HR: 0830h
AN: B41B-0882 [PDF]
TI: An Analysis of Two Amazonian Forests using an Automated Crown Detection Algorithm and IKONOS
Imagery
AU: * Palace, M
EM: palace@kaos.sr.unh.edu
AF: Complex Systems Research, Morse Hall, University of New Hampshire, Durham, NH 03824 United States
AU: Keller, M
EM: michael.keller@unh.edu
AF: International Institute of Tropical Forestry, USDA Forest Service, Rio Piedras, PR 00928-5000 United States
AU: Braswell, B
EM: rob.braswell@unh.edu
AF: Complex Systems Research, Morse Hall, University of New Hampshire, Durham, NH 03824 United States
AU: Hagen, S
EM: steve.hagen@unh.edu
AF: Complex Systems Research, Morse Hall, University of New Hampshire, Durham, NH 03824 United States
AU: Asner, G
EM: gasner@globalecology.stanford.edu
AF: Department of Global Ecology, Carnegie Institution of Washington, Stanford University, Stanford, CA
94305 United States
AB:
Spatial analysis of tree crowns in high-resolution satellite images (e.g., IKONOS) can improve estimates of carbon stocks at
regional and local scales, aid in demographic studies at the stand level, and aid in forestry efforts. These data, at 1-4
meter resolution, represent a direct view of forest structural properties at the landscape level, but it is often difficult
to convert the detail seen by eye into useful quantitative information. In this study, we developed an automated crown
detection program that combines iterative local maximum finding and local minima finding methods with analysis of extracted
ordinate data. We calibrated the algorithm at one site, Fazenda Cauaxi ($3.75\deg$ S, $48.37\deg$ W), with tree geometry and
stand data and validated it with stand data at another site, Tapajos National Forest ($3.08\deg$ S, $54.94\deg$ W).
Panchromatic IKONOS images were used from these two Amazonian undisturbed forest sites in Para, Brazil. Mean, minimum and
maximum crown widths for field data at Cauaxi are (mean 10.6 m +/- 1.6 S.D., min 1.0 m, max 40.7 m) and automated estimation
are (mean 12.7 m +/- 5.6 S.D., min 2.0 m, max 34.0 m). Mean DBH for field data at Cauaxi is 30 cm +/- 3.4 S.D. and for
automated routine is 54.0 cm +/- 19 S.D. Automated estimation at Tapajos of mean, minimum and maximum crown widths are (mean
12.3 m +/- 5.9 S.D., min 2.0 m, max 38.0 m). Mean DBH for field data is 51.9 cm +/- 16.9 S.D. and for automated routine is
53.0 cm +/- 20.2 S.D. The automated crown algorithm provides good estimates of mean values and distributions for crown width
and stand density. Average DBH estimates are similar to field data. Automated DBH and biomass distribution requires
improvement, but this estimation is also limited by lack of site-specific allometric equations. Forest structural
characteristics are extremely difficult to obtain in many tropical forests, but high-resolution images may well hold the key
to unlocking information about the functioning of these highly important ecosystems.
DE: 0400 Biogeosciences
DE: 1640 Remote sensing
DE: 1694 Instruments and techniques
SC: Biogeosciences [B]
MN: 2003 Fall Meeting