ABSTRACT Proving UV reactor performance through validation is becoming a common requirement in wastewater, reuse water and drinking water disinfection applications. However, there is often confusion in understanding the objectives of validation and when choosing an appropriate protocol to follow. This paper will visit the fundamental rationale behind validation. The primary principle behind performance validation is to ensure that public and environmental health is being safeguarded. To do this, regulations must set risk-based disinfection targets, and reactors must be shown to have adequate performance in terms of those targets. Validation must be based on empirical results to eliminate assumptions that are unsafe. Validation must be universal for a given reactor, so that it can be applied to any site where the reactor may be installed. Protocols must not be prescriptive with consequential hindrance to innovation, and they must not be too complex so that they can be accepted and implemented by the industry. This paper will expand on these major points, showing examples of how validation protocols can violate these principles, and also showing alternatives that uphold the principles, ensuring that public and environmental health is safeguarded.
Journal of Water and Environment Technology, Vol.3, No.1, 2005 - 85 - Safeguarding Public and Environmental Health: What are the Necessary Requirements of UV Reactor Validation Protocols? B. PETRI, W. CAIRNS, L. GOWMAN, T. MAO Trojan Technologies Inc., 3020 Gore Rd., London, N5V 4T7, CANADA (E-mail: BPetri@TrojanUV.com) ABSTRACT Proving UV reactor performance through validation is becoming a common requirement in wastewater, reuse water and drinking water disinfection applications. However, there is often confusion in understanding the objectives of validation and when choosing an appropriate protocol to follow. This paper will visit the fundamental rationale behind validation. The primary principle behind performance validation is to ensure that public and environmental health is being safeguarded. To do this, regulations must set risk-based disinfection targets, and reactors must be shown to have adequate performance in terms of those targets. Validation must be based on empirical results to eliminate assumptions that are unsafe. Validation must be universal for a given reactor, so that it can be applied to any site where the reactor may be installed. Protocols must not be prescriptive with consequential hindrance to innovation, and they must not be too complex so that they can be accepted and implemented by the industry. This paper will expand on these major points, showing examples of how validation protocols can violate these principles, and also showing alternatives that uphold the principles, ensuring that public and environmental health is safeguarded. KEYWORDS UV reactor, validation protocol OR UV reactor validation, protocol requirements OR UV reactor, validation protocol requirements INTRODUCTION Ultraviolet light (UV) disinfection is a process that is being applied increasingly for treating wastewater, reuse water and drinking water. UV is an economical choice as a disinfectant, and has proven to be very effective against chlorine-resistant pathogens such as Cryptosporidium and Giardia. The use of UV can lead to the minimization or eventual elimination of chlorine, and can thus help to lower levels of associated disinfection by-products (DBP). In North America, UV has been used for decades for disinfecting wastewater, but its use in drinking water and reuse water disinfection is relatively new and growing rapidly. Bioassay validation of UV reactor performance has been embraced as a method of proving UV reactor performance in these new applications, and a number of North American test protocols have arisen to give guidance on testing for these new applications. The United States National Water Research Institute (NWRI) and American Water Works Association Research Foundation (AWWARF) released the Ultraviolet Disinfection Guidelines for Drinking Water and Water Reuse in December, 2000 (NWRI/AWWARF, 2000, 2003), and the United States Environmental Protection Agency (USEPA) released the draft UV Disinfection Guidance Manual (UVDGM) in June 2003 (USEPA, 2003). Bioassay validation is accomplished by empirically measuring the reduction in numbers of microorganisms across a UV reactor, and linking this performance to operating conditions that generated the results and to the monitoring results that provide an online assessment of performance (Petri and Henry, 2004). The challenge microorganisms are also submitted to inactivation tests in a Journal of Water and Environment Technology, Vol.3, No.1, 2005 - 86 - laboratory-based collimated beam apparatus in order to develop a dose-response calibration relationship. Using this relationship, the inactivation across the reactor can be expressed as a collimated beam equivalent dose or a reduction equivalent dose (RED). In wastewater, high levels of indigenous microbes allow for the direct testing of disinfection efficacy with performance linked directly to disinfection targets. In reuse and drinking water applications, fewer indigenous microorganisms are present to monitor performance, and microbial challenges are performed by injecting challenge organisms upstream of the reactor and measuring reduction in numbers after the UV. Historically, challenges have been done with UV-resistant microorganisms (MS2 bacteriophage, Bacilus subtilis) as surrogates to link to UV-resistant virus targets. However, when microorganisms with different UV-sensitivities are the disinfection targets, it becomes difficult to correlate performance. Every UV reactor is non-ideal. Due to the complexity of hydraulic behavior and light intensity profiles through reactors, different tracks of microbes through the reactor will result in different doses for each track. From a number of tracks through any reactor, a UV dose distribution can be constructed. Due to the dose distribution, microorganisms with different sensitivities will exhibit different results: less dose than expected will be seen for UV-sensitive microorganisms than for UV-resistant microorganisms (Wright and Lawryshyn, 2000), and these differences will be amplified for reactors with wider dose distributions (Petri and Olson, 2003). Thus, the dose distribution of a reactor is fundamental to its performance, and is key to designing a proper validation protocol that ensures public and environmental safety with properly chosen regulatory targets. This paper will visit the fundamental rationale behind validation. The primary principle behind performance validation is to ensure that public and environmental health is being safeguarded. To do this, regulations must set risk-based disinfection targets, and reactors must be shown to have adequate performance in terms of those targets. Validation must be based on empirical results to eliminate assumptions that are unsafe. Validation must be universal for a given reactor, so that the validation can be applied to any site where the reactor may be installed. Protocols must not be prescriptive to avoid hindering innovation, and protocols must not be too complex so that they can be accepted and implemented by the industry. This paper will expand on these necessary aspects of validation protocols that focus on the non-ideal characteristic inherent in every real UV reactor. ADDRESSING THE BIAS BETWEEN CHALLENGE AND TARGET MICROBES Due to dose distributions in every UV reactor, there can be a discrepancy (bias) between the UV dose measured by the challenge microbes and the UV dose that would be measured had the target microbes been used as the challenge microbes. This important fact has been recognized and addressed in the USEPA draft UVDGM, and quantified by a safety factor termed the RED Bias. Although the principle is sound, the actual USEPA calculation of the RED Bias is based on a single default dose distribution that is assumed to be worst case. This assumption cannot be verified to be safe: dose distributions vary with the change in each operating variable for a UV system, and many calculated dose distributions for real UV reactors can be shown to be worse than the default dose distribution. This can translate into under-designed systems not delivering the disinfection they are thought to be delivering. Conversely, reactors producing dose distributions better than the default are unfairly penalized with an overly large safety factor, removing the incentive for developing more efficient reactors and placing an unjustified Journal of Water and Environment Technology, Vol.3, No.1, 2005 - 87 - expense (in terms of over sizing and operation with more power than necessary) on the owner and users of the UV system. There are several ways to conduct a validation of reactor performance depending on whether the challenge organism has the same or different sensitivity to UV as the target pathogen. 1. When challenge microbes have the same UV-sensitivity as the target (regulated) microbes, the effect of the dose distribution would be the same on both microbes, and measured performance can be directly related to performance in terms of the target. Thus, the best validation protocol would require the use of challenge microbes that have the same UV-sensitivity as target microbes. The RED Bias would be necessarily eliminated, and UV systems would be designed and operated safely based on their actual measured performance. When the challenge organism is not the same as the target organism and does not have the same UV sensitivity, there are several ways to assess reactor performance for the target organism. 2. In the absence of challenge microbes that match the UV-sensitivity of target microbes, an approach that ensures public and environmental health is to give performance credit that is numerically equal to the log reductions achieved with microbes more-resistant than target microbes, rather than in terms of RED. Performance expressed in terms of log reductions of microbes of given UV-resistance eliminates the issue of a dose distribution. Typically regulatory targets for drinking water and reuse water are expressed in terms of log reductions of certain microorganisms, and expressing performance in terms of log reductions provides a direct link to the regulatory targets. System designs based on this approach will be inherently safe, with the necessary caveat that the challenge microbe must be more UV-resistant than the target microbe. However, when target and challenge microbes have very different UV-resistances, the degree of over-sizing will be compounded. 3. A more refined approach is to perform challenges with two different microbes that bracket the UV- resistance of the target microbe. For any real UV reactor tested with microbes of varying UV- resistance, a plot of the measured RED versus the UV-resistance (D10, dose required to achieve 1 log reduction) will have an upward concave shape (Figure 1). Reactors with wider dose distributions will produce relationships with greater curvature. Because of the shape of the curve, a performance result for a target microbe of intermediate UV-resistance can be safely interpolated. The example in Figure 1 is based on the calculated dose distribution of a real commercial UV reactor. Based on the calculated dose distribution, testing with two microbes with UV-resistances of 2 and 20 mJ/cm 2 /log will give REDs of 9.6 mJ/cm 2 and 20.0 mJ/cm 2 respectively. Interpolating between the two actual results gives an estimate of 11.3 mJ/cm 2 for the RED of a microbe with a UV-resistance of 5 mJ/cm 2 /log at the same test conditions. The interpolated estimate is conservative because the actual RED achieved at the same operating conditions for the target microbe would be 13.0 mJ/cm 2 . Two organism tests would allow for more efficient design and operation because of the refined performance estimates. 4. A final approach to validation would be to quantify the dose distributions produced by the UV reactor. With known dose distributions, the performance in terms of microbes with any sensitivity can be calculated directly and exactly. a) Direct methods for quantifying dose distributions are being developed and involve challenging reactors with actinometric microspheres (Anderson et al., 2003). The microspheres Journal of Water and Environment Technology, Vol.3, No.1, 2005 - 88 - are collected after the reactor and the individual actinometric responses provide input to the creation of a distribution of doses. b) A broader approach is to validate calculated dose distributions, which can then be used to determine performance in terms of alternative targets. This can be done directly also by using microsphere actinometry to compare measured and calculated results. This can also be accomplished by experimentally altering dose distributions, and determining if the calculated dose distributions are predictive of the experimental results. Because dose distributions change for every change in operating conditions, bioassay performance should be predictable if dose distributions are being calculated accurately. Showing accurate calculations of bioassay performance between-tests validates the accuracy of the dose distribution calculations. Finally, the same principle applies to multi-organism testing. A given dose distribution will result in different measured performance for organisms with different UV-resistance. Showing accurate calculations of bioassay performance between-organisms similarly validates the accuracy of the dose distribution calculations. 0 4 8 12 16 20 24 0 4 8 12162024 UV Resistance (D10, mJ/cm2/log) RED (mJ/cm 2 ) RED Conserv. Interp. Figure 1 Use of a two-organism bioassay to interpolate a conservative estimate of the actual RED for a target organism with intermediate UV-sensitivity relative to the two challenge organisms. Each of these four validation approaches recognizes the impact of the dose distribution, and without making any unsafe assumptions, provides a methodology that is safe with respect to the protection of public and environmental health. Journal of Water and Environment Technology, Vol.3, No.1, 2005 - 89 - ADDRESSING THE BIAS BETWEEN VALIDATION CONDITIONS AND OPERATING CONDITIONS An additional manifestation of the effects of a dose distribution is exhibited by reactors with polychromatic light sources. Polychromatic lamps emit germicidal energy over a range of wavelengths, whereas monochromatic lamps emit germicidal energy only at 254 nm. Sleeves, water, microbes, and sensors all have polychromatic spectra or responses. As a result, dose delivery and dose monitoring can vary between the validation and a specific operating site, due to differences in any of the spectra (e.g., water absorption during validation and at the operating site). The USEPA draft UVDGM recognizes this effect and quantifies it with a safety factor termed the Polychromatic Bias. It is important to note that there are two effects layered into this issue. If spectra differ between the validation site and a given operating site, both dose delivery and dose monitoring will be impacted. If they are impacted to the same degree, there is no bias. However, if dose delivery drops less than the change in monitored dose, or if monitored dose rises more than the rise in dose delivery, unsafe operation can result. A calculator accompanies the USEPA draft UVDGM that allows one to calculate the Polychromatic Bias, based on a single lamp annular reactor with perfect mixing and a single sensor. This simplistic hypothetical reactor and monitoring system could be far from representative of real reactors. A more robust approach to validating polychromatic reactors is to design tests that measure the impacts of polychromatic effects on dose delivery and dose monitoring. Again, this speaks to understanding the dose distributions in real reactors. Testing with microbes with different action spectra (microbial response), with waters with different transmittance modifiers (with different absorption spectra), and with new and aged lamps, will provide the information necessary to evaluate the magnitude of a bias arising from polychromatic effects. It is unlikely that a single safety factor will satisfy the requirements because inputs (or test conditions) are dependent upon water spectra for sites that may not yet even be identified. As with differences in microbe sensitivity, a more robust approach to validation (e.g., 3b above) is to recognize the dose distribution as the root of this bias, and validate the ability to accurately calculate the dose distribution under different polychromatic conditions. Validation of the monitoring tool is equally as important as validation of the reactor performance, especially for applications such as drinking water where the number of indigenous microbes precludes a direct count of microbe survivors. The primary concept in monitor validation is to determine empirically that whatever combination of hardware and software are used to output the delivered dose online, the output dose is in agreement with the empirically determined bioassay validated dose. Any online monitoring tool capable of determining the delivered dose, or that the delivered dose is above the target design dose, is an acceptable monitoring tool. Note however that reactor validation methods that do not directly indicate the reactor performance with respect to the target microbe must be corrected with respect to the target microbe, and when the monitoring tool agrees with the challenge organism dose, it too must be corrected with respect to the target organism dose. ADDRESSING TESTING UNCERTAINTY Uncertainty in testing should be translated into a reduction in stated performance. The NWRI/AWWARF guidelines do this, calling for performance to be reported as a lower 75 th confidence level of microbiological results for reuse applications and as a lower 90 th confidence level of microbiological results for drinking water applications. The differences in confidence levels reflect the Journal of Water and Environment Technology, Vol.3, No.1, 2005 - 90 - greater need for risk mitigation in the drinking water applications. The USEPA draft UVDGM quantifies uncertainty of not just performance but of operating variables and monitoring equipment, and combines them into a safety factor termed the Expanded Uncertainty. However, this approach fails to recognize the differences between accuracy (how close a measured value is to a true value) and precision (how repeatable a measured value is). Accuracy is critical for measurements of operating variables, to ensure that performance is quantified correctly. Validation protocols can account for accuracy by specifying minimum acceptable levels of accuracy for monitoring equipment (flow meters, UVT meters, power meters, sensors), and ensuring the equipment meets those levels through calibration checks. Precision can be quantified by repeated measures of the same parameter, but is probably swamped out in validation efforts by actual variation in the measured parameter. However, such variability is inherently accounted for in the performance results. For example, if the flow rate has slight variations during testing, the dose delivery will also have slight variations, which will be captured and measured in replicate microbiological samples. Thus, the uncertainty of measured operating and monitoring parameters should be addressed by specifying and demonstrating the acceptable accuracy of test equipment. Variability in microbiological samples can be used, as in the NWRI/AWWARF protocol, to discount measured performance to the level of confidence required. There should be some rationalization behind the choice of confidence limits, presumably linked to risk assessments for public and environmental health protection. ADDRESSING HYDRAULIC ISSUES Hydraulics can have an effect on UV reactor performance. This effect is addressed in the German DVGW UV reactor testing protocol by mandating that testing be done following a prescribed 90 degree bend in the feed pipe immediately upstream of the UV reactor. The USEPA draft UVDGM suggests a similar approach. These concepts are based on the premise that the tested conditions are worst case, and that all installation scenarios would result in better performance than during testing. This is an unsafe assumption. Even substituting a more challenging hydraulic setup is not a robust solution, as even more challenging hydraulic applications will no doubt appear. An alternative approach is to use sound engineering tools like computational fluid dynamics to evaluate the impacts of any hydraulic regime relative to the tested regime. The CFD-based models must also be validated, in that the models should be shown to agree with actual bioassay test results. Testing should occur in a simple straight pipe configuration, so that baseline models can accurately quantify the effects through the reactor (Petri and Olson, 2002), rather than being confounded with complex hydraulics. With validated CFD models, impacts of alternative hydraulic configurations can be reasonably evaluated. BRINGING IT ALL TOGETHER: VALIDATION AND SAFETY FACTORS UV reactor validations quantify equipment performance, but contain some inherent and random uncertainty that can be quantified and applied as a safety factor. Inherent uncertainty based on the UV reactor’s dose distribution can be quantified in terms of the USEPA-termed RED Bias and Polychromatic Bias, and random uncertainty can be quantified by the variability in performance (microbiological) results. Journal of Water and Environment Technology, Vol.3, No.1, 2005 - 91 - The USEPA draft UVDGM calculates a Safety Factor (SF) for each test: SF = RED Bias x Polychromatic Bias x (1+Expanded Uncertainty) The test-specific safety factor is then applied against the microbiological target dose, to give a new and higher target dose. This approach is confused logic in that the new target dose is different for each reactor test and target organism when in fact, the target design dose should remain unchanged for any given microbe and target logs of reduction. A more logical approach is to calculate a safety factor for the equipment based on the uncertainty during validation, and apply it against the equipment performance rather than against the target UV dose that is left unchanged. This safety factor, primarily from variability in the microbiological performance results, should be applied to de-rate the measured reactor performance. The approach is logically consistent because it applies safety where there is uncertainty (reactor performance and not the dose for microbe inactivation). The performance of the reactor should be determined in one of the four ways listed above, eliminating the USEPA-termed RED Bias. For every site-specific installation, two additional safety factors need to be calculated and applied: one to account for polychromatic effects, and one to account for hydraulic effects. Both of these can only be calculated when the particulars of the specific site are known. The validation results, de-rated for microbiological variability, can then be equated to performance in the site-specific installation: Design Performance = De-Rated Validation Performance x Hydraulic SF x Polychromatic SF The approach suggested in this paper provides a sound basis for validation. The approach a) does not incorporate any unsafe assumptions, b) retains portability of the validation results from one specific site to another, and c) provides for the a logical manner in which to apply the site-specific safety factors, once they become known, to the validation results established under simplified standard conditions that have minimal hydraulic complexity upstream or downstream of the reactor being validated. REFERENCES Anderson W.A., Zhang L., Andrews S.A. and Bolton J.R. (2003). A Technique for Direct Measurement of UV Fluence Distribution. In: Proceedings of the Water Quality Technology Conference, Philadelphia, PA, Nov. 2-6, 2003. DVGW. (1997, 2003). UV Disinfection devices for drinking water supply --- Requirements and testing. German Gas and Water Management Union (DVGW), Bonn, Germany. NWRI/AWWARF. (2000, 2003). Ultraviolet disinfection guidelines for drinking water and water reuse. National Water Research Institute and AWWARF. Petri, B., and Henry, A.J. (2004). Full Scale UV Reactor Bioassays at the Lake Huron Water Supply System. In: Proceedings of the Second Asia Conference on Ultraviolet Technologies for Environmental Applications, Tokyo, Japan, Oct. 4-6, 2004. Journal of Water and Environment Technology, Vol.3, No.1, 2005 - 92 - Petri, B. and Olson, D.A. (2002). Bioassay Validation of Computational Disinfection Models Used for UV Reactor Design and Scale-up. In: Proceedings of the WEF Disinfection 2002 Specialty Conference, St. Petersburg, FL, Feb. 17-20, 2002. Petri, B. and Olson, D.A. (2003). Critical Approach to Spectral Issues During the Validation and Operation of Polychromatic UV Disinfection Reactors. In: Proceedings of the Second International Congress on Ultraviolet Technologies, Vienna, Austria, July 9-11, 2003. USEPA (2003). EPA 815-D-03-007 Ultraviolet Disinfection Guidance Manual, June 2003 Draft, USEPA, Office of Water. Wright, H.B., and Lawryshyn, Y.A. (2000). An Assessment of the Bioassay Concept for UV Reactor Validation. In: Proceedings of Disinfection 2000: Disinfection of Wastes in the New Millennium, New Orleans, LA, March 2000. . Technologies, Vienna, Austria, July 9-11, 20 03. USEPA (20 03) . EPA 815-D- 03- 007 Ultraviolet Disinfection Guidance Manual, June 20 03 Draft, USEPA, Office of Water. Wright,. 20 03) , and the United States Environmental Protection Agency (USEPA) released the draft UV Disinfection Guidance Manual (UVDGM) in June 20 03 (USEPA, 20 03) .