Rates and processes of aeolian soil erosion in West Greenland

Abstract

In arid landscapes across the globe, aeolian processes are key drivers of landscape change, but arid Arctic regions are often overlooked. In the Kangerlussuaq region of West Greenland, strong katabatic winds have removed discrete patches of soil and vegetation, exposing unproductive glacial till and bedrock. Although lake-sediment records suggest that landscape destabilization began approximately 1000 years ago, the upland soil erosion has never been directly dated. We use a novel application of lichenometry to estimate the rates and timing of soil erosion. We show that the formation of deflation patches occurred approximately 800–230 years ago, in general agreement with lake-sediment records. In West Greenland, the ‘Little Ice Age’ (AD 1350–1880) was characterized by a cold and arid climate, conditions that increased susceptibility to erosion. On average, deflation patches are expanding at a rate of 2.5 cm yr⁻¹, and variation in the rate of patch expansion cannot be explained by proximity to the Greenland Ice Sheet (GrIS), slope, aspect, elevation, or patch size. An erosional threshold exists in this aeolian system, with climate conditions necessary for patch formation likely harsher than those necessary for continued patch expansion, a result that has implications for land management in arid regions. Currently, deflation patches are expanding throughout the study region and are forming in areas close to the GrIS, but future deflation rates are dependent on projected climate and potential land-use changes. Our results stress the importance of aeolian processes in arid polar landscapes such as Kangerlussuaq, and demonstrate the use of aeolian landforms in paleoclimate reconstructions and predicting future landscape change.

Keywords

Arctic landscape change lichenometry paleoclimate wind erosion

Introduction

In arid ecosystems, aeolian processes are often drivers of geomorphology, soil nutrient cycling, and vegetation dynamics (Field et al., 2010; Ravi et al., 2011). Wind erosion is responsible for large losses of soil across arid regions, in some cases leading to irreversible shifts in ecosystem functioning (e.g. D’Odorico et al., 2012). While the drivers and feedbacks linking wind erosion to vegetation dynamics have been well studied in warm deserts (Okin et al., 2006; Schlesinger et al., 1990), arid landscapes in the Arctic are often overlooked (Bullard, 2013). Understanding past and present wind erosion dynamics is especially important in the Arctic, given the rapid environmental changes the region is currently experiencing (e.g. Hanna et al., 2012; Mernild et al., 2015).

In the Kangerlussuaq region of West Greenland, katabatic winds off the Greenland Ice Sheet (GrIS) have removed discrete patches of fine-grained soil, exposing the underlying till and bedrock. These deflation patches are mostly devoid of vegetation and are distinct from the surrounding shrub and graminoid tundra. As winds erode soils at the leading edge, or scarp, deflation patches appear to expand across the landscape. From previous remote sensing work in this region, we know that deflation patches make up roughly 22% of the terrestrial landscape, thus potentially playing an important role in vegetation dynamics and ecosystem functioning (Heindel et al., 2015). Deflation patches cover a larger percentage of the terrestrial landscape closer to the GrIS and on steep south-facing slopes, where soils are drier and more exposed to katabatic winds (Heindel et al., 2015). Either individual deflation patches expand more rapidly close to the GrIS and on south-facing slopes, or these areas are more prone to patch formation. Currently, there are no estimates of deflation-patch expansion in the literature.

In addition to a limited understanding of patch expansion rates, we also lack information about when deflation began. The few terrestrial studies of aeolian activity in the Kangerlussuaq region have mostly been confined to Sandflugtdalen, the river valley that extends from the margin of the GrIS to Søndre Strømfjord (Bullard and Austin, 2011; Dijkmans and Törnqvist, 1991; Willemse et al., 2003). Radiocarbon dates of organic material interbedded in sand and silt deposits from Sandflugtdalen suggest limited aeolian activity 3000–600 years ago, followed by a reactivation of sand sheet formation during the ‘Little Ice Age’ (AD 1350–1880; Dijkmans and Törnqvist, 1991; Willemse et al., 2003). Similarly, lake-sediment records suggest that recent landscape destabilization began approximately 1000 years ago. Sediment cores from multiple closed-basin lakes throughout the region show an increase in silt accumulation after this time (Anderson et al., 2012; Heggen et al., 2010; Olsen et al., 2013; Perren et al., 2012). With no surface inflow supplying sediment to the lakes, the authors have interpreted the increase in silt accumulation as a signal of aeolian soil erosion resulting from climate-driven landscape destabilization (Anderson et al., 2012; Heggen et al., 2010; Olsen et al., 2013; Perren et al., 2012; Rydberg et al., 2016). Prior to this study, however, there have been no attempts to directly date the upland deflation patches, and lake-sediment records lack robust terrestrial support.

In this study, we directly date the upland deflation patches and estimate patch expansion rates to test two hypotheses about soil deflation in the Kangerlussuaq region. First, we hypothesize that deflation patches were initiated by drier and windier climate conditions associated with the ‘Little Ice Age’. This hypothesis predicts that most deflation patches formed AD 1350–1880, in concert with increasing aeolian inputs into nearby lakes and the reactivation of sand sheet formation in Sandflugtdalen. Second, we hypothesize that deflation-patch formation and deflation-patch expansion are separate processes with different drivers. While we expect that rates of patch formation are related to regional climate conditions (our first hypothesis), we hypothesize that local landscape position is more important in determining rates of patch expansion. Our second hypothesis predicts that rates of patch expansion will correlate with proximity to the GrIS, aspect, and slope, with faster deflation rates on steep south-facing slopes close to the GrIS. Understanding the current rates and past dynamics of aeolian soil erosion will allow us to draw conclusions about the drivers of soil erosion and its response to a warming climate.

To test these two hypotheses, we used lichenometry, a dating technique relating the diameters of Rhizocarpon lichens to landform ages. To calculate rates of soil erosion over the timescale of a millennium and spatial scales of a few meters, lichenometry was the only dating technique available. Although the utility of lichenometry in assigning absolute numerical ages to landforms has been called into question (Osborn et al., 2015), it can be a powerful relative dating tool when used carefully (e.g. Rosenwinkel et al., 2015). Here, we used Rhizocarpon diameters in a novel way to calculate the timing and rates of soil erosion. Within a deflation patch, we predicted that lichen diameters would increase from the scarp to the trailing edge of the patch, providing estimated exposure ages that could be used to calculate patch expansion rates. Rather than relying on a few large lichen diameters to date deflation patches, we used small lichens <22 mm in diameter to calculate rates of patch expansion, which were then used to estimate the onset of soil erosion. By focusing on young lichens that still retain their circular morphology, and by comparing lichen diameters over small geographical distances, we avoided many of the issues with the technique (Osborn et al., 2015).

Regional setting

The town of Kangerlussuaq is located at the head of Søndre Strømfjord, roughly 120 km inland from the southwest coast of Greenland. Our study area extended from ~2 km west of Kangerlussuaq to the margin of the GrIS (Figure 1). Kangerlussuaq has a continental climate, with a mean summer temperature of 9.5°C, a mean winter temperature of −18.7°C, and a mean annual precipitation of 250 mm water equivalent, all for the period 1981–2011 (Hanna et al., 2012; Mernild et al., 2015).

Figure 1.

Map showing the location of the study area and Sites 1–5.

The Kangerlussuaq region was deglaciated ~7000 years ago (Levy et al., 2012; Ten Brink and Weidick, 1974). During the mid-Holocene, the GrIS retreated inland of its current margin (Young and Briner, 2015). An unvegetated moraine, located <50 m from the current GrIS margin, marks the readvance of the GrIS during historical times (Levy et al., 2012; Weidick, 1968). The exact age of this historical moraine is unknown, although records show that at least part of the moraine was being deposited by AD 1880 (Weidick, 1968). Although deglaciation occurred ~7000 years ago, soil development has occurred on a loess deposit, roughly 30 cm to 1 m thick, which mantles the region. Sandflugtdalen, the likely source region for the loess deposit, has seen two periods of heightened aeolian activity, one prior to 3500 years ago and one associated with the ‘Little Ice Age’, 550 years ago (Willemse et al., 2003). Many aspects of the loess deposit, including the exact timing and nature of the mid-Holocene loess deposition, and whether loess deposition continues today, remain poorly understood and warrant future study. The presence of the loess, a fine-grained and easily entrained sediment, is likely necessary for the formation of deflation patches (Figure 2).

Figure 2.

(a) Deflation patches dot the tundra landscape around Kangerlussuaq. These bare areas can range in size from (b) hundreds of square meters to (c) just a few square meters in size. (d) Deflation patches expand when blocks of vegetation and soil fall from the active edge or scarp.

Between Kangerlussuaq and the GrIS, the vegetation transitions from a shrub-dominated tundra to a steppe tundra (Heindel et al., 2015). Deflation patches occur on both vegetation types, with scarps undercutting both shrubs and graminoids (Heindel et al., 2015). In addition to the vegetation gradient, wind patterns also vary between Kangerlussuaq and the GrIS. Across the entire study region, easterly katabatic winds dominate, but wind speeds diminish, and wind directions conform to local topography with increasing distance from the GrIS (Bullard and Austin, 2011; Dijkmans and Törnqvist, 1991). Moist sea breezes off the fjord influence wind patterns near Kangerlussuaq but do not penetrate farther inland (Kopec et al., 2014).

Methods

Site selection

To capture the gradients in vegetation and deflation intensity across the study region, we established five study sites between Kangerlussuaq and the GrIS margin (Figure 1). Sites are numbered from east to west, with Site 1 located ~1 km from the GrIS, and Site 5 located ~20 km from the GrIS. Sites were chosen to be evenly distributed across the study region while being relatively accessible from the road, which runs the 35-km from Kangerlussuaq to the GrIS margin.

At each site, we surveyed roughly 1 km², identifying candidate deflation patches with yellow-green Rhizocarpon sp. growing on at least 10 boulders, cobbles, or regions of bedrock. From those candidate patches, we randomly selected 11–14 patches for lichenometry measurements. Close to the GrIS, where deflation patches are more common, we selected <50% of candidate patches. Farther from the GrIS, we had to select nearly 100% of candidate patches in order to keep patch numbers roughly equivalent among sites.

Lichenometry field methods

We developed our lichenometric field methods to follow those used to construct the West Greenland lichen growth curve (Forman et al., 2007; Gordon, 1981; Ten Brink, 1973; see the ‘Data analyses’ section) and to fit our specific field requirements. First, we chose not to identify to the species level, measuring all yellow-green Rhizocarpon lichens. This was the approach taken by Ten Brink (1973) and Gordon (1981) in measuring the lichens used to establish the West Greenland growth curve. In the Kangerlussuaq region, only two yellow-green Rhizocarpon species have been identified, R. geographicum and R. inarense (Hansen, 2000). Although these species may have slightly different growth rates, the fast-growing R. alpicola has not been identified in Kangerlussuaq. Second, we chose to measure the maximum lichen diameter (i.e. the long axis), following the technique described in Innes (1986). This technique allowed us to quickly and consistently measure thousands of lichen diameters. Since the majority of our lichens were <15 mm in diameter and all were circular in morphology (Figure 3), measuring the short axis would have yielded similar results.

Figure 3.

(a) The yellow-green subgenus Rhizocarpon used for lichenometry measurements. The majority of the lichen diameters used in this study were (b) <15 mm but some lichen individuals were (c) larger. We used digital calipers to measure the maximum lichen diameter.

At roughly the center of each patch, we established a transect perpendicular to the scarp. Within 1 m on either side of the transect, we measured the five largest yellow-green Rhizocarpon sp. growing on each exposed boulder, cobble, or region of bedrock. We continued measuring lichen diameters until we reached a distance from the scarp where lichen individuals had merged together or had lost their circular morphology. Transects ranged in length from 3 to 15 m. We measured the maximum lichen diameter with digital calipers with an accuracy of 0.02 mm. For each lichen diameter, we also measured the distance from the top of the scarp to the center of the lichen. At each of our five sites, we measured roughly 20 control lichen diameters outside of deflation patches on bedrock and boulders above the level of the loess.

Landscape position measurements

In the field, we measured the scarp length and the patch dimensions perpendicular and parallel to the scarp. To estimate total deflation patch size, we calculated the area of an ellipse using the patch dimensions as the two axes. For two patches, one at Site 2 and one at Site 3, the patch dimension perpendicular to the scarp was unsafe to measure in the field because of small bedrock cliffs. For an additional 10 patches, four at Site 3, four at Site 4, and two at Site 5, the patch dimension parallel to the scarp was hard to determine in the field since the entire hillside was deflated. To estimate these patch dimensions, we used a land-cover classification that identifies deflation patches with a producer’s accuracy of 93%, a user’s accuracy of 81%, and a spatial resolution of 2 m (Heindel et al., 2015). The two patches at Site 5 were outside the spatial extent of the land-cover classification, so these patches were removed from the analyses using patch size. We ran all analyses involving patch size with and without the estimated patches and report both results. For the elevation, slope, and aspect of each deflation patch, we used a Digital Elevation Model with 4-m spatial resolution constructed from WorldView-2 imagery and acquired through the Polar Geospatial Center, University of Minnesota.

Data analyses

Lichen growth curve

Pioneering work developing the method of lichenometry occurred in the Kangerlussuaq region of West Greenland during the 1950s and 1960s (Beschel, 1958; Beschel and Weidick, 1973; Ten Brink, 1973). During this time, Beschel, Weidick, and Ten Brink established and re-photographed a number of permanent stations in order to directly quantify the growth of Rhizocarpon geographicum, finding a 2-mm increase in diameter over 12 years.

More recently, Forman et al. (2007) developed a Rhizocarpon sp. growth curve for West Greenland, using the direct measurements from Ten Brink (1973), as well as two other indirect lichen measurements, one from a historical glacier retreat from a moraine (Gordon, 1981) and one from radiocarbon dating of a moraine-dammed lake (Ten Brink, 1973). Using indirect lichen measurements to construct growth curves has shown to be the most successful method since indirect measurements integrate past changes in lichen growth because of climate fluctuations (Roof and Werner, 2011). Although the West Greenland growth curve is constrained by only three points, it agrees remarkably well with Rhizocarpon sp. growth curves from Baffin Island and W. Spitsbergen (Figure 4; Forman et al., 2007; Werner, 1990). In using this lichen growth curve for moraine dating, Forman did not assign a mathematical function to the curve. Since our methods rely on converting individual lichen diameters to ages, we fit a logarithmic function to the growth curve established by Forman et al. (2007). We chose a logarithmic function because it matched the control points well and has been used previously to model the decrease in growth rate as Rhizocarpon sp. increase in size (e.g. Armstrong, 2011).

Figure 4.

Lichen growth curves for West Greenland, Baffin Island, Canada, and West Spitsbergen, Svalbard. We fit a logarithmic function to the three control points. Image modified after Forman et al. (2007) and Werner (1990).

For our lichenometry analyses, we chose to constrain our data to the single largest lichen growing on each cobble, boulder, or region of bedrock. When gravestones and other small surfaces are used, the single largest lichen has been shown to be most representative of surface age (Innes, 1985; Locke et al., 1979; Werner, 1990). Using the largest lichen on each boulder, we successfully biased the results toward the largest lichen while maintaining the high spatial resolution needed to estimate soil erosion rates. In addition, since the largest lichen diameter on the growth curve is 22 mm (~425 years old), we eliminated all lichen measurements larger than 22 mm for all analyses using lichen age to avoid extrapolating beyond the extent of the control points. Once constrained, the number of lichen measurements per patch ranged from 13 to 47.

Rates of patch expansion

To calculate the rate of deflation-patch expansion, we fit a linear model to the relationship between the distance from the scarp and lichen age for each patch. When inverted, the slope of this line is the average rate at which the scarp has expanded over the past 425 years (the time period covered by the lichen growth curve). We tested for differences in rates of patch expansion among sites using a one-way analysis of variance (ANOVA). For some deflation patches, the relationship between the distance from the scarp and lichen age appeared to be nonlinear, suggesting an accelerating or decelerating rate of deflation-patch expansion. To determine how rates of patch expansion have changed over time, we fit an exponential (accelerating rate of expansion) and logarithmic (decelerating rate of expansion) model to each patch and determined best fit using the Akaike information criterion corrected (AICc) for finite sample sizes (Anderson, 2008). To test our hypothesis that rates of patch expansion correlate with landscape position, we used multiple linear regression to test the effects of elevation, slope, aspect, proximity to the GrIS, patch size, and scarp length on the rate of soil erosion.

Deflation-patch ages

To estimate the age of each deflation patch, and thus the onset of soil deflation, we used the rate of patch expansion and the patch dimension perpendicular to the scarp. For each patch, we calculated the amount of time since formation that it would have taken the patch to expand to its current size (i.e. we multiplied the rate of expansion by the patch dimension perpendicular to the scarp). We tested for differences in deflation-patch age among sites using a one-way ANOVA.

There are a number of assumptions that go into this method of estimating deflation-patch age. First, we assumed a constant rate of expansion and direction of scarp movement since patch formation. Although some patches are better fit by an exponential or logarithmic model, we used the linear rate of patch expansion to have a consistent method for all patches. For the patches that indicated no-directionality of expansion (i.e. there was no relationship between distance from scarp and lichen diameter), we did not estimate patch age. Second, we assumed that patches have remained distinct landforms since formation. In most cases, the relatively circular morphology of the deflation patch suggested that the landform had remained distinct. However, large patches may have anomalously old ages if they formed by multiple patches merging together. Finally, we assumed that no vegetation has encroached on the deflation patches since formation. Generally, the boundaries of deflation patches are quite distinct, suggesting that vegetation has not encroached on the bare soil. However, we interpret our ages as minimum-limiting ages, since vegetation encroachment would make our estimates younger than the actual onset of soil deflation.

To test our hypothesis that deflation patches were initiated by drier and windier climate conditions associated with the onset of the ‘Little Ice Age’, we compared our patch ages with the ages of other regional landforms and with published paleoclimate records. First, we measured lichen diameters on the unvegetated historical moraine located <50 m from the current ice margin and compared these lichen diameters with the lichen diameters measured within deflation patches. While the exact age of this moraine is unknown, historical records show that at least part of the moraine was being deposited by AD 1880 (Weidick, 1968). Second, we plotted our age estimates against published sediment records from four different lakes in the Kangerlussuaq region. For two of the lake records, SS1381 and SS1220, we plot proxies for aeolian silt deposition: mass accumulation rate and magnetic susceptibility (Anderson et al., 2012; Olsen et al., 2012). For the other two lakes, SS16 and SFL4, we plot loss-on-ignition, a proxy for lake productivity (Perren et al., 2012; Willemse and Törnqvist, 1999).

Data availability

Data are available for download from the Dryad Digital Repository: https://dx-doi-org.web.bisu.edu.cn/10.5061/dryad.v82g6.

Results

Lichen diameters

Control lichen diameters, measured outside of deflation patches and above the level of the loess, did not vary among sites (Table 1). This result supports the critical assumption that past and present climate differences among sites have not been large enough to result in differences in lichen growth. This result suggests that differences in lichen diameters reflect differences in deflation dynamics rather than differences in lichen growth rates.

Table 1.

Lichen diameters, erosion rates, and estimated age ranges across sites. Lichen diameters are in millimeter. With the exception of the interquartile age ranges (IQR), all results are reported as mean or median ± standard error (number of replicates).

Site	Median control diameter	Median deflation diameter	Erosion rate (cm/yr)	Mean age (years)	IQR (years)
1 (close to GrIS)	31.9 ± 1.5 (45)	12.2 ± 0.9 (14)	2.47 ± 0.37 (14)	456 ± 72 (14)	539–237 (14)
2	36.1 ± 1.2 (70)	11.3 ± 0.5 (14)	2.76 ± 0.47 (13)	611 ± 161 (13)	730–217 (13)
3	31.6 ± 1.2 (38)	13.3 ± 0.9 (14)	3.18 ± 0.56 (13)	932 ± 222 (13)	1509–379 (13)
4	30.8 ± 2.2 (19)	15.5 ± 1.3 (11)	1.97 ± 0.35 (10)	458 ± 111 (10)	626–193 (10)
5 (far from GrIS)	33.5 ± 1.1 (30)	16.2 ± 1.1 (12)	2.28 ± 0.58 (9)	586 ± 145 (9)	761–333 (9)
Site average	32.8 ± 0.9 (5)	13.7 ± 0.9 (5)	2.53 ± 0.21 (5)	609 ± 87 (5)	800–233 (59)

Median lichen diameters within deflation patches were smaller than the control lichen diameters (Table 1). Median lichen diameters showed a trend with distance from the GrIS, with lichen diameters slightly smaller at sites close to the GrIS (Sites 1 and 2) and larger at sites farther away from the GrIS (Sites 4 and 5; Table 1).

Rates of patch expansion

There was a positive relationship between the distance from the scarp and lichen age for 59 out of the 65 deflation patches. The relatively tight fit around the linear models suggests that lichen growth rates are consistent among lichen individuals, even between the two Rhizocarpon species (Figure 5; Supplemental Figures 1–4, available online). In patches where we do see large deviations from the linear model (for instance, Patches C and D in Figure 5), there may have been episodic patch expansion, resulting in a wide distribution of lichen diameters at a single distance from the scarp. There was no relationship between distance and lichen age for one deflation patch from Site 1, one from Site 2, one from Site 4, and three from Site 5. We did not estimate rates of patch expansion or patch age for these six deflation patches.

Figure 5.

The relationship between the distance from the scarp and lichen age plotted for all 14 deflation patches from Site 2. (b, c, d, f, g, j, m, and n) Patches best fit by a linear model have only the linear model plotted. (a, e, and l) Patches best fit by an exponential model have both the exponential and the linear models plotted. (h, k) Patches best fit by a logarithmic model have both the logarithmic and the linear models plotted. (i) Patches with no relationship have no model plotted. For similar figures of the other four sites, see Supplemental Figures 1–4, available online.

Rates of patch expansion over the past ~425 years (the time period covered by the lichen growth curve) ranged from 0.63 to 8.48 cm yr⁻¹, with an average rate (±standard error (SE)) of 2.53 cm yr⁻¹ (±0.21) across all sites (Table 1). Rates were not significantly different among sites (F(4, 54) = 0.92, p = 0.46; Table 1). Multiple linear regression showed no significant effects of elevation, slope, aspect, proximity to the GrIS, patch size, or scarp length on the rate of soil erosion. The result of the multiple linear regression remained the same without the patches where land-cover classification estimates were used to calculate patch size.

In 43 out of the 59 patches with a positive relationship between distance from the scarp and lichen age, the linear model fit better than or as well as the exponential or logarithmic model, suggesting a stable rate of patch expansion over time. Of the 16 remaining patches, half were accelerating (best fit by exponential models) and half were decelerating (best fit by logarithmic models). Both accelerating and decelerating patches were found throughout the entire study area. There were no noticeable differences between accelerating and decelerating patches in terms of patch size, scarp length, elevation, aspect, slope, or estimated patch age.

Deflation-patch ages

The mean patch age (±SE) of all 59 patches with erosion rate estimates was 609 years (±87), and the median patch age was 453 years (Table 1). Removing the two patches with land-cover classification estimates for the patch dimension perpendicular to the scarp, the mean patch age (±SE) was 560 years (±57), and the median patch age was 439 years. Because the mean patch age was strongly affected by these two older patches, we use the median, 453 years ago, and the interquartile range, 800–230 years ago, for all patches as our best estimate of the timing of deflation-patch formation. We use one range for all patches because there was no significant difference in deflation-patch ages among sites (F(4, 54) = 1.69, p = 0.17; Table 1).

As an additional check on our patch age estimates, we converted 90th percentile lichen diameters within deflation patches to ages. Although many authors use the largest one or five lichen individuals measured on a landform as an age estimate, this technique has been shown to be inadequate (e.g. Jomelli et al., 2007). Instead, we use the 90th percentile of the entire lichen size distribution to reduce the influence of anomalously large lichen individuals. Using this method, median patch age was 484 years old, which agrees well with the estimate of 453 years old described above. The bottom bound of the interquartile range was 228 years ago, which matches almost exactly with the result of 230 years ago described above. Because of the shape of the lichen growth curve, the upper bound of the interquartile range was an extrapolation of the lichen growth curve and was not a meaningful comparison. The good agreement between these methods provides additional support for our deflation-patch ages.

Deflation-patch ages fall mostly within the ‘Little Ice Age’ time period (AD 1350–1880), although deflation-patch formation did begin before AD 1350. In addition, deflation patches are older than the unvegetated historical moraine. Median lichen diameters (±SE) on the historical moraine were 6.0 mm (±0.2), considerably smaller than both the median lichen diameters (13.7 ± 0.9) and the 90th percentile lichen diameters (23.9 ± 1.4) measured within deflation patches. This supports our conclusion that deflation patches have been features on the Kangerlussuaq landscape for more than the last 100 years. Finally, our deflation-patch ages agree well with lake-sediment records. Mass accumulation rate and magnetic susceptibility, both proxies for aeolian silt deposition, show a dramatic increase in lakes SS1381 and SS1220 at the time of deflation-patch formation (Figure 6). In addition, during the time period of deflation-patch formation, lakes SS16 and SFL4 show a clear decrease in loss on ignition, suggesting a reduction in productivity, which has been attributed to colder air temperatures (Willemse and Törnqvist, 1999).

Figure 6.

Comparison of deflation-patch age estimates (top panel, dots are individual patches, and diamonds and whiskers are site medians and interquartile ranges, respectively) with lake-sediment records from the Kangerlussuaq area: mass accumulation rate from Lake SS1381 (Anderson et al., 2012), magnetic susceptibility from Lake SS1220 (Olsen et al., 2012), loss on ignition from SS16 (Perren et al., 2012), and loss on ignition from SFL4 (Willemse and Törnqvist, 1999). Shaded bar is the interquartile range for all patches, 800–230 years ago.

Discussion

Past erosion dynamics

We found that there was a period of heightened deflation-patch formation in the Kangerlussuaq region 800–230 years ago. We interpret our ages as minimum-limiting ages because of our assumption that no vegetation has grown into the deflation patches since formation. The onset of soil erosion likely occurred simultaneously at all distances from the GrIS since we see no significant difference in deflation-patch ages among sites.

The timing of soil erosion from this study supports our hypothesis that deflation patches were initiated by drier and windier climate conditions associated with the ‘Little Ice Age’. We see good agreement between deflation-patch ages and lake-sediment records. Multiple lake records show an elevated sediment accumulation rate near the onset of deflation-patch formation (Anderson et al., 2012; Heggen et al., 2010; Olsen et al., 2013; Perren et al., 2012). Since the Kangerlussuaq landscape sees little surface runoff, the increase in accumulation rate has been attributed to increased aeolian activity (Anderson et al., 2012; Heggen et al., 2010; Olsen et al., 2013; Perren et al., 2012). Anderson et al. (2012) speculate that the increase in aeolian activity may be because of landscape desiccation, vegetation die-off, and the remobilization of the loess deposit. Our estimated patch ages agree well with this interpretation; the timing of patch formation coincides with the increase in mass accumulation rate at lake SS1381 and the increase in magnetic susceptibility at lake SS1220 (Anderson et al., 2012; Olsen et al., 2012). In addition, our ages agree well with patterns of aeolian activity in Sandflugtdalen, showing renewed sand sheet formation during the ‘Little Ice Age’ after a period of landscape stability and soil formation (Willemse et al., 2003).

Numerous other paleoclimate records from the Kangerlussuaq region show both increased aridity and cooling during the ‘Little Ice Age’. Lake levels over the past 700 years appear to be much lower than the prior ~6000 years (Aebly and Fritz, 2009), reflecting a change toward more evaporation than precipitation. Drier conditions may have made it easier for deflation-patch formation to occur by weakening the vegetation cover protecting the loess deposit. Drier conditions would also likely have meant a reduction in winter snow cover, exposing the vegetation cover and ground surface to high winds and entrained ice particles. In addition, paleoproductivity proxies, pollen records, and an alkenone unsaturation index all show decreasing temperatures coincident with the increase in aridity during the last 1000–800 years (Bennike et al., 2010; D’Andrea et al., 2011; Perren et al., 2012; Willemse and Törnqvist, 1999). Colder conditions may have influenced deflation-patch formation by increasing the length of the winter season, when aeolian processes tend to be more active. In addition, colder conditions could have led to more intensive frost action, which could break the ground surface and expose the loess deposit to winds. While the exact mechanisms of deflation-patch formation remain unknown, our results suggest that colder and drier conditions during the ‘Little Ice Age’ initiated the formation of many of the deflation patches present today.

The ‘Little Ice Age’ was a period of rapid environmental change in West Greenland that may have had consequences for human ecology. D’Andrea et al. (2011) link the abrupt temperature decline seen in their alkenone unsaturation index around 850 years ago to the abandonment of the western Norse settlement, which occurred roughly 650 years ago (Barlow et al., 1997). Although the western Norse settlement was located south of Kangerlussuaq, our results suggest that temperature was not the only variable changing during this time. Rapid landscape change, in the form of soil loss, occurred simultaneously, at least in the Kangerlussuaq region of West Greenland.

Current deflation dynamics

On average, deflation patches are currently expanding at a rate of ~2.5 cm yr⁻¹, comparable with lower estimates of similar soil erosion in Iceland (Arnalds, 2000). While rates of patch expansion have been roughly linear over the past ~425 years, on shorter timescales, the nature of deflation-patch expansion is likely episodic. In any given year, minimal change may occur, or a large block of soil and vegetation may fall from the undercut scarp. The episodic nature of soil deflation may explain why some patches appeared to have accelerating or decelerating rates of expansion. For instance, at Patch (l) from Site 2 (Figure 5), a large erosion event may have occurred ~300 years ago, followed by a period of relatively stability until ~100 years ago, when another episodic event exposed more surface for lichen growth. Episodic deflation could also explain deviations from the linear trends observed in many of the plots (e.g. Patch (j) in Figure 5, see also Supplemental Figures 1–4, available online).

We did not find support for our hypothesis that rates of soil erosion are highest close to the GrIS and on steep south-facing slopes. Although previous work has shown that landscape position is important in the distribution of wind erosion (Heindel et al., 2015), deflation patches seem to expand at similar rates regardless of their proximity to the GrIS, aspect, slope, elevation, or patch size. Our results suggest that the higher occurrence of deflation close to the GrIS and on steep south-facing slopes is a function of increased patch formation rather than higher rates of expansion within individual patches. Similar results have been found in Iceland, where erosional fronts, called rofabards, move across the landscape. Dugmore et al. (2009) found that periods of rapid soil loss resulted from a high density of vegetation breaches rather than higher rates of rofabard movement. In Iceland, the sensitivity of the vegetation to initial breaches increased as a result of climate change (Dugmore et al., 2009). Our results suggest that regions closer to the GrIS and on steep south-facing slopes are more prone to patch formation, possibly due to the fact that these regions are dominated by steppe tundra rather than shrub tundra. Vegetation type, itself a function of the climate gradient between the fjord and the GrIS, may play a role in the frequency of patch formation but not in patch expansion rate.

All of the deflation patches we measured formed earlier than 100 years ago, suggesting that fewer deflation patches are forming today than 800–230 years ago. To consider the possibility of selection bias, we conducted a survey of Sites 1, 3, 4, and 5 during the summer of 2015. We found numerous young deflation patches at Site 1, with minimal lichen growth and lichen diameters all <10 mm. At this location, proximal to the GrIS, it seems as though deflation patches are still actively forming. At Site 3, only a few young patches were found, and at Sites 4 and 5, we found no young deflation patches. These results further support the hypothesis that vegetation type plays a role in the frequency of patch formation. Farther away from the GrIS, shrubs dominate the landscape and may be a stabilizing force preventing current or future patch formation.

We propose that there is an erosional threshold in the Kangerlussuaq aeolian system. Climate conditions required for initial patch formation are more extreme than the climate conditions required for continued deflation. During the ‘Little Ice Age’, climate conditions were harsh enough to push the entire study region across this threshold, resulting in the current pattern of deflation expansion at all sites. Today, however, patches are forming only at sites close to the GrIS. Results from Iceland support this finding as well. In Iceland, more intense grazing or severe climate conditions are needed to trigger rofabard formation, but rofabard expansion occurs under milder grazing or climate conditions (Streeter and Dugmore, 2013). Once there is a break in the vegetation cover and soil is exposed, there is little to prevent continued deflation; the erosional threshold has been crossed, and the deflation patch may continue expanding indefinitely until it reaches a north-facing slope or some other physical barrier. This has significant implications for land management in arid regions susceptible to soil erosion. Management for the prevention of soil loss should focus on sustaining an intact landscape since curbing soil loss that is already underway may prove challenging.

Future deflation dynamics

The rapid climate changes currently occurring in Greenland and across the Arctic have direct implications for aeolian activity in the Kangerlussuaq region. During the period 1981–2011, Kangerlussuaq experienced both warming (Hanna et al., 2012) and increased precipitation (Mernild et al., 2015), although the warming trends were more robust, especially during winter months. Similar trends are projected into the future, with up to 7°C of wintertime warming and a 30% increase in wintertime precipitation by the end of the century, under the IPCC RCP4.5 scenario (IPCC, 2013). A warmer and wetter climate, directly opposite to the conditions that we propose triggered the initial formation of deflation patches, might be expected to result in a decrease in aeolian activity. However, the increase in precipitation projected for Kangerlussuaq may not be able to offset the increase in evapotranspiration because of warmer temperatures, leading to increased aridity. Drier conditions may increase the susceptibility of the vegetation cover to breaches, allowing for the formation of new deflation patches. In addition, changes in vegetation cover may also play a role in determining future deflation dynamics since vegetation type seems to play a role in the frequency of patch formation. While shrub expansion has been observed across the Arctic (Myers-Smith et al., 2011), it is unclear whether or not shrub expansion will occur in Kangerlussuaq, where low soil moisture may counter the effects of warming (Myers-Smith et al., 2015). Regardless, deflation patches that already exist will likely continue to expand, with soil deflation remaining an issue of concern for the Kangerlussuaq region.

Currently, soil deflation in the Kangerlussuaq region occurs without anthropogenic stresses. While muskoxen were introduced into the region in the 1960s (Pedersen and Aastrup, 2000), soil erosion had already been occurring for centuries, and no recent acceleration of soil erosion is evident. In addition to rapid climate change, West Greenland may also face changes in land use that could threaten soil stability, such as agriculture, mining, tourism, or grazing. Iceland serves as a severe example of the extreme soil loss possible when climate and anthropogenic factors work together to drive erosion (Aradottir et al., 2013; Dugmore et al., 2009; Streeter et al., 2012). Understanding current deflation dynamics without any anthropogenic influence is an important baseline in a region experiencing such rapid environmental change.

One gap in our knowledge of how aeolian erosion impacts the Kangerlussuaq landscape is whether or not deflation patches can re-vegetate. It is possible that soil deflation is a permanent ecosystem change, with deflation patches indefinitely remaining areas of low productivity and altered vegetation. Similar state changes occur in other arid regions, such as the southwestern United States, where positive feedbacks can sustain a shift from grasslands to woody shrubs (e.g. D’Odorico et al., 2012). However, the mixed deciduous shrubs and herbaceous species characteristic of the surrounding Kangerlussuaq landscape may eventually reclaim deflation patches. In this case, deflation patches would represent a large area for deciduous shrubs to move into, further expanding their dominance on the landscape. While the rate of vegetation growth may be too slow to measure directly, understanding the soil forming processes within deflation patches would be a first step toward distinguishing between these opposing scenarios. Understanding the potential for vascular plants to re-vegetate deflation patches would greatly improve our predictions of the long-term impacts of aeolian erosion in West Greenland and is thus an important area for future study.

Conclusion

The formation of deflation patches, prominent landscape features in the Kangerlussuaq region of West Greenland, occurred approximately 800–230 years ago, during the ‘Little Ice Age’, at a time when lake-sediment records indicate widespread aeolian activity. An erosional threshold exists in the Kangerlussuaq aeolian system, with the cold and dry climate conditions that trigger deflation-patch formation likely harsher than conditions necessary for continued erosion, a result with direct implications for land management in arid regions. Rates of patch expansion, averaging 2.5 cm yr⁻¹, are similar among patches and are not correlated with landscape position. Areas characterized by intense deflation have experienced a high density of vegetation breaches rather than higher rates of deflation-patch expansion. The impact of future deflation on the Kangerlussuaq landscape remains an issue of concern, with projected climate change potentially leading to conditions more favorable for erosion if warming continues to outpace increases in precipitation. Future land-use change, whether for agriculture, mining, tourism, or grazing, could also threaten soil stability in the region. The long-term impact of aeolian soil erosion on the Kangerlussuaq landscape depends partly on whether or not deflation patches can re-vegetate, an area for future investigation.

Footnotes

Acknowledgements

We thank the Kangerlussuaq International Science Support and Polar Field Services for field logistics. Thanks to Christine Urbanowicz, Becca Novello, and Phoebe Racine for field assistance and to Lee McDavid and Angela Spickard for administrative and logistics support. We thank Meredith Kelly and Matt Ayres for their helpful discussions about this project.

Data availability

Data are available for download from the Dryad Digital Repository: .

Funding

This work was supported by the National Science Foundation (grant numbers 0801490 to RAV and 1506155 to LEC and RAV) and Dartmouth’s Institute of Arctic Studies.

References

Aebly

Fritz

(2009) Palaeohydrology of Kangerlussuaq (Sondre Stromfjord), West Greenland during the last ~8000 years. The Holocene 19: 91–104.

Anderson

(2008) Model Based Inference in the Life Sciences: A Primer on Evidence. New York; London: Springer.

Anderson

Liversidge

McGowan

, et al. (2012) Lake and catchment response to Holocene environmental change: Spatial variability along a climate gradient in southwest Greenland. Journal of Paleolimnology 48(1): 209–222.

Aradottir

Petursdottir

Halldorsson

, et al. (2013) Drivers of ecological restoration: Lessons from a century of restoration in Iceland. Ecology and Society 18(4): 33.

Armstrong

(2011) The biology of the crustose lichen Rhizocarpon geographicum. Symbiosis 55(2): 53–67.

Arnalds

(2000) The Icelandic ‘rofabard’ soil erosion features. Earth Surface Processes and Landforms 25(1): 17–28.

Barlow

Sadler

Ogilvie

AEJ

, et al. (1997) Interdisciplinary investigations of the end of the Norse western settlement in Greenland. The Holocene 7(4): 489–499.

Bennike

Anderson

McGowan

(2010) Holocene palaeoecology of southwest Greenland inferred from macrofossils in sediments of an oligosaline lake. Journal of Paleolimnology 43(4): 787–798.

Beschel

(1958) Lichenometrical studies in West Greenland. Arctic 11: 254.

10.

Beschel

Weidick

(1973) Geobotanical and geomorphological reconnaissance in West Greenland, 1961 (Lichenometry: Dedicated to the memory of the late Roland E Beschel). Arctic and Alpine Research 5(4): 311–319.

11.

Bullard

(2013) Contemporary glacigenic inputs to the dust cycle. Earth Surface Processes and Landforms 38(1): 71–89.

12.

Bullard

Austin

(2011) Dust generation on a proglacial floodplain, West Greenland. Aeolian Research 3(1): 43–54.

13.

D’Andrea

Huang

Fritz

, et al. (2011) Abrupt Holocene climate change as an important factor for human migration in West Greenland. Proceedings of the National Academy of Sciences of the United States of America 108(24): 9765–9769.

14.

Dijkmans

JWA

Törnqvist

(1991) Modern periglacial eolian deposits and landforms in the Sondre Stromfjord area, West Greenland and their palaeoenvironmental implications. Meddelelser Om Grønland, Geoscience 25: 1–39.

15.

D’Odorico

Okin

Bestelmeyer

(2012) A synthetic review of feedbacks and drivers of shrub encroachment in arid grasslands. Ecohydrology 5(5): 520–530.

16.

Dugmore

Gisladottir

Simpson

, et al. (2009) Conceptual models of 1200 years of Icelandic soil erosion reconstructed using tephrochronology. Journal of the North Atlantic 2: 1–18.

17.

Field

Belnap

Breshears

, et al. (2010) The ecology of dust. Frontiers in Ecology and the Environment 8(8): 423–430.

18.

Forman

Marín

Van

, et al. (2007) Little Ice Age and neoglacial landforms at the Inland Ice margin, Isunguata Sermia, Kangerlussuaq, west Greenland. Boreas 36(4): 341–351.

19.

Gordon

(1981) Glacier margin fluctuations during the 19th and 20th centuries in the Ikamiut kangerdluarssuat area, West Greenland. Arctic and Alpine Research 13(1): 47–62.

20.

Hanna

Mernild

Cappelen

, et al. (2012) Recent warming in Greenland in a long-term instrumental (1881–2012) climatic context: I. Evaluation of surface air temperature records. Environmental Research Letters 7(4): 045404.

21.

Hansen

(2000) A contribution to the lichen flora of the Kangerlussuaq area, West Greenland. Cryptogamie Mycologie 21(1): 53–59.

22.

Heggen

Birks

Anderson

(2010) Long-term ecosystem dynamics of a small lake and its catchment in west Greenland. The Holocene 20(8): 1207–1222.

23.

Heindel

Chipman

Virginia

(2015) The spatial distribution and ecological impacts of aeolian soil erosion in Kangerlussuaq, West Greenland. Annals of the Association of American Geographers 105(5): 875–890.

24.

Innes

(1985) Lichenometry. Progress in Physical Geography 9(2): 187–254.

25.

Innes

(1986) Dating exposed rock surfaces in the Arctic by lichenometry: The problem of thallus circularity and its effect on measurement errors. Arctic Institute of North America 39(3): 253–259.

26.

IPCC (2013) Annex I: Atlas of Global and Regional Climate Projections [ Van Oldenborgh

Collins

Arblaster

, et al. (eds)]. In: Stocker

Qin

Plattner

G-K

, et al. (eds) Climate Change 2013: The Physical Science Basis: Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge; New York: Cambridge University Press, pp. 1322–1325.

27.

Jomelli

Grancher

Naveau

, et al. (2007) Assessment study of lichenometric methods for dating surfaces. Geomorphology 86(1–2): 131–143.

28.

Kopec

Lauder

Posmentier

, et al. (2014) The diel cycle of water vapor in west Greenland. Journal of Geophysical Research: Atmospheres 119(15): 9386–9399.

29.

Levy

Kelly

Howley

, et al. (2012) Age of the Ørkendalen moraines, Kangerlussuaq, Greenland: Constraints on the extent of the southwestern margin of the Greenland Ice Sheet during the Holocene. Quaternary Science Reviews 52: 1–5.

30.

Locke

III Andrews

Webber

(1979) A Manual for Lichenometry. London: British Geomorphological Research Group.

31.

Mernild

Hanna

McConnell

, et al. (2015) Greenland precipitation trends in a long-term instrumental climate context (1890–2012): Evaluation of coastal and ice core records. International Journal of Climatology 35(2): 303–320.

32.

Myers-Smith

Elmendorf

Beck

PSA

, et al. (2015) Climate sensitivity of shrub growth across the tundra biome. Nature Climate Change 5(9): 887–891.

33.

Myers-Smith

Forbes

Wilmking

, et al. (2011) Shrub expansion in tundra ecosystems: Dynamics, impacts and research priorities. Environmental Research Letters 6(4): 045509.

34.

Okin

Gillette

Herrick

(2006) Multi-scale controls on and consequences of aeolian processes in landscape change in arid and semi-arid environments. Journal of Arid Environments 65(2): 253–275.

35.

Olsen

Anderson

Knudsen

(2012) Variability of the North Atlantic Oscillation over the past 5,200 years. Nature Geoscience 5(11): 808–812.

36.

Olsen

Anderson

Leng

(2013) Limnological controls on stable isotope records of late-Holocene palaeoenvironment change in SW Greenland: A paired lake study. Quaternary Science Reviews 66: 85–95.

37.

Osborn

McCarthy

LaBrie

, et al. (2015) Lichenometric dating: Science or pseudo-science? Quaternary Research 83(1): 1–12.

38.

Pedersen

Aastrup

(2000) Muskoxen in Angujaartorfiup Nunaa, West Greenland: Monitoring, spatial distribution, population growth, and sustainable harvest. Arctic 53(1): 18–26.

39.

Perren

Anderson

Douglas

MSV

, et al. (2012) The influence of temperature, moisture, and eolian activity on Holocene lake development in West Greenland. Journal of Paleolimnology 48(1): 223–239.

40.

Ravi

D’Odorico

Breshears

, et al. (2011) Aeolian processes and the biosphere. Reviews of Geophysics 49: RG3001.

41.

Roof

Werner

(2011) Indirect growth curves remain the best choice for lichenometry: Evidence from directly measured growth rates from Svalbard. Arctic Antarctic and Alpine Research 43(4): 621–631.

42.

Rosenwinkel

Korup

Landgraf

, et al. (2015) Limits to lichenometry. Quaternary Science Reviews 129: 229–238.

43.

Rydberg

Lindborg

Sohlenius

, et al. (2016) The importance of eolian input on lake-sediment geochemical composition in the dry proglacial landscape of western Greenland. Arctic Antarctic and Alpine Research 48(1): 93–109.

44.

Schlesinger

Reynolds

Cunningham

, et al. (1990) Biological feedbacks in global desertification. Science 247(4946): 1043–1048.

45.

Streeter

Dugmore

(2013) Anticipating land surface change. Proceedings of the National Academy of Sciences of the United States of America 110(15): 5779–5784.

46.

Streeter

Dugmore

Vesteinsson

(2012) Plague and landscape resilience in premodern Iceland. Proceedings of the National Academy of Sciences of the United States of America 109(10): 3664–3669.

47.

Ten Brink

(1973) Lichen growth rates in West Greenland. Arctic and Alpine Research 5: 323–331.

48.

Ten Brink

Weidick

(1974) Greenland ice sheet history since the last glaciation. Quaternary Research 4(4): 429–440.

49.

Weidick

(1968) Observations on some Holocene glacier fluctuations in west Greenland. Meddelelser Om Grønland, Geoscience 165: 1–202.

50.

Werner

(1990) Lichen growth rates for the northwest coast of Spitsbergen, Svalbard. Arctic and Alpine Research 22(2): 129–140.

51.

Willemse

Tornqvist

(1999) Holocene century-scale temperature variability from west Greenland lake records. Geology 27(7): 580–584.

52.

Willemse

Koster

Hoogakker

, et al. (2003) A continuous record of Holocene eolian activity in West Greenland. Quaternary Research 59(3): 322–334.

53.

Young

Briner

(2015) Holocene evolution of the western Greenland Ice Sheet: Assessing geophysical ice-sheet models with geological reconstructions of ice-margin change. Quaternary Science Reviews 114: 1–17.