Ultraviolet disinfection kinetics for potable water production.

Irradiation with ultraviolet (UV) light is used for the disinfection of bacterial contaminants in the production of potable water, and in the treatment of selected wastewaters. However, efficacy of UV disinfection is limited by the combined effect of suspended solids concentration and UV absorbance. Limited published UV disinfection data are available that account for the combined effects of UV dose, suspended solids concentration and UV absorbance. This present lack of a rigorous quantitative understanding of the kinetics of UV disinfection limits process optimisation and wider application of UV treatment. The development and validation of an adequate model to describe UV disinfection kinetics presented in this thesis can therefore be justified by an increased confidence of reliability of design for UV disinfection. Using the published data of Nguyen (1999), four established model forms were assessed to account for the combined effect of suspended solids and/or soluble UV absorbing compounds, and UV dose on the efficacy of disinfection. The four model forms were: a log-linear form, Davey Linear-Arrhenius (DL-A), Square-Root (or Ratkowsky- Belehradek) and a general nth order Polynomial (nOP) form that was limited to a third order. Criteria for assessment of an adequate predictive model were established including: accuracy of predicted against observed values, percent variance accounted for (%V), and; appraisal of residuals. The DL-A model was shown to best fit the data for UV disinfection of Escherichia coli (ATCC 25922); followed by the nOP, log-linear and Square-Root forms. However, the DL-A form must be used in conjunction with a first-order chemical reaction equation, and was shown to predict poorly at high experimental values of UV dose (> 40,000 μWs cm-2). The DL-A model was not amenable to extrapolation beyond the observed UV dose range. To overcome the shortcomings of the Davey Linear-Arrhenius model synthesis of two new, non-linear model forms was undertaken. The two models were a modified exponentially damped polynomial (EDPm) and a form based on the Weibull probability distribution. The EDPm model has three terms: a rate coefficient (k), a damping coefficient (λ), and; a breakpoint dose ([dose]B). The rate coefficient governs the initial rate of disinfection prior to the onset of tailing, whilst the breakpoint is the UV dose that indicates the onset of tailing. The damping coefficient controls curvature in the survivor curve. The Weibull model has just two terms: a dimensionless scale parameter (β0), and; a shape parameter (β1). The scale parameter represents the level of disinfection in the tail of the survivor curve (as log10 N/N0), whilst the shape parameter governs the degree of curvature of the survivor data. Each model was assessed against the independent and published UV disinfection data of Nelson (2000) for treatment of faecal coliforms in a range of waste stabilisation pond effluents. Both models were found to be well suited to account for tailing in these UV disinfection data. Overall, the EDPm model gave a better fit to the data than the Weibull model form. To rigorously validate the suitability of the new EDPm and Weibull models a series of experimental trials were designed and carried out in a small-scale pilot UV disinfection unit. These trials included data determined specifically at low values of UV dose (