Mapping savanna tree species at ecosystem scales using support vector machine classification and BRDF correction on airborne hyperspectral and LiDAR data

Matthew S. Colgan, Claire A. Baldeck, Jean baptiste Féret, Gregory P. Asner

Research output: Contribution to journalArticle

111 Citations (Scopus)

Abstract

Mapping the spatial distribution of plant species in savannas provides insight into the roles of competition, fire, herbivory, soils and climate in maintaining the biodiversity of these ecosystems. This study focuses on the challenges facing large-scale species mapping using a fusion of Light Detection and Ranging (LiDAR) and hyperspectral imagery. Here we build upon previous work on airborne species detection by using a two-stage support vector machine (SVM) classifier to first predict species from hyperspectral data at the pixel scale. Tree crowns are segmented from the lidar imagery such that crown-level information, such as maximum tree height, can then be combined with the pixel-level species probabilities to predict the species of each tree. An overall prediction accuracy of 76% was achieved for 15 species. We also show that bidirectional reflectance distribution (BRDF) effects caused by anisotropic scattering properties of savanna vegetation can result in flight line artifacts evident in species probability maps, yet these can be largely mitigated by applying a semi-empirical BRDF model to the hyperspectral data. We find that confronting these three challenges-reflectance anisotropy, integration of pixel- and crown-level data, and crown delineation over large areas-enables species mapping at ecosystem scales for monitoring biodiversity and ecosystem function.

Original languageEnglish (US)
Pages (from-to)3462-3480
Number of pages19
JournalRemote Sensing
Volume4
Issue number11
DOIs
StatePublished - Nov 1 2012
Externally publishedYes

Fingerprint

savanna
ecosystem
pixel
imagery
biodiversity
bidirectional reflectance
detection
support vector machine
ecosystem function
herbivory
lidar
artifact
reflectance
anisotropy
flight
scattering
spatial distribution
vegetation
climate
monitoring

Keywords

  • CAO
  • Carnegie airborne observatory
  • Crown segmentation
  • Kruger national park
  • South Africa
  • Species mapping
  • SVM

ASJC Scopus subject areas

  • Earth and Planetary Sciences(all)

Cite this

Mapping savanna tree species at ecosystem scales using support vector machine classification and BRDF correction on airborne hyperspectral and LiDAR data. / Colgan, Matthew S.; Baldeck, Claire A.; Féret, Jean baptiste; Asner, Gregory P.

In: Remote Sensing, Vol. 4, No. 11, 01.11.2012, p. 3462-3480.

Research output: Contribution to journalArticle

@article{62c31c07259d4f6c9fbd4bdbbc463d8b,
title = "Mapping savanna tree species at ecosystem scales using support vector machine classification and BRDF correction on airborne hyperspectral and LiDAR data",
abstract = "Mapping the spatial distribution of plant species in savannas provides insight into the roles of competition, fire, herbivory, soils and climate in maintaining the biodiversity of these ecosystems. This study focuses on the challenges facing large-scale species mapping using a fusion of Light Detection and Ranging (LiDAR) and hyperspectral imagery. Here we build upon previous work on airborne species detection by using a two-stage support vector machine (SVM) classifier to first predict species from hyperspectral data at the pixel scale. Tree crowns are segmented from the lidar imagery such that crown-level information, such as maximum tree height, can then be combined with the pixel-level species probabilities to predict the species of each tree. An overall prediction accuracy of 76{\%} was achieved for 15 species. We also show that bidirectional reflectance distribution (BRDF) effects caused by anisotropic scattering properties of savanna vegetation can result in flight line artifacts evident in species probability maps, yet these can be largely mitigated by applying a semi-empirical BRDF model to the hyperspectral data. We find that confronting these three challenges-reflectance anisotropy, integration of pixel- and crown-level data, and crown delineation over large areas-enables species mapping at ecosystem scales for monitoring biodiversity and ecosystem function.",
keywords = "CAO, Carnegie airborne observatory, Crown segmentation, Kruger national park, South Africa, Species mapping, SVM",
author = "Colgan, {Matthew S.} and Baldeck, {Claire A.} and F{\'e}ret, {Jean baptiste} and Asner, {Gregory P.}",
year = "2012",
month = "11",
day = "1",
doi = "10.3390/rs4113462",
language = "English (US)",
volume = "4",
pages = "3462--3480",
journal = "Remote Sensing",
issn = "2072-4292",
publisher = "Multidisciplinary Digital Publishing Institute (MDPI)",
number = "11",

}

TY - JOUR

T1 - Mapping savanna tree species at ecosystem scales using support vector machine classification and BRDF correction on airborne hyperspectral and LiDAR data

AU - Colgan, Matthew S.

AU - Baldeck, Claire A.

AU - Féret, Jean baptiste

AU - Asner, Gregory P.

PY - 2012/11/1

Y1 - 2012/11/1

N2 - Mapping the spatial distribution of plant species in savannas provides insight into the roles of competition, fire, herbivory, soils and climate in maintaining the biodiversity of these ecosystems. This study focuses on the challenges facing large-scale species mapping using a fusion of Light Detection and Ranging (LiDAR) and hyperspectral imagery. Here we build upon previous work on airborne species detection by using a two-stage support vector machine (SVM) classifier to first predict species from hyperspectral data at the pixel scale. Tree crowns are segmented from the lidar imagery such that crown-level information, such as maximum tree height, can then be combined with the pixel-level species probabilities to predict the species of each tree. An overall prediction accuracy of 76% was achieved for 15 species. We also show that bidirectional reflectance distribution (BRDF) effects caused by anisotropic scattering properties of savanna vegetation can result in flight line artifacts evident in species probability maps, yet these can be largely mitigated by applying a semi-empirical BRDF model to the hyperspectral data. We find that confronting these three challenges-reflectance anisotropy, integration of pixel- and crown-level data, and crown delineation over large areas-enables species mapping at ecosystem scales for monitoring biodiversity and ecosystem function.

AB - Mapping the spatial distribution of plant species in savannas provides insight into the roles of competition, fire, herbivory, soils and climate in maintaining the biodiversity of these ecosystems. This study focuses on the challenges facing large-scale species mapping using a fusion of Light Detection and Ranging (LiDAR) and hyperspectral imagery. Here we build upon previous work on airborne species detection by using a two-stage support vector machine (SVM) classifier to first predict species from hyperspectral data at the pixel scale. Tree crowns are segmented from the lidar imagery such that crown-level information, such as maximum tree height, can then be combined with the pixel-level species probabilities to predict the species of each tree. An overall prediction accuracy of 76% was achieved for 15 species. We also show that bidirectional reflectance distribution (BRDF) effects caused by anisotropic scattering properties of savanna vegetation can result in flight line artifacts evident in species probability maps, yet these can be largely mitigated by applying a semi-empirical BRDF model to the hyperspectral data. We find that confronting these three challenges-reflectance anisotropy, integration of pixel- and crown-level data, and crown delineation over large areas-enables species mapping at ecosystem scales for monitoring biodiversity and ecosystem function.

KW - CAO

KW - Carnegie airborne observatory

KW - Crown segmentation

KW - Kruger national park

KW - South Africa

KW - Species mapping

KW - SVM

UR - http://www.scopus.com/inward/record.url?scp=84870853911&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84870853911&partnerID=8YFLogxK

U2 - 10.3390/rs4113462

DO - 10.3390/rs4113462

M3 - Article

AN - SCOPUS:84870853911

VL - 4

SP - 3462

EP - 3480

JO - Remote Sensing

JF - Remote Sensing

SN - 2072-4292

IS - 11

ER -