Introduction

Insect pests and pathogens represent the largest and most pervasive agents of natural disturbance in North American forests, with potentially significant economic, aesthetic, and ecological consequences for northern forest ecosystems (Ayres and Lombardero 2000). In order for land managers to make successful management decisions for mitigation and treatment activities, they must know the location and extent of the host resource as well as the anticipated risk of host mortality. Traditionally, land managers have relied on plot-based field sampling efforts to supply this information. Although this is useful, a comprehensive, landscape-scale, spatially continuous coverage of the resource and its vulnerability is needed to fully assess the potential impacts on the forest resource and to devise successful management strategies.

Recently, remote sensing technologies have greatly increased the amount and quality of information that is available for landscape-scale ecological risk modeling. This information includes abundance maps for individual tree species, detailed forest decline assessments (including previsual symptoms), and foliar chemical concentrations (Foody 2002, Martin and Aber 1997, Ollinger and others 2002, Plourde and others, in press, Pontius and others 2005, Smith and others 2002) . Such remote sensing-based products, combined in a GIS platform with traditional topographic-based data layers, have expanded the tools available for risk modeling. They provide the potential to greatly enhance our ability to create spatially continuous, landscape-scale models of ecosystem function and response to disturbance.

Hemlock woolly adelgid (Adelges tsugae Annand) is an invasive insect pest that is causing widespread mortality of eastern hemlock (Tsuga canadensis). Current rates of spread into uninfested areas are estimated at 10 to 15 miles per year, and all indications are that HWA will penetrate the entire range of eastern hemlock (McClure 1995a). Because HWA may infest all hemlock stands eventually, susceptibility assessments (assessments of likelihood of infestation) are not necessarily informative for long-term, landscape-scale risk modeling. However, there is evidence of differing hemlock vulnerability (ability to tolerate prolonged HWA infestation). Many infested hemlock have shown minimal resistance to A. tsugae and little chance for recovery (McClure 1995b). However, some stands remain living more than a decade after infestation (Pontius and others 2006). Indeed, two adjacent hemlock stands can often respond very differently to attack, with differences commonly attributed to topographic characteristics such as landscape position, slope, and aspect ( Bonneau and others 1997, Hunter 1993, Orwig and others 2002, Royle and Lathrop 1999). Greenhouse studies have shown that the presence of HWA alone did not cause the death of hemlock seedlings, and that it is the combined stress of drought and infestation that ultimately leads to mortality (Sivaramakrishnan and Berlyn 1999). In the field, reduced growth rates were associated with infested trees on ridge top and southwestern facing sites, but not with those on well-watered sites (Sivaramakrishnan and Berlyn 1999). Pontius and others (2006) found that several site factors could be used to predict hemlock decline across the Northeast. The severest hemlock decline was associated with markedly low growing season precipitation levels, southern and western exposures, and ridge top/side-slope positions.

All of these landscape variables are in some way related to potential soil-moisture content, indicating that water availability may be an additional stressor, accelerating decline in the drought-sensitive eastern hemlock (Bonneau and others 1997, Orwig and Foster 1999, Orwig and others 2002, Young and others 1999). Orwig and others (2002) concluded that although the duration of infestation primarily controls the intensity of hemlock decline and mortality, stands on xeric aspects succumb most rapidly.

Although significant, these landscape variables typically explain only a small portion of the overall variation in hemlock decline. Adding foliar chemistry to site factors at the plot level, Pontius and others (2006) predicted an 11-class decline rating with 98 percent one-class tolerance accuracy on an independent validation set, indicating that foliar chemistry may also play an important role in HWA dynamics and hemlock decline. Herbivory is often positively correlated with foliar nitrogen concentrations, with low nitrogen-limiting insect populations (McClure 1980, Schowalter and others 1986, White 1984). Nitrogen can be particularly limiting to insects because there is a large difference between the nitrogen concentration of plants (around 2 percent dry weight) and that of insects (approaching 7 percent) (Dale 1988). This link between nitrogen and aphid success has been documented for a variety of host species (Carrow and Betts 1973, Douglas 1993, Koritsas and Garsed 1985, McClure 1980).

For relatively immobile insects such as HWA, the nutritive quality of forage becomes even more important. McClure (1991, 1992) found that N fertilization resulted in increased relative growth rate, survivorship, and fecundity of HWA, thus increasing hemlock vulnerability and reducing the effectiveness of implanted and injected pesticides. Regionally, Pontius and others (2006) also found that foliar N concentration was the strongest correlate with HWA infestation with higher N consistently associated with higher HWA population levels.

Although useful in identifying key factors in the hemlock decline complex, such plot-level and greenhouse-based studies, do little to assist land managers in making critical planning and treatment decisions for their forests. Here, we apply a plot-level empirical model from field-based observations of hemlock decline across the northeast to a landscape-scale GIS model for hemlock woolly adelgid risk assessment. Because of the unique ecological niche occupied by hemlock stands, it is important to identify the stands that are most likely to tolerate HWA infestations so that hemlock can be preserved as a component of forest habitats in the region. At the same time, stands likely to suffer high rates of mortality can be evaluated for integrated pest management activities or conversion to other species.