Palaseanu_Lovejoy, M. E., ETI / USGS, St. Petersburg, USA, mpal@usgs.gov
Nayegandhi, A. ., ETI / USGS, St. Petersburg, USA, anayegandhi@usgs.gov
Brock, J. ., USGS, St. Petersburg, USA, jbrock@usgs.gov
Wright, C. W., NASA, Wallops Islands, USA,
Woodman, R. ., NPS, Gulf Coast Network, Lafayette, USA, Robert_Woodman@nps.gov

UNSUPERVISED CLASSIFICATION OF VEGETATION COMMUNITIES USING AIRBORNE LIDAR DATA AT JEAN LAFITTE NATIONAL PARK, LOUISIANA, USA

This study evaluates the capabilities of the NASA Experimental Advanced Airborne Research Lidar (EAARL) to delineate vegetation communities in Jean Lafitte National Park (JELA), Louisiana, using a hierarchical approach. Five-meter-resolution grids of bare earth (BE), canopy height (CH), canopy reflection ratio (CRR), and height of the median energy (HOME) were derived. We used a statistical based approach to divide the CH in 5 distinct classes and for each height class we carried out a principal component analysis (PCA) and an independent component analysis (ICA). Within each height class, original metrics, principal components (PC) and independent components (IC) were sub-classified in 4 groups either by k-means or neural-gas algorithms. Unsupervised original metrics classifications and PCA- and ICA-based classifications were compared with JELA color infrared aerial photography. Ground-truth data are being acquired at two test sites to assess species composition, canopy cover, and to identify biologically consistent vegetation patches. Our study reveals that neural-gas PCA- and ICA-based classifications performed better in categorizing different vegetation communities than direct unsupervised classification of the four metrics or k-means based classifications of PC and IC.

Poster presentation

Presentation is given by student: No
Session #:078
Date: 03-05-2008
Time: 17:30 - 19:30

Top of Page | Home Page | Contact | Help | Search | Print This Page