The set of metrics used in the wilderness mapping literature can be divided into two major dimensions of defining wilderness: the subjective or perceived wilderness experience and ecological intactness. Most wilderness metrics attempt to describe both aspects. For example, the presence of roads and human settlements indicate both easiness of access, visual impact, and the ecological impact of these infrastructures. Yet some indicators address the two dimensions separately as it is the case with apparent naturalness and biophysical naturalness (Lesslie et al. 1988). For the purposes of this chapter, we chose a series of four metrics: two that describe both the subjective human experience of wilderness and the ecological impact, and two that have mainly an ecological dimension. The metrics used here quantify human impact thus wilderness increases with the decrease of the metrics.
Remoteness from roads and human settlements is an important dimension in the feeling of solitude intrinsic to the wilderness experience. However, roads and other human access infrastructure have also a strong impact on wild populations and ecosystems. The most obvious impact is road mortality, shown to affect mammals (Philcox et al. 1999; Seiler 2005; Grilo et al. 2009), birds (Orlowski 2005), reptiles (Iosif et al. 2013) and amphibians (Patrick et al. 2012). But impacts of roads, traffic and human access can be much more profound, affecting population and community structure (Habib et al. 2007), trophic interactions (Kristan III and Boarman 2003; Whittington et al. 2011), ecosystem functioning and structure (Christensen et al. 1996; Hansen et al. 2005; Rentch et al. 2005), and environmental conditions through high pollution levels (Hatt et al. 2004). Roads can favour the expansion of invasive species (Jodoin et al. 2008; Vicente et al. 2010), and of exotic and human-favoured predators (Alterio et al. 1998). They also expose forest habitats to edge effects (Tabarelli et al. 2004). These ecological impacts of roads and human settlements alter a range of ecological conditions compared with the context that would exist without these human infrastructures. Here we evaluate human access from roads and settlements by calculating the cost distance to paved roads and settlements according to the Naismith's rule which assumes differentiated relative traveling times depending on terrain, land cover, and river networks (Carver and Fritz 1999). We extracted the data on paved roads from the Eurogeographics Road database and the Open Street Map database, land use data from Corine Land Cover 2000 and 2006, and terrain ruggedness data from the Shuttle Radar Topography Mission (SRTM) at 1 km resolution. The range of the human access score values is expressed from 0 to 1. In Europe, the mountainous areas, the Iberian Peninsula, the Balkans, Scotland, and Scandinavia are the least accessible regions and the least impacted by roads and settlements (Fig. 2.1a).
Artificial night light has a similar dimension in the definition of wilderness. Light pollution has been decried for its impact on the visibility of the natural night sky (Cinzano et al. 2000), diminishing the night wilderness experience. But artificial light has also strong ecological impacts (Longcore and Rich 2004; Navara and Nelson 2007; Hölker et al. 2010b; Gaston et al. 2013), affecting invertebrates (Davies et al. 2012, see Chap. 6), fish (Becker et al. 2013), mammals (Boldogh et al. 2007) and bird populations (Montevecchi et al. 2006). Direct mortality (Hölker et al. 2010b), impacts on trophic relations and community structure (Perkin et al. 2011), disruption of migratory routes (Gauthreaux Jr et al. 2006) by night light lead to profound modifications of ecosystems functions (Hölker et al. 2010a). Nocturnal species such as bats and moths (see also Chap. 6) receive the brunt of the impact. We assess the impact of artificial light on ecosystems and wilderness experience by using the satellite data of the upwards emitted and reflected artificial light with a spectral range of 0.5–0.9 μm in Europe from the Visible Infrared Imaging Radiometer Suite (VIIRS) of the Soumi National Polar-orbiting Partnership (SNPP) for the year 2012 (NOAA National Geophysical Data Center 2012) with a resolution of 15 arc sec (approximately 450 m). We apply a kernel function to distribute the impact over a radius of approximately 10 km (Fig. 2.1b) as a conservative approximation meant to cover the night glow effects reported in the literature (Kyba et al. 2011) along with the direct ecological impacts (Longcore and Rich 2004). In each pixel, the light impact score is the sum of all the impact scores from the surrounding light sources and it represents a relative measure aimed at encompassing both the ecological aspect and the impact on the subjective wilderness experience (Fig. 2.1b).
The last two metrics that we consider here are qualitative and quantitative measures of the human modification of ecosystems and thus they convey mainly, although not exclusively, an ecological significance. Anthropogenic change of natural habitat is one of the major drivers of biodiversity loss (Pereira et al. 2010) and it has been studied extensively for a large range of taxa (Bolliger et al. 2007). The most conspicuous element of habitat loss is the change in vegetation, and intact vegetation cover has been used before as a wilderness indicator (Bryant et al. 1997). Human changes in vegetation tips the balance in favour of species benefiting from human presence and impacts habitat-sensitive ones (Leu et al. 2008). Therefore we use here the deviation from potential natural vegetation (dPNV) as a qualitative measure of the human impact on the landscape. We used the potential natural vegetation (PNV) classes of the map developed by Bohn et al. (2000). We calculate the similarity of current land cover to PNV by estimating the probability that the CORINE 2000 land cover class in any one location in Europe belongs to the local PNV type (Bohn et al. 2000). The probability of agreement was classified in four classes with different scores: assumed = 1, most probable = 0.75, probable = 0.5 and possible = 0.1. The resulting map was combined with the grazing density data from Food and Agriculture Organization, which was previously linear transformed to a scale from 0 to 1, where 1 represents a density of 20 heads/km2 or more. We used
Fig. 2.1 Wilderness areas according to four metrics. a Access from roads and human settlements. b Artificial night light. c Deviation from potential natural vegetation. d Proportion of harvested primary productivity out of the potential primary productivity. Wilderness value increases with the decrease of the metrics
grazing density to account for human transformations in semi-natural grasslands. We expressed the dPNV value by subtracting from 1 the score calculated according to the described methodology. (Fig. 2.1c).
Through agriculture, hunting, fishing and forestry, humans are removing significant quantities of biomass from the ecosystems. Primary productivity (PP) is the foundation of trophic networks and it influences the structure and functions of ecosystems in a domino effect across trophic levels (Haberl et al. 2004). Humans have reduced drastically the PP available to other species and this has changed the composition of the ecological communities (Barnosky 2008; Pereira et al. 2012). We map the proportion of human harvested PP out of the total potential PP in Europe as another indicator of wilderness and using the data analysed in Haberl et al. (2007). We calculated the harvested PP by extracting net PP remaining in ecosystems after harvest from the net PP of the actual vegetation. We then calculated the proportion of harvested PP by dividing net harvested PP by net PP of the potential vegetation. The data are calculated based on country-level statistics of the Food and Agriculture Organization (Haberl et al. 2007) while potential PP is estimated using the Lund-Potsdam-Jena dynamic global vegetation model (Sitch et al. 2003). Some abnormalities can be noticed in the harvested PP map which are due to the assumptions of the model and the FAO national level data. The map has to be interpreted with this limitation in mind (Fig. 2.1d).
The four resulting maps based on the selected metrics show a common pattern of high human footprint in the lowlands of central Europe (Fig. 2.1). The most unaltered values of all metrics occur in high mountainous areas and Scandinavia. But the differences at intermediate values of wilderness provide a key signal to what are the strongest determinants of human footprint at regional level in Europe. For example, although the dPNV is very low in almost all of Scandinavia (Fig. 2.1c), the proportion of harvested PP is comparatively higher, consistent with high forestry harvest in the Nordic countries (Fig. 2.1d). The reverse pattern is noticeable in the Iberian Peninsula where although the drier climate restricts high harvesting of PP, the current vegetation is quite far from PNV as measured in our map and consistent with the degradation of the Mediterranean habitats (Myers et al. 2000). In the same region, the significant differences between the inland and coastal values of the night light impact and human access (Fig. 2.1a and b) indicates the high difference between the human population densities inland compared with the coastal regions. These differences in the distribution of human populations are masked in the PNV score and harvested PP maps (Fig. 2.1c and d). The map of artificial light (Fig. 2.1b) also points out to a discrepancy in the relative wilderness values in East and SouthEast Europe compared with the dPNV score map for example (Fig. 2.1c). The lower economic activity in this area results in lower light impact although the level of vegetation change is very high (Doll et al. 2006).
The lowest wilderness areas in Europe have usually low scores for all the wilderness dimensions considered, and they represent mainly areas of high human densities and intense economic activity. Conversely, high wilderness areas are the wildest from all the points of view taken here. But the areas of intermediate wilderness values are strongly impacted by only one or two metrics with very low wilderness values. Especially dPNV and harvested PP have a farther reach, affecting even ecosystems where infrastructure and artificial light impacts are reduced. These indicators are connected with more extensive land-uses such as agriculture and forestry, and less with high human population densities and infrastructure.
The synergies and interactions between the different elements of our wilderness mapping emphasize even further their ecological significance. In areas of high habitat quality the road mortality can be higher in absolute terms because it affects more abundant populations (Patrick et al. 2012) while road lighting can increase the impact of the road itself on the local ecological communities by favouring certain types of predation (Rich and Longcore 2005) or providing additional perches for improved hunting efficiency of raptors such as kestrels (Sheffield et al. 2001).