PERFORMANCE EVALUATION OF FIVE PROBABILITY DISTRIBUTION MODELS FOR THE ANALYSIS OF FLOOD DATA AT RIVER NIGER

Abstract

Accurate flood prediction is essential for hydrological planning, yet selecting the most suitable probability distribution model remains a challenge. This study evaluates five statistical models to determine the most reliable method for predicting extreme flood events at River Niger. Flood frequency analysis procedure was carried out on the annual maximum discharge data for River Niger at Onitsha bridge head from 1960 to 1991 using Normal distribution, Gumbel distribution, Log normal, Log Pearson Type III and Pearson type III. Flood discharge estimates for return periods of 2 to 200 years provide valuable insights for flood mitigation strategies, hydraulic infrastructure design, and disaster preparedness. The results shows that Gumbel distribution model predicted discharge values in range of 21997.78m³/s for 2 years return period to 37389.68m³/s for 200years return period.  For Log Normal distribution; 18620.87m³/s for 2 years return period to 32656.49m³/s for 200 years were estimated. Normal distribution; 19051m³/s for 2 years return period to 29367m³/s for 200 years. Log Pearson Type III predicted discharge values ranging between 9081m³/s for 2 years return period to 28732m³/s for 200 years return period, and Pearson Type III predicted discharges were within the range of 1996.95m³/s for 2 years return period to 24415.53m³/s for 200 years return period. The models were assessed using Mean Absolute Deviation Index (MADI), Relative Root Mean Square Error (RRMSE), and Probability Plot Correlation Coefficient (PPCC). Scaling and ranging method was used to arrange the result from the comparative method used to model the five different probability distributions.