Technical Thematic Report No. 17. - Monitoring ecosystems remotely: a selection of trends measured from satellite observations of Canada
Land Cover Change
Broad land cover change 1985 to 2005
The Canada Centre for Remote Sensing (CCRS) has produced a consistent land cover time series (1985, 1990, 1995, 2000, 2005) for Canada derived from coarse (1 km resolution) satellite data from the National Oceanic and Atmospheric Administration (NOAA) – advanced high-resolution radiometer (AVHRR) sensors (Latifovic and Pouliot, 2005). This is the longest land cover dataset that covers all of Canada, providing a unique ability to consistently track long term trends.
Satellite data are assigned to land cover categories during a process called image classification (Cihlar et al., 1998). Land cover types are differentiated based on spectral signatures (patterns of absorption and reflectance of the electromagnetic spectrum) which result from pigment content, leaf structure, and plant structure (Fleishman and Mac Nally, 2007).
CCRS developed a robust methodology to produce land cover maps from 1985 to 2005 through the detection of changes to a pre-existing, widely accepted 1995 base map produced by Cihlar et al. (1999). Areas of change detected by comparison with the 1995 base map were reclassified for each year of the series, updating the 1995 map as opposed to creating new maps. This methodology maintains high consistency in land cover between maps. The methodology for the detection and reclassification of changes is described in detail in Latifovic and Pouliot (2005).
The 12-class map produced by CCRS was simplified to nine classes for the purposes of ESTR (Appendix 1). The ‘Disturbance’ class from the CCRS 12-class legend was renamed, ‘Fire ScarsFootnote7 , the three forest classes were combined into the ‘Forest’ class and the ‘Cropland’ and ‘Cropland/Woodland’ classes were combined and renamed ‘Agricultural Land’ (Appendix 1). Land cover changes were examined for six of these classes. Footnote8 Land cover and land cover change results were summarized by ecozone+ for the purposes of ESTR.
Quality checks and limitations
Errors in the data can be introduced in many stages. Errors in the 1995 base map may be carried into the change maps derived from this base map. There may also be errors in the satellite input data (AVHRR), in the change detection process, or in the re-classification of detected changes.
A visual comparison of the land cover results against higher resolution (30 m) Landsat imagery showed that the expected dynamics of disturbance followed by revegetation were captured by the land cover change analysis (Latifovic and Pouliot, 2005). This visual comparison also showed that the spatial shape and extent of large disturbances classified by the land cover change analysis differed considerably, in some cases, from the Landsat image due to the coarse resolution of the AVHRR data.
CCRS also conducted a quantitative comparison of the land cover results presented in this report with the Satellite Database for the Land Cover of Canada (SILC), a higher resolution (30 m) Landsat database which covers 9% of the Canadian land mass in 2000. In order to make comparisons between the two datasets, the SILC database was reclassified at a 1 km resolution, classifying each 1 km pixel as the dominant 30 m pixel within. Comparisons with the SILC showed that the overall accuracy of the 2000 land cover map was 61.5%, while the accuracy of reclassified areas identified in the change detection process was only 44.0% (Latifovic and Pouliot, 2005). These accuracy values for the 2000 land cover map are likely underestimates of the overall accuracy as the analysis was restricted to areas included in the SILC database, which are primarily of highly variable topography and therefore prone to classification errors (for example, mountainous areas and transition zones between broad biomes).
Extra caution should be taken when interpreting these results for ecozones+ with large mountainous areas (such as the Pacific Maritime Ecozone+), or ecozones+ with large grassland areas (such as the Prairies Ecozone+) as large fluctuations in moisture availability from year to year result in dramatically different surface appearances of grassland areas (Davidson and Wang, 2004 in Latifovic and Pouliot, 2005). Land cover changes in areas made up of several small parcels of different classes (such as the Mixedwood Plains and the Prairies ecozones+) will not be detected as changes will likely occur at a scale too small to be detected at this coarse resolution.
In summary, this dataset is most effective at detecting large scale change in land cover. The results for every ecozone+ should be corroborated with other information, particularly for smaller scale changes.
General land cover for 2005 is shown in Figure 1. During the period from 1985 to 2005, changes were primarily detected in the Forest and Fire Scars classes (Figure 2, see Appendix 2 for a breakdown by ecozone+). Fire Scars increased from 1990 to 2000 followed by a decrease from 2000 to 2005. The net change from 1985 to 2005 was an increase of approximately 200% (~146,000 km2). A net decrease in Forest area by approximately 4% (~158,000 km2) was observed during the same period (1985 to 2005). The majority of this decrease (3.5 out of 4.0%) transitioned to Fire Scars. The remainder of the net decrease in Forest area during this period is attributed to transitions to Low Vegetation and Barren (< 1%), Shrubland (< 1%), and Agricultural Land ( <0.1%). Decreases in Forest area during the time period appear to be concentrated in the southwestern Taiga Shield, and northwestern Boreal Shield (Figure 3).
Further examination of the components of the Agricultural Land class in the Boreal Plains Ecozone+ reveals a trend of transition from the ‘Cropland/Woodland’ component into the ‘Cropland’ component (Figure 4). From 1985 to 2005, 7.7% (~5,150 km2) of Cropland/Woodland was lost, most of which (~5,020 km2) was converted to Cropland. Land cover statistics for 2005, and a summary of change from 1985 to 2005 are presented in Appendix 2.
Figure 1. Broad (1 km resolution) land cover classification for Canada, 2005.
Long Description for Figure 1
This map shows the general land cover for Canada in 2005. Agricultural Land dominates the Prairies and Mixedwood Plains Ecozones+, and along the southern boundary of the Boreal Plains Ecozone+. Several larger areas of Grassland can be found in the core of the Prairies Ecozone+. Forest and Shrubland are concentrated in the mid-latitude interior and southern coastal Ecozones+. Fire Scars are scattered through these Ecozones+ but are particularly apparent in the northwestern portion of the Boreal Shield Ecozone+ and southwestern portion of the Taiga Shield Ecozone+. They are less common near the Pacific and Atlantic coasts. The northern transitional areas of the Taiga Shield Ecozone+ and the southern portion of the Arctic Ecozone+ are almost exclusively Low Vegetation & Barren (tundra), while Snow and Ice covers a high proportion of the higher latitudes of the Arctic Ecozone+.
Fire Scars increased from 1990 to 2000 followed by a decrease from 2000 to 2005. The net change from 1985 to 2005 was an increase of approximately 200% (~146,000 km2). A net decrease in Forest area by approximately 4% (~158,000 km2) was observed during the same period (1985 to 2005). The majority of this decrease (3.5 out of 4.0%) transitioned to Fire Scars. The remainder of the net decrease in Forest area during this period is attributed to transitions to Low Vegetation and Barren (< 1%), Shrubland (< 1%), and Agricultural Land (<0.1%).
Source: Derived from Latifovic and Pouliot (2005)
Figure 2. Area by land cover class for all of Canada, 1985, 1990, 1995, 2000 and 2005.
Long Description for Figure 2
This bar chart shows the following information. Area (in 1000 km2):
|Low Vegetation & Barren||2476||2481||2494||2470||2478|
Labels are the total percent change for each land cover class between 1985 and 2005.
Source: Derived from land cover statistic produced by Latifovic and Pouliot (2005)
Figure 3. Areas of transition between Forest and other classes, 1985 to 2005.
Long Description for Figure 3
This map shows the areas of transition from Forest and other land cover classes between 1985 and 2005. Most transitions occurred from areas of Forest to areas of Fire Scars. These decreases in Forest area appear to be concentrated in the southwestern Taiga Shield and northwestern Boreal Shield.
Most transitions occurred between areas of Forest and areas of Fire Scars.
Source: Derived from land cover maps produced by Latifovic and Pouliot (2005)
Figure 4. Transitions from Cropland/Woodland to Cropland in the Boreal Plains and Prairies ecozones+, 1985 to 2005.
Long Description for Figure 4
This map shows transitions from Cropland/Woodland to Cropland in the Boreal Plains and Prairie Ecozones+ between 1985 and 2005. In the Boreal Plains Ecozone+, from 1985 to 2005, 7.7% (~5,150 km2) of Cropland/Woodland was lost, most of which (~5,020 km2) was converted to Cropland. Only a small amount of Cropland/Woodland was converted to Cropland in the Prairies Ecozone+.
Source: Derived from land cover maps produced by Latifovic and Pouliot (2005)
The primary changes in land cover shown by this analysis are in the Fire Scar and Forest classes. The occurrence of fire is influenced by several factors, primarily weather/climate, fuel, topography, and human influences (Flannigan and Wotton, 2001; Flannigan et al., 2005; Parisien et al., 2006). The increase in Fire Scars to 1995 is consistent with the data presented in the Technical thematic report on large fires in Canada (Krezek-Hanes et al., 2011) which show a general increase in area burned from the 1960s to 1990, with extreme fire years in 1989, 1994, and 1995. The increase in area burned from the 1960s to 1990 has been linked the expanded use of forests by humans during this time and to warmer temperatures across the country (Podur et al., 2002; Gillett et al., 2004; Skinner et al., 2006; Girardin, 2007). For further discussion on trends in fires in Canada, including annual area burned, fire severity, and seasonality, see Krezek-Hanes et al., (2011).
The shift in the components of Agricultural Land from Cropland/Woodland to Cropland in the Boreal Plains Ecozone+ is the result of woodlots and other woodlands being converted to cropland. This finding is supported by the Wildlife Habitat Capacity on Agricultural Land in Canada National Thematic Report (Javorek and Grant, 2011) which shows that the area of land classified as All Other LandFootnote9 in the agricultural area of the Boreal Plains Ecozone+ decreased between 1986 and 2006. The Wildlife Habitat Capacity, a multi-species assessment of broad-scale trends in the potential ability of the agricultural landscape to provide suitable habitat for populations of terrestrial vertebrates, also declined significantly in the Boreal Plains Ecozone+ during this period (Javorek and Grant, 2011).
Urbanization case studies
Most of Canada’s population growth between 1996 and 2001 occurred in four main regions: the Lower Fraser Valley and southern Vancouver Island in British Columbia (Pacific Maritime Ecozone+); the Edmonton-Calgary corridor (Prairies Ecozone+); Toronto’s Golden Horseshoe region (Mixedwood Plains Ecozone+); and the Greater Montreal area (Mixedwood Plains Ecozone+) (Gurin, 2003). Population growth can be accommodated through densification of existing urban areas or through expansion into previous non-urban areas (or ‘urbanization’ as defined for the purposes of this analysis). Unlike some other types of land cover change, urbanization is most often permanent (McKinney, 2002). Urbanization can impact biodiversity in numerous ways, including:
- direct loss of habitat, including forests, wetlands, and agricultural land;
- changes to the water cycle, including excess runoff following precipitation and lack of water during droughts as a result of the impermeability of built-up areas; and
- imposing barriers to plant and wildlife dispersal.
Two of the four main regions of population growth, the Golden Horseshoe of Ontario and the Lower Fraser Valley of British Columbia, were chosen for this case study. Landsat imagery has a greater resolution than AVHRR satellite data (used to show broad land cover changes in the previous section) and has become more readily available for studies of this scale in recent years. Landsat imagery was analysed specifically for this report.
The Golden Horseshoe
The Golden Horseshoe region in southern Ontario extends around the western end of Lake Ontario from Pickering to Niagara, including the Greater Toronto Area and the cities of Burlington and Hamilton. It is part of the Lake Erie Lowland Ecoregion of the Mixedwood Plains Ecozone+ where the natural land cover is typically broadleaf and mixedwood forests (Ecological Stratification Working Group, 1995). This region contains intensive agriculture and has undergone prolonged and extensive urban development since the 1950s. It is currently the most populous and heavily urbanized region of Canada (Statistics Canada, 2012), and is also home to the majority of Ontario’s species at risk (Wilson, 2008), with birds and species that require wetland habitats making up the majority of these at risk species (David Suzuki Foundation and Ontario Nature, 2011). Ecology and conservation issues include threats to forests and plants, a need to conserve wetlands, water quantity and quality, and invasive species (Environment Canada, 2005).
Although this area has no natural limits to growth, in 2005 the Government of Ontario established a large greenbelt area (more than 7,600 km2) surrounding the Golden Horseshoe with zoning restrictions to protect agricultural and natural lands from urban sprawl (Government of Ontario, 2005; Ministry of Municipal Affairs and Housing, 2005).
The Lower Fraser Valley
The Lower Fraser Valley of British Columbia follows the Fraser River from just past Mission and Abbotsford in the east to its outlet in the Strait of Georgia. It includes the city of Vancouver and surrounding suburbs. It is part of the Lower Mainland Ecoregion of the Pacific Maritime Ecozone+. Climate is mild and very wet and natural vegetation is dominated by forests of coastal Douglas fir, western hemlock, and western red cedar (Ecological Stratification Working Group, 1995). The flat-lying Fraser delta is ideal for agriculture, and except for a number of large peat bogs, agricultural land has replaced the natural land cover throughout most of the delta (Environment Canada, 2005). As in the case of the Golden Horseshoe, the wetlands of the Lower Fraser Valley represent the most threatened and diverse ecosystems. The area once had numerous peat bogs, but most are gone or greatly reduced. The Burns Bog, near Richmond, is the largest remaining peatland (Hebda et al., 2000).
The Lower Fraser Valley has strong geographic constraints to urbanization with elevation rising steeply to the north of the Fraser River, and the United States border stopping Canadian urbanization to the south. To the east, the Fraser valley narrows rapidly. These factors restrict further expansion of urban areas primarily to the broad Fraser River delta, with the exception of North and West Vancouver, which climbs the mountains to the north.
For both the Golden Horseshoe and the Lower Fraser Valley case studies, baseline land cover maps for 1990 with land cover boundaries already visually interpreted from Landsat TM imagery were obtained from MDA Federal (formerly EarthSat). Landsat MSS images (80 m resolution) for 1974/1975 and Landsat TM images (30m resolution) for 2005/2007 were obtained from CCRS. Both the resolution and spectral separation of the 1974/1975 Landsat MSS images are lower than the 1990 and 2005/2007 Landsat TM images due to improvements in the TM sensor over the previous MSS sensor.
A false colour infrared colour composite image was created for both the 1974/1975 Landsat MSS images and the 2005/2007 Landsat TM images to improve visual interpretation. Urban area was delineated manually through a visual comparison with the 1990 map, assuming that for 1974/1975 Urban area was contained within the 1990 extent and for 2005/2007 Urban extent expanded beyond that of 1990.
Once the map revisions were completed, changes in Urban area and the resulting changes in the other land cover classes were calculated. Changes between the other land cover classes were not detected in this process (for example, conversion of Forest or shrub to Agriculture). Confidence in the results of this process is very high for changes in Urban area and medium-high for the categorization of land cover in 1975 that was converted to urban area by 1990 (land cover converted to Urban area from 1990 to 2005 was already defined in the 1990 map obtained from MDA Federal). We estimate the accuracy of the area totals to be greater than 90%.
The Golden Horseshoe, Ontario
There was a substantial increase in Urban or barren area and a corresponding loss of Agricultural land and, to a lesser extent, Forest or shrub land in the Golden Horseshoe region from 1974 to 2005 (Figure 5, Figure 6). A total of 210 km2 of Agricultural land was converted to Urban or barren land during the first period from 1974 to 1990 (a rate of 13 km2/year) and a further 305 km2 was converted from 1995 to 2005 (a rate of 20 km2/year). When the loss of forest land and wetlands is added, the total urbanization rate between 1974 and 1990 was 20 km2/year, increasing to 23 km2/year between 1990 and 2005.
The increase in Urban area during this period was concentrated in the area surrounding Toronto, with greater growth between 1990 and 2005 (Figure 6). On the south side of Lake Ontario, the there was more modest growth. Hamilton also grew more between 1990 and 2005 than between 1974 and 1990. Urban growth on Niagara Peninsula was minimal during the study period.
Figure 5. Land cover changes in Ontario’s Golden Horseshoe, 1974, 1990 and 2005. The graph shows a 60% total increase in urban or barren land.
Long Description for Figure 5
This bar chart shows the following information. Area (km2):
|Urban or barren||1111||1436||1778|
|Forest or shrub||1008||891||859|
Note: image resolution in 1974 was lower than in 1990 and 2005
Figure 6. Changes in land cover in Ontario’s Golden Horseshoe, 1974, 1990, and 2005.
Long Description for Figure 6
These three maps show the changes in land cover in Ontario’s Golden Horseshoe between 1974, 1990, and 2005. The increase in Urban area during this period was concentrated in the area surrounding Toronto, with greater growth between 1990 and 2005. On the south side of Lake Ontario, the there was more modest growth. Hamilton also grew more between 1990 and 2005 than between 1974 and 1990. Urban growth on Niagara Peninsula was minimal during the study period.
The Lower Fraser Valley, British Columbia
Urban expansion in the Lower Fraser Valley occurred primarily in former areas of Forest or shrub land (Figure 7, Figure 8). Expansion also occurred, to a lesser extent, in Agricultural land areas. A total of 87 km2 of Forest or shrub land was converted to Urban or barren land between 1975 and 1990 (a rate of 6 km2/yr) and 62 km2 of Forest or shrub land was converted to Urban area from 1990 to 2007 (a rate of 4 km2/yr). The average urbanization rate for all land cover types between 1975 and 2007 was 6 km2/year.
Growth in the Lower Fraser Valley did not follow the same pattern as in the Golden Horseshoe. The city of Vancouver itself has grown laterally as much as it can. North Vancouver expanded to the west, but is not growing substantially. There was also not much urban expansion in Richmond during the study period. Further upstream on the Fraser River, however, expansion of Urban area is evident. On the south shore, the Surrey to Delta area expanded considerably, while on the north side, Urban area filled in between Burnaby and New Westminster, as with the area between Port Moody and Port Coquitlam. Even more distant from Vancouver, the Abbotsford-Clearbrook area expanded noticeably (Figure 8).
Figure 7. Land cover changes in the Lower Fraser Valley of British Columbia, 1975, 1990 and 2007.
Long Description for Figure 7
This bar chart shows the following information. Area (km2):
|Urban or barren||498||591||680.1|
|Forest or shrub||1647||1559||1496.9|
The graph shows a 37% total increase in urban or barren land.
Note: image resolution in 1974 was lower than in 1990 and 2007
Figure 8. Changes in land cover in the Lower Fraser Valley, 1975, 1990, and 2007.
Long Description for Figure 8
These three maps show the changes in land cover in the Lower Fraser Valley of British Columbia between 1975, 1990, and 2007. The City of Vancouver itself has grown laterally as much as it can. North and West Vancouver have expanded to the west, but are not growing substantially. There was also not much urban expansion in Richmond during the study period. Further upstream on the Fraser River, however, expansion of Urban area is evident. On the south shore, the Surrey to Delta area expanded considerably, while on the north side, Urban area filled in between Burnaby and New Westminster, as with the area between Port Moody and Port Coquitlam. Even more distant from Vancouver, the Abbotsford-Clearbrook area expanded noticeably.
These two studies show an increase in Urban area since the 1970s, particularly in the Golden Horseshoe region of Ontario. Urban expansion in the Golden Horseshoe occurred primarily at the expense of Agricultural land (also found by Cheng and Lee (2008)), and to a lesser extent, forest. Expansion is most dramatically seen centered around Toronto. The Government of Ontario’s Greenbelt Plan (Ministry of Municipal Affairs and Housing, 2005) came into effect in 2005 with an aim to protect a large area (>7,600 km2) surrounding the Golden Horseshoe from further urbanization. The population growth rate of the Greater Golden Horseshoe (which, as defined by Statistics Canada, includes areas beyond this analysis, north to the Georgian Bay and east to Kawartha) between 2001 and 2006 was 8.4%. Growth in this region accounted for 84% of Ontario’s population increase and 39% of Canada’s population increase during this period (Statistics Canada, 2012). Further analysis of urbanization in this region will be important to examine how effective the Greenbelt Plan is at slowing the increase in Urban area and encouraging “smart growth” (densification as opposed to expansion, see for example, Canadian Mortgage and Housing Corporation, 2005).
Urban expansion in the Lower Fraser Valley since 1975 has taken place primarily in areas of Forest or shrub land. In the thirty years leading up to the 2001 census, the population in Vancouver and surrounding area grew almost 70%, with the majority of that growth occurring outside the core cities of Vancouver, Burnaby, and New Westminster (Gurin, 2003). Between 2001 and 2006, the population growth rate was 6.5%, greater than the Canadian average of 5.3%. This trend matches those areas that have also shown the greatest urbanization. Future urban expansion in the Lower Fraser Valley and will likely continue in the more distant communities up the valley as the city of Vancouver and the nearby suburbs have reached geographic limits to sprawl.
- Footnote 7
The Fire Scars class contains both new (<5 years) and old (>5 years, but not yet reclassified) burned areas. It may also include very large areas of harvest or mining, and severe insect defoliation.
- Footnote 8
The Urban, Snow/ice/glacier and Inland Water classes were not part of the change analysis. See Appendix 1 for further details.
- Footnote 9
As defined by the Census of Agriculture, which roughly corresponds to the Cropland/Woodland category used here
- Date Modified: