A Java-Based, Object-Oriented Modeling System for Southern African Hydrology

Abstract

An object-oriented modeling system, ACRU2000 (Agricultural Catchments Research Unit 2000), was created to enable users to simulate the hydrological and environmental effects of various land uses in the Republic of South Africa. The ACRU2000 modeling system was entirely re-designed and re-coded from the concepts used in the previously developed ACRU agro-hydrological model. The model was designed in the Unified Modeling Language (UML) and implemented in the Java programming language. The ACRU2000 model was designed to process distributed catchments in a parallel manner during each simulation time step to facilitate the modeling of artificial water flows, particularly with regard to the implementation of the South African National Water Act of 1998. The object-oriented design methodology used to restructure the ACRU model produced a more flexible and extensible model structure, which will facilitate future model expansion. The ACRU2000 modeling system was then tested in the Mgeni River watershed, a medium-sized (4387 km2) watershed in the KwaZulu-Natal province of South Africa. The ACRU2000 model was parameterized for 137 sub-catchments to simulate daily streamflows. Monthly totals of simulated flows were compared with observed data from several management catchments for a 15-year period (1979-1993). The model gave satisfactory results in simulating upstream management catchments with the variable high and low flows that are endemic to southern African hydrology. The model had less success in areas encompassing large reservoir and transfer schemes combined with urban and peri-urban areas. Coefficient of determination, model efficiency, and root mean squared error of monthly totals of daily streamflow ranged from 0.43 to 0.85, -4.17 to 0.81, and 4% to 51% of median observed streamflow for individual management catchments, respectively. Further study is necessary to determine whether the differences between model and observed results were caused by model inadequacies, poor estimates of crucial input parameters, or faulty/missing flow data.