How LIFE works

Relating area of habitat to extinction risk

The LIFE metric is an attempt to estimate the impacts of different human actions, positive and negative, on species’ risks of extinction. No species persists forever, of course, so LIFE looks at these human-caused changes relative to a species’ probability of extinction in the absence of human impacts. And it does so by focusing on the biggest of those impacts–changes in the extent of the habitats which a species lives in, its so-called Area of Habitat or AOH. Specifically, LIFE uses and then analyses estimates of the AOH a species would have if humans were absent, the current AOH available to a species in the presence of humans, and the change in AOH caused by a further marginal (i.e., relatively small-scale) land-use change involving habitat conversion or restoration.

If we consider a theoretical species with a range (i.e., the approximate area in which experts expect the species to occur) given in light green, we derive the three quantities of interest as follows:

1. Human-absent AOH: we select the parts of the range which are within the species’ habitat and elevation preferences based on likely habitat classifications in the absence of humans; this is the region in purple.

2. Human-present or Current AOH: we select the parts of the range which are within the species’ habitat and elevation preferences based on current habitat classifications; this is the region in orange.

3. Further change in AOH: to investigate the impact of human actions, now we estimate the effect of converting (or restoring) current AOH by a relatively small amount–the area lost to a fire or gained by recreating a wetland, for example, or the hypothetical impact of changing land cover in a single cell; potential conversion and restoration scenarios are shown in red and dark green, respectively.

But how do we connect these various AOH quantities with extinction risk? If we think of a species’ current AOH relative to that in the absence of people–what we refer to as the proportion of AOH remaining–as a proxy for its current  population size relative to its average size over its evolutionary history, we can make some weak assumptions:

• If the proportion of AOH remaining is 1, then the population is (roughly) the same size as if humans were absent, so the change in its probability of extinction relative to that in the absence of people is 0%.

• If the proportion of AOH remaining is 0, then the population has collapsed, extinction is guaranteed, and the change in its relative probability of extinction is 100%.

• If the proportion of AOH remaining is greater than 1, we fix the probability of extinction at 0% since again we frame extinction risk relative to the absence of humans.

We have sorted out the intercepts, but we still need the rest of the curve. While we do not know the exact shape for each species in question, there is evidence that it should be nonlinear: we expect that any given amount of new habitat loss will have an increasing effect on a species’ relative probability of extinction as the proportion of AOH remaining decreases.

To illustrate this, imagine that so far a species has lost roughly half its human-absent AOH. Then, using our nonlinear curve, we can estimate its current probability of extinction relative to that in the absence of people.

If we then pursue a project or policy which decreases the current AOH by ΔAOH, we can convert this to a change in relative probability of extinction due to the project or policy, shown as ΔP(E) above.

If, however, a project or policy caused the same ΔAOH but the species had already lost a substantially larger proportion of its human-absent AOH (i.e., it had a much lower current AOH), we would see a much larger change in the probability of extinction attributable to the project or policy. This key feature of the AOH-extinction risk curve is a direct result of its nonlinear shape.

LIFE: Land-use change Impacts on Future Extinctions

So far, we have described the extinction risk dynamics of a single species, but this framework generalises readily to any number of species.

Suppose we now have two species (labelled 1 and 2 above) and we have calculated the pre-human and current AOHs for each (above, top left). By placing a grid over these AOH polygons (above, top right), we can estimate the impact on each species of further conversion of suitable habitat (in this example to arable land) for any cell in this grid. Both species’ AOHs overlap the red hatched cell and cannot live in arable land, so a conversion in this cell generates a ΔAOH for each species which we can convert to a change in relative probability of extinction of each species, ΔP(E), using the AOH-extinction risk curve (above, bottom left). Again, we do not know the unique shape of this curve for each species, so for now (based on the curves assumed in many previous studies) we fix all species’ curves to the same nonlinear shape: an exponential function of the form y = x^z, usually with z = 0.25.

If we sum these ΔP(E) over all species present in the cell, we get the LIFE score for that cell with respect to the land-use change scenario (in this case conversion of natural habitat to arable) and the two species in question (above, bottom right). Importantly, the LIFE score is mathematically equivalent to the expected number of extinctions attributable to that land-use change across those species. If we apply the same land-use change scenario to each cell in the grid in turn, we can then get the LIFE surface for that scenario and those species.

Global conversion and restoration maps

To create LIFE surfaces that could be used for many different kinds of conservation problems we computed human-absent and current AOHs for all terrestrial vertebrate species for which suitable range and habitat preference data exist (roughly 30,000). We mapped out the cell-by-cell effects of two scenarios: (1) the conversion of remaining natural habitats and pasture to arable land (what we call the conversion to arable scenario), and (2) the restoration of non-natural habitats (excluding urban areas) to their potential human-absent habitat (the restoration to natural scenario). 

LIFE scores for these scenarios were calculated for 1 arc-minute grid cells (approximately 1.86 km on the side or 3.46 km^2 at the equator) and divided by the area of the cell undergoing land-cover change. We further standardised values by dividing all values by the total number of species included in the analysis; this ensures the metric is agnostic to the number of species considered and remains comparable even if additional species are included. The resulting cell values reflect the per-species contribution to the change in extinction probability due to land conversion or restoration and is referred to subsequently as the change in extinction risk.

A few notable observations:

First, the impacts on extinction of converting remaining habitats and pasture to arable land are largely greater than zero (indicating increases in overall extinction risk) and those of restoring natural habitats largely less than zero (indicating decreases in extinction risk), although there are exceptions. (For example: where conversion of a forested area to cropland increases the AOHs of lots of arable-tolerant species that cannot live in forest–here conversion might reduce extinction risk and restoration would increase it.)

Next, the increases in extinction risk from conversion to arable tend to be both greater in magnitude and spatially more widely distributed than the decreases in extinction risk resulting from habitat reversion to natural. For both scenarios, LIFE scores are highly skewed, with the majority of regions having relatively low values and a few regions scoring very highly. For instance, the conversion to arable map tends to highlight species-rich regions with high levels of endemism.

Finally, direct tests show that the LIFE framework exhibits considerable scalability: published surfaces generated at 1 arc-minute resolution (3.46 km^2 at the equator) yield reasonably reliable estimates of the impacts of land-use change across areas ranging from approximately 0.5 km^2 to 1,000 km^2. Importantly this means that LIFE can be used to estimate the number of extinctions caused (or avoided) by actions in this wide area range simply by adding up or dividing the values for affected cells in the existing layers. In this way, LIFE allows the impacts of various actions–from large-scale conservation efforts down to individual dietary choices–to be measured in terms of their impact on extinctions.