FLOOD ANALYSIS IN LANGAT RIVER BASIN USING STOCHATIC MODEL

Abstract

This study analyzed the annual maximum stage readings of three rivers in Langat River Basin for
flood forecasting using Autoregressive Integrated Moving-average(ARIMA) model. Model identification was
done by visual inspection on the Autocorrelation Function(ACF) and Partial Autocorrelation Function(PACF).
The model parameters were computed using the Maximum Likelihood (ML) method. In model verification, the
chosen criterion for model parsimony was the Akaike Information Criteria Corrected(AICC) and the diagnostic
checks include residuals’ independence, homoscedasticity and normal distribution. The best ARIMA models for
the Dengkil, Kg. Lui and Kg. Rinching series were (1,1,0), (1,1,0) and (1,1,1) respectively, with their AICC
values of 133.736, 55.348 and 42.292. Homoscedasticity was confirmed with the Breusch-Pagan test giving pvalues of 0.145, 0.195 and 0.747 for the Dengkil, Kg. Lui and Kg. Rinching models respectively. Forecast series
up to a lead time of eight years were generated using the accepted ARIMA models. Model accuracy was checked
by comparing the synthetic series with the original series. Results show that the ARIMA models for the rivers
and the forecast series were adequate. In conclusion, the Box-Jenkins approach to ARIMA modelling was found
to be appropriate and adequate for the rivers. The flood forecast up to a lead time of eight years for the three
models exhibit a straight line with near constant streamflow values showing that the forecast values were similar
to the last recorded observation.