Study Area

Comprehensive photographic archives, extensive documentation of paleoclimates and past human interactions with the landscape, and considerable prior research on alpine and subalpine environments make GNP an ideal study area for repeat photography. GNP occupies 4075 km² in northwestern Montana (Fig. 1), where elevations range from 948 to 3290 m a.s.l. (Shaw and On, 1979; Rockwell, 2002). Steep topography, climatic fluctuations, and past glaciations have created a treeline that is much more spatially variable than that occurring in the southern Rocky Mountains (Cairns and Malanson, 1997). Becwar and Burke (1982) used topographic maps to estimate that 80% of the forest-tundra transition spans an elevation range of 550 m in GNP, compared to 150 m in the southern Rocky Mountains.

Figure 1

Location of Glacier National Park. Latitude and longitude are given in Figure 2.

Figure 2

Locations of the 12 sites which were rephotographed in Glacier National Park. See Table 2 for site details.

The area that is now GNP was protected as a forest reserve prior to 1910, when the limited human impact in the area was further reduced by the establishment of the park. The area was never commercially logged, and grazing has been largely limited to horses used for pack trips (Shaw and On, 1979; Vogler, 1998). During the 1930s an outbreak of whitebark pine blister rust in high elevation sites infected whitebark pines (Pinus albicaulus Engem.) and by the 1950s had killed most of these trees (Kendall and Keane, 2001). Recent human impact in alpine areas of GNP is minimal and is largely due to trail use and maintenance. Probably the greatest direct human impact on vegetation in GNP has been a policy of fire suppression over the past 80 years that has contributed to increased forest cover and reduced early successional habitats. This policy now allows some fires to burn for wildland benefit. Fires occasionally burned into alpine areas in the past and may have affected treeline ecotone dynamics (Arno, 2001). Study sites for this project were located within the treeline ecotone both at the extreme upper limit of trees and in open subalpine meadows at slightly lower elevations. Topography at the sites varied from gentle meadows to steep rocky hillsides. The dominant treeline species at all sites is subalpine fir (Abies lasiocarpa [Hook.] Nutt), with isolated instances of Engelmann spruce (Picea engelmannii Parry ex Engelm.). Observed treelines were both orographic and altitudinal as defined by Holtmeier (2003).

Selection of Historical Photographs

The historical photographs replicated for this study were obtained from GNP's archive and were taken by three early 20th century photographers: George Grant, Tomer J. Hileman, and Fred Kiser. The collection of photographs from which selections were made was extensive and contained views from throughout the park, although particularly scenic or easily accessible views are more common. Approximately 2000 photographs were examined before final selections were made. Photographs were selected based on several criteria. First, the images needed to be of high quality and to clearly document the treeline ecotone. Second, photographs were selected with the goal of isolating climate as the primary factor that would cause a change in the characteristics of the treeline ecotone shown in the photograph. Third, photographs of a hillside of roughly constant slope that extended above a non-orographic treeline were preferred. Album prints of 17.5 × 26.5 cm (8 × 12 in.) were scanned at 273 dots per cm [600 dots per inch (DPI)], and saved as Tagged Image File Format (TIFF) files. Printed copies were used in the field to locate historic photo-points.

Locating Historical Photo-points

To locate the sites from which historical photographs were taken, information accompanying the photographs or prominent landmarks in the view were used to find the general area on a topographic map. To determine the precise location of photo-points in the field, the principle of parallax was used to find lines of equal perspective and proportionate length between features (e.g. background ridgeline intersections and permanent foreground objects, such as boulders) in the historical photograph and the present landscape (Rogers et al., 1984). Photo-points were located to within ~20 m using background features, then adjusted to within ~1 m of the original point using foreground features. In a few photo-pairs new vegetation obscured the view shown in the historical photograph, so the modern photograph was taken from a slightly different location (<2 m away). Figure 2 shows the locations of all photo-points included in this study.

A Nikon F-100 35 mm single lens reflex camera with a 24–200 mm f3.5 Nikon lens was used to shoot Kodak technical-pan black and white film at an ASA of 12. For each historical photograph site, 12 frames were taken with bracketed exposures; 6 of these used a Yellow 12 filter to increase contrast. The location of each photo-point was recorded using both a GPS unit and a topographic map. Camera height and bearing as well as aperture, shutter-speed, and focal length were also recorded.

Analysis of Photo-pairs

Of the 12 modern negatives per photo-pair, the image with exposure and contrast values which most closely matched the historical image was selected for comparative analysis. All modern negatives were converted to TIFF files by scanning at 1820 dots per cm (4000 DPI). To quantitatively analyze change between the modern and historic image files, an original method was developed using ArcGIS software (ESRI, 2002). First, the paired image files were imported into ArcMAP and ortho-referenced to each other using a minimum of five common points with the majority of these located in the vicinity of the area to be analyzed. This step ensured that the scales of the two images were equal and that they were overlain exactly in ArcMAP's spatial system. Second, an ESRI “fishnet” script (ESRI, 2002; Nicholas, 2003) was used to create a grid overlying the area of interest. Each cell of the grid is an independent polygon (Fig. 3), and information about what lies under each grid cell can then be used to populate a table corresponding to the grid in a spatially explicit manner. The grid was scaled to vegetation features depicted in each photo-pair, and because the scale of each photo-pair and corresponding grid varies depending on the distance of the subject from the photo-point, the amount of ground covered by each cell differs among photo-pairs. Thus, only relative changes can be measured between photo-pairs. Additionally, within a photo-pair, foreground space will occupy a larger area in the image than background space. To mitigate this issue, the area chosen for analysis in each photo-pair was held to a roughly constant distance from the photo-point.

Figure 3

Attribute table and associated grid overlain on study site Hileman 5053. Each cell has a unique identifier (FID). Information regarding historical and modern vegetation as well as the calculated type of vegetation change populates the table.

Third, the grids scaled to each photo-pair were digitally overlain on the images, and cropped to the area from which cell-by-cell changes were to be measured. An attribute table was created for each grid file, with fields containing information regarding the vegetation visible in each cell (Fig. 3). Each cell received a value of 1, 2, or 3 indicating bare ground, partial tree cover, and complete tree cover, respectively. This information was then used to create mutually exclusive categories (no change, vegetation loss, infilling, and establishment) describing the type of tree cover change occurring within each grid cell. No change indicates no visible difference in the cell between the two photographs; vegetation loss is defined as a decrease in visible vegetative cover in the modern photograph; infilling is defined as increased or denser vegetation in areas which showed sparse vegetation in the historical photograph; and establishment is defined as new vegetation in areas which showed bare ground in the historical photograph (Table 1). It was not possible to determine the absolute change in tree cover documented in the photo-pairs because of the continuously varying scale created by the oblique perspective and the complex topography.

Table 1

Definitions of cell types showing categories of tree cover change.

Finally, to evaluate the effect cell size had on the analysis, two photo-pairs were resampled with a new grid, the cell size of which was four times larger than used in the first analysis, and these results were compared to the original analysis.

Photo-pair Hileman 8042

This photo-pair illustrated the most dramatic expansion of trees into a meadow environment, likely as a result of several factors that make this site ideal for tree growth and establishment (Fig. 4). First, stands of mature trees located just to the north (right) of the study site provide a substantial and proximal seed bank. Second, the physical environment at this site appears generally favorable for tree growth, as exemplified by the trees in the original Hileman 8042 photograph. Despite their location immediately below the upper treeline, these trees have a growth form more typical of those in lower elevation, mature forests. In contrast, other historical photographs show trees in krummholz forms or as isolated tree patches that are often located below cliffs and on more exposed mountainsides where winter desiccation and ice abrasion are greater impediments to upright tree growth (Holtmeier, 2003; Malanson et al., in press). Finally, the ground at this site is flat to gently sloping and includes many minor topographic highs, which Butler et al. (2003) observed to have a strong influence on tree position by providing well-drained soil that increases establishment.

Figure 4

Detail of photo-pair Hileman 8042 showing vegetation type superimposed on the historical photograph and the type of vegetation change superimposed on the modern photograph. Note the massive invasion of trees across the entire photograph.

Photo-pair Hileman 5053

The treeline in the photo-pair Hileman 5053, located at the base of a scree slope and near the upper elevation of tree growth, provides an example of two controls acting in unison on the treeline at a single location (Fig. 5). The photo-pair shows that the upper limit of tundra has not changed perceptibly in the intervening 72 years, suggesting that tundra vegetation is unable to colonize the talus. Upright trees however, have clearly expanded their range (approximately 10–20 vertical m), with some now occupying the upper limit of all vegetation at this site. It is possible that the functional treeline at this site may be depressed due to the presence of the talus, which is hypothesized to be the second largest control on treeline in GNP after climate (Cairns, 2001). This notion is supported by the presence of upright trees within the upper portion of the treeline ecotone at this site. Thus, the combined effect of tree establishment and infilling, and the inability of alpine meadow species to colonize the active talus, has caused a marked decrease in tundra area at this site while the density of trees has increased.

Figure 5

Detail of photo-pair Hileman 5053 showing vegetation type superimposed on the historical photograph and the type of vegetation change superimposed on the modern photograph. Note the disappearance of meadow below the talus and the preference of tree establishment to occur adjacent to historical vegetation in the three tongues of meadow which used to reach down into the forest.

The majority of the infilling and establishment in this photo-pair occurred along the edges of historical vegetation and may be a result of existing trees improving conditions for establishment in their immediate environment by altering soil characteristics and microclimates (Stevens and Fox, 1991; Holtmeier, 2003; Holtmeier and Broll, 2005). Specifically, establishment was concentrated in the three tongues of meadow that reach down into the trees in the historical photograph (Fig. 5). It is not possible to determine the direct cause of the observed vegetation change from the photo-pair alone. However, the distinct spatial pattern related to distance from both the talus and historical tree clumps suggests that protection from wind plays a role in the spatial variability of tree establishment within the treeline ecotone as was observed by Resler (2006) on Lee Ridge in GNP.

Photo-pair Kiser 11813

The change seen in photo-pair Kiser 11813 is notable because 61% of the grid cells record change (Fig. 6), but this occurred largely as infilling (41%), with a relatively low level of establishment (14%). Tree growth proceeded outward from the lower center of the site, where trees were densest in the historical photograph. The small amount of establishment occurred primarily on the rising slope at the left side of the site and represents an overall altitudinal increase of 5–10 m. Most notable is the disappearance of the ribbon forests present in the historical photograph. This forest type is thought to be caused by underlying bedrock topography (Butler et al., 2003) creating topographic highs and depressions in the landscape. The observed change indicates that the factors which caused the infilling of the ribbon forest are powerful enough to overcome geologic controls on tree location.

Figure 6

Detail of photo-pair Kiser 11813, showing vegetation type superimposed on both the historical and modern photographs. Note the disappearance of the ribbon forest environment present in the historical photograph and the slight upward movement of treeline on the far left side of the photograph.

General Trends and Interpretations

The most common trend documented by the photo-pairs is an increase in tree cover, both as infilling and establishment. Establishment occurred up to hundreds of meters away from historical vegetation, and half the sites experienced a 35% or greater increase in total tree cover. Even in the two sites where overall tree cover decreased, localized establishment and infilling occurred. Local factors certainly play a significant role in determining the magnitude of change that occurs at any given site (Holtmeier, 2003), yet the ubiquity of the observed changes indicates there is a more universal forcing mechanism that is amplified or attenuated by local conditions. Previous studies have documented an increase in tree establishment in subalpine forests in response to increased growing season temperature (Kullman, 1991; Rochefort et al., 1994; Cairns and Malanson, 1998; Kusnierczyk and Ettl, 2002; Lloyd et al., 2002), while Hofgaard et al. (1991) and Kullman (1991) in Scandinavia, and Rochefort et al. (1994) and Hessl and Baker (1997) in the American and Canadian Rockies, found a strong correlation between increased tree growth at treeline and rising summer temperatures. Data from the Kalispell WSO AP Station, part of the Historical Climate Network (Easterling et al., 1996) reveal that while the yearly mean temperature has increased only a small amount since 1900 (about 0.5°C), minimum summer temperatures have increased about 1.5°C in the past century (Fig. 7), lengthening the growing season by increasing the number of frost-free days. Watson et al. (in press) show a similar or greater increase in minimum spring and summer temperatures from 15 temperature stations along the continental divide. Butler et al. (1994) noted that increases in tetraterm temperatures (average temperature of the period June to September) observed in GNP from 1984 to 1991 could be linked to seedling establishment. And in a summary of studies examining seedling establishment in the alpine tundra, Alftine et al. (2003) found tundra invasion by forests of the Pacific Northwest to be associated with warmer, drier periods. Together these types of climatic changes would improve the carbon balance for existing trees and make it easier for seeds to germinate and prosper. While establishing direct causality for increased tree growth is not possible from the photo-pairs alone, changes in climate occur at a large enough scale to have influenced tree establishment at all sites in this study. Thus, the rise in mean summer minimum temperatures over the past several decades (Fig. 7) is a possible explanation for the observed changes, with differences in the magnitude and nature of the observed change demonstrating the importance of site specific conditions.

Figure 7

Annual spring snow pack (1 May Snow Water Equivalent) from Flattop Mountain and mean minimum summer (June–August) temperatures at the U.S. Historical Climatology Network's Kalispell WSO AP Station, 70 km from Glacier National Park. Black lines represent 10-year moving averages.

Precipitation also affects tree growth and establishment but in more complex ways depending on whether water comes in the form of snow or rain (Holtmeier, 2003). Most of the precipitation at high elevations in GNP occurs as snow (Klasner and Fagre, 2002) and has more interannual variation than temperature (Fig. 7). Kusnierczyk and Ettl (2002) documented a positive correlation between non-growing season precipitation and tree growth, and given that cold soils and high winds are the largest limiting factors on tree establishment in continental climates, a deep snow pack would reduce these negative effects and aid tree establishment (Rochefort et al., 1994). The 10-year running mean of 1 May Snow Water Equivalent (SWE) from Flattop Mountain in GNP shows a marked increase during the period 1950–1975 (Fig. 7). This increase differs from the temperature record, which is fairly stable during this period and shows its most significant gains between 1915 and 1930 and again between 1970 and 1980 (Fig. 7). Selkowitz et al. (2002) demonstrated a link between longer term (20–30 yr) cycles of increased SWE within GNP and the negative phase of the Pacific Decadal Oscillation (PDO), and Alftine et al. (2003) found high levels of establishment occurring in the treeline ecotone of Lee Ridge in GNP during the negative phase (late 1940s to late 1970s) of the PDO. It is likely then that some of the establishment seen in the photo-pairs is a result of the increased snow levels during the period 1950–1975.

Consistent with other studies in GNP, photo-pairs from this study indicate that krummholz patches appear to have changed far less than upright trees at treeline or in the subalpine meadow environment. A rephotographic study by Butler et al. (1994) showed treeline position to have been quite stable throughout the park from 1972 to 1992, including krummholz patches that were exactly the same size in historical and modern photographs.

Photo-pair Kiser 11810, which includes two separate sites, underscores the complex interrelationships between altitude, growth form, precipitation, and exposure. Tree establishment was more than four times greater (Table 2) at the exposed mountain site where upright trees are present ~245 m above the sheltered, krummholz-dominated valley site. The valley site likely has a greater potential for excessive snow drifting, which has been negatively correlated with establishment by limiting growing season length and quality (Franklin et al., 1971) and by reducing the carbon balance of the trees (Cairns and Malanson, 1998). Other studies have shown that snow accumulation can benefit trees by protecting them from winter desiccation and injury (Walsh and Kelly, 1990; Cairns and Malanson, 1997; Hattenschwiler and Smith, 1999). In this case, it seems likely the valley site receives so much wind-deposited snow that the shortened growing season outweighs the sheltering effects of the winter snow pack and limits seedling establishment.

Finally, it should be noted that fire is not thought to be a major factor in causing the changes documented by the studied photo-pairs. Forest stand-age maps were consulted in selecting the historical photographs to eliminate the possibility that the changes documented in the photo-pairs were simply due to succession following a stand-clearing fire. These stand-age maps do not necessarily provide the level of detail necessary to identify less-extensive fires, but the effects of fire would be obvious in the photo-pairs. For instance, at one site—excluded from the final analysis—a fire occurred between the historical and modern photographs, resulting in a loss of tree cover in 73% of the grid cells and limiting tree cover increase to only 6% of the grid cells. Stand-age maps showed this site as having had no stand replacing fires in the past 200 years, yet evidence from the photo-pair was unequivocal. The thin bark of subalpine fir (the dominant treeline species in the photo-pairs) makes this species particularly vulnerable to high intensity fires (Alexander et al., 1990) and as a result, such fires leave considerable visual evidence in the landscape. Furthermore, while subalpine fir establishment does occur on burned areas, fire is not an essential part of the life cycle of this shade-tolerant tree (Alexander et al., 1990). Thus the observed establishment and infilling of subalpine fir does not require fire to create suitable habitat.