Toward Digital Soil Mapping in Canada: Existing Soil Survey Data and Related Expert Knowledge

Abstract

Digital soil mapping involves the creation of new raster-based soil attribute datasets from existing soil and environmental data, coupled with other spatial knowledge of soil distribution. The GlobalSoilMap.net project is intended to provide a digital soil map of the world on a 90 m raster base derived from existing soil data sources in each country. This paper provides an overview of the exist-ing soil information holdings in the Canadian Soil Information System (CanSIS) in terms of their scale, coverage, and potential suitability for digital soil mapping applications. A description of a possible approach to the capture and transformation of legacy soil survey knowledge for digital soil mapping purposes is also provided. Most historical soil inventory maps and reports in Canada have been produced for the southernmost 20% of Canada's land area by a variety of federal and provincial agencies at scales ranging from 1:20,000 to 1:250,000. Many of these datasets are available in digital format as part of the National Soil Data Base (NSDB) within CanSIS. The NSDB detailed soil map coverages provide the most precise spatial data source for building raster based digital soil mapping products, but coverage is incomplete. The Soil Landscapes of Canada (SLC) map series provides com-plete coverage for all of Canada, at a scale of 1:1 million. SLC maps are less spa-tially precise, but provide the national coverage needed for applications like the GlobalSoilMap.net project. Pedon datasets provide spatial information at specific points, but the sampling density is very low, and not well spatially distributed. Understanding the status and relevance of the NSDB data holdings, and how they can be effectively combined with expert knowledge, digital terrain models, and other data sources are important for organizing our approach to future digital soil mapping work in Canada. Soil forms on a continuum of topography and surficial material and as such, soil spatial variability should be inherent in models describing soil genesis and develop-ment. In practice, soil survey is often characterized by treating soils as geographic bodies. In the last 80 years, soil survey in Canada has focused on the identification of bodies of related soils that can be recognized as natural units and on their prediction and delineation on maps (Coen, 1987). Although the resulting maps and reports are valuable information sources for users such as policy makers, land managers or farmers, the true nature of soil spatial distribution is not well represented. This lack of spatial specificity has meant that these map polygons are often not suitable for system modeling or other needs for continuum soil property data (Behrens and Scholten, 2006; Carre et al., 2007). Currently there exists only limited effort in field survey in Canada for the pur-poses of soil mapping, even though it is greatly acknowledged that such data are necessary and useful for organizations and individuals working at local through national scales. Recently there have been advances in techniques for the extrapo-lation of existing soil map information along with the development of inference models that can predict the spatial distribution of soil properties and/or classes at varying scales that may not require expensive and timely field soil survey (Grinand et al., 2008; Henderson et al., 2005). This field of work is broadly being termed digital soil mapping. In this paper we aim to review the data holding of the Canadian Soil Information System (CanSIS), to briefly assess the adequacy and usability of the legacy soil sur-vey data for GlobalSoilMap.net (see also Chapter 33) and other digital soil mapping applications and to explore the methods to extract accumulated expert knowledge that is embedded within the existing soil survey data. We expect that this work will provide a much-needed start to national scale research and development on digital soil mapping and will build on and apply the earlier foundational work on digital soil mapping in Canada (MacMillan et al., 2004).