Estimated numbers of bacteria in samples from non‐homogeneous bodies of water: how should MPN and membrane filtration results be reported?

Abstract

Published tables of Most Probable Numbers (MPN) used in water bacteriology make the assumption that the sample is representative of a large body of water in which the bacteria are randomly distributed. MPN values for certain dilution series are calculated without making this assumption. Although the actual MPN values obtained in this way are similar to those published, a question remains about how the MPN result should be reported. Because it is an estimate of the likely number and not a count of the organisms, a range of other plausible values could be given for that particular water sample. It is suggested that for certain results, where there are close contenders for the description ‘MPN’, a numerical range should be given in preference to a single MPN value. A further range of ‘possible’ numerical values is also suggested to replace the published approximate confidence intervals which are inappropriate for non‐homogeneous waters. Confidence intervals for the numbers of bacteria calculated from membrane filtration may be estimated from this sample only if it can be assumed that there is random distribution of bacteria in that water source. Water sources likely to contain bacteria in varying density (other than purely random variation) include pre‐treatment waters intended for drinking, and bathing waters. The only way to study the distribution of bacteria in such water sources is to take a series of samples appropriately planned in place and time.