Spatially-explicit modelling with support of hyperspectral data can improve prediction of plant traits

Alby D. Rocha, T.A. Groen, A.K. Skidmore

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

Data from remote sensing with finer spectral and spatial resolution are increasingly available. While this allows more accurate prediction of plant traits at different spatial scales, it raises concerns about a lack of independence between observations. Hyperspectral wavelengths are serially correlated provoking multicollinearity among the predictors. As collection of ground reference points for validation remains time-consuming and difficult in many environments, empirical models are trained with a limited number of observations compared to the number of wavelengths. Moreover, any set of observations collected from a continuous surface is also likely to be spatially autocorrelated. Machine learning regression facilitates the task of selecting the most informative wavelengths, and then transforming them into latent variables to avoid the problem of multicollinearity. However, these regression methods do not solve the problem of spatial autocorrelation in the model residuals. In this study we show that, when significant spatial autocorrelation is observed, models that explicitly deal with spatial information and use a spectral index as a covariate exhibit a higher prediction accuracy than machine learning regressions do. However, for these models to work, the number of (hyperspectral) bands included in the models has to be drastically reduced and the model can not be directly extrapolated to a new (unobserved) location in another area. We conclude that quantifying spatial autocorrelation a-priori in the data can help in deciding whether the spatial and the spectral dimensions should be modelled together or not.
Original languageEnglish
Article number111200
Number of pages13
JournalRemote sensing of environment
Volume231
Early online date29 May 2019
DOIs
Publication statusPublished - 15 Sep 2019

Fingerprint

prediction
autocorrelation
Autocorrelation
modeling
wavelengths
artificial intelligence
wavelength
Wavelength
Learning systems
spectral resolution
remote sensing
Remote sensing
spatial resolution
machine learning
methodology

Keywords

  • INLA
  • Machine learning
  • Spatially explicit models
  • Data simulation
  • Radiative transfer models
  • Spatial autocorrelation
  • Plant traits
  • ITC-ISI-JOURNAL-ARTICLE

Cite this

@article{13cb65e23aa7429a96cddbcdf37c8bb8,
title = "Spatially-explicit modelling with support of hyperspectral data can improve prediction of plant traits",
abstract = "Data from remote sensing with finer spectral and spatial resolution are increasingly available. While this allows more accurate prediction of plant traits at different spatial scales, it raises concerns about a lack of independence between observations. Hyperspectral wavelengths are serially correlated provoking multicollinearity among the predictors. As collection of ground reference points for validation remains time-consuming and difficult in many environments, empirical models are trained with a limited number of observations compared to the number of wavelengths. Moreover, any set of observations collected from a continuous surface is also likely to be spatially autocorrelated. Machine learning regression facilitates the task of selecting the most informative wavelengths, and then transforming them into latent variables to avoid the problem of multicollinearity. However, these regression methods do not solve the problem of spatial autocorrelation in the model residuals. In this study we show that, when significant spatial autocorrelation is observed, models that explicitly deal with spatial information and use a spectral index as a covariate exhibit a higher prediction accuracy than machine learning regressions do. However, for these models to work, the number of (hyperspectral) bands included in the models has to be drastically reduced and the model can not be directly extrapolated to a new (unobserved) location in another area. We conclude that quantifying spatial autocorrelation a-priori in the data can help in deciding whether the spatial and the spectral dimensions should be modelled together or not.",
keywords = "INLA, Machine learning, Spatially explicit models, Data simulation, Radiative transfer models, Spatial autocorrelation, Plant traits, ITC-ISI-JOURNAL-ARTICLE",
author = "Rocha, {Alby D.} and T.A. Groen and A.K. Skidmore",
year = "2019",
month = "9",
day = "15",
doi = "10.1016/j.rse.2019.05.019",
language = "English",
volume = "231",
journal = "Remote sensing of environment",
issn = "0034-4257",
publisher = "Elsevier",

}

Spatially-explicit modelling with support of hyperspectral data can improve prediction of plant traits. / Rocha, Alby D.; Groen, T.A.; Skidmore, A.K.

In: Remote sensing of environment, Vol. 231, 111200, 15.09.2019.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Spatially-explicit modelling with support of hyperspectral data can improve prediction of plant traits

AU - Rocha, Alby D.

AU - Groen, T.A.

AU - Skidmore, A.K.

PY - 2019/9/15

Y1 - 2019/9/15

N2 - Data from remote sensing with finer spectral and spatial resolution are increasingly available. While this allows more accurate prediction of plant traits at different spatial scales, it raises concerns about a lack of independence between observations. Hyperspectral wavelengths are serially correlated provoking multicollinearity among the predictors. As collection of ground reference points for validation remains time-consuming and difficult in many environments, empirical models are trained with a limited number of observations compared to the number of wavelengths. Moreover, any set of observations collected from a continuous surface is also likely to be spatially autocorrelated. Machine learning regression facilitates the task of selecting the most informative wavelengths, and then transforming them into latent variables to avoid the problem of multicollinearity. However, these regression methods do not solve the problem of spatial autocorrelation in the model residuals. In this study we show that, when significant spatial autocorrelation is observed, models that explicitly deal with spatial information and use a spectral index as a covariate exhibit a higher prediction accuracy than machine learning regressions do. However, for these models to work, the number of (hyperspectral) bands included in the models has to be drastically reduced and the model can not be directly extrapolated to a new (unobserved) location in another area. We conclude that quantifying spatial autocorrelation a-priori in the data can help in deciding whether the spatial and the spectral dimensions should be modelled together or not.

AB - Data from remote sensing with finer spectral and spatial resolution are increasingly available. While this allows more accurate prediction of plant traits at different spatial scales, it raises concerns about a lack of independence between observations. Hyperspectral wavelengths are serially correlated provoking multicollinearity among the predictors. As collection of ground reference points for validation remains time-consuming and difficult in many environments, empirical models are trained with a limited number of observations compared to the number of wavelengths. Moreover, any set of observations collected from a continuous surface is also likely to be spatially autocorrelated. Machine learning regression facilitates the task of selecting the most informative wavelengths, and then transforming them into latent variables to avoid the problem of multicollinearity. However, these regression methods do not solve the problem of spatial autocorrelation in the model residuals. In this study we show that, when significant spatial autocorrelation is observed, models that explicitly deal with spatial information and use a spectral index as a covariate exhibit a higher prediction accuracy than machine learning regressions do. However, for these models to work, the number of (hyperspectral) bands included in the models has to be drastically reduced and the model can not be directly extrapolated to a new (unobserved) location in another area. We conclude that quantifying spatial autocorrelation a-priori in the data can help in deciding whether the spatial and the spectral dimensions should be modelled together or not.

KW - INLA

KW - Machine learning

KW - Spatially explicit models

KW - Data simulation

KW - Radiative transfer models

KW - Spatial autocorrelation

KW - Plant traits

KW - ITC-ISI-JOURNAL-ARTICLE

UR - https://ezproxy2.utwente.nl/login?url=https://doi.org/10.1016/j.rse.2019.05.019

UR - https://ezproxy2.utwente.nl/login?url=https://library.itc.utwente.nl/login/2019/isi/groen_spa.pdf

U2 - 10.1016/j.rse.2019.05.019

DO - 10.1016/j.rse.2019.05.019

M3 - Article

VL - 231

JO - Remote sensing of environment

JF - Remote sensing of environment

SN - 0034-4257

M1 - 111200

ER -