Technical Comments

Response to Comment on "Roadless Space of the Conterminous United States"

See allHide authors and affiliations

Science  23 Nov 2007:
Vol. 318, Issue 5854, pp. 1240
DOI: 10.1126/science.1145424

Abstract

Girvetz et al. criticize our work and the value of the roadless volume (RV) metric, basing their arguments on the artificial imposition of wildlife habitat connectivity as the context. We counter that in our intended contexts, both practical and geographic theoretical considerations make RV an appropriate measure of the space between roads.

We do not agree with the comments of Girvetz et al. (1), who claim that the roadless volume (RV) metric that we introduced (2) is misleading. Their criticisms proceed as if habitat connectivity were the only relevant issue. The terms “fragmentation” and “connectivity” do not appear in our paper; Girvetz et al. apparently hang their arguments on a single sentence that appears amid a long listing of ecological effects of roads: “Road encroachment affects ecological resources...usually by fragmenting habitats and introducing chemical contaminants and exotic species....”

Girvetz et al. contend that any land status metric must apply to a specific ecological process. We assert that RV in fact does this, but in a different—perhaps radically different—way that the authors overlook. Humans are demonstrably the most invasive of terrestrial megafauna, and roads are a durable signal of human invasion. Measurements of the extent and rate of human invasion are ecological and geographic problems quite independent of wildlife habitat analysis. Human invasion assuredly affects wildlife habitat, but we do not claim that RV is the appropriate metric to express this influence. Our study revolved around human process, not “natural” (nonhuman) ecological process, and mentioned only briefly that there are connections between the two.

Girvetz et al. set up an artificial contest between RV and a metric introduced by Jaeger (3), effective mesh size (meff), which is a connectivity metric. Reductions of RV are caused by penetration of roads into otherwise roadless space, so RV itself is a measure of unpenetrated space. Examination of a penetration metric from a connectivity viewpoint is spurious. We illustrate the reverse, also spurious, argument against meff. Begin with the large bounding square in the right panel of figure 1 in (1) (a windowpane pattern of four small squares making up one larger square), for which meff equals the area of the large square. Now add the four interior lines (roads), but take away the smallest piece at their central junction. The entire landscape is connected and meff still equals the area of the bounding square. In this same experiment, however, RV is reduced by half. One might conclude from meff that nothing changed, and indeed connectivity did not, yet something else—penetration in this case—was altered by the addition of the four interior roads. Comprehensive description of fragmentation demands multiple metrics, such as length of roads (road density), space remaining between roads (RV), and the closure of roads into loops that separate patches (meff). One cannot use arguments couched in the language of one of these elements to invalidate another element; they are related but distinct.

In the language of stimulus and response analysis, an important type of ecological research, the expansion of humans into all corners of the terrestrial environment is a stimulus with a host of ecological responses. Analysis of stimulus-response connections requires quantitative expression of the stimulus. Roads are the durable evidence of ready human access; distance to road (DTR) is a primitive and incomplete measure of human proximity and, therefore, potential for human influence. Riitters and Wickham (4) pioneered national-scale road distance analysis, demonstrated large-scale geographic patterns of road penetration in forests, and pointed the way for our work. To our knowledge, no one has yet published a national-scale analysis of fragmentation based on meff or other polygon-derived metrics. It appears, therefore, that practical considerations point toward advantages of distance-based analysis when studying patterns of human stimulus at national scale.

The Riitters and Wickham data set mapped grid points that lay within a discrete set of nine distances to the nearest road. We introduced the notion that integrated DTR gives a measure of unpenetrated space and that changes of penetration could be measured by changes of the RV integral. Our county maps of mean DTR for obvious reasons show large-scale patterns similar to those in Riitters and Wickham's watershed and ecoregion maps. Applications of RV analysis at other scales of human geography, such as in urban and regional planning, require no modification of concept or method. We expect that RV will be used to measure, for example, urban incursion into rural space.

We conclude by addressing the concerns of Girvetz et al. related to distance decay of road-induced effects and the limits of the “road-effect zone.” It has been demonstrated that the road-effect zone is highly variable when mapped in detail (5) and is essentially unquantifiable without such mapping. Girvetz et al. suggest that the ever-increasing values of DTR imply increasing road effect, but it is quite the opposite. DTR is a coarse, uncalibrated indicator of road-effect invulnerability, the opposite or reciprocal of vulnerability. One could, for instance, use the reciprocal of mean DTR for an area (this is the same as area divided by RV) to express road-effect vulnerability. The first job is to measure spatial attributes, in this case space unpenetrated by roads; the second job is to relate spatial metrics to ecological conditions and processes. A clear advantage of leaving ecological processes out of definitions of spatial metrics is the avoidance of circular reasoning when space measurements are used in ecological models. Use of the reciprocal, or some other transformation, of mean DTR is more appropriate than redefining the spatial metric with built-in ecological assumptions.

References and Notes

View Abstract

Navigate This Article