Methods

Human Modification Framework

Here we further refine the HMF to emphasize three primary factors that can be used to characterize land use based on what human activities occur at a given location. Principally, humans go to an area to produce or remove natural resources (production/extraction), visit but do not extract significant resources (recreation/tourism), or concentrate or intensify resources by reconfiguring and constructing buildings and other infrastructure (urban/built-up). Each of these three factors begins with a value of 0.0 denoting no human influence, (i.e., a wild location), and 1.0 representing an area that is strongly influenced (developed) by human activities. We seek to fully characterize the spatial heterogeneity of threats by characterizing land uses at a relatively fine grain (<1 km² cells or minimum mapping unit), rather than attributing an average or overall value across a broader spatial extent, (e.g., a GAP stewardship value for an entire national park). We then can overlay our threat maps with a map of protected areas to assess gaps and opportunities.

For each factor, we explicitly place values for different levels of threat along a 0 to 1.0 axis (Table: Example land uses). Although often these values are arbitrary, estimated by experts, or mapped through surrogate variables, or both, it is critical to make our assumptions explicit, at the very least to communicate how threats may affect a location and their relative importance to other threats. Also note that this approach is agnostic to land ownership—that is, land use activities for each of these factors can occur on both private and public lands (though some types may be more dominated than others). For example, although built-up is often associated with urban areas, some localized areas within protected areas, (e.g., a national park visitor center) can be highly built-up as well. Below we provide a short, preliminary listing of surrogate maps that are commonly used to estimate these activities for the three factors—yet much work needs to be done to fill these critical data gaps. Again, we emphasize that often these values are arbitrary, but providing explicit values is a useful and important exercise.

Urban/Built-Up

This factor characterizes the intensity to which humans occupy a given location. Human activities often concentrate or intensify resources by reconfiguring resources or constructing buildings and other infrastructure such as roads, dams, bridges, etc. (Table: Possible surrogate spatial data for urban/built-up factor) provides a list of useful data layers that can be used to measure the degree of urban/built-up. Common surrogate spatial variables include population density and housing density. We prefer housing density because it more directly characterizes landscape and ecological changes on the ground, whereas population density is attributed typically only for primary residences. Data on residential land cover, (e.g., from NLCD) are often used to represent built-up areas, though additional data from the Census Bureau (or parcel level) are needed to characterize lower density development beyond the urban fringe. Commercial, industrial, and transportation land uses also are associated with high levels of human modification, (e.g., high levels of impervious surface), population density (less direct), etc. Data on transportation infrastructure, such as road density, (e.g., Theobald 2003), are also strong measures of land uses associated with built-up areas. We believe that housing density and road density provide complimentary information—high density of roads can occur not only in urban areas but also in remote rural areas, (e.g., forest logging roads). Recently, researchers are beginning to distinguish major road types, (e.g., interstate vs. secondary roads) when computing road density, though more work to characterize road use rather than the existence of roads is needed. Improved datasets are needed to better identify the location and intensity of structures in rural areas, particularly bridges, culverts, roadway fences, utility corridors, (e.g., power lines, pipelines), etc.

Production/Extraction

This factor characterizes the intensity of human activities associated with the amount and intensity of extracting or removing resources. Agricultural production, especially cropland, but, also, grazing and mining and timber activities are common forms of this activity type (Table: Possible surrogate spatial data for production/extraction factor). It is common to use land cover data to map production land uses that have caused land cover conversion, such as agricultural croplands and orchards. The intense changes in pattern and the high degree of modification of ecological processes mean that this is a critical component to map. A variety of additional activities also contribute to human modification for which data are less commonly available. This would include spatially explicit information about grazing (intensity and animal density) through permits on public lands and stocking rates on private lands. Extraction of natural resources in the form of quarries and mines that cause significant land surface modification can be readily identified. However, activities with extensive but more diffuse disturbance, (e.g., oil and gas wells or selective harvesting/thinning of forest products) are more difficult to identify and generate metrics from typical satellite imagery, though increasingly specialized databases, (e.g., well locations) are being developed.

Recreation and Tourism

This factor characterizes the intensity of human activities associated with use or visitation of an area. This is best represented by recreation and tourism activities, but can also include work-related activities. It includes the amount and type (mode) of recreational use, (i.e., pedestrian only, passive, motorized, etc.). There are a large number of important data gaps here—particularly whether areas are publicly accessible, motorized vs. nonmotorized use, visitor use levels, etc. (Table: Possible surrogate spatial data for recreation/visitation factor).

Spatially explicit data on this aspect of human modification are particularly underdeveloped. Although general visitor use levels may be available for whole parks or national forests, more detailed data to map the visitation levels (and the timing) are particularly poor. Use has been estimated based on trail density (Schumacher and others 2000), though this suffers from the same limitation as road density methods—namely the assumption that areas close to trail heads receive the same amount of use as those more distant. Methods to estimate visitor use via spatial models of accessibility are increasingly common (Geertman and van Eck 1995).

Assessing Landscape Context

To facilitate the characterization of landscape context, we characterize each factor along a 0 to 1.0 axis, based on explicit criteria (Table: Surrogate variables). We use a fuzzy average approach to combine our individual surrogate layers into a composite layer for each factor. This is calculated simply as the maximum value (at every cell) of any single input. These ratio values are then spatially averaged over different neighborhood sizes and shapes to establish a measure of landscape context. A challenging part of developing estimates along the three factors is to be explicit about the spatial characterization. That is, for example, when a value for wilderness is prescribed, does this apply to an entire wilderness area (as a political designation), or within some defined neighborhood, (e.g., within a 1-km circular window)? The spatial grain needs to be considered before one provides or computes estimates of any given factor. Here we summarize the spatial context using circular moving windows with radii of 1 and 5 km. These distance values are arbitrary, but attempt to capture the relatively local and midscale spatial context pattern. Again, our goal here is to provide an illustration of the broader method, and users will likely want to employ different values that reflect the species, process, or patterning that they are concerned about, which likely varies across ecoregions.