Abstract:
Flood forecasting can help to reduce the socio-economical and ecological effect of flood
hazard in Benue Sub Basin, in Nigeria. The objective of this study is to set-up parsimonious
flood forecasting model as an input to flood warning in central part of Nigeria. In this study,
high resolution (TRMM-3B42) satellite rainfall product was used for model calibration and
real time (TRMM-3B42RT) and forecasted (ECMWF) rainfall products were evaluated for
flood forecasting. The implications of the TRMM-3B42 and TRMM-3B42RT bias were
evaluated. The HEC-HMS rainfall-runoff hydrological model was selected for this study.
Performance of the rainfall-runoff model was assessed using Nash- Sutcliffe Efficiency
(NSE), Coefficient of determination (R
2
) and Relative Volume Error (RVE) and percentage
error of peak flow (PEPF) performance indices. Real time flow forecast assessment is
conducted with respect to three different flood warning threshold levels (Medium, severe and
very severe) for 1 to 6 days lead time. The forecast skill was assessed using categorical
verification statistical such as Probability Of Detection (POD), Frequency Of Hit (FOH) and
Frequency Of Miss (FOM). It is found that the model performance was satisfactory in terms
of reproducing the observed hydrograph for the Benue basin at Makurdi station though the
performance slightly deteriorated for the calibration period. For the study area, the bias of
both the operational and real time satellite rainfall product should be corrected to improve
forecast skill. When evaluating the graphical forecast skill the bias corrected TRMM-
3B42RT flood forecast was more accurate than the ECMWF flood forecast. Satellite based
rainfall product are important for data scarcity area and crucial input for real time flood
forecasting model. However, the model performance may be improved further by
considering additional rain gauges for the bias correction of satellite products.
