We undertook a scoping of candidate LWE areas by identifying the five most intact areas for each ecoregion of the world. We overlaid the most recent human footprint map (Venter et al., 2016a,b) on the most recent map of ecoregions of the world (Dinerstein et al., 2017). The human influence index (HII) can range between 0 and 50; for the purposes of this analyses we identified the best 10% of each ecoregion (lowest footprint scores), and within this subset selected the 5 largest intact polygons. HII has been widely used relative to more recent datasets of human modification (e.g., Kennedy et al., 2019) and therefore offered opportunities to compare our results with previous literature. A similar approach was used for the “Last of the Wild” (Sanderson et al., 2002), except this assessed the 10 largest areas within the best 10% of each Biome. In order to avoid polygons that had a lot of “gaps” resulting from pockets of high human activity in otherwise non-impacted landscapes, we applied a 5 km smoothing buffer to the HII map (each pixel representing the average HII score within the buffer radius). We selected this buffer radius from a range between 0 and 25 km, testing the result in ~2.5 km increments. The 5 km buffer provided the best smoothing, and appeared to stabilize both the reduction in border effects and remnant habitat selection (median effects on area perimeter and shape did not substantially vary using larger buffers). This meant that fragmented and border habitats would experience an increase in HII, while core areas would remain unaffected.

We selected a subset of these areas, designated Forest LWE, which selected only those forested ecoregions that overlapped with IFL polygons. We used these polygons to intersect with global maps of species loss (see below).

Our rationale for using anthropogenic influence to identify LWE areas is that a reduced anthropogenic influence is likely to translate to a reduced ecological impact, including species loss. Global maps of anthropogenic influence, however, are not likely detailed enough in many places of the world to capture (directly or indirectly) all pressure types that might result in species loss, or if modern measures of anthropogenic influence can still capture historical impacts. Notable pressures that are unmapped to date include hunting/poaching and will have drastically altered species composition in large landscapes; regional or local-scale maps will capture more elements of the human footprint. The regions we identify will need to go through more stages of rigorous analysis, including on-the-ground evaluation before KBA criterion C sites can be positively identified. As such, this is a preliminary scoping of regions of the world where such sites are likely to exist.

We assessed loss of faunal intactness by mapping the distribution of extinct ranges for species. We compiled maps of all species assessed from the IUCN Red List of Threatened Species (IUCN, 2018) and mapped the native range where a species was considered to be extinct or possibly extinct as classified under the range fields “Presence” for each species range polygon. The Red List assesses where species have gone extinct since the year 1500 AD, so does not include species that were extirpated prior to this date and there are recognized gaps in coverage as a result. It maps species that have gone extinct, and areas of range where extant species have lost range. We mapped all such ranges to assess the numbers of extinct species to obtain a measure of loss of faunal intactness across the world. We then intersected this map with the IFL and LWE/Forest LWE polygons to calculate the percentage area of polygons that had not lost any species using this measure.

The map of LWE areas is visually dominated by large areas of the northern ecoregions in tundra, taiga, and boreal forests (Figure 1). This is because the HII in these less populated ecoregions are mostly scored zero, so that when selecting the 10% of lowest scores, all cells with zero are selected. Human impacts in these regions include logging and other resource development (e.g., mining areas, seismic lines), but these are not available on global maps and therefore do not get incorporated in the HII (though see Kennedy et al., 2019). This highlights a compromise tied with using current global data sets, accepting data that may be of lower quality than locally available in order to reduce geographical bias across management units (countries or sub-national regions). For this reason, there have been several efforts to make regional human footprint maps (Leu and Hanser, 2008; Woolmer et al., 2008) What is clear is that for most of the ecoregions of the planet, the five most intact areas of each ecoregion are relatively small in size (average area was 6,323 km² but median area was only 696 km²), reflecting the fine scale of many ecoregions and the density of human activity. Only 340 areas out of a total of 3,852 identified were larger than 10,000 km², the recommended minimum size for a KBA criterion C site (IUCN, 2016).

The results of the intersection of the IFL and LWE areas with the ranges of the 16 species for which we had data on density or suitable habitat are shown in Table 1 (Figure S1 in supplementary material shows the overlap of LWE/IFL with suitable/high density habitat for all forest species). This table shows the areas of high density/suitable habitat in the LWE areas, the IFLs and the percentage of the species global range. On average 21% of high density/suitable habitat of a species range was captured within the LWE areas and 34% within IFLs. Given that all LWE areas (including those in non-forest ecoregions) cover only 2% of the surface covered by IFL (24.4 million km² vs. 1,163.3 million km²), we also compared the relative percentages of high density/suitable habitat captured by each approach by plotting the percentage of high density/suitable habitat within the species range against the percentage of the area of the LWE/IFL polygons where there was overlap (Figure 2). For example, only 17% of the LWE areas that overlap African forest elephant range have high elephant densities, and this drops to only 4.5% in IFL polygons, similar to the global percentage across their range (3.8%). In some cases, the LWE areas capture more of a species' high-density range while for others the IFL do a better job. For the most part both capture a larger percentage than what is available within the species' range (species above the line of 1:1 –Figure 2), although many of the comparisons are close to what would be expected if polygons were allocated randomly (1:1 line—Figure 2). For many species, however, large areas of the IFL or LWE polygons do not have high density/suitable habitat. In these areas, the species are unlikely to be at a functional density or not even present. This was particularly true for those species measured using density, which were likely to better reflect functional integrity of a site. Note that the percentage areas of suitable habitat will be on the high side for the ranges determined by range wide priority setting because the assumption was made that definite/confirmed range or high quality range would have functional densities of the species, yet at many areas this may not be the case and they are simply recorded as being present which qualifies it as definite range. This assessment clearly shows that these species, which tend to be affected by human impact, will be at low and likely non-functional densities across large areas of either IFL or the LWE areas.

Figure 3 shows the results of mapping all species assessed on the IUCN Red List where a species has been extirpated or is possibly extirpated since 1500 AD. A total area of 54.7% of the terrestrial realm of the earth (excluding Antarctica) has at least one species with range recorded as extinct or possibly extinct. The white areas in Europe and central Asia would certainly have had species that had been lost prior to 1500 AD, such as bears (Ursus arctos), wolves (Canis lupus), and beavers (Castor fiber). It is important to note that within the IUCN Red List assessments, most records do not estimate where the species has lost part of its range. Therefore, this map will very much be an underestimate of species loss across the world. Yet it is still valuable in highlighting how much of the world has lost one or more species.

Intersecting all LWE areas with this map shows that 33% of the area of LWE polygons have no recorded extinctions, compared with 31% of IFL polygons. However, of the forested LWE areas only 19% of their total area have no extinctions recorded (Table 2).

Our results show that there are few places left on the planet that are faunally intact, a result that corresponds with many assessments of global biodiversity (e.g., Secretariat of the Convention on Biological Diversity United Nations Environment Programme, 2014; Wolf and Ripple, 2017). The two measures we assessed of intactness—the IFL measure and the LWE areas—encompass reasonably large areas of the globe. However, when we look at forest species that might be using the IFLs and LWE areas, for many there are large parts of these “intact areas” where they are absent or at low densities (Figure 2) that are not likely ecologically functional (sensu Sanderson, 2006). In extreme cases, species may be present, but ecologically extinct (Estes et al., 1989; Novaro et al., 2000). Assessment of species range loss or complete extinction also shows that a large area of the earth does not have a full complement of species and cannot be thought of as faunally intact (Figure 3). More than half of LWE areas had no species recorded as having gone extinct, but this would decrease if pre 1500 data were available for ecoregions in Europe in particular, and if extinction records in the IUCN Red List were more comprehensive in general. Comparison of forested areas demonstrated that more than half of IFL and forest LWE areas had lost at least one species.

Table 1 shows that areas of suitable habitat/high density range for the species we assessed were not captured by the IFL or LWE assessments, indicating that there are important areas for species where lack of intactness, as measured by HII or from anthropogenic alteration of forest cover, is still important for these species.

Estimates of human influence could be improved by using available local and regional data at the jurisdictional (e.g., national or subnational) level (e.g., Woolmer et al., 2008). Many smaller-scale analyses will reveal considerably higher levels of human impact, however, which could reduce the area of IFL polygons or LWE areas. Our use of a 10% threshold to select the least impacted areas of each ecoregion does limit the area where candidate LWE sites could occur in some ecoregions where human impacts are negligible throughout the ecoregion. In ecoregions with extensive scores of zero (northern tundra/taiga) LWE areas were large and included more than 10% of the area of the ecoregion, however for most ecoregions we were able to limit the analysis to the best 10% or smaller area of the ecoregion (if there were more than 5 candidate sites). There is a need to assess the potential impact of this on the identification of LWE areas in future.

In addition to regional-scale human pressures mapping, accurate assessments of faunal intactness will require (1) better models of direct pressures on biodiversity such as hunting intensity, which cannot be predicted using HII-like approaches, and (2) knowledge of the extent to which abundance levels have changed relative to historical baseline for multiple species throughout their ranges. How far a hunter will move into a forest will be determined by several factors such as the importance of hunting to their livelihood (e.g., poachers in wealthier vs. poorer countries), the relative reward obtained from hunting a species (e.g., ivory vs. bushmeat), the accessibility of a site (e.g., rugged terrain vs. flat), and the intensity and likelihood of penalties that might be incurred if caught (e.g., small fine vs. jail terms). Considerably more detail than traditional range maps will be necessary to evaluate the extent to which faunal communities have retained their integrity in the face of human disturbance. Only when we have a good handle on factors such as these will we be able to start modeling faunal intactness better. A recent study that was published as this paper went to press makes the first attempt to model hunting across tropical forests (López et al., 2019) which estimates large mammals have been lost in more than 50% of IFL. Global analyses, while being useful to help with planning for conservation, must also be supplemented with site evaluations for identification of evidence-based, intact areas for species, as specified by Global Standard for the identification of Key Biodiversity Areas with respect to criterion C sites.

Intactness as measured by global datasets and functional ecological integrity of biodiversity may not always co-occur and it is important to recognize that the two have a value for conservation independently. Some well-managed protected areas, for example, will inevitably have a high degree of anthropogenic influence, while at the same time retaining a full, or nearly full, complement of species at functional densities, exactly because they are being well protected and managed (e.g., Nairobi National Park on the outskirts of Nairobi city in Kenya (Ogutu et al., 2013). On the other hand, some intact areas may not currently contain species at functional densities but numbers might be recovered with management, so that areas become ecologically functional in the future. Areas, for instance, where a keystone species has been extirpated through hunting, and could meet KBA criterion C status after reintroduction and recovery of that species to functional levels.

The message highlighted from these analyses is that faunal intactness is highly rare in the remaining large areas on earth and that we cannot easily identify this from satellite images of seemingly intact forest canopy and human disturbance (the IFL method) nor from assessments using the HII (the LWE areas). Recent papers have highlighted the small percentage of remaining wilderness or intact sites (Watson et al., 2016, 2018; Potapov et al., 2017) and yet our results indicate that truly intact sites with a full complement of species are likely to be much rarer still.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.