Cartographie de la texture de surface des sols de la montérégie á l'aide de données d'observation de la terre et de sol archivées

Translated title of the contribution: Mapping of the surface texture of the soil Monte 're' gy at using gives you 're observing the earth and soil archive' are

Mohamed Niang, M. C. Nolin, I. Perron, M. Bernier, C. Codjia

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Soil texture is one of the most important soil properties to be considered in soil classification and mapping. A multitemporal series of four RADARSAT-1 (R-1) images, two RADARSAT-2 (R-2) images, one acquired in fine quadpolarization mode (FQ4), and the other in standard dual-polarization mode (HH and HV), an IKONOS image, and a LANDSAT-7 (L-7) image were used for mapping soil surface texture of the Montérégie area. The linear discriminant analysis was used as the classification algorithm. The two soil survey datasets used for training and validation of classification models are soil polygons delineated on published soil maps and a soil profile database described in the field. The results showed a relatively weak influence of the two sources of data on the overall classification accuracies (OCA): 79% with the soil profile database and 75% when using soil maps. However, an adequate selection of training sites was required to achieve these good results. At the regional scale, the spectral information extracted from sensor L-7 gives better classification accuracy (OCA =76%) than the multitemporal R-1 and R-2 (with HH polarization) data (OCA=60%). Therefore, the combination of these two types of data has achieved the best classification results of the study (OCA =79%). At the local scale, the polarimetric information extracted from a R-2 image (FQ4) was significantly more effective (OCA - 72%) than the spectral information derived from the IKONOS image (OCA=57%) for mapping soil surface texture of a representative area (625 km2) of the Rouville county. This study demonstrates the usefulness of earth observation for updating soil maps at the regional (1: 40 000) and local scale (1: 20 000).

Original languageFrench
Pages (from-to)548-563
Number of pages16
JournalCanadian Journal of Remote Sensing
Volume37
Issue number5
Publication statusPublished - 1 Oct 2011
Externally publishedYes

Fingerprint

texture
soil
IKONOS
RADARSAT
soil profile
soil surface
polarization
soil classification
soil survey
image classification
polygon
soil texture
discriminant analysis
soil property
sensor
soil map

ASJC Scopus subject areas

  • Earth and Planetary Sciences(all)

Cite this

Cartographie de la texture de surface des sols de la montérégie á l'aide de données d'observation de la terre et de sol archivées. / Niang, Mohamed; Nolin, M. C.; Perron, I.; Bernier, M.; Codjia, C.

In: Canadian Journal of Remote Sensing, Vol. 37, No. 5, 01.10.2011, p. 548-563.

Research output: Contribution to journalArticle

Niang, Mohamed ; Nolin, M. C. ; Perron, I. ; Bernier, M. ; Codjia, C. / Cartographie de la texture de surface des sols de la montérégie á l'aide de données d'observation de la terre et de sol archivées. In: Canadian Journal of Remote Sensing. 2011 ; Vol. 37, No. 5. pp. 548-563.
@article{7c3aeda351764d0c877740ee854362a4,
title = "Cartographie de la texture de surface des sols de la mont{\'e}r{\'e}gie {\'a} l'aide de donn{\'e}es d'observation de la terre et de sol archiv{\'e}es",
abstract = "Soil texture is one of the most important soil properties to be considered in soil classification and mapping. A multitemporal series of four RADARSAT-1 (R-1) images, two RADARSAT-2 (R-2) images, one acquired in fine quadpolarization mode (FQ4), and the other in standard dual-polarization mode (HH and HV), an IKONOS image, and a LANDSAT-7 (L-7) image were used for mapping soil surface texture of the Mont{\'e}r{\'e}gie area. The linear discriminant analysis was used as the classification algorithm. The two soil survey datasets used for training and validation of classification models are soil polygons delineated on published soil maps and a soil profile database described in the field. The results showed a relatively weak influence of the two sources of data on the overall classification accuracies (OCA): 79{\%} with the soil profile database and 75{\%} when using soil maps. However, an adequate selection of training sites was required to achieve these good results. At the regional scale, the spectral information extracted from sensor L-7 gives better classification accuracy (OCA =76{\%}) than the multitemporal R-1 and R-2 (with HH polarization) data (OCA=60{\%}). Therefore, the combination of these two types of data has achieved the best classification results of the study (OCA =79{\%}). At the local scale, the polarimetric information extracted from a R-2 image (FQ4) was significantly more effective (OCA - 72{\%}) than the spectral information derived from the IKONOS image (OCA=57{\%}) for mapping soil surface texture of a representative area (625 km2) of the Rouville county. This study demonstrates the usefulness of earth observation for updating soil maps at the regional (1: 40 000) and local scale (1: 20 000).",
author = "Mohamed Niang and Nolin, {M. C.} and I. Perron and M. Bernier and C. Codjia",
year = "2011",
month = "10",
day = "1",
language = "French",
volume = "37",
pages = "548--563",
journal = "Canadian Journal of Remote Sensing",
issn = "0703-8992",
publisher = "Canadian Aeronautics and Space Institute",
number = "5",

}

TY - JOUR

T1 - Cartographie de la texture de surface des sols de la montérégie á l'aide de données d'observation de la terre et de sol archivées

AU - Niang, Mohamed

AU - Nolin, M. C.

AU - Perron, I.

AU - Bernier, M.

AU - Codjia, C.

PY - 2011/10/1

Y1 - 2011/10/1

N2 - Soil texture is one of the most important soil properties to be considered in soil classification and mapping. A multitemporal series of four RADARSAT-1 (R-1) images, two RADARSAT-2 (R-2) images, one acquired in fine quadpolarization mode (FQ4), and the other in standard dual-polarization mode (HH and HV), an IKONOS image, and a LANDSAT-7 (L-7) image were used for mapping soil surface texture of the Montérégie area. The linear discriminant analysis was used as the classification algorithm. The two soil survey datasets used for training and validation of classification models are soil polygons delineated on published soil maps and a soil profile database described in the field. The results showed a relatively weak influence of the two sources of data on the overall classification accuracies (OCA): 79% with the soil profile database and 75% when using soil maps. However, an adequate selection of training sites was required to achieve these good results. At the regional scale, the spectral information extracted from sensor L-7 gives better classification accuracy (OCA =76%) than the multitemporal R-1 and R-2 (with HH polarization) data (OCA=60%). Therefore, the combination of these two types of data has achieved the best classification results of the study (OCA =79%). At the local scale, the polarimetric information extracted from a R-2 image (FQ4) was significantly more effective (OCA - 72%) than the spectral information derived from the IKONOS image (OCA=57%) for mapping soil surface texture of a representative area (625 km2) of the Rouville county. This study demonstrates the usefulness of earth observation for updating soil maps at the regional (1: 40 000) and local scale (1: 20 000).

AB - Soil texture is one of the most important soil properties to be considered in soil classification and mapping. A multitemporal series of four RADARSAT-1 (R-1) images, two RADARSAT-2 (R-2) images, one acquired in fine quadpolarization mode (FQ4), and the other in standard dual-polarization mode (HH and HV), an IKONOS image, and a LANDSAT-7 (L-7) image were used for mapping soil surface texture of the Montérégie area. The linear discriminant analysis was used as the classification algorithm. The two soil survey datasets used for training and validation of classification models are soil polygons delineated on published soil maps and a soil profile database described in the field. The results showed a relatively weak influence of the two sources of data on the overall classification accuracies (OCA): 79% with the soil profile database and 75% when using soil maps. However, an adequate selection of training sites was required to achieve these good results. At the regional scale, the spectral information extracted from sensor L-7 gives better classification accuracy (OCA =76%) than the multitemporal R-1 and R-2 (with HH polarization) data (OCA=60%). Therefore, the combination of these two types of data has achieved the best classification results of the study (OCA =79%). At the local scale, the polarimetric information extracted from a R-2 image (FQ4) was significantly more effective (OCA - 72%) than the spectral information derived from the IKONOS image (OCA=57%) for mapping soil surface texture of a representative area (625 km2) of the Rouville county. This study demonstrates the usefulness of earth observation for updating soil maps at the regional (1: 40 000) and local scale (1: 20 000).

UR - http://www.scopus.com/inward/record.url?scp=84870037462&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84870037462&partnerID=8YFLogxK

M3 - Article

VL - 37

SP - 548

EP - 563

JO - Canadian Journal of Remote Sensing

JF - Canadian Journal of Remote Sensing

SN - 0703-8992

IS - 5

ER -