Monday, December 11, 2006 - 8:35 AM
0394

Applying probabilistic cellular automata to model county level southern pine beetle infestations

Adrian Duehl, ajduehl@ncsu.edu1, Fred Hain, fphain@ncsu.edu1, and John Bishir, bishir@ncsu.edu2. (1) North Carolina State University, Forest Entomology, 201 Buck Jones Road, Raleigh, NC, (2) North Carolina State University, Department of Mathematics, Box 8205, Raleigh, NC

The Southern Pine Beetle is the most destructive insect pest in southern forests. Its population dynamics follow a periodic eruptive pattern, with outbreaks occurring across large geographic areas. The best historic outbreak data, which has been continuously updated from 1960 to 2004, is a record of counties with infested stands. This record was geographically referenced in ARC GIS and converted into a matrix with cells about the size of the smallest county. For each year the cells were classified by the infested county area contained within them: majority uninfested (0), or majority infested (1). The patterns of infested cells over time were then analyzed to see if the patterns in previous years would influence the chance of a focal cell being infested in a given year. Moving windows from 3x3 to 7x7 were used to determine the optimal window size for analysis. One and two year timeframes, using either the number of infested cells or the pattern created by the exact location of the infested cells, were also tested. The amount of variation possible in exact pattern analysis caused many patterns to be unique, especially in the larger windows, so predictions could not be made for these patterns. The number of infested cells in each window made good predictions in the larger templates. The best predictions balance sufficient window variability and predictions of both high and low infestation severity.


Species 1: Coleoptera Curculionidae Dendroctonus frontalis (Southern Pine Beetle)

[ Recorded presentation ] Recorded presentation