Identification of specific tree species in ancient semi-natural woodland from digital aerial sensor imagery

Remote sensing has great potential as a source of information on tree species. The classification approaches used commonly to extract species information from remotely sensed imagery typically aim to optimize the overall accuracy of species identification, a target which need not satisfy the require...

Full description

Saved in:
Bibliographic Details
Published in:Ecological applications 2005-08, Vol.15 (4), p.1233-1244
Main Authors: Foody, G.M, Atkinson, P.M, Gething, P.W, Ravenhill, N.A, Kelly, C.K
Format: Article
Language:English
Subjects:
accuracy
Acer pseudoplatanus
Aerial photography
Applied ecology
Datasets
digital aerial sensor imagery
Discriminant analysis
Ecology
Foodie culture
forest ecology
forest trees
Remote sensing
Sensors
Species
species diversity
thematic mapping
thematic maps
tree species classification
Trees
woodlands
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Remote sensing has great potential as a source of information on tree species. The classification approaches used commonly to extract species information from remotely sensed imagery typically aim to optimize the overall accuracy of species identification, a target which need not satisfy the requirements of a particular user. Often users are interested in a specific species or subset of species, and these may not be accurately identified in a conventional classification. Here, a two-phase classification approach was used to map specific species from aerial sensor imagery of an ancient British woodland. Particular attention was focused on the identification of sycamore since this is displacing the native ash and information on its distribution would enhance basic understanding and management activities. The results show that the classification approach can be adapted to focus on a specific species of interest and used to increase classification accuracy significantly. For example, sycamore was classified to a low accuracy when a conventional approach to classification with a neural network was used (46.6-63.6%, depending on perspective), but the adoption of the two-phase approach increased its accuracy significantly (82.3-93.3%). The results demonstrate the ability to map specific class(es) of interest accurately from remotely sensed imagery. The approach used also highlights the ability to tailor an analysis to the specific requirements of the ecological study in hand and is of broad applicability.
ISSN:1051-0761
1939-5582
DOI:10.1890/04-1061