The stage-discharge relationship or rating curve at a river cross-section is a fondamental technique in hydrology employed for determining discharge from catchments. While river cross-sections are inherently variable, it generally takes quite some time for a change in cross-section to occur. Thus, where good quality data on stage and discharge are available, a stable and representative rating curve may be established. However, all too often, the natural variability is aggravated by data errors of human origin. These measurement errors occur because the routine of periodic flow measurement may be undertaken by illtrained personnel. Such errors resulting from this practice may lead to erroneous estimates of available water resources. This study discusses the derivation of a single rating curve for the Okavango River at Mohembo from measurements containing errors and outliers. It uses data transformation procedures, regression techniques and removal of outliers to derive a useful rating curve from suspect data. Finally, it Investigates the possibility of fitting a stochastic model to the discharge series. The results suggest that it may be possible to estimate missing data with an AR1MA model.