GEOSCAN Search Results: Fastlink

GEOSCAN Menu


TitlePerformance analysis of RapidEye multispectral land-cover mapping for the western coast of Hudson Bay, Nunavut
DownloadDownloads
AuthorOldenborger, G A; Bellehumeur-Génier, O; LeBlanc, A -M; Tremblay, T
SourceGeological Survey of Canada, Open File 8380, 2018, 25 pages (1 sheet), https://doi.org/10.4095/311216
Image
Year2018
PublisherNatural Resources Canada
Documentopen file
Lang.English
MapsPublication contains 1 map
Map Info.physiographic, land cover, 1:800,000
ProjectionUniversal Transverse Mercator Projection, UTM zone 15 (WGS84)
Mediaon-line; digital
File formatreadme
File formatpdf (Adobe® Reader®); rtf; txt (ASCII); xls (Microsoft® Excel®); tif
ProvinceNunavut
NTS55E/01; 55E/02; 55E/03; 55E/06; 55E/07; 55E/08; 55E/09; 55E/10; 55E/11; 55E/14; 55E/15; 55E/16; 55F; 55J/12; 55J/13; 55K; 55L/01; 55L/02; 55L/07; 55L/08; 55L/09; 55L/10; 55L/16
AreaKivalliq; Hudson Bay; Rankin Inlet; Whale Cove; Arviat
Lat/Long WENS -95.2500 -91.5000 63.0000 61.0000
Subjectsgeophysics; economic geology; surficial geology/geomorphology; remote sensing; satellite imagery; mapping techniques; vegetation; sediments; glacial deposits; glacial landforms; tills; eskers; marine sediments; raised beaches; boulders; wetlands; bedrock geology; surface waters; lakes; topography; land use; mineral deposits; aggregates; sands; gravels; mineral potential; permafrost; ground ice; land cover; RapidEye; data processing; training set; classification; land use planning; thaw-sensitive permafrost; plants; lichen; moss; tundra
Illustrationslocation maps; geoscientific sketch maps; tables; satellite images
Viewing
Location
 
Natural Resources Canada Library - Ottawa (Earth Sciences)
 
ProgramPermafrost, Climate Change Geoscience
Released2018 09 20
Abstract(Summary)
This Open File presents results of investigations into land-cover mapping using multispectral RapidEye satellite imagery along the western coast of Hudson Bay in the Kivalliq region of Nunavut. Land-cover mapping is similar to remote predictive mapping for surficial materials, with the distinction that for land-cover mapping, establishment of the training set and classification are performed in terms of the bio-physical conditions of the land surface as opposed to surficial geology. A series of land-cover maps are developed for the western coast of Hudson Bay based on a training set of land-cover classes developed through expert visual examination of the satellite images. Using the initial training set, classification of a band-limited mosaic image exhibits limited prediction accuracy and poor separability of classes related to extensive tundra vegetation over glacial till. Subsequent grouping of classes results in improved prediction accuracy, increased separability, and reduced confusion between classes. After class grouping, confusion persists between some land-cover types including wetlands, bedrock, boulders and dry ground in various settings with partial coverage by black lichen or moss. Addition of topographic position to the multispectral data improves classification statistics, and resolves some confusion between wetlands and other land-cover types. However, addition of topography also results in over-mapping of bedrock, and misclassification of water as vegetated land, and therefore, should only be used to resolve potential confusion of land-cover types in localized areas of interest. The land-cover maps are not intended to necessarily represent what may actually be observed by a scientist on the ground. Rather, they provide a high-resolution accompaniment in the interpretation of other available information and may be used for assessing aggregate potential, for land-use planning, or for mapping ground ice distribution and thaw-sensitive permafrost.
Summary(Plain Language Summary, not published)
This Open File presents results of investigations into land cover mapping using multi-spectral RapidEye satellite imagery along the western coast of Hudson Bay in the Kivalliq region of Nunavut. Land cover mapping uses expert training information to perform semi-automated classification of satellite imagery in terms of the bio-physical conditions of the land surface. A series of land cover maps are developed, incorporating different groupings of the training classes and topography. Classification performance is moderate and confusion persists between some land cover types. Addition of topography improves classification, but also results in erroneous mapping, and should only be used to resolve potential confusion of land cover types in localized areas of interest. The land cover maps provide a high-resolution accompaniment in the interpretation of other available information and may be used for assessing aggregate potential, for land-use planning, or for mapping ground ice distribution and thaw-sensitive permafrost.
GEOSCAN ID311216