Quantitative Microbiology 1, 111±136, 1999
# 2000 Kluwer Academic Publishers. Manufactured in The Netherlands.
Resolving UV Bioactinometric Results with Intensity
and Time Distributions
BLAINE F. SEVERIN, PH. D, PE.
Director Environmental Technologies, MBI International, PO Box 27609, Lansing MI 48909
PETER F. ROESSLER
Research Microbiologist, Amway Incorporated, 7575 East Fulton Rd., Ada, MI 49355-0001
Doctoral Candidate, Department of Statistics and Probability, Michigan State University, E. Lansing, MI.
Received July 13, 1998; Accepted April 27, 1999
Abstract. A UV reactor with an annular design, a total liquid volume of 460 ml, and out®tted with a single lamp
with 1690 mW of germicidal power was tested. Coliphage MS2 was used as a bioactinometer to measure the UV
dose at a ¯ow rate of 56.7 mlasec in water with a very low absorbance. The Beers Law coef®cient was
` 0X003. The measured dose (MS2 bioactinometry) was 35X2 Æ 1X1 mW-secacm
A retention time distribution was generated with a dye tracer study. The reactor was modeled as if ¯ow was
con®ned to ten equal volume paths existing as concentric rings around the lamp. The UV intensity along each path
(ith intensity) was calculated to generate a simulated distribution of UV intensity in the reactor. The retention time
distribution was subdivided to estimate the retention time associated with each decile jth time) of the total ¯ow.
Seven methods of associating the ith intensity with the jth retention time were used to produce simulated dose
distributions for the reactor. The average UV dose for each distribution was calculated as the average of the
products of I and t (AP protocol) and by the apparent survival (AS protocol), in which the predicted survival along
each path was averaged to back-calculate dose from the reference batch inactivation curve. The average dose
predicted assuming that time and intensity were independent was 51.5 mW-secacm
based on the arithmetic
average (AP protocol). Using the apparent survival method, the predicted dose for the independent distribution
(I independent of t) was 36.4 mW-secacm
. Three methods of developing dependent structure between time and
intensity were tested. In the best possible case for strati®ed ¯ow (I negatively correlated with t) the calculated (AS)
intensity was 46.3 mW-secacm
In the worst case for strati®ed ¯ow (I positively correlated with t) the AS
intensity was 32.0 mW-secacm
. In a rational case where ¯ows were assumed to be distributed parabolically (low
¯ow at the wall and at the lamp) produced an AS intensity of 37.7 mW-secacm
. When either time or intensity was
averaged, while the other variable was allowed to keep its distribution, the (AS) dose (time averaged 43.3 mW-
, intensity averaged 41.0 mW-secacm
), yielded a poor prediction compared to the measured value.
The errors associated with averaging time, intensity, or both, far outweigh the errors associated with choosing a
rational distribution or an independent distribution of time and intensity in the prediction. This observation is
generally true whenever an organism is exposed to UV light in a ¯ow through reactor such that the range of doses
is within the portion of the inactivation curve exhibiting strong exponential decay.
Key words: ultraviolet light, disinfection, coliphage MS2, tracer studies, intensity pro®les
Most ultraviolet light (UV) disinfection reactors are constructed to achieve intimate contact
between the water and the light source. Due to the nature of this design, light intensity