Technical Comments

Comment on “Roadless Space of the Conterminous United States”

Science  23 Nov 2007:
Vol. 318, Issue 5854, pp. 1240b
DOI: 10.1126/science.1145349

Abstract

Watts et al. (Reports, 6 May 2007, p. 736) introduced a metric of landscape pattern called roadless volume (RV). However, as with most previous metrics, RV does not explicitly address ecological processes. We demonstrate that RV can produce results inconsistent with the notion of landscape connectivity and contend that more ecologically relevant metrics are available.

A major research trend in landscape ecology is to identify relationships between metrics of landscape structure and ecological processes (1, 2). Li and Wu (3) state that the proper use of landscape metrics should include addressing relations between observed landscape patterns and ecological processes. Dozens of landscape metrics have been proposed that are commonly used to quantify pattern and permit comparison across landscapes (4). However, they often fail to correlate with ecological processes (5). We are concerned that Watts et al. (6) have proposed another landscape metric—roadless volume (RV)—which may be useful for describing pattern but is problematic for use in analyzing landscape ecological processes. We recognize that the questions posed by the authors are “(i) how much space is there between the roads, and (ii) how much is lost as new roads are added to the network, penetrating roadless space?” (6). However, the authors explicitly set their questions in the context of using the metric to assess the effects of roads on ecological conditions, including the fragmentation of habitats. Although RV can be a quantification of remoteness or inaccessibility of areas in a landscape, it does not explicitly relate to any ecological process. In particular, we demonstrate that this metric can produce misleading results if applied to assess habitat fragmentation and landscape connectivity as related to ecological processes associated with species movement. We contend that better metrics are available.

Several landscape metrics have been proposed that explicitly incorporate ecological processes. These include the ecologically scaled landscape index average patch connectivity (1), which is the probability that a patch is colonized based on species-specific movement distances and the spatial configuration of habitat patches. This metric addresses a key ecological process impacted by roads, namely, the movement of individuals between habitat patches in the landscape. Another metric that explicitly incorporates this process is effective mesh size (meff), an expression of the probability that any two locations in the landscape are connected (i.e., not separated by barriers such as roads) (7, 8). This metric can also be interpreted as the average size of the area that an animal placed randomly in the landscape would be able to access without crossing barriers.

Our main concern is that RV may produce misleading results if it is used to quantify habitat fragmentation and landscape connectivity as related to ecological processes associated with species movements. To illustrate this issue, we present two theoretical landscapes (Fig. 1) that have exactly the same length of roads but different fragmentation patterns. The landscape with one small and one large patch (left) is less fragmented and more connected than the landscape with four small patches (right) because there is a higher probability that two animals randomly located in that landscape can encounter each other without having to cross a barrier. However, the RV gives results counter to this, with the less connected four-patch landscape having an RV 12% higher than the more connected two-patch landscape (Fig. 1). This assessment is inconsistent with the notion of landscape connectivity. In contrast, the effective mesh size is 60% lower for the less connected landscape, which agrees with the intuitive understanding of landscape connectivity (8). This example shows that although RV may quantify remoteness from roads, its use in analyzing landscape ecological processes should be treated with caution and explicitly examined for its utility in quantifying the process of interest.

Fig. 1.

Two theoretical landscapes fragmented by roads. The left landscape has higher connectivity (lower fragmentation) than the right landscape. However, the RV calculated for the less connected landscape (right) is higher than the RV for the other landscape, which is counterintuitive. In contrast, the effective mesh size (meff) gives results consistent with the notion of connectivity. RV is calculated as the volume beneath the pseudotopographic surface defined by distance to the nearest road following Watts et al. (6). Size of each landscape is 10 km × 10 km; width of the roads is 10 m.

The ecological utility of the RV metric could be improved by more accurately modeling the “road-effect zone.” RV assumes that the ecological “value” of every point location increases linearly ad infinitum away from roads. However, recent reviews suggest that most road effects occur within 1000 m of roads and that the slope of the road effect levels off with increasing distance from a road (9). Thus, rather than using a linear model, the road-effect zone should be modeled with a diminishing curve (e.g., negative exponential), which can be thought of graphically as shaving off the tops of the pyramids in Fig. 1 at some distance.

Including this distance decay in the calculation of ecologically relevant landscape metrics is straightforward. For example, this can be accomplished for the effective mesh size by subtracting the area of road-effect zone from the size of the patches. In addition, a function specific to a particular species or ecological process can be applied to this calculation to model the diminishing effect of roads at further distances, as well as to incorporate the influence of traffic volume on the width of the road-effect zone, as has been done by Jaeger et al. (10). The positive effect of wildlife crossing structures on landscape connectivity can also be included in metrics of connectivity, whereas RV cannot address this issue.

The goal of using landscape metrics to assess landscape fragmentation is to gain insight into landscape-level ecological processes associated with species movements, such as foraging, dispersal, genetic connectivity, and metapopulation dynamics, which depend on the ability to move through the landscape and between habitat patches (8). We argue that the RV landscape metric proposed by Watts et al. (6), while potentially useful for measuring landscape patterns, does not address ecological processes relevant to landscape fragmentation. Moreover, we have demonstrated that RV gives an assessment of two theoretical landscapes that results in a higher RV for the more fragmented, less connected, and therefore less easily traversed landscape. Landscape metrics need to move beyond assessing simple first-order characteristics about habitat patches and move toward second-order statistics that assess ecological relationships occurring both within and among habitat patches.

References and Notes

View Abstract

Navigate This Article