Liana canopy cover mapped throughout a tropical forest with high-fidelity imaging spectroscopy

David C. Marvin, Gregory P. Asner, Stefan A. Schnitzer

Research output: Contribution to journalArticle

13 Citations (Scopus)

Abstract

Increasing size and abundance of lianas relative to trees are pervasive changes in Neotropical forests that may lead to reduced forest carbon stocks. Yet the liana growth form is chronically understudied in large-scale tropical forest censuses, resulting in few data on the scale, cause, and impact of increasing lianas. Satellite and airborne remote sensing provide potential tools to map and monitor lianas at much larger spatial and rapid temporal scales than are possible with plot-based forest censuses. We combined high-resolution airborne imaging spectroscopy and a ground-based tree canopy census to investigate whether tree canopies supporting lianas could be discriminated from tree canopies with no liana coverage. Using support vector machine algorithms, we achieved accuracies of nearly 90% in discriminating the presence-absence of lianas, and low error (15.7% RMSE) when predicting liana percent canopy cover. When applied to the full image of the study site, our model had a 4.1% false-positive error rate as validated against an independent plot-level dataset of liana canopy cover. Using the derived liana cover classification map, we show that 6.1%-10.2% of the 1823. ha study site has high-to-severe (50-100%) liana canopy cover. Given that levels of liana infestation are increasing in Neotropical forests and can result in high tree mortality, the extent of high-to-severe liana canopy cover across the landscape may have broad implications for ecosystem function and forest carbon storage. The ability to accurately map landscape-scale liana infestation is crucial to quantifying their effects on forest function and uncovering the mechanisms underlying their increase.

Original languageEnglish (US)
Pages (from-to)98-106
Number of pages9
JournalRemote Sensing of Environment
Volume176
DOIs
StatePublished - Apr 1 2016
Externally publishedYes

Fingerprint

tropical forests
tropical forest
lianas
spectroscopy
canopy
image analysis
Spectroscopy
Imaging techniques
Carbon
census
Ecosystems
Support vector machines
Remote sensing
Satellites
tree mortality
growth form
ecosystem function
carbon sequestration
carbon sinks
remote sensing

Keywords

  • Carnegie Airborne Observatory
  • Hyperspectral imaging
  • Panama
  • Support vector machine

ASJC Scopus subject areas

  • Soil Science
  • Geology
  • Computers in Earth Sciences

Cite this

Liana canopy cover mapped throughout a tropical forest with high-fidelity imaging spectroscopy. / Marvin, David C.; Asner, Gregory P.; Schnitzer, Stefan A.

In: Remote Sensing of Environment, Vol. 176, 01.04.2016, p. 98-106.

Research output: Contribution to journalArticle

@article{472e47512e0443d3946dbf31e148d55a,
title = "Liana canopy cover mapped throughout a tropical forest with high-fidelity imaging spectroscopy",
abstract = "Increasing size and abundance of lianas relative to trees are pervasive changes in Neotropical forests that may lead to reduced forest carbon stocks. Yet the liana growth form is chronically understudied in large-scale tropical forest censuses, resulting in few data on the scale, cause, and impact of increasing lianas. Satellite and airborne remote sensing provide potential tools to map and monitor lianas at much larger spatial and rapid temporal scales than are possible with plot-based forest censuses. We combined high-resolution airborne imaging spectroscopy and a ground-based tree canopy census to investigate whether tree canopies supporting lianas could be discriminated from tree canopies with no liana coverage. Using support vector machine algorithms, we achieved accuracies of nearly 90{\%} in discriminating the presence-absence of lianas, and low error (15.7{\%} RMSE) when predicting liana percent canopy cover. When applied to the full image of the study site, our model had a 4.1{\%} false-positive error rate as validated against an independent plot-level dataset of liana canopy cover. Using the derived liana cover classification map, we show that 6.1{\%}-10.2{\%} of the 1823. ha study site has high-to-severe (50-100{\%}) liana canopy cover. Given that levels of liana infestation are increasing in Neotropical forests and can result in high tree mortality, the extent of high-to-severe liana canopy cover across the landscape may have broad implications for ecosystem function and forest carbon storage. The ability to accurately map landscape-scale liana infestation is crucial to quantifying their effects on forest function and uncovering the mechanisms underlying their increase.",
keywords = "Carnegie Airborne Observatory, Hyperspectral imaging, Panama, Support vector machine",
author = "Marvin, {David C.} and Asner, {Gregory P.} and Schnitzer, {Stefan A.}",
year = "2016",
month = "4",
day = "1",
doi = "10.1016/j.rse.2015.12.028",
language = "English (US)",
volume = "176",
pages = "98--106",
journal = "Remote Sensing of Environment",
issn = "0034-4257",
publisher = "Elsevier Inc.",

}

TY - JOUR

T1 - Liana canopy cover mapped throughout a tropical forest with high-fidelity imaging spectroscopy

AU - Marvin, David C.

AU - Asner, Gregory P.

AU - Schnitzer, Stefan A.

PY - 2016/4/1

Y1 - 2016/4/1

N2 - Increasing size and abundance of lianas relative to trees are pervasive changes in Neotropical forests that may lead to reduced forest carbon stocks. Yet the liana growth form is chronically understudied in large-scale tropical forest censuses, resulting in few data on the scale, cause, and impact of increasing lianas. Satellite and airborne remote sensing provide potential tools to map and monitor lianas at much larger spatial and rapid temporal scales than are possible with plot-based forest censuses. We combined high-resolution airborne imaging spectroscopy and a ground-based tree canopy census to investigate whether tree canopies supporting lianas could be discriminated from tree canopies with no liana coverage. Using support vector machine algorithms, we achieved accuracies of nearly 90% in discriminating the presence-absence of lianas, and low error (15.7% RMSE) when predicting liana percent canopy cover. When applied to the full image of the study site, our model had a 4.1% false-positive error rate as validated against an independent plot-level dataset of liana canopy cover. Using the derived liana cover classification map, we show that 6.1%-10.2% of the 1823. ha study site has high-to-severe (50-100%) liana canopy cover. Given that levels of liana infestation are increasing in Neotropical forests and can result in high tree mortality, the extent of high-to-severe liana canopy cover across the landscape may have broad implications for ecosystem function and forest carbon storage. The ability to accurately map landscape-scale liana infestation is crucial to quantifying their effects on forest function and uncovering the mechanisms underlying their increase.

AB - Increasing size and abundance of lianas relative to trees are pervasive changes in Neotropical forests that may lead to reduced forest carbon stocks. Yet the liana growth form is chronically understudied in large-scale tropical forest censuses, resulting in few data on the scale, cause, and impact of increasing lianas. Satellite and airborne remote sensing provide potential tools to map and monitor lianas at much larger spatial and rapid temporal scales than are possible with plot-based forest censuses. We combined high-resolution airborne imaging spectroscopy and a ground-based tree canopy census to investigate whether tree canopies supporting lianas could be discriminated from tree canopies with no liana coverage. Using support vector machine algorithms, we achieved accuracies of nearly 90% in discriminating the presence-absence of lianas, and low error (15.7% RMSE) when predicting liana percent canopy cover. When applied to the full image of the study site, our model had a 4.1% false-positive error rate as validated against an independent plot-level dataset of liana canopy cover. Using the derived liana cover classification map, we show that 6.1%-10.2% of the 1823. ha study site has high-to-severe (50-100%) liana canopy cover. Given that levels of liana infestation are increasing in Neotropical forests and can result in high tree mortality, the extent of high-to-severe liana canopy cover across the landscape may have broad implications for ecosystem function and forest carbon storage. The ability to accurately map landscape-scale liana infestation is crucial to quantifying their effects on forest function and uncovering the mechanisms underlying their increase.

KW - Carnegie Airborne Observatory

KW - Hyperspectral imaging

KW - Panama

KW - Support vector machine

UR - http://www.scopus.com/inward/record.url?scp=84955508653&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84955508653&partnerID=8YFLogxK

U2 - 10.1016/j.rse.2015.12.028

DO - 10.1016/j.rse.2015.12.028

M3 - Article

AN - SCOPUS:84955508653

VL - 176

SP - 98

EP - 106

JO - Remote Sensing of Environment

JF - Remote Sensing of Environment

SN - 0034-4257

ER -