Image for The case of the declining eastern quoll: correlation equals causation – or does it?

Eastern quoll populations have declined by more than 50% over the past decade, but their decline correlates with a number of possible causal factors

The case of the declining eastern quoll: correlation equals causation – or does it?

In a word, no. It is human nature to try and draw links between events that happen in tandem – it’s how we try and make sense of the world and understand why things happen the way they do. But we have to be careful not to prematurely mistake correlation as unequivocal evidence. We see this with circumstantial evidence in our legal system every day. And there are good reasons why this is the case, as there is often more than one logical conclusion that can be inferred from the same set of circumstances. Acting solely on circumstantial evidence may lead to innocent bystanders being wrongly accused of crimes they didn’t commit, just because they were unlucky enough to be in the wrong place at the wrong time.

Sure, sometimes correlative evidence is all we have to go on. But it should only ever be seen as a starting point for investigating potential links between cause and effect.

I’ll unashamedly steal an example that was given to our first-year ecology class, many years ago, in an attempt to make this important distinction to a class full of budding young scientists. Once upon a time, a Martian landed on Earth. The first creature he encountered was a dog named Ralph. He called out to Ralph, and like all good dogs, Ralph ran to the Martian when called. So, being a budding young scientist himself, the Martian did an experiment and chopped off one of Ralph’s legs, and then called out to Ralph again. Ralph still responded, although his response was a bit slower. So finally, the Martian chopped off Ralph’s remaining 3 legs and called out for a final time. This time, Ralph did not respond to his calls. The Martian’s conclusion: clearly Ralph needed his legs to hear, and by chopping them off, Ralph was no longer able to hear the Martian calling him, and therefore did not respond when called.

Now anyone with a basic understanding of anatomy and a little common sense would know that Ralph’s legs were in no way responsible for his hearing. But the loss of his legs did correlate with his lack of response to the Martian’s verbal commands, so it is easy to see how the Martian (who had no knowledge of a dog’s anatomy) may have made that mistake. It seems blatantly obvious when we look at things we already understand, but the challenge is to not make the same mistake in situations where we have incomplete knowledge.

Such is my life, or has been for the past 4 years. No, I haven’t had my legs chopped off by a Martian. I am a PhD candidate at the University of Tasmania, investigating the cause of decline of the eastern quoll in Tasmania. Eastern quolls are presumed extinct on the Australian mainland, and Tasmania is their last remaining refuge. In Tasmania, their numbers have been declining over the past 12-13 years, with declines of more than 50% statewide over the past decade ( for more details, see: http://www.publish.csiro.au/paper/AM13004.htm ). Trying to understand what is driving their numbers down is a mammoth task, and not one I am realistically expecting to answer single-handedly during the course of my PhD. The best I can hope for is to try and narrow down the list of likely suspects by collecting evidence to eliminate some potential suspects or (ideally) incriminate others.

So, where to start? First, I looked into why the eastern quoll disappeared from the mainland a bit over 50 years ago. The bottom line – we don’t know why they disappeared, but disease, foxes, poisoning, persecution, feral cats and even rabbits have all been implicated to varying degrees, with some variables carrying more evidential weight than others. But unfortunately, all this information is correlative – any one of these factors may have single-handedly been responsible for the demise of the mainland eastern quolls, or even in combination with other factors, but in the absence of any direct evidence (and the unfortunate absence of any time machines to go back in time and find out what actually happened), we can only hypothesise as to what drove them to extinction.

Next, I looked at what has occurred in Tasmania over the past 10-12 years – the period when the quolls were rapidly declining down here. Some of the potential causal factors mirror those implicated on the mainland (e.g. foxes, feral cats, disease) but there are also some additional factors to consider in Tasmania during that period, such as the decline of devils due to DFTD, severe and prolonged drought, introduction of novel poisons such as fox baits and new second-generation anticoagulants in rodent baits, and the extensive conversion of large tracts of foraging and denning habitat to plantation. There may also be unknown variables such as new emerging diseases or pathogens that are not even on our radar just yet. The quoll decline correlates with each and every one of these potential factors – so if we were to look at any one of these factors in isolation, we could conclude that it was responsible for the quoll decline. Alternatively, and probably more realistically, it could be a combination of 2 or more of these factors that are leading to the current quoll declines. But as this research is still ongoing, I will address only a few of these factors here.

Drought

To cut a long story short, I have ruled this one out. Some in-depth modelling that pulled in all the observational data for the quolls over the past 60 years, together with a suite of climatic and weather variables for each of those locations, has illustrated that quoll numbers have historically responded to weather fluctuations - not really surprising to anyone who knows the species and may have observed their numbers fluctuating a bit over the years. But what was surprising was that while the weather explained a lot of their fluctuations over time, it did not predict what has happened over the past decade. Quite the opposite, with quoll numbers predicted to increase over the past 10 years – the complete inverse of what we have actually seen in the landscape. Conclusion: there is something else driving their numbers down.

Fox baits

Again, I have ruled this one out. This was originally one of my highest candidate causal factors as, on paper, there is definitely a theoretical risk to the species. For an animal to be killed by fox baiting, there are a series of steps it must complete. The animal has to be exposed to the bait, encounter the bait in the landscape, consume and ingest the bait, then die (or suffer indirect effects, such as consuming a sub-lethal dose that may not kill a mother but could kill her pouch young). In Tasmania, fox baiting has been undertaken in much of the core eastern quoll habitat (predominantly the drier eastern half of the state) and hence would expose quolls to baits. We know that eastern quolls can dig up baits buried at depths of 20cm ( see http://www.publish.csiro.au/?paper=WR95077 ) and video on YouTube also attests to their ability to dig up buried pieces of meat ( this is a piece of beef, not a toxic fox bait: http://www.youtube.com/watch?v=Q0xiSWzFYww&feature=youtu.be ).  What we don’t know is whether they actually then consume the baits, or whether they die from that consumption. Again, in theory, this is possible as there are two studies that established an LD50 (the theoretical dose at which 50% of individuals will die) for the species’ sensitivity to 1080; the first is 3.73mg/kg ( see http://www.publish.csiro.au/?paper=WR9810385 ) and the second is 1.5mg/kg (see http://www.publish.csiro.au/?paper=WR9890131). Now these 2 studies were both done under laboratory conditions and used different methods. But if we take the worst-case scenario, an average 800gm adult female quoll would only need to consume less than half a single 3mg fox bait to receive a lethal dose. Even if she doesn’t die, there may be a risk to her losing her nursing offspring. So the theoretical risk is concerning to say the least.

But we know from extensive research undertaken on the mainland ( with spotted-tailed quolls and 1080 baits ) that the theoretical laboratory-derived risk does not necessarily equate to an equivalent risk in the landscape. Spotted-tailed quolls have a low laboratory-derived tolerance to 1080 of 1.85mg/kg ( see: http://www.publish.csiro.au/?paper=WR9810385 ). However, studies in areas with high-density populations of spotted-tailed quolls in NE NSW have shown that while quolls do encounter baits, dig them up and even chew them, those quolls don’t die. This has been investigated by radio-collaring quolls before, during and after baiting operations, and injecting a non-toxic biomarker ( Rhodamine B ) into the baits that leaves a fluorescent stain in the mouth of the quoll ( and its faeces ) if it chews the bait. The marker even grows out as fluorescent bands that can be detected microscopically in the quoll’s hairs and whiskers. These studies found that when cameras detected the quolls digging up and consuming the baits, nearly every time, the bait was later spat out and found out of view of the camera. Many quolls were found alive with Rhodamine B staining in their mouth and/or hairs, indicating that they had at least chewed the bait but had survived. Other collared quolls either did not encounter baits, or did not chew them. Some of these studies looked at buried baits, others at aerial baiting, some used fox baits ( 3mg ) while others used wild dog (dingo) baits ( 6mg ). And yet the quolls survived each of these baiting operations, even though baits were typically distributed in much higher densities than are used in Tasmania. So while the theoretical risk to spotted-tailed quolls looked concerning, it did not equate to a real risk in the landscape. In fact, the mainland research site has been heavily baited for both wild dogs and foxes for around 40 years, and it supports what is probably the highest density of spotted-tailed quolls anywhere in the country. So one could argue that the baiting has probably been of considerable benefit to the spotted-tailed quolls through removal of many of its predators and competitors ( dingoes and foxes ), enabling them to breed up to numbers not seen at sites where baiting has not occurred ( also alluded to in this paper http://rzsnsw.metapress.com/content/n3846m893483504h/ but this hypothesis is only one of several alternative explanations and requires further testing ).

For 1080 fox and dingo baiting impacts on spotted-tailed quoll related papers, see: http://www.publish.csiro.au/?paper=WR02107, http://www.publish.csiro.au/paper/WR05014.htm , http://www.publish.csiro.au/?paper=WR06076, http://www.publish.csiro.au/?paper=WR06151

And studies showing how the concentration of 1080 in buried baits rapidly deteriorates in a matter of just days ( http://www.publish.csiro.au/?paper=WR99031 )

But this is a different quoll species with a different sensitivity to 1080 and a different life history. We cannot extrapolate observations from spotted-tailed quolls to form conclusions about risks to eastern quolls. Hence, part of my PhD plan was to test the impact of 1080 fox baiting on eastern quoll populations in the landscape. Given the theoretical risk, I was not prepared to put any additional poisoned baits into the landscape. So I had planned to follow on the tails of the scheduled baiting operations already being undertaken by the Fox Eradication Branch. The plan was to replicate the studies of spotted-tailed quolls on the mainland, by radio-tracking eastern quolls before, during and after baiting, and injecting Rhodamine B into the baits to enable us to determine if animals had consumed baits and survived or died. Unfortunately, while this was still in the planning stages, quoll numbers crashed at the sites I was planning to use, before any baiting was done. Over the past 2 years, I have undertaken over 30 infrared camera surveys at a range of sites across the state and unfortunately, with the exception of Bruny Island ( which will not be baited ), there is no site where reasonable quoll numbers are persisting that would enable me to assess the impacts of baiting in the landscape. One or two individuals would not provide enough evidence to assess the impact on a population. So for now, whether the theoretical risk for eastern quolls equates to a real risk in the landscape remains an unknown.

However, in terms of whether the fox baiting is responsible for the recent decline, while the baiting operations correlate temporally with the declines (both starting around 10-12 years ago), they do not correlate spatially. Declines have occurred in eastern quoll populations at sites where no baiting has been done, or in some places a year or two before the baiting occurred at that site (as happened to both my proposed study sites). While we cannot rule out the possibility that fox baiting would affect some eastern quolls, we can rule it out as the cause of the recent population decline at numerous sites. If quolls have declined long before baiting has occurred at these sites, then something else has driven their numbers down.

That’s not to say whether it may have had an impact had the quolls still been there when the baiting went through later, but in terms of trying to identify the responsible culprit that is currently wreaking havoc on our eastern quolls, the evidence from many sites suggests it clearly isn’t the fox baits causing the decline at those sites.

Toxoplasmosis

Again, I’ve managed to rule this one out. But it did look very likely for a while, and we know this disease can be fatal to many native Tasmanian marsupials including eastern barred bandicoots, pademelons and wallabies. I have tested quolls for toxoplasmosis at several monitoring sites over the past 4 years. Interestingly, there is pretty much a 100% prevalence of this parasite ( Toxoplasma gondii) in the quoll populations at my 2 sites with declining populations, yet only around 15% prevalence at my one and only stable site on Bruny Island. If we were to rely solely on correlative evidence, then this would strongly suggest that toxoplasmosis ( the disease caused by Toxoplasma gondii ) was to blame for the quoll decline, with really high prevalence where quolls are declining, and really low prevalence where quolls are stable. But being the stubborn bugger that I am, I pursued this further, and after 4 years of surveying these sites nearly every 2 months, I have collected long-term survival data for each quoll ( which, if they are lucky, live between 2-4 years at most in the wild ). And the interesting thing is this - the animals that were positive for the parasite lived just as long as those that were negative. So clearly, the parasite and its associated disease are not reducing quoll survival. So while the correlative evidence is screaming that this disease has to have something to do with the quoll declines, long-term survival data indicates that this is not the case.

Feral Cats

But what the high prevalence of Toxoplasma gondii at the declining sites is showing us is a much higher exposure to the definitive host of the parasite, the feral cat. Camera surveys at each of these sites have confirmed this association with feral cats, so my current working hypothesis is that feral cats are looking like they may be contributing to eastern quoll declines. Now cats have been in Tasmania for over 200 years, so why have they waited until now to make their move? Well, there may be a link with the decline in devils. There is a hypothesis (currently being tested by some of my colleagues) that devils may have historically limited feral cat numbers through aggressive encounters and possibly predation of kittens. With the decline in devils, this may be allowing feral cats to possibly increase in abundance. But probably more importantly, even without an increase in cat numbers, cats may now be able to modify their behaviour in the absence of devils, shifting their activity from around dawn and dusk (to avoid devils) to now being active right throughout the night. This could be exposing nocturnal animals such as eastern quolls to a new predator and competitor that they rarely encountered before. So the loss of devils may have triggered new threats to quolls that were not present before.

image
Infrared camera surveys around Tasmania have revealed higher densities of feral cats at sites where eastern quolls have declined

One of my sites where I am monitoring quoll declines (near Judbury) has provided a working example of how cats can impact wildlife populations in a very short period of time. For the first 12 months of the study, we had really good quoll numbers at this site (18 different quolls trapped in 3 nights in May 2011). There was no sign of cats in traps, on camera or in spotlight surveys. A year into the study, around 6 feral cats suddenly appeared (in traps, on camera and in spotlight surveys) and the quoll numbers crashed. We trapped only 4 quolls, as well as 3 feral cats in 3 nights in May 2012 – a 75% quoll decline in only a year. Quoll numbers have continued to decline, but we still have a couple of quolls hanging on out there.

Again, this is purely correlative evidence, and it requires further investigation. But we have also seen a similar (if not worse) crash in bettong numbers at this site. Camera surveys in February 2012 ( prior to the cats arriving ) recorded 62 bettong detections in 3 weeks. Two months after the cats arrived, the number of bettong detections on camera had dropped to 52, and by October 2012 was down to 28. Subsequent camera surveys in December 2012, April, June and October 2013 have all recorded zero bettongs. They have completely disappeared. Again, this is just correlative, but with both quolls and bettongs declining at the same time as cats appearing, it definitely remains our main line of investigation.

There are still many variables that need more investigation, but time is a-wasting and quoll numbers are continuing to decline. So my primary focus to date (and in the immediate future) is to try and pin down the cause of this ongoing population decline. While the research is still continuing, I have managed to rule out 3 of the potential culprits in the case of the declining eastern quoll: drought, fox baits and toxoplasmosis. To have prematurely attributed the declines to any of these 3 factors based on correlative evidence alone would have been akin to locking up the wrong suspect, while the real criminal remains at large wreaking havoc on what is left of our eastern quolls.

So no, correlation does not always equal causation.

Cassy O’Connor: Cats laws: A toothless tiger

• Ethical Gambler, in Comments: Been in Las Vegas for a few days but better gambling odds here I think. Hope discussion has not closed? I hereby up the offer (ref #26 and #28)to Fox Eradication Program people or independent fox program supporters as follows - “I hereby bet $10,000 that the Tasmanian Fox Eradication Program cannot prove, beyond reasonable doubt, that its Foxoff fox poisoning has NOT CAUSED AT LEAST 15% OF THE POPULATION DECLINE OF THE EASTERN QUOLL in Tasmania.” Betting Rules: 1 All argument from both sides be in the form of written presentations in Tasmanian Times. 2 A jury of independent scientists will decide the winner.  The jury selection and its verdict decision process will be as in criminal trials, with scope for both sides to have a say in the jury scientists selected.  3. No new research to be conducted, but unpublished DPIPWE data can be used.