0% found this document useful (0 votes)
65 views19 pages

Leclerc 2017

This study investigates land cover change near the abandoned Pine Point Mine in Canada's Northwest Territories using remote sensing analysis and landscape metrics. Landsat imagery from 1988-2008 was classified into 7 land cover classes to track changes over 20 years since mine closure. The mine site showed little vegetation regrowth while nearby Wood Buffalo National Park had larger, more contiguous forest patches. While remote sensing allowed broad changes to be monitored, higher resolution data is needed to effectively monitor land cover dynamics at abandoned mines.

Uploaded by

tonmoy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
65 views19 pages

Leclerc 2017

This study investigates land cover change near the abandoned Pine Point Mine in Canada's Northwest Territories using remote sensing analysis and landscape metrics. Landsat imagery from 1988-2008 was classified into 7 land cover classes to track changes over 20 years since mine closure. The mine site showed little vegetation regrowth while nearby Wood Buffalo National Park had larger, more contiguous forest patches. While remote sensing allowed broad changes to be monitored, higher resolution data is needed to effectively monitor land cover dynamics at abandoned mines.

Uploaded by

tonmoy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 19

Environ Monit Assess (2017) 189:185

DOI 10.1007/s10661-017-5893-7

Assessing post-industrial land cover change at the Pine Point


Mine, NWT, Canada using multi-temporal Landsat analysis
and landscape metrics
Emma LeClerc & Yolanda F. Wiersma

Received: 9 October 2016 / Accepted: 14 March 2017


# Springer International Publishing Switzerland 2017

Abstract This study investigates land cover change industrial mineral development. We found that the dense
near the abandoned Pine Point Mine in Canada’s North- conifer class was dominant in the park and exhibited
west Territories. Industrial mineral development trans- larger and more contiguous patches than at the mine site.
forms local environments, and the effects of such dis- Bare land at the mine site showed little conversion
turbances are often long-lasting, particularly in subarc- through time. While the combination of raster algebra
tic, boreal environments where vegetation conversion and landscape metrics allowed us to track broad changes
can take decades. Located in the Boreal Plains Ecozone, in land cover composition and configuration, improved
the Pine Point Mine was an extensive open pit operation access to affordable, high-resolution imagery is neces-
that underwent little reclamation when it shut down in sary to effectively monitor land cover dynamics at aban-
1988. We apply remote sensing and landscape ecology doned mines.
methods to quantify land cover change in the 20 years
following the mine’s closure. Using a time series of Keywords Land cover change . Abandoned mines .
near-anniversary Landsat images, we performed a su- Remote sensing . Landscape metrics
pervised classification to differentiate seven land cover
classes. We used raster algebra and landscape metrics to
track changes in land cover composition and configura-
tion in the 20 years since the mine shut down. We Introduction
compared our results with a site in Wood Buffalo Na-
tional Park that was never subjected to extensive anthro- Canada’s boreal forest is increasingly recognized as a
pogenic disturbance. This space-for-time substitution conservation priority (Carlson et al. 2010; Valeria et al.
provided an analog for how the ecosystem in the Pine 2012; Badiou et al. 2013; Powers et al. 2013). Powers
Point region might have developed in the absence of et al. (2013) argue that large reserves of intact boreal
forest are important for maintaining biodiversity and
long-term ecological processes. Additionally, the Inter-
national Boreal Conservation Science Panel (IBCSP)
formally recognized the boreal forest as an ecologically
E. LeClerc (*)
Department of Geography, Memorial University of significant environment, citing in particular its role as
Newfoundland, St. John’s, NL, Canada habitat for endangered and threatened species and its
e-mail: [email protected] importance for mitigating climate change through car-
bon sequestration. A report from the panel recommends
Y. F. Wiersma
Department of Biology, Memorial University of Newfoundland, the conservation of a minimum of 50% of Canada’s
St. John’s, NL, Canada intact boreal forest (Badiou et al. 2013). However,
185 Page 2 of 19 Environ Monit Assess (2017) 189:185

identifying Bintact boreal forest^ for the purposes of monitor this change after closure. A mine site is consid-
conservation and management is complicated by the ered fully reclaimed when it has been stabilized,
widespread presence of historical and ongoing anthro- revegetated, returned to ecological productivity, and no
pogenic disturbances such as industrial resource extrac- longer requires monitoring (CCSG Associates 2001).
tion. This problem is compounded by the dynamic Reclamation can be accomplished through active inter-
nature of forest ecosystems and the challenge of defin- vention or, in the case of many historical mines, by
ing baselines to establish conservation targets (Coppin passively waiting for the ecosystem to reestablish eco-
et al. 2004; Alagona et al. 2012). logical productivity (Hodacova and Prach 2003; Worrall
The boreal forest has supported an array of socioeco- et al. 2009). For remote mines, like many of those in
nomic activities through time. As with natural distur- northern Canada, monitoring reclamation progress in
bance regimes, these activities have varied in frequency situ is costly and usually undertaken by mining compa-
and magnitude, leaving legacies that can often be traced nies or government over a short-term basis (MMSD
on the land. Aboriginal1 people have relied on the boreal 2002). At large and/or remote mine sites with extensive
ecosystem to support subsistence activities and the fur landscape change and loss of vegetation, remote sensing
trade, both historically and in contemporary times is a more cost-effective platform for monitoring
(Fumoleau 1974; Ray 1990; Usher et al. 2003; Abel landscape-level reclamation over long periods of time.
2005; Natcher 2009; LeClerc and Keeling 2015). In the Remotely sensed data are commonly used to monitor
twentieth century, resource development activities in- land cover change at mine sites (Antwi et al. 2008;
cluding timber harvesting and industrial mineral extrac- Townsend et al. 2009; Li et al. 2012; Li et al. 2015);
tion spread throughout northern Canada’s boreal eco- Schmidt and Glaesser (1998) and Gillanders et al.
system (Boucher et al. 2009; Keeling and Sandlos 2009; (2008) indicate that remote sensing provides an impor-
Piper 2009). Technological advances made it possible to tant means of monitoring land cover change at mine
mine lower grade ores at massive scales, generating sites throughout the process of reclamation.
more waste and higher potential for environmental dam- The combination of remote sensing and landscape
age (Cooke and Johnson 2002; Lecain 2009). When metrics has immense potential for communities interest-
economic benefits at a particular site dwindled, mining ed in monitoring landscape changes in the wake of
operations ceased and environmental damage was often industrial development. Because many abandoned
not addressed (Cooke and Johnson 2002; Worrall et al. mines leave large footprints and are located in places
2009). Prior to the introduction of more stringent regu- that are difficult to access, remote sensing can be a
lations and environmental protection mandates (e.g., powerful tool for collecting time series data on land
Northern Inland Waters Act 1972, Guidelines for the cover post-closure. Landscape metrics quantify patterns
Abandonment and Restoration for Mines in the North- in land cover composition and configuration that can
west Territories 1990) which required companies to yield insight to the ecological processes that shape post-
articulate reclamation plans from the inception of mine industrial landscapes. For instance, land cover patches
development, the deleterious effects of historical indus- tend to be less complex in places that have hosted
trial resource extraction, particularly from mining, were anthropogenic disturbances (Narumalani et al. 2004).
frequently overlooked (Bowman and Baker 1998; In theory, disturbed sites become more similar in struc-
MMSD 2002; Wenig et al. 2005; Otto 2009; Worrall ture and function to the surrounding environment
et al. 2009). The long-term environmental legacies of through time as the reclamation process, whether active
abandoned mines have been identified as a key chal- or passive, progresses (Bradshaw 2000). In this study,
lenge for an industry seeking to become more sustain- we assess the application of remote sensing and land-
able (Worrall et al. 2009). scape metrics as a means of monitoring land cover
While mineral development has created extensive change at abandoned mines.
and persistent landscape change in northern Canada The Pine Point Mine is one example of an industrial
(Piper 2009; Keeling 2010), it is often difficult to extraction operation whose abandonment pre-dated
stricter reclamation policy. Although Pine Point is located
1 in the Boreal Plains Ecozone in the Northwest Territories
We use the term BAboriginal^ to collectively refer to First Nations
and Métis people, in accordance with the current terminology used in (ESWG 1995), the development, operation, and closure
Canada (Desbiens and Rivard 2014) of the mine occurred before the boreal environment’s
Environ Monit Assess (2017) 189:185 Page 3 of 19 185

ecological and socioeconomic importance was formally ecological data are lacking (Pickett 1989; Fukami and
acknowledged by policy-makers. Indeed, operations to Wardle 2005; Lecomte et al. 2005; Bond-Lamberty and
extract lead and zinc ore began in 1964, and the first laws Gower 2008). We use remote sensing and landscape
governing habitat protection and water use in the area metrics to compare the post-industrial mine site at Pine
were enacted in the 1970s (Boyd 2003). Bowman and Point to a site in nearby Wood Buffalo National Park.
Baker (1998) point out that Bthere is no single piece of The space-for-time substitution allowed us to compare
legislation that directly addresses the reclamation of mine the composition and configuration of land cover at the
sites and surrounding impacted areas^ (p. 8) in the mine site with a reference landscape ostensibly unaffect-
Northwest Territories; instead, reclamation is guided by ed by industrial development. This space-for-time sub-
discretionary power established through regulations of stitution serves as an analog for how land cover in the
land leases and water use (Bowman and Baker 1998; Pine Point region might have changed under natural
Wenig et al. 2005). The first Reclamation and Abandon- conditions had the mine never been developed. We
ment Plan for the Pine Point Mine was drafted in 1985 in hypothesized that the Pine Point site would become
fulfillment of the requirements of the mine’s water li- more similar to the park site in terms of composition
cense, which was governed by the Northwest Territories and configuration through time as a result of passive
Water Board established by the Northern Inland Waters reclamation.
Act in 1972. Two years after milling operations ceased in The analysis of land cover change in the Pine Point
1988, the Northwest Territories Water Board and the region offers insight to the long-term effects of industrial
Department of Indian Affairs and Northern Development development on landscape patterns in a boreal environ-
released their Guidelines for Abandonment and Restora- ment. Using a time series of near-anniversary remotely
tion Planning for Mines in the Northwest Territories sensed images spanning 1989 to 2009, we classify land
(Bowman and Baker 1998). cover types and compute a series of landscape metrics at
Upon closure, the Pine Point Mine site was reclaimed the Pine Point site and comparison site. The comparative
to the standards of the time, which prioritized site stabi- nature of this study elucidates how ecosystem dynamics
lization and pollutant containment (Bowman and Baker operate at a subarctic site subjected to extensive anthro-
1998; MMSD 2002; Otto 2009), but left an extensive pogenic disturbance and whether such dynamics deviate
post-industrial landscape approximately 1900 km2 in from those at work in a Bnatural^ context (Fukami and
area, dominated by 46 open pits, piles of waste rock, Wardle 2005; Soverel et al. 2010). Given the growing
haul roads, cutlines, and a 570-ha tailings pond (Sandlos call for boreal conservation in Canada and the efforts of
and Keeling 2012). The degraded environmental state of the mining industry to become more sustainable, it is
this post-industrial site is keenly felt by land users from necessary to explore the utility and limitations of using
the nearby community of Fort Resolution (Sandlos and remotely sensed data and landscape metrics to monitor
Keeling 2012; LeClerc and Keeling 2015). The ecolog- salient effects of industrial mineral extraction in a sub-
ical impacts of industrial anthropogenic disturbance are arctic, boreal environment.
particularly pronounced in the subarctic, where it can
take decades for altered landscapes to reestablish former
composition and functionality (Bowman and Baker Materials and methods
1998; Lee and Boutin 2006; Deshaies et al. 2009). For
instance, while some cutlines which were cleared during Study area
Pine Point’s exploration phase may have begun the
process of vegetation succession, others continue to The Pine Point Mine site is located on the southern shore
crisscross the boreal forest decades after closure. of Great Slave Lake in the Northwest Territories
At historical mine sites, it is important to understand (60.8336° N, 114.4545° W), just north of the boundary
how the lack of reclamation has affected the local envi- of Wood Buffalo National Park in the Boreal Plains
ronment through time. While remote sensing can pro- Ecozone (Fig. 1). The average temperature in the region
vide a time series dataset, its limited historical depth is ranges from −22 °C in January to 16 °C in July, and
sometimes insufficient to establish pre-industrial base- annual precipitation is approximately 336 mm
lines. Space-for-time substitutions have been used as a (Environment Canada 2010). The study site’s elevation
strategy to assess vegetation change when long-term ranges from approximately 150 m to approximately
185 Page 4 of 19 Environ Monit Assess (2017) 189:185

Fig. 1 Map of the study region showing the proximity of the Pine the northern boundary of Wood Buffalo National Park, and the
Point Mine to Wood Buffalo National Park and Great Slave Lake. dark gray lines represent waterways. Map of northern North
The star indicates the town site that was abandoned when the mine America showing the location of Pine Point (inset). (Map by
shut down in 1988. The thin black lines show the roads, including Emma LeClerc)
the network of haul roads and the highways. The thick black line is

300 m, making topographic variability negligible for the (Kendall 1992). The town of the same name, built
purposes of this study. The dominant tree species in the expressly to serve the mine, was home to approximately
area are black spruce (Picea mariana) and jack pine 2000 residents at its height (Macpherson 1978; Sandlos
(Pinus banksiana), and the understory is characterized and Keeling 2012; Sandlos 2015). When the mine shut
by shrubby cinquefoil (Potentilla fruticosa), reindeer down, 46 pits were left open, the town site was cleared
lichen (Cladonia spp.), and feather moss (Calliergon of buildings, and berms or barricades were built to block
spp.). Poplar (Populus spp.), white birch (Betula access to the site (Sandlos and Keeling 2012). A pilot
papyrifera), tamarack (Larix laricina), and alder (Alnus study was undertaken to investigate the viability of
spp.) are also common in the region. Fire is the domi- stabilizing the 570-ha tailings pond using vegetation;
nant natural disturbance regime in the area, although the however, the project was not successful. Nutrient and
anthropogenic effects of timber harvesting and mining moisture deficiencies as well as high concentrations of
in the boreal environment are becoming a disturbance salt, lead, and zinc were all identified as factors that
regime in their own right. limited vegetation germination and growth. The sparse
Pine Point was an extensive open pit mine that began vegetation that did establish on the tailings was found to
operations in 1964 and was abandoned in 19882 when have accumulated lead and zinc at levels considered
lead and zinc prices dropped below a profitable level hazardous for foraging wildlife. As a result, attempts at
revegetation were scrapped (Gardiner 1990).
2
Although closure and abandonment activities continued into the early The reclamation efforts at Pine Point were deemed
1990s, Pine Point halted milling operations in 1988 (Cominco Ltd. 1991). adequate by governing bodies at the time. For example,
Environ Monit Assess (2017) 189:185 Page 5 of 19 185

the Chairman of the Northwest Territories Water Board, represents how the land around Pine Point might have
the regulatory body responsible for licensing industrial been used and changed in the absence of the mine.
water use under the Northern Inland Waters Act,
commended Cominco Ltd. on its reclamation efforts Data and processing
(Nickerson 1992). The mine was later featured as an
exemplar of successful closure by Environment Canada We used Landsat 5 Thematic Mapper (TM) data from
(1996). However, the reclamation efforts, which focused 1989, 1994, 2005, and 2009 to document land cover
primarily on physical hazards and which did not success- change since the Pine Point Mine shut down. Landsat
fully revegetate the mine site, would be considered unsat- TM images have a spatial resolution of 30 m and are
isfactory by today’s standards. Since the closure and terrain-corrected (level 1T) products. To ensure consisten-
decommissioning of the Pine Point Mine, environmental cy across the time series, the co-registered images were
policy and regulation in Canada have evolved to require verified visually prior to analysis. Regular, repeat coverage
somewhat more comprehensive closure procedures. How- provides Landsat images that are archived; the US Geo-
ever, most of these regulations center on preventing pollu- logical Survey (USGS) has made data in this archive freely
tion at current operations and fail to address the long-term available since 2009 (Woodcock et al. 2008). This makes
environmental consequences of historical mines, including Landsat products an affordable means for monitoring
the Pine Point Mine (MMSD 2002; Bridge 2004; Wenig broad-scale land cover change and forest dynamics, espe-
et al. 2005). cially in remote or inaccessible regions (Valeria et al. 2012;
In order to assess the long-term environmental effects of Ahmed et al. 2013; Czerwinski et al. 2014). A visual
the Pine Point Mine, we compared land cover change at assessment of the data showed that major mining features
the mine with land cover change at a comparison site (e.g., pits, roads) were visible; however, finer resolution
located in Wood Buffalo National Park. The absence of features (e.g., cutlines) were sometimes obscured. This
industrial resource extraction and the close proximity to the limitation notwithstanding, the temporal depth offered by
mine site make Wood Buffalo National Park an ideal the Landsat archive made it the most appropriate dataset
location for the space-for-time substitution. We apply re- for this analysis.
mote sensing and landscape metrics to determine whether The study region is located in Worldwide Reference
land cover change near Pine Point differs from land cover System (WRS) path 45, row 18 of Landsat 5 TM. For the
change in an ecologically comparable area that never study period, there were four images with minimal cloud
hosted industrial resource development. Wood Buffalo cover in the Landsat archive (Table 1). The selected images
National Park is a UNESCO World Heritage Site located were all recorded in the late summer (within the May–
in northern Alberta and the southern part of the Northwest October growing season; Larsen (1980)) in order to min-
Territories. Established in 1922 to protect wood bison imize phenological variation and thus maximize the spec-
(Bison bison athabascae), today the park’s mandate has tral stability of land cover types, as season is an important
expanded to protect whooping crane (Grus americana) factor in the study of vegetation change detection (Miller
and their nesting areas, within Parks Canada’s broader goal et al. 1997; Beaubien et al. 1999; Coppin et al. 2004;
of preserving natural and cultural heritage. At Gillanders et al. 2008; Fichera et al. 2012; Ahmed et al.
44,807 km2—just larger than Denmark—Wood Buffalo 2013). For this analysis, we chose a 4-5-3 band combina-
National Park is Canada’s largest national park (Parks tion, representing the near infrared (NIR), mid-infrared
Canada 2010). The 1964 National Parks Policy stated that (MIR), and red (R) bands of the electromagnetic spectrum.
industrial land uses (e.g., mine development, timber ex- Beaubien et al. (1999) and Valeria et al. (2012) recommend
traction3) are no longer permitted in national parks, and
Aboriginal hunting and trapping rights are protected in
accordance with treaties 8 and 11 (Boyd 2003; Sandlos Table 1 Landsat 5 TM
datasets, from USGS Path/row Acquisition date
2007; Parks Canada 2010; Will 2015). Thus, the park GloVis
45/18 10 July 1989

3
45/18 24 July 1994
While timber extraction continued in Wood Buffalo National Park
until the early 1990s, the logging occurred in the Peace River Low- 45/18 23 August 2005
lands, well south of the comparison site delineated in this study 45/18 3 September 2009
(Timoney and Peterson 1996; Boyd 2003)
185 Page 6 of 19 Environ Monit Assess (2017) 189:185

this band combination for research in boreal environments developed a Global Human Footprint based on the un-
with many cover types. Gratto-Trevor (1996) indicates the derstanding that anthropogenic land uses have differen-
utility of these bands for differentiating vegetation types, tial impacts on the surrounding ecosystem depending on
boundaries between land and water, and moisture content the type of use. Woolmer et al. (2008) found that adapting
in soil and vegetation, making the bands particularly suit- the Global Human Footprint to account for locally avail-
able for the characterization of land cover in the Pine Point able, finer resolution geospatial data improved the utility
region. of the Human Footprint for regional conservation plan-
Radiometric correction of the images was performed ning. Similarly, Wiersma et al. (2004) employed a grad-
with the ATCOR model for absolute radiometric correction uated buffer protocol to approximate the ecological ef-
in PCI Geomatica 2013 (PCI Geomatics, Richmond Hill, fects of anthropogenic infrastructure, and Wiersma and
ON). The correction established spectral consistency Simonson (2010) showed that variation in buffers did not
across the images by calibrating for sensor settings, atmo- significantly change the interpretation of ecological ef-
spheric and illumination conditions, which served to min- fects. Guided by the human influence scoring framework
imize errors during the classification and comparison pro- detailed in Woolmer et al. (2008), we placed a 5-km
cesses. This ensures that the change detected in land cover buffer around the open pits, a 1-km buffer around the
over time corresponds meaningfully to change in the spec- major roads, and a 500-m buffer around the cutlines, the
tral response patterns between pixels (Lillesand and Kiefer abandoned rail track, and the local trails. The open pits,
1994; Jensen 2005; Gillanders et al. 2008). We then re- roads, cutlines, abandoned rail, and trail datasets were all
moved remaining visible clouds and cloud shadows; this retrieved from GeoGratis, which is maintained by Natu-
masking was performed manually because the data were ral Resources Canada. The size of each buffer reflects the
acquired before USGS began offering Landsat surface differential ecological impacts of each of the mine-related
reflectance products that include a cloud mask. Once these anthropogenic features and is easily replicable for future
data were available, we performed a post hoc comparison land cover change monitoring. Woolmer et al. (2008)
which showed that the manual mask removed more than present a range of human influence scores for each type
90% of the cloud and cloud shadow identified in the cloud of anthropogenic feature; the buffers for the features
mask provided by Landsat. included in this study were selected from the outer limit
In order to focus on mine-related land cover changes, of this range to account for the shorter growing season of
we placed a series of digital buffers (i.e., polygons in the boreal study site. By placing these buffers around
GIS) around the geographic features associated with the mine-related features, we ensure that changes in land
Pine Point Mine (Fig. 2). Sanderson et al. (2002) cover related to the mine are not subsumed by broader

Fig. 2 2009 Landsat 5 TM false


color composite (4-5-3) image of
the Pine Point site (A) and
comparison site (B). The northern
boundary of Wood Buffalo
National Park is represented in
black
Environ Monit Assess (2017) 189:185 Page 7 of 19 185

trends in land cover change. Additionally, in order to We used raster algebra and landscape metrics to quan-
control for shape and edge effects when calculating land- tify land cover composition and configuration, using the
scape metrics, the same polygon created by the cumula- image from 1989 as the common reference year to deter-
tive buffer around all built features on the mine site was mine change since closure at the three subsequent points
used to delineate the comparison site in the park. We in time. Raster algebra in ArcGIS (v. 10.1, ESRI, Red-
compared the resultant land cover rasters to assess land lands, CA) allowed us to perform a pixel-by-pixel com-
cover change at Pine Point and the Wood Buffalo Na- parison of the land cover rasters and generate a map of
tional Park site through time. change for each of the two study sites (Fig. 3). To
complete the analysis, we used FRAGSTATS software
version 4 (McGarigal et al. 2012) to calculate a series of
Analysis methods metrics which quantify patterns in land cover composi-
tion and configuration (Table 3). The background was
In order to document land cover change, we clas- considered no data, the eight-neighbor rule was used, and
sified the satellite images into seven land cover the borders of the selected study sites were not considered
types based on the spectral response pattern of each edges for this analysis.
pixel in the images. We performed a supervised Guided by the principles for selecting meaningful
classification to identify land cover in the study metrics laid out by Cushman et al. (2008), we selected
region. For each image, we selected training sam- a suite of metrics that work in concert to quantify
ples representing each of the target land cover landscape structure at the Pine Point and Wood Buffalo
classes and used the maximum likelihood classifi- National Park sites. The metrics included in this study
cation algorithm in PCI Geomatica 2014 to gener- were chosen to describe different components of land-
ate mutually exclusive land cover categories for the scape structure and to minimize any redundancy inher-
four Landsat images. The land cover categories are ent to the metrics themselves. Percentage of landscape
(1) water, (2) regenerating disturbance, (3) dense (PLAND) provided a comparison of the proportional
coniferous forest, (4) open coniferous forest, (5) abundance of each land cover type at the Pine Point site
meadow, (6) exposed/bare land, and (7) background and the comparison site in Wood Buffalo National Park
(Table 2). A qualitative, post hoc validation of the over the 20-year study period. Largest patch index (LPI)
cover types was performed based on image inter- served as a measure of dominance for the cover types.
pretation and comparison with 80 ground-truth data We used mean patch area (AREA_MN) and number of
points collected in July 2012. Historical fire data patches (NP) to quantify the size and number of patches
from the Canadian National Fire Database, made in each land cover class. We used mean contiguity
available by the Canadian Forest Service, were also (CONTIG_MN) and percentage of like adjacencies
considered in the validation. (PLADJ) to measure the degree of aggregation in the

Table 2 Classification scheme with land cover classes and corresponding descriptions

Class Description

(1) Water Lakes, rivers, and other water bodies


(2) Regenerating disturbance Former disturbances, e.g., burns, that have begun the process
of revegetation; characterized by coarse woody debris, young
jack pines, and fireweed
(3) Dense coniferous forest >65% canopy density, dominated by black spruce
(4) Open coniferous forest <65% canopy density, characterized by black spruce and jack pine
(5) Meadow Open, vegetated areas dominated by sedges, alder, and willow
(6) Exposed/bare land Bare rock or built features like roads, waste rock piles, etc.; extremely
sparse vegetation
(7) Background Pixels with null values
185 Page 8 of 19 Environ Monit Assess (2017) 189:185

Fig. 3 Map showing no land


cover change (black), change
from vegetated to bare (red),
change from bare to vegetated
(green), and water (blue) between
1989 and 2009 at the Pine Point
site (A) and the comparison site
(B). The lack of clear spatial
trends in land cover change
reflects the dynamism of the
boreal environment; the exception
is the black spot at the Pine Point
site (A), which shows that the
tailings pond has not undergone
land cover conversion in the
20 years since mine closure

land cover classes. Mean perimeter-area ratio land cover, and we expected that the Pine Point site
(PARA_MN) indicates the complexity of patches for would be more fragmented, with a higher number of
each land cover type. For landscape-level metrics, we land cover patches and less aggregation. Due to the
calculated Simpson’s diversity (SIDI) and Simpson’s linear features associated with the mine (e.g., roads
evenness (SIEI) to measure the relative diversity and and cutlines), we expected that the Pine Point landscape
dominance of each of the landscapes as a whole. Taken would display less patch complexity within individual
together, these metrics are descriptive tools that yield land cover classes than the comparison site in Wood
insight to the ecological processes occurring at each Buffalo National Park. At the Pine Point site, we ex-
study site. pected the landscape to become less fragmented through
We expected the comparison site in Wood Buffalo time, with an increase in aggregation in the vegetation
National Park to exhibit large, aggregated patches of classes and a decrease in aggregation in the exposed

Table 3 List of landscape metrics calculated using FRAGSTATS (v. 4, Amherst, MA; McGarigal et al. 2012)

Landscape metric Abbreviation Range Units Description

Percentage of landscape PLAND 0–100 Percent (%) Measures the proportional abundance of each land cover class
Largest patch index LPI 0–100 Percent (%) Percent of the landscape comprised by largest patch of each land
cover class
Mean patch area AREA_MN 0–∞ Hectares (ha) Average area of patches in each land cover class
Number of patches NP 1–∞ n/a Measures the extent of subdivision or fragmentation of each land
cover class
Mean contiguity CONTIG_MN 0–1 n/a Average spatial connectedness of patches in each land cover class;
patches are increasingly contiguous as value approaches 1
Percentage of like adjacencies PLADJ 0–100 Percent (%) Measure of aggregation for each land cover class
Mean perimeter-area ratio PARA_MN 0–∞ n/a Measure of patch complexity in each land cover class
Simpson’s diversity SIDI 0–1 n/a Measure of diversity; the probability that any two pixels will be
different patch types
Simpson’s evenness SIEI 0–1 n/a Inverse measure of dominance
Environ Monit Assess (2017) 189:185 Page 9 of 19 185

Fig. 4 Percentage of landscape (%) showing the proportional abundance of land cover types at the Pine Point and Wood Buffalo National
Park study sites

ground class. We also expected patch complexity to Results from the LPI also indicate that the dominant
increase at the Pine Point site through time, as revege- cover type at the WBNP site is dense conifer (Fig. 5).
tation starts to occur around the edges of the linear, Open conifer is dominant at the Pine Point site until
anthropogenic features. 2009, when regenerating vegetation becomes the cover
type with the largest patches. At WBNP, there was a
sharp increase in open conifer in 2005 and an increase in
regenerating vegetation.
Results High mean patch area (AREA_MN) and standard
deviation (AREA_SD) show that dense conifers occur
Class-level metrics in large patches in WBNP and that there is high vari-
ability in the average size of dense conifer patches
The relative abundance of land cover types at Pine Point (Table 4). Dense conifer also showed a relatively low
and the comparison site in Wood Buffalo National Park NP. The average patch sizes for open conifers,
(WBNP) shows notable differences over the 20-year regenerating vegetation, and exposed land in the park
study period (Fig. 4). At Pine Point, meadows and open are considerably lower and exhibit much less variability
conifer are the most abundant cover types. Regenerating compared to dense conifer. Patches of regenerating veg-
vegetation decreased until 2005, after which it increased etation in the park became larger overall, and the num-
again due to a 2008 forest fire. Exposed ground ber of patches decreased over the 20-year study period.
remained approximately constant4 over the 20-year Mean patch size and number of patches exhibited minor
study period. By contrast, dense conifer was consider- fluctuations in the open conifer class.
ably more abundant at the WBNP site than at Pine Point. At the Pine Point site, there was little fluctuation in
A slight increase in the abundance of regenerating veg- mean patch size and relatively stable number of patches
etation in 2005 in WBNP corresponds to a fire that in both conifer classes. For exposed ground, the mean
occurred in the park in 2004. Open conifer increased patch area decreased and the number of patches in-
in WBNP between 1989 and 1994 then decreased, creased (Fig. 6).
which corresponds to the increase in regenerating veg- At Pine Point, the mean contiguity and percentage of
etation in 2005 after the 2004 fire. like adjacencies of exposed land both showed a distinct
downward trend, indicating that patches became more
4
The 2005 image had a considerable amount of haze, likely due to disaggregated in the 20 years following mine closure.
smoke from a forest fire. While we masked out as much as possible, the
spike in the exposed land cover type is a result of high albedo from the Regenerating vegetation also showed an overall down-
haze rather than changes on the ground. ward trend in mean contiguity. The percentage of like
185 Page 10 of 19 Environ Monit Assess (2017) 189:185

Fig. 5 Largest patch index showing the dominance of land cover forest fire. Dense conifer is the dominant cover type in Wood
types at the study sites. At Pine Point, there is no clear dominant Buffalo National Park during the study period. Exposed land is
cover type, but regenerating vegetation increases following a 2008 less dominant in the park than at Pine Point

adjacencies for regenerating vegetation decreased until constant over the study period, and that of dense conifers
2005 and increased sharply by 2009, which showed shows a slight increase until 2005, followed by a slight
large, contiguous patches of regenerating vegetation. decrease. At WBNP, the perimeter-area ratio of exposed
The mean contiguity of open conifer remained fairly land showed a distinct upward trend, indicating that
constant over the study period; this class also had the patches became more complex. Regenerating vegetation
highest values for percentage of like adjacencies and a and dense conifers both showed a slight upward trend,
low range. and regenerating vegetation varied slightly between
At WBNP, the mean contiguity and percentage of downward and upward trends (Table 6).
like adjacencies of exposed ground also both showed a
downward trend. The mean contiguity of open conifer Landscape-level metrics
remained fairly constant, and percentage of like adja-
cencies showed a slight downward trend. Mean conti- At Pine Point, Simpson’s diversity decreased in 1994
guity of dense conifer decreased slightly, and the per- and 2005, then increased dramatically in 2009. This
centage of like adjacencies was high with a fairly narrow indicates that the landscape became less diverse for a
range of values. The percentage of like adjacencies in time, then more diverse by 2009. At WBNP, Simpson’s
regenerating vegetation increased sharply in 2005 be- diversity decreased slightly, then increased in 2005, then
fore slowly decreasing again (Table 5, Fig. 7). decreased slightly in 2009. The values for Simpson’s
At Pine Point, the perimeter-area ratios of regenerating evenness exhibited a very similar pattern to the values of
vegetation and exposed land both show an upward trend, Simpson’s diversity, indicating more dominance and
indicating increasingly complex patch shapes. The mean less evenly dispersed land cover classes at the Pine Point
perimeter-area ratio of open conifers remains fairly site than the park site (Fig. 8).

Table 4 Mean patch area (ha) with standard deviations

1989 1994 2005 2009

Cover type Site Mean SD Mean SD Mean SD Mean SD

Regenerating vegetation PP 0.7 6.3 0.5 2.4 0.2 0.7 0.9 54.6
WBNP 0.6 7.9 0.5 3.9 1.8 46.2 1.3 24.7
Dense conifer PP 1.9 21.5 1.5 16.8 2.1 30.9 2.2 37.6
WBNP 4.2 104.3 3.6 103.1 5.8 156.3 4.2 120.3
Open conifer PP 2.8 58.2 3.2 64.9 4.0 94.7 2.5 45.2
WBNP 1.1 13.1 1.8 24.9 1.5 50.4 1.3 25.1
Bare PP 1.2 41.1 1.0 45.9 0.5 14.8 0.5 24.1
WBNP 1.0 22.5 0.6 6.0 0.8 12.5 0.4 3.5

Precision is listed to the nearest tenth because one 30 × 30 m pixel in the Landsat imagery equals 0.09 ha
Environ Monit Assess (2017) 189:185 Page 11 of 19 185

different from the park site. At the Pine Point site,


the downward trend in the proportional abundance of
regenerating vegetation corresponded with an up-
ward trend in open conifer, indicating that
regenerating vegetation was converting to open co-
nifer until the 2008 fire. The overall slight increase
in the abundance of dense conifer likely indicates
that some open conifer was converting to dense
conifer. The relatively constant proportion of ex-
posed land suggests that little vegetation conversion
occurred at the mine site itself. At the park site,
dense conifer was considerably more abundant than
at Pine Point. Both study sites underwent changes in
land cover composition indicative of fire activity,
specifically an increase in regenerating vegetation
after a 2004 fire in Wood Buffalo National Park
and a 2008 fire closer to the Pine Point Mine. How-
ever, the site in Wood Buffalo National Park is
characterized by consistently abundant dense coni-
fers, while Pine Point is characterized by open coni-
Fig. 6 Number of patches showing fewer patches of dense conifer fers and meadows.
in Wood Buffalo National Park compared to Pine Point In Wood Buffalo National Park, dense conifer is
clearly the dominant cover type. At Pine Point, the
Discussion overall differences in largest patches among the land
cover classes were smaller, indicating less domi-
Interpretation of metrics nance by a particular cover type than at the park site.
While the proportional abundance of exposed land
While we expected that the land cover composition was comparable for Pine Point and Wood Buffalo
and configuration at the Pine Point site would be- National Park, the largest patch index indicates that
come more similar to the comparison site in Wood bare land is a more dominant cover type at the mined
Buffalo National Park through time due to natural landscape than in the park. At Pine Point, the spike
revegetation, our analysis shows that, 20 years after in large patch sizes of regenerating vegetation in
mine closure, the Pine Point site is still considerably 2009 was caused by the 2008 fire near the mine site.

Table 5 Mean contiguity with standard deviations

1989 1994 2005 2009

Cover type Site Mean SD Mean SD Mean SD Mean SD

Regenerating vegetation PP 0.135 0.179 0.127 0.171 0.072 0.123 0.078 0.136
WBNP 0.130 0.169 0.130 0.171 0.113 0.173 0.117 0.182
Dense conifer PP 0.201 0.219 0.184 0.210 0.169 0.204 0.188 0.212
WBNP 0.193 0.219 0.184 0.212 0.180 0.217 0.178 0.215
Open conifer PP 0.166 0.203 0.176 0.206 0.165 0.201 0.170 0.203
WBNP 0.165 0.193 0.181 0.200 0.171 0.196 0.172 0.196
Bare PP 0.180 0.195 0.150 0.179 0.114 0.157 0.093 0.146
WBNP 0.165 0.185 0.138 0.170 0.125 0.167 0.101 0.146

Pecision is listed to the nearest thousandth because the measure of contiguity is not constrained by pixel size
185 Page 12 of 19 Environ Monit Assess (2017) 189:185

ground, the mean patch area decreased and the number


of patches increased. This means that bare land is be-
coming more fragmented, possibly as a result of con-
version to vegetation at the mine site. Patches of
regenerating vegetation in the park became larger over-
all, and the number of patches decreased, indicating that
this cover type became less fragmented over the 20-year
study period. The high mean patch area and relatively
low number patches show that dense conifers are less
fragmented than the other cover types in the park.
At Pine Point, the distinct downward trend in
mean contiguity and percentage of like adjacencies
of exposed land indicates that patches became more
disaggregated through time as vegetation conversion
started to occur at some parts of the mine site. The
overall downward trend in mean contiguity of
regenerating vegetation is indicative of vegetation
conversion. The sharp increase in percentage of like
adjacencies for regenerating vegetation in 2009 is
likely the result of a 2008 fire which created large,
Fig. 7 Percentage of like adjacencies showing aggregated conifer contiguous patches of regenerating vegetation. The
classes and increasingly disaggregated bare land high values for percentage of like adjacencies and
low variability in mean contiguity indicate that open
conifer remained consistently aggregated over the
At Wood Buffalo National Park, the sharp increase 20-year study period. Mean contiguity and percent-
in open conifer in 2005 coupled with an increase in age of like adjacencies for exposed land, then, would
regenerating vegetation indicates that the 2004 fire decrease due to the slight seasonal gradient of the
reduced the size of some dense conifer patches. four images. The slight downward trend in percent-
Little fluctuation in mean patch size and a relatively age of like adjacencies of open conifer shows that
stable number of patches indicates that fragmentation at this cover type gradually became less aggregated.
the Pine Point site remained constant for both conifer The values for mean contiguity and percentage of
classes in the 20 years following closure. For exposed like adjacencies for dense conifers in the park

Table 6 Mean perimeter-area ratio with standard deviations

1989 1994 2005 2009

Cover type Site Mean SD Mean SD Mean SD Mean SD

Regenerating vegetation PP 1131.133 267.711 1141.724 257.486 1222.222 195.350 1216.302 210.908
WBNP 1138.414 255.445 1136.269 258.986 1162.723 257.915 1157.789 269.543
Dense conifer PP 1030.326 318.666 1057.150 307.506 1077.919 298.234 1049.349 309.687
WBNP 1043.459 318.189 1056.950 309.631 1062.398 315.166 1065.793 312.650
Open conifer PP 1082.921 297.731 1067.889 302.444 1084.909 294.824 1077.017 296.979
WBNP 1084.881 285.867 1061.806 294.822 1076.472 289.354 1074.739 288.671
Bare PP 1061.488 289.592 1107.078 269.659 1160.403 239.942 1192.346 224.637
WBNP 1082.876 276.376 1125.457 257.502 1146.122 251.277 1180.879 225.228

Precision is listed to the nearest thousandth because the measure of perimeter-area ratio is not constrained by pixel size
Environ Monit Assess (2017) 189:185 Page 13 of 19 185

Fig. 8 Simpson’s diversity index


showing that the study site in
Wood Buffalo National Park is a
more heterogeneous landscape
than its post-industrial counterpart

indicate a consistently aggregated cover type. The large, contiguous patches of dense conifer and a
sharp increase in 2005 of percentage of like adjacen- higher degree of fragmentation, similar to the park
cies for regenerating vegetation was likely the result site. It is clear that the passive reclamation strategy
of larger, contiguous patches created by the 2004 adopted at Pine Point has effected little vegetation
fire, and the subsequent decrease is consistent with conversion at the mine site in the 20 years since
what we would expect to see as vegetation conver- closure (Fig. 9).
sion occurred.
At Pine Point, the perimeter-area ratios of regenerating Monitoring reclamation at abandoned mines
vegetation and exposed land both show an upward trend,
indicating increasingly complex patch shapes that could By combining remote sensing and landscape metrics,
be the result of vegetation conversion at the mine site. we were able to track broad land cover changes at the
Finally, the results from Simpson’s diversity and abandoned Pine Point Mine. However, despite the ad-
Simpson’s evenness indicate that the site in Wood Buf- vantages the remote sensing has for monitoring remote
falo National Park is a more heterogeneous landscape and/or expansive sites such as this one, there are note-
than the Pine Point site. worthy shortcomings. Some of these, like spatial and
In summary, the park site is dominated by large, temporal resolution of remote sensing datasets, can be
contiguous patches of dense conifer and subject to addressed. Other constraints, such as the launch date of
the natural fragmentation and vegetation conversion satellites, are inherent limitations to the application of
typical of boreal forests as a result of fire. By con- remote sensing and landscape metrics at historical mine
trast, the Pine Point site is comprised of smaller sites.
patches of vegetation with overall much less dense Due to its history of intensive industrial extraction
conifer. The post-industrial landscape differs from and poor reclamation, we expected patches at the Pine
the natural control in that it exhibits less fragmenta- Point site to exhibit less complexity (lower perimeter-
tion in the conifer classes than the park site. There is area ratio) in the open conifer and exposed land classes
some land cover conversion at the Pine Point Mine, than at the park site; however, this was not the case. It is
evidenced by the increasingly disaggregated ex- possible that the 30-m spatial resolution of the imagery
posed land cover class and the change from exposed was insufficient to observe less complex patches of
to vegetated land, but the stability of the proportion- vegetation (e.g., cutlines less than 30 m wide). This is
al abundance of exposed land indicates that this a technical limitation of using Landsat TM data to
conversion is happening very slowly. There is very monitor land cover change at abandoned mine sites.
little conversion of exposed mine land to vegetation, Because the spectral response pattern of each pixel is
mostly occurring at the edges of mine features comprised of the average reflectance of land cover with-
(Fig. 9). While subject to a similar fire regime as in that pixel (Bhatta 2008; Chen et al. 2004), the spectral
the park site, the Pine Point site is overall less response pattern from each pixel—and consequently the
heterogeneous and dynamic at this scale of analysis. reliability of the supervised classification—is dependent
In the absence of industrial mineral development, on the resolution of the satellite image. The ecological
the Pine Point site might have been dominated by processes at work at abandoned mines provide a unique
185 Page 14 of 19 Environ Monit Assess (2017) 189:185
Environ Monit Assess (2017) 189:185 Page 15 of 19 185

R Fig. 9 a Close-up of part of the Pine Point Mine showing land effect, valuable information about landscape pattern is
cover in 2009. Some cutlines are visible as linear patches of dense obscured by broader scale trends in land cover due to
and open conifer, but many that were visible in the field are too
fine to be detected with 30-m Landsat data. b Close-up of part of image resolution. Because metrics are dependent on the
the Pine Point Mine showing changes in land cover between 1989 resolution of the imagery from which they are derived,
and 2009. Some conversion to vegetation is occurring around the the resolution must be sufficiently fine to quantify pat-
edges of the pits and roads, but most of the mine site has remained terns relevant to the process of reclamation on post-
exposed over the 20-year study period
industrial landscapes.
As is often the case with satellite-derived data in
challenge to remote monitoring. With the 30-m spatial remote locations, collecting ground-truth data to vali-
resolution of Landsat TM images, it is possible that land date the land cover classes is time-consuming, costly,
cover change occurring at a finer resolution was not and limited by road access (Ferguson 1991; Foody
detected. Indeed, while many cutlines were observed 2002; Franklin et al. 2003; Barnett et al. 2004). In this
on the ground during the 2012 field season, relatively study, each validation point collected for this study was
few are visible on the 2009 image. Although broad-scale collected within 200 m of a road. In addition, the GPS
or medium-scale products like Landsat TM data are ground-truth data were collected during the 2012 field
immensely useful for tracking broad-scale forest chang- season and as such can only inform the land cover
es (Wulder et al. 2008), it is possible that land cover classification of the most recent satellite image, which
changes indicating ecological processes at abandoned was recorded 3 years beforehand. Contemporaneous
mines are not captured at this resolution. validation of the historical images was not possible,
Long-term monitoring of land cover dynamics at and we relied on image interpretation to inform the
abandoned mines would benefit from affordable, high- supervised classification. Finally, the study was also
resolution satellite imagery. Given that the USGS pro- limited by coarse temporal resolution, as there were
vides a freely accessible data archive which offers repeat few cloud-free, near-anniversary images of study area.
coverage of the study region, Landsat TM data were the DeFries et al. (2007) have stated that imagery cost is a
best option for this study. Indeed, Landsat has often barrier to monitoring deforestation, and Turner et al.
been used to study land cover change at mine sites (2015) call for improved access to affordable, high-
(e.g., Schmidt and Glaesser 1998; Antwi et al. 2008; resolution remotely sensed data to inform conservation
Gillanders et al. 2008; Townsend et al. 2009; Li et al. strategies. The same need exists for monitoring post-
2012). At Pine Point, the Landsat product allowed us to industrial landscapes. Remotely sensed data have the
document overall trends in land cover composition and potential to be a cost-effective means for local commu-
configuration. However, some ecological processes oc- nities to monitor nearby industrial developments, both
curring at Pine Point at fine scales were likely obscured during operations and after closure, but barriers remain.
due to the coarse grain size. There are many active satellites in orbit, often with
Landscape metrics are also susceptible to scale ef- multiple sensors per satellite. Of these, only some are
fects (Lausch and Herzog 2002; Wu et al. 2002; Li and built for Earth observation, and few are available at no
Wu 2004). At Pine Point, linear features like roads and cost (see the ITC database of satellites and sensors 2015).
cutlines are present but not always discernible with Local communities and land managers may be reluctant
imagery that has a 30-m grain size. While less important to allocate limited funds to expensive imagery and the
for analyses of landscape composition, this missing specialized software and skill set required to process the
information is crucial for understanding how and where data and generate a useful product. The application of
land cover conversion happens after passive reclama- remote sensing and landscape metrics to monitoring mine
tion. For example, a cutline that is 12 m wide may show reclamation requires careful consideration of the geo-
up in 30-m imagery as individual pixels in a stepladder graphic coverage, spatial resolution, temporal resolution,
pattern that is not necessarily contiguous, while with and historical depth of free satellite imagery in order to be
finer grained imagery, we might be able to detect the certain that the patterns reflecting ecological processes at
same cutline as a narrow, contiguous patch of land cover work are discernible in the remotely sensed dataset.
that is less complex than its surroundings. The overall Finally, the use of landscape metrics to monitor
measure of composition may reflect little change, but change at abandoned mines would benefit from a stan-
the configuration would be meaningfully different. In dardized protocol for assessing the statistical significance
185 Page 16 of 19 Environ Monit Assess (2017) 189:185

of land cover change. While this paper presents a case for environmental monitoring. Broad-scale or
study that employed landscape metrics as a descriptive medium-scale products like Landsat TM data can
tool for evaluating change at an abandoned mine and be used effectively to monitor broad-scale forest
comparison site, statistical analyses based on a much changes; however, these datasets are not necessarily
larger sample of mine sites in similar ecosystems would adequate for monitoring the legacy effects of aban-
be required to incorporate the methodology into a formal doned mines because some ecological processes are
monitoring strategy that could be applied at mines across taking place at a scale finer than the scale of obser-
Canada. If such a monitoring strategy were to exist, it vation available from such imagery. Landscape met-
could inform the creation of dynamic, adaptive reclama- rics are also influenced by scale effects and resolu-
tion plans that could be tailored to the unique reclamation tion; in this case, we expected to see simpler patch
needs of derelict mines. Despite its potential utility for shapes at the mine site. We hypothesize that the 30-
monitoring land cover change at abandoned mines, de- m resolution of Landsat imagery is insufficiently
veloping such a protocol is far from straightforward. fine to detect simplicity in patches when features
Statistical analyses of metrics for abandoned mines of interest (e.g., cutlines) are often only a pixel or
would require distributions of expected values for land- two wide. Satellite remote sensing is potentially a
scape metrics at appropriate reference landscapes powerful means of monitoring mine reclamation;
(Remmel and Csillag 2003). Appropriate reference land- however, the land cover dynamics at abandoned
scapes would have to mirror the environmental condi- mined sites might be operating at scales that remain
tions of each abandoned mine to capture undisturbed inaccessible with 30-m Landsat data.
land cover change through time. Any statistical compar- While certain obstacles to using remote sensing for
isons would also have to account for metrics’ sensitivities monitoring historical mines will always exist due to the
to imagery resolution and extent, as well as issues of limited temporal depth of satellite technology, remote
spatial autocorrelation (Wu et al. 2002; Remmel and sensing is well-suited for monitoring the closure and
Csillag 2003; Wu 2004). reclamation success of mines currently in operation.
However, the issues of affordability and access still pose
significant challenges. Land managers or local commu-
Conclusions nities interested in monitoring the reclamation progress
of abandoned mines must weigh the relative importance
At Pine Point, it is clear that the industrial development of spatial resolution, temporal resolution, historical
of the mine and the lack of active reclamation after its depth, and cost of imagery before land cover can be
abandonment altered the land cover composition and classified and landscape metrics calculated. The effec-
configuration in the area. While there appears to be tive monitoring of long-term land cover change on post-
some vegetation conversion happening at the edges of industrial landscapes requires an improvement in the
the mine site, the slow rate of this conversion indicates accessibility and affordability of high-resolution time
that passive reclamation has so far been insufficient to series datasets.
restore ecological function at the site. The results from
the comparison site in Wood Buffalo National Park
suggest that the Pine Point site might have been domi- Acknowledgements The authors would like to thank the
nated by large, contiguous patches of dense conifer and Deninu Kue First Nation for their support of this research. Thanks
a higher degree of fragmentation had the mine not been also to Dr. Arn Keeling for his thoughts on earlier drafts of this
paper, Dr. Zhaohua Chen (C-CORE) and Dr. Vincent Lecours for
developed.
their insight regarding remote sensing and scale, and the anony-
We have shown that it is possible to trace overall mous reviewers for their constructive comments. This research
trends in the composition and configuration of land was supported by the Abandoned Mines in Northern Canada
cover on post-industrial landscapes using 30-m sat- Project, which was funded through the Social Sciences and Hu-
manities Research Council (SSHRC) grant number 866-2008-16.
ellite imagery. We posit that monitoring the legacy
effects of historical mines in northern Canada would
Compliance with ethical standards
be improved by better access to affordable, high-
resolution imagery. Satellite remote sensing and Conflict of interest The authors declare that they have no con-
landscape metrics are unquestionably valuable tools flict of interest.
Environ Monit Assess (2017) 189:185 Page 17 of 19 185

References Cooke, J. A., & Johnson, M. S. (2002). Ecological restoration of


land with particular reference to the mining of metals and
industrial minerals: a review of theory and practice.
Abel, K. M. (2005). Drum songs: glimpses of dene history (2nd Environmental Review, 10, 41–71.
ed.). Montreal: McGill-Queen’s University Press. Coppin, P., et al. (2004). Digital change detection methods in
Ahmed, O. S., Franklin, S. E., & Wulder, M. A. (2013). ecosystem monitoring: a review. International Journal of
Interpretation of forest disturbance using a time series of Remote Sensing, 25(9), 1565–1596.
Landsat imagery and canopy structure from airborne lidar. Cushman, S. A., McGarigal, K., & Neel, M. C. (2008). Parsimony
Canadian Journal of Remote Sensing, 39(6), 521–542. in landscape metrics: strength, universality, and consistency.
Alagona, P. S., Sandlos, J., & Wiersma, Y. F. (2012). Past imperfect: Ecological Indicators, 8, 691–703.
using historical ecology and baseline data for conservation and Czerwinski, C. J., King, D. J., & Mitchell, S. W. (2014). Mapping
restoration projects in North America. Environmental forest growth and decline in a temperate mixed forest using
Philosophy, 9(1), 49–70. temporal trend analysis of Landsat imagery, 1987-2010.
Antwi, E. K., Krawczynski, R., & Wiegleb, G. (2008). Detecting Remote Sensing of Environment, 141, 188–200.
the effect of disturbance on habitat diversity and land cover
DeFries, R., et al. (2007). Earth observations for estimating green-
change in a post-mining area using GIS. Landscape and
house gas emissions from deforestation in developing coun-
Urban Planning, 87, 22–32.
tries. Environmental Science & Policy, 10, 385–394.
Badiou, P., et al. (2013). Conserving the world’s last great forest is
Desbiens, C., & Rivard, É. (2014). From passive to active dia-
possible: here’s how. Science/policy briefing note by the
logue? Aboriginal lands, development, and métissage in
International Boreal Conservation Science Panel (IBCSP)
Québec, Canada. Cultural Geographies, 21(1), 99–114.
and associates.
Deshaies, A., Boudreau, S., & Harper, K. A. (2009). Assisted
Barnett, P. J., et al. (2004). Methods for remote engineering
revegetation in a subarctic environment: effects of fertiliza-
geology terrain analysis in boreal forest regions of Ontario,
tion on the performance of three indigenous plant species.
Canada. Environmental& Engineering Geoscience, X(3),
Arctic, Antarctic, and Alpine Research, 41(4), 434–441.
229–241.
Beaubien, J., et al. (1999). Land cover from multiple thematic Ecological Stratification Working Group (ESWG). (1995). A
National Ecological Framework for Canda. Agriculture
mapper scenes using a new enhancement-classification meth-
odology. Journal of Geophysical Research, 104(D22), 27, and Agri-Food Canada, Research Branch, Centre for Land
909–27, 920. and Biological Resources Research and Environment
Canada, State of the Environment Directorate, Ecozone
Bhatta, B. (2008). Remote sensing and GIS. New Delhi: Oxford
Analysis Branch, Ottawa/Hull. Report and national map at
University Press.
1:7,500,000 scale.
Bond-Lamberty, B., & Gower, S. T. (2008). Decomposition and
Environment Canada. (1996). The state of Canada’s environment
fragmentation of coarse woody debris: re-visiting a boreal
1996. Ottawa: Government of Canada.
black spruce chronosequence. Ecosystems, 11, 831–840.
Boucher, Y., et al. (2009). Logging pattern and landscape changes Environment Canada. (2010). 1981–2010 Climate Normals and
over the last century at the boreal and deciduous forest transition Aver ages . http:// climat e.we ather.gc.ca/ cli m ate_
in Eastern Canada. Landscape Ecology, 24, 171–184. normals/index_e.html. Accessed 2 October 2014.
Bowman, B., & Baker, D. (1998). Mine reclamation planning in Ferguson, R. S. (1991). Detection and classification of muskox
the Canadian North. Northern Minerals Program Working habitat on Banks Island, Northwest Territories, Canada, using
Paper No. 1. Prepared for the Canadian Arctic Resources Landsat Thematic Mapper data. Arctic, 44(1), 66–74.
Committee (CARC). Fichera, C. R., Modica, G., & Pollino, M. (2012). Land cover
Boyd, D. R. (2003). Unnatural law: rethinking Canadian envi- classification and change-detection analysis using multi-
ronmental law and policy. Vancouver: UBC Press. temporal remote sensed imagery and landscape metrics.
Bradshaw, A. (2000). The use of natural processes in reclama- European Journal of Remote Sensing, 45, 1–18.
tion—advantages and difficulties. Landscape and Urban Foody, G. M. (2002). Status of land cover classification accuracy
Planning, 51(2–4), 89–100. assessment. Remote Sensing of Environment, 80, 185–201.
Bridge, G. (2004). Contested terrain: mining and the environment. Franklin, S. E., et al. (2003). Discrimination of conifer height, age
Annual Review of Environment and Resources, 29, 205–259. and crown closure classes using Landsat-5 TM imagery in
Carlson, M., et al. (2010). Maintaining the role of Canada’s forests the Canadian Northwest Territories. International Journal of
and peatlands in climate regulation. The Forestry Chronicle, Remote Sensing, 24(9), 1823–1834.
86(4), 434–443. Fukami, T., & Wardle, D. A. (2005). Long-term ecological dynamics:
CCSG Associates. (2001). Financial options for the remediation reciprocal insights from natural and anthropogenic gradients.
of mine sites: a preliminary study. Ottawa: Prepared for Proceedings of the Royal Society B, 272, 2105-2115.
Mining Watch Canada. Fumoleau, R. (1974). As long as this land shall last: a history of
Chen, D., Stow, D. A., & Gong, P. (2004). Examining the effect of treaty 8 and treaty 11, 1870–1939. Toronto: The Canadian
spatial resolution and texture window size on classification Publishers.
accuracy: an urban environment case. International Journal Gardiner, R. T. (1990). Tailing vegetation studies at the Pine Point
of Remote Sensing, 25(11), 2177–2192. Mine. Vancouver: Progress Report for Pine Point Mines Ltd.
Cominco Ltd. (1991). Restoration and abandonment plan: millsite Gillanders, S. N., et al. (2008). Application of Landsat satellite
and tailings containment areas. imagery to monitor land-cover changes at the Athabasca Oil
185 Page 18 of 19 Environ Monit Assess (2017) 189:185

Sands, Alberta, Canada. The Canadian Geographer, 52(4), Resource Use and Land Use Policy Study (pp. 65–110).
466–485. Ottawa: Canadian Arctic Resources Committee.
Gratto-Trevor, C. L. (1996). Use of Landsat TM imagery in McGarigal, K., Cushman, S.A., & Ene, E. (2012). FRAGSTATS
determining important shorebird habitat in the Outer v4: Spatial Pattern Analysis Program for Categorical and
Mackenzie Delta, Northwest Territories. Arctic, 49(1), 11–22. Continuous Maps. Computer software program produced by
Hodacova, D. T., & Prach, K. (2003). Spoil heaps from brown coal the authors at the University of Massachusetts, Amherst.
mining: technical reclamation versus spontaneous revegeta- Available at the following web site: http://www.umass.
tion. Restoration Ecology, 11(3), 385–391. edu/landeco/research/fragstats/fragstats.html.
ITC database of satellites and sensors (2015). Maintained by the Miller, J. R., et al. (1997). Seasonal change in understory reflec-
Faculty of Geo-Information Science and Earth Observation, tance of boreal forests and influence on canopy vegetation
University of Twente, Netherlands. http://www.itc. indices. Journal of Geophysical Research, 102(D24), 29,
nl/research/products/sensordb/AllSatellites.aspx. Accessed 7 475–29, 482.
May 2015. Mining, Minerals and Sustainable Development (MMSD). (2002).
Jensen, J. R. (Ed.). (2005). Introductory digital image processing: Research on mine closure policy. International Institute for
a remote sensing perspective (3rd ed.). Upper Saddle River: Environment and Development. No. 44 (pp. 1–91).
Pearson Prentice Hall. Narumalani, S., Mishra, D. R., & Rothwell, R. G. (2004). Change
Keeling, A. (2010). ‘Born in an atomic test tube’: landscapes of detection and landscape metrics for inferring anthropogenic
cyclonic development at Uranium City, Saskatchewan. The processes in the greater EFMO area. Remote Sensing of
Canadian Geographer, 54(2), 228–252. Environment, 91, 478–489.
Keeling, A., & Sandlos, J. (2009). Environmental justice goes Natcher, D. C. (2009). Subsistence and the social economy of
underground? Historical notes from Canada’s northern min- Canada’s Aboriginal north. The Northern Review, 30, 83–98.
ing frontier. Environmental Justice, 2(3), 117–125. Nickerson, D. (1992). RE: Approval of the Abandonment and
Kendall, G. (1992). Mine closures and worker adjustment: the case Restoration Plan-Pine Point. Letter to D.L. Johnston, Vice
of pine point. In C. Neil, M. Tykklainen, & J. Bradbury President of Northern Operations, Cominco Metals,
(Eds.), Coping with closure: an international comparison of Cominco Ltd.
mine town experiences (pp. 131–150). London: Routledge. Otto, J. M. (2009). Global trends in mine reclamation and closure
Larsen, J. A. (1980). The boreal ecosystem. New York: Academic regulation. In J. P. Richards (Ed.), Mining, Society, and a
Press, Inc.. sustainable World (pp. 251–288). London: Springer-Verlag.
Lausch, A., & Herzog, F. (2002). Applicability of landscape Parks Canada. (2010). Wood Buffalo National Park of
metrics for the monitoring of landscape change: issues of Canada Management Plan. http://www.pc.gc.ca/pn-
scale, resolution and interpretability. Ecological Indicators, np/nt/woodbuffalo/plan/plan1.aspx. Accessed 22
2, 3–15. February 2015.
Lecain, T. J. (2009). Mass destruction: the men and giant mines Pickett, S. T. A. (1989). Space-for-time substitution as an alterna-
that wired America and scarred the planet. New Brunswick: tive to long-term studies. In G. E. Likens (Ed.), Long Term
Rutgers University Press. Studies in Ecology: Approaches and Alternatives (pp. 110–
LeClerc, E., & Keeling, A. (2015). From cutlines to traplines: post- 135). New York: Springer-Verlag.
industrial land use at the Pine Point mine. The Extractive Piper, L. (2009). The industrial transformation of subarctic
Industries and Society, 2(1), 7–18. Canada. Vancouver: UBC Press.
Lecomte, N., et al. (2005). Effects of fire severity and initial tree Powers, R. P., et al. (2013). Integrating accessibility and intactness
composition on understorey vegetation dynamics in a boreal into large-area conservation planning in the Canadian boreal
landscape inferred from chronosequence and paleoecological forest. Biological Conservation, 167, 371–379.
data. Journal of Vegetation Science, 16, 665–674. Ray, A. J. (1990). The Canadian fur trade in the industrial age.
Lee, P., & Boutin, S. (2006). Persistence and developmental transition Toronto: University of Toronto Press.
of wide seismic lines in the western Boreal Plains of Canada. Remmel, T. K., & Csillag, F. (2003). When are two landscape
Journal of Environmental Management, 78, 240–250. pattern indices significantly different? Journal of
Li, H., & Wu, J. (2004). Use and misuse of landscape indices. Geographical Systems, 5(4), 331–351.
Landscape Ecology, 19, 389–399. Sanderson, E. W., Jaiteh, M., Levy, M. A., Redford, K. H.,
Li, X., Li, S., & Gao, J. (2012). The monitoring of ecological Wannebo, A. V., & Woolmer, G. (2002). The human footprint
reclamation evolution and landscape ecology remodeling and the last of the wild. BioScience, 52(10), 891.
technology of mining subsidence area. Paper presented at Sandlos, J. (2007). Hunters at the margin: native people and
the Second International Workshop on Earth Observation wildlife conservation in the Northwest Territories.
and Remote Sensing Applications. Vancouver, BC: UBC Press.
Li, J., et al. (2015). Reconstructing disturbance history for an inten- Sandlos, J. (2015). BA mix of the good and the bad^: community
sively mined region by time series analysis of Landsat imagery. memory and the Pine Point Mine. In A. Keeling & J. Sandlos
Environmental Monitoring and Assessment, 187, 557. (Eds.), Mining and Communities in Northern Canada:
Lillesand, T. M., & Kiefer, R. W. (1994). Remote sensing and History, Politics, and Memory (pp. 137–166). Calgary:
image interpretation (3rd ed.). New York: John Wiley and University of Calgary Press.
Sons, Inc.. Sandlos, J., & Keeling, A. (2012). Claiming the new north: devel-
Macpherson, J. E. (1978). The Pine Point mine. In E. B. Peterson opment and colonialism at the Pine Point mine, Northwest
& J. B. Wright (Eds.), Northern Transitions, Vol. I: Northern Territories, Canada. Environment and History, 18, 5–34.
Environ Monit Assess (2017) 189:185 Page 19 of 19 185

Schmidt, H., & Glaesser, C. (1998). Multitemporal analysis of satellite comparison with selected Canadian and U.S. jurisdictions
data and their use in monitoring of the environmental impacts of (pp. 1–113). Canadian Institute of Resources Law, Calgary;
open cast lignite mining areas in Eastern Germany. International Canadian Arctic Resources Committee; and Department of
Journal of Remote Sensing, 19(12), 2245–2260. Indian and Northern Affairs Canada.
Soverel, N., Coops, N. C., White, J. C., & Wulder, M. A. Wiersma, Y. F., & Simonson, C. (2010). Parks as islands: species
(2010). Characterizing the forest fragmentation of loss in Canadian national parks. Parks Science, 27(2), 70–77.
Canada’s national parks. Environmental Monitoring and Wiersma, Y. F., Nudds, T. D., & Rivard, D. H. (2004). Models to
Assessment, 164, 481–499. distinguish effects of landscape patterns and human popula-
Timoney, K. P., & Peterson, G. (1996). Failure of natural regener- tion pressures associated with species loss in Canadian na-
ation after clearcut logging in Wood Buffalo National Park, tional parks. Landscape Ecology, 19, 773–786.
Canada. Forest Ecology and Management, 87, 89–105. Will, A. (2015). Resident attitudes and beliefs toward bison,
Townsend, P. A., et al. (2009). Changes in the extent of surface disease and management in Wood Buffalo National Park.
mining and reclamation in the Central Appalachians detected Master’s thesis, Memorial University.
using a 1976-2006 Landsat time series. Remote Sensing of Woodcock, C. E., et al. (2008). Free access to Landsat imagery.
Environment, 113, 62–72. Science, 320(5879), 1011.
Turner, W., et al. (2015). Free and open-access satellite data are Woolmer, G., et al. (2008). Rescaling the Human Footprint: a tool
key to biodiversity conservation. Biological Conservation, for conservation planning at an ecoregional scale. Landscape
182, 173–176. and Urban Planning, 87, 42–53.
Usher, P. J., Duhaime, G., & Searles, E. (2003). The household as Worrall, R., et al. (2009). Towards a sustainability criteria and
an economic unit in arctic Aboriginal communities, and its indicators framework for legacy mine land. Journal of
measurement by means of a comprehensive survey. Social Cleaner Production, 17, 1426–1434.
Indicators Research, 61, 175–202. Wu, J. (2004). Effects of changing scale on landscape pattern
Valeria, O., Laamrani, A., & Beaudoin, A. (2012). Monitoring the analysis: scaling relations. Landscape Ecology, 19, 125–138.
state of a large boreal forest region in eastern Canada through Wu, J., et al. (2002). Empirical patterns of the effects of changing scale
the use of multitemporal classified satellite imagery. on landscape metrics. Landscape Ecology, 17, 761–782.
Canadian Journal of Remote Sensing, 38(1), 91–108. Wulder, M. A., et al. (2008). Monitoring Canada’s forests. Part 1:
Wenig, M. M., O’Reilly, K., & Chambers, D. (2005). The mining Completion of the EOSD land cover project. Canadian
reclamation regime in the Northwest Territories: a Journal of Remote Sensing, 34(6), 549–562.

You might also like