Spectroscopic determination of leaf traits using infrared spectra

Maria Buitrago Acevedo (Corresponding Author), Thomas A. Groen, Christoph A. Hecker, Andrew K. Skidmore

Research output: Contribution to journalArticleAcademicpeer-review

6 Citations (Scopus)
2 Downloads (Pure)

Abstract

Leaf traits characterise and differentiate single species but can also be used for monitoring vegetation structure and function. Conventional methods to measure leaf traits, especially at the molecular level (e.g. water, lignin and cellulose content), are expensive and time-consuming. Spectroscopic methods to estimate leaf traits can provide an alternative approach. In this study, we investigated high spectral resolution (6612 bands) emissivity measurements from the short to the long wave infrared (1.4-16.0 μm) of leaves from 19 different plant species ranging from herbaceous to woody, and from temperate to tropical types. At the same time, we measured 14 leaf traits to characterise a leaf, including chemical (e.g., leaf water content, nitrogen, cellulose) and physical features (e.g., leaf area and leaf thickness). We fitted partial least squares regression (PLSR) models across the SWIR, MWIR and LWIR for each leaf trait. Then, reduced models (PLSRred) were derived by iteratively reducing the number of bands in the model (using a modified Jackknife resampling method with a Martens and Martens uncertainty test) down to a few bands (4–10 bands) that contribute the most to the variation of the trait. Most leaf traits could be determined from infrared data with a moderate accuracy (65 < View the MathML source < 77% for observed versus predicted plots) based on PLSRred models, while the accuracy using the whole infrared range (6612 bands) presented higher accuracies, 74 < View the MathML source < 90%. Using the full SWIR range (1.4–2.5 μm) shows similarly high accuracies compared to the whole infrared. Leaf thickness, leaf water content, cellulose, lignin and stomata density are the traits that could be estimated most accurately from infrared data (with View the MathML source above 0.80 for the full range models). Leaf thickness, cellulose and lignin were predicted with reasonable accuracy from a combination of single infrared bands. Nevertheless, for all leaf traits, a combination of a few bands yields moderate to accurate estimations.
Original languageEnglish
Pages (from-to)237-250
Number of pages14
JournalInternational Journal of Applied Earth Observation and Geoinformation (JAG)
Volume69
DOIs
Publication statusPublished - Jul 2018

Fingerprint

Infrared radiation
Cellulose
Lignin
Water content
cellulose
lignin
Spectral resolution
Water levels
Nitrogen
water content
Monitoring
stomata
vegetation structure
emissivity
spectral resolution
leaf area
water level
nitrogen
monitoring

Keywords

  • Infrared spectra
  • Leaf spectra
  • Leaf traits
  • Leaf water content
  • Lignin
  • Spectroscopy
  • ITC-ISI-JOURNAL-ARTICLE
  • ITC-HYBRID

Cite this

@article{e24d60a4fbc44ddaa2b21345cebdcb3a,
title = "Spectroscopic determination of leaf traits using infrared spectra",
abstract = "Leaf traits characterise and differentiate single species but can also be used for monitoring vegetation structure and function. Conventional methods to measure leaf traits, especially at the molecular level (e.g. water, lignin and cellulose content), are expensive and time-consuming. Spectroscopic methods to estimate leaf traits can provide an alternative approach. In this study, we investigated high spectral resolution (6612 bands) emissivity measurements from the short to the long wave infrared (1.4-16.0 μm) of leaves from 19 different plant species ranging from herbaceous to woody, and from temperate to tropical types. At the same time, we measured 14 leaf traits to characterise a leaf, including chemical (e.g., leaf water content, nitrogen, cellulose) and physical features (e.g., leaf area and leaf thickness). We fitted partial least squares regression (PLSR) models across the SWIR, MWIR and LWIR for each leaf trait. Then, reduced models (PLSRred) were derived by iteratively reducing the number of bands in the model (using a modified Jackknife resampling method with a Martens and Martens uncertainty test) down to a few bands (4–10 bands) that contribute the most to the variation of the trait. Most leaf traits could be determined from infrared data with a moderate accuracy (65 < View the MathML source < 77{\%} for observed versus predicted plots) based on PLSRred models, while the accuracy using the whole infrared range (6612 bands) presented higher accuracies, 74 < View the MathML source < 90{\%}. Using the full SWIR range (1.4–2.5 μm) shows similarly high accuracies compared to the whole infrared. Leaf thickness, leaf water content, cellulose, lignin and stomata density are the traits that could be estimated most accurately from infrared data (with View the MathML source above 0.80 for the full range models). Leaf thickness, cellulose and lignin were predicted with reasonable accuracy from a combination of single infrared bands. Nevertheless, for all leaf traits, a combination of a few bands yields moderate to accurate estimations.",
keywords = "Infrared spectra, Leaf spectra, Leaf traits, Leaf water content, Lignin, Spectroscopy, ITC-ISI-JOURNAL-ARTICLE, ITC-HYBRID",
author = "{Buitrago Acevedo}, Maria and Groen, {Thomas A.} and Hecker, {Christoph A.} and Skidmore, {Andrew K.}",
year = "2018",
month = "7",
doi = "10.1016/j.jag.2017.11.014",
language = "English",
volume = "69",
pages = "237--250",
journal = "International Journal of Applied Earth Observation and Geoinformation (JAG)",
issn = "1569-8432",
publisher = "Elsevier",

}

Spectroscopic determination of leaf traits using infrared spectra. / Buitrago Acevedo, Maria (Corresponding Author); Groen, Thomas A.; Hecker, Christoph A.; Skidmore, Andrew K.

In: International Journal of Applied Earth Observation and Geoinformation (JAG), Vol. 69, 07.2018, p. 237-250.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Spectroscopic determination of leaf traits using infrared spectra

AU - Buitrago Acevedo, Maria

AU - Groen, Thomas A.

AU - Hecker, Christoph A.

AU - Skidmore, Andrew K.

PY - 2018/7

Y1 - 2018/7

N2 - Leaf traits characterise and differentiate single species but can also be used for monitoring vegetation structure and function. Conventional methods to measure leaf traits, especially at the molecular level (e.g. water, lignin and cellulose content), are expensive and time-consuming. Spectroscopic methods to estimate leaf traits can provide an alternative approach. In this study, we investigated high spectral resolution (6612 bands) emissivity measurements from the short to the long wave infrared (1.4-16.0 μm) of leaves from 19 different plant species ranging from herbaceous to woody, and from temperate to tropical types. At the same time, we measured 14 leaf traits to characterise a leaf, including chemical (e.g., leaf water content, nitrogen, cellulose) and physical features (e.g., leaf area and leaf thickness). We fitted partial least squares regression (PLSR) models across the SWIR, MWIR and LWIR for each leaf trait. Then, reduced models (PLSRred) were derived by iteratively reducing the number of bands in the model (using a modified Jackknife resampling method with a Martens and Martens uncertainty test) down to a few bands (4–10 bands) that contribute the most to the variation of the trait. Most leaf traits could be determined from infrared data with a moderate accuracy (65 < View the MathML source < 77% for observed versus predicted plots) based on PLSRred models, while the accuracy using the whole infrared range (6612 bands) presented higher accuracies, 74 < View the MathML source < 90%. Using the full SWIR range (1.4–2.5 μm) shows similarly high accuracies compared to the whole infrared. Leaf thickness, leaf water content, cellulose, lignin and stomata density are the traits that could be estimated most accurately from infrared data (with View the MathML source above 0.80 for the full range models). Leaf thickness, cellulose and lignin were predicted with reasonable accuracy from a combination of single infrared bands. Nevertheless, for all leaf traits, a combination of a few bands yields moderate to accurate estimations.

AB - Leaf traits characterise and differentiate single species but can also be used for monitoring vegetation structure and function. Conventional methods to measure leaf traits, especially at the molecular level (e.g. water, lignin and cellulose content), are expensive and time-consuming. Spectroscopic methods to estimate leaf traits can provide an alternative approach. In this study, we investigated high spectral resolution (6612 bands) emissivity measurements from the short to the long wave infrared (1.4-16.0 μm) of leaves from 19 different plant species ranging from herbaceous to woody, and from temperate to tropical types. At the same time, we measured 14 leaf traits to characterise a leaf, including chemical (e.g., leaf water content, nitrogen, cellulose) and physical features (e.g., leaf area and leaf thickness). We fitted partial least squares regression (PLSR) models across the SWIR, MWIR and LWIR for each leaf trait. Then, reduced models (PLSRred) were derived by iteratively reducing the number of bands in the model (using a modified Jackknife resampling method with a Martens and Martens uncertainty test) down to a few bands (4–10 bands) that contribute the most to the variation of the trait. Most leaf traits could be determined from infrared data with a moderate accuracy (65 < View the MathML source < 77% for observed versus predicted plots) based on PLSRred models, while the accuracy using the whole infrared range (6612 bands) presented higher accuracies, 74 < View the MathML source < 90%. Using the full SWIR range (1.4–2.5 μm) shows similarly high accuracies compared to the whole infrared. Leaf thickness, leaf water content, cellulose, lignin and stomata density are the traits that could be estimated most accurately from infrared data (with View the MathML source above 0.80 for the full range models). Leaf thickness, cellulose and lignin were predicted with reasonable accuracy from a combination of single infrared bands. Nevertheless, for all leaf traits, a combination of a few bands yields moderate to accurate estimations.

KW - Infrared spectra

KW - Leaf spectra

KW - Leaf traits

KW - Leaf water content

KW - Lignin

KW - Spectroscopy

KW - ITC-ISI-JOURNAL-ARTICLE

KW - ITC-HYBRID

UR - https://ezproxy2.utwente.nl/login?url=https://doi.org/10.1016/j.jag.2017.11.014

UR - https://ezproxy2.utwente.nl/login?url=https://webapps.itc.utwente.nl/library/2018/isi/groen_spe.pdf

UR - http://www.scopus.com/inward/record.url?scp=85051646900&partnerID=8YFLogxK

U2 - 10.1016/j.jag.2017.11.014

DO - 10.1016/j.jag.2017.11.014

M3 - Article

VL - 69

SP - 237

EP - 250

JO - International Journal of Applied Earth Observation and Geoinformation (JAG)

JF - International Journal of Applied Earth Observation and Geoinformation (JAG)

SN - 1569-8432

ER -