A two-step approach to modeling urban host tree distributions for forest insects

Tuesday, November 18, 2014: 9:16 AM
E145 (Oregon Convention Center)
Frank Koch , Eastern Forest Environmental Threat Assessment Center, USDA - Forest Service, Research Triangle Park, NC
Mark Ambrose , Department of Forestry and Environmental Resources, North Carolina State University, Research Triangle Park, NC
Denys Yemshanov , Landscape Analysis and Applications, Natural Resources Canada, Canadian Forest Service, Sault Ste. Marie, ON, Canada
P. Eric Wiseman , Department of Forest Resources and Environmental Conservation, Virginia Polytechnic Institute and State University, Blacksburg, VA
Many alien insect species currently impacting forested ecosystems in North America first appeared in urban forests. Unfortunately, despite serving as critical gateways for the human-mediated spread of these and other forest pests, urban forests remain less well documented than their “natural” forest counterparts: only a small percentage of the more than 26,000 communities in the US and Canada have completed any sort of urban forest inventory, and these inventories have commonly been restricted to street trees. We devised a two-step approach that utilizes the available inventory data to comprehensively model urban host tree distributions at a regional scale. We illustrate the approach for three tree genera – ash (Fraxinus), maple (Acer), and oak (Quercus) – that are associated with high-profile forest insect pests. First, based on existing inventories, we use a suite of explanatory spatial variables to estimate the proportion of the total basal area (as a proxy for forest volume) occupied by each genus in non-inventoried communities. Second, we apply a similar suite of spatial variables to estimate the total basal area of these communities. We then combine these estimates to construct region-wide urban distribution maps for each genus. By merging these maps with similar data on natural forests (e.g., distribution maps developed from Forest Inventory and Analysis plot data), we are able to provide a more complete host setting for spread modeling efforts.