ABSTRACT
Monthly water levels for four years in Lokko river were studied statistically using the time series theory, and an ARMA (1.1) model was identified as suitable to describe the data. An estimation model was developed, the results obtained were found to be satisfactory, except in the rainy season when the error could be as high as 37 percent on the average, and as high as 50 percent for a single event, due to random storm rainfall input to the streamflow which the model is unable to account for fully. A factor of safety of 1.6 is therefore recommended for the multiplication of the estimated levels for early warning and for engineering works a factor of safety of 2.2 is recommended.