Thursday, May 24, 2007

A Risky Conversation: Collignon & Khan.

I awoke this morning to find another response from Prof Peter Collignon in my inbox. I think it provides a very good opportunity for us to start discussing the issue of ‘risk’ in a little more detail. That’s where Peter started and I have tried to take it a little further in my ‘right of reply’ at the bottom of this post. I hope readers will find it interesting and feel encouraged to contribute to this public conversation. All points of view are welcome!

I’m also glad to post some of the detail on Peter’s thoughts about the specific IPR proposal for Canberra and his evaluation of its need. It would be great to receive some feedback from Actew on this...

From Prof Peter Collignon
Infectious Diseases Physician and Microbiologist
Director Infectious Diseases Unit and Microbiology Department, The Canberra Hospital. Professor, School of Clinical Medicine, Australian National University.

Dear Stuart you wrote,

"This is Prof. Collignon’s interpretation of the ‘risk matrix’ provided in the Australian Drinking Water Guidelines (and in many other risk-based guidelines). The risk matrix requires risk assessors to consider the combined implications of the ‘likelihood’ and the ‘consequences’ of a potential hazardous event. Doing so provides a ‘risk rating’ of low, medium, high, etc..

One disadvantage of the risk matrix is that it is quite qualitative in the nature of the ratings that it provides and thus very open to interpretation or opinion. However, I feel that Prof. Collignon’s use of the risk matrix is overly simplistic and thus incorrect. His simplistic approach would be expected to deliver a risk rating of “high” or “very high” for any water source whatsoever. In other words, while the likelihood of something going wrong is very low, the consequences are potentially devastating. But this is not an appropriate use of the matrix (which is fortunate or cities may have to cease delivering potable water!).

The risk matrix should be used to evaluate risks posed by the occurrence of specific “hazardous events”. A hazardous event may be something such as a high rainfall period, a failure of reverse osmosis pump, a loss of electricity to the treatment plant, a cross-connection between potable and non-potable water supplies, etc. Only once such an event is identified can the risk matrix be used to evaluate it. Risks associated with hazardous events that are not assessed as ‘low’, require the institution of additional barriers or management practices until a ‘low’ risk can be determined. The risk matrix is a planning tool intended to ensure that planners and scheme operators identify the sources of risk in their systems and manage them effectively with multiple barriers."

Stuart how you would define the level of risk of recycling water from sewage into potable water supplies if you believe my interpretation of the risk as “high” is not correct? If we just take one part of a system (eg reverse osmosis as you suggest) then surely if that goes wrong for some reason ( a membrane rupture that is undetected for a period of time), then at least for faecal viruses the consequence could be potentially be considerable for a very large number of people would they not? Does that not make this "high" risk? Yes there are steps that can be taken after this such as a UV step after the RO, but if large numbers of viruses are present can we be sure that UV will cope? In any case does that still not make the use of RO as “high” risk? It is just then as you say, the planners and scheme operators after identifying the sources of risk in their systems need to be able to manage them effectively with multiple barriers.

I attach the tables from the Australian water guidelines which I hope you might post along with these comments. I would be interested to see what others think and how they would rank the "risk" in Canberra of the current “water2water” recycling proposal.

The trouble for the Canberra proposal is that there is not good data available from elsewhere on what is proposed for Canberra to make a quantitative risk assessment as no similar proposal has been used anywhere else. Therefore it is likely that only a qualitative risk assessment can be done. I still think it is "high" risk. This however does not mean we should never take a "high" risk, just we need to be sure that it is necessary to take this risk and that there are not other reasonable alternatives available. The drinking water guidelines are available from here.

For the benefit of your readers, I will outline what I understand is the Canberra sewage recycling proposal. It is going to start with recycling 9 GL per year of water from sewage and then go to 20 GL relatively quickly (I presume within a few years). Water will be pumped from the Molonglo sewage outflow after RO treatment etc to new artificial terraced wetlands in the upper reaches of the end of the valley where the creeks/streams that eventually enter the Cotter Dam (the Cotter Dam is lowest of the three Dams on the Cotter river). The Cotter dam is only very small at 3.8 GL. It also has to be kept at about 90% full at all times so that the breeding ground of threatened fish is not prejudiced. This has meant however that even in droughts after rainfall this Dam frequently overflows. There are plans to also build a bigger Cotter Dam (to 80 GL and which I think is a good idea). However this has still not been approved and completion will be quite a number of years after the recycling plant will go into operation.

I have been told that it will only be about 2 or 3 days before the recycled water pumped into the artificial terraced wetlands enters the Dam water. Because water in the dam runs the risk of overflowing, my understanding is that water will need to be pumped out almost as fast as it enters the reservoir. This means it will be pumped up to the treatment works at Stromlo and then put directly into the reticulated water system of Canberra. What is not used along the way in the city while it traverses the reticulated water system will be pumped into the Googong dam (which has over 100 Gl capacity but is on the other side of the city). However I presume this will mean, given that Canberra uses currently about 50 Gl of water per year, that if 20 Gl of water is recycled, then about 40% of our reticulated water may be recycled water. One can argue that we might access other sources of water from other Dams and this percentage will be lower, but if we do that why then produce such energy expensive recycled water? While the Cotter dam is so small, unless the added water is pumped out it will likely just overflow over the top of the dam and be lost from the catchment.

I would also appreciate people’s view on how this proposal should be defined. This is currently being called an “Indirect” potable water usage. However if it is really only takes 2 or 3 days before this water hits the small Cotter dam and then is also likely to be relatively quickly pumped into the reticulated water system, is this not really a “direct” potable use? (if not, does it not miss out on that definition by just a couple of days).

This rapid return of this recycled water to our reservoir and then into our reticulated water system seems to me to be taking short-cuts with a lot of normal biological and natural safety barriers that are usually in place, namely major dilution factors and long retention times in shallow wetlands, aquifers or dams.

As a last note, with level 3 water restriction in Canberra, we use 50 Gl or less from our dams per year for domestic/industrial consumption. In every one of the recent drought years, except for 2006, we have had much larger inflows than 50 GL into our dams (in 2005 it was over 100 GL’s). Indeed at the beginning of 2006 our dams’ levels were close to 70% despite the drought. In Canberra therefore, I can't see why we need to recycling of water from sewage into our drinking water, especially given the compromises to natural safety barriers that will occur if it is done before an enlarged Cotter Dam is built. Stuart I don’t know if you can attach graphs but this I think shows our situation.


Stuart’s Response:

The risk matrix is a useful tool to assist in assessing the risks associated with specific, well defined ‘hazardous events’. The risks arising from such ‘hazardous events’ are described in terms of their ‘likelihood’ and ‘consequences’ in order to assign a ‘risk rating’ from ‘low’ to ‘very high’.

As mentioned in the previous post, a major limitation of the risk matrix is its rather qualitative nature and thus the fact that interpretations may be left largely open to opinion or even ‘gut feeling’. However, I find that it is useful to try to use quantitative (or at least semi-quantitative) assessments as much as possible. This can be a little cumbersome with the risk matrix, but there are a number of workable approaches.

Table A4 above gives ‘qualitative measures of likelihood’. The recent National Guidelines for Water Recycling (Phase 1) take this a step further by providing some semi-quantitative descriptions for assigning likelihood. These involve describing expected frequency of occurrence by the following descriptors. Rare: “May occur only in exceptional circumstances. May occur once in 100 years”, Unlikely: “Could occur within 20 years or in unusual circumstances”, Possible: “Might occur or should be expected to occur within a 5- to 10-year period”, Likely: “Will probably occur within a 1- to 5-year period”, Almost certain: “Is expected to occur with a probability of multiple occurrences within a year”. With such an approach, it is then possible to examine performance histories of other plants using similar technologies (or other predictive means) to make a reasonable judgement of the frequency (and thus likelihood) of specific hazardous events occurring.

Assessing the ‘consequences’ of a hazardous event can be undertaken in a similarly quantitative manner. Considering Peter’s example of viruses, we can consider the ‘consequences’ of a hazardous event in terms of the impact it has on the number of infections. For a given virus, this is directly related to the level of human exposure to the virus. So a useful approach is to assess the consequences of a hazardous event in terms of its impact on exposure.

To quantify the impact of a hazardous event on exposure, we first need to have a quantitative description of ‘normal’ exposure. This will be defined in terms of virus concentrations (and variability) in the source water and how well it is removed at each of numerous subsequent barriers (eg. microfiltration, reverse osmosis, advanced oxidation, environmental residence, drinking water flocculation, chlorination). Again, these removals (or ‘decimal reductions’ or ‘log reductions’) can be informed by experience with existing plants using the same or comparable technologies. By considering a large number of hazardous event scenarios, we can examine the impact that they can be expected to have on the removal of the virus, and thus on human exposure to it.

All significant international health organisations (eg the WHO), accept some concept of ‘tolerable risk’. For viruses, this is typically defined in terms of infection likelihood (or more recently, a concept called DALYs). Advanced water recycling schemes will be designed to achieve sufficient removal of viruses, such that accepted infection likelihoods will not be exceeded. The multiple barrier approach ensures that it must be achieved with significant safety margins of numerous orders of magnitude.

Consideration of a hazardous event by its impact on human exposure and thus by impact on infection likelihood, now allows us make a reasonable quantitative assessment of risk.

Peter has asked me “Stuart how you would define the level of risk of recycling water from sewage into potable water supplies if you believe my interpretation of the risk as “high” is not correct?”. However, it was not his assessment of “high” that I necessarily disagreed with. I was really commenting on the simplistic reasoning that was used to arrive at it. I hope that after the above very brief introduction to the concept of quantitative risk assessment, readers will appreciate that I can not provide a simple answer to how I would rate the risks associated with a particular IPR scheme without access to significant plant design information. Even once that is available, it is a formidable task to systematically consider possible hazardous events and assess them in terms of their likelihood and consequences.

Nonetheless, I will go as far as to say that it is not necessarily the case that a ruptured RO membrane presents a significant risk. Even if we ignore the ‘likelihood’ for a moment and just consider ‘consequences’, as we have described, in terms of impact on human exposure. A typical RO plant will not have just one RO membrane module, but hundreds of them installed in parallel (and even in series in some cases). If one RO membrane out of 100 has a complete rupture, that would represent a 1% increase in exposure. A complete rupture would be instantly noticed by a dramatic loss of pressure, so a much smaller fraction of this 1% is actually more realistic. This may mean that the schemes overall performance for reducing virus concentrations may be reduced from 10 log units to 9.9 log units [Note: this calculation is incorrect. See comments below]. For a safely designed scheme, this would result in an elevated level of exposure, but one that was still well within tolerable risk levels.

The issue of ‘indirect potable recycling’ and ‘direct potable recycling’ is one that we have dealt with a few times on this blog (eg. see the long discussion between ‘Mark’ and myself here). I see that it really is a sticking point for many people!

If you want to go by the accepted definitions, the scheme proposed for Canberra is an indirect potable recycling scheme. This is because the water is not plumbed directly from the advanced water treatment plant to the distribution system. Instead, it discharges to a reservoir and then will be treated (with other source water in the reservoir) at a drinking water treatment plant. Only after the drinking water treatment plant will it be plumbed to the distribution system.

However, before we go down that path again, I wish to state that simply being named or defined as an ‘indirect potable recycling scheme’ doesn’t automatically make a scheme safe (a rose by any other name…). In order to be judged as ‘safe’ it is necessarily for each scheme to be carefully considered in terms of its actual design attributes…not just its name.


Anonymous said...

I’d prefer a risqué conversation myself.

Mark said...

Hi Stuart,

If one RO membrane out of 100 has a complete rupture, that would represent a 1% increase in exposure. A complete rupture would be instantly noticed by a dramatic loss of pressure, so a much smaller fraction of this 1% is actually more realistic. This may mean that the schemes overall performance for reducing virus concentrations may be reduced from 10 log units to 9.9 log units.

Math Check: Lets assume that the RO barrier had a capability of 4 log units on its own (6 log units from the other barriers). That would mean that if 10,000 virus particles were in the feed to RO, 1 virus particles would pass through when all is functioning correctly. But if 1% of the flow bypassed the RO due to a leak, that would allow 1% of the virus particles to pass untreated i.e. 100 virus particles through the leak. The remaining 99% of the flow would allow approximately 1 virus particle to pass. Therefore, a total of 101 virus particles would pass through RO from a starting number of 10,000. That is a log 2 reduction.

So I figure that the reduction would be around log 10 to log 8 (not log 9.9). Or have I miscalculated?


Stuart Khan said...


Yes, of course you are correct. No matter what log reduction we assign to a row of 100 parallel membranes, if we assume that one fails completely we can never achieve more than a 2 log reduction across the whole row. Fortunately such a situation is very unlikely and could be quickly picked up by an observable conductivity increase in the permeate from that particular pressure vessel or bank. Nonetheless, it does reinforce the need to implement ‘redundant’ log reduction capacity across a multiple barrier system.

Thanks for keeping me honest, Mark. I hereby vow never again to attempt mental calculations on this blog before eating breakfast!

Mark said...

It's my turn to attempt pre-breakfast logic 8~<

If 1% of the RO membranes ruptured completely, I would imagine almost 100% of the flow would pass though that gaping hole - so things like pressure, conductivity and flow sensors would turn the control-room screens bright flashing red. So granted, this situation would not last for long.

But in the case of a 1% flow leak, the pressure or flow or conductivity may only vary by a mere 1% from it's normal value. Would that necessarily stop the show? I'm imagining that these parameters would vary a little over the course of the day in any case e.g. due to membrane fouling. And from memory, 1% accuracy on a pressure sensor used to be pretty good.

The obvious thing would be to have a reliable on-stream pathogen counter. Does such a thing exist, and if not, what is the state of the art in pathogen detection?


Stuart Khan said...

Hi Mark,

It depends how you define a 1% flow leak. If you mean that the permeability of the membranes increases by 1%, then I suspect no, you probably wouldn’t see it.

However, if you mean that 1% of the surface area of all the membranes are now gaping holes, then depending on how large those holes are (and thus how many there are), we may find that you’ve effectively turned RO membranes into something more like microfiltration (MF) membranes (this is a purely qualitative statement!). In such a case, I expected that we would certainly observe a difference in conductivity since salt rejection would be significantly reduced.

An on-stream pathogen counter would be a wonderful thing. But no, I am fairly sure that no such thing exists. Nonetheless, pathogens are much, much larger than salts, so observing salts as a surrogate for pathogens is very useful.

I’m no microbiologist, but the “state of the art” in pathogen detection depends very much on which type of pathogen you are looking for (bacteria, viruses, protozoa, etc). All of these relatively difficult to detect at the low concentrations that we would be concerned about and all of them require numerous days to do. I think an on-line tool for direct monitoring of pathogens is probably a long way off. I’m sure Prof. Collignon could provide a more detailed answer to this question!

Mark said...

Hi Stuart,

It depends how you define a 1% flow leak.
I was imagining around 1% of the flow passing through a single tear - say a few millimeters in size, or (more plausibly) through faulty seals at the ends of the RO tubes.

I think an on-line tool for direct monitoring of pathogens is probably a long way off.
I guess that the next best thing would be a rapid and cheap off-line analysis method. Would you know of any promising research findings in this area, or is this also a long way off?


Stuart Khan said...

G’day Mark,

As you know, I’m no microbiologist, so I can’t answer your question with any real authority or expertise. However, after having sat through my fair share of conferences and seminars, I can tell you where I think things appear to be heading.

Probably the most significant development for rapid virus detection has been the development of advanced PCR techniques. PCR stands for ‘polymerase chain reaction’ and is the name for the process that is used to enumerate DNA strands for DNA analysis. Most viruses are essentially the same thing as single strands of DNA, so they can be measured by the same technologies. New PCR techniques have led to the rapid detection of numerous viruses. If you’re interested in reading up on some recent research in this area, you could check out the following papers (if you have access to a university library):

Hwang, YC; Leong, OM; Chen, W; et al.
Comparison of a reporter assay and immunomagnetic separation real-time reverse transcription-PCR for the detection of enteroviruses in seeded environmental water samples

Stedtfeld, RD; Baushke, S; Tourlousse, D; et al.
Multiplex approach for screening genetic markers of microbial indicators

Parida, MM; Santhosh, SR; Dash, PK; et al.
Rapid and real-time detection of Chikungunya virus by reverse transcription loop-mediated isothermal amplification assay

Stuart Khan said...

The following is a comment that I was asked to post by Prof. Peter Collignon...

Dear Stuart i am not sure my posting using Danish instructions on the web worked. Happy if you want to post this plus your viewpoint. Catching plane home in about 10 hrs so may not see any replies till Sunday.

Hi, Stuart,

I am in Denmark at a WHO expert panel meeting (on antimicrobial resistance), but have been looking at the comments re the recycling of water from sewage proposal on your website (when I hit the response button on your website however, the site's instructions are then in Danish, so I hope this gets to you OK!)

I was very interested in the discussion you and Mark had about math's. I share Marks' concerns and I think my math's came out similar to what I think Mark (and now you) are saying. My worry is what happens if 1% of the water does not go through the reverse osmosis membranes. That is different to 1% of the membrane failing. If 1% of the membrane failed, I presume large volumes of water would go through any rupture, as the high pressure in the system would drive the water that way. This presumably would be readily picked up by continuous pressure measurements etc. However it may only take very small leaks, tears etc to have 1% of the water volume go via some alternate pathways. I then don't see how any in-line measuring system will pick up such a small loss (eg pressure etc).

I thus agree with your comment that we need some type of regular measuring system developed that detects micro-organisms (especially viruses) rapidly and efficiently (presumably some type of viral molecular PCR testing– however even when we get over the practicalities of getting rapid results, PCR only picks up what you suspect is there. It won't pick up viruses etc that you don't have their genetic code included in the primers for your testing).

If 1% of water bypassed the RO system, then it likely means the numbers of viruses removed will be log 2 less that what when the system had no leaks. This means that if there was say log 6 viruses per 100 litres of water (1 million viruses) in the original sample, then log 4 virus (10 thousand) would still be coming out the other end.

This is why I believe that you should not recycle water from sewage into our drinking water if there are other reasonable options to obtain water or decrease the amounts of water being taken from our potable water supplies (eg sewer mining for irrigation etc). If we have no choice but to recycle sewage for drinking water, then we firstly need to have some type of monitoring of viruses operating fairly regularly (I would think at least twice daily - but such systems do not seem to currently exist for everyday use). There also as you have said, needs to be multiple other barriers in place after the RO system, so that if something does go wrong you have added safety barriers in place. That is why I am so vocally against the current Canberra proposal - I do not think the proposal is needed plus it is not safe enough.

In Canberra, we have enough water from other sources. We thus don't have to take this risk. However even if this proposal was to proceed, nearly all of the natural safety barriers that should be in place, will have been removed. People should note that in the recently released draft environmental report that the implications of membrane and system failure are commented on (more so than in the draft health report). In the environmental report, concerns are raised re the large volumes of water that will be put upstream of the very small Cotter dam. Because of these reasonable environmental concerns, I note that there is a proposal to consider putting the recycled water directly into the small Cotter reservoir ( 3.8 GL) instead of into artificial wetlands (which don't look to be able to work very well in the Canberra proposal anyway). This will mean that the sewage recycling proposal is then really a "direct" potable recycling scheme, that the recycled water will only have very short retention times and only relatively small dilution effects. Also there will be no slow exposure via shallow marshes, wetlands etc where UV light and other factors might have a protective and polishing effect on any viruses or other pathogens that might be in the water if a mishap with the equipment occurred. To go ahead with this proposal without finding better ways to test to ensure firstly that micro-organisms such as viruses may have slipped through (eg from small membrane leaks etc as per your previous math's discussion) and then also remove as many natural safety barriers as possible, strikes me as leaving this as a "high risk" proposal but without now any safety nets.

None of the discussion about Canberra's water2water proposal I have seen so far, have made me fee l any happier about its overall merits and safety. I think short-cuts on health and safety look like they are going to be taken. Even if this proposal goes ahead, it in my view should not start until we have a much bigger dam available to capture the recycled water. This will allow a much bigger dilution effects and much longer retention times to be available as natural protection barriers. A larger dam that can be kept "offline" for periods will also allow us to presumably quarantine any recycled water until we know it is "safe" by appropriate test results. Even without the bigger dam, we need some type of accredited monitoring system for viruses to be readily available and in regular use so that if a failure in the system occurred, we firstly know about it and we then can then try as best we can to keep any contaminated water out of our drinking water supplies.

Peter Collignon

Post a Comment