ABSTRACT: -Remote predictive mapping of surficial geology in aid of regional land-use planning in the Far North of Ontario, Canada.

CGRG Bibliography of Canadian Geomorphology
Search Results

Author :

Date :

Title :

Publication :

Issue :

Page(s) :

Abstract

In 2008, the Ontario government announced plans to permanently protect half of the Far North region of Ontario and launched a planning process to support this goal. During the initial stages of planning the need for primary landscape data became apparent. For example, existing surficial geology map coverage is at a scale of 1:250 000 and not sufficient for regional land-use planning. A project to remotely predict surficial materials was initiated by the Ontario Geological Survey in response to this information need. Remotely sensed imagery which is available at various resolutions and degrees of coverage, and existing digital elevation models (DEMs) can be used as proxies for the landscape. Combining information gained from the imagery, such as vegetation type and moisture conditions, with landform recognition from DEMs, various landform/sediment and vegetation cover/material relationships can be determined and applied to remotely predict surficial material distribution. Object based image analysis is being used in the processing of imagery and DEM data. On-screen digitization of select landforms supplements the predicted material distribution and will occur as symbols on the maps. Ground verification is based on data collected during previous field programs and a modest amount of field work. Key aspects of the field work include making observations of landform/sediment relations, observations of the associated vegetation communities and attempting to understand the spatial and stratigraphic relationships between materials. The understanding of these associations is vital to the project’s goal of remotely predicting surficial materials and producing maps of these in the Far North of Ontario. In the Hudson Bay Lowland, the Quaternary sediment succession is usually thick and consists of several glacial and nonglacial sediment sequences, the details of which are exposed primarily along the rivers and creeks of the lowlands (Skinner 1973; Thorleifson, Wyatt and Warman 1993). A few river and sea exposures were visited during field work: some to become familiar with the stratigraphy as it is known and other exposures to collect new information. Field stops to investigate the geological material and vegetation cover associated with various landforms throughout the lowlands were also made. The distribution of field sites visited is depicted in Figure 1. In addition to stops at various types of wetlands and the extensive marine plains; marine shorelines (Figure 2), drumlins and ice margin depositional features were also investigated. SPOT imagery (4 colour bands and the panchromatic band), SRTM 1-arcsecond digital elevation data and its derivatives, in particular SAGA wetness index, and CanVec vector drainage shape files are the primary data sources for this remote predictive mapping exercise. Multiresolution segmentation algorithm, using different image layer weights, scale parameters and homogeneity criterion, within Definiens eCognition Developer 8 object-based image analysis software is used to achieve meaningful objects representing various surficial material types. Objects are then classified based on digital signature, internal variability of signature and proximity to certain vector layers and certain adjacent material types. For example in Figure 3, a SPOT image of an area along eastern James Bay, coastal mudflat deposits are adjacent to oceans and salt marsh deposits are adjacent to mudflats. Elsewhere, alluvium and older alluvium on abandoned terraces occur adjacent to the vector coverage of river features. The remote predictive mapping process that is being developed to produce surficial material maps to aid land use planning in the far north of Ontario appears to be working well. It combines observations collected from field work (both past and present) with the development of a semi-automated, object-based image analysis process for the classification of remotely sensed imagery, to predict and map the distribution of surficial materials. The classification of objects, based on digital signature, internal variability of signature and proximity to certain vector layer and certain adjacent material types, shows promise in improving the existing surficial material maps (scale 1:250 000) of the region. The new remotely predicted maps will be published at a scale of 1:100 000. Figure 4 is an example of the results of the remote predictive mapping process. It displays the predicted surface materials based on the Spot image shown in Figure 3. It shows the spatial relationship between several surficial material types and vegetation communities within this part of the Hudson Bay Lowland. In the future, it is planned to develop a remote predictive mapping process to be applied to the Boreal forest region of the Far North of Ontario. In summary, the initial results from the remote predictive mapping project are promising. There is an improvement on the level of detail of the predicted information when compared to existing maps. This increased detail will greatly improve the base information of surficial materials in the Far North and will help during the land use planning exercise to come.

Bibliography of Canadian Geomorphology