Effect of spatial heterogeneity on the validation of remote sensing based GPP estimations

Gy Gelybó, Z. Barcza, A. Kern, N. Kljun

Research output: Contribution to journalArticle

17 Citations (Scopus)

Abstract

Satellite based remote sensing provides an efficient way to estimate carbon balance components over large spatial domains with acceptable temporal and spatial resolution. In the present study remote sensing based gross primary production (GPP) estimations were evaluated using data from a tall eddy-covariance flux tower, located over a heterogeneous agricultural landscape in Hungary. Four different methods were used to simulate 8-day mean GPP for the tower site based on the MOD17 light use efficiency model. Additionally, we present a novel approach for model validation to exploit the advantage of footprint size similarity between remote sensing and the hourly eddy covariance signal measured at the tall tower. Besides using footprint information for the model validation we performed downscaling of MOD17 using 250m resolution MODerate resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) dataset in order to address land use heterogeneity. The results showed that GPP was underestimated by MOD17 especially in years with average precipitation during the growing season, while model performance was better during dry years. Our downscaling technique significantly improved agreement between the MOD17 model results and the eddy covariance measurements (modeling efficiency (ME) increased from 0.783 to 0.884, root mean square error (RMSE) decreased from 1.095g Cm-2day-1 to 0.815g Cm-2day-1), although GPP remained underestimated (bias decreased from -0.680gC m-2day-1 to -0.426g Cm-2day-1). Model evaluation results suggest that model performance should be optimally evaluated using RMSE, index of agreement (IA), ME and bias. The presented method is applicable to any eddy-covariance tower with limitations depending of the complexity of landscape around the flux tower. As incorporation of footprint information clearly impacts validation results, future model validation and/or calibration should also involve source area estimation which can be easily implemented following the presented approach.

Original languageEnglish
Pages (from-to)43-53
Number of pages11
JournalAgricultural and Forest Meteorology
Volume174-175
DOIs
Publication statusPublished - Jun 15 2013

Fingerprint

remote sensing
primary production
eddy covariance
model validation
footprint
downscaling
light use efficiency
carbon balance
NDVI
MODIS
modeling
spatial resolution
moderate resolution imaging spectroradiometer
growing season
agricultural land
effect
Hungary
calibration
land use
methodology

Keywords

  • Cropland
  • Eddy covariance
  • Footprint
  • Light use efficiency model
  • MODIS

ASJC Scopus subject areas

  • Agronomy and Crop Science
  • Forestry
  • Atmospheric Science
  • Global and Planetary Change

Cite this

Effect of spatial heterogeneity on the validation of remote sensing based GPP estimations. / Gelybó, Gy; Barcza, Z.; Kern, A.; Kljun, N.

In: Agricultural and Forest Meteorology, Vol. 174-175, 15.06.2013, p. 43-53.

Research output: Contribution to journalArticle

@article{48d1cf6edb83417a90ec5925e34c335e,
title = "Effect of spatial heterogeneity on the validation of remote sensing based GPP estimations",
abstract = "Satellite based remote sensing provides an efficient way to estimate carbon balance components over large spatial domains with acceptable temporal and spatial resolution. In the present study remote sensing based gross primary production (GPP) estimations were evaluated using data from a tall eddy-covariance flux tower, located over a heterogeneous agricultural landscape in Hungary. Four different methods were used to simulate 8-day mean GPP for the tower site based on the MOD17 light use efficiency model. Additionally, we present a novel approach for model validation to exploit the advantage of footprint size similarity between remote sensing and the hourly eddy covariance signal measured at the tall tower. Besides using footprint information for the model validation we performed downscaling of MOD17 using 250m resolution MODerate resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) dataset in order to address land use heterogeneity. The results showed that GPP was underestimated by MOD17 especially in years with average precipitation during the growing season, while model performance was better during dry years. Our downscaling technique significantly improved agreement between the MOD17 model results and the eddy covariance measurements (modeling efficiency (ME) increased from 0.783 to 0.884, root mean square error (RMSE) decreased from 1.095g Cm-2day-1 to 0.815g Cm-2day-1), although GPP remained underestimated (bias decreased from -0.680gC m-2day-1 to -0.426g Cm-2day-1). Model evaluation results suggest that model performance should be optimally evaluated using RMSE, index of agreement (IA), ME and bias. The presented method is applicable to any eddy-covariance tower with limitations depending of the complexity of landscape around the flux tower. As incorporation of footprint information clearly impacts validation results, future model validation and/or calibration should also involve source area estimation which can be easily implemented following the presented approach.",
keywords = "Cropland, Eddy covariance, Footprint, Light use efficiency model, MODIS",
author = "Gy Gelyb{\'o} and Z. Barcza and A. Kern and N. Kljun",
year = "2013",
month = "6",
day = "15",
doi = "10.1016/j.agrformet.2013.02.003",
language = "English",
volume = "174-175",
pages = "43--53",
journal = "Agricultural and Forest Meteorology",
issn = "0168-1923",
publisher = "Elsevier",

}

TY - JOUR

T1 - Effect of spatial heterogeneity on the validation of remote sensing based GPP estimations

AU - Gelybó, Gy

AU - Barcza, Z.

AU - Kern, A.

AU - Kljun, N.

PY - 2013/6/15

Y1 - 2013/6/15

N2 - Satellite based remote sensing provides an efficient way to estimate carbon balance components over large spatial domains with acceptable temporal and spatial resolution. In the present study remote sensing based gross primary production (GPP) estimations were evaluated using data from a tall eddy-covariance flux tower, located over a heterogeneous agricultural landscape in Hungary. Four different methods were used to simulate 8-day mean GPP for the tower site based on the MOD17 light use efficiency model. Additionally, we present a novel approach for model validation to exploit the advantage of footprint size similarity between remote sensing and the hourly eddy covariance signal measured at the tall tower. Besides using footprint information for the model validation we performed downscaling of MOD17 using 250m resolution MODerate resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) dataset in order to address land use heterogeneity. The results showed that GPP was underestimated by MOD17 especially in years with average precipitation during the growing season, while model performance was better during dry years. Our downscaling technique significantly improved agreement between the MOD17 model results and the eddy covariance measurements (modeling efficiency (ME) increased from 0.783 to 0.884, root mean square error (RMSE) decreased from 1.095g Cm-2day-1 to 0.815g Cm-2day-1), although GPP remained underestimated (bias decreased from -0.680gC m-2day-1 to -0.426g Cm-2day-1). Model evaluation results suggest that model performance should be optimally evaluated using RMSE, index of agreement (IA), ME and bias. The presented method is applicable to any eddy-covariance tower with limitations depending of the complexity of landscape around the flux tower. As incorporation of footprint information clearly impacts validation results, future model validation and/or calibration should also involve source area estimation which can be easily implemented following the presented approach.

AB - Satellite based remote sensing provides an efficient way to estimate carbon balance components over large spatial domains with acceptable temporal and spatial resolution. In the present study remote sensing based gross primary production (GPP) estimations were evaluated using data from a tall eddy-covariance flux tower, located over a heterogeneous agricultural landscape in Hungary. Four different methods were used to simulate 8-day mean GPP for the tower site based on the MOD17 light use efficiency model. Additionally, we present a novel approach for model validation to exploit the advantage of footprint size similarity between remote sensing and the hourly eddy covariance signal measured at the tall tower. Besides using footprint information for the model validation we performed downscaling of MOD17 using 250m resolution MODerate resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) dataset in order to address land use heterogeneity. The results showed that GPP was underestimated by MOD17 especially in years with average precipitation during the growing season, while model performance was better during dry years. Our downscaling technique significantly improved agreement between the MOD17 model results and the eddy covariance measurements (modeling efficiency (ME) increased from 0.783 to 0.884, root mean square error (RMSE) decreased from 1.095g Cm-2day-1 to 0.815g Cm-2day-1), although GPP remained underestimated (bias decreased from -0.680gC m-2day-1 to -0.426g Cm-2day-1). Model evaluation results suggest that model performance should be optimally evaluated using RMSE, index of agreement (IA), ME and bias. The presented method is applicable to any eddy-covariance tower with limitations depending of the complexity of landscape around the flux tower. As incorporation of footprint information clearly impacts validation results, future model validation and/or calibration should also involve source area estimation which can be easily implemented following the presented approach.

KW - Cropland

KW - Eddy covariance

KW - Footprint

KW - Light use efficiency model

KW - MODIS

UR - http://www.scopus.com/inward/record.url?scp=84876344433&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84876344433&partnerID=8YFLogxK

U2 - 10.1016/j.agrformet.2013.02.003

DO - 10.1016/j.agrformet.2013.02.003

M3 - Article

AN - SCOPUS:84876344433

VL - 174-175

SP - 43

EP - 53

JO - Agricultural and Forest Meteorology

JF - Agricultural and Forest Meteorology

SN - 0168-1923

ER -