Amazon Forest Structure from IKONOS Satellite Data and the Automated Characterization of Forest Canopy Properties
Article first published online: 25 SEP 2007
2008 The Author(s) Journal compilation © 2008 by The Association for Tropical Biology and Conservation
Volume 40, Issue 2, pages 141–150, March 2008
How to Cite
Palace, M., Keller, M., Asner, G. P., Hagen, S. and Braswell, B. (2008), Amazon Forest Structure from IKONOS Satellite Data and the Automated Characterization of Forest Canopy Properties. Biotropica, 40: 141–150. doi: 10.1111/j.1744-7429.2007.00353.x
- Issue published online: 25 SEP 2007
- Article first published online: 25 SEP 2007
- Received 16 March 2007; revision accepted 24 May 2007.
- automated algorithm;
- crown delineation;
- crown width;
- rain forest;
- tropical forest
We developed an automated tree crown analysis algorithm using 1-m panchromatic IKONOS satellite images to examine forest canopy structure in the Brazilian Amazon. The algorithm was calibrated on the landscape level with tree geometry and forest stand data at the Fazenda Cauaxi (3.75° S, 48.37° W) in the eastern Amazon, and then compared with forest stand data at Tapajos National Forest (3.08° S, 54.94° W) in the central Amazon. The average remotely sensed crown width (mean ± SE) was 12.7 ± 0.1 m (range: 2.0–34.0 m) and frequency of trees was 76.6 trees/ha at Cauaxi. At Tapajos, remotely sensed crown width was 13.1 ± 0.1 m (range: 2.0–38.0 m) and frequency of trees was 76.4 trees/ha. At both Cauaxi and Tapajos, the remotely sensed average crown widths were within 3 percent of the crown widths derived from field measurements, although crown distributions showed significant differences between field-measured and automated methods. We used the remote sensing algorithm to estimate crown dimensions and forest structural properties in 51 forest stands (1 km2) throughout the Brazilian Amazon. The estimated crown widths, tree diameters (dbh), and stem frequencies differed widely among sites, while estimated biomass was similar among most sites. Sources of observed errors included an inability to detect understory crowns and to separate adjacent, intermingled crowns. Nonetheless, our technique can serve to provide information about structural characteristics of large areas of unsurveyed forest throughout Amazonia.