Publication Cover
Canadian Journal of Remote Sensing
Journal canadien de télédétection
Volume 44, 2018 - Issue 5
162
Views
3
CrossRef citations to date
0
Altmetric
Original Articles

Mapping Surficial Materials in Nunavut using RADARSAT-2 C-HH and C-HV, Landsat-8 OLI, DEM and Slope Data

, , , &
Pages 491-512 | Received 04 Jul 2018, Accepted 05 Nov 2018, Published online: 12 Jan 2019
 

Abstract

The Canadian Arctic is currently subject to increased mapping activities for providing better knowledge to assist in making informed decisions for sustainable development. Surficial material maps are one of the required maps. For an area located in Nunavut, we produced a map with 21 surficial material classes by applying a non-parametric classifier, Random Forests (RF), to a combination of RADARSAT-2 C-HH and C-HV with Landsat-8 OLI, digital elevation model, and slope data. We also tested the All-polygon and Sub-polygon scripts of RF. Validation accuracies were determined by comparing the resulting maps to more than 1000 field sites. By adding RADARSAT-2 dual-polarized images, the classification overall accuracy increases from 90.6% to 96.4% with the Sub-polygon script and from 92.8% to 98.1% with the All-polygon script. The overall validation accuracy increases from 76.3% to 88.9% with the Sub-polygon script and from 76.4% to 93.3% with the All-polygon script. With the All-polygon script, the validation accuracies are above 85% for all classes, except the user’s accuracy of gravelly till (76.7%) and the producer’s accuracy of sand and gravel with vegetation (70%), both classes being confused with thin till over bedrock.

RÉSUMÉ

L’Arctique canadien fait actuellement l’objet d’une cartographie intensive pour améliorer les connaissances du milieu et prendre des décisions éclairées en vue d’un développement durable. Les cartes de matériaux superficiels sont un des types de cartes à réaliser. Pour une région localisée au Nunavut, nous avons réalisé une carte ayant 21 classes de matériaux superficiels, en appliquant un classificateur non paramétrique, Random Forests (RF), à une combinaison d’images RADARSAT-2 (C-HH et C-HV) et Landsat-8 OLI avec un modèle numérique d’élévation et des données de pente. Nous avons aussi testé deux différents codes (All-polygon et Sub-polygon) de RF. Les précisions de validation ont été déterminées en comparant les cartes produites à plus de 1000 sites de terrain. En ajoutant des images RADARSAT-2 à double polarisation, la précision globale de classification passe de 90,6% à 96,4% avec le code Sub-polygon et de 92,8% à 98,1% avec le code All-polygon. La précision globale de validation augmente de 76,3% à 88,9% avec le code Sub-polygon et de 76,4% à 93,3% avec le code All-polygon. En utilisant finalement le code All-polygon, la précision de validation est supérieure à 85% pour toutes les classes, sauf pour la précision d'utilisateur pour le till graveleux (76,7%) et la précision du producteur pour la classe « sable et gravier couvert de végétation » (70%), les deux classes étant confondues avec le till mince sur roc.

Acknowledgments

We thank the Canadian Space Agency via the Geological Survey of Canada for providing the RADARSAT-2 images to the UNB (University of New Brunswick) team. Special thanks to Deborah Lemkow (GSC) for helping with the RADARSAT-2 ordering, the Janet Campbell review, and the anonymous reviewers.

Additional information

Funding

J. Byatt was funded by a post-graduate Master scholarship from Natural Sciences and Engineering Research Council of Canada (NSERC), New Brunswick Innovation Foundation (NBIF) (STEM & Social Innovation award), and he received a W. Garfield Weston Award in Northern Research (Doctoral) managed by the Association of Canadian Universities for Northern Studies (ACUNS). He was also funded by a NSERC Discovery Grant awarded to Dr. Brigitte Leblon. The field work was funded by the Natural Resources Canadaʼs Geo-mapping for Energy and Minerals (GEM-1) program, as part of the Wager Bay surficial geology activity.

Log in via your institution

Log in to Taylor & Francis Online

PDF download + Online access
  • 48 hours access to article PDF & online version
  • Article PDF can be downloaded
  • Article PDF can be printed
USD 61.00 Add to cart
* Local tax will be added as applicable

Related Research

People also read lists articles that other readers of this article have read.

Recommended articles lists articles that we recommend and is powered by our AI driven recommendation engine.

Cited by lists all citing articles based on Crossref citations.
Articles with the Crossref icon will open in a new tab.