty method. Ungauged tributary inflow contributions at the Anambra river junction are estimated from a regression equation relating monthly mean discharges to rainfall. Values of 'K' obtained range between o.6 and 1.52 while 'Z' values of between 0.45 and 0.5 are obtained. 'X' on the other hand assumed a value of 0.25 for all river system as a consequence of the constraint element. But for the constraint introduced unreasonable lower values of 'X' would have resulted. The model is found to be adequate for discharge predictions in the lower Niger. It is therefore recommended for calibration in other Nigerian river systems as soon as data requirements and other forecasting facilities are available.