Effect of snow cover on the vulnerability of lemmings to mammalian predators in the Canadian Arctic

Rapport de stage effect of snow cover on the vulnerability of lemmings to mammalian predators in the Canadian Arctic, tutoriel & guide de travaux pratiques en pdf.

Materials and methods

Study area

The study was conducted on the south plain of Bylot Island, Sirmilik National Park, Nunavut Territory, Canada (73°08’N, 80°00’W). The main study area (70 km2) consists of tundra polygons, thaw lakes and ponds forming wetlands in parts of the valley bottom and drier mesic habitat in the upland areas and nearby slopes and hills (see Bilodeau et al. 2013b and Duchesne et al. 2011b for details). Small, intermittent streams running through upland areas are often located at the bottom of small gullies where vegetation along the stream edge is similar to that of wetland habitat. We considered this habitat (gully) as a 3rd habitat category because it is conducive to high snow accumulation and can be heavily used by lemmings in winter (Duchesne at al. 2011b). Only 2 species of rodents are present, the brown (Lemmus trimucronatus) and the collared lemming (Dicrostonyx groenlandicus), though the former species is more abundant, especially in peak years (Gruyer et al. 2010). Arctic foxes and ermines are present throughout the winter on Bylot Island. Average fox density is approximately 0.1 individual/km2 on the south plain of Bylot, where up to 25 pairs may breed in a good lemming year (Legagneux et al. 2012).
Average air temperature at our study site from October to June is -23.4 ± 0.4 (SE) °C and average temperature under the snow, where lemmings live in winter, is -15.8 ± 0.4 °C (Bilodeau et al. 2013b). A permanent snow cover is usually established in late September and melting occurs in mid-June. Maximum snow depth typically occurs in May and average snow depth at the end of the winter is 33.8 ± 2.5 cm (Bilodeau et al. 2013b), although snow can be as much as 2m deep in areas conductive to high snow accumulation (e.g., gullies). The snow pack is very dry, owing to the very cold temperature prevailing during the winter.

Fox hunting behavior

We sampled fox predation events over a 980 ha area from 19 May to 2 June 2010 and over a 240 ha area from 17 to 26 May 2011. Fresh fox tracks were found opportunistically while travelling on foot or on snowmobile through the study area. When encountered, they were back-tracked to find signs of predation attempts on lemmings. Tunnels in the snow pack that were going straight down to the subnivean space and that were about the size of a fox body were considered to be predation attempts. These tunnels were easy to distinguish from other holes, such as those made by foxes to locate a den, which were associated with characteristic topographical features (Szor et al. 2008). We could recognize 2 hunting techniques used by foxes. The first one involved digging into the snow, as evidenced by digging marks, a small pile of snow near the tunnel entrance, and the rough edges of these entrances. The attempts were primarily directed at lemming nests as nest remains were always found at the bottom of these holes. The 2nd type of tunnel was narrower with no digging marks and often no signs of lemming nests at their bottom. No obvious pile of snow appeared near the tunnel entrance, which had rather smooth edges. Observations of foxes at distance confirmed that these occurred when immobile foxes listened intensely before jumping and pounding through the snow cover, most likely to catch a lemming that they heard moving under the snow. After classifying each type of tunnel encountered (dig or jump), snow depth and hardness were measured at the site (see details below).

Ermine predation

We sampled ermine winter predation annually from 2007 to 2011 by sampling lemming winter nests across the entire study area shortly after snow-melt. We used 500 m transects randomly distributed among our 3 habitats (wet, mesic, and gully habitats; 25, 10, and 20 transects per habitat in 2007, 2008, and 2009-2011, respectively). All winter nests found while walking along the transect line were removed and their position and perpendicular distance to the transect were noted. Nests found opportunistically, away from transects, were also collected and used for the determination of ermine predation rate.
All winter nests were dissected in the laboratory to determine whether they were used by brown or collared lemmings. We based species identification on the length, form, and color of feces found in nests (Duchesne et al. 2011a; MacLean et al. 1974). Collared lemmings have 4-6 mm long, dark reddish feces, while brown lemmings produce 6-10 mm long, green-brown feces. Lemmings reproduce under the snow in their winter nests (Millar 2001; Stenseth and Ims 1993). We determined if reproduction (defined as the production of weaned young) had occurred based on the presence of small juvenile feces in nests (Duchesne et al. 2011a). We also examined nests for signs of ermine predation. When ermines prey on lemmings, they often line the nest with their prey’s fur and use the nest themselves, or they leave rodent body parts and partially eaten carcasses in the nest (MacLean et al. 1974; Sittler 1995). Fur lining and lemming body parts were thus used as evidence of predation.

Snow measurements

We measured snow depth with a graduated rod and snow hardness with a custom-made penetrometer (McClung and Schaerer 2006). The penetrometer was a conical-shaped object of dimensions similar to a fox head and about the same mass as an adult fox (3 kg—Audet et al. 2002). The penetrometer was dropped from 60 cm above the snow and we noted how deep it penetrated the snow. Recorded depths of penetration were always <30 cm. For ease of interpretation, we express snow hardness as 30 – penetration depth (highest values represent hardest snow).
Depth and hardness measurements were done at 85 locations where we had found a lemming winter nest in the previous year in 2010 (n = 81 in 2011), as well as at 93 random locations in 2010 and 90 in 2011. These measurements were taken during the same period as those taken at the fox tunnels, before any significant snow-melt had occurred. There was negligible precipitation during the measurement periods and average air temperature during this period was -8.8 ± 1.4 °C. Measurements could not be taken at nest locations of the current year because nests can only be found after snow-melt. To verify if areas of deep accumulation were consistent between years, we selected 30 random locations where snow depth was sampled in 2010 and 2011. Snow depth at these locations was highly correlated between years (r = 0.76, t28 = 6.15, P < 0.001).
Snow depth at winter nests where predation by ermine had been detected was estimated in the following spring using the same procedure as described above. The hardness of snow covering nests predated by ermines could not be estimated because it is unlikely to be consistent between years. Indeed, snow hardness can be affected by multiple climatic factors such as snowfall, air temperature, wind erosion, and deposition (Pielmeier and Schneebeli 2002), which vary annually.

Lemming nest density

We estimated nest density on each transect to obtain a local index of lemming density in the area of the transect. Nest density was based on all nests found within 10 m on either side of the transect. This distance corresponds to the effective detection distance for sampling winter nests in this area (Duchesne et al. 2011b). We estimated nest density by dividing the number of nests found by the 1 ha (500 m x 20 m) transect area. Nests found opportunistically, away from transects, were not used to estimate nest densities. Based on the overall abundance of winter nests (Table 1) and live-trapping conducted during the summer (see Bilodeau et al. 2013d), lemming abundance was high in 2008, 2010, and 2011 (3.8 ± 0.8, 4.7 ± 0.5 and 8.0 ± 0.9 lemmings/ha respectively), intermediate to low in 2007 (0.3 ± 0.1 lemmings/ha), and very low in 2009 (0.2 lemmings/ha).

Statistical analyses

To test if snow depth and hardness had an impact on digging and jumping predation attempts by foxes, we compared depth and hardness measurements at tunnels made by foxes, at non-predated winter nest locations, and at random locations using ANOVA (separate analyses for digs and jumps). To test if snow depth had an effect on predation by ermines, we compared snow depth at predated nest sites, non-predated nest sites, and random locations also using an ANOVA. Interaction between year (2010-2011 for fox and 2007-2011 for ermine) and treatment (predated, non-predated, and random) was always tested. When our treatment was significant, we used Tukey’s test to compare individual means (or Least Square Means, LSM, when there was a significant interaction).
To test the hypotheses that predation by ermines was greater in habitats preferred by lemmings in winter and in areas of high nest density, we used a logistic regression to determine if the presence or absence of predation on a nest varied according to year (2007-2011), habitat (wet, mesic, gully), average nest density along the transect, and presence or absence of reproduction and species (brown, collared lemming or both). Two-way interactions among habitat, year, density, reproduction, and species were examined. We determined the most important variables using Akaike’s Information Criterion adjusted for small sample size (AICc) and AICc weight (Burnham and Anderson 2002). We used model averaging to calculate the slope parameter and associated error of the most influential variables. Because there were 9 times less predated than non-predated nests, we verified if this could bias our results by conducting a 2nd analysis where we reduced (through random sub-sampling) the number of non-predated nests to that of the predated ones. Due to the small resulting sample size, we repeated the sub-sampling and associated analysis 10 times. We then calculated the average cumulative AICc weight of each variable across the 10 analyses. Since the analysis based on sub-sampling did not change our results, we only report those based on the full samples. Analyses were conducted in R 2.13.1 (R Development Core Team 2010).

Results

Effects of snow cover on fox hunting behavior

We found 56 fox tunnels in 2010 (47 digs and 9 jumps) but only 9 in 2011 (all digs). Snow depth varied significantly among treatments for digs (F2,385 = 5.73, P = 0.004). Lemming winter nests (whether predated or not) had a deeper snow cover than random locations (P < 0.001), but snow depth did not differ according to whether nests had been predated or not (P = 0.988; Fig. 1) and this was consistent across years (F2,385 = 0.74, P = 0.480). Snow depth also varied significantly among treatments for jumps in 2010 (F2,184 = 14.84, P < 0.001) but this time snow was shallower at predated than at non-predated nests (P < 0.001; Fig. 1).
Snow was harder in 2011 than in 2010 (random locations, 2010: LSM = 5.17 [3.65-6.69, 95% CI]; 2011: LSM = 11.08 [9.55-12.61]) and we detected a significant treatment x year interaction for the effect of snow hardness on digs (F2,385 = 5.88, P = 0.003; Fig.2). In 2010, snow hardness at predated nests (LSM = 8.58 [6.81-10.35]) did not differ from that at non-predated nests (LSM = 7.73 [6.53-8.93]), but all nests were under harder snow than random locations (LSM = 5.17 [4.03-6.32]; Fig. 2). In 2011, hardness tended to be lower at predated than non-predated nests but the difference was not quite significant (LSM = 4.89 [-1.13-10.91] versus 10.48 [8.46-12.49], respectively; Fig. 2). Snow hardness also varied significantly between jumps, non-predated nests, and random locations (F2,180 = 6.15, P = 0.003). Lemming winter nests had a harder snow cover than random locations (P = 0.004) but snow hardness at jumps did not differ from that at non-predated nest locations (P = 0.850; Fig. 2).

Effects of snow cover on predation by ermine

The proportion of transects with winter nests was higher in both mesic and gully habitats than in wet habitat (χ22 = 10.8, P = 0.004) and the density of nests was generally highest in gully habitat, intermediate in mesic, and lowest in wet habitat (F2,8 = 6.33, P = 0.022; Table 1). On all transects sampled from 2007 to 2011, 886 winter nests were found and 33 of them (3.7%) were predated by ermine. Snow depth varied significantly among treatments (F2,360 = 5.55, P = 0.004) because lemming winter nests had a deeper snow cover than random locations (P < 0.001; Fig. 1) but snow depth at nests predated by ermines did not differ from non-predated ones (P = 0.330) and this was consistent across years (F2,360 = 1.70, P = 0.185).