The recognition of the need for improved integration of human dimensions with biophysical dimensions in ecosystems analysis and management is not new (Brumell et al. 1990, Liu et al. 2007, Rindfuss et al. 2004). Ecosystems management has become more complex as it continues to adapt to broader questions and issues that are continuously changing and increasingly involve aspects of human-environment interaction (Szaro et al. 2005). Among many aspects that such integration entails, a core requirement pertains to the need for improved flexibility in spatial analysis and modelling of human interaction with ecosystems across multiple spatial and temporal scales. A number of themes discussed in the literature are identified as follows:

1. Coupled human-natural systems: The first is the recognition that there is a need for improved integration between human systems (socio-economic) and natural systems (ecological/biophysical) (Liu et al. 2007, McKinnon and Webber 2005, Rammel et al. 2007, Stevens et al. 2007). Such integration needs to consider how human and natural systems coevolve (Catton 1994, Stepp et al. 2003), and how new information can be obtained from an integrated analysis of their interaction that might not otherwise be determined individually within their respective domains (Christensen et al. 1996, Eddy et al. 2014).
2. Defining of custom study areas: Data and methods employed in studying particular human–environment relationships are highly contextual, depending on the spatial and temporal boundaries used to delineate a study area. There needs to be greater flexibility in working outside of pre-fixed geographical boundaries such as political, census, and other administrative-type spatial frameworks (Bachtler 2010, Bizikova and Waldick 2010, Marsden 2013, Marsden and Farioli 2015, McKinnon and Webber 2005, Morin et al. 2015, Reed et al. 2006, Stevens et al. 2007).
3. Multiscalar focus in space and time: Closely related to defining custom study areas, is the need to accommodate variables associated with human and natural systems that operate on different spatial and temporal scales (Holling 2001, Rindfuss et al. 2004, Satake et al. 2008, Stevens et al. 2007). Whereas many socio-economic analyses often focus on short-term timeframes (sometimes referred to as fast variables), there is a need to also support upscaling of the same type of data for longer term, regional-scale analyses and models that are synchronized with longer term ecosystems change (slow variables) (Holling 2004, Kangas and Kangas 2004).
4. Multiperspective modelling: Human–environment interaction is a complex system that inevitably involves having to work with incomplete data, knowledge, and related uncertainties (Hoogstra and Schanz 2008, Meitner et al. 2009). There is often no one correct answer to a particular question, but there are multiple ways of exploring questions depending on the data, methods, and models available. This is best achieved by providing the flexibility needed to model scenarios that explore different assumptions and model criteria.
5. Multiuser orientation: The use of scientific information in evidence-based and consensus-based decision making is also an important factor. Scientific results need to be presented in terms that laypersons can readily understand (Eddy et al. 2014, Meitner et al. 2009), as well as make explicit all uncertainties and assumptions associated with any particular scenario or perspective (Hardi and Zdan 1997, Hoogstra and Schanz 2008).

Taking all of these requirements into account, the manner in which both human and ecosystems data are represented geographically is fundamental to any scientific analysis of human-environment relationships. However, a fundamental barrier to addressing these requirements is the incompatibility between spatial data frameworks used for georeferencing socio-economic and ecosystems data. Ecosystems data are commonly portrayed using natural boundaries and complex spatial patterns, whereas socio-economic data commonly use more Euclidean-type administrative boundaries. When using GIS overlay processes to integrate data from these different spatial frameworks, a number of problems are encountered. Most fundamental are problems associated with scale mismatches and boundary alignment (Cumming et al. 2006, Huck et al. 2015), the most serious of which constrains our ability to spatially synchronize quantitative data variables for more advanced spatial analyses. The importance of this problem cannot be understated; socio-economic data systematically collected in many countries is referenced to administrative-type spatial frameworks that do not align with natural boundaries. Here, we examine the problem in more detail as experienced in Canada.

In Canada, the challenge in integrating socio-economic and ecosystems data is particularly challenging in this regard. Canadian census data is one of the most commonly used sources of socio-economic data, particularly at the level of census subdivisions (CSDs) which most closely serve as proxies for individual communities. However, CSD boundaries are delineated by geometric features derived from road networks and other administrative boundaries (e.g., municipal, county, or political jurisdictions), and therefore rarely follow natural features. For example, Fig. 1 shows a region in southern Manitoba in central Canada. It illustrates how an analysis that requires a spatial overlay of a socio-economic variable (e.g. population) with an ecosystems variable (e.g. forest cover) is confronted with a challenging overlay problem. If the aim is to estimate the human population associated with forested areas, one might be forced to split data values in individual census units to relate proportionally to corresponding areas of the forest pattern. Doing so not only risks falsifying data, but many other problems are encountered with using administrative frameworks in this manner.

Figure 1.  

Illustration of a spatial overlay problem between socio-economic and ecological data using census boundaries and forested areas. Example study area is located in southern Manitoba, Canada.

First, in the case of Canadian census data, CSDs are delineated on the basis of how municipalities are defined in each province, which results in a lack of normalization across provinces. As illustrated in Fig. 2, some provinces define municipalities at a fine scale for individual small towns and hamlets (e.g., Quebec, Saskatchewan), whereas other provinces define municipalities as regional clusters of communities (e.g., Ontario, Nova Scotia). Second, the geometries of CSD boundaries also vary from province to province in relation to the different frameworks used for delineating municipal boundaries. Some provinces use complex (irregular) boundary types that follow natural features or regions (e.g., Newfoundland and Labrador, British Columbia), whereas other provinces use more simplified (straight) administrative boundary geometries (e.g., Alberta, New Brunswick). The result is that municipalities are not only defined differently, but the shape and size of CSDs vary widely.

Figure 2.  

Map showing the distribution of Census Sub-divisions (CSDs) for all of Canada. Note the differences in the shape and size of boundaries among different provinces.

A third problem, of no lesser importance, relates to the identification and names of individual communities. The CSD framework is designed and structured primarily for the purpose of dissemination of census data to municipalities, and therefore the names of many CSDs do not always correspond with the names of communities where people live. This is an important issue in terms of considering sense of place and cultural identity (Sack 2003, Vodden et al. 2015), which is often the case for smaller communities that are grouped into larger regions. Data values for smaller communities are sometimes diluted in higher level aggregates at the CSD level. As a result, people who live in smaller communities do not find the higher level aggregated data meaningful because the data may incorporate trends happening in adjacent communities that contrast sharply with their own community (e.g. demographic, employment or labour force statistics). 

Such limitations do not negate the important value that census data have to offer. The limitations identified here are associated more with the geospatial representation of communities than the quality of the data collected. Ideally, what is needed is a spatial representation of communities in which: 1) residents identify with on a local level, and 2) are delineated by boundaries that can be more naturally aligned and integrated with ecosystems data. A standardized approach would improve the normalization across provinces.

To address some of these problems, one method developed for this approach uses the concept of human habitats, wherein local community boundaries are mapped using natural boundaries delineated by the physical footprints, as opposed to using census or administrative boundaries (Eddy and Dort 2011). In that study, key socio-economic attributes that characterize each populated area as a human habitat are re-assigned to each community, and include physical size, population growth, physical development, demographic changes, education, employment, and income (ibid). When mapped collectively, the distribution of communities resembles a natural habitat map for the human species akin to habitat mapping for other animal species, and for which habitat characteristics can then be analysed. Building on this approach, the following section describes the data integration methodology used to map the whole ecumene of Canada in this manner.

Forest communities are an integral component of forest management in Canada, and in the development of Canada’s forest sector. Given that the forest industry is one of the most geographically distributed economic sectors in the country, it should be no surprise that forest communities made a significant contribution to the development of Canada’s human settlement pattern over the past 200 years. The close relationship Canadians have with the forest has resulted in complex patterns of human activity within and adjacent to Canada’s forest ecosystems. The interaction can have both positive and negative effects to varying degrees, which makes consideration of multiple human factors critically important in forest science, policy, and management. Identifying and estimating the population of forest communities in Canada is therefore an important scientific policy input, and requires an unambiguous definition of what constitutes a ‘forest community’. Criteria used may vary depending on the application. Factors such as economic dependency, population characteristics, and physical proximity in relation to forest areas are key considerations. The approach taken here first begins by considering all communities in Canada (the whole ecumene) to which different criteria may be applied depending on the definition used. For comparison, whereas the map in Fig. 4 shows the whole ecumene of Canada, the map in Fig. 5 shows one rendering of the forest ecumene by where populated places, transportation and utility lines intersect forest-dominated ecozones. Quantitative estimates, however, of the number of communities and corresponding population can vary widely depending on the definition of forest community used. 

Figure 5.  

Map of Canada's "Forest Ecumene", as a sub-set of the whole ecumene of Canada (see Fig. 4). In this rendering, the forest ecumene is shown as the portions of the whole ecumene of Canada that intersect with forest dominated ecosystem zones.

In our analysis, we examine three definitions combined with two ways of representing forested areas. First, a ‘forest zones’ layer was created by combining physical forest areas with major ecological zones. The forested areas layer was derived from a 250 m MODIS kNN land cover analysis (Beaudoin et al. 2014) by reclassifying each land cover type into a binary raster of ‘forest’ and ‘non-forest’. This layer was then combined with a derivative of the national ecological zones layer (CCEA 2014) which was classified into ‘Forest-dominated’ and ‘Non-forest dominated’ ecological zones. The resulting forest zones map contains four classes: 1 – Non-forest/Non-forest Ecozones (NF/NFE), 2 – Forested/Non-forest Ecozones (F/NFE), 3 – Non-forest/Forest Dominated Ecozones (NF/FDE), and 4 – Forested/Forest Dominated Ecozones (F/FDE). By using a zonal analysis overlay method in ArcGIS, each ecumene polygon was assigned the highest intersecting forest zone class value. Population values were then aggregated for each forest zone category for each census year. The map in Fig. 6 shows examples of communities that correspond to each definition, and technical details on the overlay GIS process used to generate these results are provided in Supplement A (Suppl. material 1). Table 2 presents resulting population estimates and percentages for each forest zone (2a, 2b), and additional aggregation estimates using three definitions (D1, D2, and D3) of a forest community (2c, 2d).

Figure 6.  

Map showing the spatial relationships of communities to forest areas and forest-dominated ecosystem zones for different definitions of a forest community. See text for elaboration.

Each definition in Table 2 is successively more inclusive of communities in terms of the spatial relationship to forests and forest-dominated ecosystems. The first definition (D1) is limited to communities that are located exclusively within forested areas and within forest-dominated ecozones (location A in Fig. 6). The second definition (D2) includes all communities in D1 plus communities outside of the forest-dominated ecosystems but located either within or are in close proximity to forested areas (location B in Fig. 6). There are also many communities within forest-dominated ecozones that are not located within or in close proximity to forested areas (location C in Fig. 6). Such communities may be considered for inclusion in the forest ecumene by virtue of being surrounded by forest-dominated ecosystems (D3), with the rationale that they may be subject to hazards from forest areas (e.g., smoke from wildfire) or have greater economic, social, or cultural dependence than similar nonforest communities in nonforest-dominated ecozones. The only communities that are not considered to be forest communities are those that are not located within or in close proximity to forest areas, nor within forest-dominated ecosystems (location D in Fig. 6).

In terms of the results of this analysis, for comparison, Natural Resources Canada used CSD level data to estimate the total population of forest communities to be approximately 11 million people, or approximately 31% of Canada’s total population for 2016 (NRCan 2019). By contrast, quantitative differences shown in Table 2c in terms of both population estimates and number of communities illustrate the sensitivity in choice of definition and the analysis method used. In our analysis, population estimates range from approximately 8.2 million (D1) to nearly 19.5 million (D3), or a range of 23% to 55% of Canada’s total population for 2016. The total number of forest communities range from around 2500 under D1 (60%) to as many as over 3400 (or 80%) communities under the D3 definition.

A number of trends and relationships among community categories can be observed from this analysis, starting with the relative increase in the nonforest population (Forest Zone 1 – NF/NFE) (Table 2b). A continuation of the rural–urban population shift is evident in comparing changes in these values with the values of forest communities under either of the D1-D3 definitions, particularly in terms of percentages of total population (Table 2d), and in considering that many of the major urban areas do not qualify as forest communities. A trend worth noting is how the population of the forest ecumene has also increased during the same time period, despite the economic downturn that has affected many forest communities over the past twenty years. It illustrates that although the rural–urban population shift continues, the population of forest communities also continues to increase, but at a comparatively slower rate. The overall 20 year % change for the Forest Zone categories 2, 3, and 4 in Table 2b range from 9% to 25% increase, in comparison to a 32% increase for the strictly non-forest (and urban dominant) communities.

Information extracted from this type of analysis provides an important input for integrating regional economic development, and other socio-economic considerations in ecosystems management and planning. Whereas the use of spatial overlay analysis in mapping the forest ecumene and deriving quantitative measures on population captures the extensiveness of human interaction with forest ecosystems of Canada, it is also possible to use ecumene data to model spatially intensive socio-economic patterns that relate to the proximity of ecosystems resources and services. To demonstrate such an approach, our second example application focuses on mapping labour force distribution as an indicator of economic dependency on the forest industry within the forest ecumene.

Mapping labour force data can be useful for identifying regions that are economically reliant on natural resources by virtue of their physical proximity to the resource base. It is one means of including an economic dimension in ecosystems management, and by extension, associated social, demographic and cultural factors. As is the case with population data described above, in Canada, labour force data and other social, demographic and economic variables are also collected and aggregated according to census and administrative boundaries. In the Canadian forest sector, for example, economic dependency of forest communities has been a topic of interest for many years (Korber et al. 1998, Parkins et al. 2001, Stedman et al. 2004, Stedman et al. 2005, Stedman et al. 2012, Stedman et al. 2007). In such studies, economic dependency measures use labour force data linked with CSDs to identify and rank forest-dependent communities (FDCs). FDCs are measured on the basis of a proportional contribution of the forest industry to the overall structure of a local economy. The outputs of such studies are tabular lists of communities with corresponding dependency values. This approach has been useful for general tracking purposes; however, there are limitations with mapping such data using CSD boundaries for ecological applications. 

In cartography, mapping data variables using homogeneous units, such as census or other administrative boundaries, is known as choropleth mapping (Slocum et al. 2005). Choropleth maps are intended for mapping numerical or categorical data only in cases where it is known that the value of the data being mapped is homogeneous throughout the geographical area covered by the choropleth unit, and when the size and shape of the units are similar (ibid., 286). The lack of similarity in size and shape of CSDs in Canada is therefore a challenge in this regard (see Fig. 2). Mapping ratio or proportional data values is not considered to be cartographically valid unless it is known for certain that the data values are relatively similar for all communities within an administrative unit. In spite of these limitations, choropleth mapping has been a common practice due to the convenience in linking data to the administrative or census units in which the data was collected and/or distributed.

Using an ecumene as an alternative spatial framework addresses some of these limitations. For comparison, Fig. 7 and Fig. 8 show maps of the labour force distribution for the forestry sector using the CSD boundaries and ecumene, respectively for the year 2006. The method used for the CSD map is straightforward choropleth mapping. Labour force ratios are calculated for each CSD, then mapped using a standardized legend showing a range of threshold intervals with a monochromatic colour ramp. Custom tabulations of the labour force data were provided by StatsCan (2018). As an alternative to the approach used in the calculation of FDCs, which is specific for the forest sector only, the approach taken here uses a standardized calculation of 'Labour Force Distribution (LFD)' which can be applied for the forestry sector or any other sector or sub-sector for direct comparison. In this approach, labour force ratio values are calculated as the total income for the forestry sector as a proportion of the total income for all base economic sectors, where the base sectors include the natural resource, utilities and construction, and manufacturing sectors.

Figure 7.  

Choropleth map showing labour force distribution for the forestry sector using census sub-division (CSD) boundaries for the year 2006. Locations of forest product mills and the northern limit of the managed forest area is shown for additional reference. Compare with Fig. 8.

Figure 8.  

A spatial interpolation map showing the labour force distribution for the forestry sector using the ecumene framework for the year 2006. Locations of forest product mills and the northern limit of the managed forest area are shown for reference. Compare with Fig. 7.

The resulting map in Fig. 7 shows a wide variation of class values associated with the different sizes and shapes of the CSD boundaries. As an area-based representation of the data, it is prone to misinterpretation as the larger CSDs in the northern areas have lower population values and smaller communities, whereas the smaller CSDs in the southern areas have higher numbers of communities with higher population values. The resulting map pattern gives the impression that very large areas of northern Canada are highly dependent on the forest sector, yet in reality, there are very few communities in these large areas. There is also no clear spatial relationship between the labour force distribution within the limit of the managed forest area and the location of mill facilities.

By comparison, the same data mapped using the ecumene framework in Fig. 8 shows labour force distribution as a more realistic geographical representation. The same labour force data aggregated to forest ecumene communities (as opposed to CSDs) also provides a means for normalization. In this approach, labour force ratios are first calculated for individual ecumene communities, from which a spatial interpolation is applied to map clusters of forest dependency on a regional level. Technical details on the method used for this map are provided in Supplement B (Suppl. material 2). In terms of the result, one pattern worth noting is how the higher dependency values tend to cluster in some regions more than others, notably the southern and central interior of BC, the north-central areas of the prairie provinces, northern Ontario, through Quebec, northern New Brunswick, and western-central Newfoundland. The higher values in these areas are not surprising given these regions follow the southern edge of Canada’s forested areas. Although these regions are already well known to be highly dependent on the forest industry, this map makes this knowledge spatially explicit. It reveals relative economic importance of the forest industry in many regions of Canada, and how much regional variation there is in the labour force distribution. An earlier version of this map was used in a national integrated assessment of potential impacts of climate change on Canada’s forest sector (NRCan 2016, Ste-Marie et al. 2015) where it is observed that regions with the highest level of forest industry dependency correspond with transitional areas along the southern limit of the boreal forest that are potentially vulnerable to the impacts of climate change (Boulanger et al. 2016).

It is also worth noting that because the ecumene allows for spatial normalization (among provinces) and temporal normalization (across census periods), a series of maps may be generated for each census period, or combined for time-series analysis. In addition, because the formula used is a simple calculation of the labour force income for the forestry sector as a proportion of the total base economic sectors, the same method can be used as a standardized approach for mapping other natural resource sectors such as fisheries, agriculture, mining, and petroleum industries. A series of these maps are provided in Supplement C (Suppl. material 3) for illustration.