Content uploaded by Scott Hotaling
Author content
All content in this area was uploaded by Scott Hotaling on Jul 02, 2019
Content may be subject to copyright.
1
Rolling stones gather moss: Movement and longevity of moss balls on an Alaskan 1
glacier 2
3
Scott Hotaling1, Timothy C. Bartholomaus2, and Sophie L. Gilbert3
4
5
Affiliations: 6
1 School of Biological Sciences, Washington State University, Pullman, WA, USA; ORCID = 7
0000-0002-5965-0986 8
2 Department of Geology, University of Idaho, Moscow, ID, USA; ORCID = 0000-0002-1470-9
6720 10
3 College of Natural Resources, University of Idaho, Moscow, ID, USA; ORCID = 0000-0002-11
9974-5146 12
13
Correspondence: Sophie L. Gilbert, College of Natural Resources, University of Idaho, 14
Moscow, ID, 83844, USA; Email: sophiegilbert@uidaho.edu; Phone: (208) 885-8605 15
16
Abstract: 17
Glaciers support diverse ecosystems that are largely comprised of microbial life. However, at 18
larger, macroscopic scales, glacier moss balls (sometimes called “glacier mice”) develop from 19
impurities in the ice and represent a relatively rare biological phenomenon. These ovoid-shaped 20
conglomerations of dirt and moss are only found on some glacier surfaces and provide key 21
habitats for invertebrate colonization. Yet, despite their development and presence being widely 22
reported, no targeted studies of their movement and longevity have been conducted. This 23
knowledge gap is particularly important when considering the degree to which glacier moss 24
balls may represent viable, long-term biotic habitats on glaciers, perhaps complete with their 25
own ecological succession dynamics. Here, we describe the movement and longevity of glacier 26
moss balls on the Root Glacier in southcentral Alaska, USA. We show that glacier moss balls 27
move an average of 2.5 cm per day in herd-like fashion, and their movements are positively 28
correlated with glacier ablation. Surprisingly, the dominant moss ball movement direction does 29
not align with the prevailing wind nor downslope directions; instead, we propose that it depends 30
on the dominant direction of solar radiation. We also show that glacier moss balls are relatively 31
long-lived, with a lifespan in excess of 6 years and annual survival rates similar to large 32
vertebrates. Finally, we observed moss ball formation on the Root Glacier to occur within a 33
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
2
narrow, low albedo stripe downwind of a nunatuk, a potential key source of moss spores and/or 34
fine-grained sediment that interact to promote their formation. 35
36
Keywords: cryobiology, glacier mice, glacier biology, jokla-mys, Root Glacier, Wrangell-St. 37
Elias National Park 38
39
Introduction: 40
Glaciers have long been overlooked as important components of global biodiversity, but 41
it is now clear that they host thriving, multi-trophic ecosystems (Anesio and Laybourn-Parry 42
2012), supporting taxa from microbes to vertebrates (Rosvold 2016; Dial et al. 2016; Hotaling et 43
al. 2019; Hotaling et al. 2017b). Most biological activity on glaciers occurs within surface ice 44
where microorganisms take advantage of nutrients that are either wind-delivered or generated 45
in situ (Hotaling et al. 2017b). In addition to a nutrient input, impurities on the glacier surface can 46
drive the development of at least two potential “hotspots” of biological diversity on glaciers: well-47
studied cryoconite holes (depressions in the ice surface caused by local melt, Anesio et al. 48
2017) and glacier moss balls (ovular conglomerations of moss and sediment that move on the 49
glacier surface, Coulson and Midgley 2012). 50
Under the right conditions, a small piece of rock or other impurity can set in motion the 51
formation of a glacier moss ball [also referred to as “jokla-mys” (Eythórsson 1951), “glacier 52
mice” (e.g., Coulson and Midgley 2012), or “moss cushions” (e.g., Porter et al. 2008)]. On a 53
local scale, glacier moss balls are typically distributed with some degree of local clustering (e.g., 54
~1 glacier moss ball/m2; Fig. 1). While more immobile moss aggregations have been observed 55
on glaciers elsewhere (e.g., East Africa, Uetake et al. 2014), true glacier moss balls appear to 56
be particularly rare, having only been described on a few geographically disparate glaciers in 57
Alaska (Shacklette 1966; Heusser 1972), Iceland (Eythórsson 1951), Svalbard (Belkina and 58
Vilnet 2015), and South America (Perez 1991). Many different moss species have been found in 59
glacier moss balls (Shacklette 1966; Heusser 1972; Perez 1991; Porter et al. 2008), suggesting 60
that they are not dependent on specific taxa, but instead their development is likely driven by 61
the interaction of suitable biotic (e.g., availability of moss spores) and abiotic (e.g., growth 62
substrate) factors. However, the specific steps and timeline underlying glacier moss ball genesis 63
remains unclear. 64
An intriguing aspect of glacier moss balls, and one that is almost certainly partially 65
responsible for their “glacier mice” namesake, is their movement. Though the speed and 66
direction of glacier moss ball travel has not been studied, it has been posited that they move by 67
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
3
inducing the formation of an ice pedestal, then rolling or sliding off of it (Porter et al. 2008). Moss 68
balls first reduce local albedo by shielding the ice beneath them from sunlight and locally 69
reducing the ablation rate. As the surrounding ice melts, the glacier moss ball is left on an 70
elevated pedestal. Eventually, a threshold is reached where the moss ball falls from its pedestal 71
and the process begins anew, potentially including a “flip” of the moss ball that exposes what 72
was previously their underside (Porter et al. 2008). However, the speed and direction of moss 73
ball movement has not been tested, though it has been suggested that their movements 74
generally track the downslope direction of their local habitat (Porter et al. 2008). 75
Where they do occur, glacier moss balls contribute to glacier biodiversity by offering a 76
thermally buffered, island-like habitat on the glacier surface that can host a wide array of 77
invertebrates (Coulson and Midgley 2012). On Icelandic glaciers, moss balls contain 78
invertebrate communities dominated by springtails (Collembola), tardigrades (Tardigrada), and 79
nematodes (Nematoda; Coulson and Midgley 2012). While an array of potential food resources 80
are available on glacier surfaces (Hotaling et al. 2017b), these are typically only exploited by 81
invertebrates on the margins of glaciers (e.g., springtails, spiders, grylloblattids) because 82
suitable on-glacier habitat is lacking (Mann et al. 1980). Glacier moss balls may therefore 83
provide key habitable islands on the glacier that facilitate wider resource exploitation versus 84
glaciers without moss balls (Coulson and Midgley 2012). It is also possible that glacier moss 85
balls, which have not been shown to be inhabited by larger predatory insects (e.g., 86
grylloblattids) may provide prey refuge that are sufficiently removed from the typical foraging 87
areas of their predators. Either way, it is clear that glacier moss balls represent important habitat 88
for glacier-associated fauna yet basic aspects of their ecology (e.g., longevity and movement) 89
are unknown. 90
In this study, we took an integrated behavioral ecology and geophysical approach to the 91
study of glacier moss balls to answer three basic questions about their life history: (1) What is 92
the lifespan of a glacier moss ball? (2) How quickly do they move and is their movement 93
idiosyncratic or herd-like? (3) Are the movements of glacier moss balls linked to the ablation of 94
the glacier itself? The answers to these questions have implications for invertebrate fauna in 95
glaciated ecosystems, nutrient cycling (both directly via moss ball decomposition and indirectly 96
as supporting habitat for biotic communities), and feedback between glacier moss balls and 97
local ablation rates. Beyond biotic interactions and ecosystem dynamics, glaciers are rapidly 98
receding worldwide (Gardner et al. 2013; Larsen et al. 2015; Roe et al. 2017) and their 99
diminished extents will almost certainly affect the persistence of glacier moss balls on local and 100
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
4
global scales. Thus, it is important to better understand these unique micro-ecosystems before 101
their habitats are lost. 102
103
104
Fig. 1. a) Our study site (solid green square) on the Root Glacier in southcentral Alaska, USA, within
105
Wrangell-St. Elias National Park. Contour lines are spaced every 100 m in elevation. The dashed square
106
represents the field of view shown in panel (b). The inset map shows the location of the Root Glacier
107
(white star) within Alaska. b) Satellite image of the study site (green square) showing the confluence of
108
the Root and Kennicott Glaciers with the Donoho nunatak to the northwest. The image was recorded on
109
19 June 2013. c) A landscape view looking northwest of the study site dotted with glacier moss balls. d
) A
110
close-up view of a glacier moss ball with the type of bracelet tag used in this study.
111 112
Materials and methods: 113
Study area 114
A
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
5
We conducted fieldwork over four years (July, 2009 - July, 2012) on the lowest portion of 115
the Root Glacier, a major tributary to the Kennicott Glacier, in the Wrangell Mountains in 116
Wrangell-St. Elias National Park, Alaska, USA (Fig. 1a). Our study area (61.5076° N, 142.9172° 117
W, ~700 m elevation) spanned a ~15 x ~40 m (600 m2) area of glacier ice selected for its 118
especially high concentration of moss balls. The site has a gentle slope, dipping 3° east-119
northeast (N75°E) and is found between two medial moraines (Fig. 1b), each ~100 m away. 120
Moss ball concentrations decrease both up- and down-glacier and are absent from the coarse-121
grained (> 5 cm) rock that covers the adjacent medial moraines. 122
We estimated the proportion of fine-grained sediment cover on the ice within our study 123
area by applying image processing techniques in the Python package scikit-image (Van der 124
Walt et al. 2014) to two vertical photographs taken of representative ice surfaces. Pixel 125
brightness contrasts between ice and sediment are most distinct within the blue band of the red-126
green-blue images, so we differentiated between sediment (dark pixels) and ice (bright pixels) 127
by binarizing the blue band with Otsu’s thresholding method. We then performed a 128
morphological opening to diminish the influence of light-colored sediment grains set within the 129
otherwise dark sediment cover. Finally, we quantified the areal sediment cover as being 130
approximately equal to the number of dark colored pixels relative to the total number of pixels in 131
the binarized images. 132
133
Mark-recapture 134
During the summer of 2009, we tagged 30 glacier moss balls with a bracelet identifier 135
(Fig. 1d). Each bracelet consisted of a unique combination of colored glass beads (~2-3 mm in 136
diameter) threaded on aluminum wire. Bracelets were threaded through the moss ball center 137
and pulled snug so as to not protrude beyond the moss ball’s exterior and interfere with 138
movement. We returned eight times during the 2009 season to re-survey moss balls and record 139
their movements. We followed up our initial surveys with annual visits from 2010-2012. During 140
each survey, we visually inspected in and around the core study area multiple times in an effort 141
to recapture moss balls. As part of this process, we visually inspected each moss ball in the 142
area for any sign of a bracelet tag. After inspection, we replaced each moss ball in the exact 143
location and orientation as it was found. 144
145
Moss ball movement and glacier ablation 146
We assessed moss ball movement over 54 days in 2009. As benchmarks for their 147
movement, we installed three ~1.3 cm PVC tubes into the glacier. Each stake was drilled ~60 148
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
6
cm into the glacier. Stakes were installed in a triangle that spanned the study area and served 149
two purposes. First, the stakes provided a reference against which the location of each moss 150
ball was measured. Second, they allowed us to measure glacier ablation (i.e., the distance the 151
ice surface moves vertically down) over the same study period so we could test for links 152
between moss ball movement and the rate of glacier ablation. 153
To track glacier moss ball movement, we measured the distance between re-identified 154
moss balls and each of the reference stakes for each visit to the site. Next, for each moss ball, 155
we calculated three independent positions within our field site--one for each of the three pairs of 156
reference stakes. We assigned the location of a surveyed moss ball to the mean of these three 157
relative positions and constructed a location covariance matrix for each measurement, to assign 158
uncertainties to surveyed locations. After diagonalizing the covariance matrix, we identified the 159
size (eigenvalues) and orientation (eigenvectors) of an uncertainty ellipse around each mean 160
location. Major and minor axes of the uncertainty ellipse were defined as twice the square root 161
of the eigenvalue lengths, such that each error ellipse represented a 2 error window. Thus, 162
assuming independent, normal errors, we are 95% confident that the true location of each moss 163
ball fell within its error ellipse. While we used stakes for most of the measurement period, we 164
were forced to switch to washers (~5 cm in diameter) laid flat on the ice surface later in the 165
season, during a period when we were unable to drill the benchmark stakes sufficiently deep to 166
avoid melting out between visits to the study area. Before transitioning from benchmark stakes 167
to washers, we tested the stability of the washers to ensure that they did not slide over the ice 168
surface. Over a 5-day period, we did not detect significant washer movement (outside of 2 169
uncertainty). Final measurements (11 August 2009) and calculations were made relative to the 170
washers. 171
For the purposes of quantifying glacier ablation, the height of each stake above the local 172
ice surface was re-measured during each visit and periodically re-drilled into the ice as 173
necessary. Ablation reported in this study is the mean ice surface lowering rate calculated for 174
each of the three stakes. As an assessment of ablation uncertainty, we also calculated the 175
maximum deviation of any single stake’s ablation rate from the overall mean. 176
177
Longevity 178
We sought to understand how long glacier moss balls survive, particularly across 179
individual seasons. We hypothesized that moss balls might survive better or worse in some 180
years due to variation in environmental conditions (e.g., precipitation, freeze-thaw cycles) or 181
random chance (e.g., a crevasse opening within a key area). Furthermore, we wanted to know 182
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
7
not only how likely we are to detect glacier moss balls, given that they had persisted within the 183
study area, but also if our detection probability varies among years. To do this, we fit capture-184
recapture models of annual survival to each glacier moss ball included in the study. Because 185
moss balls were individually marked but were not equipped with radio-transmitters or other 186
devices which would allow us to know their ultimate fates, we applied Cormack-Jolly-Seber 187
(CJS; Lebreton et al. 1992) survival models. These CJS models develop a “capture history” of 188
each moss ball to estimate apparent survival (i.e., the probability that an individual is in the 189
population at time i and still in the population at time i+1) and probability of detection if they 190
persisted within our study area. Survival estimates from CJS models only represent apparent 191
survival because emigration cannot be estimated from survival data with unknown fates (i.e., we 192
did not know if a tagged moss ball had disaggregated, lost its identifying bracelet, or was no 193
longer in the study area). Therefore, our estimates of apparent survival are likely to 194
underestimate true survival (e.g., a moss ball might have lost its bracelet or moved out of the 195
study site). In addition, CJS models also account for imperfect detection, which in our case 196
would be if a moss ball persisted within our study area but was overlooked. 197
Using our individual moss ball annual detection data (1 = detected, 0 = not detected), we 198
fit four competing CJS survival models, including the null model [no effect of year on apparent 199
survival ( ) or detection probability (p); Model 1)], an effect of year on (Model 2), an effect of 200
year on p (Model 3), or an effect of year on both and p (Model 4). We then selected the 201
model(s) best supported by our data using Akaike’s information criterion (AIC; Akaike 1998), 202
adjusted for small sample size (AICc; Hurvich and Tsai 1989). Our model selection approach 203
was based on model likelihoods and models were penalized for extra parameters to favor 204
parsimony. 205
Finally, we calculated the average life expectancy of a glacier moss ball. To do this, we 206
used annual survival rates based on life-table analysis (Deevey Jr 1947; Millar and Zammuto 207
1983), in which average life expectancy was calculated as -1/ln(Annual Survival Rate). Because 208
this estimation of life expectancy is quite sensitive to annual survival rate, we calculated it for 209
both the lowest annual survival rate and the mean annual survival rate. Thus, the true average 210
life expectancy might be substantially greater than the conservative values estimated here. This 211
framework for estimating average life expectancy does not account for potentially higher or 212
lower mortality rates when glacier moss balls are first forming or nearing the end of their 213
lifespans. 214
215
Results: 216
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
8
Study area 217
Our study area was located on a “bare ice” glacier surface, between two medial 218
moraines covered by coarse-grained, angular, rock debris. However, two types of sediment 219
distinguish the study area surface from what would be considered clean, pure, water ice. First, 220
glacier moss balls were found amidst gravel and small boulders (< 30 cm diameter), spaced 221
every ~1 m. Second, the ice surface within the study area has an unusually pervasive, fine-222
grained sediment cover, ~1-3 mm thick, which partially blankets the otherwise bare ice. Image 223
processing indicated that this fine sediment covers approximately 70% of the study area 224
surface. This low albedo sediment cover is visible in all inspected satellite imagery of the site 225
and first appears at lower concentrations emerging from cleaner ice ~1 km northwest of the 226
study site (Fig. 1b). Down-glacier of the study site, the low albedo region extends ~1.7 km as a 227
~300-m-wide, rounded finger that spans adjacent medial moraines, in a manner consistent with 228
wind-deposited dust, draping over underlying geomorphic features. Therefore, we interpreted 229
the southeast (135°) trend direction of this low albedo finger to be the prevailing, down-glacier, 230
katabatic wind direction. 231
During the 26 days of glacier ablation measurements, the ice surface lowered by 1.91 m
232
due to melt and sublimation. Ablation rates ranged from 5.8-9.6 cm per day (cm/d) between 233
measurement times, and averaged 7.3 cm/d. 234
235
236
Fig. 2.
(A) Locations of surveyed glacier moss balls throughout the survey period. Most likely locations of
237
each moss ball are shown with small filled circles relative to an arbitrary, local grid system. Ellipses
238
surrounding each moss ball indicate 2 uncertainty (i.e., 95% confidence) of their location. Thin black
239
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
9
lines connect consecutive surveyed locations for individual moss balls. The red rectangle identifies the
240
location of the large-scale view in panel (B). (B) A zoomed in view of movement patterns for six glacier
241
moss balls (red square in A), showing their similar azimuths.
242 243
244
Fig. 3.
(A) A comparison of glacier moss ball movements versus the dominant wind (dashed red line) and
245
downslope (dashed blue line) directions. Direction of each moss ball’s motion between measurement
246
times is shown with thin gray lines, while the bold black line indicates the median direction of all glacier
247
moss ball movements. (B) Glacier moss ball movement versus ablation rate. Median ablation rate is
248
indicated with a bold red line, while the mean +/- the maximum absolute deviation from the mean are
249
shown with thin red lines. The median speed of glacier moss balls is shown with the bold blue line, while
250
the 25th and 75th percentile speeds are shown with thin blue lines. Numbers in circles along the bottom
251
of the plot represent the number of moss balls surveyed at each timepoint (single measurements not
252
indicated).
253
d
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
10
254
Movement 255
Glacier moss ball movements varied systematically over the study period, with increases 256
and decreases that coincided with changes in direction (Figs. 2-3). Median moss ball speed was 257
2.5 cm/d, but their rates varied widely throughout the season. The median speed started at 1.8 258
cm/d in late June, increased to 4.0 cm/d at the start of July, then slowed to 2.0 cm/d during late 259
July/early August. The maximum observed speed for any glacier moss ball was 7.8 cm/d during 260
the 5-day period from July 9-14 (excluding two outlier speeds that were more than 8 interquartile 261
ranges greater than the median, 14.2 and 21.0 cm/d, and which were based upon particularly 262
uncertain moss ball positions). The interquartile range of moss ball speeds was approximately 263
50% of the median speed; thus, these observed increases and decreases in speed reflect 264
changes in the entire population of moss balls. 265
The direction of glacier moss ball movements was not random. Rather, glacier moss 266
balls underwent clear changes in their direction of motion (i.e., azimuth) throughout the summer 267
season (Fig. 3a). While individual moss balls moved in many directions, when viewed in 268
aggregate, azimuths of the population clearly cluster over time. Early in the season, median 269
moss ball motion was south-southeast (165°) but over the ensuing weeks azimuths 270
progressively increased, such that at the end of the measurement period the median azimuth 271
was west-southwest (240°; Fig. 3a). 272
Considering speeds and azimuths together, we see the moss ball population initially 273
moving at 2 cm/d to the south for 9 days, then the group nearly doubles its speed to 4 cm/d 274
while deviating slightly to the right (towards the west). After a week at these maximum speeds, 275
speeds drop by 25% to 3 cm/d while also deviating 45 degrees further towards the west for five 276
days. During the next 5-day measurement period, speeds drop further, back to 2 cm/d while the 277
azimuths turn another 10-15 degrees further west. Over the final 28-day measurement period, 278
the azimuths remain stable, while speeds continued to fall. This decrease in speed is apparent 279
in the decline of the upper quartile of speeds, despite our not making sufficient new 280
measurements to influence the median speed. 281
Our fine-scale movement and ablation data allowed us to compare glacier moss ball 282
speeds and azimuths with potential drivers of their motion. The southern and western directions 283
of moss ball movement are clearly distinct from both the prevailing wind direction as inferred 284
from the dust plume (towards the southeast) or the downhill direction of the gently sloping ice 285
surface (towards the east-northeast; Fig. 3a). Instead, we find more rapid moss ball speeds are 286
associated with more rapid ablation; an ordinary least squares model between ablation rate and 287
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
11
speed indicates that, on average, for every 1 cm of surface ablation, the glacier moss balls 288
move horizontally 0.34 cm (Fig. 3b). However, the relationship between ablation rate and speed 289
is relatively weak (R2 = 0.40). It should also be noted that during the course of our study, 290
participants in a program hosted by the Wrangell Mountains Center, McCarthy, Alaska, visually 291
confirmed the posited primary movement method described by Porter et al. (2008), when a 292
glacier moss ball was observed rolling off its elevated pedestal and inverting in the process. 293
294
Table 1. Apparent survival models for glacier moss balls tested in this study with their 295 corresponding Akaike’s Information Criterion scores that have been adjusted for small sample 296 sizes (AICc). Relative AICc scores (ΔAICc) model weight are also given. Lower ΔAICc and 297 higher model weight indicate greater support for a given model. Model components: probability 298 of detection (p), apparent survival (φ). 299 300 Model Description AICc
Δ
AICc Weight
1 Null; No year effect on p or φ 107.09 1.56 0.26
2 Year effect on φ 105.53 0 0.58
3 Year effect on p 108.92 3.39 0.10
4 Year effect on both p and φ 110.25 4.72 0.05
301 302 Table 2. Estimates of the apparent survival (φ) and detection probability (p) of glacier moss balls 303 for the two best-fit models. Parentheses after estimates indicate 95% confidence intervals. 304 305 Model Parameter Estimate
1 p 0.84 (0.70-0.92)
φ 0.86 (0.75-0.93)
2 p 0.82 (0.69-0.91)
φ (2009-2010) 0.74 (0.55-0.87)
φ (2010-2011) 0.98 (0.27-0.99)
φ (2011-2012) 1.0 (0.99-1)
306
Longevity 307
We initially tagged 30 glacier moss balls in 2009. We subsequently recaptured 18 moss 308
balls each in 2010, 2011, and 2012 (although this was not the same 18 moss balls each year). 309
Recapture rates for individual glacier moss balls were highly variable with some never seen 310
again after the first year (n = 8) and others detected every year (n = 13). The best-fit survival 311
model included differing apparent survival ( ) among years, but with constant detection 312
probability (p; Model 2; Table 1). This model received 58% of AICc weight, compared to 26% for 313
the null model (Model 1), and less than 10% for the other models (Models 3 & 4; Table 1). The 314
average annual rate of apparent survival, , based on the null model, was 0.86 [95% 315
confidence interval (CI) = 0.75-0.93], and the average detection rate was 0.84 (95% CI = 0.70-316
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
12
0.92). When parameterized by year, the annual apparent survival rate ranged from 0.74 in 317
2009-2010 to 1.0 in 2011-2012 with a particularly large 95% CI for 2010-2011 (Table 2; Fig. 4).
318
319
Fig. 4. Estimates of apparent moss ball survival (
; dark circles) with 95% confidence intervals (thin dark
320
lines) from model 2, the best-fit model, which included a year effect on . Year-long, bracketed time
321
intervals labeled on the x-axis are identified by their starting year. For instance, apparent survival for
322
2009-2010 is shown as 2009.
323 324
Our detection rate estimates may underestimate actual glacier moss ball survival for 325
several reasons. First, at least four glacier moss balls lost their marking bracelet after the first 326
year because we found the marking bracelet on the ice, separate from a moss ball. Second, 327
another moss ball partially obscured its bracelet by growing to cover the beads, but we were 328
able to detect a single bead and then delicately “excavate” the bracelet. Since we did not 329
destructively search glacier moss balls that did not have an obvious bracelet, it is possible that 330
additional instances of lost marking bracelets or growth to cover beads may have impacted our 331
detection. Third, between 2009 and 2010, two tagged moss balls fell inside of a shallow 332
crevasse within the study area. The two crevasse-
bound glacier moss balls persisted, and likely
333
continued to photosynthesize and grow to some capacity for the remainder of the study. We 334
continued to c
heck crevasses in the study area carefully, but some moss balls could have fallen
335
into deeper crevasses, or into shallow crevasses in a way that obscured their markings, and 336
therefore persisted without detection. 337
Our estimate of average life expectancy var
ied depending on whether the lowest overall
338
or mean annual survival rate were used. If using the lowest annual survival rate (0.74), average
339
life expectancy was 3.3 years (95% CI = 1.67-7.18). However, we expect this life expectancy to 340
be biased low to some extent, because we were only able to estimate apparent survival (e.g., 341
some insecure tags fell off moss balls that likely still persisted). If using the mean annual 342
k
ly
n
ll
e
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
13
apparent survival rate across the entire study (0.86), average life expectancy rose to 6.63 years 343
(95% CI = 3.48-13.78), although this may be biased high because we did not tag any new moss 344
balls in years 2 and 3 (2010 and 2011), but simply re-captured existing (and therefore high 345
survival probability) glacier moss balls. When thinking of lifespan, it is relevant to note that we 346
also observed a glacier moss ball split roughly in half during the course of the study along its 347
intermediate axis. The moss ball had become elongated and essentially pulled apart. This 348
mechanism may contribute to both keeping glacier moss balls ovular as well as a form of 349
cloning, propagating new moss balls on the landscape. 350
351
Discussion: 352
Glacier moss balls are intriguing components of glacier ecosystems that integrate 353
physical (e.g., debris cover) and ecological (e.g., invertebrate colonization) factors into a unique 354
habitat type. Glacier moss balls have a global distribution, with colonies identified in Iceland, 355
North and South America, and Asia (Eythórsson 1951; Heusser 1972; Perez 1991; Porter et al. 356
2008; Coulson and Midgley 2012) and more dispersed moss aggregations on glaciers have 357
been described from an even broader area (e.g., Uganda, Uetake et al. 2014). Previous studies 358
have revealed a great deal about glacier moss ball biology (e.g., their invertebrate colonizers, 359
Coulson and Midgley 2012) yet their movement and longevity has remained unknown. It has 360
been speculated that glacier moss ball movement patterns likely follow the general downward 361
slope of the glacier (Porter et al. 2008) and that they represent an ephemeral habitat type on 362
glaciers, a factor that may limit colonization by specific invertebrate taxa (e.g., spiders; Coulson 363
and Midgley 2012). Our results do not align with these predictions of movement and longevity. 364
Rather, we show that glacier moss balls, at least on a relatively gently sloped Alaskan glacier, 365
exhibit relatively quick (2.5 cm/d) herd-like movements that do not align with the downward 366
slope of the glacier nor the dominant wind direction. Glacier moss balls are also relatively long-367
lived with a mean lifespan of more than 6 years. 368
369
Movement 370
On the Root Glacier, glacier moss balls move relatively quickly (~2.5 cm/d) in similar 371
general directions and at similar general speeds. Directions of motion do not align solely with 372
either the downhill direction nor the direction of the prevailing wind. The rate of glacier moss ball 373
movements is also positively correlated, albeit weakly, with overall glacier ablation (Fig. 3b). It 374
appears likely that the dominant direction of solar radiation, which melts exposed ice 375
surrounding glacier moss balls more rapidly than the insulated ice below them (Porter et al. 376
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
14
2008), is the major force driving glacier moss ball movement. However, the relative 377
contributions of gravity in the downslope direction versus solar radiation is almost certainly 378
dependent on glacier steepness. Porter et al. (2008) posited a considerable effect of gravity on 379
glacier moss ball movement for a relatively steep (9.6°) Icelandic glacier which contrasts with 380
our much flatter study area on the Root Glacier (~3°). Still, regardless of steepness, differential 381
melt patterns create pedestals that glacier moss balls rest upon and, eventually, enough ice 382
melts below the moss ball causing it to fall and potentially flip (Porter et al. 2008). Assuming 383
glacier moss balls are, on average, ~10 cm in their intermediate axis, and their only means of 384
movement is melt-induced flipping driven by pedestal emergence at the rate of 6-9 cm/d, their 385
rates of movement would imply each glacier moss ball flips every ~2-4 days. However, we 386
cannot rule out alternative modes of glacier moss ball movement. Many glacier moss balls have 387
one side that is flattened and commonly faces down, while the other, more rounded and 388
vegetated side faces skyward (Shacklette 1966). Given this orientation, an alternative scenario 389
is that glacier moss balls also move by basal sliding over the wet glacier surface below. 390
One movement-related question remains puzzling: why do the azimuths of glacier moss 391
balls appear to shift simultaneously throughout the summer season, resulting in the moss ball 392
“herd” synchronously changing directions (Fig. 3a)? They begin the season moving generally 393
south and slowly transition towards the west. Given their independence from the dominant wind 394
direction and downhill direction of the glacier, we can speculate that a slow shift in the dominant 395
direction of solar radiation or shifting weather patterns during summer drives this pattern. 396
Perhaps the weather transitioned from clear mid-day skies during late June and early July 397
(associated with the most rapid motion and southerly azimuths), to a different weather pattern in 398
late July consisting of morning clouds and afternoon sun, that drives enhanced ablation on the 399
west sides of moss balls, and therefore preferential rolling towards the west. Furthermore, the 400
interaction between rate of movement and ablation may depend additionally on the degree to 401
which the dominant solar radiation and downslope directions align. If they are in the same 402
direction, glacier moss balls should move rapidly, downslope, towards the sun. However, if melt 403
is being driven from a direction perpendicular to the downslope direction, then glacier moss 404
balls will move considerably less per unit melt. 405
406
Longevity 407
Glacier moss balls show considerable potential to persist across multiple years as stable 408
ecological units. On average, 86% of the marked glacier moss balls included in this study 409
survived annually which translates to a lifespan of more than 6 years. This longevity is on par 410
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
15
with large, long-lived vertebrates (e.g., adult female ungulates in typically have annual survival 411
probabilities 0.85; Gaillard and Yoccoz 2003; Gaillard et al. 1998). Thus, with high rates of 412
survival across multiple years, and relatively high detection rates, we consider glacier moss 413
balls to be long-lived, rather than ephemeral, glacier features. Unlike living individual organisms 414
which can show senescence as they age (e.g., Loison et al. 1999), moss ball survival rates are 415
unlikely to decline with time in the traditional sense, nor are they likely to exhibit density 416
dependence in survival (e.g., Festa Bianchet et al. 2003), however the factors that control 417
disaggregation may be the most important factor for moss ball longevity. At any rate, the 418
temporal stability of individual moss balls on the glacier surface means that they likely exist for 419
long enough to develop complex biotic communities, a fact supported by the distinctive 420
invertebrate communities inhabiting moss balls (Coulson and Midgley 2012). However, the 421
degree to which geographic location (e.g., distance to a glacier margin), and not persistence, 422
influences invertebrate colonization remains to be tested. 423
The limited scope of our mark-recapture data collection precludes us from drawing 424
conclusions about the inter-annual drivers of moss ball apparent survival. However, we can 425
highlight potential factors that may influence it. First, it is possible that glacier moss balls moved 426
more frequently out of the study area in one year versus others, perhaps due to exceptionally 427
clear skies (and thus higher rates of glacier ablation). Second, we observed a number of 428
fragmented moss balls. This fragmentation may be part of normal glacier moss ball growth 429
trajectories, too frequent or intense freeze thaw cycles, perhaps as a product of moss ball water 430
content, or some other as yet unknown factor. It is also unclear at what rate fragmented glacier 431
moss balls continue to move and grow, eventually developing back into mature, ovoid, full-sized 432
moss balls. If glacier moss balls did survive within our study area, they had an 84% probability 433
of being detected in a given year. This indicates that our bracelet and colored beads marking 434
scheme was relatively successful. However, for future studies, more robust marks should be 435
considered. One possibility is the use of passive integrated transponder (PIT) tags which are 436
commonly used for mark-recapture studies of a variety of organisms (e.g., fish; Castro-Santos 437
et al. 1996), and allow researchers to scan study organisms electronically rather than rely on 438
visual ID. 439
440
Genesis, growth, and disaggregation 441
Our combined movement and longevity analyses allow us to add new speculation about 442
patterns of glacier moss ball growth as well as additional evidence for previous hypotheses 443
regarding their genesis and disaggregation (e.g., Heusser 1972; Perez 1991). In terms of 444
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
16
growth, our documentation of glacier moss balls rolling over a fine-grained, wet, sedimentary 445
substrate is consistent with growth through the adherence of fine-grained sediment to an 446
existing moss ball. When a moss ball rolls from its elevated pedestal, sediment grains stick to 447
the moss (or potentially to the cohesive sediment itself). We visually observed such “dirty” moss 448
on some glacier moss balls in our study area. As the moss itself grows, this adhered sediment 449
may then become integrated within the fibrous material, increasing the size of the glacier moss 450
ball. Field observation of moss growth over and around our identification bracelets indicates that 451
several millimeters of growth can occur within years. However, the observation that most 452
bracelets were not engulfed by sediment accumulation and moss growth during our four-year 453
study period suggests an upper limit on moss ball growth. Consistent with a greater than 6-year 454
lifespan, moss ball growth to an observed size of 10 cm must take years, and given the 455
conservative nature of our longevity estimates, potentially much longer. 456
Understanding year-to-year moss ball growth, however, does not explain moss ball 457
genesis, nor disaggregation. It is well-established that fibrous moss provides the skeletal 458
structure that allows moss balls to be cohesive, ovoid structures. A source of moss spores is 459
there essential for initial glacier moss ball genesis (in our study, putatively, the Donoho 460
nunatak). The question, then, is how glacier moss balls begin to grow in the first place, and on 461
what substrate. (Eythórsson 1951) suggested that a “stone kernel” is likely at their centers. 462
However, later investigations (e.g., Shacklette 1966; Gremmen 1982) found mixed results that 463
largely reflected a consensus that there is no general rule about rock cores at the center of 464
glacier moss balls. Our exploratory testing of moss balls also indicated that some, but not all, 465
moss balls contained a ~1-cm gravel “kernel” at their centers. Potentially, these kernels, with 466
adhered fine-grained sediment, provide a growth substrate for initially wind-deposited moss 467
spores. In our study area, the co-occurrence of moss balls with an unusually extensive, fine-468
grained “plume” of sediment cover (Fig. 1b) aligns with a similar observation by Heusser (1972) 469
for the Gilkey Glacier in southeastern Alaska, USA. The fine-grained sediment may be essential 470
proto-soil for moss ball growth and may explain the unusual density of moss ball occurrence at 471
our study area. The origin of this fine-grained sediment is unknown, but in satellite imagery (Fig. 472
1b), it appears to originate from the ice itself. It may be a volcanic ash layer emerging from the 473
ice after being carried down from the high, volcanic, Wrangell Mountain peaks. Once moss 474
growth initiates, the moss itself, and the abundant moisture of the ablating glacier surface, may 475
provide the necessary cohesion for the incipient moss ball formation, and continued growth 476
thereafter. 477
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
17
In this study, we identified very few glacier moss balls greater than ~15 cm on their long 478
axis. Generally, moss balls appear to rarely exceed ~10 cm except for rare cases in Alaska 479
where they have been reported up to 18 cm (Heusser 1972; Benninghoff 1955). Why glacier 480
moss balls in Alaska appear to grow larger than elsewhere in the world remains an open 481
question but, regardless of location, there appears to be some size limiting process within the 482
moss ball lifecycle at work. Shacklette (1966) suggested that the tensile strength of moss stems 483
may be the key factor controlling their size. Exceeding this tensile limit appears to occur when 484
the moss ball major axis grows too great relative to the intermediate axis. For instance, when a 485
moss ball becomes too elongated, subtle variations in ice surface topography may lead the two 486
ends of a moss ball to move in different directions, leading to a tear in the middle when the 487
moss ball’s tensile strength is exceeded. We observed such a tearing-in-two of a long, linear 488
moss ball during the course of our study. Again, while this process applies an upper-limit to 489
moss ball size it also circles back to inform questions regarding the presence or absence of a 490
rock kernel. If the upper size limit is reached and a moss ball splits, only one of the two 491
remaining moss balls involved in this “cloning” process will retain the gravel kernel. This may 492
explain why a number of moss balls do not appear to have any coarse-grained rock at their 493
cores. 494
495
Conclusions 496
Since their first description nearly 75 years ago (Eythórsson 1951), a general 497
understanding of the physical and biological composition of glacier moss balls and their 498
ecological role as drivers of small-scale ecosystem development has been established. In this 499
study, we extended this previous work to quantify the movement and longevity of glacier moss 500
balls on an Alaskan glacier. In light of these results, we discussed a potential life cycle for moss 501
balls. We showed that glacier moss balls move relatively quickly, at a rate of centimeters per 502
day, and that moss balls within the surveyed colony speed up, slow down, and change direction 503
in synchrony. This finding suggests that moss ball motion is controlled by some broadly applied 504
external forcing. However, for our study, this external forcing was surprisingly not solely 505
associated with either the wind or downslope direction. Instead, the glacier moss ball movement 506
patterns we observed likely depend on a combination of the intensity of glacier ice ablation, the 507
direction of solar radiation, and the physical surface of the glacier (i.e., the downslope direction). 508
Future studies that take a similar mark-recapture approach to the study of glacier moss ball 509
movements on multiple glaciers throughout the world, including a range of steepness, will shed 510
important light on the general nature of their movements. 511
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
18
We also showed that glacier moss balls are long-lived, with an average life expectancy 512
of more than 6 years, on par with many relatively long-lived vertebrates (e.g., large mammalian 513
herbivores, Gaillard and Yoccoz 2003). This potential for glacier moss balls to act as relatively 514
stable, long-term ecological units further confirms their potential to act as key habitat for 515
invertebrates. Coulson and Midgley (2012) previously described invertebrate colonization of 516
glacier moss balls and suggested that a lack of Enchytraeidae and Aranea may be the result of 517
the ephemeral nature of moss balls in glacier habitats. Our results contrast with this assumption. 518
Instead, we postulate that selective invertebrate colonization of glacier moss balls may depend 519
instead on their geographic locations and their frequent movements or, as Coulson and Midgley 520
(2012) noted, may simply reflect the variable dispersal capacities of potential colonizers. 521
Given the importance of microbial diversity to carbon cycling (Anesio et al. 2009), 522
ecosystem function (Anesio et al. 2017; Hotaling et al. 2017a; Hotaling et al. 2017b), and even 523
albedo (Ganey et al. 2017), future efforts to understand the microbial ecology of glacier moss 524
balls could shed important light on their ecological role in glacier ecosystems. Like cryoconite, 525
the granular, darkly pigmented dust that accumulates on the surface of glaciers and drives 526
hotspots of biological activity and microbial diversity (Cook et al. 2016), glacier moss balls may 527
have similar value at the ecosystem scale. Indeed, glacier moss balls may act as reservoirs of 528
microbial diversity, seeding and re-seeding a glacier surface with key, colonizing microbial life 529
as they move around its surface. 530
531
Acknowledgements: 532
We thank the Wrangell Mountains Center for logistical support and assisting with field 533
measurements, and Dr. Billy Armstrong for providing the orthoimage of the study area. 534
535
References: 536
Akaike H (1998) Information theory and an extension of the maximum likelihood principle. In: 537 Selected papers of Hirotugu Akaike. Springer, pp 199-213 538 Anesio AM, Hodson AJ, Fritz A, Psenner R, Sattler B (2009) High microbial activity on glaciers: 539 importance to the global carbon cycle. Global change biology 15:955-960 540 Anesio AM, Laybourn-Parry J (2012) Glaciers and ice sheets as a biome. Trends Ecol Evol 541 27:219-225 542 Anesio AM, Lutz S, Chrismas NA, Benning LG (2017) The microbiome of glaciers and ice 543 sheets. NPJ biofilms and microbiomes 3:10 544 Belkina OA, Vilnet AA (2015) Some aspects of the moss population development on the 545 Svalbard glaciers. Czech Polar Reports 5:160-175 546 Benninghoff WS (1955) Jökla mýs. Journal of Glaciology 2:514-515 547 Castro-Santos T, Haro A, Walk S (1996) A passive integrated transponder (PIT) tag system for 548 monitoring fishways. Fisheries Research 28:253-261 549
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
19
Cook J, Edwards A, Takeuchi N, Irvine-Fynn T (2016) Cryoconite: the dark biological secret of 550 the cryosphere. Progress in Physical Geography 40:66-111 551 Coulson S, Midgley N (2012) The role of glacier mice in the invertebrate colonisation of glacial 552 surfaces: the moss balls of the Falljökull, Iceland. Polar Biology 35:1651-1658 553 Deevey Jr ES (1947) Life tables for natural populations of animals. The Quarterly Review of 554 Biology 22:283-314 555 Dial RJ, Becker M, Hope AG, Dial CR, Thomas J, Slobodenko KA, Golden TS, Shain DH (2016) 556 The role of temperature in the distribution of the glacier ice worm, Mesenchytraeus 557 solifugus (Annelida: Oligochaeta: Enchytraeidae). Arctic, Antarctic, and Alpine Research 558 48:199-211 559 Eythórsson J (1951) Correspondence. Jökla-mys. Journal of Glaciology 1:503 560 Festa Bianchet M, Gaillard JM, Côté SD (2003) Variable age structure and apparent density 561 dependence in survival of adult ungulates. Journal of Animal Ecology 72:640-649 562 Gaillard J-M, Festa-Bianchet M, Yoccoz NG (1998) Population dynamics of large herbivores: 563 variable recruitment with constant adult survival. Trends in Ecology & Evolution 13:58-63 564 Gaillard J-M, Yoccoz NG (2003) Temporal variation in survival of mammals: a case of 565 environmental canalization? Ecology 84:3294-3306 566 Ganey GQ, Loso MG, Burgess AB, Dial RJ (2017) The role of microbes in snowmelt and 567 radiative forcing on an Alaskan icefield. Nature Geoscience 10:754 568 Gardner AS, Moholdt G, Cogley JG, Wouters B, Arendt AA, Wahr J, Berthier E, Hock R, Pfeffer 569 WT, Kaser G (2013) A reconciled estimate of glacier contributions to sea level rise: 2003 570 to 2009. Science 340:852-857 571 Gremmen N (1982) The vegetation of Subantarctic Islands Marion and Prince Edward. The 572 Hague, Boston, London, W. Junk Publishers, 573 Heusser CJ (1972) Polsters of the moss Drepanocladus berggrenii on Gilkey Glacier, Alaska. 574 Bulletin of the Torrey Botanical Club:34-36 575 Hotaling S, Finn DS, Joseph Giersch J, Weisrock DW, Jacobsen D (2017a) Climate change and 576 alpine stream biology: progress, challenges, and opportunities for the future. Biological 577 Reviews 92:2024-2045 578 Hotaling S, Hood E, Hamilton TL (2017b) Microbial ecology of mountain glacier ecosystems: 579 biodiversity, ecological connections and implications of a warming climate. 580 Environmental microbiology 19:2935-2948 581 Hotaling S, Shain DH, Lang SA, Bagley RK, Lusha M, Weisrock DW, Kelley JL (2019) Long-582 distance dispersal, ice sheet dynamics, and mountaintop isolation underlie the genetic 583 structure of glacier ice worms. Proceedings of the Royal Society B: Biological Sciences 584 286:20190983 585 Hurvich CM, Tsai C-L (1989) Regression and time series model selection in small samples. 586 Biometrika 76:297-307 587 Larsen C, Burgess E, Arendt A, O'neel S, Johnson A, Kienholz C (2015) Surface melt 588 dominates Alaska glacier mass balance. Geophysical Research Letters 42:5902-5908 589 Lebreton J-D, Burnham KP, Clobert J, Anderson DR (1992) Modeling survival and testing 590 biological hypotheses using marked animals: a unified approach with case studies. 591 Ecological Monographs 62:67-118 592 Loison A, Festa-Bianchet M, Gaillard J-M, Jorgenson JT, Jullien J-M (1999) Age specific 593 survival in five populations of ungulates: evidence of senescence. Ecology 80:2539-2554 594 Mann D, Edwards J, Gara R (1980) Diel activity patterns in snowfield foraging invertebrates on 595 Mount Rainier, Washington. Arctic and Alpine Research 12:359-368 596 Millar JS, Zammuto RM (1983) Life histories of mammals: an analysis of life tables. Ecology 597 64:631-635 598
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
20
Perez FL (1991) Ecology and morphology of globular mosses of Grimmia longirostris in the 599 Paramo de Piedras Blancas, Venezuelan Andes. Arctic and Alpine Research 23:133-600 148 601 Porter P, Evans A, Hodson A, Lowe A, Crabtree M (2008) Sediment–moss interactions on a 602 temperate glacier: Falljökull, Iceland. Annals of Glaciology 48:25-31 603 Roe GH, Baker MB, Herla F (2017) Centennial glacier retreat as categorical evidence of 604 regional climate change. Nature Geoscience 10:95 605 Rosvold J (2016) Perennial ice and snow-covered land as important ecosystems for birds and 606 mammals. Journal of biogeography 43:3-12 607 Shacklette HT (1966) Unattached moss polsters on Amchitka Island, Alaska. Bryologist:346-352 608 Uetake J, Tanaka S, Hara K, Tanabe Y, Samyn D, Motoyama H, Imura S, Kohshima S (2014) 609 Novel biogenic aggregation of moss gemmae on a disappearing African glacier. PloS 610 one 9:e112510 611 Van der Walt S, Schönberger JL, Nunez-Iglesias J, Boulogne F, Warner JD, Yager N, Gouillart 612 E, Yu T (2014) scikit-image: image processing in Python. PeerJ 2:e453 613 614
.CC-BY 4.0 International licenseIt is made available under a
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which. http://dx.doi.org/10.1101/687665doi: bioRxiv preprint first posted online Jul. 2, 2019;
