LIFE: A metric for mapping the impact of land-cover change on global extinctions

Abstract

Human-driven habitat loss is recognized as the greatest cause of the biodiversity crisis, yet to date we lack robust, spatially explicit metrics quantifying the impacts of anthropogenic changes in habitat extent on species' extinctions. Existing metrics either fail to consider species identity or focus solely on recent habitat losses. The persistence score approach developed by Durán et al. (Durán et al. 2020 Methods Ecol. Evol. 11, 910-921 (doi:10.1111/2041-210X.13427) represented an important development by combining species' ecologies and land-cover data while considering the cumulative and non-linear impact of past habitat loss on species' probability of extinction. However, it is computationally demanding, limiting its global use and application. Here we couple the persistence score approach with high-performance computing to generate global maps of what we term the LIFE (Land-cover change Impacts on Future Extinctions) metric for 30 875 species of terrestrial vertebrates at 1 arc-min resolution (3.4 km² at the equator). These maps provide quantitative estimates, for the first time, of the marginal changes in the expected number of extinctions (both increases and decreases) caused by converting remaining natural vegetation to agriculture, and restoring farmland to natural habitat. We demonstrate statistically that this approach integrates information on species richness, endemism and past habitat loss. Our resulting maps can be used at scales from 0.5-1000 km² and offer unprecedented opportunities to estimate the impact on extinctions of diverse actions that affect change in land cover, from individual dietary choices through to global protected area development.This article is part of the discussion meeting issue 'Bending the curve towards nature recovery: building on Georgina Mace's legacy for a biodiverse future'.

Keywords: biodiversity metrics; extinction; habitat loss; land cover; persistence; restoration.

Figure 1.

Illustrative example of calculating the change in probability of persistence across 100 years and the LIFE score associated with a change in land cover of a single cell (i), for the simple cases of three example species (A, B and C). Species A currently still has one-third of its human-absent AOH and loses 25% upon conversion of cell $i$. Species B has already lost a larger portion of its AOH, so the conversion of $i$ has a greater impact on its probability of persistence than for A. The final species C demonstrates that not all species have the same historic range and the influence this has on changes in persistence. The LIFE score for the conversion of the cell is the sum of the changes in the probabilities of extinction (which is equal and opposite to the changes in their probabilities of persistence) for all species present in the cell. LIFE can be calculated in the same way for any land-use change including changes that result in increases in habitat such as restoration (see electronic supplementary material, figure S1).

Figure 2.

Histogram of the proportion of original AOH remaining for terrestrial vertebrate species (n = 30 875). For migratory species, we plot data only for the season with the lower value. Species with increased AOH as a result of human activity have values >1 and their distribution is depicted in the inset. The black line in the inset marks a value of 1. The geometric mean across all species and for those that have lost AOH as a result of human activity are depicted by solid and dashed grey lines, respectively.

Figure 3.

Global maps of LIFE scores associated with (a) conversion of remaining natural habitats to arable land and (b) restoration of cropland and pasture to natural habitats. The maps show changes in extinction risk realized over a 100 year time period summed across all study taxa (amphibians, reptiles, birds and mammals), aggregated to 1 arc-min grid cells and expressed as the average impact per km² of conversion or reversion. Changes in probability of extinction are derived assuming a power-law persistence–habitat loss curve with z = 0.25. Break points divide the scores into octiles, except for the uppermost octile, where we also show the top 2.5% of most impactful land-use changes. Positive values indicate an increase in extinction risk, while negative values show a decrease in extinction risk. Negative values in the conversion map arise where species can inhabit arable, farmed land but not the natural habitat that it replaces. Conversely, positive values occur in the restoration map where species can inhabit farmed land but not the natural habitat that replaces it. Where land-use changes have the opposite from the predominant effect, they are shown in grey.

Figure 4.

The modelled mean deviation (solid line) from the mapped LIFE scores for (a) actions that affect multiple pixels, with the 10% threshold marked by a dashed line, and (b) actions covering only a fraction of a pixel, with dashed lines marking the standard error ranges.