Re-Review of PONE-D-15-15642 “Wolf lethal control and livestock depredations: counter-evidence from respecified models” by Niraj Poudyal, Nabin Baral, and Stanley T. Asah.

My initial concerns and objections remain.

The authors claim that the major concerns expressed (strongly by reviewer #1, somewhat by reviewer #2, very strongly by reviewer #4 - myself) regarding co-linearity between their “most important” time series variable and: # of wolves, # of breeding pairs, # of wolves killed are of no real concern. I quote “The explanatory variables were somewhat correlated but not perfectly correlated, so multicollinearity is not a serious problem”.

Please see the below attached plots and R2 values of the collinearity. The R2 for wolves vs. year ranges from .91 to 0.97 – a near perfect correlation.












The same goes for breeding pairs vs. year:

And wolves killed vs. year:


In this case, as in many other ecological problems, time is auto-correlated with the biological parameters of real interest (population size, population growth rate, mortality rate, predation rate, depredation rate, etc. etc). That is why no ecologists use time as an explanatory variable for number of animals next year, harvestable surplus, or livestock depredations. Instead, ecologists use numbers of animals last year, reproductive rate, and number of deaths to explain numbers of animals or depredations this year. As I previously stated: time explains everything, but explains nothing.

Indeed, I quote the authors again “Nowhere in this rebuttal have we mentioned that time explains anything. We used the time index as a control variable”

But because time is so strongly auto-correlated with the real variables of interest – it cannot be used as a control either. For example, on Page 7, bottom paragraph, the authors state “Only the number of wolves killed remained significant in our replication: the other two explanatory variables of the original model were no longer significant. The sign of the number of wolves killed was negative in our case”.

In addition they forgot to mention – the sign of the number of breeding pairs at time t was positive. More breeding pairs actually means fewer depredations!

In other words – their model showed that number of wolves and number of breeding pairs of wolves had no effect of livestock depredations by wolves at all, or had a benefitial effect, if any (Table 1). Wow! This is a major paradigm shift! Predators have no effect (or a good effect) on the number of prey killed! It makes no difference if there are 10 wolves, 100 wolves, 1,000 wolves, or 10,000 wolves. Wolves have nothing to do with wolf depredations on livestock. Revise all the ecology textbooks and wildlife management plans. It’s a new world!

Of course, this is all absolute nonsense. The only reason these results were obtained was because the real biological variables of interest (# wolves, # breeding pairs, # wolves killed) were strongly auto-correlated with time, and time subsumed all of them. The wolf variables were simply smoothing variables added to improve the fit of the completely synthetic, non-biological time model.

Before submitting my original paper for publication in this journal, it was reviewed by wolf biologists from the Washington Department of Fish & Wildlife, Idaho Department of Fish and Game, Montana Fish, Wildlife and Parks, Wyoming Department of Fish and Game, and the US Fish and Wildlife Service. All of these biologists know that wolves and livestock depredations were auto-correlated with time, and I explained that is why I was leaving that variable out of the analyses. All understood my approach, and to my knowledge, agreed (since none suggested using time as a causative biological variable).

The authors of this paper conclude and argue that killing wolves reduced depredations next year - contradictory to my findings ….but they could have just as easily concluded and argued that number of live wolves had no effect (or a benefitial effect) – also contradictory to my findings.

The authors piled on layers and layers of statistical mumbo-jumbo and jiggery-pokery to try to explain their results, get published, and perhaps slay a well-known carnivore population specialist – but the bottom line is that this paper is biological nonsense. There are very good reasons that ecologists do not use time as an explanatory or controlling variable in their investigations – unfortunately these authors seem unaware of that practice.

Even some reviewers were swayed by their statistical jargon – so many readers would also likely be. Wolf haters and policy makers would jump on this paper as justification for widespread killing of wolves. This is a dangerous paper for wolves and ecology and I sincerely hope it is not published.

Sincerely
Robert Wielgus