1	Modeling & Interpretation of Cryptosporidium Inactivation with Ozone Issam Najm Water Quality & Treatment Solutions, Inc. Jeff Rosen & Jose Sobrinho Technology Planning & Management Corp. Steve Via American Water Works Association
2	"Basic Approach – From bench..." Basic Approach – From bench testing to CT Table Approach used by EPA AWWA’s suggestions Impact on the viability of ozone as a primary disinfectant
3	"Bench Testing" Bench Testing Data Analysis & Modeling Development of a CT Table
4	"Start with a water sample..." Start with a water sample containing cryptosporidium oocysts at a concentration “N_o”. Expose the water to multiple CT values. For each CT, measure the concentration of the surviving cryptosporidium oocysts “N”.
5	"For each CT" For each CT, calculate Log(N/N_o), Plot Log(N/N_o) versus CT
6	"Determine K₁₀ at different temperatures" Determine K₁₀ at different temperatures Plot Ln(K₁₀) vs Temp.
7	"Knowing the K₁₀ values at..." Knowing the K₁₀ values at different temperatures, then for each temperature: Log Kill = K₁₀ x (CT)
8	"Databases Used" Databases Used Modeling Approach Uncertainty Factor
9	"Finch Database (Li et..." Finch Database (Li et al., 2000) Mariñas Database (Rennecker et al., 1999) MWH Database (Oppenheimer, et al., 2000) EPA Database (Owens et al., 2000)
10
11
12
13	"Exclude data with Kill <..." Exclude data with Kill < 0.5 logs Because of high uncertainty in that range Because the CT tables do not go below 0.5 logs Utilize K₁₀ values from individual experiments, not individual data points
14
15
16
17
18
19
20	"Due" Due, in part, to the perceived high degree of variability in the data, EPA decided that an uncertainty factor should be applied to the K₁₀ value The uncrtainty factor was determined by defining the 90% confidence envelope around the data [Ln(K₁₀) vs Temp], and selecting the lower 90^th %ile K₁₀ value for developing the CT table.
21
22	"Variability in K₁₀ value due..." Variability in K₁₀ value due to: Varibility in batches of oocysts used Unknown impacts, if any, of water matrix Experimental error Analytical error EPA’s confidence region envelopes all these errors We need to separate these errors and account only for the variability in K₁₀ value
23
24
25
26
27
28
29	"Use a single K value..." Use a single K value for each experiment, not for each datapoint Exclude datapoints below 0.5 Log inactivation The regression line is the best estimate of the true K₁₀ value… it is not just the “average” The variability in the data is largely due to analytical and experimental inaccuracies, not uncertainty in the true K₁₀ value If a confidence region were to be used for an uncertainty factor, it should be defined around the K₁₀ value, not around the experimental datapoints
30