Eastern hemlock is an ecologically important tree species which helps modulate soil temperature, moisture, nutrient levels, and headwater stream temperature. It is an exceptionally shade tolerant, late-successional, and long-lived tree. Kentucky hemlock grows in patches usually in moist, shaded coves throughout the eastern part of the state. Unfortunately, this vital species is under attack from the hemlock woolly adelgid.

Hemlock woolly adelgids (HWA) are very small insects, native to Asia, which feed on all hemlock species. HWA has been spreading through hemlock forests along the east coast of the U.S. for the last two decades. These harmful parasites can kill a tree in a matter of a few years.....much too quickly for hemlocks to replace themselves. In 2006, HWA was found on Pine Mountain in southeastern Kentucky. As of February 2009, there were 11 counties in Kentucky with infestations.

To slow the spread of this invader, my research goal is to create maps of eastern hemlock distribution and areas at high risk of potential HWA infestations. This will be accomplished using GIS and remote sensing technologies.

Locations of eastern hemlock trees have been recorded for several years by a number of state employees. A geodatabase of hemlock locations is maintained for use in this project. I am using Landsat 7 ETM+ satellite imagery to classify eastern hemlock cover. The wavelengths detected by the satellite, along with several environmental variables, are used in a classification and regression tree analysis (CART). The output of this statistical model will determine the areas covered by eastern hemlock. The accuracy of the resulting map will be checked with known locations of eastern hemlock tree, captured with GPS units.

As a result of its small size, hemlock woolly adelgids can be easily overlooked or feeding at the top of the tree, invisible to an observer on the ground. Because of this, we cannot be significantly certain that a tree is not infested with HWA. Therefore, I am using a presence-only species distribution model.....specifically MaxEnt. GPS coordinates of known HWA infestations are modeled on independent variables representing likely corridors of dispersal. For example, distances from trails and power lines are variables included in the model because is it possible for HWA to be transported by humans and birds. The result of the MaxEnt model is a map of continuous probability of likely HWA infestation. Using an appropriate threshold, I will convert that product into a high-risk/low-risk map of HWA spread. This map will be useful in helping land managers and state agencies combat the invasion of the hemlock woolly adelgid.