Open access

Anomalous surface elevation, velocity, and area changes of Split Lake Glacier, western Prince of Wales Icefield, Canadian High Arctic

Publication: Arctic Science
30 August 2022


Here we use a variety of remote sensing data sets to characterize the evolving extent, surface features, dynamics, and surface elevations of Split Lake Glacier, a small outlet of the Prince of Wales Icefield, Nunavut. The glacier started advancing between 1959 and 1975, with a continued increase in terminus area up to the present day, coincident with significant upper elevation thinning and lower elevation thickening that cannot be accounted for by surface mass balance. The highest velocities reach >600 m year−1, with the region of fastest ice motion focused around an icefall that occurs in a bedrock constriction. Distinctive ogives are present in a 1975 air photo of the glacier for the first time, which suggests that rapid motion started by 1970. These patterns are anomalous when compared with the geometry, velocity, and area changes of all other nearby areas of western Prince of Wales Icefield and suggest that Split Lake Glacier may be a slowly surging glacier. The surge duration of 50+ years is longer than any other previously described surge within the Canadian Arctic Archipelago. These results give further information concerning the wide variety of dynamic and geometrical changes of glaciers across this region.


Les auteurs utilisent ici divers ensembles de données de télédétection pour caractériser l’étendue évolutive, les caractéristiques de surface, la dynamique et les élévations de surface du glacier du lac Split, un petit exutoire du champ de glace Prince-de-Galles, au Nunavut. Le glacier a commencé à avancer entre 1959 et 1975, avec une augmentation continue de la zone terminale jusqu’à aujourd’hui, coïncidant avec un amincissement significatif de l’élévation supérieure et un épaississement de l’élévation inférieure qui ne peuvent être expliqués par le bilan de masse de surface. Les vitesses les plus élevées atteignent >600 m an-1, la région où le mouvement de la glace est le plus rapide étant concentrée autour d’une chute de glace qui se produit dans une constriction du substratum rocheux. Des ogives distinctes apparaissent pour la première fois sur une photo aérienne du glacier prise en 1975, ce qui suggère que le mouvement rapide a commencé en 1970. Ces patrons sont anormaux lorsqu’on les compare à la géométrie, à la vitesse et aux changements de surface de toutes les autres zones voisines de l’ouest du champ de glace Prince-de-Galles et suggèrent que le glacier du lac Split pourrait être un glacier à écoulement lent. La durée de l’écoulement de plus de 50 ans est plus longue que tout autre écoulement décrit précédemment dans l’archipel arctique canadien. Ces résultats fournissent des informations supplémentaires concernant la grande variété des changements dynamiques et géométriques des glaciers dans cette région. [Traduit par la Rédaction]

1. Introduction

The Canadian Arctic Archipelago (CAA) has been experiencing an overall trend of glacier mass loss since in situ mass balance measurements began in the 1960s, which has become particularly pronounced since the 1990s (Lenaerts et al. 2013; Millan et al. 2017; Thomson et al. 2017; Noël et al. 2018; Derksen et al. 2019). As a result of these negative mass balances, glacier retreat (Cook et al. 2019) and surface elevation loss (Hugonnet et al. 2021; Tepes et al. 2021) have become ubiquitous across the CAA in the 21st century.
The release of open access data products (e.g., Hugonnet et al. 2021) allows for the study of geometry changes for individual glaciers and enables process-based studies to investigate the mechanisms that drive surface elevation changes at the basin scale. Similarly, the repeated annual mapping of glacier velocity from remote sensing imagery (synthetic aperture radar (SAR) and optical data sets) allows for variations in ice motion to be investigated at both regional and glacier basin scales. For example, the systematic collection and processing of SAR data, including ALOS/PALSAR, Radarsat-1, and Radarsat-2, has allowed for the flow characteristics of glaciers within the CAA to be established in detail (Van Wychen et al. 2014; 2016; Millan et al. 2017; Sánchez-Gámez and Navarro 2017; 2021). Collectively, these studies, combined with the open availability of pre-derived velocity maps from optical imagery from the ITS_LIVE project (Gardner et al. 2018, 2021) provide a robust catalogue of data to investigate dynamic change within the CAA. However, most previous studies have focused on investigations of the large tidewater outlet glaciers within the region (Van Wychen et al. 2016; 2021; Millan et al. 2017; Harcourt et al. 2020), with little work completed on process-based investigations for smaller outlet glaciers. The objective of this study is to use remote sensing data sets to characterize the changes in the area, dynamics, and geometry of Split Lake Glacier since the 1950s, a small outlet glacier in the eastern CAA (Fig. 1a)
Fig. 1.
Fig. 1. (a) Location of Split Lake Glacier and the Prince of Wales (POW) Icefield in the Canadian High Arctic. (b) Overview of the lower portion of Split Lake Glacier (RGI ID: RGI60-30.0713) with centerline distances marked in purple dots, the RGI V. 6.0 1999–2003 outline is in red, and the area sampled for the elevation change plots in Fig. 7 is delineated in black dashed lines. The background image is a Landsat-8 image acquired 2020–07–28. (c) Split Lake Glacier and surrounding area with the connection to Makinson Inlet annotated. Taggart Lake Glacier and Kooetigeto River Glacier are previously identified surge-type glaciers on Eastern POW and are *unofficial names. Each glacier basin along the western margin of POW is numbered to correspond to their areas in Table 2a. The red lines indicate the glacier basins of POW Icefield from RGI V. 6.0. (d) Elevation profile of Split Lake Glacier from ArcticDEM 2013–03–13, extracted along the centerline shown in part (b). (e) and (f) inset of the upper icefall of Split Lake Glacier that shows the change in crevasse patterns between 2002 (e) and 2020 (f). Imagery is from Landsat-5 2002–07–19 (e) and 2020–07–24 (f).

2. Study site

Split Lake Glacier (77°44.401′N, 81°47.384′W) is located on the south-western margin of the Prince of Wales (POW) Icefield, Ellesmere Island (Fig. 1a). The glacier flows from the broad interior portion of the POW Icefield through a narrow bedrock constriction and associated icefall that is 1.5 km wide to a 6.5 km wide piedmont terminus (Fig. 1b). At the present day, the ice surface becomes heavily crevassed from ∼4.5 km up-glacier from this constriction point, while the lowermost portion of the terminus is covered by ogives and supraglacial streams (Fig. 1b). The terminus itself is land-terminating along its main western boundary, but it meets water in two other locations: the northern tip calves ice into a freshwater lake, while the southern tip calves into a small lake which is connected (albeit weakly) to the ocean at the northern edge of Makinson Inlet. The elevation profile of the terminus is relatively gentle (7° average slope), with surface elevation gradually increasing from ∼55 m a.s.l. to ∼200 m a.s.l. over the lower 6 km of the glacier (Fig. 1d). Above this point, the surface becomes steeper, with elevations increasing from <200 m a.s.l. to >1000 m a.s.l. along the next 12 km section of the glacier. Of note is the steeper region (14° average slope) that occurs between ∼9 and 10 km from the terminus (Fig. 1d), which corresponds with a region of intense crevassing. This contrasts with the relatively flat topography of the terminus and more gentle topography (2° average surface slope) in the upper parts of the glacier as it extends into the POW Icefield.
The model results of Noël et al. (2018) indicate a long-term trend of increasingly negative surface mass balance (SMB) across POW Icefield, including the part which includes Split Lake Glacier. In the lower ablation zone below the bedrock constriction, the annual average SMB was −2302 ± 163 mm w.e. year−1 between 1996 and 2015, while in the upper reservoir region it was −641 ± 161 mm w.e. year−1 (Noël et al. 2018). The equilibrium line altitude (ELA) of Split Lake Glacier is estimated to be 980 m a.s.l. based on the field measurements of Mair et al. (2009) and the modelling work of Noël et al. (2018).
Several previous studies have indicated that Split Lake Glacier is one of the fastest flowing outlet glaciers of the POW Icefield, with surface velocities ranging from 300 to 600 m year−1 in the constricted region above the piedmont glacier tongue (Van Wychen et al. 2014; 2021; Millan et al. 2017; Strozzi et al. 2017; Sánchez-Gámez and Navarro 2017), although none of these have discussed the dynamics of the glacier in detail. In comparison, nearby geomorphologically similar glaciers along the western margin of POW Icefield (i.e., smaller primarily land terminating glaciers that drain relatively small basins) have maximum reported velocities of approximately 100 m year−1 (Van Wychen et al. 2014; 2016; 2021; Strozzi et al. 2017; Millan et al. 2017).

3. Methods

3.1. Ice extents and area change

To quantify the associated area and terminus changes of Split Lake Glacier since the late 1950s, we manually digitized extents from historical air photos and Landsat imagery. To obtain our earliest observations, we georeferenced two digitized air photos at 5–10 m resolution from July 26 and 27, 1959 (A16692-108 and A16690-009, respectively), acquired by the Royal Canadian Air Force and obtained from the National Air Photo Library, Ottawa, Canada. Next, we digitized terminus extents from level-1 satellite images collected by Landsat-1 (May 27, 1975), Landsat-5 MSS (August 10, 1987), Landsat-5 TM (July 9, 1993), and Landsat-8 (April 1, 2015; July 28, 2020) at a resolution of 60, 60, 30, and 30 m, respectively. The satellite data sets were preprocessed, which accounts for geometric and radiometric corrections and orthorectification. The air photos were manually georeferenced to the Landsat-8 image from July 28, 2020, using surrounding stable bedrock features with an average of 23 tie points for 1959.
The box method of Moon and Joughin (2008) was used to measure the terminus area changes of Split Lake Glacier and three adjacent glaciers flowing from the same region of POW Icefield (boxes 2, 3, and 4 in Fig. 2a). The basin outline from the Randolph Glacier Inventory (RGI) version 6.0 was used (Pfeffer et al. 2014) and was modified to include the changing terminus extents. Terminus extents were digitized from 1959 to 2020 for box 2 and Split Lake Glacier, while the glaciers in boxes 3 and 4 were only digitized for the years 1975 to 2020 due to the lack of earlier air photos.
Fig. 2.
Fig. 2. (a) Terminus extents of Split Lake Glacier and glacier 2 for 1959–2020 and extents of glaciers 3 and 4 for 1975–2020. Background: Landsat-8, 2020–07–28 (UTM Zone 17 N); (b, c, d, e) Change in the area over time of the glacier terminus within each black box; vertical lines indicate errors. Note the difference in y-axis scales between plots.
The primary sources of error for ice front extents are from manual digitization and georeferencing errors for the air photos. For manual digitization, we estimate the error to be ±1 pixel, following the methodology of Krumwiede et al. (2014), which is defined by the 60 m resolution for Landsat-1 imagery, and 30 m resolution for Landsat-5 and Landsat-8 imagery. Therefore, the error for the glacier area is proportional to the sensor resolution and estimated by the percentage area (Aer):
where n is the number of pixels of the perimeter, m is sensor spatial resolution expressed as an area, and Agl is the area of the glacier within the delineated box. The resolution of the air photos varies by flightline but is always better than 10 m, so we used a digitization error of 10 m for this source. Additional errors for the air photos are from georeferencing during the alignment process in ArcGIS, with average positional uncertainty of ± 66.4 m for the 1959 photo.

3.2. Glacier velocities

We use various velocity data sets to characterize the dynamic evolution of Split Lake Glacier between the 1990s and 2020, derived from analysis of optical and radar satellite images.

3.2.1. Radarsat velocity data set

This data set consists of a pre-existing record of ice velocities (Van Wychen et al. 2016; 2021) derived from speckle-tracking of pairs of Radarsat-1 (for the year 2000) and Radarsat-2 (from 2011 to 2020) SAR data acquired during the winter months (November to April in each year). From hereon, we refer to this record as the Radarsat velocity data set. It is important to note that Radarsat velocities provided for each year are determined from a single pair of images and thus represent a single snapshot within a particular year rather than the motion of the glacier for the entirety of the year (image dates provided in Table 1).
Table 1.
Table 1. Details of the images used to derive the “Radarsat” velocity record and the mean velocity error for each image pair. Standard deviation in error is provided in brackets in the error column.
The speckle-tracking processing methodology used for this data set was implemented in MATLAB and is well detailed in earlier work (Van Wychen et al. 2014, 2016), so is only briefly reviewed here. The method begins by accurately co-registering pairs of SAR images acquired with the same imaging geometry (multiples of 24-day repeat orbits in the case of Radarsat-1 and Radarsat-2). In the case of Radarsat-2, the orbital parameters provided in the image metadata are good enough to facilitate the co-registration of the image pairs, but in the case of Radarsat-1 an additional manual co-registration step is required. Once this is completed, a cross-correlation algorithm is used to determine displacements between image chips in both the reference and search images. In this case, image chips with sizes of ∼400 m in both azimuth and range were used. The 1:250 000 Canadian Digital Elevation Dataset with 100 m pixel spacing was used to convert the slant range displacement to ground range displacement. Displacements are then calibrated by selecting non-moving regions (i.e., bedrock outcrops) and removing a constant from the magnitude in velocity. This removes artificial displacements in the velocity results that are caused by co-registration errors between the images that are present when only orbital information is used for the original co-registration of the images. All velocities are then standardized to annual values (m year−1).
Changes in the ionosphere between SAR image acquisitions can cause incorrect displacements in the azimuth direction to be determined in velocity maps using this speckle-tracking method, a phenomenon known as azimuth streaking (Gray et al. 2000). To exclude this effect from our derived Radarsat-2 velocity maps, we inspected each map and only included those that did not have any streaking. Our speckle tracking methodology has been widely used in the CAA before (e.g., Van Wychen et al. 2012, 2014, 2017, 2021), and provides comparable results (in terms of precision) to commercial software packages such as Gamma InSAR (Schellenberger et al. 2016).
The Radarsat velocity data set was manually scrutinized, with erroneous points (those deemed to deviate implausibly from their neighbours in either direction or magnitude) identified and removed. Finally, an Inverse Distance Weighting (IDW) interpolation was run on the filtered displacement data set to create a continuous raster surface of glacier velocities with a grid spacing of 100 m. This velocity raster was then clipped to the extents of the glacierized area using the outline provided by the RGI version 6.0 (Pfeffer et al. 2014). A full error analysis for the data sets used in this study is provided in Van Wychen et al. (2016, 2021), with error values for each of the individual Radarsat velocity maps provided in Table 1. The highest error is associated with the velocity map determined from Radarsat-1 data, with mean apparent motion over bedrock areas of 12.0 m year−1 in 2000. For the velocity maps generated from Radarsat-2 data, the mean apparent motion over bedrock ranged from a minimum of 4.1 m year−1 for 2018 data to a maximum of 8.0 m year−1 for 2014 data (Table 1; Van Wychen et al. 2016, 2021). Furthermore, Van Wychen et al. (2012) compared velocities determined with the speckle tracking method used here with coincident in situ GPS displacements at two locations on Devon Ice Cap and found a maximum difference of 3.05 m year−1. A similar analysis using Radarsat-2 data acquired over Svalbard indicated similar accuracies (see, e.g., Fig. 2 in Schellenberger et al. 2016). Based on these assessments, we conservatively assume that changes of >15 m year−1 between our Radarsat-derived velocity products are required to indicate real changes in ice motion.

3.2.2. ITS_LIVE velocity data set

This data set uses velocities provided by the ITS_LIVE project (, based on optical feature matching of Landsat imagery. The processing scheme utilizes the Jet Propulsion Laboratory's open-source autonomous Repeat Image Feature Tracking (autoRIFT) algorithm, which is automatically applied to all Landsat image pairs available within the CAA for a given year. This scheme uses a normalized cross-correlation function to find matches in the panchromatic band between reference and search images, with full details of the processing scheme provided by Gardner et al. (2018) and Gardner et al. (2018) state that uncertainty in surface ice velocities can be up to 20–30 m year−1, although inspection of velocities extracted for non-moving bedrock regions adjacent to Split Lake Glacier provides values that are generally <10 m year−1.
In this study, we utilize the annual ITS_LIVE velocity composites (standardized to m year−1) for the Canadian High Arctic for 2000–2001 and 2004–2018, with a resolution of 240 m. The paucity of earlier Landsat data collected over the CAA, together with the lack of any Landsat data with a resolution better than 60 m prior to 1999, means that earlier ITS_LIVE velocities were patchy and unreliable, so instead we used a manual method to derive velocities over the lower terminus for this period (see the next section). For the period between 2000 and 2009, reliable results from the ITS_LIVE record are mainly restricted to the lower portions of the glacier, with frequent data gaps in snow-covered upper elevation regions where there are few well-defined features that can be tracked. After the launch of Landsat-8 in 2013, there is much better data coverage over Split Lake Glacier, resulting in velocities becoming more continuous even to upper elevation regions.
When comparing the velocity data sets, it is possible that there may be systematic variations in velocities from the same “year”. This arises because the ITS_LIVE data set is assembled from images collected outside of the polar night (i.e., spring, summer, and fall), while the Radarsat velocities are derived from images collected during the winter (November to April). A previous comparison of glacier velocities across the CAA derived from winter Radarsat imagery versus summer Landsat imagery suggested an average summer speed-up of 13.6%, although this was highly spatially variable (Van Wychen et al. 2016). We do not have any information about the presence or magnitude of seasonal velocity variability on Split Lake Glacier, but our focus here is on the long-term dynamics of this glacier. We therefore believe that our velocity data sets can be combined to effectively understand the long-term evolution of this ice mass given that their uncertainty margins are comparable, although we are careful to consider the potential impact of seasonal variability in our interpretations. We also note that there is a difference in resolution between our ITS_LIVE velocity products (240 m) and Radarsat velocity products (100 m), but since we are looking for broad changes over a multi-year period, we do not believe that these differences influence our interpretations.

3.2.3. Derivation of velocities over the lower terminus

Distinctive ogive banding is present across the lower terminus of the northern portion of Split Lake Glacier below the bedrock constriction (Fig. 3), which we use to map velocities along a central flow line with Landsat imagery, similar to the methodology of Wilson et al. (2016). For Landsat-7 and Landsat-8 images collected between 1999 and 2021, we derived velocities using an automated normalized cross-correlation algorithm implemented in Matlab, using 300 m × 300 m image chips centred along the flow line from approximately 1-year separated images (with final displacements standardized to m year−1). This methodology is similar to that used for ITS_LIVE and the velocities were consistent for overlapping periods, but our velocity derivations provide higher resolution information (approximately every 150 m along the flowline) with lower errors (approximately ±5 m year−1) due to the use of quality controlled image pairs.
Fig. 3.
Fig. 3. Images of the northern terminus of Split Lake Glacier showing changes in ogive banding over time: (a) Air photo, 1959–07–26. (b) Air photo, 1975–08–06. For (a) and (b), the blue line indicates terminus extent in 1993 and the magenta line indicates terminus extent in 2021. (c) Landsat-5, 1993–07–09. (d) Landsat-8, 2021–08–28; For (c) and (d), spacing of black dots show yearly displacements from Figure 5, year labels illustrate approximately when the ogives originated from their source at the base of the ice fall, and red labels represent distance from the terminus in 2018.
For Landsat-5 imagery, the 60 m resolution was too low to obtain reliable results using our automated cross-correlation algorithm over a 1-year period, although it was possible to visually match the same ogive bands in images with a 2-year separation. We therefore used manual tracking of the displacement of individual ogive band boundaries in 2-year separated Landsat-5 images from summer 1991 and 1993 to produce mean annual velocities along the central flow line, with an estimated error of ±10 m year−1. This provides information for a period when ITS_LIVE coverage is unavailable. Unfortunately, no other cloud-free, summer image pairs with sufficient resolution were available to use for feature matching during the 1990s or earlier.

3.3. Digital elevation data

3.3.1. Digital elevation models differencing

Until the start of the 21st century, high-quality digital elevation models (DEMs) were sparse in the CAA, especially ones from which reliable DEM differencing could be conducted to quantify glacier changes. However, the release of the ArcticDEM (Porter et al. 2018), as well as the recent work of Hugonnet et al. (2021), have provided a number of high-quality DEM and pre-derived surface elevation change products that now enable us to characterize surface elevation change for Split Lake Glacier over the last two decades. First, we utilized the pre-derived surface elevation change products at 100 m resolution provided by Hugonnet et al. (2021) to quantify elevation change for Split Lake Glacier over the periods 2000–2004, 2005–2009, and 2010–2014. These surface elevation changes were derived from differencing of DEMs primarily derived from ASTER stereo satellite images, with the full methodology described in Hugonnet et al. (2021). We obtained the Canadian Northern Arctic data products, and their associated uncertainly values, from the Theia data portal (, with all data being openly available under the International Creative Commons License 4.0.
To quantify elevation changes for the latter part of our study period (2012–2017), we utilized two ArcticDEM StripMap products with 2 m resolution from March 13, 2012, and March 23, 2017. These were identified from Release 7 of the index of StripMap footprints provided by the Polar Geospatial Centre, University of Minnesota (; Porter et al. 2018). These DEMs provide good spatial overlap for Split Lake Glacier, and the 5 years elapsed between them allows for detectable surface change to occur. To reduce noise, we resampled both StripMap DEMs from their original 2 m resolution to a 5 m resolution using a nearest neighbour approach in ESRI ArcGIS Desktop version 10.8.1. These resampled DEM products were then co-registered using the approach of Berthier et al. (2007), which utilizes an open source ENVI/IDL workflow ( which shifts the secondary DEM into an xy, and z direction relative to the primary DEM. To quantify uncertainty, we calculated the mean difference between our reference (March 13, 2012) and secondary (March 23, 2017) ArcticDEMs over stable bedrock regions adjacent to the glaciers. This analysis of nearly 27 000 000 points over bedrock yielded a mean difference of −0.03 m (standard deviation of 2.5 m) between the DEMs. As such, we use the standard deviation of 2.5 m to bound our uncertainty.
Taken together, the DEM differencing of the ASTER and ArcticDEM products allows us to quantify the geometry change of Split Lake Glacier for four distinct epochs: 2000–2004, 2005–2009, 2010–2014, and 2012–2017. Collectively, these measurements allow us to characterize the surface elevation change over a nearly two decade long period, which overlaps with our record of glacier-wide velocity change.

3.3.2. ICESat altimetry

To further explore the geometry change of Split Lake Glacier, we also use ICESat laser altimetry observations of the glacier surface between 2003–2009 (obtained by ICESat-1) and 2018–2021 (obtained by ICESat-2), all obtained from Due to the relatively limited number of observations from overlapping tracks within the glacier basin, we did not compare individual ICESat observations with each other. Rather, we compared all of the ICESat surface elevations with the ArcticDEM StripMap product from March 13, 2012 and use this as our reference from which surface elevation change is calculated. All surface elevation change estimates are standardized to values of m year−1.
To assess the uncertainty between the ICESat and ArcticDEM data, we compared surface elevations over stationary bedrock areas adjacent to Split Lake Glacier for both data sets. Using this approach, we found a median difference of 0.18 m year−1 for the ICESat-2 observations based on ∼29 000 point comparisons. For ICESat-1, for which there are much less data available, there is a median difference of −0.08 m year−1 based on 650 observations. In both instances, there is an agreement between the elevations at overlapping locations in the ICESat-1/2 and ArcticDEM observations, meaning that these data sets provide an effective way of monitoring surface geometry changes over time.

4. Results

4.1. Changes in terminus area

The Split Lake Glacier basin in 2003 covered a total area of 437.8 km 2 (Table 2a). The results here focus on the surface area changes within the boxes outlined in Fig. 2a. Between 1959 and 2020, the surface area of Split Lake Glacier within the black box expanded from 15.18 ± 0.31 to 18.29 ± 0.49 km2 (Fig. 2d, Table 2b). The greatest rate of increase occurred between 1959 and 1975, by a total of 1.89 ± 1.34 km2, with the terminus area continuing to increase after that from 17.08 ± 1.04 km2 in 1987 to 18.29 ± 0.49 km2 in 2020 (Table 2b).
Table 2.
Table 2. (a) Area, RGI Id, and coordinates of each glacier basin on the western margin of the Prince of Wales Icefield. Basin # refers to the basin numbers from Fig. 1c. (b) Change in terminus area as shown in Fig. 2 over the study period for Box 4, Box 3, Split Lake Glacier, and Box 2. Total is the difference in area between the first and last year area change values.
Between 1975 and 2020, the area of the glacier termini in Box 3 decreased from 0.72 ± 0.21 km2 to 0.16 ± 0.03 km2 (Fig. 2c, Table 2b) and in Box 4 it decreased from 1.39 ± 0.06 km2 to 0.70 ± 0.03 km2 (Fig. 2b). The glacier in Box 2 had small fluctuations in its terminus position over the study period, although most of these were within error limits and indicate that this glacier has remained generally stable since 1959 (Fig. 2d, Table 2b).
Inspection of the air photos and satellite imagery of the lower terminus (Fig. 3) illustrates how this region has dramatically changed over time. In 1959 the terminus was in its most retreated position, with a generally smooth surface (Fig. 3a). By 1975, the terminus had advanced by almost 1 km along parts of its northern and eastern margins, and the glacier surface had become extensively crevassed, with ogives now present near the base of the icefall (Fig. 3b). The terminus had continued to expand by 1993 (Fig. 3c), and further by 2021 (Fig. 3d), with the ogives becoming much more extensive. Comparison of the crevasse patterns in and above the icefall using the 15 m panchromatic band from Landsat 7 and Landsat 8 imagery (Fig. 1e and 1f) shows that in the 18 years between the observations the crevasses became more extensive and extended to higher elevations.

4.2. Velocity observations

Figure 4 provides an overview of surface velocities for Split Lake Glacier since 2000 from both the ITS_LIVE and Radarsat records. The region of elevated velocities corresponds with a highly crevassed section of the glacier that is evident in Fig. 1b, with this portion of the glacier steepest along the centreline, descending from 500 m a.s.l. to 100 m a.s.l. between the top and bottom of the constriction point (Fig. 1d). Up-glacier of where the glacier connects to the broader POW Icefield, velocities are generally low (<50 m year−1) and indicative of ice frozen to its bed, meaning that ice deformation is the process governing motion. In the lower terminus region, velocities are also generally low (<100 m year−1).
Fig. 4.
Fig. 4. (a) Surface velocities of Split Lake Glacier from the ITS_LIVE and Radarsat (indicated with *) records, standardized to values of m year−1 (background image is a Landsat-8 image acquired 2019–09–03; UTM Zone 17 N); (b) Centreline velocities of Split Lake Glacier extracted from the Radarsat record; (c) Centreline velocities of Split Lake Glacier extracted from the ITS_LIVE record. The location of the centreline is shown in Fig. 1b.
Figures 4b and 4c provide the centreline velocity profile (extracted along the centreline shown in Fig. 1b) from both the ITS_LIVE and Radarsat data sets for 2000–2020. In the year 2000, both the ITS_LIVE and Radarsat data sets indicate that the middle portion of Split Lake Glacier was flowing at speeds of 200–300 m year−1. However, by 2001 (Fig. 4c), velocities in the same section had increased to up to 500 m year1, and from 2005 to 2020 they were consistently 400–550 m year−1.
Figure 5 illustrates the changes in velocity along the central flow line over the lower terminus. Individual annual velocities derived between 2015 and 2021 varied little, so the mean is plotted as a single line with the vertical bars indicating standard deviation. All velocity curves show similar patterns of longitudinal variability, but differences in magnitude between periods. Velocities in 1992 and 2015–2021 were similar, but values were significantly lower between 1999 and 2002.
Fig. 5.
Fig. 5. Surface velocity of Split Lake Glacier (standardized to m year−1) along the central flow line (location shown in Fig. 3c and 3d). Velocities since 1999 were derived using automated cross-correlation analysis on annually separated images, with the line for 2015–2021 showing the average of six individual yearly displacements. The line for 1992 was derived from the manual matching of ogive banding in images from 1991 and 1993.
From visual inspection of the Landsat imagery and pattern of annual velocities (indicated by the spacing of black dots in Fig. 3c and 3d), it is clear that each ogive band reflects an annual displacement, similar to the findings of other studies such as Wilson et al. (2016). While longitudinal compression and extension may change slightly with time along the profile, this enables the bands to be used as an indication of historical movement. The ogives originate at the bottom of the icefall that occurs in the bedrock constriction (lower right in Fig. 3), with a comparison of the image sources indicating that no ogives were present across the lower terminus in 1959 (Fig. 3a), they were starting to form at the base of the icefall in 1975 (Fig. 3b), they extended across about half of the lower terminus in 1993 (to about the 2.3 km marker in Fig. 3c), and extended across the entire terminus in 2021 (Fig. 3d). Based on this pattern and their annual nature, we estimate that the earliest identifiable band was created around 1970 (Fig. 3c). The spacing between the bands inferred to have formed between 1970 and 1992 (Fig. 3c) is slightly greater than those which have formed more recently (Fig. 3d), which suggests that the lower terminus was moving more quickly in the earlier period, although we do not have direct measurements or high enough resolution imagery to confirm this.

4.3. Elevation changes

For the four time periods that geometry changes were determined for Split Lake Glacier (2000–2004; 2005–2009; 2010–2014; 2012–2017), there is a pattern of significant surface elevation lowering (∼2 to 3 m year−1) in the upper portions of the glacier compared to modest elevation gain across the lower terminus (∼0.25 to 1 m year−1) (Fig. 6). Total surface lowering of the upper glacier was ∼50 m from 2000 to 2014, which is approximately an order of magnitude higher than adjacent areas of western POW Icefield at this elevation over the same period (c.f., Hugonnet et al. 2021). There is some evidence that the area of surface elevation loss has increased in its spatial extent recently, with the region of maximum surface lowering (areas of pink and purple) being more extensive during 2012–2017 (Fig. 6d) than in 2000–2004 (Fig. 6a).
Fig. 6.
Fig. 6. Surface elevation change of Split Lake Glacier for (a) 2000–2004, (b) 2005–2009, (c) 2010–2014 (data for (a)–(c) derived by Hugonnet et al. 2021: data are openly available under the International Creative Commons License 4.0 and via the Theia data portal, and (d) 2012–2017. The dashed black line indicates the approximate location of the dynamic balance line. Background image (a)(d) is a Landsat-8 Image acquired 2019–09–03; UTM 17 N.
The ICESat altimetry analysis (Fig. 7) shows similar patterns of surface elevation change to those found for the DEM differencing analysis (Fig. 6). The ICESat-1 results (Fig. 7a and 7b), although limited in their spatial coverage, indicate that there was a high degree of thinning (∼3 and 4 m year−1) between 2003/2009 and 2012 in the upper region of Split Lake Glacier, while the lower terminus region experienced modest (0.25–1 m year−1) surface elevation gain over this period. The ICESat-2 results (Fig. 7c and 7d) provide better spatial coverage and indicate upper elevation thinning rates of 2–4 m year−1 and lower elevation thickening rates of 0.5–1.75 m year−1 across the terminus between 2012 and 2018/2021.
Fig. 7.
Fig. 7. Surface elevation change of Split Lake Glacier based on differencing of (a) 2003–2009 ICESat-1 surface elevations from a March 13, 2012 ArcticDEM StripMap product; (c) 2018–2021 ICESat-2 surface elevations differenced from the same March 13, 2012 ArcticDEM StripMap product. (b) and (d) show the elevation changes over a larger region than (a) and (c), respectively, to show the thinning rates within the context of adjacent areas. All rates standardized to values of m year−1. Background image (a)–(d) is a Landsat-8 Image acquired 2019–09–03; UTM 17 N.
All of our geometry change observations (Figs 6 and 7) consistently indicate significant surface lowering in the upper elevations of Split Lake Glacier. This region is centred just upstream of the most heavily crevassed portion of the glacier trunk in the bedrock constriction. When this region is compared with the surrounding icefield, it is clear that the rates of change are distinctly higher than the more modest and gradual surface lowering (0.25–0.5 m year−1) commonly observed over adjacent areas (Figs 7b and 7d). To further quantify the surface elevation changes of Split Lake Glacier with those of adjacent areas, we calculated the mean elevation change in 100 m elevation bins for each of the four epochs presented in Fig. 6. This indicates that the geometry change of Split Lake Glacier is anomalous when compared with adjacent areas of western POW Icefield (Fig. 8). For example, in all epochs the lowermost portion of Split Lake Glacier (0–100 and 100–200 m elevation bins) is increasing in elevation, while adjacent glaciers are decreasing in surface height in the same elevation range. Similarly, at the higher parts of Split Lake Glacier (400–500 m, 500–600 m, and 600–700 m elevation bins) the glacier is thinning at rates that are approximately double those observed in the same elevation bins on adjacent areas.
Fig. 8.
Fig. 8. Mean thinning rates within 100 m elevation bins on Split Lake Glacier (red lines) and adjacent areas of western Prince of Wales Icefield (blue lines) for each epoch presented in Fig. 6. The area used to calculate 100 m elevation bins on Split Lake Glacier is shown in Fig. 1b.

5. Discussion

The general pattern of slower motion over the upper accumulation and lower ablation areas of Split Lake Glacier, and the fastest motion occurring as the ice is funnelled through the bedrock constriction (located ∼5 to 10 km along the centreline shown in Fig. 1b), is similar to other studies that have derived ice motion within the region (Millan et al. 2017; Strozzi et al. 2017; Sánchez-Gámez and Navarro 2017). However, our results indicate that the geometry and area changes of Split Lake Glacier are anomalous when compared to nearby regions of POW Icefield. Regarding area change, the terminus of Split Lake Glacier has continued to gradually expand after the rapid 1959 to 1975 advance of the northern terminus (Fig. 2; Table 2). In contrast, two land terminating glaciers at similar elevations to the north have both been experiencing retreat since 1970. Indeed, observations from the regional mapping of terminus positions beyond the study area indicate that retreating glacier termini is generally the norm across the CAA (e.g., Cook et al. 2019; Derksen et al. 2019), suggesting that the continuous advance of Split Lake Glacier is atypical.
With regard to geometry changes, the rates of surface lowering in the upper elevations (400–700 m a.s.l.) of Split Lake Glacier are significantly higher than the rates of surface lowering across adjacent glaciers along the western margin of POW Icefield (Figs. 6–8). Modelling by Noël et al. (2018) indicates that the average annual SMB of the upper elevation of Split Lake Glacier (pink/purple area in Fig. 6) was −0.64 ± 0.16 m w.e. yr−1 between 1996 and 2015, compared to our measurements of mean surface lowering of 2–4 m yr−1 in this region over the period 2000–2014 (Fig. 6). In the terminus region below the bedrock constriction, the SMB results of Noël et al. (2018) indicate that there should be significant thinning (average −2.30 ± 0.16 m w.e. year−1), whereas the elevation changes show a mean surface thickening of 0.25–1 m year−1 (Figs 68). The dynamic balance line (mean elevation of long-term zero elevation change) on Split Lake Glacier occurred at approximately 280–300 m a.s.l. in 2012–2017 (Figs 6d, 8), well below the long-term ELA of 980 m a.s.l. from measured and modelled SMB (Mair et al. 2009; Noël et al. 2018). This discrepancy between modelled SMB and observed surface elevation changes, and the offset between the dynamic balance line and equilibrium line, implies that glacier dynamics must be responsible for the anomalous geometry changes of Split Lake Glacier.

5.1. Evidence in support of a slowly surging glacier

Surge-type glaciers are characterized by ice motion that typically oscillates between short periods (months to years) of fast flow which can be several orders of magnitude greater than the balance velocity, followed by prolonged periods of slow flow (years to decades) that is at least an order of magnitude below the expected balance velocity if the glacier was experiencing normal flow (Jiskoot 2011). Surge-type glaciers were first identified in the northern CAA in the 1960s (Hattersley-Smith 1964,1969; Müller 1969), but it was not until decades later that the first major catalogue of surge-type glaciers in this region was completed by Copland et al. (2003). In this catalogue, a total of 51 glaciers were classified as a surge type, with 15 observed in the active phase in 1959/60 and (or) 1999/2000. Similarly, Sevestre and Benn (2015) identified a total of 46 glaciers as a surge type in the CAA based on a comprehensive literature review, compared to 322 in Alaska/Yukon and 345 in Svalbard.
Copland et al. (2003) classified two glaciers as surge types along the western margin of POW Icefield, but not Split Lake Glacier. Sevestre and Benn (2015) also did not classify Split Lake Glacier as surge type. Taggart Lake Glacier (unofficial name), located 80 km north of Split Lake Glacier (Fig. 1c), was identified by Copland et al. (2003) as actively surging into a proglacial lake in 1999 Landsat imagery, with shear margins, looped surface moraines, and extensive surface crevassing over the lower part of the glacier and a 3 km advance of the terminus between 1959 and 1999. Another glacier near Kooetigeto River, located 35 km north of Taggart Lake, was classified as a surge type based on large moraine loops, although it was not actively surging in 1959 or 1999. Copland et al. (2003) identified a further three glaciers on eastern POW Icefield as confirmed surge type, and one as possibly surge type, but none of these were near Split Lake Glacier. However, a recent automated analysis of SAR backscatter signatures in 2018–2019 Sentinel-1 imagery identified Taggart Lake Glacier as actively surging (Leclercq et al. 2021), one of only two such glaciers with this classification in the CAA over that period.
The changes in surface elevation and terminus extents reported here are consistent with what would be expected during the active phase of a glacier surge (Meier and Post 1969; Jiskoot 2011): thinning of the upper reservoir area, thickening of the lower receiving area, and sometimes terminus advance. Although we do not have comprehensive velocity maps available prior to the year 2000, our manual tracking of ogive banding over the lower terminus suggests that glacier motion was significantly higher in 1991–1993 than during 1999–2002, before increasing again towards the present day (Fig. 5). Long-term changes in the presence and distribution of ogives across the lower terminus of Split Lake Glacier suggest that rapid flow started there in about 1970.
For the period when comprehensive velocity maps are available, the Radarsat-derived maximum flow speeds in 2000 were 340 ± 15 m year1 compared to maximum speeds of 625 ± 15 m year1 in 2015 (Fig. 4a). The input imagery used to create the Radarsat velocity record was all acquired during the winter (November to May of each year, Table 1) and as such, unlikely to be influenced by increases in ice motion induced by meltwater transmission to the bed, meaning that seasonal flow variations cannot explain the observed changes. Additionally, when comparing velocities across POW Icefield (Van Wychen et al. 2014, 2016, 2021), it is clear that Split Lake Glacier is moving ∼200 m year−1 faster than nearby glaciers, suggesting that its motion is enhanced by internal dynamics and out of balance with current SMB.
If Split Lake Glacier is indeed undergoing a surge, the results presented here are most aligned with the interpretation that it is undergoing a slow surge (Jiskoot 2011). In the case of slowly surging glaciers, the surge phase is expected to last 20 years or more with flow rates 5–10 times higher than those observed during the quiescent phase, while the glacier front experiences relatively little to no positional change (Jiskoot 2011). As a comparison, Frappé and Clarke (2007) detailed the slow surge of Trapridge Glacier (Yukon) and found that the active phase occurred from around 1980 to around 2000, although the velocities only increased from pre-surge rates of 16 m year−1 to a maximum of 42 m year−1 in 1984 (see fig. 11 in Frappé and Clarke 2007). Despite these slight variations in velocities, the prolonged period that they occurred over (20 years) led to the advance of the terminus of Trapridge Glacier. The cause of the apparent slowdown of Split Lake Glacier over the period 1999–2002 is unknown, but the velocities during this period were still elevated, the ogive patterns are continuous, and the terminus continued to advance (Fig. 2), suggesting that the slow surge has occurred continuously since the 1970s. Slow surges have also been identified in Svalbard (Sund et al. 2009), but have not been previously described for the CAA. However, the area, geometry, and velocity changes we describe here are consistent with those reported for slowly surging glaciers in other geographic regions, so we believe it likely that Split Lake Glacier falls within this classification.

5.2. Comparisons with surges on other glaciers

The evidence presented above suggests that Split Lake Glacier started surging around 1970 and that the surge is continuing at the present day. This is a long active phase compared to surge-type glaciers found in mid-latitude regions (Jiskoot 2011), but is not entirely without precedence for the CAA. For example, Van Wychen et al. (2016,2021) found that the Middle Glacier of Axel Heiberg Island has had elevated flow velocities lasting over a decade, while velocity maps for Mittie Glacier of southern Ellesmere Island suggest a surge duration of 12 years (Copland et al. 2003; Van Wychen et al. 2016; 2021; Millan et al. 2017). Similarly, Otto Glacier located on Northern Ellesmere Island was found to be surging from ∼2000 to ∼2014, although its flow speed increased and decreased over the ∼14-year long active phase, and the areas of elevated velocities propagated spatially over this time (Van Wychen et al. 2016; 2021). This is in contrast to Split Lake Glacier, where increased flow speeds have remained relatively high throughout the entire active phase and are spatially confined.
Perhaps the closest analogue is that of Good Friday Glacier on Axel Heiberg Island, which has experienced sustained terminus advance and high flow speeds since the 1950s, although it is arguable whether this meets the traditional definition of a surge (Medrzycka et al. 2019). For both Good Friday and Split Lake glaciers, the observed terminus advance and elevated surface velocities have been at odds with the behaviour of nearby glaciers and occurred within the wider context of sustained regional mass loss. Medrzycka et al. (2019) suggested that the behaviour of the Good Friday Glacier might be driven by a delayed response to a positive mass balance condition of the Little Ice Age. However, this is unlikely to explain the behaviour of Split Lake Glacier given that adjacent glacier basins of a similar size, elevation range, and aspect on western POW Icefield are not behaving in the same way. Van Wychen et al. (2016,2017), as well as Medrzycka et al. (2019), have postulated that perturbations in basal topography can cause variable dynamic behaviours for glaciers within the CAA. It is suggested that these perturbations cause localized pinning points that are responsible for driving dynamic fluctuations. However, we do not have observations of the basal topography of Split Lake Glacier to investigate whether this might provide a likely explanation here. It is possible, however, that the narrow bedrock constriction through which ice is channelled from the icefield to the glacier terminus acts as a lateral pinning point that modulates ice motion in the same way that basal pinning points have been suggested by earlier studies (Van Wychen et al. 2016, 2017; Medrzycka et al. 2019), although more intensive field measurements of basal and lateral topography are necessary to determine if this is truly the case.
In other regions, nearby glaciers have been shown to experience contrasting behaviours. A good example of this is found in Alaska, where Taku Glacier (tidewater terminating) and Lemon Creek Glacier (land terminating) are separated by about 30 km but experienced differing behaviours over the period 1946–2018 (McNeil et al. 2020). For much of this period, Taku Glacier experienced a positive mass balance and terminus advance, while Lemon Glacier experienced mass loss and terminus retreat. It was only after 2013 that Taku Glacier began to experience thinning and terminus retreat. In this case, it was found that the divergence in response was related to distinct differences in local climate and glacier hypsometry. It is possible that the changes observed on Split Lake Glacier are also a manifestation of localized climate and hypsometry conditions that allow for its terminus to respond differently to other regions along the western margin of POW Icefield. This explanation might explain why the small glacier to the south of Split Lake Glacier underwent little significant change over the observation period (Box 2 in Fig. 2). However, if there was a localized positive mass balance situation Split Lake Glacier would be expected to undergo widespread thinning in the 400–700 m elevation bin (Fig. 8), rather than the observed widespread thickening. Furthermore, a long-term record of mass balance for POW Icefield, which included the collection of shallow ice cores near Split Lake Glacier, does not indicate any evidence of anomalous positive SMB in this portion of the icefield (Mair et al. 2009).

6. Conclusions

Here we used a combination of remote sensing methods and data sets to characterize changes in glacier surface elevation, terminus area, and surface velocities of Split Lake Glacier located on the western margin of the POW Icefield in the Canadian High Arctic. Collectively, the velocity data sets and ogive analysis suggest that Split Lake Glacier started surging around 1970, with relatively higher velocities in 1971–1973 than in the late 1990s and early 2000s. Furthermore, since 2004, accelerated ice motion has continued until at least winter 2021. However, the region of higher velocities has remained spatially restricted to a relatively narrow portion of the glacier trunk focused around the location of an icefall, with limited evidence of either up-glacier or down-glacier propagation. Coincident with this speed-up, there has been the transport of mass between the upper elevation reservoir area to the lowermost terminus receiving area, with the dynamic balance line near the top of the bedrock constriction at an elevation of 280–300 m a.s.l. The surface elevation changes have been significantly different from that which can be attributed to SMB, with thinning in the upper elevation region over the last two decades of up to 2–3 m year−1 greater than expected surface lowering of 0.64 ± 0.16 m w.e. year−1 predicted from SMB modelling (Noël et al. 2018). Similarly, surface elevation gain (0.25–1 m year−1) over the glacier terminus is markedly different from the modelled SMB at similar elevations of −1.01 to −2.10 m w.e. year−1 (Noël et al. 2018). As such, glacier dynamics are likely responsible for driving these geometry changes.
In addition, our results indicate that the changes we describe for Split Lake Glacier are anomalous when compared to other nearby glaciers on the western margin of POW Icefield. It seems most likely that Split Lake Glacier is currently undergoing a slow surge, and our surface elevation change and terminus area analysis results are consistent with that explanation. If this is the case, it would be the first time a slow surge has been reported in the CAA. However, because the surge characteristics described here are somewhat distinct from any other glacier in the CAA, and because we lack in situ observations to further investigate the processes governing the observed variations in flow speed, we cannot rule out that other mechanisms might be driving the dynamic changes. Aspects of our results are similar to those reported for Good Friday Glacier (Medrzycka et al. 2019) and Taku Glacier (McNeil et al. 2020), where the geometry, velocity, and area changes of those glaciers contrast with those located nearby. The results reported here provide another layer of nuance to the dynamic behaviour of glaciers within the CAA and provide further evidence of the need to think widely in terms of the types of geometry, dynamics, and terminus area oscillations that glaciers can undergo within this region. Furthermore, this work shows the richness of combining a variety of remote sensing data sets to identify and characterize glacier change, particularly in locations where field campaigns are limited.


We gratefully acknowledge support from the Natural Sciences and Engineering Research Council of Canada (Discovery Grant), the Canada Foundation for Innovation (John Evan's Leadership Fund), Environment and Climate Change Canada (Climate Research Division), ArcticNet Network of Centres of Excellence Canada, and the University of Waterloo and the University of Ottawa. The authors also sincerely thank Dr. David Burgess of Natural Resources Canada for aid in the acquisition of Radarsat-2 imagery used in this work. Finally, the authors thank the editors of Arctic Science, particularly Derek Mueller, and the comments from three anonymous reviewers who provided valuable input which helped improve this work.


Berthier E., Arnaud Y., Kumar R., Ahmad S., Wagnon P., Chevallier P. 2007. Remote sensing estimates of glacier mass balances in the Himachal Pradesh (Western Himalayas, India). Remote Sensing of the Environment, 108(3): 327–338.
Cook A., Copland L., Noël B., Stokes C., Bentley M., Sharp M., 2019. Atmospheric forcing of rapid marine terminating glacier retreat in the Canadian Arctic Archipelago. Science Advances, 5(3): 1–10.
Copland L., Sharp M., Dowdeswell J. 2003. The distribution and flow characteristics of surge-type glaciers in the Canadian High Arctic. Annals of Glaciology, 36: 73–81.
Derksen C., Burgess D., Duguay C., Howell S., Mudryk L., Smith S., 2019. Changes in snow, ice, and permafrost across Canada. In Canada's changing climate report. Edited by E. Bush, D.S. Lemmen. Government of Canada, Ottawa. pp.194–260.
Frappé T.P., Clarke G.K.C. 2007. Slow surge of Trapridge Glacier, Yukon Territory, Canada. Journal of Geophysical Research: Earth Surface. 112(F3): 1–17.
Gardner A.S., Fahnestock M.A., Scambos T.A. 2021. ITS_LIVE regional glacier and ice sheet surface velocities. Data archived at National Snow and Ice Data Center, Boulder, Colorado.
Gardner A.S., Moholdt G., Scambos T., Fahnestock M., Ligtenberg S., van den Broeke M., 2018. Increased west antarctic and unchanged east antarctic ice discharge over the last 7 years. The Cryosphere, 12(2): 521–547.
Gray A.L., Mattar K.E., Sofko G. 2000. Influence of ionospheric electron density fluctuations on satellite radar interferometry. Geophysical Research Letters, 27(10): 1451–1454.
Harcourt W.D., Palmer S.J., Mansell D.T., Le Brocq A., Bartlett O., Gourmelen N., 2020. Subglacial controls on dynamic thinning at Trinity-Wykeham Glacier, Prince of Wales Icefield, Canadian Arctic. International Journal of Remote Sensing, 41(3): 1191–1213.
Hattersley-Smith G. 1964. Rapid advance of glacier in northern Ellesmere Island. Nature, 201(4915): 176.
Hattersley-Smith G. 1969. Recent observations of surging Otto Glacier, Ellesmere Island. Canadian Journal of Earth Science, 6(4): 883–889.
Hugonnet R., McNabb R., Berthier E., Menounos B., Nuth C., Girod L., 2021. Accelerated global glacier mass loss in the early twenty-first century. Nature, 592: 726–731.
Jiskoot H., Singh V.P., Singh P., Haritashya U.K. 2011. Glacier surging. In Encyclopedia of snow, ice and glaciers. Edited by. Springer, Dordrecht. pp. 415–428.
Krumwiede B.S., Kamp U., Leonard G.J., Kargel J.S., Dashtseren A., Walther M. 2014. Recent glacier changes in the Mongolian Altai mountains: case studies from Munkh Khairkhan and Tavan Bogd. In Global land ice measurements from space. Edited by J. Kargel, G. Leonard, M. Bishop, A. Kääb. Springer, Berlin, pp.481–508.
Leclercq P.W., Kaab A., Altena B. 2021. Brief communication: detection of glacier surge activity using cloud computing of Sentinel-1 radar data. The Cryosphere, 15(10): 4901–4907.
Lenaerts J.T.M., Van Angelen J.H., Van Den Broeke M.R., Gardner A.S., Wouters B., Van Meijgaard E. 2013. Irreversible mass loss of Canadian Arctic archipelago glaciers. Geophysical Research Letters, 40(5): 870–874.
Mair D., Burgess D., Sharp M., Dowdeswell J.A., Benham T., Marshall S., 2009. Mass balance of the Prince of Wales icefield, Ellesmere Island, Nunavut, Canada. Journal of Geophysical Research Earth Surface. 114(F2): F02011.
McNeil C., O'Neel S., Loso M., Pelto M., Sass L., Baker E.H., 2020. Explaining mass balance and retreat dichotomies at Taku and Lemon Creek Glaciers, Alaska. Journal of Glaciology, 66(258): 530–542.
Medrzycka D., Copland L., Van Wychen L., Burgess D. 2019. Seven decades of uninterrupted advance of Good Friday Glacier, Axel Heiberg Island, Arctic Canada. Journal of Glaciology, 65(251): 440–452.
Meier M.F., Post A. 1969. What are glacier surges? Canadian Journal of Earth Sciences, 6(4): 8907–8917.
Millan R., Mouginot J., Rignot E. 2017. Mass budget of the glaciers and ice caps of the Queen Elizabeth Islands, Canada, from 1991 to 2015. Environmental Research Letters, 12(2): 024016.
Moon T., Joughin I. 2008. Changes in ice front position on Greenland's outlet glaciers from 1992 to 2007. Journal of Geophysical Research: Earth Surface, 113(F2): F02022.
Müller F. 1969. Was the Good Friday Bay Glacier on Axel Heiberg Island surging? Canadian Journal of Earth Science, 6(4): 891–894.
Noël B., van de Berg W.J., Lhermitte S., Wouters B., Schaffer N., van den Broeke M.R. 2018. Six decades of glacial mass loss in the Canadian Arctic archipelago. Journal of Geophysical Research: Earth Surface, 123(6): 1430–1449.
Pfeffer W.T., Arendt A.A., Bliss A., Bolch T., Cogley J.G., Gardner A.S., 2014. The Randolph Glacier Inventory: a globally complete inventory of glaciers. Journal of Glaciology, 60(221): 537–552.
Porter C., Morin P., Howat I., Noh M., Bates B., Peterman K., 2018. ArcticDEM. Available from, Harvard Dataverse, V1 [Accessd January 2020].
Sánchez-Gámez P., Navarro F. 2017. Glacier surface velocity retrieval using D-InSAR and offset tracking techniques applied to ascending and descending passes of Sentinel-1 data for southern Ellesmere ice caps, Canadian Arctic. Remote Sensing, 9(5): 442.
Schellenberger T., Van Wychen W., Copland L., Kääb A., Gray L. 2016. An inter-comparison of techniques for determining velocities of maritime Arctic glaciers, Svalbard, using Radarsat-2 wide fine mode data. Remote Sensing, 8(9): 785.
Sevestre H., Benn D.I. 2015. Climatic and geometric controls on the global distribution of surge-type glaciers: implications for a unifying model of surging. Journal of Glaciology, 61(228): 646–662.
Strozzi T., Paul F., Wiesmann A., Schellenberger T., Kääb A. 2017. Circum-Arctic changes in the flow of glaciers and ice caps from satellite SAR data between 1990s and 2017. Remote Sensing, 9(9): 947.
Sund M., Eiken T., Hagen J.O., Kääb A. 2009. Svalbard surge dynamics derived from geometric changes. Annals of Glaciology, 50(52): 50–60.
Tepes P., Gourmelen N., Nienow P., Tsamados M., Shepherd A., Weissgerber F. 2021. Changes in elevation and mass of Arctic glaciers and ice caps, 2010-2017. Remote Sensing of Environment, 261: 112481.
Thomson L., Zemp M., Copland L., Cogley G., Ecclestone M. 2017. Comparison of geodetic and glaciological mass budgets for White Glacier, Axel Heiberg Island, Canada. Journal of Glaciology, 63(237): 55–66.
Van Wychen W., Burgess D., Kochtitzky W., Nikolic N., Copland L., Gray L. 2021. Radarsat-2 derived glacier velocities and dynamic discharge estimates for the Canadian High Arctic: 2015-2020. Canadian Journal of Remote Sensing, 46(6): 695–714.
Van Wychen W., Burgess D.O., Gray L., Copland L., Sharp M., Dowdeswell J., 2014. Glacier velocities and dynamic ice discharge from the Queen Elizabeth Islands, Nunavut, Canada. Geophysical Research Letters, 41(2): 484–490.
Van Wychen W., Copland L., Gray L., Burgess D., Danielson B., Sharp M. 2012. Spatial and temporal variation of ice motion and ice flux from Devon ice cap, Nunavut, Canada. Journal of Glaciology, 58(210): 657–664.
Van Wychen W., Davis J., Burgess D.O., Copland L., Gray L., Sharp M., 2016. Characterizing interannual variability of glacier dynamics and dynamic discharge (1999-2015) for the ice masses of Ellesmere and Axel Heiberg Islands, Nunavut, Canada. Journal of Geophysical Research Earth Surface, 121(1): 39–63.
Van Wychen W., Davis J., Copland L., Burgess D.O., Gray L., Sharp M., 2017. Variability in ice motion and dynamic discharge from Devon ice cap, Nunavut, Canada. Journal of Glaciology, 63(239): 436–449.
Wilson R., Mernild S.H., Malmros J.K., Bravo C., Carrion D. 2016. Surface velocity fluctuations for Glaciar Universidad, central Chile, between 1967 and 2015. Journal of Glaciology, 62(235): 847–860.

Information & Authors


Published In

cover image Arctic Science
Arctic Science
Volume 8Number 4December 2022
Pages: 1288 - 1304


Received: 23 July 2021
Accepted: 26 April 2022
Accepted manuscript online: 2 June 2022
Version of record online: 30 August 2022

Key Words

  1. glacier surging
  2. slow surge
  3. Split Lake Glacier
  4. Prince of Wales Icefield
  5. glacier dynamics


  1. montée des glaciers
  2. montée lente
  3. glacier du lac Split
  4. champ de glace Prince-de-Galles
  5. dynamique des glaciers



Department of Geography and Environmental Management, University of Waterloo, Waterloo, ON N2L 3G1, Canada
Danielle A.M. Hallé
Department of Geography and Environmental Management, University of Waterloo, Waterloo, ON N2L 3G1, Canada
Luke Copland
Department of Geography, Environment and Geomatics, University of Ottawa, Ottawa, ON K1N 6N5, Canada
Luke Copland served as an Associate Editor at the time of manuscript review and acceptance; peer review and editorial decisions regarding this manuscript were handled by Derek Mueller and Melissa Lafrenière.
Laurence Gray
Department of Geography, Environment and Geomatics, University of Ottawa, Ottawa, ON K1N 6N5, Canada

Metrics & Citations


Other Metrics


Cite As

Export Citations

If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.

There are no citations for this item

View Options

View options


View PDF

Get Access





Share Options


Share the article link

Share on social media