A methodology to derive global maps of leaf traits using remote sensing and climate data

Álvaro Moreno-Martínez*, Gustau Camps-Valls, Jens Kattge, Nathaniel Robinson, Markus Reichstein, Peter van Bodegom, Koen Kramer, J.H.C. Cornelissen, Peter Reich, Michael Bahn, Ülo Niinemets, Josep Peñuelas, Joseph M. Craine, Bruno E.L. Cerabolini, Vanessa Minden, Daniel C. Laughlin, Lawren Sack, Brady Allred, Christopher Baraloto, Chaeho Byun & 2 others Nadejda A. Soudzilovskaia, Steve W. Running

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

10 Citations (Scopus)

Abstract

This paper introduces a modular processing chain to derive global high-resolution maps of leaf traits. In particular, we present global maps at 500 m resolution of specific leaf area, leaf dry matter content, leaf nitrogen and phosphorus content per dry mass, and leaf nitrogen/phosphorus ratio. The processing chain exploits machine learning techniques along with optical remote sensing data (MODIS/Landsat) and climate data for gap filling and up-scaling of in-situ measured leaf traits. The chain first uses random forests regression with surrogates to fill gaps in the database (> 45% of missing entries) and maximizes the global representativeness of the trait dataset. Plant species are then aggregated to Plant Functional Types (PFTs). Next, the spatial abundance of PFTs at MODIS resolution (500 m) is calculated using Landsat data (30 m). Based on these PFT abundances, representative trait values are calculated for MODIS pixels with nearby trait data. Finally, different regression algorithms are applied to globally predict trait estimates from these MODIS pixels using remote sensing and climate data. The methods were compared in terms of precision, robustness and efficiency. The best model (random forests regression) shows good precision (normalized RMSE≤ 20%) and goodness of fit (averaged Pearson's correlation R = 0.78) in any considered trait. Along with the estimated global maps of leaf traits, we provide associated uncertainty estimates derived from the regression models. The process chain is modular, and can easily accommodate new traits, data streams (traits databases and remote sensing data), and methods. The machine learning techniques applied allow attribution of information gain to data input and thus provide the opportunity to understand trait-environment relationships at the plant and ecosystem scales. The new data products – the gap-filled trait matrix, a global map of PFT abundance per MODIS gridcells and the high-resolution global leaf trait maps – are complementary to existing large-scale observations of the land surface and we therefore anticipate substantial contributions to advances in quantifying, understanding and prediction of the Earth system.

Original languageEnglish
Pages (from-to)69-88
JournalRemote Sensing of Environment
Volume218
DOIs
Publication statusPublished - Dec 2018

Fingerprint

remote sensing
Remote sensing
moderate resolution imaging spectroradiometer
climate
methodology
MODIS
leaves
Learning systems
Phosphorus
artificial intelligence
Landsat
Pixels
Nitrogen
Processing
Ecosystems
phosphorus
pixel
nitrogen
Earth (planet)
dry matter content

Keywords

  • Climate
  • Landsat
  • Machine learning
  • MODIS
  • Plant ecology
  • Plant traits
  • Random forests
  • Remote sensing

Cite this

Moreno-Martínez, Á., Camps-Valls, G., Kattge, J., Robinson, N., Reichstein, M., van Bodegom, P., ... Running, S. W. (2018). A methodology to derive global maps of leaf traits using remote sensing and climate data. Remote Sensing of Environment, 218, 69-88. https://doi.org/10.1016/j.rse.2018.09.006
Moreno-Martínez, Álvaro ; Camps-Valls, Gustau ; Kattge, Jens ; Robinson, Nathaniel ; Reichstein, Markus ; van Bodegom, Peter ; Kramer, Koen ; Cornelissen, J.H.C. ; Reich, Peter ; Bahn, Michael ; Niinemets, Ülo ; Peñuelas, Josep ; Craine, Joseph M. ; Cerabolini, Bruno E.L. ; Minden, Vanessa ; Laughlin, Daniel C. ; Sack, Lawren ; Allred, Brady ; Baraloto, Christopher ; Byun, Chaeho ; Soudzilovskaia, Nadejda A. ; Running, Steve W. / A methodology to derive global maps of leaf traits using remote sensing and climate data. In: Remote Sensing of Environment. 2018 ; Vol. 218. pp. 69-88.
@article{80f8bb8307574f039395b79d8a3916f1,
title = "A methodology to derive global maps of leaf traits using remote sensing and climate data",
abstract = "This paper introduces a modular processing chain to derive global high-resolution maps of leaf traits. In particular, we present global maps at 500 m resolution of specific leaf area, leaf dry matter content, leaf nitrogen and phosphorus content per dry mass, and leaf nitrogen/phosphorus ratio. The processing chain exploits machine learning techniques along with optical remote sensing data (MODIS/Landsat) and climate data for gap filling and up-scaling of in-situ measured leaf traits. The chain first uses random forests regression with surrogates to fill gaps in the database (> 45{\%} of missing entries) and maximizes the global representativeness of the trait dataset. Plant species are then aggregated to Plant Functional Types (PFTs). Next, the spatial abundance of PFTs at MODIS resolution (500 m) is calculated using Landsat data (30 m). Based on these PFT abundances, representative trait values are calculated for MODIS pixels with nearby trait data. Finally, different regression algorithms are applied to globally predict trait estimates from these MODIS pixels using remote sensing and climate data. The methods were compared in terms of precision, robustness and efficiency. The best model (random forests regression) shows good precision (normalized RMSE≤ 20{\%}) and goodness of fit (averaged Pearson's correlation R = 0.78) in any considered trait. Along with the estimated global maps of leaf traits, we provide associated uncertainty estimates derived from the regression models. The process chain is modular, and can easily accommodate new traits, data streams (traits databases and remote sensing data), and methods. The machine learning techniques applied allow attribution of information gain to data input and thus provide the opportunity to understand trait-environment relationships at the plant and ecosystem scales. The new data products – the gap-filled trait matrix, a global map of PFT abundance per MODIS gridcells and the high-resolution global leaf trait maps – are complementary to existing large-scale observations of the land surface and we therefore anticipate substantial contributions to advances in quantifying, understanding and prediction of the Earth system.",
keywords = "Climate, Landsat, Machine learning, MODIS, Plant ecology, Plant traits, Random forests, Remote sensing",
author = "{\'A}lvaro Moreno-Mart{\'i}nez and Gustau Camps-Valls and Jens Kattge and Nathaniel Robinson and Markus Reichstein and {van Bodegom}, Peter and Koen Kramer and J.H.C. Cornelissen and Peter Reich and Michael Bahn and {\"U}lo Niinemets and Josep Pe{\~n}uelas and Craine, {Joseph M.} and Cerabolini, {Bruno E.L.} and Vanessa Minden and Laughlin, {Daniel C.} and Lawren Sack and Brady Allred and Christopher Baraloto and Chaeho Byun and Soudzilovskaia, {Nadejda A.} and Running, {Steve W.}",
year = "2018",
month = "12",
doi = "10.1016/j.rse.2018.09.006",
language = "English",
volume = "218",
pages = "69--88",
journal = "Remote Sensing of Environment",
issn = "0034-4257",
publisher = "Elsevier",

}

Moreno-Martínez, Á, Camps-Valls, G, Kattge, J, Robinson, N, Reichstein, M, van Bodegom, P, Kramer, K, Cornelissen, JHC, Reich, P, Bahn, M, Niinemets, Ü, Peñuelas, J, Craine, JM, Cerabolini, BEL, Minden, V, Laughlin, DC, Sack, L, Allred, B, Baraloto, C, Byun, C, Soudzilovskaia, NA & Running, SW 2018, 'A methodology to derive global maps of leaf traits using remote sensing and climate data', Remote Sensing of Environment, vol. 218, pp. 69-88. https://doi.org/10.1016/j.rse.2018.09.006

A methodology to derive global maps of leaf traits using remote sensing and climate data. / Moreno-Martínez, Álvaro; Camps-Valls, Gustau; Kattge, Jens; Robinson, Nathaniel; Reichstein, Markus; van Bodegom, Peter; Kramer, Koen; Cornelissen, J.H.C.; Reich, Peter; Bahn, Michael; Niinemets, Ülo; Peñuelas, Josep; Craine, Joseph M.; Cerabolini, Bruno E.L.; Minden, Vanessa; Laughlin, Daniel C.; Sack, Lawren; Allred, Brady; Baraloto, Christopher; Byun, Chaeho; Soudzilovskaia, Nadejda A.; Running, Steve W.

In: Remote Sensing of Environment, Vol. 218, 12.2018, p. 69-88.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - A methodology to derive global maps of leaf traits using remote sensing and climate data

AU - Moreno-Martínez, Álvaro

AU - Camps-Valls, Gustau

AU - Kattge, Jens

AU - Robinson, Nathaniel

AU - Reichstein, Markus

AU - van Bodegom, Peter

AU - Kramer, Koen

AU - Cornelissen, J.H.C.

AU - Reich, Peter

AU - Bahn, Michael

AU - Niinemets, Ülo

AU - Peñuelas, Josep

AU - Craine, Joseph M.

AU - Cerabolini, Bruno E.L.

AU - Minden, Vanessa

AU - Laughlin, Daniel C.

AU - Sack, Lawren

AU - Allred, Brady

AU - Baraloto, Christopher

AU - Byun, Chaeho

AU - Soudzilovskaia, Nadejda A.

AU - Running, Steve W.

PY - 2018/12

Y1 - 2018/12

N2 - This paper introduces a modular processing chain to derive global high-resolution maps of leaf traits. In particular, we present global maps at 500 m resolution of specific leaf area, leaf dry matter content, leaf nitrogen and phosphorus content per dry mass, and leaf nitrogen/phosphorus ratio. The processing chain exploits machine learning techniques along with optical remote sensing data (MODIS/Landsat) and climate data for gap filling and up-scaling of in-situ measured leaf traits. The chain first uses random forests regression with surrogates to fill gaps in the database (> 45% of missing entries) and maximizes the global representativeness of the trait dataset. Plant species are then aggregated to Plant Functional Types (PFTs). Next, the spatial abundance of PFTs at MODIS resolution (500 m) is calculated using Landsat data (30 m). Based on these PFT abundances, representative trait values are calculated for MODIS pixels with nearby trait data. Finally, different regression algorithms are applied to globally predict trait estimates from these MODIS pixels using remote sensing and climate data. The methods were compared in terms of precision, robustness and efficiency. The best model (random forests regression) shows good precision (normalized RMSE≤ 20%) and goodness of fit (averaged Pearson's correlation R = 0.78) in any considered trait. Along with the estimated global maps of leaf traits, we provide associated uncertainty estimates derived from the regression models. The process chain is modular, and can easily accommodate new traits, data streams (traits databases and remote sensing data), and methods. The machine learning techniques applied allow attribution of information gain to data input and thus provide the opportunity to understand trait-environment relationships at the plant and ecosystem scales. The new data products – the gap-filled trait matrix, a global map of PFT abundance per MODIS gridcells and the high-resolution global leaf trait maps – are complementary to existing large-scale observations of the land surface and we therefore anticipate substantial contributions to advances in quantifying, understanding and prediction of the Earth system.

AB - This paper introduces a modular processing chain to derive global high-resolution maps of leaf traits. In particular, we present global maps at 500 m resolution of specific leaf area, leaf dry matter content, leaf nitrogen and phosphorus content per dry mass, and leaf nitrogen/phosphorus ratio. The processing chain exploits machine learning techniques along with optical remote sensing data (MODIS/Landsat) and climate data for gap filling and up-scaling of in-situ measured leaf traits. The chain first uses random forests regression with surrogates to fill gaps in the database (> 45% of missing entries) and maximizes the global representativeness of the trait dataset. Plant species are then aggregated to Plant Functional Types (PFTs). Next, the spatial abundance of PFTs at MODIS resolution (500 m) is calculated using Landsat data (30 m). Based on these PFT abundances, representative trait values are calculated for MODIS pixels with nearby trait data. Finally, different regression algorithms are applied to globally predict trait estimates from these MODIS pixels using remote sensing and climate data. The methods were compared in terms of precision, robustness and efficiency. The best model (random forests regression) shows good precision (normalized RMSE≤ 20%) and goodness of fit (averaged Pearson's correlation R = 0.78) in any considered trait. Along with the estimated global maps of leaf traits, we provide associated uncertainty estimates derived from the regression models. The process chain is modular, and can easily accommodate new traits, data streams (traits databases and remote sensing data), and methods. The machine learning techniques applied allow attribution of information gain to data input and thus provide the opportunity to understand trait-environment relationships at the plant and ecosystem scales. The new data products – the gap-filled trait matrix, a global map of PFT abundance per MODIS gridcells and the high-resolution global leaf trait maps – are complementary to existing large-scale observations of the land surface and we therefore anticipate substantial contributions to advances in quantifying, understanding and prediction of the Earth system.

KW - Climate

KW - Landsat

KW - Machine learning

KW - MODIS

KW - Plant ecology

KW - Plant traits

KW - Random forests

KW - Remote sensing

U2 - 10.1016/j.rse.2018.09.006

DO - 10.1016/j.rse.2018.09.006

M3 - Article

VL - 218

SP - 69

EP - 88

JO - Remote Sensing of Environment

JF - Remote Sensing of Environment

SN - 0034-4257

ER -

Moreno-Martínez Á, Camps-Valls G, Kattge J, Robinson N, Reichstein M, van Bodegom P et al. A methodology to derive global maps of leaf traits using remote sensing and climate data. Remote Sensing of Environment. 2018 Dec;218:69-88. https://doi.org/10.1016/j.rse.2018.09.006