Plant Ecology

, Volume 214, Issue 11, pp 1309–1319

Spatial contexts for temporal variability in alpine vegetation under ongoing climate change

Authors

    • Department of GeographyUniversity of Iowa
  • Daniel B. Fagre
    • U.S. Geological Survey Northern Rocky Mountain Science Center
Article

DOI: 10.1007/s11258-013-0253-3

Cite this article as:
Malanson, G.P. & Fagre, D.B. Plant Ecol (2013) 214: 1309. doi:10.1007/s11258-013-0253-3

Abstract

A framework to monitor mountain summit vegetation (The Global Observation Research Initiative in Alpine Environments, GLORIA) was initiated in 1997. GLORIA results should be taken within a regional context of the spatial variability of alpine tundra. Changes observed at GLORIA sites in Glacier National Park, Montana, USA are quantified within the context of the range of variability observed in alpine tundra across much of western North America. Dissimilarity is calculated and used in nonmetric multidimensional scaling for repeated measures of vascular species cover at 14 GLORIA sites with 525 nearby sites and with 436 sites in western North America. The lengths of the trajectories of the GLORIA sites in ordination space are compared to the dimensions of the space created by the larger datasets. The absolute amount of change on the GLORIA summits over 5 years is high, but the degree of change is small relative to the geographical context. The GLORIA sites are on the margin of the ordination volumes with the large datasets. The GLORIA summit vegetation appears to be specialized, arguing for the intrinsic value of early observed change in limited niche space.

Keywords

Alpine tundraOrdinationClimate changePlant communitySimilarityWestern USA

Introduction

Alpine tundra may be threatened by global warming. It is associated with cold temperatures and short growing seasons, and it is spatially isolated and surrounded by forest to the point where some have envisioned that it could be pushed off the tops of mountains into nonexistence by rising treelines (e.g., Dirnbock et al. 2003; cf. Lenoir et al. 2008). Climate change projections indicate areal losses of the type of climate associated with alpine tundra (Diaz and Eischeid 2007; Ackerly et al. 2010), fueling this concern. Studies of changes in vegetation and specific responses (e.g., phenology) support this concern (e.g., Huelber et al. 2006; Lesica and McCune 2004; Inouye 2008). These potential changes threaten ecosystem services, such as water resources, biodiversity, and aesthetic and recreational opportunities (Malanson et al. 2007; Gret-Regamey et al. 2008). Anticipating these changes (and superseding earlier efforts; e.g., Lesica and Steele 1996), in 1997 Grabherr et al. (2000) initiated an international framework to monitor vegetation change in alpine tundra on mountain summits (The Global Observation Research Initiative in Alpine Environments, GLORIA). This framework has now expanded to 104 active target regions, and the network is beginning to bear fruit. Erschbamer et al. (2011) reported upward migration of species in the Italian Dolomites, including an advance of the forest boundary at the lowest sites. Gottfried et al. (2012) reported an increase in plant species adapted to warmer climates in GLORIA sites across all of Europe, and Pauli et al. (2012) also found upward migration with increasing species richness for this area, except in Mediterranean regions also known to be high in endemic species.

The meaning of these initial (2001–2008) GLORIA results is complicated, however. As Pauli et al. (2007, 2012) showed, the trends are not uniform, and Gottfried et al. (2012) found statistically significant change only when all European ranges were pooled. Studies across alpine tundra are not definitive in change (e.g., Sandvik et al. 2004; Walther et al. 2005; Bjork and Molau 2007; Randin et al. 2009a; Wipf et al. 2009; Kudo et al. 2010; Evju et al. 2012) or illustrate significance for non-climatic factors (e.g., Callaway et al. 2002; Körner 2002; Boyce et al. 2005; Kikvidze et al. 2005; Bruun et al. 2006; Vonlanthen et al. 2006; Litaor et al. 2008; Loffler and Pape 2008; Randin et al. 2009b; Rose and Malanson 2012). Danby et al. (2011) found significant changes in alpine tundra in the Yukon, Canada by repeating a study done 40 years earlier; while the general change was consistent, some variation among aspects was found. Elmendorf et al. (2012a, b) found regional variations in comparisons of experimental warming in alpine and arctic tundra. Scherrer and Körner (2011) have illustrated that considerably more variation than is likely to exist within GLORIA summit plots is maintained within similarly small areas on heterogeneous alpine slopes. Temperature is only one of many variables affecting this vegetation (e.g. nitrogen; Smith et al. 2012), and sites very different from summits may be more likely to change (Malanson et al. 2012). Moreover, Kammer et al. (2007) found that species on mountain summits may now be changing as a disequilibrium response to earlier climate change.

The GLORIA framework is well described at http://www.gloria.ac.at/?a=5; its initiation in North America was in 2003 (Millar and Fagre 2007). To summarize, a GLORIA target region consists of an area with four summits, one each within the nival/subnival, subnival/upper alpine, upper alpine/lower alpine, and lower alpine/subalpine ecotones. Each summit is divided into four quadrants facing the cardinal directions. Within each quadrant, a quadrat cluster with four 1 × 1 m permanent plots at the corners of a 3 × 3 m quadrat is the basis for sampling. Within the 1 m2 quadrats, the percent cover of each vascular plant species is estimated visually. The intent is to resample these areas at regular, usually 5 year, intervals in order to document the possible effects of climate change. The interpretation of results of resampling GLORIA target regions needs development for two reasons, one practical and one theoretical.

For the former, land managers and others need to know the direction and significance of any observed change, how this can be communicated to stakeholders and the public, and, if desired, mitigated. Having broader context for these summit trajectories will improve interpretation and suggest outcomes and possible actions. Because the range of variation in alpine tundra is far beyond what is captured on GLORIA summits and the direction of change that summits may experience is unknown, establishing context for GLORIA results will improve their utility. Malanson et al. (2011) examined the differences among hundreds of vegetation surveys concentrated in Glacier National Park, Montana and the Indian Peaks of Colorado, USA, without observing temporal change, relative to climatic and geographic differences; Hong and Ayako (2012) examined 17 sites with better geographic spread; but while neither had clear cut differentiation of the effects of climate versus geography, both provide the type of context needed.

For the theoretical concern, we can be guided by the theory of island biogeography (TIB; MacArthur and Wilson 1967). While Nagy and Grabherr (2009) have pointed out its limitations for mountain summits, a core point raised by the theory is not island specific: species turnover happens (just slower with increasing isolation and smaller areas). TIB leads us to expect that species in GLORIA plots will change in the absence of climate change. This hypothetical background rate of change (hereafter, background turnover) will depend, in TIB terms, on the isolation of the summit from the “mainland” pool of species; this isolation will depend on how precisely one defines differences in alpine tundra. Higher precision would define the greatest isolation, where the species pool is restricted to only those species that are adapted to similar conditions (for GLORIA, summit conditions), a lower theoretical background turnover, and an easier threshold for finding change above the background rate. Less precision would lump all alpine tundra as the pool, isolation would be minimal, and the background turnover rate would be highest.

We are not in a position to define and map species pools and examine turnover rates, but we can provide practical information, and indicate how precisely to classify alpine tundra, by examining change in GLORIA sites relative to the species pools we observe at multiple scales. The comparison of experiments by Elmendorf et al. (2012b) indicated the regional variation seen in responses to warming requires a broad context. Stoeckli et al. (2011) made this point in relation to historical plant survey data for 300 mountain summits in Europe.

We report changes observed at GLORIA sites in Glacier National Park, Montana, USA (GNP) relative to the range of variability in species composition that might be found in alpine tundra across the western USA and southwestern Canada. We examine dissimilarity among pairs of sites based on their species composition and then examine ordinations that are based on these dissimilarity measures to provide a quantitative description of the context within which observed change can be interpreted and to indicate the degree to which mountain top GLORIA vegetation is distinct from any potential pools of species. For the repeated samples of the GLORIA sites the trajectories of change can then be visualized as well as quantified within the statistical space (e.g., Malanson and Trabaud 1987; Robbins and Matthews 2010). The 525 plots in GNP that we use for landscape context were sampled in the 1990s, and the 418 samples for regional context were measured at different times with different methods. Most of the samples were taken as a visual estimate of percent cover in quadrats ranging from 1 to 16 m2, but some were sampled on line intercept transects. We contend that we can use them for spatial context if we assume that each represents a type of alpine tundra that could exist somewhere in this region. Our rationale is that the ordination space created by these sites, when taken together, represents a background range of variability that exists in alpine tundra; that the sites may differ now does not change this background status. As noted by Ross et al. (2010), the spatial variability of sites is a useful context within which to examine repeated samples (although matching locations accurately is not a problem in GLORIA). We only directly compare pairs of GLORIA samples in the foreground of the ordination space; therefore, sampling >400 plots across the region at close to the same time as the GLORIA sampling is unnecessary as well as impractical (cf. Stoeckli et al. 2011).

Methods

Data sources

We took descriptive and quantitative data for alpine tundra from a variety of studies, and we consider these at three scales.

GLORIA

Holzer and Fagre (2004) described the GLORIA North American site in GNP. The locations are on the summits of Bison Mtn., Dancing Lady Mtn., Pitamakan Peak, and Seward Mtn. (Table 1). The Bison and Dancing Lady sites were sampled in 2003 and the Pitamakan and Seward sites in 2004. All were resampled in 2009. Because of empty quadrats we have 14 matched pairs.
Table 1

Locations of GLORIA sites in Glacier National Park, MT

Summit

Latitude

Longitude

Elevation (m)

Dancing Lady Mtn.

48.425

113.312

2245

Bison Mtn.

48.465

113.311

2387

Pitamakan peak

48.520

113.445

2493

Seward Mtn

48.871

113.680

2717

Landscape

Damm (2001) did the most extensive classification and description of alpine vegetation in GNP. His study recorded cover in over 700 Braun-Blanquet (1932) relevés in all types of alpine vegetation across the whole of GNP. For other purposes Malanson et al. (2011, 2012) reduced this dataset to 525 relevés, which we use.

Region

Our data come from a variety of studies. The general location, approximate geographic coordinates, and sources are shown in Table 2. In some cases we used individual site data and in others only data that summarized several sites were available. We used 418 cases.
Table 2

Sources of background data used for the regional analyses

Mountain(s)

State/province

Coordinates

 

# Plots

Source

Medicine Bow

Wyoming

41.3539

−106.311

2

Billings and Bliss (1959)

Medicine Bow

Wyoming

41.3539

106.311

6

Billings (1988)

Beartooth

Wyoming

45.0234

109.402

3

Johnson and Billings (1962)

Sierra

California

37.23

−118.624

1

Chabot and Billings (1972)

Flint Creek

Montana

46.0876

−113.09

6

Bamberg and Major (1968)

Big Snowy

Montana

46.0876

−113.09

6

Bamberg and Major (1968)

Sierra

California

38.699

−119.994

6

Taylor (1976)

Ruby

Nevada

40.613

−115.384

4

Loope (1970)

Sierra

California

37.5855

−118.874

5

Pemble (1970)

RockyMtnNP

Colorado

40.4132

−105.733

16

Willard (1979)

Jefferson

Oregon

44.7417

−121.795

6

Ingersoll (1991)

GlacierNP

Montana

48.7162

−113.629

36

Damm (2001)

Cascades

Washington

47.5049

−120.824

19

del Moral (1979)

White

California

37.63527

−118.253

3

Mooney (1973)

N Rockies

Idaho

44.0073

−113.6

74

Moseley (1985, 1993)

Indian Peaks

Colorado

40.1768

−105.622

56

Komarkova (1979)

Front Range

Colorado

39.262

−106.14

319

Stanton et al. (1994),Cooper and Sanderson (1997), Phillips (1982)

Lemhi

Idaho

44.5208

−113.509

66

Urbanczyk (1993)

NCascades

Washington

48.732

−121.488

1

Douglas and Ballard (1971)

Wheeler Pk

New Mexico

36.558

−105.414

10

Baker (1983)

Eagle Cap

Oregon

45.1619

−117.301

34

Johnson (2004)

Cerro Potosi

Mexico

24.86

−100.231

2

Beaman and Andresen (1966)

Big Snowy

British Columbia

49.738

−118.932

14

Eady (1971)

Banff NP

Alberta

51.318

−116.128

2

Beder (1967)

Plateau Mtn

Alberta

50.221

−114.524

4

Bryant (1968)

Highwood Pass

Alberta

50.595

−114.984

12

Trottier (1972)

Prospect Mtn

Alberta

52.222

−117.058

3

Mortimer (1978)

Jasper NP

Alberta

52.706

−117.691

3

Kuchar (1975)

Jasper NP

Alberta

52.792

−118.108

8

Hamilton (1981)

Uinta

Utah

40.744

−110.687

1

Ostler et al. (1982)

Uinta

Utah

40.779

−110.483

18

St Clair (1984)

Mesa Seco

Colorado

38.035

−107.243

11

Johnson (1970)

We organized the data for all three levels. First, we reconciled the taxonomy using the online USGS National Biological Information Infrastructure (www.nbii.gov; this no longer exists: http://www.usgs.gov/core_science_systems/Access/p1111-1.html) and Flora North America (floranorthamerica.org/). Second, we deleted lichen and mosses and kept only vascular plants for analysis because many studies only recorded the latter, which are also the standard for GLORIA sites. Third, we deleted sites to reduce over-weighting by some studies. We reduced the data from Komarkova (1979) and Damm (2001) from over 500 relevés each by retaining only the first listed from each of their classified associations. We reduced the data from Moseley (1985, 1993) and Urbanczyk (1993) by removing the second listed of any pair with ≤0.15 dissimilarity (see more below). Where the data were in the form of Braun-Blanquet cover classes we transformed the class number to the midpoint of its range of percent cover. After initial analysis we deleted outliers (observations farther than in 2 U in NMDS ordination space from any other); given our approach this is a conservative choice in that these are sites of small area (~1 m2) with few species [sampled by Komarkova (1979) or Damm (2001) and chosen because of their rarity].

Analyses

We examine dissimilarity measures primarily to demonstrate the need for ordination. Dissimilarity measures are limited in that pairs of sites that share no species have the same degree of dissimilarity. In most instances, this would be 1 or 100 %, and all degrees of dissimilarity are within this range. Ordination allows the dissimilarity among all pairs of sites to be used so that two sites with no shared species that have a species in common with a third site are closer than a fourth with no species in common with the others. We used Sorensen’s index of dissimilarity:
$${\text{D}} = 1 - {\text{(shared}}\;{\text{abundance}}/{\text{total}}\;{\text{abundance)}}$$
for a pair of sites (Czekanowski 1909; Sorensen 1948; Bray and Curtis 1957), “which has repeatedly been shown to be one of the most effective measures…” (McCune and Grace 2002, 54). We compute this measure for all sites in the data at the three scales and for each GLORIA site for their 2003/4 and 2009 observations. We examine the difference between the GLORIA pairs and the average dissimilarity across all observations.

We ordinated the site × species data using nonmetric multidimensional scaling in PC-ORD (NMDS; McCune and Mefford 2011). NMDS provides a mapping of sites in statistical space based on the similarity of their plant community composition and thus is quantitatively related to similarity per se. NMDS is a preferred method for many types of exploratory ordination, but it is especially well suited to data wherein the values for the species are imprecise, and to cases where many sites share no species, because it uses the rank order of distances (McCune and Grace 2002). It makes minimal assumptions about the form and relations of the data. This approach minimizes the problem of the differences in precision among our data sources, and Otypkova and Chytry (2006) showed that some ordination results were not sensitive to sample size. Danby et al. (2011) used NMDS to elucidate differences in community structure among alpine tundra sites over a 40 year interval.

We initialized NMDS in PC–ORD using the criteria in Table 3, but in the end we used three ordination axes for comparison even when the algorithm recommended fewer (this only means that the improvement in stress, a measure of the difference in ordination space relative to the dissimilarity measure for all possible pairs, was small with additional axes).
Table 3

Initial conditions set for the nonmetric multidimensional scaling

Maximum number of iterations

300

Number of runs, real data

250

Number of runs, randomized data

250

Instability criterion

0.00001

Starting number of axes

6

To analyze the change in GLORIA sites relative to other sites in ordination space we compare two measures in ordination space. Ross et al. (2010) recommended assessing repeat samples in comparison to spatial variation using by correlating similarity with distance for assessing uncertainty in locating revisited plots; here, we compare similarity directly and in ordination space. First, we examine the distance between the GLORIA pairs relative to the average distance among all pairs and the longest ordination axis. Second, for each GLORIA pair we compute the volume of a sphere for which the distance in ordination space would be the diameter, and we compare this to the volume of an ellipsoid with the lengths of the three ordination axes taken as the semi-axes. The comparison of volumes captures the imprecision of representing the possible trajectories of a site through time as a sphere rather than a line relative to the full range of possibilities for alpine tundra (that could exist) represented by the total volume. We also examined the direction in which the GLORIA sites moved and their overall location in the ordination space relative to the others examined.

Results

The average dissimilarity of the 84,666 possible pairs across the entire region is 0.966. But this is hardly more than the average dissimilarity within the 137,550 possible pairs from landscape level sites in GNP (0.915) or even among the 91 possible pairs from the GLORIA summit sites alone (0.855). The 14 2003, 2004 to 2009 GLORIA matched pairs are substantially more similar (0.326) but still notably changed (Table 4).
Table 4

Similarity between GLORIA sites sampled in 2003, 2004 and 2009; the codes refer to the site names, cardinal direction, and year of sample (e.g., BiE×03 is Bison Mountain, East, 2003)

Before

After

Sorenson

BiE×03

BiE49×09

0.3380

BiN×03

BiN×09

0.3272

BiS×03

BiS49×09

0.2741

BiW×03

BiW×09

0.2784

DLE×03

DLE×09

0.1244

DLN×03

DLN×09

0.6000

DLS×03

DLS×09

0.2438

DLW×03

DLW×09

0.0571

PiE×04

PiE×09

0.4724

PiN×04

PiN×09

0.2413

PiS×04

PiS×09

0.3280

PiW×04

PiW×09

0.3770

SeE×04

SeE×09

0.4168

SeN×04

SeN×09

0.5205

For the landscape and regional data, NMDS produced a roughly spherical cloud of points (each point representing the vegetation at a site arranged to reflect the dissimilarity in a few dimensions) that would be difficult to correlate with underlying environmental gradients. For the purpose of analyzing dissimilarities, however, this form best minimizes stress, the measure of goodness of fit of the distance in ordination space to the original dissimilarity of pairs (Table 5).
Table 5

Statistics on the outcome of nonmetric multidimensional scaling

 

Final stress

Final instability

Iterations

GLORIA

7.006

0.00000

147

Landscape

13.977

0.00000

248

Region

12.513

0.00000

265

The ordinations of the GLORIA sites show their trajectories in this limited ordination space (Fig. 1). There is no consistent direction of change, but the connected pairs show the small degree of change.
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-013-0253-3/MediaObjects/11258_2013_253_Fig1_HTML.gif
Fig. 1

Trajectories of the GLORIA sites in their own ordination space

Within the landscape context of GNP, the GLORIA sites are distributed at the margin of the cloud of points (Fig. 2). In most cases the distances between the pairs are small.
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-013-0253-3/MediaObjects/11258_2013_253_Fig2_HTML.gif
Fig. 2

Ordination of plots–525 from Damm (blue), 28 from GLORIA (red). (Color figure online)

For the regional context, the GNP sites are somewhat peripheral, but not exclusively so, while the GLORIA sites are definitely at the margin (Fig. 3). Here, distances between matched pairs are negligible.
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-013-0253-3/MediaObjects/11258_2013_253_Fig3_HTML.gif
Fig. 3

Ordination of all sites (28 GLORIA, green; 36 Damm GNP, red; 374 other, blue). (Color figure online)

Within the ordinations, scale matters (Tables 6, 7). Trivially, the distances between GLORIA pairs, and their corresponding volumes, tend to be smaller proportions of the whole as one expands the geographic context. The relative measures for the change in GLORIA sites are small; even for all GLORIA sites in GNP, the spatial variability exceeds the temporal change (Fig. 1) (cf. Ross et al. 2010), they are undetectable against the spatial variation revealed in ordination space. The relative lengths of the matched pairs average 0.092, 0.166 and 0.041 for the GLORIA, landscape, and regional data; the corresponding volumes are 0.013, 0.166 and 0.001 (one pair, Pitamakan Peak South (PiS09), was placed far apart in the landscape scale NMDS although the dissimilarity was a moderate 0.328).
Table 6

Overall dimensions of the ordination space determined in nonmetric multidimensional scaling in raw units (e.g., a length of 0.5 would indicate 50 % dissimilarity)

 

Length

Total volume

Axis 1

Axis 2

Axis 3

GLORIA

2.634

1.855

1.650

4.221

Landscape

3.957

2.853

3.281

19.391

Region

3.662

3.352

3.846

24.724

Table 7

Differences in paired GLORIA sites in 3-D NMDS ordination space represented by distance (Dist) and volume (Vol) with values relative to the longest axis (RelD) and total volume (RelV) of the 3D space

Pair

GLORIA

Landscape

Region

Dist

RelD

Vol

RelV

Dist

RelD

Vol

RelV

Dist

RelD

Vol

RelV

BiE×09

0.122

0.046

0.012

0.003

0.214

0.056

0.214

0.036

0.136

0.035

0.014

0.001

BiN×09

0.242

0.092

0.046

0.011

0.295

0.078

0.295

0.068

0.082

0.021

0.005

0.000

BiS×09

0.207

0.079

0.034

0.008

0.135

0.036

0.135

0.014

0.111

0.029

0.010

0.000

BiW×09

0.217

0.083

0.037

0.009

0.147

0.039

0.147

0.017

0.030

0.008

0.001

0.000

DLE×09

0.218

0.083

0.037

0.009

0.119

0.031

0.119

0.011

0.262

0.068

0.054

0.002

DLN×09

0.477

0.181

0.179

0.042

0.349

0.092

0.349

0.095

0.471

0.122

0.174

0.007

DLS×09

0.203

0.077

0.032

0.008

0.087

0.023

0.087

0.006

0.110

0.029

0.010

0.000

DLW×09

0.128

0.049

0.013

0.003

0.195

0.051

0.195

0.030

0.103

0.027

0.008

0.000

PiE×09

0.346

0.131

0.094

0.022

0.289

0.076

0.289

0.066

0.137

0.036

0.015

0.001

PiN×09

0.146

0.056

0.017

0.004

0.223

0.059

0.223

0.039

0.052

0.014

0.002

0.000

PiS×09

0.129

0.049

0.013

0.003

2.233

0.588

2.233

3.917

0.031

0.008

0.001

0.000

PiW×09

0.192

0.073

0.029

0.007

0.084

0.022

0.084

0.006

0.076

0.020

0.005

0.000

SeE×09

0.285

0.108

0.064

0.015

0.163

0.043

0.163

0.021

0.331

0.086

0.086

0.003

SeN×09

0.470

0.178

0.174

0.041

0.265

0.070

0.265

0.055

0.284

0.074

0.063

0.003

Mean

0.242

0.092

0.056

0.013

0.343

0.090

0.343

0.313

0.158

0.0412

0.032

0.001

StdDev

0.116

0.044

0.056

0.013

0.550

0.145

0.550

1.037

0.130

0.034

0.049

0.002

The GLORIA sites were all near the edge of the 3-D clouds of points in NMDS with both the GNP and regional datasets. The GLORIA sites are thus identified as a specific type of vegetation differing in species composition from alpine tundra in general.

Discussion

The change in similarity in the GLORIA matched pairs (dissimilarity = 0.33) is relatively large compared to the standard value for replicated samples of vegetation found in the literature (= 0.15–0.20; Bray and Curtis 1957, discussed by Goodall 1978) but the earlier figures are not for alpine vegetation and replication standards are rare. The change is relatively small compared to the dissimilarity across the range of sites at landscape and regional scale. This result might imply that the change seen is unimportant (cf. Ross et al. 2010), but this conclusion would depend on acceptance of the idea that all tundra is equivalent. Our analysis shows that the GLORIA summits are on the margins of the distribution of all our sites in 3-D ordination space, and so supports a restricted niche interpretation of alpine biogeography. The changes observed may be large relative to those that could occur in a limited portion of niche space.

These results depend on accepted taxonomy. Congeners identified at different places and different times have different names and so have the same weight in the ordination as species that are as distantly related as possible. If a more thorough analysis was to lump some species together, then the marginality of the GLORIA sites in GNP might not be so marked. Second, the present distributions might not be in equilibrium (cf. Harris 2007).

The temporal trajectories of the GLORIA sites make up successively smaller proportions of the ordination space as scale increases from local to landscape to region, but it is the size of the steps that matters. The changes represented in relative ordination space are reduced substantially when the geographical and ecological space expands from the GLORIA summits to the landscape of Glacier National Park. Within the regional scale, the shifts seen in the GLORIA sites over 4–5 years are small, but these changes cannot provide reliable information on the long-term trajectories of these sites in a future of climate warming. The vegetation appears to be sensitive to local, short term climate variability (Holzer and Fagre 2004), and background turnover may mask any trends. At this point consistent directions of change are not evident in our data. The shifts over 40 years seen in the NMDS ordination space presented by Danby et al. (2011) show larger changes than our 5 year interval would be expected to reveal, but these too are not consistent in direction; although sites of four different aspects tended to shift left in a 2-D ordination space, some aspect groups also shifted up while others moved down and the distances varied. Background rates can be better understood once more GLORIA results are compared because the degree of directional climate change over any 5 year interval will vary among locations and we can expect that vegetation change regressed against climate change will have a non-zero intercept.

Our analysis does not capture the importance of individual species or extinctions (cf. Ross et al. 2012; Kapfer et al. 2013). Individual extinctions could be weighted more highly than changes in vegetation similarity in evaluation of climate change impacts. Pauli et al. (2012) noted the potential for loss where endemics are higher in southern Europe, but actual extinctions will be difficult to document given the extent of alpine tundra that is not within GLORIA monitoring sites. Although GLORIA sites differ from alpine vegetation in general, they could, as intended, help identify those mountaintop species that are most threatened by climate change, especially those at the nival/subnival ecotone that will have fewer other locations of their habitat. Patience is needed with a program such as GLORIA in order to assess the degree and direction of change that may occur over decades, and a broad perspective is required in order to evaluate the importance of change across regions and globally.

Acknowledgments

This research was supported by a seed grant from the University of Iowa Center for Global and Regional Environmental Research to GPM, by a US Geological Survey Park-Oriented Biological Support grant to DBF, and by NSF award 1121305. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. This is a contribution from the Mountain GeoDynamics Research Group.

Copyright information

© Springer Science+Business Media Dordrecht (outside the USA) 2013