Modelling highly variable daily maximum water temperatures in a perennial South African river system

Abstract

Thirty-three months of observed hourly water temperatures were used to calculate daily maximum water temperatures for nine sites within the Sabie-Sand River system, Mpumalanga Province, South Africa. A suite of statistical models for simulating daily maximum water temperatures, of differing complexity and using inputs of air temperature, flow rates, rainfall and relative humidity, were developed and verified. Whilst all models performed well, the most suitable was a site-specific multiple linear regression model using inputs of mean daily air temperature, minimum daily air temperature and relative humidity. The inclusion of a flow rate term would greatly enhance the utility value of such models, but insufficient flow rate data is often a limiting factor. For pragmatic purposes, a simple non- inear regression model using mean daily air temperatures is probably adequate for many areas of South Africa. A generic statistical water temperature model at a daily time step is difficult to achieve, whereas catchment- or site-specific, models, were found to be more appropriate.

African Journal of Aquatic Science 2005, 30(1): 55–63