Retrieving vegetation canopy water content from hyperspectral thermal measurements

Research output: Contribution to journalArticleAcademicpeer-review

8 Citations (Scopus)
1 Downloads (Pure)

Abstract

The retrieval of vegetation canopy water content using thermal hyperspectral (TIR, 8–14 μm) measurements is investigated in this study. Vegetation water content indicators such as fuel moisture content (FMC, %, mass-based) and equivalent water thickness (EWT, g cm−2, area based) play significant roles in plant physiology, as well as in the modelling of fire risk and behavior, particularly in forests. Although retrieval of these parameters, in particular EWT, has been demonstrated from optical and TIR measurements, to our knowledge their prediction at canopy level in the thermal part of the electromagnetic spectrum has not yet been investigated. Therefore, the application of hyperspectral TIR data for predicting FMC and EWT parameters at canopy level is explored here. The emissivity of spectral data in the TIR region is measured for four species (Azalea japonica, Buxus sempervirens, Euonymus japonicus, and Ficus benjamina) under controlled laboratory conditions, using a portable MIDAC Fourier transform infrared spectrometer. EWT, FMC, and their corresponding canopy emissivity measurements are assessed by destructive sampling of the leaves. Leaf area, as well as fresh and dry mass of the harvested leaves, is determined for all four species. Partial least square regression and artificial neural networks, using various spectral subsets, are used to predict the two variables of interest. Higher estimation accuracies have been obtained for both FMC and EWT at canopy level using artificial neural networks. Unexpectedly, the FMC at canopy level, as a mass-based variable, more accurately retrieved using either method. This is contrary to previous findings using multispectral and hyperspectral data. Our results suggest that plant mass may play a greater role in determining spectral emissivity than plant area does.
Original languageEnglish
Pages (from-to)365-375
JournalAgricultural and forest meteorology
Volume247
DOIs
Publication statusPublished - 1 Dec 2017

Fingerprint

water content
canopy
heat
vegetation
emissivity
artificial neural network
neural networks
Euonymus japonicus
Buxus sempervirens
Ficus benjamina
Rhododendron
plant physiology
Fourier transform infrared spectroscopy
spectral analysis
leaf area
Fourier transform
least squares
leaves
moisture content
spectrometer

Keywords

  • ITC-ISI-JOURNAL-ARTICLE

Cite this

@article{ec58359ef38a4058a2483b2767db99e2,
title = "Retrieving vegetation canopy water content from hyperspectral thermal measurements",
abstract = "The retrieval of vegetation canopy water content using thermal hyperspectral (TIR, 8–14 μm) measurements is investigated in this study. Vegetation water content indicators such as fuel moisture content (FMC, {\%}, mass-based) and equivalent water thickness (EWT, g cm−2, area based) play significant roles in plant physiology, as well as in the modelling of fire risk and behavior, particularly in forests. Although retrieval of these parameters, in particular EWT, has been demonstrated from optical and TIR measurements, to our knowledge their prediction at canopy level in the thermal part of the electromagnetic spectrum has not yet been investigated. Therefore, the application of hyperspectral TIR data for predicting FMC and EWT parameters at canopy level is explored here. The emissivity of spectral data in the TIR region is measured for four species (Azalea japonica, Buxus sempervirens, Euonymus japonicus, and Ficus benjamina) under controlled laboratory conditions, using a portable MIDAC Fourier transform infrared spectrometer. EWT, FMC, and their corresponding canopy emissivity measurements are assessed by destructive sampling of the leaves. Leaf area, as well as fresh and dry mass of the harvested leaves, is determined for all four species. Partial least square regression and artificial neural networks, using various spectral subsets, are used to predict the two variables of interest. Higher estimation accuracies have been obtained for both FMC and EWT at canopy level using artificial neural networks. Unexpectedly, the FMC at canopy level, as a mass-based variable, more accurately retrieved using either method. This is contrary to previous findings using multispectral and hyperspectral data. Our results suggest that plant mass may play a greater role in determining spectral emissivity than plant area does.",
keywords = "ITC-ISI-JOURNAL-ARTICLE",
author = "Elnaz Neinavaz and Skidmore, {Andrew K.} and Roshanak Darvishzadeh and Groen, {Thomas A.}",
year = "2017",
month = "12",
day = "1",
doi = "10.1016/j.agrformet.2017.08.020",
language = "English",
volume = "247",
pages = "365--375",
journal = "Agricultural and forest meteorology",
issn = "0168-1923",
publisher = "Elsevier",

}

TY - JOUR

T1 - Retrieving vegetation canopy water content from hyperspectral thermal measurements

AU - Neinavaz, Elnaz

AU - Skidmore, Andrew K.

AU - Darvishzadeh, Roshanak

AU - Groen, Thomas A.

PY - 2017/12/1

Y1 - 2017/12/1

N2 - The retrieval of vegetation canopy water content using thermal hyperspectral (TIR, 8–14 μm) measurements is investigated in this study. Vegetation water content indicators such as fuel moisture content (FMC, %, mass-based) and equivalent water thickness (EWT, g cm−2, area based) play significant roles in plant physiology, as well as in the modelling of fire risk and behavior, particularly in forests. Although retrieval of these parameters, in particular EWT, has been demonstrated from optical and TIR measurements, to our knowledge their prediction at canopy level in the thermal part of the electromagnetic spectrum has not yet been investigated. Therefore, the application of hyperspectral TIR data for predicting FMC and EWT parameters at canopy level is explored here. The emissivity of spectral data in the TIR region is measured for four species (Azalea japonica, Buxus sempervirens, Euonymus japonicus, and Ficus benjamina) under controlled laboratory conditions, using a portable MIDAC Fourier transform infrared spectrometer. EWT, FMC, and their corresponding canopy emissivity measurements are assessed by destructive sampling of the leaves. Leaf area, as well as fresh and dry mass of the harvested leaves, is determined for all four species. Partial least square regression and artificial neural networks, using various spectral subsets, are used to predict the two variables of interest. Higher estimation accuracies have been obtained for both FMC and EWT at canopy level using artificial neural networks. Unexpectedly, the FMC at canopy level, as a mass-based variable, more accurately retrieved using either method. This is contrary to previous findings using multispectral and hyperspectral data. Our results suggest that plant mass may play a greater role in determining spectral emissivity than plant area does.

AB - The retrieval of vegetation canopy water content using thermal hyperspectral (TIR, 8–14 μm) measurements is investigated in this study. Vegetation water content indicators such as fuel moisture content (FMC, %, mass-based) and equivalent water thickness (EWT, g cm−2, area based) play significant roles in plant physiology, as well as in the modelling of fire risk and behavior, particularly in forests. Although retrieval of these parameters, in particular EWT, has been demonstrated from optical and TIR measurements, to our knowledge their prediction at canopy level in the thermal part of the electromagnetic spectrum has not yet been investigated. Therefore, the application of hyperspectral TIR data for predicting FMC and EWT parameters at canopy level is explored here. The emissivity of spectral data in the TIR region is measured for four species (Azalea japonica, Buxus sempervirens, Euonymus japonicus, and Ficus benjamina) under controlled laboratory conditions, using a portable MIDAC Fourier transform infrared spectrometer. EWT, FMC, and their corresponding canopy emissivity measurements are assessed by destructive sampling of the leaves. Leaf area, as well as fresh and dry mass of the harvested leaves, is determined for all four species. Partial least square regression and artificial neural networks, using various spectral subsets, are used to predict the two variables of interest. Higher estimation accuracies have been obtained for both FMC and EWT at canopy level using artificial neural networks. Unexpectedly, the FMC at canopy level, as a mass-based variable, more accurately retrieved using either method. This is contrary to previous findings using multispectral and hyperspectral data. Our results suggest that plant mass may play a greater role in determining spectral emissivity than plant area does.

KW - ITC-ISI-JOURNAL-ARTICLE

UR - https://ezproxy2.utwente.nl/login?url=https://webapps.itc.utwente.nl/library/2017/isi/neinavaz_ret.pdf

U2 - 10.1016/j.agrformet.2017.08.020

DO - 10.1016/j.agrformet.2017.08.020

M3 - Article

VL - 247

SP - 365

EP - 375

JO - Agricultural and forest meteorology

JF - Agricultural and forest meteorology

SN - 0168-1923

ER -