Journal list menu

Volume 66, Issue S1 p. S157-S168
Original Article
Open Access

Effects of prolonged sedimentation from permafrost degradation on macroinvertebrate drift in Arctic streams

Brianna Levenstein

Corresponding Author

Brianna Levenstein

Department of Biology and Canadian Rivers Institute, University of New Brunswick, Fredericton, New Brunswick, Canada

Correspondence: [email protected]

Search for more papers by this author
Jennifer Lento

Jennifer Lento

Department of Biology and Canadian Rivers Institute, University of New Brunswick, Fredericton, New Brunswick, Canada

Search for more papers by this author
Joseph Culp

Joseph Culp

Department of Biology and Canadian Rivers Institute, University of New Brunswick, Fredericton, New Brunswick, Canada

Environment and Climate Change Canada, Wilfrid Laurier University, Waterloo, Ontario, Canada

Search for more papers by this author
First published: 03 December 2020
Citations: 6
Associate editor: Suzanne E. Tank
Special Issue: Biogeochemistry and Ecology across Arctic Aquatic Ecosystems in the Face of Change
Edited by: Peter J. Hernes, Suzanne E. Tank and Ronnie N. Glud

Abstract

Retrogressive thaw slumps are areas of unstable degraded permafrost that often drain into nearby watersheds, leading to increased sediment loads and changes in water quality. Thaw slumps are prevalent across the Arctic, including western Canada, Alaska, and Russia, and high-altitude areas of western China. Over the past several decades, increased temperatures and precipitation in the Arctic have led to increases in the size and frequency of thaw slumps. Our study explored the effects of prolonged sedimentation from thaw slumps in the Peel Plateau, NWT, Canada on benthic macroinvertebrate drift, an important biological function of stream ecosystems. Though sedimentation is known to initiate a catastrophic drift response, studies have generally not considered the drift response to ongoing, long-term perturbation. Drift densities and sediment loads were measured using drift nets and sediment traps at paired sites upstream and downstream of thaw slumps. We compared drift densities and sediment loads between sites and examined how drift differed over a fine-sediment gradient. The amount of suspended and settling fine sediments increased significantly at downstream sites. Drift densities decreased at downstream sites; however, when drift was corrected for benthic abundance at each site, there was an increase in proportional drift density associated with increased fine sediments. These results indicate that prolonged impacts from thaw slumps result in lower macroinvertebrate abundance and higher proportional drift relative to undisturbed sites. Ultimately, increased sediment loads from thaw slumps represent a chronic stressor that will continue to prevent recovery of macroinvertebrate communities at impacted sites until these features stabilize.

Increases in precipitation and warming air temperatures have destabilized permafrost in many areas of the Arctic, leading to widespread thermokarst formation (Lacelle et al. 2010; Kokelj et al. 2015; Turetsky et al. 2019), and associated changes to water quality in nearby waterbodies, including increased sedimentation, nutrient, and solute loads (Bowden et al. 2008; Kokelj et al. 2013; Moquin et al. 2014; Chin et al. 2016; Levenstein et al. 2018). Among the more conspicuous features of thermokarst are retrogressive thaw slumps (Burn and Lewkowicz 1990; Lantz and Kokelj 2008). Retrogressive thaw slumps are visible depressions in the landscape that occur in areas of ice-rich permafrost (Lewkowicz and Way 2019) and are initiated by a variety of physical processes including large precipitation events (Lacelle et al. 2010; Kokelj et al. 2015) and erosion by rivers or wave action along coastlines (Kokelj et al. 2013). As the permafrost degrades, it falls to the base of the slump, forming a mud slurry that can flow downslope, often into nearby stream systems (Kokelj et al. 2013). This mud slurry can drastically alter stream environments by increasing concentrations of sediments (Kokelj et al. 2013), solutes (Malone et al. 2013), and contaminants (St. Pierre et al. 2018). Over the past few decades, retrogressive thaw slumps have increased in size and frequency in the western Canadian Arctic (Lantz and Kokelj 2008) and the Russian Arctic (Murton et al. 2017), which has led to the formation of massive thaw slumps, termed megaslumps. Thaw slumps are categorized as megaslumps if they are greater than 5 ha in area, have headwalls greater than 4 m in height, and have retreated beyond the slope of the fluvial valley (Kokelj et al. 2013). These large thermokarst formations are prevalent throughout the ice-rich permafrost landscape of the Peel Plateau and the eastern foothills of the Mackenzie Mountains, Northwest Territories, Canada (Lacelle et al. 2015; Segal et al. 2016; Kokelj et al. 2017), and the largest known megaslump in the world is found in Yakutia in Siberian Russia (Murton et al. 2017). Upon reaching watercourses, the debris flows from retrogressive thaw slumps can partially or completely fill the interstitial spaces of stream beds (Chin et al. 2016). These slumps change the physical and chemical characteristics of the stream ecosystem by driving diurnal fluctuations in water level and by drastically increasing the timing and magnitude of fine-sediment and solute loads (Kokelj et al. 2013; Malone et al. 2013) as well as contaminants such as heavy metals (St. Pierre et al. 2018).

The effect of permafrost thaw slumps on the chemical and physical habitat of streams has a high potential to impact in-stream processes. For example, benthic macroinvertebrate communities are used extensively to monitor change in freshwater ecosystems (Wallace and Webster 1996; Milner et al. 2006; see review in Buss et al. 2014, Moquin et al. 2014) because they show functional and structural responses to changes in water quality (Buss et al. 2014). Increased sediment loads from slump debris can have multiple impacts on macroinvertebrate communities including changes in habitat structure through the filling of interstitial spaces on the streambed (Lenat et al. 1981; Lemly 1982) or loss of food resources such as benthic algae (Levenstein et al. 2018). Moreover, scouring sediments can damage fragile respiratory organs of certain macroinvertebrate taxa (Lemly 1982) and deposited sediments can lead to burial and oxygen depletion along the streambed (see review in Jones et al. 2012).

One way that macroinvertebrate larvae respond to unfavorable changes in the surrounding environment is through drift, a common dispersal method in which macroinvertebrates actively or passively enter the stream flow and settle downstream where conditions may be more suitable for habitation (see review in Naman et al. 2016). Downstream drift can increase with changes in stream conditions, such as increased turbidity, suspended solids, and fine sediment movement (Rosenberg and Wiens 1978; Culp et al. 1986; Suren and Jowett 2001; Molinos and Donohue 2009; Larsen and Ormerod 2010). The disturbance regime caused by slumping may lead to increased macroinvertebrate drift, with macroinvertebrates searching for more favorable environmental conditions further downstream (see review in Brittain and Eikeland 1988; Suren and Jowett 2001) or being removed by catastrophic disturbance (e.g., Culp et al. 1986; Gibbins et al. 2007). Catastrophic drift—where a large number of macroinvertebrates enter the drift simultaneously—is a common occurrence at the initiation of a disturbance (Rosenberg and Wiens 1978; Culp et al. 1986; Suren and Jowett 2001) and can lead to an initial decrease in benthic densities (Suren and Jowett 2001; Larsen and Ormerod 2010). Where disturbance is ongoing, the density of drifting organisms may decrease as an equilibrium in the system is reached and only tolerant organisms remain, although drift may continue to respond to disturbance pulses. However, studies of macroinvertebrate drift have generally focused on short-term disturbances, with no evidence of the patterns in drift that might be expected with ongoing perturbation (e.g., longer than a year).

This study aimed to evaluate the impacts of increased sediment loads from ongoing thaw slump disturbances on macroinvertebrate drift by collecting drift samples upstream and downstream of thaw slumps. Although many drift studies have been performed in temperate regions, relatively few have been done in Arctic streams (e.g., Müller 1954, 1982; Johansen et al. 2000), and to our knowledge, none have documented the impacts of thermokarst features such as retrogressive thaw slumps on macroinvertebrate drift (see review in Vonk et al. 2015). Furthermore, our study streams provide a unique situation for examining drift under ongoing and chronic sedimentation events, conditions which few drift studies have documented. Impacts of retrogressive thaw slumps on macroinvertebrate community structure were explored by Chin et al. (2016) and Levenstein (2016), both of whom found significantly lower macroinvertebrate abundance associated with high levels of total suspended solids (TSS). These studies showed that thaw slumps affect a range of taxa from those sensitive to water quality change to other, non-sensitive taxa. Therefore, with the onset of thaw slump impacts, a catastrophic drift response (e.g., large drift volume following the onset of impact) might be expected due to scouring sediments (sensu Culp et al. 1986), lower food quality (sensu Herbst 1980 and Levenstein et al. 2018), and habitat homogenization through the loss of interstitial spaces (sensu Lenat et al. 1981), all caused by sediment release from thaw slumps. However, the effects of prolonged sediment inputs from ongoing thaw slump disturbances are not clear. We hypothesize that drift could be lower at these sites after prolonged exposure if only tolerant macroinvertebrates remained in low abundances, which would reduce the pool of macroinvertebrates present to enter the drift.

Methods

Study area

Drift samples were taken from five stream reaches upstream and downstream of thaw slumps across the Peel Plateau, Northwest Territories, Canada to determine the impacts of ongoing thaw slumping on macroinvertebrate drift. Sites were located in first- to third-order streams with gravel-cobble substrate in the Stony Creek (catchment area ~ 1100 km2; Kokelj et al. 2013) and Vittrekwa River (catchment area ~ 2000 km2; Paquette 2015) watersheds, both of which are within the Peel River basin (Fig. 1). This area is part of the Tundra Cordillera Subarctic ecoregion (Ecosystem Classification Group 2010), which is characterized by tundra landscapes at higher altitudes and subalpine forests within deep fluvial valleys. Winters in the area are long and cold, with long-term average (1981–2010) daily minimum temperatures below −30°C in January, and daily maximum temperatures below freezing from October through April, whereas summers are short and cool, with long-term average daily maximum temperatures greater than 15°C in June through August, and reaching a peak of only 20°C in July (Fort McPherson station; Environment and Climate Change Canada 2019). Streamflow in the study area is driven primarily by snowmelt (Prowse et al. 2006). Such landscapes are susceptible to retrogressive thaw slump formation because they are underlain by ice-rich permafrost (Kokelj et al. 2013), which is easily destabilized by rainfall and warm air temperatures (Lacelle et al. 2010; Kokelj et al. 2015). Lacelle et al. (2015) found 189 thaw slumps in the Richardson Mountain-Peel Plateau that had been active since 1985, with a high density of these slumps found in the Peel Plateau. They also found an average 20-yr headwall retreat rate of 14.2 ± 5.0 SD m yr−1 from 11 randomly chosen thaw slumps in the Stony Creek and Vittrekwa River watersheds between 1990 and 2010 (Lacelle et al. 2015).

Details are in the caption following the image
Map of drift study sites in the Peel Plateau, Northwest Territories, Canada. Upstream sites are followed by a “U” and downstream sites are followed by a “D”. Paired sites share the same reach number. Sites numbers correspond with those in Chin et al. (2016). The slump upstream of site 38D corresponds to slump SF that is described in Zolkos and Tank (2019); other slumps are previously undescribed. The black star in the inset represents the study area in relation to the rest of Canada.

A control/impact design was used to assess biological changes associated with thaw slumps by sampling at paired upstream undisturbed and downstream disturbed sites (Fig. 1). Undisturbed sites had no upstream thaw slumps while disturbed sites had one thaw slump upstream. Morphology data have only been published for one of the slumps in our study: slump SF, upstream of site 38D (Fig. 1), which was found to have an average headwall height of 5.58 m (Zolkos and Tank 2019; Fig. 1). All slump disturbances were ongoing prior to and during sampling. Although the age of each slump (and therefore the exact duration of the disturbance event prior to sampling) was unknown (S. Kokelj pers. comm., August 2013), observations from previous studies confirm that some of the smaller slumps were active prior to 2010 (Chin et al. 2016) and a study by Lacelle et al. (2015) showed that slumps similar in size to our larger slumps were already active in the late 1990s and early 2000s, indicating that slump disturbances in our study were at least 3 yr old and in some cases had been active for decades. The study region covered a small geographic area, with all sites likely experiencing similar weather patterns, so within the sampling season, there should have been a similar frequency of precipitation-driven pulses from all the slumps (Kokelj et al. 2013, 2015).

Data collection

Benthic macroinvertebrate drift was assessed in August 2014 at five paired undisturbed and disturbed sites. Three drift nets (mesh size of 400 μm; frame dimensions of 15 cm × 15 cm) were placed at least five meters apart longitudinally along the stream reach and offset laterally to cover as much of the stream width as possible. Nets were left in the stream for 22.75–30.25 h; retrieval times varied due to the logistical difficulty of reaching some sites. Depth and velocity measurements were taken at the mouth of the drift net upon deployment and retrieval using the velocity head ruler method (modified from Carufel 1980). Drift samples were preserved in 99% isopropyl alcohol (in lieu of ethanol) and shipped to the University of New Brunswick for sorting and identification.

Sedimentation from thaw slump inputs was assessed by deploying sediment traps (n = 3 per site) alongside the drift nets in August 2014 (19–49 h). Containers (500 mL) filled with coarse gravel (approximately 2–3 cm in width) were placed in each stream flush with the sediments in the streambed (method modified from Welton and Ladle 1979). Depth and velocity measurements were taken at the rim of each container using the velocity head ruler method (modified from Carufel 1980). Once retrieved, debris from each trap was run through a 4 mm and 75 μm sieve to separate the gravel and stream sediments, and was stored in 99% isopropyl alcohol until processing.

Laboratory processing

Invertebrates known to exhibit drift behavior were sorted and identified to family (or subfamily for chironomids) from each drift sample, including aquatic insects from the orders Ephemeroptera (mayflies), Plecoptera (stoneflies), Trichoptera (caddisflies), Diptera (true flies), and Coleoptera (beetles) as well as Amphipoda (scuds) and Hydracarina (water mites) (Supporting Information Table S1). Sample QA/QC required that if the drift sorting error was greater than 5%, another sample was checked. If the average error of the two samples was still above 5%, all samples were re-sorted.

After drift samples were sorted and QA/QC requirements were met, the remaining debris from each sample was processed to determine the amount and composition of debris being transported downstream at each site (sensu O'Hop and Wallace 1983). Drift debris samples were washed through a 1 mm sieve to separate out coarse particulate matter (defined as sediment > 1000 μm), and because samples were collected using 400-μm-mesh drift nets, the remaining fine-sediment particulate matter was defined as covering the size range of 400–1000 μm. Samples were dried (at 60°C) to constant weight, dry-ashed (at 550°C), and weighed again to determine the organic and inorganic components of each sample. Sediments from the sediment traps were processed in the same manner, although samples were filtered through both a 1 mm and a 75 μm sieve, so fine-sediment classification for settling debris was 75–1000 μm. Although this resulted in different size fractions for fine particulate matter for drifting and settling sediments, these data were not combined or directly compared in the analysis. The data on fine sediments from drift nets, although covering a more restrictive size range, were retained to allow quantification of fine sediments in the drift, but this is acknowledged to likely represent an underestimate of fine particulate matter.

Statistical analyses

Sedimentation was assessed at drift sites to quantify drifting and settling material at each stream site and to see if it was associated with macroinvertebrate drift. The amount of organic and inorganic coarse and fine particulate matter was compared at paired undisturbed and single-disturbance sites to assess whether the amount of sedimentation was higher at thaw-slump impacted sites. Debris density (per net-hour) was calculated as mg 100 m−3 (Allan and Russek 1985) and summed for each site, and sediment measurements (from sediment traps) were converted to sedimentation rates (mg h−1) and also summed for each site. Velocities measured at each sediment trap were not significantly different at downstream sites, so sedimentation rates were not expected to be influenced by differences in stream flow between undisturbed and disturbed sites. Paired t-tests were run to determine if there was a significant difference in the amount of drifting debris or the rate of sedimentation between undisturbed and disturbed sites. Debris drift and sedimentation were further quantified by size fraction and organic content to evaluate whether thaw slumps led to increases in certain types of particulate matter. Coarse particulate organic matter (CPOM), coarse particulate inorganic matter (CPIM), fine particulate organic matter (FPOM), and fine particulate inorganic matter (FPIM) content in debris drift and sediment traps (calculated separately for debris density and sedimentation rate measurements) were compared at undisturbed and single-disturbance using one-tailed paired t-tests.

Macroinvertebrate drift was quantified as drift density to standardize drift measurements by accounting for differences in water volume sampled and deployment time at each site (Allan and Russek 1985) and summed across nets at each site. Drift density is the standard metric used to quantify and compare macroinvertebrate drift among stream sites (Smock 2006). To determine if there were differences in the number of drifting organisms upstream and downstream of thaw slumps, drift density (per net-hour) was compared at undisturbed and single-disturbance sites using a two-tailed paired t-test (as it was not clear a priori whether drift would be expected to be higher or lower with ongoing disturbance). To evaluate whether drift density was related to the number of benthic invertebrates inhabiting the stream reach, a correlation was run to see if drift density was associated with catch per unit effort from benthic samples collected in August 2014 from the same sites. Benthic samples were taken using Canadian Aquatic Biomonitoring Network protocols, which requires a 3-min traveling kick with a 400-μm-mesh kick net to collect the sample (Environment Canada 2012). Because thaw slumps have been shown to cause a significant decrease in macroinvertebrate abundance (Chin et al. 2016), macroinvertebrate drift was also quantified as proportional drift, or the drift density divided by catch per unit effort from benthic kick samples from August 2014, to correct for low abundance at disturbed sites. A two-tailed paired t-test was run to see if proportional drift changed between paired upstream and downstream sites.

Regression analysis was completed to determine if drift density and proportional drift were influenced by the rate of settling and density of drifting fine-particulate matter (FPM = FPOM + FPIM) at each site. We tested the regression models: Drift density = Drifting FPM, Drift density = Settling FPM, Proportional drift = Drifting FPM and Proportional drift = Settling FPM. Assumptions of the regression model were checked prior to analysis, and drift and FPM variables were log-transformed to stabilize variance in the models. There remained some evidence of heteroscedasticity in the relationship between log drift density and log drifting FPM that was due to one site with a low value of FPM, but regressions are robust to non-extreme violations of assumptions (Zar 2010). Paired t-tests and regressions were completed using RStudio (Version 0.98.501) running R 3.2.2 using the stats package established by the R Core Team (2017).

Results

Sediment movement

Comparison between undisturbed and disturbed sites showed evidence of significant increases in organic matter and inorganic sediments downstream of disturbance. There was a significant increase in the density of drifting FPOM moving downstream from undisturbed to disturbed sites (t = −2.26, df = 4, p = 0.043), but there was no change in the density of drifting CPOM (t = 0.70, df = 4, p = 0.74) or CPIM (t = 0.38, df = 4, p = 0.64), nor was there a change in FPIM (t = −1.99, df = 4, p = 0.059; Fig. 2a). In contrast, deposition of FPOM (t = −3.58, df = 4, p = 0.012) and FPIM (t = −3.00, df = 4, p = 0.020) showed a significant increase at disturbed sites compared to undisturbed sites. Both CPOM (t = −1.47, df = 4, p = 0.108) and CPIM (t = −1.77, df = 4, p = 0.076) showed no difference between disturbed and undisturbed sites, although when an ecological outlier was removed (the only site with higher CPOM and CPIM at the undisturbed than disturbed site), CPOM and CPIM were both significantly higher at disturbed sites (Fig. 2b). Average TSS (mg L−1) at undisturbed sites was 14.6 ± 19.2 SD and average TSS (mg L−1) at disturbed sites was 418 ± 292.6 SD (Supporting Information Table S2).

Details are in the caption following the image
Coarse particulate organic matter (CPOM), CPIM, FPOM, and FPIM (a) densities (mg 100 m−3) measured from drift nets at undisturbed (no-fill) and single-disturbance (gray-fill) sites and (b) sedimentation rates (mg h−1) measured from sediment traps at undisturbed (no-fill) and single-disturbance (gray-fill) sites in streams on the Peel Plateau, NT, Canada. Standard error bars are plotted. A * signifies a p value < 0.05 from paired t-tests.

Macroinvertebrate drift

Mean benthic abundance (from kick samples) at upstream sites was 1723 ± 1019 SD and at downstream sites was 397 ± 368 SD. Macroinvertebrate composition in drift samples at both undisturbed and disturbed sites were generally dominated by Chironomidae (midges), though other Diptera (true fly) families were abundant in some sites, as were families of Ephemeroptera (mayflies), Plecoptera (stoneflies), Coleoptera (beetles), and Acari (mites) (Table 1). There were no strong differences in macroinvertebrate community composition of drift samples between undisturbed and disturbed sites (Table 1), despite the disparity in total abundance from benthic kick samples.

Table 1. Mean percent Ephemeroptera, Plecoptera, Chironomidae, non-chironomid Diptera, Acari, and Coleoptera, and taxa richness (with range in parentheses) of drift samples collected from undisturbed and retrogressive thaw slump disturbed sites in August 2014 from the Peel Plateau, NWT, Canada.
Disturbance % Ephemeroptera % Plecoptera % Chironomidae % Non-chironomid Diptera % Acari % Coleoptera Richness
Undisturbed 2.4 (0–9.5) 8.8 (2.0–17.5) 64.6 (26.6–76.9) 7.1 (1.1–20.4) 11.7 (0.4–40.8) 0.7 (0–1.9) 11 (9–15)
Disturbed 6.0 (0–13.6) 15.1 (2.3–57.6) 44.3 (16.3–73.6) 10.5 (3.1–21.8) 19.0 (3.0–41.8) 2.2 (0–8.0) 13 (9–19)

Macroinvertebrate drift density was higher at all undisturbed sites (mean drift density = 56.7 organisms/100 m3) than disturbed sites (mean drift density = 15.1 organisms/100 m3). Average drift density of Ephemeroptera, Plecoptera, and Trichoptera taxa (EPT, a metric of water quality) was not significantly different between undisturbed sites (12.98 ± 8.49, 95% confidence interval) and disturbed sites (9.79 ± 12.92, 95% confidence interval). Average drift density of chironomids was also not significantly different between undisturbed sites (115.89 ± 113.25, 95% confidence interval) and disturbed sites (24.27 ± 23.95, 95% confidence interval). A comparison of paired undisturbed and disturbed sites indicated that there was a significant decrease in drifting macroinvertebrates downstream of a disturbance (t = −3.48, df = 4, p = 0.026; Fig. 3a). A significant negative relationship was evident between drift density and the rate of settling FPM (slope = −0.633, r2 = 0.47; Table 2; Fig. 4a), and the density of drifting FPM (slope = −0.608, r2 = 0.45; Table 2; Fig. 4b), although the latter was driven in part by one site with high drift and low drifting FPM.

Details are in the caption following the image
(a) Drift densities (organisms/100 m3) and (b) proportional drift (standardized no. of organisms/100 m3) at paired undisturbed (no-fill) and single-disturbance (gray-fill) sites measured in streams on the Peel Plateau, NT, Canada. Paired sites are separated by dotted vertical lines.
Table 2. Results from invertebrate drift density/proportional drift and settling/drifting FPM least-squares regressions measured in the Peel Plateau, NT, Canada. The regression models with coefficients, regression t-statistics, and p values for the slope term, the model r2, and residual mean squares (RMS) are shown. p values that were significant at α = 0.05 are indicated in bold.
Analysis Model t p value r2 RMS
Drift density vs. settling FPM log10(drift density) = −0.633*log10(FPM) + 2.99 −2.69 0.028 0.47 0.349
Drift density vs. drifting FPM log10(drift density) = −0.608*log10(FPM) + 2.04 −2.58 0.033 0.45 0.356
Proportional drift vs. settling FPM log10(proportional drift) = 0.441*log10(FPM) − 1.88 2.30 0.055 0.43 0.239
Proportional drift vs. drifting FPM log10(proportional drift) = 0.421*log10(FPM) − 1.23 1.69 0.136 0.29 0.267
Details are in the caption following the image
Simple linear regressions of drift density (organisms/100 m3) as a function of (a) rate of settling FPM (mg h−1) and (b) density of drifting FPM (mg 100 m−3), and proportional drift (standardized no. of organisms/100 m3) as a function of (c) rate of settling FPM (mg h−1) and (d) density of drifting FPM (mg 100 m−3) at undisturbed sites (white triangles) and disturbed sites (black dots) in streams on the Peel Plateau, NT, Canada.

Drift density and benthic catch showed a strong positive correlation (r = 0.82 on log-transformed data), which suggested that sites with high benthic abundance had high drift densities and vice versa. When drift samples were standardized based on the proportion of drifting invertebrates to total benthic abundance, we saw a proportionally higher number of drifting macroinvertebrates at disturbed than undisturbed sites. Proportional drift (represented as standardized no. of organisms/100 m3) was higher on average at disturbed sites (mean proportional drift = 0.18) than undisturbed sites (mean proportional drift = 0.10), particularly in sites with the highest rate of settling sediments (sites 38, 40, and 73), which had proportional drift as much as 2–4 times higher downstream of disturbance (Fig. 3b). However, the difference in proportional drift between disturbance levels was not significant (one-tailed paired t-test, p = 0.43) because two sites (71 and 76, both with low rates of settling sediments) had similar or higher proportional drift upstream of disturbance (Fig. 3b). Similar to drift density, average proportional drift of Ephemeroptera, Plecoptera, and Trichoptera taxa was not significantly different between undisturbed sites (0.08 ± 0.08, 95% confidence interval) and disturbed sites (0.09 ± 0.11, 95% confidence interval), nor was average proportional drift of chironomids significantly different between undisturbed sites (0.13 ± 0.06, 95% confidence interval) and disturbed sites (0.19 ± 0.12, 95% confidence interval). Site 73U showed unusually high proportional drift (proportional drift = 0.19) for the low rate of sedimentation occurring at this site (mean FPM rate = 13 mg h−1) compared to other undisturbed sites (mean FPM rate = 57.5 mg h−1) and was a statistical outlier in the regressions of proportional drift as a function of FPM. Sedimentation of FPM increased in August compared to July (mean sedimentation rate from July 9th to 10th, 2014 = 8.0 mg h−1) and turbidity increased in the week before sampling took place at this site (mean turbidity levels [± SD] of 13.0 ± 13.0 NTU before 13 August 2014 and 65.1 ± 30.1 NTU between 13 August 2014 and 20 August 2014). This short-term increase in turbidity and fine sedimentation prior to our sampling event may have triggered catastrophic drift leading to the higher proportional drift seen at this site. The samples at this site were also dominated by the subfamily Diamesinae, a group of chironomids common in cold waters (Milner et al. 2001) and prone to drift in Arctic streams (Saltveit et al. 2001). Due to the abiotic and biotic conditions found at this site, it was removed as an outlier from the regression of proportional drift and FPM. There was a strong positive relationship between proportional drift and settling FPM (slope = 0.441, r2 = 0.43; Table 2; Fig. 4b) and a weaker positive relationship between proportional drift and the density of drifting FPM (slope = 0.421, r2 = 0.29; Table 2; Fig. 4d).

Discussion

Retrogressive thaw slumps, which are increasing in magnitude and frequency across Arctic and alpine regions, notably change the physical and chemical properties of downstream reaches by increasing sediment loads (Kokelj et al. 2013; Chin et al. 2016). In our study, retrogressive thaw slumps were a significant source of drifting and settling fine sediments in downstream reaches. Under exposure to short-term perturbation, macroinvertebrate drift studies have shown increases in drift densities in response to increased sediment loads (Rosenberg and Wiens 1978; Culp et al. 1986; Suren and Jowett 2001; Molinos and Donohue 2009; Larsen and Ormerod 2010). However, the thaw slump disturbances in our study represented ongoing perturbation, with effects lasting at least 3 yr, a time period over which the effects of sedimentation on invertebrate drift have generally not been studied. We showed that chronic increases in fine sediments from long-term disturbance events were associated with lower drift densities downstream of thaw slumps, corresponding to low total abundances at these sites. In contrast, proportional drift (corrected for local invertebrate abundance) was higher at sites downstream of disturbance, suggesting that macroinvertebrate drift response to sediment inputs was still occurring long after the onset of the disturbance. Notably, these results suggest that drift density, the standard metric for measuring drift, does not work well when abundances at impacted sites are already low due to chronic disturbance. This finding has important implications for how other studies should investigate macroinvertebrate drift in response to chronic impacts.

Thaw slump impacts on organic and inorganic sediment fractions

Thaw slumps strongly influenced sediment loads downstream of debris flows, with increased levels of deposited fine (75–1000 μm) and suspended fine (400–1000 μm) sediments, and increased settling coarse sediments (> 1000 μm) at thaw-slump impacted sites. Average levels of TSS (spot measurements) were also an order of magnitude higher downstream of thaw slumps. This is consistent with previous studies that have described significant transport of fine sediments downstream of permafrost thaw slumps (see review in Vonk et al. 2015), with increases in TSS of several orders of magnitude observed downstream of the largest slumps in the Peel Plateau region (Kokelj et al. 2013; Chin et al. 2016). Fine sediment deposition and suspension is a major driver of degradation in aquatic systems as it can impair ecosystem structure and function (Wood and Armitage 1997; see review in Bilotta and Brazier 2008). Deposition of fine sediments can cause burial of organisms, lead to filling of interstitial spaces (Lenat et al. 1981; Lemly 1982), or limit oxygen levels on the streambed (Jones et al. 2012), whereas mobile, suspended sediments can limit light penetration and cause abrasion (Bilotta and Brazier 2008). Furthermore, the transport of sediments from permafrost thaw slumps has been associated with increased solute levels downstream of disturbance (Kokelj et al. 2013), and may lead to the release of bound contaminants from suspended or deposited sediments (Vonk et al. 2015). The combined effects of increased fine sediments from permafrost thaw slumps on the chemical and physical habitat of streams can create an inhospitable environment for benthic macroinvertebrates and trigger invertebrate drift (Naman et al. 2016).

A review by Wood and Armitage (1997) emphasized the importance of considering both the inorganic and organic components of sediments as they can behave differently in stream ecosystems. For example, inorganic sediments can clog the nets of net-spinning caddisflies and the guts of macroinvertebrates if ingested, while organic sediments may provide a food source for suspension feeders and collector-gatherers (Jones et al. 2012). However, O'Donnell et al. (2019) found that assimilation of depleted old carbon from permafrost thaw slumps into Arctic stream food webs by lower trophic levels was associated with declines in fish health (lower growth and energy density), which suggests that contributions of organic matter from slumps may negatively impact stream food webs. Organic components of sediments may also contribute to oxygen depletion in the system due to increased microbial activity (Bilotta and Brazier 2008; Jones et al. 2012). In our study, suspended sediments from thaw slump inputs appeared to be composed of similar amounts of inorganic and organic material (although this could be due to the loss through the nets of the finest sediment classes, which are generally inorganic), but far more inorganic than organic sediment was settling along the streambed, with an order-of-magnitude difference between the amount of settling organic and inorganic sediments at disturbed sites. Though our results suggested that inorganic sediments played a larger role in these slump-impacted systems, the relative importance of inorganic and organic fractions may differ in other regions where the composition of slumping permafrost differs (Frey and McClelland 2009; O'Donnell et al. 2016), with different implications for ecosystem health.

Sediments from ongoing thaw slumps as drivers of macroinvertebrate drift

The strong relationship between fine particulate matter and proportional drift indicated that fine sediment deposition was a driver of drift at these thaw-slump impacted stream sites. Notably, increases in proportional drift downstream of disturbance were found in sites with the highest rates of sedimentation, which suggests that there might be a threshold above which sedimentation from thaw slumps induces macroinvertebrates to enter the drift. Sedimentation can promote invertebrate drift entry directly through abrasion from suspended sediments (Culp et al. 1986) or burial from settling sediments (Jones et al. 2012). For example, mobile sediments can damage sensitive structures such as the external gills of many families of Ephemeroptera, or inhibit proper function of these organs through clogging, whereas sediment deposition can cause burial of sedentary invertebrates (e.g., Trichoptera, mollusks) or vulnerable life stages (e.g., pupa) (see review in Jones et al. 2012). Conroy et al. (2018) showed that many invertebrates experience more difficulty escaping finer sediment fractions than coarser ones, and that even mobile invertebrates (e.g., Ephemeroptera) can become entrapped in fine sediments when burial depth reaches 10 mm. There may also be indirect impacts of sedimentation through a reduction in habitat quality due to changes in substrate composition (Lenat et al. 1981; Suren and Jowett 2001) or changes to food availability (Hildebrand 1974; Suren and Jowett 2001), particularly through scouring or burial of biofilm (Levenstein et al. 2018). In our study, higher levels of both suspended and deposited fine sediments were associated with increased proportional drift densities, which suggests that both contributed to the lower benthic abundances at impacted sites. However, deposited fine sediments explained much higher variation in proportional drift, which indicates that fine sediment deposition may be a stronger driver of drift at these sites than drifting sediments.

Many researchers have examined the impacts of short-term (< 1 yr) sedimentation events on macroinvertebrate drift from disturbances such as construction projects or logging (e.g., Rosenberg and Wiens 1978; Doeg and Milledge 1991; Suren and Jowett 2001; Larsen and Ormerod 2010). In these short-term investigations, invertebrate drift densities increased in response to sedimentation events (Bilotta and Brazier 2008). In contrast, the impacts of prolonged sedimentation on macroinvertebrates have been less well defined, with few studies showing the effects of extended sedimentation disturbances such as those associated with ongoing inputs from permafrost degradation. At the beginning of a sediment pulse or disturbance, there are often episodes of catastrophic drift, an event where many invertebrates enter the drift once a sediment threshold is reached. It is possible that catastrophic drift (sensu Brittain and Eikeland 1988) occurred at our disturbed sites upon initiation of thaw slump activity (which likely occurred within the past few decades as retrogressive thaw slumps began to increase in size and frequency in the Peel Plateau; Lantz and Kokelj 2008), but this was not observed in our study, as these streams have experienced prolonged exposure to thaw-slump inputs. Mean abundance in upstream, undisturbed sites was roughly four times that of downstream (impacted) ones, and there were thus fewer individuals that could enter the drift in downstream sites. Consequently, abundance of macroinvertebrates was so low at sites downstream of disturbance that drift from these sites appeared to be lower than drift from upstream sites. However, when drift densities were corrected for macroinvertebrate abundance, proportional drift densities indicated that macroinvertebrate drift was higher downstream of disturbance when sedimentation rates were high, which suggested an ongoing response to prolonged thaw slump inputs. Correcting drift density for abundance accounts for catastrophic losses that occur at disturbance onset, as well as continued loss of individuals if perturbation does not cease, and it may provide a more appropriate measure of drift where impacts have been ongoing for a long period of time.

Sedimentation will continue to impart a physical stressor to aquatic systems with active thaw slumps, and we hypothesize that this stressor will impair macroinvertebrate communities until the disturbance stabilizes. Suren et al. (2005) predicted that long-term exposure to high turbidity levels would have significant effects on macroinvertebrate communities and lead to increased drift rates and reduced benthic densities. This was further supported by Molinos and Donohue (2009), who noted that the exposure time and sediment dose can have additive or synergistic effects on macroinvertebrate drift, which may further explain the strong negative relationship seen between sedimentation and macroinvertebrate densities. Since macroinvertebrate drift occurs on small spatial scales (Elliott 2003), it is unlikely that drift from upstream undisturbed areas will contribute to recovery of slump-impacted macroinvertebrate communities, particularly when the effects of permafrost thaw slumps extend far downstream of the disturbance. We hypothesize that macroinvertebrates drifting from upstream sites perish when exposed to thaw slump inputs as it is a non-selective pressure (Chin et al. 2016). Recovery of these systems seems unlikely as long as the system remains active, as Narf (1985) showed that insect colonization via drift will not occur until appropriate habitat and food resources are available. The results of our drift study along with those of Levenstein et al. (2018) show that habitat and food resources are impaired in thaw-slump impacted streams, thus low macroinvertebrate benthic abundance and drift densities will likely persist until the thaw slump becomes inactive.

Conclusions

Retrogressive thaw slumps are often long-lived, chronic stressors to stream systems that can change the structure and function of macroinvertebrate communities in impacted stream reaches. In contrast to the standard drift response at the onset of disturbance, these systems with prolonged impacts from permafrost thaw slumps did not display a typical “catastrophic” stressor response. Instead, our results indicate that prolonged periods of high sediment loading associated with ongoing thaw slumps contribute to low macroinvertebrate abundance and increased proportional macroinvertebrate drift in Arctic streams, relative to undisturbed systems. The implication of these findings for other studies of chronic disturbance is that it is more appropriate to consider proportional drift densities, which quantify the magnitude of response relative to the total population and correct for variation in benthic abundances across sites. With the intensity of retrogressive thaw slump impact expected to increase with a warming climate, we predict that recolonization of impacted stream reaches through invertebrate drift from upstream, undisturbed reaches will not occur until these features stabilize. Therefore, macroinvertebrate abundances will likely remain low in thaw slump impacted reaches, as will the (uncorrected) density of drifting invertebrates. This may have negative effects on stream ecosystem function, as macroinvertebrates represent a critical link in the flow of energy from lower to higher trophic levels of stream food webs. Furthermore, with retrogressive thaw slump formation accelerating in other parts of the Arctic, including the high and low Canadian Arctic (Segal et al. 2016; Ward Jones et al. 2019), northern Alaska (Balser et al. 2014), central Russia (Séjourné et al. 2015), and the Qinghai-Tibet Plateau region of western China (Luo et al. 2019) similar impacts on macroinvertebrate drift and stream food webs may soon be evident across the circumpolar and alpine regions.

Acknowledgments

We are grateful for the superb technical assistance of K. Roach, D. Hryn, D. Halliwell, K. Heard, and E. Luiker. We are particularly indebted to K. Chin, S Kokelj, G. Vaneltsi, S. Tetlichi, C. Firth, and D. Colin. Special thanks to the Aurora Research Institute in Inuvik for providing field and logistical assistance. We acknowledge T. Lantz for providing GIS data. Constructive comments from R. Cunjak, M. Gray, A. Alexander, S. Tank, and anonymous reviewers were much appreciated and helped improve this manuscript. Funding to J. Culp and J. Lento was provided by the Cumulative Impacts Monitoring Program (CIMP-Government of the Northwest Territories), as well as funding from Environment and Climate Change Canada and a Natural Sciences and Engineering Research Council of Canada Discovery Grant to J. Culp. The Polar Continental Shelf Program provided funding to support transportation and logistical support.

    Conflict of Interest

    None declared.