A voxel matching method for effective leaf area index estimation in temperate deciduous forests from leaf-on and leaf-off airborne LiDAR data

Xi Zhu*, Jing Liu, A.K. Skidmore, Joe Premier, Marco Heurich

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

5 Downloads (Pure)

Abstract

The quantification of leaf area index (LAI) is essential for modeling the interaction between atmosphere and biosphere. The airborne LiDAR has emerged as an effective tool for mapping plant area index (PAI) in a landscape consisting of both woody and leaf materials. However, the discrimination between woody and leaf materials and the estimation of effective LAI (eLAI) have, to date, rarely been studied at landscape scale. We applied a voxel matching algorithm to estimate eLAI of deciduous forests using simulated and field LiDAR data under leaf-on and leaf-off conditions. We classified LiDAR points as either a leaf or a woody hit on leaf-on LiDAR data by matching the point with leaf-off data. We compared the eLAI result of our voxel matching algorithm against the subtraction method, where the leaf-off effective woody area index (eWAI) is subtracted from the effective leaf-on PAI (ePAI). Our results, which were validated against terrestrial LiDAR derived eLAI, showed that the voxel matching method, with an optimal voxel size of 0.1 m, produced an unbiased estimation of terrestrial LiDAR derived eLAI with an R2 of 0.70 and an RMSE of 0.41 (RRMSE: 20.1%). The subtraction method, however, yielded an R2 of 0.62 and an RMSE of 1.02 (RRMSE: 50.1%) with a significant underestimation of 0.94. Reassuringly, the same outcome was observed using a simulated dataset. In addition, we evaluated the performance of 96 LiDAR metrics under leaf-on conditions for eLAI prediction using a statistical model. Based on the importance scores derived from the random forest regression, nine of the 96 leaf-on LiDAR metrics were selected. Cross-validation showed that eLAI could be predicted using these metrics under leaf-on conditions with an R2 of 0.73 and an RMSE of 0.27 (RRMSE: 17.4%). The voxel matching method yielded a slightly lower accuracy (R2: 0.70, RMSE:0.41, RRMSE: 20.1%) than the statistical model. We, therefore, suggest that the voxel matching method offers a new opportunity for the estimating eLAI and other ecological applications that require the classification between leaf and woody materials using airborne LiDAR data. It potentially allows transferability to different sites and flight campaigns.
Original languageEnglish
Article number111696
Pages (from-to)1-14
Number of pages14
JournalRemote sensing of environment
Volume240
Early online date6 Feb 2020
DOIs
Publication statusE-pub ahead of print/First online - 6 Feb 2020

Fingerprint

temperate forests
temperate forest
deciduous forests
deciduous forest
leaf area index
leaves
methodology
statistical models
method
Statistical Models
biosphere
flight
taxonomy

Keywords

  • Effective leaf area index
  • Airborne LiDAR
  • Leaf-off
  • Leaf-on
  • Voxel matching
  • ITC-ISI-JOURNAL-ARTICLE
  • ITC-HYBRID
  • UT-Hybrid-D

Cite this

@article{6f7a7f4e7e3644d1a77c3d838bf015c1,
title = "A voxel matching method for effective leaf area index estimation in temperate deciduous forests from leaf-on and leaf-off airborne LiDAR data",
abstract = "The quantification of leaf area index (LAI) is essential for modeling the interaction between atmosphere and biosphere. The airborne LiDAR has emerged as an effective tool for mapping plant area index (PAI) in a landscape consisting of both woody and leaf materials. However, the discrimination between woody and leaf materials and the estimation of effective LAI (eLAI) have, to date, rarely been studied at landscape scale. We applied a voxel matching algorithm to estimate eLAI of deciduous forests using simulated and field LiDAR data under leaf-on and leaf-off conditions. We classified LiDAR points as either a leaf or a woody hit on leaf-on LiDAR data by matching the point with leaf-off data. We compared the eLAI result of our voxel matching algorithm against the subtraction method, where the leaf-off effective woody area index (eWAI) is subtracted from the effective leaf-on PAI (ePAI). Our results, which were validated against terrestrial LiDAR derived eLAI, showed that the voxel matching method, with an optimal voxel size of 0.1 m, produced an unbiased estimation of terrestrial LiDAR derived eLAI with an R2 of 0.70 and an RMSE of 0.41 (RRMSE: 20.1{\%}). The subtraction method, however, yielded an R2 of 0.62 and an RMSE of 1.02 (RRMSE: 50.1{\%}) with a significant underestimation of 0.94. Reassuringly, the same outcome was observed using a simulated dataset. In addition, we evaluated the performance of 96 LiDAR metrics under leaf-on conditions for eLAI prediction using a statistical model. Based on the importance scores derived from the random forest regression, nine of the 96 leaf-on LiDAR metrics were selected. Cross-validation showed that eLAI could be predicted using these metrics under leaf-on conditions with an R2 of 0.73 and an RMSE of 0.27 (RRMSE: 17.4{\%}). The voxel matching method yielded a slightly lower accuracy (R2: 0.70, RMSE:0.41, RRMSE: 20.1{\%}) than the statistical model. We, therefore, suggest that the voxel matching method offers a new opportunity for the estimating eLAI and other ecological applications that require the classification between leaf and woody materials using airborne LiDAR data. It potentially allows transferability to different sites and flight campaigns.",
keywords = "Effective leaf area index, Airborne LiDAR, Leaf-off, Leaf-on, Voxel matching, ITC-ISI-JOURNAL-ARTICLE, ITC-HYBRID, UT-Hybrid-D",
author = "Xi Zhu and Jing Liu and A.K. Skidmore and Joe Premier and Marco Heurich",
year = "2020",
month = "2",
day = "6",
doi = "10.1016/j.rse.2020.111696",
language = "English",
volume = "240",
pages = "1--14",
journal = "Remote sensing of environment",
issn = "0034-4257",
publisher = "Elsevier",

}

A voxel matching method for effective leaf area index estimation in temperate deciduous forests from leaf-on and leaf-off airborne LiDAR data. / Zhu, Xi; Liu, Jing; Skidmore, A.K.; Premier, Joe; Heurich, Marco.

In: Remote sensing of environment, Vol. 240, 111696, 04.2020, p. 1-14.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - A voxel matching method for effective leaf area index estimation in temperate deciduous forests from leaf-on and leaf-off airborne LiDAR data

AU - Zhu, Xi

AU - Liu, Jing

AU - Skidmore, A.K.

AU - Premier, Joe

AU - Heurich, Marco

PY - 2020/2/6

Y1 - 2020/2/6

N2 - The quantification of leaf area index (LAI) is essential for modeling the interaction between atmosphere and biosphere. The airborne LiDAR has emerged as an effective tool for mapping plant area index (PAI) in a landscape consisting of both woody and leaf materials. However, the discrimination between woody and leaf materials and the estimation of effective LAI (eLAI) have, to date, rarely been studied at landscape scale. We applied a voxel matching algorithm to estimate eLAI of deciduous forests using simulated and field LiDAR data under leaf-on and leaf-off conditions. We classified LiDAR points as either a leaf or a woody hit on leaf-on LiDAR data by matching the point with leaf-off data. We compared the eLAI result of our voxel matching algorithm against the subtraction method, where the leaf-off effective woody area index (eWAI) is subtracted from the effective leaf-on PAI (ePAI). Our results, which were validated against terrestrial LiDAR derived eLAI, showed that the voxel matching method, with an optimal voxel size of 0.1 m, produced an unbiased estimation of terrestrial LiDAR derived eLAI with an R2 of 0.70 and an RMSE of 0.41 (RRMSE: 20.1%). The subtraction method, however, yielded an R2 of 0.62 and an RMSE of 1.02 (RRMSE: 50.1%) with a significant underestimation of 0.94. Reassuringly, the same outcome was observed using a simulated dataset. In addition, we evaluated the performance of 96 LiDAR metrics under leaf-on conditions for eLAI prediction using a statistical model. Based on the importance scores derived from the random forest regression, nine of the 96 leaf-on LiDAR metrics were selected. Cross-validation showed that eLAI could be predicted using these metrics under leaf-on conditions with an R2 of 0.73 and an RMSE of 0.27 (RRMSE: 17.4%). The voxel matching method yielded a slightly lower accuracy (R2: 0.70, RMSE:0.41, RRMSE: 20.1%) than the statistical model. We, therefore, suggest that the voxel matching method offers a new opportunity for the estimating eLAI and other ecological applications that require the classification between leaf and woody materials using airborne LiDAR data. It potentially allows transferability to different sites and flight campaigns.

AB - The quantification of leaf area index (LAI) is essential for modeling the interaction between atmosphere and biosphere. The airborne LiDAR has emerged as an effective tool for mapping plant area index (PAI) in a landscape consisting of both woody and leaf materials. However, the discrimination between woody and leaf materials and the estimation of effective LAI (eLAI) have, to date, rarely been studied at landscape scale. We applied a voxel matching algorithm to estimate eLAI of deciduous forests using simulated and field LiDAR data under leaf-on and leaf-off conditions. We classified LiDAR points as either a leaf or a woody hit on leaf-on LiDAR data by matching the point with leaf-off data. We compared the eLAI result of our voxel matching algorithm against the subtraction method, where the leaf-off effective woody area index (eWAI) is subtracted from the effective leaf-on PAI (ePAI). Our results, which were validated against terrestrial LiDAR derived eLAI, showed that the voxel matching method, with an optimal voxel size of 0.1 m, produced an unbiased estimation of terrestrial LiDAR derived eLAI with an R2 of 0.70 and an RMSE of 0.41 (RRMSE: 20.1%). The subtraction method, however, yielded an R2 of 0.62 and an RMSE of 1.02 (RRMSE: 50.1%) with a significant underestimation of 0.94. Reassuringly, the same outcome was observed using a simulated dataset. In addition, we evaluated the performance of 96 LiDAR metrics under leaf-on conditions for eLAI prediction using a statistical model. Based on the importance scores derived from the random forest regression, nine of the 96 leaf-on LiDAR metrics were selected. Cross-validation showed that eLAI could be predicted using these metrics under leaf-on conditions with an R2 of 0.73 and an RMSE of 0.27 (RRMSE: 17.4%). The voxel matching method yielded a slightly lower accuracy (R2: 0.70, RMSE:0.41, RRMSE: 20.1%) than the statistical model. We, therefore, suggest that the voxel matching method offers a new opportunity for the estimating eLAI and other ecological applications that require the classification between leaf and woody materials using airborne LiDAR data. It potentially allows transferability to different sites and flight campaigns.

KW - Effective leaf area index

KW - Airborne LiDAR

KW - Leaf-off

KW - Leaf-on

KW - Voxel matching

KW - ITC-ISI-JOURNAL-ARTICLE

KW - ITC-HYBRID

KW - UT-Hybrid-D

UR - https://ezproxy2.utwente.nl/login?url=https://library.itc.utwente.nl/login/2020/isi/zhu_vox.pdf

U2 - 10.1016/j.rse.2020.111696

DO - 10.1016/j.rse.2020.111696

M3 - Article

VL - 240

SP - 1

EP - 14

JO - Remote sensing of environment

JF - Remote sensing of environment

SN - 0034-4257

M1 - 111696

ER -