A3.11 Interpretation of calculated LRVs for practical treatment guidance

The calculated LRVs presented in Table A3.8 are selected outputs of the QMRA analysis based on the input values described in this appendix. There is however uncertainty associated with characterising the input assumptions. The final defining assumptions are based on the best available published scientific evidence from the Australian context, as collated and summarised by WaterRA (Deere et al. 2014) and piloted by the Water Services Association of Australia (WSAA 2015) (Walker et al. 2015). These assumptions have been made in the context of what relevant treatment processes can achieve and what is currently accepted national and international practice. In presenting recommended LRVs for typical catchment types, Table A3.9 summarises those treatment process trains, LRVs and a reasonable nominal treatment target. These have been drawn from the Australian context, noting that this is within or very close to the range given in Table A3.8.

The calculated LRVs for Cryptosporidium are likely to be highly conservative because:

• the estimated concentrations by source water category were based on total oocyst counts enumerated based on immunofluorescent antibody staining and did not account for viability (i.e. whether the oocysts are “dead” or “alive”) and other determinants of infectivity in humans (i.e. whether the oocysts are of a strain that is capable of causing infections in humans).

• the dose-response function assumes that all oocysts are highly infectious, with the estimated probability of infection per oocyst close to 20%. It is unlikely that all oocysts are this infectious, and many individuals may have some form of acquired immunity that is not accounted for in this model.

• the cited dose-response studies forming the basis of the QMRA used total oocysts numbers as the dose variable. The studies did not adjust for in vitro infectivity and the oocysts were often stored for many weeks before being fed to volunteers. Furthermore, the use of both live animal and in vitro infectivity assays are highly variable between oocysts (see for instance Rochelle et al., 2002; Slifko et al. 1997; 1998; and 1999).

Infectivity of Cryptosporidium was examined in two drinking water sub-catchments vulnerable to localised contamination by livestock and human faecal wastes (vulnerability class 4) (Swaffer et al. 2014, 2018). In the first investigation 12 of 383 oocysts (3.1%) tested from a drinking water catchment were infective by cell culture assay (Swaffer et al. 2014). In the second investigation 1162 of almost 15000 oocysts (7.8%) from 578 samples collected over 4 years were infective (Swaffer et al. 2018). Average infectivity at individual sampling locations ranged from 0-18% with a maximum infectivity of 65.4% recorded from one event coinciding with the presence of newborn livestock at a location high in the catchment. C. parvum was detected in 107 of 475 PCR positive results (23%) but C. hominis was not detected. The samples were collected during rainfall run-off events and were upstream of drinking water reservoirs. Longer retention times in reservoirs or in river runs with limited impacts close to points of abstraction (e.g. vulnerability classes 1 and 2) are likely to provide further reductions of infectivity. However, these studies show variable results. Based on the results for speciation and infectivity, the LRV targets for Cryptosporidium shown in Table A3.9 have been reduced by 1 $\text{log}_{10}$.

In more vulnerable catchments where water abstraction points in rivers may be closer to sources of faecal contamination (e.g. Category 4 catchments) it may be prudent to only reduce LRV targets by 0.5 $\text{log}_{10}$. As shown by Swaffer et al. (2018) average and maximum infectivities can be higher than 10% in impacted catchment streams.

Changes in LRV targets based on infectivity can be varied depending on local circumstances (e.g. low or high ambient temperatures, short or long travel time of oocysts) in consultation with the relevant health authority or drinking water regulator.

The LRVs presented in Table A3.9 for viruses and bacteria are point values selected from the range represented in Table A3.8. This is done by taking the upper limit and rounding to the nearest half $\text{log}_{10}$. A LRV of 4 for bacteria is recommended for Category 1 due to the need to protect against local bacterial contamination from wildlife (e.g. birdlife) which may carry human-infectious bacterial enteric pathogens.

Table A3.9 Recommended LRVs by source water category (also see Chapter 5 Table 5.5)

Source water category	Source water type (vulnerability classification)	Maximum or 95ᵗʰ percentile⁽¹⁾ E. coli results from raw water monitoring (number/100 mL)	LRV target to achieve 1 µDALY per person per year - Protozoa⁽²⁾	LRV target to achieve 1 µDALY per person per year - Virus⁽³⁾	LRV target to achieve 1 µDALY per person per year - Bacteria⁽⁴⁾	Indicative specified treatment technologies
Category 1	Fully protected surface water or groundwater (Class 1)	<20 (E. coli band 1)	0	0	4.0	Chlorination
Category 2	Groundwater or surface water with moderate levels of protection (Class 2)	20 to 2000 ⁽⁵⁾ (E. coli band 2)	3.0	4.0	4.0	Direct filtration and chlorination
Category 3	Groundwater or surface water with poor levels of protection (Class 3)	20 to 2000 ⁽⁵⁾ (E. coli band 2)	4.0	5.0	5.0	Conventional filtration and chlorination
Category 4	Unprotected surface water (Class 4)	>2000 to 20,000 (E. coli band 3)	5.0	6.0	6.0	Conventional filtration, UV and chlorination

1. Maximum E. coli results from raw water representative of inlet to treatment plant should be used unless data set is robust enough to use 95ᵗʰ percentile.

2. The LRV was based on the estimated arithmetic mean of total oocyst counts for different Australian source waters. Total counts typically overestimate human infectious oocysts, and therefore an infectivity discount factor of -1.0 $\text{log}_{10}$ was applied to obtain the LRV target. For high-risk sources (e.g. Category 4) the discount factor may need to be reduced (e.g. to -0.5 $\text{log}_{10}$) which will change the required LRV. This should only be done in consultation with the relevant health authority or drinking water regulator.

3. The LRV was based on the ratio between protozoa concentration and the adenovirus concentration (MPNIU/L) found in sewage (viruses being present at approximately 0.5 log higher concentrations than protozoa as shown in Table A3.3). For Category 1 sources which have no humans in the catchment a 0 log reduction is set as humans are the predominant source of enteric viruses.

4. The LRV was based on the ratio between protozoa and bacteria found in sewage (bacteria being present at approximately 0.5 log higher concentrations than protozoa as shown in Table A3.3) for Category 4 sources. For Category 2 and 3 sources, an additional log reduction was included to allow for the presence of non-human inputs and zoonotic pathogens. For the Category 1 source, a 4 log reduction was included to inactivate zoonotic pathogens from non-human inputs.

5. Maximum E. coli results for raw water monitoring for source water Categories 2 and 3 are within the same range and distinguishing between these two categories is confirmed based on the results of the vulnerability classification.