A phenology-based approach to the classification of Arctic tundra ecosystems in Greenland

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningfagfællebedømt

Standard

A phenology-based approach to the classification of Arctic tundra ecosystems in Greenland. / Karami, Mojtaba; Westergaard-Nielsen, Andreas; Normand, Signe; Treier, Urs A.; Elberling, Bo; Hansen, Birger U.

I: ISPRS Journal of Photogrammetry and Remote Sensing, Bind 146, 2018, s. 518-529.

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningfagfællebedømt

Harvard

Karami, M, Westergaard-Nielsen, A, Normand, S, Treier, UA, Elberling, B & Hansen, BU 2018, 'A phenology-based approach to the classification of Arctic tundra ecosystems in Greenland', ISPRS Journal of Photogrammetry and Remote Sensing, bind 146, s. 518-529. https://doi.org/10.1016/J.ISPRSJPRS.2018.11.005

APA

Karami, M., Westergaard-Nielsen, A., Normand, S., Treier, U. A., Elberling, B., & Hansen, B. U. (2018). A phenology-based approach to the classification of Arctic tundra ecosystems in Greenland. ISPRS Journal of Photogrammetry and Remote Sensing, 146, 518-529. https://doi.org/10.1016/J.ISPRSJPRS.2018.11.005

Vancouver

Karami M, Westergaard-Nielsen A, Normand S, Treier UA, Elberling B, Hansen BU. A phenology-based approach to the classification of Arctic tundra ecosystems in Greenland. ISPRS Journal of Photogrammetry and Remote Sensing. 2018;146:518-529. https://doi.org/10.1016/J.ISPRSJPRS.2018.11.005

Author

Karami, Mojtaba ; Westergaard-Nielsen, Andreas ; Normand, Signe ; Treier, Urs A. ; Elberling, Bo ; Hansen, Birger U. / A phenology-based approach to the classification of Arctic tundra ecosystems in Greenland. I: ISPRS Journal of Photogrammetry and Remote Sensing. 2018 ; Bind 146. s. 518-529.

Bibtex

@article{8d62fed56718447dbb05475534491b24,
title = "A phenology-based approach to the classification of Arctic tundra ecosystems in Greenland",
abstract = "The disproportionate warming in the Arctic and the resulting adverse ecosystem changes underline the importance of continued monitoring of these ecosystems. Land-cover classification maps of the Arctic regions are essential for monitoring and change detection purposes, as well as upscaling of various ecosystem processes. However, large-scale land cover maps of the Arctic regions are often too coarse to properly capture the heterogeneity of these landscapes. In this study, we bridge this gap through incorporating multi temporal Landsat-8 OLI data in a large-scale land cover classification, and subsequently produce a tundra classification map for the entire Greenland. An algorithm is developed that allows for the extraction of vegetation phenology from single-year time series of 4169 OLI scenes at 30 m resolution despite the low revisit frequency of the satellite and persistent cloud cover. The phenological metrics, satellite-derived wetness, and terrain information are then used to separate land surface classes using a random forest classifier. The optimal algorithm parameters and input layers are identified, ultimately yielding a cross-validation accuracy of 89.25% across the studied area. Finally, we have conducted a comprehensive analysis on the resulting land-cover map and for the first time presented the geographical distribution, latitudinal gradients, and climate linkages of the various tundra vegetation classes across the ice-free part of Greenland. With a resolution of 30 m and Greenland-wide spatial coverage, the produced land-cover map can support various applications at scales ranging from the landscape to regional level.",
author = "Mojtaba Karami and Andreas Westergaard-Nielsen and Signe Normand and Treier, {Urs A.} and Bo Elberling and Hansen, {Birger U.}",
note = "CENPERM[2018]",
year = "2018",
doi = "10.1016/J.ISPRSJPRS.2018.11.005",
language = "English",
volume = "146",
pages = "518--529",
journal = "ISPRS Journal of Photogrammetry and Remote Sensing",
issn = "0924-2716",
publisher = "Elsevier",

}

RIS

TY - JOUR

T1 - A phenology-based approach to the classification of Arctic tundra ecosystems in Greenland

AU - Karami, Mojtaba

AU - Westergaard-Nielsen, Andreas

AU - Normand, Signe

AU - Treier, Urs A.

AU - Elberling, Bo

AU - Hansen, Birger U.

N1 - CENPERM[2018]

PY - 2018

Y1 - 2018

N2 - The disproportionate warming in the Arctic and the resulting adverse ecosystem changes underline the importance of continued monitoring of these ecosystems. Land-cover classification maps of the Arctic regions are essential for monitoring and change detection purposes, as well as upscaling of various ecosystem processes. However, large-scale land cover maps of the Arctic regions are often too coarse to properly capture the heterogeneity of these landscapes. In this study, we bridge this gap through incorporating multi temporal Landsat-8 OLI data in a large-scale land cover classification, and subsequently produce a tundra classification map for the entire Greenland. An algorithm is developed that allows for the extraction of vegetation phenology from single-year time series of 4169 OLI scenes at 30 m resolution despite the low revisit frequency of the satellite and persistent cloud cover. The phenological metrics, satellite-derived wetness, and terrain information are then used to separate land surface classes using a random forest classifier. The optimal algorithm parameters and input layers are identified, ultimately yielding a cross-validation accuracy of 89.25% across the studied area. Finally, we have conducted a comprehensive analysis on the resulting land-cover map and for the first time presented the geographical distribution, latitudinal gradients, and climate linkages of the various tundra vegetation classes across the ice-free part of Greenland. With a resolution of 30 m and Greenland-wide spatial coverage, the produced land-cover map can support various applications at scales ranging from the landscape to regional level.

AB - The disproportionate warming in the Arctic and the resulting adverse ecosystem changes underline the importance of continued monitoring of these ecosystems. Land-cover classification maps of the Arctic regions are essential for monitoring and change detection purposes, as well as upscaling of various ecosystem processes. However, large-scale land cover maps of the Arctic regions are often too coarse to properly capture the heterogeneity of these landscapes. In this study, we bridge this gap through incorporating multi temporal Landsat-8 OLI data in a large-scale land cover classification, and subsequently produce a tundra classification map for the entire Greenland. An algorithm is developed that allows for the extraction of vegetation phenology from single-year time series of 4169 OLI scenes at 30 m resolution despite the low revisit frequency of the satellite and persistent cloud cover. The phenological metrics, satellite-derived wetness, and terrain information are then used to separate land surface classes using a random forest classifier. The optimal algorithm parameters and input layers are identified, ultimately yielding a cross-validation accuracy of 89.25% across the studied area. Finally, we have conducted a comprehensive analysis on the resulting land-cover map and for the first time presented the geographical distribution, latitudinal gradients, and climate linkages of the various tundra vegetation classes across the ice-free part of Greenland. With a resolution of 30 m and Greenland-wide spatial coverage, the produced land-cover map can support various applications at scales ranging from the landscape to regional level.

U2 - 10.1016/J.ISPRSJPRS.2018.11.005

DO - 10.1016/J.ISPRSJPRS.2018.11.005

M3 - Journal article

VL - 146

SP - 518

EP - 529

JO - ISPRS Journal of Photogrammetry and Remote Sensing

JF - ISPRS Journal of Photogrammetry and Remote Sensing

SN - 0924-2716

ER -

ID: 208856242