Reading group: An Italian wolf in Switzerland

wolfIn reading group this fortnight (March 16th) we talked about the social acceptability of Italian wolves in Switzerland based on a recent article available online at Journal of Applied Ecology: Behr, D.M., Ozgul, A. & Cozzi, G. (2017). Combining human acceptance and habitat suitability in a unified socio-ecological suitability model: a case study of the wolf in Switzerland.

I (David Duncan) chose this paper because many QAEco folk have more than a passing interest in models that explain and/or predict the current or potential distribution of species, and their application for the conservation of biodiversity.  Fortunately, the article was well written and in general clearly explained so it was not an onerous read.

I was struck by the strong opening gambit – i) that a Habitat Suitability Model (HSM) without sociological data on human acceptance toward the focal species is deficient; ii) that such HSM misrepresent observed processes (I think this is a reference to colonisation); and iii) that such HSM lead to inappropriate management – and I wanted to hear what people thought about the article.

Behr and co-authors specifically and uniquely referred to Habitat Suitability Models (HSM), but from a technical standpoint HSM are indistinguishable from what might be called Predictive Habitat Models (Guisan & Zimmerman 2000) or Species Distribution Models (Elith & Leathwick 2009), so I took the whole class of models to be tarred.

First things first: a premise well and truly overcooked
The premise was demolished in the first moments of the discussion. Whereas the authors assert a deficiency of HSMs and a solution via the “integration” (combination would be a more accurate term) of sociological data, the response is of course that one would choose an approach depending on the scientific objective or question at hand.

Everyone could see the value in a surface that i) permits consideration and analysis of where a species range expansion could generate conflict with the populace, and ii) helps generate management responses – in fact some amongst our lab have been involved in studies with similar root concerns (e.g., Whitehead et al 2014 and related blog post). However the idea that HSMs as a rule are deficient without spatial representation of human acceptance was rejected.
How did this paper get through peer review with such bluster intact?

Behr et al’s study
They combine the published HSM with a model of human acceptance (of the wolves) the “HAM” to create a Socio-Ecological Suitability Model (SESM).  The SESM affords power of veto to the human acceptance data, where a binary variable for acceptance cancels out otherwise suitable habitat through multiplication (1 for habitat suitability x 0 for acceptance = 0).  Powerful stuff, so what’s in the HAM?

A survey about the wolves was distributed to 10 000 dwellings located near points dropped randomly over the country, a process that saw 63% of all Swiss settlements included.  The authors state that spatially representative sampling was thus ensured, but of course it’s the survey returns that decide whether it was achieved or not. One third of the surveys came back, which seems good for social survey returns, and accounts for 42% of settlements by my calculations. However, we don’t learn about the spatial representativeness of those returns.

A suite of interesting questions about social acceptability were asked, including experience or encounters with wolves, perceptions of harm that wolves could do to people and livestock, perceived proximity to wolves, etc.  However, and it took a while (for me) to realise this, the response to only one question was used in the spatial model of social acceptance. Rendered in English that question was:

Are you in favour or against wild living wolves in Switzerland?

for which a Likert style five-point response scale was available ranging from in favour [1] to against [0], stopping all stations via ambiguous [0.5].

Zombie wolf jokes to one side, those with more exposure and experience in collection and analysis of social data in ecology felt that the paper lacked a strong theoretical framework to underpin the social component of the modelling. This absence was more striking given the research high-ground asserted in that all-important opening gambit.

On that combination of HSM and HAM for SESM
The authors were at pains to say that the two scales for probability of habitat suitability and probability of social acceptability were not directly comparable and thus couldn’t be multiplied to obtain a continuous socio-ecological suitability score PSESM. Their solution was to convert the scales into binary variables based on the threshold value where the models classification error was lowest (the sum of specificity and sensitivity were highest), and then combine these to identify areas where habitat suitability and social acceptability were both high. However, in effect they multiplied the two binary variables to calculate PSESM. No one in our discussion could speak to how that improves on multiplying the original scales.

The combination of HSM and HAM might have lost its way trying to stay too close to the analytical framework of a HSM. An alternative analysis that could tackle the implied trade-off head-on might have been to seek a certain area of total habitat that would optimise HSM and HAM according to pareto efficiency. Such approaches have been widely implemented to assess trade-offs and synergies between different goals and features (a quick sweep of the last few years suggests Bryan et al 2016 for carbon storage and biodiversity; Jacobsen et al. 2017 for conservation and profit from a fishery; LaFond et al 2017 for timber production and biodiversity conservation).

A bit parochial
In lots of studies for reasons of practicality we researchers work at a smaller scale than might be ideal for the question. In this case, given that the original HSM spanned the Alpine Range of which Switzerland forms a central part, and that the authors in the introduction contrasted the wolf’s distribution in Switzerland compared with its neighbours, it would have been really cool to see how the acceptability data and model performance varied over that same geographic range.

Having said that, no doubt a range of non-trivial practical and social science considerations would arise in trying to extract uniform or comparable social preference data across the entire Alpine range.

What should a manager do about this?
The article’s discussion was surprisingly mute on how the results of this analysis and the map products should be interpreted and acted upon by managers. This stung a bit as a few present had been rejected by this very journal on that criterion!

What should researchers do about this?
Instead, and despite not really exploring the end-use, the authors suggested that on the basis of this paper, many researchers would re-evaluate their HSMs where polemical species such as large carnivores are involved (I can imagine that someone may well do this for the dingo in Australia, if they haven’t begun the process already).

The implications and take home messages presented in the article were for other researchers, whereas Behr and coauthors presented their study as being without significant limitations.  From this summary you’ll gather that our group wasn’t of similar mind.

A consensus?
No one was in any doubt about the importance of the social acceptability to biodiversity and conservation.  I for one was oblivious to some of the disputable technical issues, and was happy enough to overlook the ones I could see.  Most would have preferred that HSMs were not routinely imbued with human acceptability data, as HSMs and SESMs to use the nomenclature of the authors, have distinct purposes.
I think everyone bristled at the premise and pitch. Oversell attracts the ire of a reader and puts a great deal of pressure on authors to make sure their article can back it up, and in this case we felt that the article didn’t have the substance to justify the rhetoric employed.

What can we learn from this as a group?
It was suggested that a useful exercise for our group might be to conduct quick, mock peer reviews of articles to later share and compare. Through this process we might gain some insight into the different aspects of a manuscript that catch people’s attention, and we can learn to be more diligent and efficient reviewers.

* In fact, I see that Falcucci and coauthors in the 2013 HSM article noted that human population density might inadequately speak for heterogeneity in human attitudes, that “sharp differences in tolerance towards wolves … [could] be extremely important in shaping the distribution of the species”. Curiously Behr and coauthors didn’t reference that directly in building the case for their own study.
This also makes it clear that that human density was used as a predictor in both the HSM and the HAM.

This entry was posted in Reading Group and tagged , , , . Bookmark the permalink.

One Response to Reading group: An Italian wolf in Switzerland

  1. We studied this paper in our Journal Club two weeks ago. Its merit, according to the audience, is above all to show where there are gaps in the country in terms of population education regarding large predators. Actually it points to areas within Switzerland where the cantonal authorities communicate negatively about large predators (wolf, lynx) since decades (by doing so they are ‘illegal’). Then we should first educate these local authorities even before the populace. This concerns Valais and Central Switzerland in particular. But we know that it in this country already.

    Raphaël Arlettaz, University of Bern

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s