Kansas Academy of Science

Practical and theoretical basis for mapping landscape sensitivity in the Southern Canadian Interior Plains1

David J. Sauchyn

Department of Geography, University of Regina, Regina, Saskatchewan, Canada S4S 0A2 (sauchyn@leroy.cc.uregina.ca)

This article is published in the Transactions of the Kansas
Academy of Science, vol. 100, no. 1/2, p. 61-72 (1997).

Table of Contents
Introduction Study Area and the GIS Database
Spatial Scale Landscape Sensitivity
Methodology Conclusion
Acknowledgments Literature


The Canadian Climate Centre's general circulation model predicts that, with increased CO2 concentrations, the largest rise in mean surface temperature in southern Canada will occur in the Interior Plains. For five years the Palliser Triangle Global Change Project has focused on geomorphic responses to climate in the driest part of this region. A major component of this project is the GIS modeling and mapping of landscape sensitivity, the potential for change in rates of surface processes. Soil landscapes, defined by superimposing 1:100,000 digital soil maps with slope polygons derived from 1:50,000 digital topographic maps, are combined with digital maps of climate, surficial geology, land cover and hydrography to evaluate the climatic and geopotential energy for geomorphic work (e.g., rainfall erosivity, relief) and resistance to geomorphic activity (e.g., soil texture, land cover). The spatial relations between resistance and potential disturbance are the basis for determining the sensitivity of soil landscapes. Dimensionless, smoothed and synthetic data are the basis for expressing the properties of slopes and streams as regional landscape parameters. This methodology is better suited to a regional scale and a subhumid poorly integrated plains landscape than are empirical soil loss models or the more demanding physically based models.
1 Palliser Triangle Global Change Contribution no. 32


With the availability of digital geographic data, widespread use of GIS, and greater emphasis on social relevance of scientific research, modeling of biophysical processes and landscape change is increasingly spatially distributed over larger areas. In geomorphology, the traditional one-dimensional models of slope profiles and stream channels have been extrapolated to two-dimensional landscape-scale models of landforms and geomorphic processes (e.g., Moore et al. 1988; Dietrich et al. 1992). This generally is achieved by coupling process models and relatively high-resolution digital geographic data, using a GIS (e.g., Desmet and Govers, 1995; Mitasova et al. 1996).

This paper describes a conceptual framework for the GIS-modeling of landscape sensitivity to climatic variability and change. The identification and mapping of metastable (sensitive) soil landscapes at a regional scale (1:100,000) differs significantly from the prediction of soil erosion at local scales over much smaller areas, the typical function of geomorphic models. Therefore existing models of geomorphic processes and landscape change are not transferable to our study without adjustments for scale. Furthermore the southern Canadian Interior Plains lack the fluvial dissection, drainage development and integration of slopes and channels that most models assume. The conceptual framework for our regional evaluation of landscape sensitivity has two parts: a strategy to characterize regional soil landscapes with parameters measured at local scales, and a model to convert these parameters into an expression of landscape sensitivity.

This research is a major component of the Palliser Triangle Global Change Project initiated by the Geological Survey of Canada to examine geomorphic responses to climatic change and variability in the driest part of the Interior Plains (Lemmen et al. 1993). According to the Canadian Climate Centre's general circulation model (GCM), the largest CO2-induced rise in mean surface temperature in southern Canada will occur in the Interior Plains (Boer et al. 1992). The impacts of this climatic change will be greatest at the margins of land and climate suitable for annual crop production, that is, the subhumid Palliser Triangle--see Fig. 1. The sustainability of dryland agriculture depends, to a large degree, on adjustment of land use and production systems to climatic variability, the periodic fluctuation of atmospheric conditions (e.g., drought, early frosts, major storms) and to climatic change, a significant departure from previous average conditions (Environment Canada, 1995). Historically prairie agriculture has succeeded in adapting to climatic variability (Hill and Vaisey, 1995:52).

Various studies have examined the potential direct or 'first-order' impacts of global climatic change on prairie agriculture (e.g., Schweger and Hooey, 1991). However little attention has been paid to the links among land and water resources, climatic variability and change, and rates of earth surface processes (e.g., Wheaton, 1984; Favis-Mortlock and Boardman, 1995). Any attempt to anticipate the degree and distribution of landscape sensitivity must first consider the distinctive physical and cultural landscapes of the southern Interior Plains. The unique character of this region strongly influences the suitability of existing models of surface processes and the application of geomorphic research to institutional land management.

Return to Table of Contents


The driest part of the southern Interior Plains is the brown soil zone of southwestern Saskatchewan and southeastern Alberta--see Fig. 1, also known as the Palliser Triangle. It is the only major region of Canada where seasonal water deficits characterize the physical environment and limit economic activity. Most landforms and surficial deposits are the product of late-Pleistocene deglaciation. Due to the dry climate and recent geomorphic history, the landscape is poorly integrated, there are few permanent streams, and large areas are internally drained. In general the region lacks the order and characteristic landforms of a well-developed fluvial landscape and rather consists of largely disconnected soil landscapes with various geomorphic histories. Persistence of Tertiary and Pleistocene morphology is juxtaposed with landforms of recent origin (Sauchyn, 1993a).

The surficial geology is dominated by glacial drift derived primarily from underlying argillaceous, poorly consolidated Cretaceous sediments. The glacial soils support a vast area of cropland and pasture, but are susceptible to erosion where plant cover is lacking. Rates of erosion on cropland are 2-3 times higher than sediment yields in small watersheds (Carson & Associates, 1990), because wind and water erosion mostly redistribute soil within local landscapes, especially in the hummocky moraine (Martz and de Jong, 1991; Pennock and de Jong, 1990). Fields (single crops) are typically a quarter section (65 ha) and the average farm is about 800 ha (almost 2000 acres). Cattle ranches encompass many sections (1000s ha) and represent the only land use over large areas. The rural population density is less than one person per km². Five Canadian census divisions, representing 104,043 km² of the brown soil zone, have a population of 133,116, but 44% of this total is in the cities of Swift Current and Medicine Hat--see Fig. 1 (Statistics Canada, 1992a, 1992b). In this context, detailed geomorphic studies have little immediate socio-economic value.

Table 1 lists the digital geographic data available for the study area. The basic components are 1:100,000 soil maps and 1:50,000 topographic maps. The basic spatial units are the soil map polygons, as these are defined in terms of soil and landform. Digital topographic data are preferred to the categorical (slope class) and descriptive (landform type) data provided with a soil survey. Assigned to each soil landscape polygon are the inherent physical characteristics that govern geomorphic response to climatic variability and change. Over 100s of km² and 100s of years (steady time), the soil landscape boundaries and attributes are considered independent of climate change.

Table 1. Contents of the Palliser Triangle digital geographic database.
Digital map Scale* Source Region or Map Sheet
Soil 1:1000 CLBBR Alberta and Saskatchewan
1:100 SIP S.W. Saskatchewan
Surficial geology 1:250 David (1964) 72K
1:250 Klassen (1991, 1992) 72F, 72G
1:250 Shetson (1987) southern Alberta
1:500 SRC 72J, 72L, 72H
Land cover 1:50 SRC S.W. Saskatchewan
Topography 1:50 CSMA S.W. Saskatchewan
1:20 ABSM S.E. Alberta
Climate 1:250 AES complete
DEM 1:20 U of C RMs 138 and 139
1:20 CSMA Cypress Hills, Sask.

* Scale of hardcopy map from which data were derived or scale at which data
are output, expressed as a representative fraction with denominator in 1000's.

Abbreviated phrases: Saskatchewan Institute of Pedology (SIP), Central Survey and Mapping Agency (CSMA), University of Calgary (U of C), digital elevation model (DEM), Geological Survey of Canada (GSC), Centre for Land and Biological Resource Research (CLBBR), Alberta Bureau of Surveys and Mapping (ABSM), Atmospheric Environment Service (AES), Rural Municipality (RM).

The digital land cover data are classified Landsat TM data from 1991-92. They give the distribution of crop, pasture, treed areas, bare ground and water. Whereas the interaction between vegetation, climate and geomorphic processes could be modeled (Kirkby, 1995), land cover in this region is mostly a function of human activities. It changes annually and seasonally. Soil conservation practices (e.g., minimum tillage) reflect the continuous adjustment of land use and management to climatic variability. Since human responses to climate are beyond the scope of this investigation, there are only three land cover scenarios: 1991-92 (applicable over a longer period bounded by major land use adjustment in response to institutional and economic factors), pre-settlement (native) cover, and native cover for a given climatic change scenario.

Return to Table of Contents


The central issue of spatial scale in geomorphology (Schumm, and Lichty, 1965; de Boer, 1992) includes the dichotomy between the slope- and channel-scale of geomorphic processes and the regional scale of landscapes and institutional land and water management. Conventional geomorphic modeling has been profile-based, or at the scale of small catchments, where individual hillslopes and channels can be represented (e.g., Kirkby, 1989). Despite the use of GIS in the earth sciences (Bonham-Carter, 1994; Vitek et al. 1996) and the application of geomorphology to environmental issues (Eybergen and Imeson, 1989; Patrono, 1995), the spatial dimensions of geomorphic modeling have remained relatively small. For example, Mitasova et al. (1996:630) developed "methods for computation of topographic factors ... suitable for complex terrain and applicable to large areas." Their methods were applied to a "small region" (500 m by 500 m) and a second area, 3.6 km by 4 km. Similarly, Desmet and Govers (1995) simulated patterns of erosion and deposition over an area of 1.23 km². With objectives closely related to ours, Kerenyi and Csorba (1991) examined "the sensitivity of the landscape to climatic conditions" in a 9 km² area. By comparison, our study area is nine 1:250,000 map sheets (138,600 km²), an area larger than England and about the size of Arkansas.

Disaggregating this large region into component slopes is technically feasible with the functions of a GIS (Sauchyn, 1993b), however, this scale of analysis serves little purpose relative to the scale of land use activity, as described. Even if there was a purpose, the data processing, storage and analysis would require major computing resources. When 1:100,000 soil surveys and slope polygons generated from 1:50,000 topographic data were overlaid for one of the nine 1:250,000 map sheet areas, slightly less than 100,000 unique soil landscapes were mapped (Ambrosi, 1995). This was after small map units (< 1 ha) had been eliminated.

The evaluation of agricultural soil loss commonly has involved extrapolation of empirical soil loss equations (e.g., Logan et al. 1982; Snell, 1985). This requires judicious interpretation of soil loss predictions, because parameter values averaged over heterogeneous map units represent a misuse of process models derived from plot-scale research (Wischmeier, 1976; Roels, 1985). Anderson and Knapic (1984) mapped soil erosion risk for western Canada at a scale of 1:2,500,000. Whereas these maps provide a broad geographic overview, they are based necessarily on a limited number of factors and data of variable quality, given the size and diversity of the soil landscapes. Our objective and possible results are not maps of soil loss and gain, but rather the identification of sensitive soil landscapes, those combinations of soil, landform and land cover that may respond to disturbance of a particular magnitude. We aim to maintain the rigor of a theoretical approach, while evaluating landscape sensitivity at regional scales (1:50,000 - 1:100,000). Applying geographic data from large scale maps to the evaluation of landscape sensitivity at smaller map scales requires three strategies for linking spatial scales and the corresponding use of ratio, smoothed and synthetic data (Thorn, 1988).

Ratio or dimensionless data are scale-independent. For example, the ratios of contributing area (slope area per unit contour width) to slope length and local relief to drift thickness (or regional relief) have no units to imply the scale of observation. Relations among morphometric parameters can be maintained over a range of scales. The second strategy is to generalize or smooth the spatial data by trend surface analysis or classification. Either approach leaves broad patterns, although the resulting parameters are either statistical indices (e.g., eigenvectors) or classified (interval) data.

The third strategy for linking scales is to construct a synthetic spatial model based on the statistical relationships among topography, soil and land cover in each soil landscape unit. The dominant (modal) attributes captured by the synthetic landform have the dimensions and physical meaning of the original data. Uncharacteristic slopes and soils (azonal) are excluded from the evaluation of landscape sensitivity, much like the intuitive process of survey and mapping, where map units are delimited and classified according to dominant features. The complexity of a soil landscape, and in turn the storage and processing of its attributes, is reduced to the analysis of a synthetic landform with the correlative soil and land cover. The GIS enables a two-dimensional extension to the synthetic slope profile (Caine, 1979).

Return to Table of Contents


As understood and applied here, landscape sensitivity is "the likelihood that a given change in the controls of a system will produce a sensible, recognizable and persistent response" (Brunsden and Thornes, 1979:476). Figure 2 is a reproduction of Brunsden and Thornes' (1979:477) graphic analogy of the unstable, metastable (sensitive) and stable states in a dynamic geomorphic system. This view of sensitivity led to the concept of the landscape change safety factor: "the ratio of the magnitude of barriers to change [resistance] to the magnitude of the disturbing forces [energy for geomorphic work]" (Brunsden and Thornes, 1979:476). Therein lies the basis for evaluating landscape sensitivity. In fact, Brunsden and Thornes (1979:478) recommended that "A project for future studies will be to map these safety factor distributions as a predictive aid to landform change studies."

The landscape change safety factor is both a conceptual and mathematical model. It implies that the distribution of landforms and geomorphic processes reflects the spatial covariation of disturbing and resisting forces. As a mathematical expression, it requires that the barriers and disturbing forces be identified and quantified. Table 2 is list of parameters which represent either measures of disturbance by water, wind or gravity or barriers to these forces. It does not include parameters, barriers in particular, that cannot be evaluated from commonly mapped data because, for example, they reflect the geomorphic history of a soil landscape (Brunsden, 1993).

Table 2. Parameters for the evaluation of the landscape change sensitivity factor
Parameter Process Reference
annual storm fluvial
10-yr storm fluvial
soil water storage fluvial, eolian de Ploey, Kirkby and Ahnert (1991)
contributing area fluvial
slope gradient fluvial, mass wasting
wind magnitude-frequency eolian Muhs and Maat (1993)
local relief mass wasting
land cover fluvial, eolian
soil texture fluvial, eolian
precipitation/ potential evapotranspiration eolian Wolfe, in press
surficial geology mass wasting
bedrock geology mass wasting

Even though the parameters in Table 2 correspond to data in our GIS database, the landscape safety factor cannot be solved numerically, because for some parameters, especially barriers such as land use and soil texture, the data are categorical. Considerable field data would be required to convert soil and vegetation types to quantitative measures of resistance (e.g., shear strength, boundary layer roughness). There are empirical expressions of soil erodibility and land cover, in particular, the K and C factors in the Universal Soil Loss Equation. However these factors, and the USLE in general, have been applied to the Canadian plains with limited success, given the different soil, climate and topography from western Kansas where the USLE was derived (Pennock and de Jong, 1990).

Return to Table of Contents


The quantifiable parameters are almost exclusively climatic and topographic. Thus the disturbing forces can be expressed as numerical indices. Climatic erosion potential (de Ploey, Kirkby, and Ahnert, 1991; Kirkby, and Cox., 1995) and an index of wind magnitude-frequency (Muhs and Maat, 1993; Wolfe, in press) can be computed for meteorological stations and assigned to corresponding elevation-weighted Theissen polygons. The synthetic landforms for each soil landscape have dimensionless gradients, contributing areas and local relief. Some barriers also are related to topography and climate: landscape disorder, the ratio precipitation to potential evapotranspiration, and persistent water deficit (drought) or surplus which reduce resistance to surface processes and landsliding, respectively.

The methodology must at least differentiate between unstable, stable and metastable landscapes. The unstable landscapes are active sand dunes, badlands and unstable slopes that are the focus of field studies in process geomorphology (Campbell, 1982; Wolfe et al. 1995; Sauchyn and Lemmen, 1996). Differentiating between the stable and metastable (sensitive) soil landscapes then is the crucial function of the model. Our present understanding of the plains glacial landscape is not sufficient to define the topographic, soil and land cover thresholds that separate stability and instability, although the many studies of agricultural soil loss in this region (e.g., Pennock et al. 1995; Mermut et al. 1983; Martz and de Jong, 1991; Pennock and de Jong, 1990) provide data for the verification of predicted potential for change. The thresholds fall somewhere on the continuum between the most metastable and most stable soil landscapes. We can at least establish this continuum and place soil landscapes on it.

In lieu of a numerical value for the landscape change safety factor for each soil landscape, the various disturbance and resistance parameters are individually classified and mapped, using appropriate scales and units. Landscape sensitivity is then determined and mapped by superimposing the relevant parameter maps for fluvial, eolian and mass wasting processes. A classification of relative sensitivity emerges from the spatial covariation in barriers and disturbances, from the unstable and most metastable soil landscapes to those that exhibit the greatest resistance to change.

Return to Table of Contents


This attempt to model landscape sensitivity differs from the typical geomorphic modeling of soil loss by overland flow on individual slopes in small fluvially dissected catchments. The study area is 138,600 km² of subhumid glaciated plains with poorly integrated drainage and weakly linked slopes and channels. The basic spatial units are aggregations of landforms and soils. The objective is to identify soil landscapes where there is potential for change in the rates of fluvial, eolian and mass wasting processes. Ratio (dimensionless) data, synthetic (characteristic) landforms and the landscape change safety factor are the basis for linking slope and landscape scales and recognizing sensitive (metastable) soil landscapes.

Maps of landscape sensitivity are not reproduced here, because the product of this research is digital spatial information, rather than a specific map series or paper maps. The results will be distributed on CD-ROM, displaying various scenarios of landscape sensitivity. Of greatest interest is the sensitivity of soil landscapes to weather conditions of a certain probability (e.g., annual or 10-year rainfall) or, that is, various scenarios of climatic variability, particularly in relation to climatic change (Changnon and Changnon, 1992). Another option is to identify sensitivity to simulated changes in land use. Over most of the study area, human activities determine surface resistance to disturbance by wind and water. Thus the nature and rate of adjustment of land use and management practices are the most significant factors determining the response of agricultural landscapes to climate (Jones, 1993). However it is not the only factor. A knowledge of the physical characteristics of the soil landscapes, their geographic distribution and relation to disturbance will guide the adjustment of human activities to climatic change and variability.

Return to Table of Contents


This research is a component of the Geological Survey of Canada's Palliser Triangle Global Change Project. I gratefully acknowledgment the role of Project Coordinator D.S. Lemmen in this research. J.S. Aber, G.A. Ambrosi, M. Black, I.A. Campbell, M.J. Kirkby, D.J. Pennock, and S.A. Wolfe also provided much useful advice. Computing facilities were provided by the University of Regina.


Return to Table of Contents

Return to Geospatial Analysis Symposium

Comments to sauchyn@leroy.cc.uregina.ca