1
Brynna Bolger
December 2012
Vulnerability assessment of forests to invasion by the Asian Long Horned Beetle: a pilot study of Worcester County, MA.
One of the biggest threats to Massachusetts’ trees in both urban and forested areas is the destruction by invasive pests. The Hemlock Woolly Adelgid (HWA) has already decimated
Massachusetts’ hemlocks. Our ash, maple, birch, elm, willow and many others will be next if pests like the Emerald Ash Borer (EAB) and the Asian Longhorned Beetle (ALB) are allowed to spread into the state. Massachusetts has already had 2 close encounters with the Asian
Longhorned Beetle – a large infestation in Worchester identified in 2008 (currently in the process of eradication) and a very small infestation in Boston identified in 2010. While there are some options for control of HWA (both chemical and biological) the only real option for the
ALB is cutting down and destroying the infested trees. Unfortunately, this is not a viable solution in large forests, but only in urban and residential areas. Therefore, prevention or early detection is the best hope of saving Massachusetts’ forests.
One of the most common ways invasive pests spread to new areas is through firewood or other wood material. For example, both ALB and EAB initially spread to North America from
Asia, where they are native insects, in wooden shipping materials. Despite campaigns promoting
“burn it where you buy it” many people either do not realize the level of threat or are not exposed to the message. In addition, the areas where the largest number of out of state firewood is coming in is to campgrounds, state parks, etc. which are commonly surrounded by large stands of trees, and so the job of controlling an infestation would be extremely difficult. Therefore, making sure that the message of limiting the movement of firewood is a high priority, but especially in areas where the presence of host trees coincides with a large influx of out of state wood.
In this pilot study I worked to create a method of identifying forested areas at highest risk of invasion by the ALB. I chose to focus on Worcester County for this initial analysis because it is centrally located in the state and because it contains a city that already has a known infestation.
For the scope of this project I am focusing on Massachusetts forests and not trees in residential/urban areas because infestations in forests are much harder to initially identify and then to eliminate. This project is meant to be a pilot study – to design a method of identifying at risk areas in Massachusetts or the Northeast by using Worcester County as test run.
I looked only at forested areas – deciduous and mixed evergreen and deciduous – because the host trees of ALB are of this type. Additionally, I looked at population density of areas adjacent to forested areas because studies have found that introduction and success of invasives is correlated with human population density. Another important aspect to consider is that invasives typically thrive in disturbed habitats or edge habitats, so I selected for forested areas around roads because these forests are both disturbed and have large edges. Lastly, I located plant nurseries and campgrounds adjacent to the selected forest types. Using raster analysis I determined which areas have the most overlap of these factors as a method of assessing risk.
Data Sources
Census data
American Factfinder at http://census.gov

2
TOTAL POPULATION, 2010 SF1 100% Data
Metadata downloadable with census data
Census data in table format
Census geography
TIGER at http://census.gov
2012 TIGER Line/Shapefiles
Massachusetts Census Tracts, vector
Worcester County
MassGIS (M drive)
COUNTIES_POLY, vector
2000 http://www.mass.gov/anf/research-and-tech/it-serv-and-support/application-serv/office-ofgeographic-information-massgis/datalayers/counties.html
Forest type
Landfire, http://www.landfire.gov/
Existing Vegetation Type, raster, cells 30m
2008
Metadata: http://landfire.cr.usgs.gov/distmeta/servlet/gov.usgs.edc.MetaBuilder?TYPE=HTML&DATASE
T=f4m
Campgrounds
Google Maps, http://maps.google.com/
Search: Massachusetts Campgrounds
.Kml to Google Earth
Plant Nurseries
Google Maps, http://maps.google.com/
Search: Massachusetts Plant Nurseries
.Kml to Google Earth
Roads
MassGIS (M drive)
EOTMajorRoads
2000
Methods
Acquiring the Campgrounds & Nurseries layers
Using Google Maps I saved each campground and then each plant nursery (to eliminate those points that were not actually campgrounds or plant nurseries) to a Google drive map, which I then converted to a .kml file in Google Earth. I then used the Converting KML (Google Earth,
Google Maps) data to shape files for ArcGIS tips and tutorials sheet to open the files in ArcMap where I exported the data to a shapefile.

3
Initial Processing
For the Environment I set the snap raster and cell size to the Landfire data layer. The cell size was 30m. I set the mask and extent to the Worcester County layer, which I made by using the select by attribute tool on the COUNTIES_POLY layer to create a new layer of only Worcester
County.
After bringing each layer into the map I set the data frame projected coordinate system to
NAD_1983_StatePlaneMassachusettsMainland_FIPS_2001. For each individual vector layer
(not including the census tracts, major roads, and counties layer) I used the Project tool to set each layer on the PCS NAD_1983_StatePlaneMassachusettsMainland_FIPS_2001.
However, for the Landfire vegetation layer I used the Raster Project tool.
Creating the RISKFACT layer
The RISKFACT layer was composed of different factors that I found maybe related to the vulnerability of an area to invasion. ALB is primarily spread by human activities that involve the movement of untreated wood, such as firewood or wood packaging materials. Thus the
RISKFACT layer included campgrounds and plant nurseries. Furthermore, I chose to include major roads as they see high human traffic and because disturbed and edge habitats have been found more vulnerable to non-native species. I also included areas of highest population density because that has been correlated to introduction and success of invasives.
For each of the data layers I used the select by location tool to select for vector data inside the
Worcester County layer. I then created a new data layer out of each selection.
For the population density layer I first used the calculate geometry tool in the attribute table to calculate the area of each tract in the county tracts layer. I then joined the data from the Total
Population, 2010 SF1 100% Data table to the county tracts layer. I normalized population by the area of the tract and then selected for the tracts with the highest densities to create a layer of only those tracts. I then used the Euclidean Distance tool to make the layer a raster layer and the
Reclassify tool where 200 m=4, 400 m=3, 600 m=2, and >600 m=1 (Fig. 1).
Figure 1. Reclassified raster population density data layer of Worcester County.

4
For both the campground and plant nursery layers I followed the same set of steps. First I used the Buffer tool to create a buffer of 500 m to take the area of each point into account. I then used the Euclidean Distance tool to make the buffer layers raster layers and the Reclassify tool where
200 m=4, 400 m=3, 600 m=2, and >600 m=1 (Fig. 2a and b). a.
b.
Figure 2. a) Reclassified raster campground buffer data layer of Worcester County. B) Reclassified raster plant nurseries buffer data layer of Worcester County.
For the Major roads layer I used the Euclidean Distance tool to make the layer a raster layer and the Reclassify tool where 200 m=4, 400 m=3, 600 m=2, and >600 m=1 (Fig. 3).
Figure 3. Reclassified raster roads data layer of Worcester County.
Last, I used the Raster Calculator tool to combine the 4 reclassified raster data layers-
“Reclass_camp” + “Reclass_nurs” + “Reclass_road” + “Reclass_pop”. This layer was called
RISKFACT.

5
Creating the Vegetation layers
I used the Existing Vegetation Type data from Landfire to determine the vegetation types. For this layer I created two reclassified layers, one where deciduous closed forest=4, deciduous open forest=3, mixed evergreen/deciduous closed forest=2, mixed evergreen/deciduous open forest=1, and everything else was NoData. This layer I called ALB Habitat. The second layer, Vegetation type, was similarly reclassified except that the “everything else” was labeled 1, and the others were bumped up a number. This was so I could visualize only vulnerability in areas of forests that are habitat for the ALB (with some discrepancy in suitability) and to visualize vulnerability where vegetation type is simply another risk factor.
I used the Raster Calculator tool to combine the ALB Habitat layer and the RISKFACT layer and to combine the Vegetation type layer and the RISKFACT layer.
Final Products
The 5 maps I included in my final poster were RISKFACT, ALB Habitat, Vegetation type, and then the combination of the latter 2 with RISKFACT. For each map I used the Extraction by
Mask tool to clip the maps to the Worcester County layer.
Difficulties & Limitations
One of the major limitations and difficulties with this project had to do with the campground and plant nursery data. Firstly, by using Google Maps I was able to get a more complete list of the campgrounds and nurseries in the state, however I had to go through each one individually to check if it was actually a campground or plant nursery. In many cases similar business types would be incorrectly included in the Google Maps search.
Furthermore, this data was somewhat limiting because I did not know the area of each business.
Additionally, I did not know how many people come to each, how many people came from out of state to each campground, or how much plant material came to the nurseries from Asia or from infested states in the U.S. All of these components would have allowed me to better assess the risk associated with the campground or plant nursery. Dealing with the road data presented a similar problem. I did not take the widths of the different roads into account because it was not included on the attribute table and I did not use the classification column to look up average widths of each road type as I was limited by time.
Because of the large number of layers involved in this project and because I had to project, rasterize, and reclassify most of them it used up a lot of space. I constantly had to check and delete contents of my H drive when tools would not work due to lack of space. Furthermore, I had a lot of trouble with ArcMap shutting down while I was using the reclassify tool.
A major limitation of this pilot study was that I concentrated forested areas, however, as is clear by the risk factors map, human activities in non-forested areas provide many opportunities for introduction. Furthermore, there are additional factors that increase the risk of introduction, such as points where industrial or commercial products are brought to the county (airports, train stations, etc.) I was also limited by the specificity of the vegetation. There is a wide variety of tree types ALB inhabit, but they do so with varying degrees of preference and success.

6
Additionally, forests with increased heterogeneity are less susceptible to invasion. Therefore, knowledge of tree species would have enabled me to better determine risk.
Results
Based on this initial analysis, it appears that there are several areas throughout the county where the risk factors (campgrounds, plant nurseries, roads, and highest population density) are concentrated (Risk Factors). For the most part these areas are around high population densities.
The largest area of high risk is the Worcester area. However, when vegetation type and habitat are taken into account, the area of highest vulnerability appears to be the Northwest region of the state. Furthermore, although there are small areas of vulnerability, the areas shown to be at highest risk by the human-related factors are no longer of concern once vegetation type is taken into consideration.
Conclusions
The Risk Factors map shows that it may not be necessary to combine all of the risk factors – simply using highest population density may be sufficient. This has been shown historically in the state, as the two areas of infestation have been Boston and Worcester, two major cities and thus areas of high human density. On the other hand, the vulnerabilities maps show that these areas of highest influence by human-related risks are not significant once vegetation type is considered. This exhibits the interaction of anthropogenic and ecological factors in determining where pests are introduced and then where they can become established.
As a whole I think that the project was an effective first step towards assessing areas of vulnerability in Massachusetts. Obviously, there are many factors that could deter or promote introduction and establishment, but since this project was termed a “pilot study” the methods rather than the results were important. I believe the idea of looking for areas of overlap of known risk factors of invasion was good, especially because most parts of the Northeastern U.S. are already known to be suitable habitat for the ALB as far as climate and presence of host trees
(Townsend 2004).
Annotated Sources
Hu, J., Angeli, S., Schuetz, S., Luo, Y., Hajek, A.E. 2009. Ecology and management of exotic and endemic Asian longhorned beetle Anoplophora glabripennis . Agricultural and Forest
Entomology. 11: 359-375.
This article includes an extensive review of research regarding the biology and geographic range of the ALB. It lists the types of trees that act as hosts of the pest and how good of a host they are. Furthermore, it provides a lot of insight on what types of areas are conducive to infestations and how far the beetle can move naturally. It also includes some information on several projects that utilized GIS to determine the spread of the beetle. This includes maps of where current infestation are in North America and a map looking at the climate of North America and predicting areas of possible infestation.

7
Townsend Peterson, A. and Vieglais, D.A. 2001. Predicting species invasions using ecological niche modeling: new approaches from bioinfomatics attack a pressing problem.
BioScience. 51(5): 363-371.
This article is useful because it depicts current methods of dealing with invasive species – such as mapping spread. However, it explains the necessity of developing methods of predicting where infestations can occur and suggests the combination of biological knowledge and geographic information – which is basically what I am hoping to do. Furthermore, it presents maps on some other invasive species and explains how the ecological niche information was gathered and projected onto a new area (North America).
Townsend Peterson, A., Scachetti-Pereira, R., Hargove, W.W. 2004. Potential geographic distribution of Anoplophora glabripennis (Coleoptera: Cerambycidae) in North America.
American Midland Naturalist. 151(1): 170-178.
In this study the researchers used ecological niche modeling to predict where future invasions were likely to occur in North America. They did so by determining the ecological niche factors in its native range and projecting them onto North America. By doing so they concluded that much of the Northeast is at risk while the Northwest is not as vulnerable. However, they determined that the most at risk area is just south of the Great Lakes – which is indeed where many of the infestations are. Although this is obviously a much larger scale and focuses mainly on climate and land cover, it does provide useful insight as to what factors are important in determining risk.
Smith, M.T., Bancroft, J., Gao, R., Li, G. Dispersal potential of Anoplophora glabripennis
Motsch. USDA Agricultural Research Service, Beneficial Insects Introduction Research
Lab 501.
This study looks specifically at the dispersal rates of adult Asian longhorned beetles. They contend that previous efforts at mapping potential spread and areas for monitoring have chosen arbitrary distances, which ignore the biology of the species. This will be useful in determining the distances from points of possible introduction that ALB could spread and thus the areas of highest risk.