Kullu, in a recently published scientific paper, you’ve suggested that most snow leopard population studies existing today are biased. Can you explain what that means?
First, it’s important to clarify that by themselves, these studies we analyzed are robust. They’re not wrong. However, most of them tend to be biased in terms of the choice of their study areas. They mostly focus on the very best patches of habitat, and cover areas that are too small to be truly representative of the larger landscape they are set in. As a result of this, the snow leopard densities they report tend to be very high – because they are close-ups, rather than long shots.
Can you give an example of this?
Perhaps you can think of it this way: Assume you were to estimate the population density of King County, where Seattle is located. Now, let’s say you set up 30 camera traps in a couple of blocks in downtown Seattle to do so. When you analyze your data, you’ll probably arrive at a population density of something like 4,000 people per square kilometer. There’s nothing wrong with this calculation, all the math adds up. But is it representative of King County? I’d suggest it’s not, because large parts of the county are rural and relatively sparsely populated, and you haven’t sampled those at all. So, to get a more meaningful result, you’ll need to sample in spots in downtown Seattle, in the suburbs, and towards the Cascade Mountains – or choose your sampling location completely randomly in the county. If you did that and then applied the same maths to your data, your population density estimate will suddenly look very different; probably closer to 400 people per square kilometer – which is much more representative of King County.
Now, of course that’s an extreme example, but it’s not far off the mark. We all understand immediately that humans aren’t distributed evenly across the map, and that there are population centers as well as areas that are relatively deserted. From our data, it seems pretty clear that snow leopards aren’t distributed evenly across their habitat either. It’s just not as obvious, and we don’t understand it nearly as well. But we need to take this fact into account to avoid bias in our studies.
Why is this type of bias a problem?
The snow leopard’s global range is huge, and less than 1% of it has ever been surveyed with solid scientific methods to estimate how many cats there are. These studies are a really small snapshot of reality. But they’re also all we’ve got right now. People often ask how many snow leopards there may be in the world, and the best answer we can give them based on the very limited data we have is somewhere between 3,500 to perhaps 8,000, but we don’t really know. Now, if there is any bias in our data, our picture of reality will be very skewed! At the same time, major conservation decisions are being made on the basis of this data – decisions that will impact snow leopards and their habitat for generations. What if there are actually fewer snow leopards than we think because most existing studies are biased towards high-density areas that don’t accurately reflect reality? It’s impossible to say for sure, but our research shows it’s at the very least a real concern. We can’t afford to be wrong.
How can this bias be avoided in the future?
There are two things researchers can do: Firstly choose study sites that are representative of the larger landscape, rather than just the best part of the landscape. Secondly, sample in larger areas, ideally larger than 500 square kilometers. Or sample several randomly chosen small parts of a larger landscape, and look at the aggregate data from those places, rather than extrapolating from one small area to an entire landscape.