Using Widely Spaced Observations of Land Use, Forest Attributes, and Intrusions to Map Resource Potential and Human Impact Probability

-Scant information exists about the spatial extent of human impact on forest resource supplies, i .e. , depreciative and nonforest uses. I used observations of ground-sampled land use and intrusions on forest land to map the probabil i ty of resource use and human impact for broad areas. Data came from a seven-state survey region (Alabama, Arkansas, Louisiana, Mississippi, east Oklahoma, Tennessee, and east Texas) containing 32,000 land-use plots, with detailed at tr ibute information for about half of these plots classed as forest land. Forest land at tr ibutes included human-associated intrusions (beverage containers, garbage, l ivestock grazing, timber management activities), proximity to nonforest land, forest fragment size, ownership, and forest type. Tools included geographic information software, a 100 MHz Pentium I processor, and 0.4-ha land-use and forest resource sample plots nominally spaced at 4.8~km intervals. I transferred information from sample plot locations to grid cells sized large enough to minimize computer memory storage and computat ion requirements, and small enough to conservat ively model information from adjacent cells with plot information and include no more than one sample plot per cell . Results used spatial ly moving averages, with examples, to assess the spatial context of forest resources. Maps displayed regions of high and low probabil i ty of altered forest resources, forest attr ibutes, and patterns quali tatively correlated with nonforest land-use neighborhoods. Findings suggested land areas with potential for mult iple resource uses and forest land vulnerable to nonforest conversion. Regional resource inventories (e.g. , the USDA Forest Service’s Forest Inventory and Analysis [FIA] and the Natural Resources Conservation Service’s National Resources Inventory [NRI] program) document the status and change in land use and resource production from widely spaced observations at ground-sampled locat ions. Limited spatial information exists about the extent and potential for human impact, i.e., depreciative and multiple uses, on resource supplies. These inventory f indings tradi t ional ly provide tabular s ta t is t ics and note a t t r ibute locat ions spat ia l ly by pol i t ical subdivis ion, e .g . , by s ta te and county. I conducted a study to (1) document an approach to more uniformly i l lustrate spat ial relat ionships of inventoried at tr ibutes from widely spaced ground observations, and (2) make an init ial , coarse-scale evaluation of the extent and probabil i ty of regional resource potential from observat ions of human intrusions. Examples that follow use land-use surveys, forest resource inventories, and ground observations, but the methods of this s tudy could be appl ied to most other systematically sampled, widely spaced earth surface surveys. Ground-based regional forest resource inventory sample plots in the United States, which range in size from 0.1 to 8 ha, are generally too small or otherwise an inadequate Research Forester , Forest Inventory and Analysis Unit , Southern Research Stat ion, U.S. Department of Agriculture, Forest Service, Starkville, MS 39760-0928, USA. sample for quantifying resources other than forest vegetation, such as wildlife populations and recreation opportunities. Resources requiring larger samples must be aggregated to larger land divisions. Sample information aggregated by county, for example, yields estimates of black bear habitat and primitive-oriented recreation opportuni t ies (Rudis and Tansey 1995). Although aggregat ion by county or other administrat ive divis ion in inventory planning is often straightforward, natural resources and the processes that affect them may not be al igned with those divis ions. An intui t ively appeal ing approach is to aggregate information with a uniformsized, large-area division, or “window,” that encompasses several sample plots. Incorporating the context (location, adjacency, and neighborhood) with spatially referenced sample plot locat ions entai ls calculat ing spat ia l ly moving averages, i .e. , moving the window across the earth surface so that the average at any location always depends on the nearest plot values. “Neighborhoods” are broad regions with s imilar a t t r ibutes or values . Of particular interest are forest resources adjacent to nonforest uses or in nonforest neighborhoods named by their predominant land use. Forest land in urban and buil t-up land neighborhoods or adjacent to roads may yield lower forest land est imates than est imates that ignore these contexts (Befort et al. 1988). Forest land near urban areas has reduced timber harvests (Barlow et al. 1998) and primitive-oriented recreation opportunities (Rudis 1987). In regions of high human population