Evaluation of using local information from donor catchments to improve the performance of selected deterministic and empirical design flood estimation methods in South Africa

There are a number of design flood estimation methods routinely used in South Africa. Flood Frequency Analysis (FFA) remains the preferred technique in instances where adequate records of observed data are available. However, in many parts of South Africa, rivers are not gauged for continuous streamflow monitoring. In the case of ungauged catchments, hydrologists and engineers generally make use of methods based on deterministic and empirical approaches for design flood estimation. Due to the limitations and performance of these methods, improved approaches need to be developed for design flood estimation in ungauged catchments. International practice has shown that making use of local information transfer from nearby gauged catchments, also referred to as donor catchments, can improve flood estimation in ungauged catchments. The main considerations for implementing such methods are the type of information transferred and selection of suitable donor catchments. Approaches for donor catchment selection include physical similarity, spatial proximity and integrated similarity. Some studies have also shown that the use of multiple donor catchments can offer further improvements. Thus, a methodology is proposed and evaluated in this paper for using local information transfer in the form of a method and location-specific adjustment factor from gauged donor catchments to improve the performance of selected deterministic and empirical flood estimation methods widely used in South Africa. The pilot study was undertaken at 48 catchments located in the north-eastern part of South Africa, with river flow data provided by the Department of Water and Sanitation. The results show that the degree of improvement for each of the methods is largely dependent on the approach used for donor catchment selection. Even though the best approach for donor catchment selection varied for the different design flood estimation methods, the integrated similarity approach performed consistently well for all methods considered. Similar to the use of single donor catchments, the results obtained from the use of multiple nearest donor catchments also varied for each of the methods used. Due to the general improvements, and the promising results and success of the simple approach used in this study, it is recommended that further refinements of the proposed methodology and approaches to donor catchment selection be considered for future research projects in order for the method to be used by engineers and hydrologists in practice.