Minimizing Collimated Beam Uncertainty
Shawn Verhoeven, Conrad Odegaard and Shawn Hess, GAP EnviroMicrobial Services Ltd.
1020 Hargrieve Rd., Unit 14, London, ON, Canada
Collimated beam-generated standard curves are an important step in determining the fluence applied by an ultraviolet (UV) disinfection system whereby a bioassay-measured microbe log inactivation is translated to a reduction equivalent UV fluence (RED). This RED then is used to properly size a system to meet disinfection requirements for a target pathogen or pathogen of interest at an installation site. Proper calculation of this fluence (UV dose) when performing a collimated beam tests is thus extremely important for making decisions on UV reactor sizing; wherein there lies a problem. Without proper quality control, each of the factors used in the collimated beam fluence calculation equation introduces error. Minimizing the error in each of these factors is thus pivotal to producing an accurate fluence response curve for proper UV system sizing. Additionally, there are three equations for the calculation of collimated beam fluence that exist in North American guidance. This adds to the confusion as to how to properly perform a collimated beam test.
Keywords: Collimated beam, petri factor, UVDGM, NWRI, NSF, reflection, fluence, UV dose
Regulations around the world require UV system manufacturers to validate the performance of their UV reactors. Various guidance manuals and validation protocols exist from groups including in the US Environmental Protection Agency with their Ultraviolet Disinfection Guidance Manual for the Final Long Term 2 Enhanced Surface Water Treatment Rule (UVDGM), the German Technical and Scientific Association for Gas and Water (DVGW), the National Water Research Institute’s (NWRI) Ultraviolet Disinfection Guidelines and the NSF International Standard/American National Standard for Drinking Water Treatment Units – Ultraviolet microbiological water treatment systems (NSF 55). All of these documents outline protocols for the validation of drinking water systems; some also cover water reuse and wastewater systems. These protocols use the [log] inactivation of non-pathogenic surrogate microorganism(s), measured under a defined set of conditions, to characterize the system performance. Alone, this log inactivation does not provide a great deal of information regarding performance, as these surrogates generally do not display identical UV sensitivity to the pathogens targeted in a particular installation. In order to relate the performance of the UV system to inactivation of a target pathogen, the measured log inactivation of the surrogate microorganism is converted to a fluence, or UV dose, by “back calculating” using a collimated beam generated fluence response standard curve to give a reduction equivalent fluence (RED). The collimated beam procedure is thus an intimately important process in properly validating a UV system.
A collimated beam apparatus, more accurately called a quasi-collimated beam apparatus, typically contains one or more low-pressure mercury lamps mounted horizontally within an enclosure. The bottom side of the enclosure has a circular opening, or aperture, that allows UV light to be transmitted to a suspension of microbes located below the opening. The angle of incidence of the UV light on the suspension of microbes is minimized using a series of apertures or collimating tube located below the lamps. Examples of both of these configurations can be found in the paper titled “Standardization of Methods for Fluence (UV Dose) Determination in Bench-Scale UV Experiments” (Bolton et al. 2003). The fluence delivered to the suspension is calculated using an equation that takes into account a number of measurable factors, including the incident irradiance, uniformity of the irradiance field, the ultraviolet transmittance of the sample, reflection off the sample surface and divergence of the beam upon entering the sample.
The protocol or guidance chosen for validating a UV reactor will impact the equation that is selected for calculating UV fluence in the collimated beam test. All fluence calculations are based on the Beer-Lambert Law, where the fluence is defined as the average irradiance through a water layer multiplied by the exposure time in seconds. This takes into account the incident irradiance, the UV transmittance of the water sample and the exposure time, with each protocol implementing the definition in different ways. The implementation of this definition can be found in equations ,  and  below.
In the UVDGM approach outlined in equation , D is the fluence (or UV dose), I0 is the incident irradiance (in mW/cm2) as measured using a radiometer at a location corresponding to the center of the surface of the sample, Pf is termed the Petri factor and defined as the ratio of the irradiance measured at the center of the sample surface to the average irradiance measured across the sample surface, (1 – R) is termed the reflection factor where R is the reflection coefficient of UV light at 253.7 nm at the air-surface interface (typically R = 0.025), L/(d + L) is termed the divergence factor where L is the distance from the lamp centerline to the sample surface and d is the sample depth, and (1 – 10-ad)/(adln(10)) is termed the absorbance factor where a is the UV absorption coefficient (absorbance in 1 cm) of the suspension at 253.7 nm. Time is represented by t in seconds (UVDGM 2006).
The NSF implementation in Equation  uses some of the same factors as in Equation  but assumes the reflection is only 2% (resulting in a 0.98 factor) and does not take into account a divergence factor or petri factor. The NWRI implementation in Equation , similar to Equation , also does not take into account the divergence or petri factors, and uses an absorbance coefficient k, which is equivalent to 2.3a (NWRI 2012); but through the rearrangement of the equation, there is a high level of similarity to the UVDGM equation .
Inherent in the number of factors taken into account in these three fluence calculations lies the possibility that a large amount of error can be introduced with imprecise or incorrect measurement of one, or a number of the factors. The UVDGM outlines the suggested levels of uncertainty for all factors used in Equation , but there is opportunity to reduce the uncertainty and increase measurement accuracy even further than is suggested in this guideline. Other sources of error also exist that are not factored into any of these equations that may lead to miscalculation of the UV fluence. These factors include reflected light off the reaction vessel walls to increase the incident light at the periphery. A parallax effect also may be introduced when trying to match the height of a sample surface to the calibration plane of a radiometer. It is not necessary to include these factors in the fluence calculation but rather, every effort should be made to eliminate them altogether such that they have no impact on the calculation.
With careful consideration, and sufficient quality control procedures, it is possible to reduce the uncertainty and increase the accuracy of the collimated beam calculation. Doing this allows for more accurate sizing of UV reactors and allows us to carefully scrutinize the historical procedures used for determining the sensitivities of target pathogens.
Which fluence calculation to use?
All collimated beam fluence calculations are based on the Beer-Lambert Law, but unfortunately the equations apply this in a few different ways. Although all of the equations outlined in ,  and  appear to differ greatly, if the correction factors are removed, Equation  and  do provide very similar results, with Equation  also providing similar results in many instances. If conditions are selected to nearly eliminate the effect of the petri and divergence factors, we can see exactly how similar the equations truly are. If one assumes a thin film (1 mm) irradiation with a perfectly uniform irradiance field (petri factor equals 1), in nearly transparent water (99% transmittance at 253.7nm), the NSF Equation  calculates a fluence 0.5% higher and the NWRI Equation  calculates an fluence less than 0.1% higher than the UVDGM Equation . At first glance any one of these iterations of the fluence calculation in a collimated beam setup seems like it would be acceptable, but this assumption deteriorates once more commonly encountered factors are applied to the equations. Table 1 below outlines a few results using a typical petri factor found at GAP of 0.99, a sample depth of 0.5 cm, a distance of 45 cm from the lamp centerline to the sample surface, and an incident irradiance of 150 μW/cm2.
From these numbers it becomes evident that the UVDGM equation is always providing a lower fluence when using the equivalent factors among the equations; this is to say that the UVDGM equation is the most conservative when it comes to calculating the fluence provided by a UV reactor. If the divergence factor and petri factor are added to the NWRI calculation, the resulting fluence is corrected by 0.979, completely eliminating the 2.1% fluence increase seen in Table 1.
For the NSF calculation there isn’t a single factor or set of factors that can be isolated as the reason for the higher calculated fluence we are observing. This is not necessarily an issue though, as the NSF protocol is looking at a defined set of conditions that are identical for all units being certified using this procedure. As a result of these conditions being static for all reactors, the exact form of the equation becomes somewhat irrelevant, as we are simply comparing all reactors to the same pass/fail criteria. With this being said, it would not be appropriate to use this equation for a fluence calculation outside of the NSF protocol, including making decisions on reactor sizing or in a regulatory situation to determine the sensitivity of a pathogen.
Although each of the three fluence calculations used in North American UV validations provide unique results, this does not mean that one is best for all situations. They all have their applications in their given protocols, with the UVDGM calculation providing the most conservative fluence due to the number of correction factors employed. One recommendation is to use the NWRI and NSF calculation only where these standards are stipulated and to use the UVDGM calculation in all other instances. This approach is suggested because of the conservative nature of the fluence calculation presented by the UVDGM, there is an inherent conservatism (or a safety factor) when designing and sizing UV reactors using this equation to err on the side of public safety. The UVDGM fluence calculation also recently has been endorsed by the IUVA Board of Directors as part of the “Protocol for the determination of fluence (UV Dose) using a low-pressure or low-pressure, high-output UV lamp in bench-scale collimated beam ultraviolet experiments” (Bolton et al., 2015a).
Unrelated to the equation chosen to calculate fluence, it is incredibly important to properly implement the equation. If the factors entered into the equation are incorrect, then it does not matter which equation has been selected – an incorrect fluence is still being calculated. Each factor has a level of uncertainty associated with it, and care must be taken to ensure that this uncertainty is minimized. The UVDGM outlines what level of uncertainty is expected for each factor used in Equation  and Bolton et al. (2015a) have provided a step-by-step protocol in an effort to limit uncertainty; however, in many instances it is possible to reduce uncertainty below suggested levels. Steps have been outlined in the following sections as a guideline to improve accuracy and uncertainty.
The incident irradiance as measured by a radiometer is the single greatest source of uncertainty in the collimated beam fluence calculation. The UVDGM suggests that the error associated with the “Average incident irradiance” (I0) in the fluence calculation in Equation  be less than or equal to 8%. As long as a radiometer is calibrated by the manufacturer it should be received with a transfer uncertainty of 6.5% and a NIST uncertainty of 1% for a combined uncertainty of 6.58%. This is well within the recommended 8% (UVDGM 2006), but any error introduced here translates one to one as error in the final fluence calculation. The UVDGM does go on to suggest that the radiometer used for measuring the incident irradiance should be verified using a second radiometer at least at the beginning and end of the collimated beam test, with a third radiometer used under certain circumstances (UVDGM 2006). What is not specified, but should be assumed, is that all of these radiometers are to be calibrated against a NIST standard within the past 12 months (Bolton et al. 2015a). Using this comparison, the radiometers are to read within 5% of one another, otherwise it is suggested that at least one radiometer is out of calibration. The main issue with this statement is that having two radiometers that read more than 5% different does not necessarily mean that one is out of calibration, but what is likely meant by this statement is that one radiometer is reading less accurately.
When the radiometers are compared using the method outlined in the UVDGM, there is no definitive way to determine which radiometer is reading most accurately, but there are some procedures that can be implemented to improve the confidence in the incident irradiance reading, including use of three calibrated radiometers at all times. As long as the difference from the highest reading radiometer to the lowest reading radiometer is less than 5%, then a correction factor should be calculated to adjust the incident irradiance reading used in the fluence calculation to the average reading of all three radiometers. If the highest to lowest reading radiometers do not agree within 5%, then the UVDGM (2006) outlines how to select the radiometers to use to obtain the correction factor which will provide the most conservative result.
Although using the average radiometer reading does help improve confidence in the accuracy of the irradiance readings, simply using this procedure blindly will not provide the best estimate of the true incident irradiance. If all radiometers are from the same manufacturer, calibrated by the same source, at the same time and under the same conditions there may be bias. To ensure random error, it is suggested that a selection of radiometers from different manufacturers, calibrated by different parties at different times throughout the year be used to minimize any potential bias. It is also critical that the calibrations are current, with an interval of one year at maximum.
A second way that the radiometer readings can be verified is by using chemical actinometry. A simple method for performing this check is outlined in a paper titled Determination of the Quantum Yields of the Potassium Ferrioxalate and Potassium Iodide-Iodate Actinometers and a Method for the Calibration of Radiometer Detectors (Bolton et al. 2011). This method determines a correction factor that can be applied to the radiometer to obtain the true incident irradiance on a sample. This method does require special care to be taken, including tight user control of measurements using calibrated instruments, such as a caliper, balance and thermometer, and the use of new reagents. It has been the experience of GAP that use of reagents that have been allowed to sit for an extended period of time does influence the initial A300 and A352 measurements. This in turn appears to have an effect on the final calculated correction factor, unrelated to the photochemistry of the actinometer. Uncertainty is still associated with this procedure due to calibrated instrumentation being used, and it would need to be taken into account in the final collimated beam uncertainty calculation. The advantage of this method is that it does not require expensive annual calibration of radiometers, but it does pose a problem when attempting to calibrate a radiometer over a range of wavelengths, since the quantum yield is wavelength-dependent.
Another concept that has recently been suggested by Li et al. (2011) is the use of a micro fluorescent silica detector to replace the radiometer entirely in the collimated beam procedure. Calibration of this instrument would be achieved using chemical actinometry, as noted above, but an advantage is that continuous measurement of the incident irradiance would be possible. This would allow for more accurate measurement of the fluence applied to the sample over the entire exposure time without the assumption of the linear response currently used where the average of the pre and post exposure radiometer readings are taken and used in the fluence calculation. This technology is still in development and more time is required before it can be fully implemented in collimated beam procedures. This will include comparison to current methods, and round robin testing.
Irrelevant of the way in which accuracy of the radiometers is verified, it is clear the measurement of the incident irradiance is integral in the calculation of fluence. If an error of 5% is introduced at this step, then this directly translates to an error of 5% in the calculated fluence. If only a single factor should be more closely monitored and controlled during in the collimated beam, this should be the target for improvement.
The petri factor is another collimated beam correction factor that shows a one to one correlation in affecting the final calculated fluence. Fortunately, the chances of seeing a high magnitude of error in the petri factor are small compared to the incident irradiance measurement using a radiometer. To calculate the petri factor, repeated measurements are made on two perpendicular axes with these measurements used to estimate the intensity across the entire suspension surface. This assumption has been confirmed to be correct by measurement of the petri factor in this “X-Y” manner in comparison to radiometer measurements taken in a grid format over the entire petri dish area. Although the error here may be minimal, there is still opportunity to reduce any error that may exist. The UVDGM states that the error for this factor should be less than 5% (UVDGM 2006). At GAP we have calculated the error in our setup to be 1% using repeated calculations of the petri factor over time using the same L measurement (Equation ) and petri dish diameter. There are factors not accounted for using this method of error calculation though, and all portions of the calculation are looked at individually.
The two most obvious sources of error would be to ensure the spacing of measurements on the axes be accurate; this can be accomplished using a calibrated ruler. Secondly, it is important to ensure that the axes are perpendicular. This can be accomplished using the 3, 4, 5 technique commonly used in construction and based on the Pythagorean theorem, where the square of the side measurements of a right triangle equal the square of the hypotenuse. Again, the calibrated ruler would be important for this calculation.
Another way to minimize error in the petri factor calculation is to reduce the viewing window of the radiometer detector using a mask if the viewing window is greater than the distance between axes measurements, typically 5 mm (Bolton et al. 2003). This will reduce the magnitude of the readings measured by the radiometer, but this not a concern, as the petri factor compares readings relative to the origin, not absolute readings. This does not significantly impact the petri factor calculation if the irradiance field is relatively uniform without much decrease in irradiance at the edges of the petri dish. This would become an important step for a non-uniform irradiance field and when there is a significant drop in the irradiance when approaching the edge of the petri dish.
The largest opportunity for introducing error with this factor is in the forecasting of the irradiance filed based on measurements on the X and Y axis. A theoretical irradiance field is calculated to encompass the size of the petri dish, with measurements typically being taken at 5 mm increments a theoretical 5 mm grid is calculated. A two-fold problem arises by doing this. With all of the care being taken to accurately measure parameters, the size of the petri dish may be measured to a decimal precision. If there is an intensity decrease at the dish edge it may not be detected as the petri factor will always be determined by rounding down. Also, if there is a significant decrease in irradiance at the petri dish extremities, then the full impact of this decrease may not be accurately accounted for, again due to the 5 mm grid. Using a theoretical irradiance field created on a 5 mm x 5 mm grid, without interpolation to a smaller grid, provides a good example of how this can happen and is illustrated in Table 2 below.
In this example two scenarios are looked at; one where the 5 mm x 5 mm grid without interpolation is calculated and one where the incident irradiance is measured in 5mm increments on the X and Y axis and forecast to a 1 mm x 1 mm grid. When using a diameter in multiples of 5 mm and well within the uniform portion of the irradiance field there is not much difference between both methods of calculation as can be seen with the 4.5 and 5.0 cm petri dish diameters being only 0.15% and 0.09% different from one another respectively. On the other hand, when the petri dish diameter is measured at 4.974 cm using a calibrated caliper, the calculated petri factors differ by 1.57%. This is significant and would contribute to an error in the fluence calculation of exactly 1.57%. With the second scenario, of the irradiance field dropping off significantly at the edges of the petri dish, we can see the effect with the 6.0 cm dish. Using the 1 mm grid calculates a petri factor 1.62% lower than a 5 mm grid, as this calculation more fully captures the irradiance decrease at the edges over the entire perimeter of the dish.
The effects of improperly applying the petri factor calculation or not applying a mask to the radiometer detector can become insignificant sources of error so long as the irradiance field is sufficiently uniform and large over the entire area encompassed by the petri dish used for the collimated beam exposures. As long as these two conditions are met, then the petri factor itself likely is approaching 1.0 and that accomplishes two goals. First, it makes the petri factor calculation on a 5 mm x 5 mm grid without interpolation valid and narrows the difference between the UVDGM fluence calculation equation  and the NWRI equation . This concern was addressed in the new IUVA guidance and spreadsheet for low-pressure UV collimated beams experiments (Bolton et al., 2015a) as interpolation from a 5 mm grid to a 2.5 mm grid is now used that provides four times the points with which to calculate the petri factor compared to past protocols.
The UVDGM does not specify an error for the absorbance measurement with a spectrophotometer, but for the water factor, which included absorbance and is a maximum of 5% (UVDGM 2006). Having an error of this magnitude would have a large impact on the final fluence calculation, but an error of this magnitude is unlikely as long as appropriate controls are in place. The error calculated for the setup at GAP was only 0.41%, at maximum, and was dependent on the spectral absorbance of the sample and the sample absorbance at 253.7 nm.
The main sources of error – which many people often overlook – are to ensure that the cuvette being used for the absorbance measurement is properly cleaned prior to placing it in the spectrophotometer to take a reading and using the proper path length. Simply having a fingerprint, condensation or a streak of dirt on the outside of the cuvette for the zeroing procedure or while taking a reading can alter the result drastically. To prevent this from happening two good habits to practice are to look at the cuvette against a backlight to catch any obvious problems and to take more than one independent absorbance measurement to ensure agreement. The UVDGM (2006) and Bolton et al. (2015a) also suggest using a quartz cuvette with a path length of at least 4 cm whenever the sample UVT is greater than 90% per centimeter (absorbance of less than 0.0458).
Another source of error that can be introduced in the absorbance measurement is from the spectrophotometer itself. As with all instruments used in the collimated beam apparatus, it is important to ensure that the spectrophotometer is calibrated annually against NIST traceable standards. Calibration is by no means a guarantee that the instrument will read correctly over the calibration interval, so it is important to maintain a routine of verifications at appropriate intervals. The UVDGM outlines two recommended verifications – one wavelength verification using a certified holmium oxide standard and one absorbance verification using a certified potassium dichromate standard. At GAP we also employ a second wavelength verification using a certified rare earth standard that displays an absorbance peak more comparable to the low pressure mercury lamp output of 253.7 nm. All of these standards are available from Starna Cells as certified standards, which they recommend re-certifying every two years. Performing these additional verifications does add cost to the collimated beam procedure but are important if working with a sample with a large absorbance slope at 254 nm.
Reaction vessel wall reflection
Over 10 years ago Kuo et al. (2003) suggested that reflection of UV light off the petri dish walls may be an important factor not taken into account in the standardized collimated beam fluence calculation, and this may lead to an underestimation of the UV light irradiating the sample and this has been restated a number of times since (Kuo et al. 2003, Bolton and Linden 2005, Kuo et al. 2005, Wright et al. 2015). This reflection may be a result of reflection within the water layer, but it has been observed here at GAP that reflection of the non-perpendicular light off the petri dish glass above the water layer has a greater impact on the fluence applied to the sample as is demonstrated in Figure 1.
Determining a correction factor to apply to a fluence equation would not be the appropriate way to deal with this error; rather, the test should be sufficiently controlled to ensure that reflection does not play a role in the underestimation of the applied UV fluence. Depending on specific conditions it has been observed at GAP that sidewall reflection can account for an error greater than 10 % with bacteriophage T1UV. The vessel diameter was 8.5 cm and height was 4.9 cm. The suspension depth was 0.87 cm. The UV sensitivity of T1UV phage with and without side wall reflection was 4.49 and 4.99 mJ/cm2 per log inactivation. The ratio of the UV sensitivities with and without wall reflections with this example was 1.11.
One way to eliminate this effect would be to use a dish material that is non-reflective or to coat the dish with a non-reflective coating. A second method would be to ensure that the walls of the reaction vessel do not extend significantly past the solution surface (Wright et al. 2015). Combining these two strategies would provide the best strategy for eliminating the effect of reflection as it would also eliminate any possible reflection that may occur within the water layer.
Another method that has been suggested to eliminate side wall reflection is through the use of a mask or washer placed on top of the petri dish (Bolton et al. 2015b, Li et al. 2011). This effectively eliminates the potential for sidewall reflection by preventing any angular light from hitting the dish walls above the water line. All light that passes through the washer will impact the surface of the water, as the mask blocks any light that would impact the dish wall above the water line. If using this method to eliminate side wall reflection, a correction factor must be applied to the dose equation based on the area of the hole in the mask as compared to the water surface area. A photon based collimated beam fluence calculation (Bolton et al. 2015b) can also be used. This method recently has been suggested, and work is currently being performed to validate the method.
Sidewall reflection is one factor that is often overlooked when it comes to the design of a collimated beam apparatus. Simply by performing a collimated beam test under a variety of circumstances using changes in material or small changes in apparatus, then comparing results is a good way to determine if any error does exist in the method and allows you to break down the components to find out where the error is originating. This method is how it was determined that reflection was playing an important role in the underestimation of the UV does under certain circumstances at GAP.
Simply put, a parallax effect would be introduced any time there is an attempt to match the height of the sample surface to the calibration plane of the radiometer detector when switching between taking incident irradiance readings and performing the collimated beam exposures. This has traditionally been done by taking a measurement to match distances when switching between the detector and the sample; at GAP we use a lab jack to move the stir plate up and down to achieve this. When using the measurement method there is a chance that a small error can be made when making this switch and it is not possible to quantify this error, so all efforts should be made to eliminate this factor. One way to do this is by using a laser level that is set to an arbitrary height and using lab jack to adjust the sample surface and radiometer calibration plane to this level. This method eliminates the parallax effect and is a much faster way to switch between sample irradiation and incident measurement using a radiometer.
The length measurements are often overlooked for the importance they play and number of times they appear in the fluence calculations. A slight error in this measurement does not necessarily result in a large error in the final fluence calculation but, like all measurements used in the fluence calculation, there are precautions and quality control steps that can be taken to ensure the error introduced by this measurement is minimized. One source of error is with the width measurement of petri dishes. These can be purchased in a variety of widths, with this diameter being used in the calculation of the petri factor and possibly sample depth. As with any manufacturing process these measurements are not exact, so it is important to measure the width of the dishes with a calibrated Vernier caliper. One suggestion is to take three random measurements across each dish and average them to obtain the average dish width. The standard deviation of the widths of all dishes then should be determined, and any dish that falls outside two times the standard deviation (95% confidence interval) should be removed from circulation. The average width of the remaining petri dishes then can be used in the fluence calculation.
A calibrated ruler is also an important instrument in measurement of collimated beam fluence calculation factors. It would be required for measurement of the petri factor X and Y axis intervals, to ensure that the entire irradiance field exposed in the reaction vessel is represented in the petri factor calculations. It also would be required for measurement of the distance from the lamp centerline to the sample surface used in the divergence factor of the UVDGM fluence calculation.
Length measurement instruments are and added expense, but fortunately routine calibration of a metal ruler is not required unless excessive wear or trauma to the instrument is evident, and calibration of a Vernier caliper is a widely performed, inexpensive, calibration procedure.
As with length measurements, a calibrated instrument, in this case a stopwatch or timer, is required to minimization uncertainty. The collimated beam design, and whether it uses a manual or automatic shutter, plays a role in the uncertainty associated with time. A manual shutter will always have a larger uncertainty associated with it, as user input is required, but the effect can be minimized by ensuring that exposure time is sufficiently long so the time required to manually move the shutter is a small percentage of the total irradiance time. The UVDGM suggests exposures no shorter than 20 seconds, but longer minimum exposures should be used if possible.
If a manual shutter is used, or an automatic shutter that is calibrated to seconds and not milliseconds, then all final fluence values should be recalculated to this precision. Typically, the time required for an irradiance is calculated to the fraction of a second, but in practice all dosing is done to a one second precision. Performing this recalculation will provide minimal differences in the actual applied fluence, but is nonetheless important in the accuracy and uncertainty of the calculated fluence.
No matter how much effort and emphasis is placed on controlling for all possible sources of error when performing a collimated beam test there is still opportunity for an error to occur. This may be due to user error, water quality or some other factor that has never been accounted for in the past. To catch any of these unexpected errors, there needs to be some check in place to analyse the final results to see if they fall within the expected range. The UVDGM and the NWRI guidance both provide upper and lower limits for MS2 irradiations, but these limits are very wide, and a large amount of error is possible within these bounds. A better approach would be for each individual lab to produce their own set of upper and lower bounds for the surrogates being irradiated routinely. Figures 2 and 3 show what the limits are for MS2 and T1UV irradiations at GAP. For the MS2 (Figure 2) it can be seen how much more restrictive GAP’s limits are when compared to the NWRI and especially the UVDGM limits.
How each individual lab handles these limits is not specified, but a suggestion is to see if the resulting dose response curve falls within the limits; individual points can fall outside the limits but the curve must fall inside the bounds. If the curve falls outside of the bounds then all of the measured parameters are verified, corrections are made to the calculated fluence values if necessary, and if the curve still falls out of bounds then the collimated beam test is repeated where time and sample volume permit. If the repeated collimated beam standard curve is still not within limits the client is made aware of the issue and results issued appropriately. The limits calculated at GAP are 95% confidence limits, so there is still a 5% opportunity for the curve to fall outside of the limits, but these limits do give us the opportunity to perform a final check to ensure that the entire procedure has been performed correctly and that no unforeseen error was introduced along the way.
Unfortunately, producing limits is not possible for all collimated beam surrogates that may be irradiated at a lab. With MS2, T1UV and other bacteriophage the surrogate being worked with is from a single parent organism which is spiked into test water. It is known that all of the MS2 (or other bacteriophage) is descended from a single source and it can be assumed that there has been little chance for mutation over the small number of generations present within a batch. Since mutation is not expected, the dose response relationship is expected to be the same for all members of the measured population. This is not the case when irradiations are being performed on indigenous organisms in a water source. For example, the fecal coliform population consists of a diverse number of bacteria that may each exhibit a unique dose response relationship to UV radiation that may vary by location. In these instances, all of the checks performed before initiating the collimated beam test have to be considered a best effort to produce an accurate collimated beam dose response curve.
Since most people rarely perform collimated beam tests, they are not necessarily aware of the care and accuracy that is required to properly perform collimated beam irradiations. If sufficient effort is taken to reduce uncertainty in all measurements entered in the collimated beam fluence calculation, there is still a question as to what fluence calculation to use. The UVDGM calculation provides the largest amount of safety by calculation of a conservative (lower) fluence applied by a UV reactor, but the NWRI equation is equivalent if working with a completely uniform irradiance field and if light divergence is minimized using a long “lamp centerline to sample surface” distance. The NSF equation is also appropriate when being used in the context of certifying a system to the NSF 55 standard as it allows direct comparison between systems where a pre-defined fluence is the target. With this being said, the NSF equation may not be appropriate for determining the dose response relationship on an emerging pathogen for regulatory purposes, and there is also indication that some NSF standards are heading towards the UVDGM fluence calculation in some of their updated methods.
By breaking down the elements that go into calculating the fluence in a collimated beam test the hope is that more care will be taken by labs and individuals by taking an in depth look at their collimated beam apparatus and their procedures to ensure that the data being produced using their system is as accurate as possible. Future revisions of the UVDGM and NWRI guidance should outline all of the possible sources of error and how they can be mitigated so a collimated beam dose response curve produced at one lab would agree with one produced by another lab. The IUVA Board of Directors has shown their commitment to this process with the release of the “Protocol for the determination of fluence (UV Dose) using a low-pressure or low-pressure high-output UV lamp in bench-scale collimated beam ultraviolet experiments” (Bolton et al., 2015a) and further discussion may bring other sources of error to light; the preceding analysis of collimated beam uncertainty can be used as a basis for this discussion. The field of UV water disinfection is a growing industry, and all effort should be made to ensure knowledge and research continues to grow as well.
- Bolton, J.R. and Linden, K.G. 2003, Standardization of methods for fluence (UV dose) determination in bench-scale UV experiments. J. Environ. Eng. 129: 209-215.
- Bolton, J.R. and Linden, K.G. 2005. Discussion of “Standardized collimated beam testing protocol for water/wastewater ultraviolet disinfection” by Jeff Kuo, Ching-lin Chen and Margaret Nellor. J. Environ. Eng. 131: 827.
- Bolton, J.R., Stefan, M.I., Lykke, K.R. and Shaw, P.-S. 2011. Determination of the quantum yield of the potassium ferrioxalate and potassium iodide-iodate actinometers and method for the calibration of radiometer detectors. J. Photochem. Photobiol. A: Chem. 222: 166-169.
- Bolton, J.R., Beck, S.E. and Linden, K.G. 2015a. Protocol for the determination of fluence (UV dose) using a low-pressure or low-pressure high-output UV lamp in bench-scale collimated beam ultraviolet experiments. IUVA News. 17(1): 11-16.
- Bolton, J.R., Mayor-Smith, I. and Linden, K.G. 2015b. Rethinking the concepts of fluence (UV dose) and fluence rate: the importance of photon-based units – a systemic review. Photochem. Photobiol. 91: 1252-1262.
- EPA 2006. Ultraviolet Disinfection Guidance Manual for the Final Long Term 2 Enhanced Surface Water Treatment Rule. United States Environmental Protection Agency, Washington, DC.
- Kuo, J., Chen, C. and Nellor, M. 2003. Standardized collimated beam testing protocol for water/wastewater ultraviolet disinfection.” J. Environ Eng. 129: 774-779.
- Kuo, J., Chen, C. and Nellor, M. 2005, Closure to “Standardized collimated beam testing protocol for water/wastewater ultraviolet disinfection” by Jeff Kuo, Ching-lin Chen and Margaret Nellor. J. Environ Eng. 131: 828-829.
- Li, M., Qiang, Z., Li, T., Bolton, J.R. and Liu, C. 2011. In situ measurement of UV fluence rate distribution by use of a micro fluorescent silica detector. Environ. Sci. Technol. 45: 3034-3039.
- NWRI 2012. Ultraviolet Disinfection Guidelines for Drinking Water and Water Reuse. Third Edition. National Water Research Institute, Fountain Valley, CA, in collaboration with American Water Works Association Research Foundation.
- Wright, H., Heath, M. and Bandy, J. 2015. Yikes! What the UVDGM Does Not Address on UV Disinfection. IUVA News. 17(2): 12-18.