Application of QuickBird and aerial imagery to detect Pinus radiata in remnant vegetation
Article first published online: 29 AUG 2010
© 2009 The Authors. Journal compilation © 2009 Ecological Society of Australia
Volume 35, Issue 6, pages 624–635, September 2010
How to Cite
HABY, N., TUNN, Y. and CAMERON, J. (2010), Application of QuickBird and aerial imagery to detect Pinus radiata in remnant vegetation. Austral Ecology, 35: 624–635. doi: 10.1111/j.1442-9993.2009.02070.x
- Issue published online: 29 AUG 2010
- Article first published online: 29 AUG 2010
- Accepted for publication September 2009.
- accuracy assessment;
- Pinus radiata;
- remnant vegetation community;
- remote sensing;
- weed control
The invasion of Pinus radiata from long-term established plantations is contributing to the degradation of fragmented and isolated remnants of native vegetation. Within the south-east of South Australia, the 20 vegetation communities that occur within 500 m of a plantation edge are at risk, including nine state threatened communities. To plan effective mitigation strategies, the current extent and distribution of P. radiata needs to be ascertained. High spatial resolution, multispectral QuickBird imagery and aerial photography were used to classify P. radiata within eucalypt and acacia woodlands, melaleuca shrubland, modified pasture and an Eucalyptus globulus plantation. Unsupervised classification of aerial photography gave the best result showing reasonable conformity with the observed distribution of P. radiata at the site scale. However, the 9.4 ± 13.5 (SD) cover classified in the quadrats sampled for the accuracy assessment exceeded the 1.4 ± 2.4 (SD) P. radiata cover determined from an independent dataset. Only 30.1 ± 37.9% (SD) of trees within the quadrats and 9.40 ± 13.49% (SD) of their foliage cover were classified. Trees detected by partial classification of canopy were positively correlated with both tree height and canopy diameter. Overall, the low detection rates were attributed to insufficient spectral resolution. Using higher resolution imagery, together with an object-based image analysis or combination of multispectral and airborne digital image classification, restricted to large emergent adult trees using LiDAR analysis, is likely to improve adult P. radiata detection accuracy.