CORINE land cover database of the Netherlands: monitoring land cover changes between 1986 and 2000

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

This paper describes the methodology for updating the CORINE 1986 database for the Netherlands to CLC2000 and the accuracy of the resulting database. The methodology consisted of computer-assisted visual interpretation of satellite images. Furthermore, topographic maps (analogue and digital), aerial photographs and the national land cover database of the Netherlands (LGN4) were used as additional information in the interpretation and verification processes. During the first step, the geometry and thematic contents of the CORINE 1986 land cover interpretation were revised. Next, the CORINE 1986 was updated on the basis of Landsat 7 ETM images of 1999 and 2000. Land cover changes larger than 5 ha were digitized into the database and labeled to ensure that the land cover changes could be discriminated from other changes in the database. Furthermore, each polygon has an attribute label for the land cover class in 1986 as well as 2000. This procedure ensured consistency between the three databases, because the CORINE 1986 classification, the CORINE 2000 classification as well as the land cover changes could be generated from the same database. The validation was based on a stratified random sample whose true land cover types were taken from aerial photographs. The validation revealed an overall accuracy of nearly 95% for the CLC2000 database (level 3) when taking into account that patches smaller than 25 hectares are not allowed in the CLC database. Omitting this condition reduces the accuracy by 30%. The changes in the CLCchange database have a user and producer's accuracy of 76.1 and 91.1%. The producer's accuracy indicates an overestimation of changes of almost 10%. Comparing the number of changes with the Dutch National database also suggests a slight overestimation of changes.
Original languageEnglish
Pages (from-to)382-387
JournalEARSeL eProceedings
Volume3
Issue number3
Publication statusPublished - 2004

Fingerprint

Netherlands
land cover
Monitoring
monitoring
photographs
Landsat 7
methodology
aerial photograph
polygons
Antennas
analogs
Landsat
polygon
Labels
geometry
Satellites

Cite this

@article{cd949cedabe240fb9cd50488a496b0ca,
title = "CORINE land cover database of the Netherlands: monitoring land cover changes between 1986 and 2000",
abstract = "This paper describes the methodology for updating the CORINE 1986 database for the Netherlands to CLC2000 and the accuracy of the resulting database. The methodology consisted of computer-assisted visual interpretation of satellite images. Furthermore, topographic maps (analogue and digital), aerial photographs and the national land cover database of the Netherlands (LGN4) were used as additional information in the interpretation and verification processes. During the first step, the geometry and thematic contents of the CORINE 1986 land cover interpretation were revised. Next, the CORINE 1986 was updated on the basis of Landsat 7 ETM images of 1999 and 2000. Land cover changes larger than 5 ha were digitized into the database and labeled to ensure that the land cover changes could be discriminated from other changes in the database. Furthermore, each polygon has an attribute label for the land cover class in 1986 as well as 2000. This procedure ensured consistency between the three databases, because the CORINE 1986 classification, the CORINE 2000 classification as well as the land cover changes could be generated from the same database. The validation was based on a stratified random sample whose true land cover types were taken from aerial photographs. The validation revealed an overall accuracy of nearly 95{\%} for the CLC2000 database (level 3) when taking into account that patches smaller than 25 hectares are not allowed in the CLC database. Omitting this condition reduces the accuracy by 30{\%}. The changes in the CLCchange database have a user and producer's accuracy of 76.1 and 91.1{\%}. The producer's accuracy indicates an overestimation of changes of almost 10{\%}. Comparing the number of changes with the Dutch National database also suggests a slight overestimation of changes.",
author = "G.W. Hazeu and {de Wit}, A.J.W.",
year = "2004",
language = "English",
volume = "3",
pages = "382--387",
journal = "EARSeL eProceedings",
issn = "1729-3782",
publisher = "European Association of Remote Sensing Laboratories",
number = "3",

}

CORINE land cover database of the Netherlands: monitoring land cover changes between 1986 and 2000. / Hazeu, G.W.; de Wit, A.J.W.

In: EARSeL eProceedings, Vol. 3, No. 3, 2004, p. 382-387.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - CORINE land cover database of the Netherlands: monitoring land cover changes between 1986 and 2000

AU - Hazeu, G.W.

AU - de Wit, A.J.W.

PY - 2004

Y1 - 2004

N2 - This paper describes the methodology for updating the CORINE 1986 database for the Netherlands to CLC2000 and the accuracy of the resulting database. The methodology consisted of computer-assisted visual interpretation of satellite images. Furthermore, topographic maps (analogue and digital), aerial photographs and the national land cover database of the Netherlands (LGN4) were used as additional information in the interpretation and verification processes. During the first step, the geometry and thematic contents of the CORINE 1986 land cover interpretation were revised. Next, the CORINE 1986 was updated on the basis of Landsat 7 ETM images of 1999 and 2000. Land cover changes larger than 5 ha were digitized into the database and labeled to ensure that the land cover changes could be discriminated from other changes in the database. Furthermore, each polygon has an attribute label for the land cover class in 1986 as well as 2000. This procedure ensured consistency between the three databases, because the CORINE 1986 classification, the CORINE 2000 classification as well as the land cover changes could be generated from the same database. The validation was based on a stratified random sample whose true land cover types were taken from aerial photographs. The validation revealed an overall accuracy of nearly 95% for the CLC2000 database (level 3) when taking into account that patches smaller than 25 hectares are not allowed in the CLC database. Omitting this condition reduces the accuracy by 30%. The changes in the CLCchange database have a user and producer's accuracy of 76.1 and 91.1%. The producer's accuracy indicates an overestimation of changes of almost 10%. Comparing the number of changes with the Dutch National database also suggests a slight overestimation of changes.

AB - This paper describes the methodology for updating the CORINE 1986 database for the Netherlands to CLC2000 and the accuracy of the resulting database. The methodology consisted of computer-assisted visual interpretation of satellite images. Furthermore, topographic maps (analogue and digital), aerial photographs and the national land cover database of the Netherlands (LGN4) were used as additional information in the interpretation and verification processes. During the first step, the geometry and thematic contents of the CORINE 1986 land cover interpretation were revised. Next, the CORINE 1986 was updated on the basis of Landsat 7 ETM images of 1999 and 2000. Land cover changes larger than 5 ha were digitized into the database and labeled to ensure that the land cover changes could be discriminated from other changes in the database. Furthermore, each polygon has an attribute label for the land cover class in 1986 as well as 2000. This procedure ensured consistency between the three databases, because the CORINE 1986 classification, the CORINE 2000 classification as well as the land cover changes could be generated from the same database. The validation was based on a stratified random sample whose true land cover types were taken from aerial photographs. The validation revealed an overall accuracy of nearly 95% for the CLC2000 database (level 3) when taking into account that patches smaller than 25 hectares are not allowed in the CLC database. Omitting this condition reduces the accuracy by 30%. The changes in the CLCchange database have a user and producer's accuracy of 76.1 and 91.1%. The producer's accuracy indicates an overestimation of changes of almost 10%. Comparing the number of changes with the Dutch National database also suggests a slight overestimation of changes.

M3 - Article

VL - 3

SP - 382

EP - 387

JO - EARSeL eProceedings

JF - EARSeL eProceedings

SN - 1729-3782

IS - 3

ER -