A comparison of predictive models for the combined effect of UV dose and solids concentration on disinfection kinetics of Escherichia Coli for potable water production

Copyright © 2001 The Institution of Chemical Engineers Published by Elsevier B.V.A present lack of a rigorous quantitative understanding of the kinetics of ultraviolet (UV) inactivation of bacterial contaminants is shown to limit wide application and optimization of UV disinfection efficacy in the provision of potable water and treatment of wastewaters. Four appropriate mathematical model forms were synthesized from extensive published data for UV disinfection of Escherichia coli in reverse osmosis (RO) water with the addition of either a shielding agent or an absorbing agent and evaluated in a comparative study. Celite 503™ (with a median particle size of 23 μm) in suspended solids concentrations in the range 0.0 to 0.3gl-1 was the UV shielding agent, and powdered coffee (International Roast™) concentrations in the range 0.0 to 0.03 gl-1 was the UV absorbing agent. The models were the classical log-linear, Davey linear-Arrhenius, Ratkowsky-Belehradek or square-root and a third order polynomial model (nOP). The test criteria for model selection and ranking included goodness of fit and accuracy in prediction; relative complexity (i.e., parsimony); ease of synthesis and use; and potential for physiological interpretation. The percent variance accounted for (%V) in prediction was used as a stringent test of goodness of fit. The Davey linear-Arrhenius model explained an overall mean of 97.2%V. This compared with, respectively, 95.6, 93.0 and 86.7%V for the nOP, log-linear and square-root models. The poorer fit of the square-root model suggests that suspended solids concentration and UV dose act independently on disinfection kinetics as implied in the Davey linear-Arrhenius model. This model is quadratic in UV dose and linear in suspended solids concentration. It has three terms: [dose], [dose]2 and [conc]. It best explained deviation in tailing from classically implied log-linear survivor kinetics for the rate of UV disinfection of viable bacterial cells. Overall it best fulfilled the criteria for model selection. Further, from a practical view, it is of a form that is readily integrated with equations describing the rheology and hydrodynamics of liquid flow that permit a simulation of a UV disinfection unit operation for the provision of potable water. © Institution of Chemical Engineers.S.A. Amos, K.R. Davey and C.J. Thoma