Wednesday, January 16, 2013

Are Foxes Widespread And Established In Tasmania?

Advance Summary

1. A recent refereed paper claims that foxes are widespread and established in Tasmania and likely to become increasingly so.

2. This paper uncritically accepts some data items that, while not discredited entirely, are in some cases very dubious.

3. This raises concerns about whether the authors are familiar enough with all the data they use to be very confident that most of the data items are valid.

4. Similar concerns relate to the continued presentation of some dubious fox data items as hard evidence by the Tasmanian government.

5. On this basis, both scientists publishing on the question of foxes, and the Tasmanian government, need to present data accurately and with reservations properly discussed and explored if their claims about foxes are to be considered credible.

6. At present neither the claims that foxes are widespread in the state, nor the claims by some "sceptics" of a conspiracy to import fox evidence in order to fabricate fox presence, are convincing.




 ---------------------------------------------------------------------------------------------------------------------

This may seem a strange subject for a psephology/politics site to be covering.  At least, to interstate and overseas audiences, it might appear very obscure.  But in Tasmania, the presence or absence of foxes in the state is the subject from time to time of front-page articles (most of them abysmal), and debates in parliament, and even opinion polls! (Alas, I didn't keep the findings, but I think the fox's existence had around a zero netsat.)   There is a lot at stake either way, with a potential risk of extinction of small mammal species already lost or nearly so from the Australian mainland if we listen to the "sceptics" and the sceptics are wrong.  On the other hand, if we listen to the "believers" and the believers are wrong, that's tens of millions of dollars of public money down the drain on an overpaid equivalent of Donkey-Work-For-The-Dole, to say nothing of the inconvenience and the possibility of collateral damage to native species arising from poisoning programs, and the eventual damage to the credibility of government departments.

In The Blue Corner ...

This is an issue on which I have very different opinions of the average qualities of the people on both sides of the debate, but it doesn't stop me from thinking that both extremes are unconvincing. For instance, of my many friends in the Tasmanian natural history community, only three, to my knowledge, openly question the official story about a widespread Tasmanian fox presence (while twice as many believe that they have sighted a Tasmanian-resident fox, and in some cases I even believe them.)  The "believers" include some of Tasmania's best government scientists, including some excellent people with whom I am on good terms both personally and professionally.  The "sceptics", with a few similarly fine exceptions, are a rabble of old-style conservatives, shooters, conspiracy theorists, and some people with some very odd ideas about science (in most, but not all, cases with the excuse of having no scientific education).  As just one example of the kind of strange output we sometimes get from Team No-Fox, I give you Ivan Dean, MLC for Windermere:

"In fact there is a Tasmania fox yarn that hit the headlines in the same month the Twin Towers of the World Trade Centre in New York was attacked in September 2001, the only tragic difference being that the collapse of the Twin Towers in New York was a reality."

He comes out with this sort of stuff about foxes on unrelated issues too. There is even a drinking game joke about it.


Another common class of fox-sceptic (especially on Tasmanian Times) is the anti-1080 obsessive.  1080 is a poison that has in the past been used to kill common species of cute fluffy mammals during forestry replanting, and is now used to "poison foxes".  Instead of sticking to a credible line of objection to the toxin based on animal welfare arguments, or trying a serious scientific argument about the impact on species at actual risk, the anti-1080 obsessive prefers to blame the toxin for anything and everything (especially, Tasmanian Devil Facial Tumour Disease), accuses anyone who disputes their arguments of supporting the use of 1080, and then charmingly invites them to consume the substance in question.  The cognitive dissonance involved in accepting that 1080 even might have a valid use being far too much to bear, the obsessive concludes that if 1080 is used to control a threat, the threat cannot exist. 

The breakdown does become a bit easier if I split the class of "sceptics" into two groups: (i) those who are unconvinced of the case for widespread fox presence, without necessarily taking any firm view as to how that case has come about and (ii) those who assert foxes are definitely absent from the state and that there has been a conspiracy to import and concoct fox evidence in order to obtain Commonwealth funding.  Most of those I don't abide go in category (ii).  Clive Marks essentially refers to group (i) as "critics" and group (ii) as "sceptics".  Another approach is to remember that scepticism (or, at least, "skepticism") is a movement with a fine tradition and to label group (ii) as "denialists".

The Battlegrounds

There is very little common ground between the furthest extremes of the debate.  Two things practically everyone involved agrees on are the following:

(i) At least two individual foxes escaped within the state of Tasmania within the past 15 years.  A fox was seen by numerous witnesses in Burnie in 1998 and footprints were found soon after.  Another fox was apparently seen by witnesses escaping from a container at the Agfest event near Longford in 2001 and this was followed by high-quality sightings (one of which is universally accepted) and a nearby footprint collection in subsequent months.

(ii) Of the various carcasses claimed to have been collected in the state since, some are hoaxes (carcasses imported from interstate then claimed to be local), including some confessed hoaxes.

The evidence claimed by the "believers" and generally rejected or disputed by the "sceptics" includes:

* Four carcasses - the "Symmons Plains fox" (2001, supposedly shot), the "Burnie fox-in-box" (2003, supposed roadkill), the "Lillico fox cub" (2006, supposed roadkill) and the "Glen Esk fox" (2006, supposed roadkill).

* One skull purportedly found at Interlaken (2008/9 and reported 2010, the subject of this piece of overblown reporting which hugely exaggerated its significance to the issue.)

* Blood claimed to be tested as fox blood collected following a poultry kill at Old Beach in 2006.

* At last count, about 61 scats tested as positive for fox DNA, of which one (Burnie 2002) contained fox grooming hairs.  Some were collected during generalised carnivore scat surveys and some during follow-ups to specific reports of fox presence.

* The higher-credibility end of thousands of public "sightings".

The trigger for this article - which I've considered writing for a while now - was my second round so far of Twitter discussion following the publication of this paper.

Sarre, S. D., MacDonald, A. J., Barclay, C., Saunders, G. R., Ramsey, D. S. L. (2012), Foxes are now widespread in Tasmania: DNA detection defines the distribution of this rare but invasive carnivore. Journal of Applied Ecology. doi: 10.1111/1365-2664.12011 (henceforth "Sarre et al") was supposed to be the long-awaited refereed paper that settled the question of fox presence in Tasmania once and for all, showing both that foxes are widespread in the state and that they are spreading.

With a longstanding interest in the various physical evidence incidents (most of it, as it happens, coming from arguing against conspiracists who claimed to have proof they were hoaxes), I was interested to see this paper's treatment of the various claimed evidence incidents.  Here's an example:

"Although the deliberate or accidental nature of this introduction remains unsubstantiated (Saunders et al. 2006), the discovery of four fox carcasses including males, females and immature individuals (Fig. 1) has confirmed the presence of foxes and suggests breeding populations in the central north coast and midland areas. "

At all times in the Sarre et al paper, the four disputed carcasses and remaining disputed non-scat evidence items are assumed to be hard evidence and accepted as modelling inputs with no questions asked whatsoever.   But for those familiar with the debates surrounding these items of evidence this is just not credible. 

The Symmons Plains fox is the most dubious of the four claimed carcasses.  There is no publicly available proof it is a hoax, but the very many caution flags include the colorful nature of the claimed shooter (who had been prosecuted for poaching) and the claimed shooter's belief he was entitled to a large reward. Even the piece of evidence that seemed to nail it as Tasmanian, the claimed presence of partial remains of a Tasmanian endemic long-tailed mouse , has over time become a Fawlty Towers rodent story, starting with questions about whether it would have occurred in the area where the fox was supposed to have been shot (by no means a conclusive objection, but enough to trigger speculation that the fox might have been shot well away from the site and moved) and ending ignominiously with the star exhibit's apparent disappearance.

The Lillico fox cub is a juvenile fox found thoroughly squashed and well and truly dead on a major road, apparently following an anonymous call reporting seeing it there months earlier. Apart from a shooter who reported seeing two foxes in the area at the time, and then later claimed he had shot one and placed it on the road to be found by others (something not stated in his original claim), there is no known claim about when this fox died (ie nobody has reported running over it.)

The Burnie fox-in-box is a purported roadkill found on a major road near the site of the original 1998 escape (but too long afterwards to be the same fox).  Apart from debate about the freshness of the carcass, it is not clear whether the fox was really roadkilled or killed elsewhere (eg on board a boat) and dumped.  Even if is assumed the fox was roadkilled, it remains unclear whether this is a fox that had been free-living in Tasmania for a substantial period or else one that had died shortly after coming ashore.  If it had been free-living, its having never bred counts against it having met any great number of fellow foxes in its time in the state.

(Both the above foxes have been the subject of persistent strawmanning from some of the "sceptics", who have claimed that there has been official acceptance that these carcasses were moved (in the Lillico case, Nick Mooney's testimony to the Public Accounts Committee is the supposed basis for the claim).  In both cases the "official" comments referred to indicate only that the carcasses may have been moved, not that it is officially concluded that they were.)

The Glen Esk fox is another much-debated apparent roadkill.  After it was run over and reported as such, initial reports that it had been run over alive by those who reported it turned out to be at best fog-of-war confusion from an uninformed staffer.  It was later decided it had probably died the day before, perhaps in a separate roadkill incident, which might not have occurred in the same place.

In summary, for all four foxes there is doubt that they were killed where they were found and for two of the foxes there is serious doubt that they were even killed anywhere near where they were found or even inside Tasmania at all. (All this is even more dubious in the case of a skull which was not reported at the time of its supposed finding on top of a tree stump.)   Yet the Sarre et al paper accepts them without comment as having died where they were found:

"Locations of physical evidence of foxes (scats, carcasses, skulls) were used to build a model of fox habitat suitability using a maximum-likelihood method developed for presence-only data (Royle et al. 2012) with the ‘maxlike’ R package (Royle et al. 2012)."

So?  Do A Few Dodgy Data Points Matter?

I assume - in the co-authors' favour - that they accepted these dubious data points as modelling inputs without comment because they did not know they were so contentious.  And it can easily be argued that since it is only a few data points out of 65 (most coming from scat records), if the rest are fine, the conclusions will be much the same, so there's no problem.

However, if the authors of a paper are willing to accept dubious data points as inputs without comment because they are not properly across the reasons for caution about doing so, this raises the question of whether their knowledge of the remainder of the issue is up to scratch.  In particular, do they know enough about the scat data to be strongly confident of correct handling at all stages of the chain of custody of the tested items?  Perhaps they do know this, but if an author handles one aspect of an issue too casually for any reason, it should not breed confidence that they will handle any problems associated with the rest of it as stringently as needs to be the case.  This problem also arises with state government output listing dubious carcasses like the Symmons Plains case as "hard evidence" without reservation. 

The Scat Issue

Fox faeces "confirmed" as such through DNA testing are central to the question of whether foxes are widespread and established in Tasmania.  If all or most of the scats collected are in fact fox scats (or scats of fox-eaters) collected from the Tasmanian environment (without being put there) then the central question is answered - at least to the point of establishing that foxes have been and probably are very widespread in the state - and we don't need to worry about the source of other forms of alleged "hard evidence".

The scat issue is also important because only the potential confirmation of scats as fox scats stands between Tasmania and a rerun of the Isle of Man fox farce, in which the prevalence of foxes on another ecologically sensitive island was hugely overestimated.

Determining whether each of the scat results should be considered reliable is beyond either my level of knowledge of the program or my understanding of DNA-testing issues.  However, again, many questions have been raised about the reliability of scat results.

One of these questions concerns a scat supposedly collected on Bruny Island.   Bruny has a population of about 600 people and is serviced by a regular vehicle ferry from Kettering.  Some of the "sceptics" have attempted to write off this single scat on the basis of the island being too far from the Tasmanian mainland for a fox to swim.  But we already know that at least two foxes found their way into Tasmania generally despite Bass Strait being much wider than the D'Entrecasteaux Channel, so Bruny Island becomes a sort of microcosm of the whole Tasmanian fox question.  If the result is real, then that means either that a fox was deliberately introduced to Bruny Island alive (raising the question of why anyone would bother) or that a fox accidentally introduced itself to Bruny Island, presumably by stowing away in a large vehicle crossing on the ferry.  The latter event seems extremely unlikely if foxes are very rare on the Tasmanian mainland to begin with.

"Unlikely" isn't "impossible", but other things aren't impossible either, such as contamination, recording errors, or mixing up of samples - even assuming perfect lab methods when scats are sent off and then tested by the methods discussed in the Sarre et al paper.  Has there been a sufficient review of processes to eliminate all possible alternative explanations for the presence of this lone scat on Bruny Island?  Who knows?  What is known is that there has been no other evidence published of fox presence on Bruny, except "sightings", which are not reliable evidence.  Too many other things look at least vaguely foxlike and it is too easy for people to mistake what they see, or to think that they have seen what they fear.

The Bruny Island scat is a problem for modelling because if it is an error, it raises the possibility that other scats are also errors.  It could be an example of a more systematic error, affecting many scat results, that only stands out as suspicious because of its unusual location.  At the very least, given the problems already discussed with Tasmanian government presentation of the carcass evidence, it presents a case for a thorough review of all scat evidence to ensure that there are few if any errors at any stage.  Once again, in the Sarre et al paper, the Bruny Island scat anomaly isn't even directly discussed.

Other concerns that have been raised concerning the scat findings include:

* That to this stage, where fox scat testing has been sufficient to determine whether scats came from the same fox or different foxes, all scats have come from different foxes.  There have been no events of multiple scat collections in the same area that have tested as coming from the same fox, although this would seem likely to occur (especially if there were breeding foxes that were anchored to a given den location).

* That there has been no properly confirmed instance of fox scats being found at baiting stations where baits are laid to attempt to kill foxes.

* That no endemic Tasmanian mammal (for instance the very common pademelon) has been found in testing of hairs examined from fox scats.  I think this objection is extremely weak.  Only a relatively small number of native mammal species have been identified from hairs inside scats, and of these none are absent from Tasmania either (as might be expected if the scats were all imported).  Indeed, one species included in the list of species of which hairs were collected is the barred bandicoot, which is practically extinct in Victoria but reasonably common in Tasmania (the latter fact being completely ignored by the televised version of a recent 7:30 story, the online version of which has been corrected.) 

Another curiosity is the lack of fox presence in some urban areas.  The Sarre et al paper testifies to this:

"The strategic scat survey comprised three phases, each of which covered a single geographic region on the eastern and northern Tasmanian mainland and contained between 200 and 300 9-km2 units. Urban areas were not considered in the survey as these areas are heavily populated, making sightings, and the gathering of hard evidence of presence more likely than in less densely populated regions, reducing the need for formal survey."

Quite aside from the absurdity of considering claimed "sightings" to really reduce the need for formal survey, this raises the question of why hard evidence of presence has not been recorded in the biggest urban areas, which foxes elsewhere are known to inhabit readily.  This especially applies when we consider the well-known fox evidence map.  Nothing from Launceston and the single (and also disputed) blood record on the outskirts of Hobart.  Would foxes really exist on both sides of Hobart and yet have never been detected in the city, instead sneaking around non-core habitat in the Wellington Range? Seems unlikely.

GIGO

I do not intend to discuss the way in which the Sarre et al paper has determined its conclusions based on the records it has admitted as valid.  A potential problem for any model is that of GIGO (garbage in, garbage out) - if many/most of a model's data inputs are incorrect, then it doesn't matter how great the maths behind the modelling is, the model's outputs are rubbish.  Given that the paper accepts dubious carcass records as data inputs without comment, and given that there are valid concerns about the possibility of error in the scat data, I do not accept that it has scrutinised other data sufficiently to be confident that the data used are valid.

This does not mean that I am writing off all claimed "evidence" for fox presence in Tasmania beyond the two undisputed specimens, by any means.  It is possible, for instance, that some of the hard evidence is valid, while much of it is affected by as yet unidentified errors.  It is possible, for all I know, that much of the hard evidence is valid with only a few items that are debatable enough to need to be rejected, and that the many discrepancies can be resolved.  And I take seriously a very small minority of sighting reports, such as the following, by Jim Nelson, one of Tasmania's best field naturalists:

"Allegdly [sic] Russell? You should know me better than that. I have spent my life observing wildlife, and I would not say I saw a fox unless I actually saw one! I grew up with foxes and I know how incredibly elusive and cryptic they can be, and I certainly know a fox when I see one at 30 M away.

The sighting was at Epping Forest in the Tom Gibson reserve. It was a field naturalists outing looking at the orchids and other features in the reserve. 

The fox was moving in a typical crouching manner trying to use as much cover as possible. At first I couldn’t figure out what it was, but then I had a full on side view for a couple of seconds and the last thing I saw was the unmistakable fox’s tail. 

Another field naturalist saw the same fox independently from me as it moved away through the bush towards him. As I looked around for scats, prints or a kill, he came upon me looking shocked and said “I have just seen a fox!”

We reported it immediately to the fox task force by mobile phone. Then we looked around for scats, but didn’t find anything but wallaby scats. I believe the task force went out the next day with a dog, but didn’t find anything."

But the possibility of correctness of most data should not be enough for formally published science that draws a major conclusion.  All aspects of the program including the chain of custody of data sources, need to be dealt with much more comprehensively for any presentation that attempts to establish its findings as scientific fact.  Frankly, the Sarre et al paper's ready acceptance as fact of disputed data items that have been debated in Tasmania for years is an indictment of the standard of external refereeing of the paper.  But that's scientific refereeing for you.  Unfortunately it's nowhere near as reliable or consistent an institution as it's cracked up to be.

Some Further Reading/Viewing

Most of the ongoing fox "debate" on Tasmanian Times - over 200 threads of it - is rubbish.  Actually, for the most part it has been not so much a debate as an ongoing strategy by a few "sceptics" to overwhelm opposition with the sheer volume of belligerent repetition of essentially the same posts.

The Fox That Wasn't There? by Dr Clive Marks stands out, generally, as an intelligent and scientifically informed presentation of reasons to be concerned about the case for fox presence in Tasmania.  See also Nick Mooney's response,  and Clive Marks' follow-up .  The debate contains useful arguments about the pitfalls of invoking the "precautionary principle" too readily.  (One of my views is that the "precautionary principle" even when correctly invoked is extremely overrated, and often foolishly mistaken for science when it is actually a moral claim that is sometimes self-contradictory, but that's a story for another post.)

The Fox Files Uncut by Simon de Little provides an introduction to many hours of footage de Little compiled and then released, having worked on a documentary about the issue which was ultimately abandoned.  de Little's main conclusion is very broadly similar to mine - that there is no actual evidence of systemic hoaxing but that there is plenty of bumbling on both sides of the debate.

Addendum 2-4 May 2014:  Clive Marks et al have now gone live with a website at tasmanianfox.com which is about to add links to their peer-reviewed publications.  The very thoroughly prepared and presented website holds, among other things, that the fox scat results were false positives that could have arisen from testing errors or contamination.  Defenders of the program have claimed in response that the genetic testing done by the tasmanianfox.com group tested only one part of a two-stage testing method.  I await and will report replies to this.  Of course, that reply does not exclude contamination.

A slightly disappointing aspect of the reporting of the tasmanianfox.com effort has been the depicting of it as an "independent review", in a way that could be read as implying it was one commissioned as such by the government.  While it is indeed a review that is indeed independent, that's not the full story.  It is in fact a review spearheaded by two scientists who have long-known disagreements with aspects of the official story, though they do not necessarily share all the same disagreements, and though some of the scientists involved have no prior history of involvement in the issue.

9 May 2014: The transcript of the admirably thorough Background Briefing story sheds light on the two-stage testing method story, with Stephen Sarre stating that there is the possibility for false positives at the second stage as well, albeit at a probability more like 0.4% or lower, which is lower than the implied probability from the actual scat tests (56 claimed positives from nearly 10,000). 

Bearing in mind that the scat evidence is the only evidence thus far advanced that, if it stands up, proves the existence of foxes in the state, I think it might be interesting to imagine that the scat evidence was the whole of the debate and that there was no possibility of contamination.

We might ask the question: what is the probability that foxes are widespread and present, given that we have 56 claimed positives? It might be as high as 99%.  Next, we might ask what would be the a priori chance that a breeding population of foxes had established in Tasmania in the last decade, assuming we knew nothing about the claimed evidence for whether they had done so.  This is a very difficult question to assess but given that we have had many decades of opportunity, most of them with laxer biosecurity than now and even with proven releases of foxes, we would probably put it at less than 10%. 

Bayesians will know where this is going: the chance that the whole scat test result is one big false positive rises when it, if true, is evidence for something that was otherwise improbable.  Not by enough to alone make the absence of foxes the most probable explanation, but enough to say that the scats are just evidence of some degree of fox presence and not proof of foxes being widespread.  But if contamination is even slightly increasing the number of false positives, even that disappears.  And that is before we even consider the P (B|A) component of Bayes' Theorem: the question of the probability of the given test results given the assumption of widespread fox presence and the claims that a widespread fox population would almost certainly produce different findings (such as multiple scats from the same fox, clusters of scats and so on.)

Throw that in the mix and the probability of widespread established fox presence may well become very low.  And honest error combined with sloppiness are more than enough to explain the pattern of apparent "evidence" without needing to resort to any tinfoil-hat material.


------------------------------------------------------------------------------------------------------------
Note to potential commenters on this article: no comments alleging fraud by any individual or body will be accepted.  If you have firm evidence to back such claims, which past experience suggests you don't, you can take it to far more influential venues than this, and if you don't, I'd rather you were sued, not me!

1 comment:

  1. Hi Kevin, I stumbled across this article while preparing a piece for TasTimes. Well done! I congratulate you on this excellent summary/comments of the Sarre contribution, and of the broad sceptics' position. - Garry Stannus.

    ReplyDelete