PS 53-179 - Remote sensing and GIS to model endangered American burying beetle abundance across a landscape and to determine the optimal spatial scale for habitat samples

Wednesday, August 10, 2011
Exhibit Hall 3, Austin Convention Center
Douglas R. Leasure, Biological Sciences, University of Arkansas, Fayetteville, AR

An economical and effective new method was developed to assess and monitor endangered American burying beetle (Nicrophorus americanus) habitat across large spatial scales using GIS and remote sensing.  Fort Chaffee Maneuver Training Center, a 25,000 hectare military installation in western Arkansas, supports one of the largest remaining N. americanus populations.  Landsat satellite imagery and a digital elevation model were used to derive predictor variables for multiple regression analysis of N. americanus abundance across Fort Chaffee.  Habitat variables derived from Landsat imagery included normalized difference vegetation index, Kauth-Thomas tasseled cap, and principal components.  Slope and aspect were determined from a digital elevation model.  Mean and standard deviations of these habitat variables within various sample radii around trap locations were used as predictor variables.  The response variable was a 3-year average N. americanus relative abundance (4th root transformed) from 60 locations at Fort Chaffee with baited pitfall trap transects.  The optimal sample radius around traps was determined by comparing models using Akaike information criteria (AIC).  The optimal sample radius was expected to approximate the sample range of N. americanus pitfall traps, conservatively estimated to be 800 m by the U.S. Fish and Wildlife Service.  All variable combinations were compared using AIC.


Principal components ordination of original Landsat TM bands produced predictor variables that were best suited for multiple regression analysis because of collinearity issues with other variable sets.  The best model contained three predictor variables: principal component 3 mean, slope mean, and slope standard deviation (adj r2 = 0.7172, P < 0.0001).  The best model used a 1600 m habitat sample radius, twice the estimated sample range of the traps.  This model was used to predict N. americanus abundance for each 30 m x 30 m pixel in the Fort Chaffee imagery and elevation model.  This provided an index of habitat suitability across Fort Chaffee that was in general agreement with a previous study of N. americanus habitat preference at Fort Chaffee based on vegetation communities and soils.  The previous study was repeated using various sample radii and the 1600 m optimal habitat sample radius was confirmed.  These results emphasize the importance of considering scale in habitat studies, providing an economical method to determine a suitable scale.  This analysis is now being replicated at the regional scale using historical N. americanus datasets from Arkansas and Oklahoma, Landsat TM imagery, digital elevation models, GAP land cover data, and NRCS soil maps.

Copyright © . All rights reserved.
Banner photo by Flickr user greg westfall.