PALSAR Wide-Area Mapping of Borneo: Methodology and Map Validation

D.H. Hoekman, M.A.M. Vissers, N. Wielaard

Research output: Contribution to journalArticleAcademicpeer-review

64 Citations (Scopus)

Abstract

This paper describes the operational radar mapping processing chain developed and steps taken to produce a provisional wide-area PALSAR forest and land cover map covering Borneo for the year 2007, compliant with emerging international standards (CEOS guidelines, FAO LCCS). A Bayesian approach based on (unsupervised) mixture modeling followed by Markov Random Field (MRF) classification has been selected for its suitability and flexibility to deal with a situation where ground truth is sparse and sometimes ambiguous. The methodology is based on the classification of Fine Beam Single (FBS) and Fine Beam Dual (FBD) polarization (path) image pairs. To cover Borneo the equivalent of 554 standard images is required. Qualitative and quantitative validation results and findings are reported. The final overall accuracy assessment result shows the demonstration map product is in 85.5% full agreement with the independent reference dataset and in 7.8% 'partial agreement'. The accuracy achieved is widely considered adequate, a very promising result for a sub-continental high resolution map based on just single-year radar data. Approaches for further improvement of the accuracy of less accurately classified thematic classes such as grassland, cropland and shrubland are suggested. This work has been undertaken in part within the framework of the ALOS Kyoto & Carbon Initiative.
Original languageEnglish
Pages (from-to)605-617
JournalIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Volume3
Issue number4
DOIs
Publication statusPublished - 2010

Fingerprint

PALSAR
methodology
Radar
radar
ALOS
accuracy assessment
Food and Agricultural Organization
shrubland
forest cover
land cover
Demonstrations
polarization
grassland
Polarization
Carbon
carbon
Processing
modeling

Keywords

  • cover
  • images

Cite this

@article{f32f5c8097a84e7ea9dbfc72455a1988,
title = "PALSAR Wide-Area Mapping of Borneo: Methodology and Map Validation",
abstract = "This paper describes the operational radar mapping processing chain developed and steps taken to produce a provisional wide-area PALSAR forest and land cover map covering Borneo for the year 2007, compliant with emerging international standards (CEOS guidelines, FAO LCCS). A Bayesian approach based on (unsupervised) mixture modeling followed by Markov Random Field (MRF) classification has been selected for its suitability and flexibility to deal with a situation where ground truth is sparse and sometimes ambiguous. The methodology is based on the classification of Fine Beam Single (FBS) and Fine Beam Dual (FBD) polarization (path) image pairs. To cover Borneo the equivalent of 554 standard images is required. Qualitative and quantitative validation results and findings are reported. The final overall accuracy assessment result shows the demonstration map product is in 85.5{\%} full agreement with the independent reference dataset and in 7.8{\%} 'partial agreement'. The accuracy achieved is widely considered adequate, a very promising result for a sub-continental high resolution map based on just single-year radar data. Approaches for further improvement of the accuracy of less accurately classified thematic classes such as grassland, cropland and shrubland are suggested. This work has been undertaken in part within the framework of the ALOS Kyoto & Carbon Initiative.",
keywords = "cover, images",
author = "D.H. Hoekman and M.A.M. Vissers and N. Wielaard",
note = "ISI:000285357400006",
year = "2010",
doi = "10.1109/JSTARS.2010.2070059",
language = "English",
volume = "3",
pages = "605--617",
journal = "IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing",
issn = "1939-1404",
publisher = "Institute of Electrical and Electronics Engineers",
number = "4",

}

PALSAR Wide-Area Mapping of Borneo: Methodology and Map Validation. / Hoekman, D.H.; Vissers, M.A.M.; Wielaard, N.

In: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol. 3, No. 4, 2010, p. 605-617.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - PALSAR Wide-Area Mapping of Borneo: Methodology and Map Validation

AU - Hoekman, D.H.

AU - Vissers, M.A.M.

AU - Wielaard, N.

N1 - ISI:000285357400006

PY - 2010

Y1 - 2010

N2 - This paper describes the operational radar mapping processing chain developed and steps taken to produce a provisional wide-area PALSAR forest and land cover map covering Borneo for the year 2007, compliant with emerging international standards (CEOS guidelines, FAO LCCS). A Bayesian approach based on (unsupervised) mixture modeling followed by Markov Random Field (MRF) classification has been selected for its suitability and flexibility to deal with a situation where ground truth is sparse and sometimes ambiguous. The methodology is based on the classification of Fine Beam Single (FBS) and Fine Beam Dual (FBD) polarization (path) image pairs. To cover Borneo the equivalent of 554 standard images is required. Qualitative and quantitative validation results and findings are reported. The final overall accuracy assessment result shows the demonstration map product is in 85.5% full agreement with the independent reference dataset and in 7.8% 'partial agreement'. The accuracy achieved is widely considered adequate, a very promising result for a sub-continental high resolution map based on just single-year radar data. Approaches for further improvement of the accuracy of less accurately classified thematic classes such as grassland, cropland and shrubland are suggested. This work has been undertaken in part within the framework of the ALOS Kyoto & Carbon Initiative.

AB - This paper describes the operational radar mapping processing chain developed and steps taken to produce a provisional wide-area PALSAR forest and land cover map covering Borneo for the year 2007, compliant with emerging international standards (CEOS guidelines, FAO LCCS). A Bayesian approach based on (unsupervised) mixture modeling followed by Markov Random Field (MRF) classification has been selected for its suitability and flexibility to deal with a situation where ground truth is sparse and sometimes ambiguous. The methodology is based on the classification of Fine Beam Single (FBS) and Fine Beam Dual (FBD) polarization (path) image pairs. To cover Borneo the equivalent of 554 standard images is required. Qualitative and quantitative validation results and findings are reported. The final overall accuracy assessment result shows the demonstration map product is in 85.5% full agreement with the independent reference dataset and in 7.8% 'partial agreement'. The accuracy achieved is widely considered adequate, a very promising result for a sub-continental high resolution map based on just single-year radar data. Approaches for further improvement of the accuracy of less accurately classified thematic classes such as grassland, cropland and shrubland are suggested. This work has been undertaken in part within the framework of the ALOS Kyoto & Carbon Initiative.

KW - cover

KW - images

U2 - 10.1109/JSTARS.2010.2070059

DO - 10.1109/JSTARS.2010.2070059

M3 - Article

VL - 3

SP - 605

EP - 617

JO - IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing

JF - IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing

SN - 1939-1404

IS - 4

ER -