Quantifying historical changes in habitat availability for endangered species: use of pixel- and object-based remote sensing.

Published online
03 Jun 2009
Content type
Journal article
Journal title
Journal of Applied Ecology
DOI
10.1111/j.1365-2664.2009.01637.x

Author(s)
Pringle, R. M. & Syfert, M. & Webb, J. K. & Shine, R.
Contact email(s)
pringle@stanford.edu

Publication language
English

Abstract

Establishing medium- to long-term trends in habitat availability for endangered species is important for determining the causes of historical population declines and for designing effective management plans. For some animal species, relative habitat availability can be estimated using time series of aerial photographs, but the limited information in old black-and-white images makes it challenging to estimate accurately at large spatial scales. Australia's most endangered snake, the broad-headed snake Hoplocephalus bungaroides, requires unshaded, exfoliated sandstone rocks for shelter. Using digitized aerial photographs of four sites from 1941 and 1971, and a Quickbird satellite image from 2006, we estimated the trend in habitat availability for a well-studied population of H. bungaroides in New South Wales. We did this using both traditional, pixel-based classifications and a more recently developed object-based approach. Both classification methods revealed substantial, monotonic decreases (by 24-77%) in the percentage cover of bare rock from 1941 to 2006 at all sites, with concomitant (although non-monotonic) net increases in percentage tree-canopy cover of 3-70%. The cover of herbaceous and low-shrub vegetation appeared variable and was poorly resolved in the classifications. Both classification methods yielded adequate (>70%) accuracy for most images, but the object-based classifications were consistently more accurate than the pixel-based classifications and performed better when analysing images (from 1971) with comparatively low resolution. Synthesis and applications. Our study indicates that habitat availability for broad-headed snakes has decreased considerably over 65 years due to increased shading from vegetation, and that object-based image analysis is a promising tool for assessing habitat trends using historical photographs. Multiple studies have reported woody-vegetation increases over multi-decadal time-scales in Australia and elsewhere, but few have tied these trends to habitat availability for animal species of concern. The phenomenon we document may therefore be quite widespread, and advances in remote sensing (including applications and refinements of the methods presented here) will be invaluable for revealing these trends and designing ameliorative management strategies. Management plans to conserve broad-headed snakes and similarly threatened species should include either controlled burns or regular targeted removal of vegetation, preferably after a rigorous comparison of the relative economic and biological costs and benefits of each.

Key words