A new Maximum Likelihood Criterion (MLE) suitable for data which are recorded at unequal time intervals and contain auto-correlated errors is developed. Validation of the new MLE criterion has been carried out both on a simple two - parameter reservoir model using synthetical data and on a more complicated hillslope model using real data from the Pukeiti Catchment in New Zealand. Comparison between the new MLE criterion and the Simple Least Squares (SLS) criterion reveals the superiority of the former over the latter. Comparison made between the new MLE and the MLE for auto-correlated case proposed by Sorooshian in 1978 has shown that both criteria would yield results with no practical difference if equal time interval data were used. However, the new MLE can work on variable time interval data which provide more information than equal time interval data, and therefore produces better visual results in hydrologic simulations.

Demands on the water resources of South Africa are ever increasing owing to population growth and increased development of urban, peri-urban and rural communities. Problems in terms of water quantity and quality are likely to be experienced during baseflow recessions. It is therefore imperative that water resources managers not only understand these baseflow periods of streamflow, but are able to model them with confidence. Research for this study thus included a comprehensive literature survey of the factors which affect baseflow as well as the approaches that previous studies have utilised to analyse and model baseflow recession. The primary aims of this study were to establish a streamflow database, to construct master recession curves (MRCs) for each catchment under consideration, evaluate the assumption that South African rivers recede exponentially, to determine a representative set of catchment characteristics for use in the baseflow recession analysis, to attempt to explain the MRC trends using these catchment characteristics and to investigate the feasibility of establishing a rule based model for baseflow recession. A streamflow database for South Africa was therefore established. This consisted initially of 202 catchments which were deemed to be recording natural streamflow. MRCs were established for 134 of these catchments. Those MRCs which were established indicate that the majority of South African rivers do not conform to an exponential model of recession. In order to account for the trends defined by the MRCs, catchment area, average catchment slope, drainage density, mean annual precipitation, rainfall concentration, rainfall seasonality, two independent estimates of groundwater recharge and a geological index were calculated for each catchment. Limited success was achieved when the data set was divided into subsets in order to group catchments with similar baseflow recession responses. The geological composition of the catchments appeared to provide the best results in that those trends exhibited by the MRCs could be explained by the types and proportions of the lithologies present. Owing to the lack of readily useable results it was concluded that until further results were forthcoming the development of a rule based model for baseflow recession analysis in South Africa would be premature. The establishment of a readily accessible database containing streamflows and associated catchment characteristics lends itself to future research. / Thesis (M.Sc.)-University of Natal, Pietermaritzburg, 1997.