Developing National-Scale Surveys for Exotic Pests and Defining Their Reliability

John W. Coulston, William D. Smith, Frank H. Koch, and Frank J. Sapio

North Carolina State University Department of Forestry and Environmental Resources (1,3)
USDA Forest Service Southern Research Station (2)
USDA Forest Service Forest Health Technology Enterprise Team (4)

Every year, a large number of non-native insects and diseases affecting forest tree species are intercepted at U.S. ports of entry from shipments of live plants, solid wood packing materials, logs and raw wood products, and other commercial goods.  Pests that evade the inspection process may be accidentally introduced into nearby forests or, via interstate shipment, into other locations across the country.  Potentially serious exotic pests that have recently made inroads into U.S. forested landscapes include sudden oak death (caused by Phytophthora ramorum), detected in coastal California in 1995, emerald ash borer (Agrilus planipennis) in 2002 and the sirex woodwasp (Sirex noctilio), which was first detected in upstate New York in 2005.  For these three pests and others, national-scale early detection/rapid response survey protocols were implemented soon after their discovery.  The purpose of such surveys is to substantiate that areas outside initial invasion zones remain free of the pest of interest.  Typically, the statistical reliability of such surveys has not been defined, nor has a desired level of statistical reliability been used in designing the surveys.  Here, we adapt techniques used in preventative veterinary medicine for substantiating freedom from disease in animal herds to the spatial domain for forest pests.  To accomplish this, we use a bootstrap approach to estimate the conditional probability of survey success in the spatial domain.  The techniques may be applied in two different ways.  First, they may be used to determine optimal sample sizes for new surveys given a desired level of confidence as well as acceptable pest detection probability and prevalence thresholds.  Alternately, the techniques may be used to determine the detection probability and prevalence thresholds or the level of confidence based on sample size of existing surveys.  We also discuss techniques to adapt the bootstrap estimation when the sensitivity or specificity of the insect trap or disease test is imperfect.  The implementation of these techniques for future regional or national-scale surveys allows for optimal allocation of sample plots for maximizing reliability given limited funds to support future survey efforts.

Exotic Pests Session - Tuesday Afternoon

corresponding author:

William D. Smith
Southern Research Station
3041 E. Cornwallis Road
Research Triangle Park, NC 27709
(919) 549-4067
bdsmith@fs.fed.us