Destructive sampling. It’s a term that makes you squirm a little isn’t it? I feel the same when I think about certain marking techniques or the need to sacrifice specimens for Museum collections. Each presents a dilemma that most ecologists will confront at some point in their careers: the need to do something that is immediately at odds with our core values in pursuit of the greater good for species conservation.
In our latest paper* we tackle the dilemma posed by destructive sampling using a decision-theoretic approach. To be clear, the dilemma here is that some species are very difficult to detect by any other means than pulling apart their favored microhabitats. Hence, gaining the information we need to manage these species is nigh on impossible without some impact on their habitat. We can’t find out how widely they are distributed, we can’t establish the range of habitats on which they rely, and we can’t quantify how they are responding to the management we apply. The focal species of our paper is an excellent example. The Earless Skink (Hemiergis millewae) is an obligate denizen of Spinifex hummocks in the Mallee regions of southern Australia. In Victoria the species is listed as critically endangered, and there is real concern that fuel reduction burning poses a threat to remnant populations. Yet we also don’t know where all the remnant populations are, let alone how persistence relates to fire regimes. To gain that information requires surveys across the Victorian Mallee, but the only practical way of doing those surveys is to prize apart the Spinifex hummocks on which the species relies.
Stefano Canessa took up the challenge of designing surveys for Hemiergis, using a hard-earned survey dataset acquired by Peter Robertson and Ian Sluiter at select sites in the Murray-Sunset National Park. The approach Stef devised finds the number of Spinifex hummocks a surveyor must sample at a site to ensure a threshold detection probability is reached (0.9, 0.95 etc), using a weighting system to reflect a surveyors choice between minimizing the number of hummocks searched and minimizing the quality of hummocks searched. This last bit is important, because Hemiergis don’t use hummocks at random; they like the big ones. So a surveyor can reduce the number of hummocks searched at a site by targeting the biggest ones available, because detection probabilities are higher on a per hummock basis for the bigger ones. But of course this removes the best microhabitats for the species. Instead, a surveyor can target small or medium-sized hummocks and leave the big ones, but this requires removing more hummocks to achieve the desired probability of detection and perhaps leads to greater impacts in the long run (because big hummocks senesce and need to be replaced by adolescent hummocks).
It’s a tricky problem, but Stef’s technique makes the trade-offs and choices clear, and enables repeatable and transparent decisions to be made (as decision theory is intended to do). Stef even provides an Excel worksheet to run the decision analysis, making the approach immediately accessible to managers. But the technique also has wider appeal. Destructive sampling is used to survey a range of species, and the fundamentals of Stef’s approach apply in each case, because all involve a central trade-off between maximizing detection probabilities and minimizing impacts on the focal species’ habitat.
*This link will take you to a post-print version of the paper, rather than the published version, because PLOS completely mangled our figures and refuse to fix them. The published version can be found here.