SEARCH

SEARCH BY CITATION

Keywords:

  • alpine tree line rise;
  • Alaska;
  • shrubs;
  • tundra;
  • climate change

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Site Description
  5. 3. Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions and Implications
  9. Acknowledgments
  10. References
  11. Supporting Information

[1] The complex response of the forest-tundra ecotone (FT) to climate change may not generalize well geographically. We document FT changes in a nonpermafrost region of southcentral Alaska during a known warming period. Using 1951 and 1996 orthophotos overlain on digital elevation models across 800 km2 of the west Kenai Mountains, we identified cover classes and topography for 978 random points and the highest closed-canopy conifer patches along 205 random altitudinal gradients. Results show 29% of FT area increased in woodiness, with closed-canopy forest expanding 14%/decade and shrubs 4%/decade; unvegetated areas decreased 17.4%/decade and tundra 5%/decade. Area of open woodland remained constant but changed location. Timberline, estimated using both the 205 altitudinal gradients and the upper quartile elevations of closed-canopy forest among the 978 points, rose very little. Tree line, identified using upper quartiles of open woodland, rose ∼50 m on cool, northerly aspects, but not on other aspects. Dendrochronology on high-elevation seedlings showed a congruence between decadal recruitment and regional changes in climate from 1945 to 2005. Patterns observed in the climatic FT of the Kenai Mountains corroborate other studies that show regional and landscape specificity of the structural response of FT to climate change. FT shifted upwards on cooler, presumably more mesic aspects near seed sources; however, on warm aspects the density of shrubs and trees increased, but FT did not rise. If current conditions continue for the next 50–100 years, the Kenai FT will markedly change to a far woodier landscape with less tundra and more closed-canopy forest.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Site Description
  5. 3. Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions and Implications
  9. Acknowledgments
  10. References
  11. Supporting Information

[2] The alpine forest-tundra ecotone is a transition zone between alpine tundra and subalpine forest and is often considered a likely indicator of climate change [Payette et al., 2001; Grace et al., 2002]. Within this ecotone, the altitudinal boundaries between shrubland and woodland (tree line) and between forest and woodland (timberline), might both be expected to shift in response to changing climate as the ecotone itself shifts. Shifting boundaries that are defined by trees could indicate climate change because of the physiological dependence of tree growth on variables that vary with both altitude and climate, such as temperature and moisture [Sveinbjornsson, 2000; Grace et al., 2002; Sveinbjornsson et al., 2002; Holtmeier and Broll, 2005]. Even so, the suite of processes that limit tree growth at tree line are known to be complex, not fully understood in all cases, and particularly dependent on regional, landscape, and even microsite processes that bear on the establishment, growth, reproduction, and mortality of trees [Payette et al., 2001; Grace et al., 2002; Holtmeier and Broll, 2005]. For example, a warming climate may offer new germination opportunities as alpine soils are warmed and wetted [Korner, 1998], or may lead to a reduction in opportunities if moisture stress increases [Barber et al., 2000; Gedalof and Smith, 2001; Wilmking et al., 2004], if competition with shrubs that respond favorably to warming is strong [Hobbie and Chapin, 1998], or if snowpack thins in a continental climate or thickens in a maritime one [Holtmeier and Broll, 2005]. Moreover, tree line rise must lag climate change due to the temporal scale of the mechanisms involved in the spread of trees into treeless communities [Lloyd et al., 2003]. Indeed, the complexities of interacting factors known to determine the forest-tundra ecotone imply that the simple rising or falling of tree line can not be a universal indicator of climate change [e.g., Payette et al., 2001; Grace et al., 2002; Holtmeier and Broll, 2005]. Nevertheless, reviews of climate change's effects on organisms and ecosystems often site numerous studies in many parts of the world that show tree line rising with temperature [e.g., Walther et al., 2002], and doing so independent of local anthropogenic effects.

[3] Alaska offers clear evidence of warming and climate-induced vegetation and geomorphologic change with only negligible effects from local human activity. Annual mean temperatures in Alaska have risen approximately 2°C since the 1950s, while mean winter temperatures of the interior have risen approximately 4°C [Alaska Regional Assessment Group (ARAG), 1999; Serreze et al., 2000; Arctic Climate Impact Assessment (ACIA), 2004; Chapin et al., 2005]. Documented effects of vegetative change in Alaska and nearby Canada concurrent with recent warming include conversion of wetlands to dry-land habitats in the southcentral region [Klein et al., 2005] and elsewhere [Riordan et al., 2006]; loss of conifer cover due to spruce-bark beetle infestation [Berg et al., 2006; Boucher and Mead, 2006] in southern Alaska and Yukon Territory; tree line rise in western and central Alaska (reviewed by Lloyd [2005]) and in southwestern Yukon [Danby and Hik, 2007]; and the expansion of shrubs in the arctic [Sturm et al., 2001; Stow et al., 2004; Tape et al., 2006]. The propensity for shrubs, a key structural component of the forest-tundra ecotone, to decrease albedo has been postulated as a positive feedback mechanism in the changing climate of arctic Alaska [Sturm et al., 2005; Chapin et al., 2006]. As reviewed by Lloyd [2005], most studies that have investigated tree line changes in Alaska [e.g., Jacoby and D'Arrigo, 1997; Suarez et al., 1999; Lloyd et al., 2002; Lloyd and Fastie, 2003; Chapin et al., 2005] have focused on arctic regions and others influenced by permafrost.

[4] The overall goal of this study was to fill a geographic gap in the understanding of vegetation change at the alpine forest-tundra ecotone in Alaska, as most studies have been more northerly both for trees (reviewed by Lloyd [2005]) and shrubs [Tape et al., 2006]. In particular, using the forest-tundra ecotone of the Kenai Mountains in southcentral Alaska, we hoped to provide information on how observed changes at altitudinal tree line might be used as indicators of climate change over a half-century, as such responses are known to be rather specific to region, landscape, and locality [Holtmeier and Broll, 2005]. Five terrestrial cover types that are easily identifiable from black and white, aerial orthophotos—unvegetated, tundra, shrub, open woodland, and closed-canopy forest—are important structural elements, interrelated by succession in most forest-tundra ecotones. However, few studies have quantified changes in all five elements and how they interrelate in a warming climate [Danby and Hik, 2007]. Our study questions concerned the transitions among these five cover types during a 45-year (a) period and the shifting altitudinal distribution and limits of the three woody cover types: shrub, open woodland, and closed-canopy forest. We wished to determine how the Kenai Mountains might inform us of changes generally in the forest-tundra ecotone, vis a vis climate changes that are warming and humid vs. warming and drying. We were motivated by several models developed to conceptualize the response of the forest-tundra ecotone to changes in climate [Payette et al., 2001; Grace et al., 2002; Holtmeier and Broll, 2005; Zeng and Malanson, 2006]. It may well be the case that the response of the forest-tundra ecotone to climate change is too complex to generalize outside of any well-studied region, and will reduce to site-specific models requiring detailed local knowledge of the particular ecotone example and its pertinent processes.

2. Site Description

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Site Description
  5. 3. Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions and Implications
  9. Acknowledgments
  10. References
  11. Supporting Information

[5] The 800 km2 study area was located 40–70 km from the nearest tidewater in the western Kenai Mountains of the Kenai Peninsula, southcentral Alaska (Figure 1a). The varied topography of the entire peninsula includes lowlands, rounded hills, rugged peaks to 2000 m above sea level (asl), large icefields and glaciers, and is subject to a maritime climate to the east and a continental climate to the west. Major habitats include treeless alpine, shrubby subalpine, boreal forests, and wetland. Dominant trees and tall shrubs in the forest tundra ecotone include Tsuga mertensiana, Picea sp., Populus sp., Betula sp., Alnus sp., Salix sp., and Sorbus sp. [Boucher and Mead, 2006; Berg et al., 2006]. While spruce-bark beetle infestations and fire are important processes in the Kenai lowlands [Boucher and Mead, 2006; Berg et al., 2006], the abundance of mountain hemlock at tree line suggests that fire plays little, if any role, in maintaining the forest-tundra ecotone there. Alpine trees are generally of a size class too small for beetle attack; principle browsers (moose, caribou, hares) feed little on conifers in Alaska; anthropogenic tree lines (in the sense of Holtmeier and Broll [2005]) are absent, and orographic tree lines less prominent on the western slope of the Kenai Mountains, where this study took place, than the interior or eastern portion with its higher snowfall and steeper topography.

image

Figure 1. (a) Study area and locations of local and regional study sites on the Kenai Peninsula, southcentral Alaska. (b) Climatic data (1944–2005) from Kenai, Alaska, 60.563 N 151.244 W. (top) Mean growing season temperatures. (middle) Water balance as precipitation (P) minus potential evapotranspiration (PET). (bottom) Drought index calculated using standard-normalized mean May–August temperatures minus standard-normalized October–September precipitation for each year. Positive values indicate potential drought stress for vegetation. Means given in each panel on left are averaged over years previous to 1968; means on right are averaged over years 1968 and following.

Download figure to PowerPoint

[6] The western Kenai is warmer than Alaska's interior and drier than most of coastal Alaska, with a mean annual precipitation of 46 cm/a [Klein et al., 2005]. Figure 1b presents climate data for the city of Kenai. Located 40–100 km from the study site, Kenai is the nearest weather station. Like other nearby stations (Anchorage, 50–130 km distant; Homer, 70–150 km distant; data from http://climate.gi.alaska.edu/Climate/Location/TimeSeries/index.html), Kenai is at sea level and not representative of absolute temperatures in the mountains. Nevertheless, relative May–August temperature patterns during the last half-century in Kenai nearly match those in Anchorage (95 km NE) and Homer (100 km S). All three sites show increasingly warmer decades since the 1940s–1950s, reflecting the north Pacific sea surface temperatures captured by the Pacific Decadal Oscillations index (PDO [Mantua et al., 1997]; data from http://jisao.washington.edu/pdo/PDO.latest). These similarities imply that the relative temperature trends are regional and likely apply across the western Kenai Peninsula. In particular, all three sites show an abrupt warming in the early 1970s that is consistent with climate data from much of Alaska [ARAG, 1999] (http://climate.gi.alaska.edu/ClimTrends/Change/TempChange.html). Extrapolations to mountain environments using sea level data must be applied with caution, although the rapid mass wasting of glaciers in Alaska [e.g., Arendt et al., 2002; Rabus and Echelmeyer, 1998] often matches climate trends at sea level, such as the PDO.

3. Methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Site Description
  5. 3. Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions and Implications
  9. Acknowledgments
  10. References
  11. Supporting Information

3.1. GIS Data Sources and Orthophoto Sampling

[7] We used two aerial photosets covering a 2540 km2 area and taken in the early 1950s and in 1996; a recently corrected digital elevation model; and a satellite-based land cover classification. Aerial, black and white, digitally scanned orthophotos taken June–August 1950–1952 (3 m resolution; referred to here as “1951”) and June–August 1996 (1 m resolution) provided data for this study. Topographic data were derived from a United States Geological Survey (USGS) National Elevation Dataset digital elevation model (NED DEM at 30 m resolution [National Digital Elevation Program (NDEP), 2004]). A vegetation classification (http://agdc.usgs.gov/data/usgs/erosafo/ak_lcc/ak_lcc.html) produced by the Alaska Land Cover Mapping Project (ALCMP) of the Kenai National Wildlife Refuge was used as ancillary data. All data were projected onto UTM NAD27 using ArcInfo 9.1 (ESRI).

[8] We made three orthophoto comparisons of 1951 and 1996. (1) A regional comparison from the western Kenai Mountains north of Tustumena Lake used 978 random points from an 800 km2 alpine region that were classified into one of the five cover classes on digital orthophotos for each year, then stored and displayed on a GIS (ArcInfo 9.1, ESRI). (2) An intensive local study used subsamples from three areas chosen for their substantial increase in woody cover between 1951 and 1996. Incoming solar radiation values were calculated for both the local and regional study sites for each sample point based on slope, aspect, latitude, and neighboring topography. (3) A comparison was made of individual patches of woody growth, both tree and shrub, to document expansions or contractions of uppermost patches along 205 altitudinal gradients.

[9] We used five land cover classes as vegetative states: unvegetated (snow patches, bare ground, rock); tundra (light in tone, fine textured, low stature); shrub (darker in tone, rougher textured); open woodland (distinct trees, coarsely textured, open canopy), and closed-canopy forest (darkest tone, continuous forest cover). All sample points were attributed with a cover class in each photoset to determine changes in vegetative states over the 45-a period. As each point was classified in both years, we were able to construct transition matrices to tally the total number of observed vegetative state changes within each of the 25 different state changes (=5 states × 5 states). Matrix rows gave each point's state in 1951 and columns its state in 1996. We defined the matrix rows as entry states and the matrix columns as exit states when describing dynamics among cover classes. We also ordered cover classes from least woody state to most woody state as: unvegetated (least woody), tundra, shrub, open woodland, and closed-canopy forest (most woody). Any point that did not differ in 1996 from 1951 was classed as no change; any point classified as a higher ordered state in 1996 than in 1951 (e.g., unvegetated to any other state; tundra to shrub, open woodland, closed-canopy; etc.), was reclassified as increased woodiness; otherwise a point was reclassed as decreased woodiness. No distinction in woodiness was made for multiple state changes over the time interval (e.g., increased woodiness included both shrub-to-open-woodland and tundra-to-closed-canopy forest).

3.2. Regional Study

[10] We began with 7000 initial random points spread over 2540 km2 of aerial photo coverage. USGS maps of the 1950s showed that the forest-tundra ecotone was above 450 m asl and studies of T. mertensiana [Means, 1990; Nienstaedt and Zasada, 1990], the highest elevation conifer in Alaska, identified its upper elevation at 1100 m asl. Eliminating points outside the 450–1100 m asl interval; those classed by the ALCMP as dwarf scrub peatland, string bogs-wetlands, or water; and those with poor photo resolution, shadows, or cloud cover left n = 978 suitable points within an 800 km2 alpine region for the regional study (Figure 1a). We assigned a cover class to each point on a 1951 photo and again on a 1996 photo, and analyzed the cover class and woodiness changes in terms of slope, aspect, elevation, and solar radiation.

3.3. Timberline, Tree Line, and Shrub-Line Rise

[11] We used two approaches to quantify tree line, timberline, and shrub-line rise. The first method attempted to quantify the commonly used definition of tree line, timberline, and shrub-line as “uppermost lines” on the landscape. The second method considered shifts in the statistical, altitudinal distribution of shrubs, woodland, and closed-canopy forest as a metric for changing shrub-line, tree line, and timberline.

[12] In method 1 the highest elevation patches of closed-canopy forest were identified along altitudinal gradients based on the NED DEM. Random points (n = 205) were chosen on the DEM, then a sample altitudinal gradient was constructed through each random point between the local summit and the 450 m contour line. The highest elevation patches of closed-canopy conifer trees were identified along each of the 205 gradients on both the 1951 and 1996 photos. The mean elevation of the upper and lower edges of each patch along the gradient for a given year provided the timberline elevation for that gradient and year. The disadvantage of this approach is that if a patch expanded equally downward and upward, then the mean elevation would not change. We calculated the rise in timberline as the mean difference between 1951 and 1996 timberline estimates from each of the 205 gradients. A similar approach was used for shrub patches.

[13] Method 2 considered the distribution of elevations for each vegetated class across the ecotone. To detect the upward movement of woody vegetation, we constructed histograms of elevation collected from the NED DEM for the 978 regional points for each of the woody states (shrub, open woodland, and closed-canopy forest). Because the literature on tree line emphasizes that both tree infilling and altitudinal rise characterize a shifting forest-tundra ecotone, we settled on the upper elevation quartile value (Q75) for each of the woody classes (shrub, open woodland, closed canopy) and the lower quartile for tundra (Q25; to determine if tundra was retreating upward) as altitudinal metrics for each year. In an orographic tree line, the upper quartile could shift over time as density increases, while the upper limit does not change, a limitation of this approach. Subtracting the upper quartile values (1996 Q75 minus 1951 Q75) for closed-canopy forest gave a measure of timberline rise; for open-woodland a measure of tree line rise; and for shrub cover a measure of shrub-line rise. In a climatic tree line, the upper quartile metric would likely capture both an increase in density and a rise in uppermost elevation. This method is analogous to determining the “LC75” in toxicology, the amount of a toxin that leads to mortality in 75% of the individuals from a sample.

3.4. Local Study

[14] Within the regional study, three local areas displayed substantial changes in land cover from 1951 to 1996. These areas (Figure 1) were selected for intensive local analysis to investigate processes that might lead to more rapid ecotone changes. The additional observations were not included in any regional estimates. Within a 64 ha square at each site (site details given below in section 4.4), points were sampled systematically at 30 m intervals using a GIS for a total of 729 points per site. As in the regional study, we documented the cover class of each point in 1951 and 1996 using digital orthophotos; however, because each site was small, only solar radiation (defined below) was used to compare local points that changed in woodiness from those that did not.

3.5. Influence of Slope, Aspect, Elevation, and Solar Radiation (Insolation)

[15] The influence of slope, aspect and elevation on vegetation changes were investigated both separately and combined as solar radiation for the regional points. Slope, aspect and elevation for all sample points were taken from the NED DEM. Solar radiation (WH/m2) was calculated for each sample point using elevation, slope, aspect, latitude, shadows from adjacent topography, daily and seasonal shifts in solar angle, and atmospheric attenuation using Solar Analyst 1.0 [Fu and Rich, 2000], an extension to ArcView 3.1 (ESRI). Solar Analyst calculates direct solar radiation, diffuse solar radiation, and their sum for any specified time interval. The program multiplies the solar irradiance constant (1,367 W/m2) by time duration (e.g., day length in hours, H, giving WH/m2) and by factors that reduce radiation such as latitude, date, transmittivity (a function of elevation), surrounding topography casting shadows, slope angle, and aspect. The program's output is an insolation value representing the sum of direct and indirect solar radiation in WH/m2 calculated for each sample point. The NED DEMs provided each point's latitude, slope, aspect, elevation, and surrounding topography. The same mask was used for both the regional and the local analysis. Using the spatial values from the DEM, solar radiation values were calculated for the equinox, summer, and winter solstices for each sample point.

3.6. Statistical Analyses

[16] To compare frequencies of cover classes and conduct tests of independence, we used a log likelihood G-test. Slope was classified into three categories (flat, moderate, and steep) and a G-test between woodiness change category (increased woodiness, no change, decreased woodiness) and slope category tested independence. Aspect was treated using circular statistics of angular mean and standard deviation. Elevation of variously classified sample points were compared using a Mann-Whitney test and insolation of points using a Kolmogorov-Smirnov test.

3.7. Field Methods and Their Analysis

[17] In 2005 we selected seven sites of mature T. mertensiana stands near the road system in the western Kenai Mountains at a distance of 40–45 km from the Turnagain Arm of Cook Inlet, then sampled nearby but higher-elevation alpine sites using belt-transects for recruitment of Picea and Tsuga. Width of belt-transects varied by site. Transects began at the highest elevation recruitment-sized conifer (<1 m in height) that we could find and continued downslope, until either the terrain was impassable or we found no additional recruitment-sized trees. We counted all tree individuals within these belt-transects and all individuals were <1 m. None of our sample belt-transects had individuals >1 m high or individuals growing as krummholz; we did not sample from mature stands. All individuals from all transects apparently recruited directly by seed. Our justification for excluding lower elevation, older trees was that we sought to age only the uppermost recruits and compare their ages to climate data. While we did find very small individuals (<3 cm in height), it is likely that we missed other individuals of the smallest-sized recruits, thereby underestimating the youngest cohort. Individual recruit locations were recorded with a GPS (Trimble GEO XT). Recruits were rare to nonexistent at three sites, despite T. mertensiana stands near the road. Recruits (113 total = 70 P. glauca + 43 T. mertensiana) were not sufficiently abundant for an age census. Only 63 (38 P. glauca and 25 T. mertensiana) recruits were used for aging. Samples were oven-dried, a cross-section removed directly above each root collar, sanded, then aged by ring count. One-tailed Spearman's rho correlation tested the hypothesis that age decreased as elevation increased.

3.8. Vegetation Change Modeling

[18] Markov models are broadly applied in plant ecology [e.g., Horn, 1975; Lippe et al., 1985; Valverde and Silvertown, 1997; Tucker and Anand, 2003]. Using orthophoto sample points we constructed stochastic transition matrices, P. Rows represented the five cover classes for 1951 and columns represented cover classes for 1996. The entries, Pij, gave the fraction of points that changed from cover class i in 1951 to cover class j in 1996. Diagonal elements gave the fraction of each cover class that remained unchanged; supradiagonal elements gave increasingly woody transitions; and subdiagonal elements gave points less woody in 1996, interpreted as disturbance. The distribution of cover classes at time k, equation imagek, was then determined after n time periods by the matrix powers of P or equation imagek = equation imagek-nPn. In this study, the time interval was 45 a (1951 to 1996) and land class distributions for the years 2041 and 2086 were forecast assuming that transitions do not change over the next century.

4. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Site Description
  5. 3. Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions and Implications
  9. Acknowledgments
  10. References
  11. Supporting Information

4.1. Regional Patterns of Vegetative Change

[19] We constructed a transition table (Table 1), to highlight changes among cover classes. From 1951 to 1996 nearly one-third of all sample points (29%) increased in woodiness, over two-thirds (68%) remained the same, and less than 3% (2.8%) reverted to a less woody state. In particular, the number of points classified as either open woodland or closed canopy increased by nearly half (Figure 2). Points that were tundra in 1951 served as an entry state for shrub, open woodland, and closed-canopy in 1996. Shrubs increased from 28% in 1951 to 33% in 1996, with tundra representing the overwhelming entry state for new shrub points. Half (51%) of open woodland points in 1996 had been shrub and 21% had been tundra in 1951. The number of points classified as open woodland remained steady (7.6% of all points in 1951 vs. 7.7% in 1996), but with less than a third (27%) of the original open woodland remaining woodland in 1996; the remainder had become closed-canopy forest. Of the original sample points classified as closed-canopy points in 1951, 98% remained so in 1996; however, the total number of closed-canopy points nearly doubled (87% increase), with tundra (5%), shrub (10%), and open woodland (33%) acting as entry states for closed-canopy points. The distribution of vegetative classes was not independent of year (G = 50.68, df = 4, p = .001). Rates of change per decade (last row of Table 1) varied from a loss of −17.4% per decade for the unvegetated class to a gain of +14.0% per decade for the closed canopy class. Overall, the regional results indicate an increasingly woody forest-tundra ecotone in the western Kenai Mountains. The greatest absolute decrease in area was tundra and greatest absolute increase was shrub; the greatest relative decrease was unvegetated and greatest relative increase was closed-canopy forest.

image

Figure 2. Distribution of five cover classes by year for 978 random point samples taken in the Kenai Mountains (450 m asl) from black and white, aerial, georeferenced digital orthophotos.

Download figure to PowerPoint

Table 1. Transition Matrix for Forest-Tundra Ecotone Vegetation Transitions From 1951 to 1996a
Classification Year1996Total Points (1951)
UnvegetatedTundraShrubOpen WoodlandClosed Canopy
  • a

    Transition matrix gives cover classifications in 1951 as rows and in 1996 as columns. Entries refer to number of sample points (978 total) that were a given row cover class in 1951 and a given column cover class in 1996. Diagonal entries give number of points that remained unchanged. Supradiagonal entries give points that increased in woodiness, and subdiagonal points give points that decreased in woodiness from 1951 to 1996, as through disturbance. Row titled “Total” gives frequency distribution for 1996 (i.e., column totals); column titled “Total” gives frequency distribution for 1951 (i.e., row totals). Last row gives percent change per decade for each cover class calculated as r = 100 × ln (n1996/n1951)/4.5, where n1996 = number of sample points in a given cover class in 1996, n1951 = number of sample points in a given cover class in 1951, and 4.5 = the number of decades from 1951 to 1996.

1951Unvegetated252660057
 Tundra0343125168492
 Shrub1211963815271
 Open woodland020205274
 Closed canopy01018284
Total points (1996)2639332775157978
Rate of change, %/decade−17.4−5.0+4.2+0.3+14.0 

4.2. Timberline, Tree Line, and Shrub-Line Rise

[20] In method 1, nearly all closed-canopy forest patches identified as the highest on 205 altitudinal gradients were located at 400–900 m asl, as seen in the regional sample (Figure 3). In 1951 the mean of highest patches was 612 m asl and in 1996 it was 618 m asl. We observed no patch reduction or dieback of closed-canopy forest along the 205 randomly located altitudinal gradients. More than twice as many closed-canopy forest patches (69% of total) expanded upslope and/or down as patches that did not expand (31% of total). Closed-canopy forest patches on 31 of the gradients (15% of total) expanded down, but not upslope; 86 patches (42% of total) expanded upslope to some degree. New closed-canopy forest patches, present in 1996 but not 1951, made up 12% of the total sample. The mean difference in closed-canopy forest patch elevations across the 205 gradients was 5.7 m. New patches were 35 m higher (s.e. = 6.1 m) in elevation on average than the 1951 patches along the same gradient, a rise of 7.8 m/decade for these 24 gradients. The mean elevation (mean ± s.e.) of new closed-canopy forest patches (567 ± 18.2 m asl, n = 24) was lower than the mean of patches that did not change in size (656 ± 15.0 m asl, n = 64). This detailed look at closed-canopy patches suggests that timberline rose on average ∼1.5 m/decade, but that new patches could establish more rapidly (∼5–10 m/decade) across 10% of the landscape at mostly lower elevations.

image

Figure 3. Frequency distributions for elevation of 978 sample points classified on georeferenced digital orthophotos into one of five cover classes in (left) 1951 and (right) 1996. All elevations were taken from digital elevation models and then binned in 50 m intervals.

Download figure to PowerPoint

[21] Shrub patches also did not retreat nor dieback, with half (50%) showing noticeable expansion along the altitudinal gradients. In contrast to the closed-canopy forest patches, we detected no new uppermost patches. The mean difference in shrub patch elevation was <1 m.

[22] In method 2, open woodland did not increase in total area (Table 1) but did shift noticeably upward in its altitudinal distribution (Figure 3). Shrubs both increased in abundance (Table 1) and became more common at higher elevation (Figure 3); however, the data do not provide an accurate account of shrub-line, as we did not capture the upper tail of the distribution. Tundra decreased in area (Table 1) and the lower, truncated end of the distribution (Q25) shifted upward from 1951 (Q25 = 632 m asl) to 1996 (Q25 = 671 m asl). Using the highest quartile elevations (Table 2) to estimate changes in timberline, tree line and shrub-line from 1951 to 1996 suggested that shrub-line rose by 59 m (13 m/decade), tree line rose 49 m (11 m/decade), and timberline 6 m (∼1.3 m/decade). Using the difference in Q25 for tundra in 1996 and 1951 implied that tundra retreated upward 39 m (8 m/decade). Taken together, Figure 3 and Table 2 imply that while the forest-tundra ecotone is shifting upward, its constituent parts are doing so at differing rates, with shrub-line likely moving most rapidly, then tree line, then tundra retreat, and timberline moving slowest.

Table 2. Extreme Quartile Elevations for Four Vegetation Classes in 1951 and 1996a
Vegetation Class1951 Upper Quartile (Q75) Elevation, m asl1996 Upper Quartile (Q75) Elevation, m asl
Closed canopy550556
Open woodland570619
Tall shrub755814
 1951 Lower Quartile (Q25) Elevation, m asl1996 Lower Quartile (Q25) Elevation, m asl
  • a

    Here asl, above sea level.

Tundra632671

4.3. Influence of Slope, Aspect, and Elevation at the Regional Scale

[23] Change in woodiness was independent of slope category (flat, moderate, steep; G = 4.33, df = 4, p = 0.363), suggesting that slope was not important in determining change in woodiness regionally. Change in cover class from 1951 to 1996 varied with aspect. Sample points that increased in woodiness tended to be more frequent on northerly aspects (N, NE, NW) and less frequent on E, W, and S slopes than expected by chance (Figure 4). Unchanging points tended to fall on W facing slopes more than expected. Change in woodiness was not independent of aspect (G = 32.2, df = 14, p = .004).

image

Figure 4. Deviation from expected counts (measured in percent deviation from an expected, independent distribution) by aspect for sample points that increased in woodiness (n = 286) from 1951 to 1996. Aspect derived from NED digital elevation models.

Download figure to PowerPoint

[24] Changes in the structure of forest-tundra ecotone from 1951 to 1996 also varied by elevation with most observed changes below 750 m asl. Sample points that increased in woodiness were significantly (Mann-Whitney U = 77787, p < .001) lower in elevation (median = 618 m asl, n = 286) than those that did not change. For points that did not change in woodiness, the median elevation (700 m asl) was identical (one-sample t-test, t = 1.062, df = 971, p = .14) to the mean of all regional sample point elevations (701 m asl). The few sample points that decreased in woodiness were higher than (median = 784 m asl, n = 26), but not significantly different from (Mann-Whitney U = 8073, p = .56), the elevation of unchanged points.

[25] Cool aspects (NW, N, NE) increased more in woodiness than warm aspects (W, SW, S, SE, E) but mostly below 750 m asl (Figure 5). On both cool and warm aspects, closed-canopy forest nearly doubled in density (measured as points per area; Table 3), but increased little in elevation. As with the closed-canopy study along the 205 altitudinal gradients, there was little change in density of patches above 650 m asl. In contrast, open woodland increased both in density and elevation (∼50 m) on cool aspects; on warm aspects open woodland increased in density below 550 m asl but did not increase in elevation. On cool aspects shrub density decreased below ∼650 m asl, corresponding to the increase in density and elevation of open woodland, an important exit state for shrubs on cool aspects, whereas shrub density increased between 700 and 1000 m asl. On warm aspects at 550–1000 m asl, shrub density increased. Tundra on cool slopes decreased substantially below 700 m asl and on warm slopes tundra decreased mostly below 850 m asl. Neither aspect lost tundra above ∼1000 m asl. In summary, there was evidence for upward movement of the forest-tundra ecotone on cool, northern exposures (N, NW, NE), less so or none for warm, nonnorthern exposures (S, SE, SW, E, W).

image

Figure 5. Distribution of five cover classes by elevation and years for points sampled from (left) cool aspects (N, NW, NE) and (right) warm aspects (S, SW, SE, E, W). Because there were more warm aspect samples (n = 580) than cool aspect samples (n = 398), both were scaled to 1000 sample points.

Download figure to PowerPoint

Table 3. Transition Matrices by Aspecta
AspectClassification Year1996Total Points (1951)
UnvegetatedTundraShrubOpen WoodlandClosed-canopy
  • a

    See Table 2 for details of interpretation.

N, NE, NW1951Unvegetated101810029
  Tundra013153122198
  Shrub1764227101
  Open woodland01062532
  Closed-canopy00003838
 Total points (1996) 111571184072398
 Rate of change, %/decade −21.5%−5.2%+3.5%+5.0%+14.2% 
E, SE, S, SW, W1951Unvegetated15850028
  Tundra02127246294
  Shrub014132168170
  Open woodland010142742
  Closed-canopy01014446
 Total points (1996) 152362093585580
 Rate of change, %/decade −13.9%−4.9%+4.6%−4.1%+13.6% 

[26] Cover class transitions also differed between the warm and cool aspects, with qualitatively different rates of change per decade for the unvegetated and open woodland classes (Table 3). Overall, the entry states of closed-canopy points on cool aspects were essentially the same as on warm aspects, with the exception that tundra as an entry state for closed-canopy forest was slightly more important and open woodland slightly less important on warm aspects. Tundra was nearly three times as important as an entry state of open woodland points on cool aspects as on warm aspects, with 30% of all 1996 woodland points on cool aspects having been tundra points in 1951, compared to 11% on warm aspects. Shrub was also a more important entry state (55%) of woodland points on cool aspects than on warm aspects (45%). Only 15% of open woodland points were self-replacing on cool aspects, compared to 40% on warm aspects. The primary entry state of shrubs on both aspects was self-replacement; however, tundra accounted for 45% of 1996 shrub points on cool aspects and 34% on warm aspects. Warm aspects also appeared to have greater disturbance, with more nonzero subdiagonal elements than cool aspects.

[27] Taken together, the elevation distributions of cover classes and their transitions show that the forest-tundra ecotone on cooler aspects shifted upward ∼50 m. Much of the difference in woodiness between cool and warm aspects resulted from less self-replacement of all cover classes (other than closed-canopy forest), implying that warm aspects were more stable in their structure than cool aspects. Also important was the frequent replacement of tundra by open woodland and shrub cover on cool aspects, a dynamic less pronounced or absent on warm aspects. In summary, on cooler slopes tundra played a more important entry role for open woodland and open woodland played a more important exit role for both tundra and shrub than on warmer slopes. This result was also seen in the local, intensive studies.

4.4. Local Studies

[28] Over the period 1951 to 1996 the three local study sites known to have substantial changes (Figure 6) showed dramatic decreases in tundra and increases in shrub and closed canopy, with a relatively constant open woodland. Across all local sites, tundra was the primary entry state both for new shrub points and for new open woodland points, providing an average of 33% of the open woodland points, in contrast to the regional entry role of tundra providing 21% of the open woodland points. The amount of shrub replaced by open woodland was about a third, varying from 16–38% among the three sites, greater than the regional average of 14% (Table 1) and nearer the cool slope value of 22% (Table 3). Shrub was the primary exit state for tundra points with almost half of the tundra replaced by shrub (30–54% among the three sites), greater than that seen regionally (25%; Table 1) or on cooler slopes (27%; Table 3). Generally about half of open woodland (41–59% among the three sites) was replaced by closed-canopy. As seen regionally, the most stable class was closed-canopy, with greater than 95% of all closed-canopy forest sample points in the 1951 photos remaining so in the 1996 photos. Changes in vegetative cover were well mixed spatially across each local area (Figure 6), displaying an outward creeping and infilling by woody classes. In summary, these local sites were similar to the cooler slope points, in that tundra, as well as shrub, was an important entry state for open woodland. However, the local sites stood out from the cooler slopes in the large amount of tundra lost to shrub. Table 4 indicates that all three local sites had NE or NW aspects and that two of the three were lower than the mean regional point elevation and all were below the 1951 Q75 elevation representing the 1951 shrub-line (Table 2). In summary, the aspect and elevation of these substantially changed areas were consistent with the overall regional result that showed an increase in woodiness.

image

Figure 6. Location of cover classes at three local study sites in the Kenai Mountains draped over digital elevation models. (left) Cover classes in 1951 (matrix_50). (right) Cover classes in 1996 (matrix_96). Red, tundra; light green, shrub; olive green, open woodland; black, closed canopy. Pixel sizes are 30 m. Longitude and latitude are in UTM m; elevation is in m asl.

Download figure to PowerPoint

Table 4. Physical Characteristics for Local and Regional Study Areas
SiteLocation Centroid, degSlope, deg, Mean, ±s.d.Aspect, deg, Mean, ±s.d.Elevation, m, MeanElevation, m, RangeW,a %ΔS,b WH/m2
  • a

    Percent of landscape increasing in woodiness from 1951 to 1996.

  • b

    Difference between median insolation values for unchanging and increasingly woody points, summed over equinox, summer, and winter solstices.

Skilak60.362°N1855465321–67958171
River150.132°W±6±17    
Russian60.457°N19305586426–73040184
Mountain150.078°W±6±14    
Mystery60.520°N1755729584–9073924
Hills150.133°W±4±51    
Regional60.306°N13288700450–11002958
study150.357°W±10±94    

4.5. Solar Radiation

[29] Insolation values varied by date and location. Generally, median winter solstice values ranged as 2–4 WH/m2, equinox values as 1.0–1.5 × 103WH/m2, and summer solstice values as 4.5–5.5 × 103WH/m2. At both the local and regional scale, the median insolation values for sample points that increased in woodiness from 1951 to 1996 were slightly but consistently lower than insolation values for sample points that remained unchanged across the three seasons (summer and winter solstices, and equinox). The difference in summed insolation values between unchanged and increasingly woody points (Table 4) was greatest at the midelevation, NW facing Russian Mountain site, a site intermediate in the degree of change among vegetative classes, followed by the lowest elevation site, NE facing Skilak River. The relatively high, NE facing local sample site, Mystery Hills, showed the least difference between irradiation values. We applied Kolmogorov-Smirnov tests to compare differences in solar radiation values between those points that increased in woodiness and those that did not change among the dozen site-by-season combinations [(3 local sites + 1 regional site) × 3 seasons]. Of the twelve combinations, six were significant (Bonferroni corrected to p ≤ .001). Those that were not significant included all seasons regionally (.01 ≤ p ≤ .08) and all seasons at Mystery Hills (.02 ≤ p ≤ .10); summer solstice values were especially similar (p ≥ .08 for both sites). A key finding is that insolation and aspect results support the observation that cooler slopes showed greater increase in woodiness than warmer slopes.

4.6. Field Measures of Age, Elevation, and Abundance of Recruits

[30] Only short seedlings were present in the belt-transects sampled; there were no trees >1 m tall or layered krummholz. The nearest older trees, mostly mountain hemlock (Tsuga mertensiana) krummholz, grew at 650–800 m asl, usually more than 100 m below the belt-transects. Much of the krummholz at 650–800 m asl showed recent flagging. Densities of seedling trees observed in the field were similar for hemlock (3.3 seedlings/ha) and spruce (3.4 seedlings/ha). The mean elevation of white spruce (Picea glauca) recruits (901 ± 6.1 m, n = 70) and hemlock recruits (904 ± 10.5 m, n = 43) were higher in elevation than the highest closed canopy forest patches located along altitudinal gradients (Kolmogorov Smirnof Z = 7.668, p < .001). Recruit density increased steeply with elevation for both species from 790 m, reaching a peak at 900 m, then decreasing more gently to 1100 m asl. Recruits ranged from 10 to 51 a old and represented the only trees at these elevations at these study sites. Hemlock recruits were marginally older (23.5 ± 1.3 y, n = 25 individuals) than spruce recruits (20.2 ± 1.3 y, n = 38 individuals).

[31] Figure 7 shows the distribution of seedling recruits by decade and elevation in comparison to growing season temperatures in Kenai, Anchorage, and Homer, and the Pacific Decadal Oscillation index for May–August (PDO [Mantua et al., 1997]) over the last six decades. There was a weak, positive rank correlation between seedling age and elevation (rs = .293, p = .010, n = 63). No recruits were found from the coldest decade, 1945–1955. The oldest recruits established during the second coldest decade and grew at ∼900 m asl. Recruits from 1965–1975, a noticeably warmer decade than previously, established between 850–950 m, while recruits from 1975–1985, a still warmer decade, colonized 800–1000 m asl. From 1985 to 1995, the second warmest decade but one of relatively high water balance (Figure 1b), recruits established at all elevations sampled and represented the most numerous cohort. The last, hottest, and driest decade (Figure 1b), 1995–2005, showed a sharp reduction in recruits that could be due to a sampling failure to find very small recruits. Nevertheless, these results are consistent with tree establishment that is sensitive both to temperature and to moisture stress.

image

Figure 7. Distribution of recruits by decade of recruitment and elevation (m asl) for individual trees (n = 63 recruits aged by cross-section of stems) in the Kenai Mountains above 800 m asl compared to mean summer temperatures for the cities of Homer, Kenai, and Anchorage, Alaska, and the Pacific Decadal Oscillation Index averaged for May–August temperatures. Shaded region shows time periods with years with above average values.

Download figure to PowerPoint

4.7. Future Projections of Alpine Vegetation by Slope Aspect

[32] We used transition matrices based on Table 3 to construct Markov models of stochastic matrices P, to estimate future distributions of cover classes, equation imagek, for the years 2041 and 2086, as equation imaget+n = equation imagetPn (Table 5). While the distributions for both aspects were similar in 1951 (G-test for independence of aspect and 1951 cover class, G = 4.68, df = 4, p = 0.321), the distributions diverged markedly over the next three time steps. Most dramatic was the increase in closed-canopy forest, which tripled on cool slopes but only doubled on warm slopes from 1951 to 2086. Shrub density increased on both aspects, then decreased on cool aspects and leveled off on warm aspects. The fraction of the landscape classified as open woodland changed little from the 1950s to the end of the 21st century for both aspects. According to the model, by the end of the 21st century, cool aspects are likely to be more than half tree-covered, while warm aspects will be about one-third tree covered, the difference in cover being principally shrub. These model results are suggestive at best, based on an assumption of unchanging transition probabilities. Their principle purpose is to highlight the differences between warm and cool aspects as their transitions compound into the future.

Table 5. Present and Future Forest-Ecotone Structure by Slope Aspecta
YearUnvegetatedTundraShrubOpen WoodlandClosed-Canopy
  • a

    Table entries give percent in each cover class by year. Future estimates determined using Markov model for one (2041) and two time steps (2086).

Warm Aspects
19514.850.729.37.27.9
19962.640.736.06.014.7
20411.433.538.46.020.5
20860.828.038.07.126.0
 
Cool Aspects
19517.349.725.48.09.5
19962.839.429.610.118.1
20411.330.129.410.828.4
20860.723.126.810.339.1

5. Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Site Description
  5. 3. Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions and Implications
  9. Acknowledgments
  10. References
  11. Supporting Information

[33] Twenty-nine percent of the area covered by forest-tundra ecotone in the west Kenai Mountains of southcentral Alaska increased in woodiness from 1951 to 1996. These changes, when compared to the PDO, occurred during a period of warming and drying across the Kenai Peninsula. Cool aspects changed more dramatically than warm aspects. Both aspects showed a decrease in tundra below 1000 m asl with corresponding increases in shrub cover above 700 m asl; however, below 650 m asl, cool slopes decreased in shrub as well as tundra cover, with open woodland and closed-canopy forest increasing markedly. On both aspects, tundra cover decreased from 50% of the area in 1951 to 40% in 1996, while the amount of area covered by closed-canopy forest nearly doubled. Because both shrub and tundra cover were an entry state for new tree-covered areas on cool aspects, cool slopes increased more in tree-cover than warm slopes, where shrub was the primary entry state for new open woodland cover. While the density of trees increased on both aspects, only cool slopes showed evidence of an upwardly moving forest-tundra ecotone, including tree line, of about 10 m/decade, similar to the rate theorized by Grace et al. [2002] and near that observed on south-facing slopes for spruce in southwestern Yukon [Danby and Hik, 2007]. Local residents of Homer and Kachemak Bay have also observed that tree line has visibly risen “at least several hundred feet” since the 1940s (Y. Kilcher, personal communication, 1997). Timberline, however, did not appear to move more than ∼5 m over the 4.5 decades. Taken together the data suggest that on cool aspects, the forest-tundra ecotone is rising, but on warm aspects it is becoming denser with trees and shrubs. Regional shrub cover increased from 28% to 33%, a change in shrub cover of 6% and a relative change in shrub cover of 21%. These observations are similar in absolute change, but somewhat lower in relative change, to that reported by Tape et al. [2006] for northern Alaska shrubs on arctic tundra slopes. There, shrubs increased by 5–8% absolutely and 33–36% relatively on low hillsides. Finally, we observed recruitment by white spruce and mountain hemlock seedlings at 100 m and higher above flagging krummholz hemlock in the Kenai Mountains at 40–45 km from the coast. The decades of establishment of these seedlings parallel warming and drying trends recorded over a half century of weather records from three stations spanning 200 km and related to the PDO. The three locations differ in their mean summer temperatures by up to 2.5 C, yet show consistent periods of above and below mean summer temperatures, indicating a region-wide trend. These results differ from a study of northern Montana tree establishment at tree line >600 km from the coast, which showed that establishment was not correlated with PDO [Alftine et al., 2004].

[34] Together, the field and remotely sensed results reported here are consistent with woody vegetation responding to an increase in temperature that is also sensitive to drought stress [Barber et al., 2000; Payette et al., 2001; Holtmeier and Broll, 2005; Wilmking et al., 2004]. Payette et al. [2001] present a model for advancing tree line in the forest-tundra ecotone of eastern Canada. Their model captures the essence of a nonalpine tree line, advancing through the production of erect trees by relictual, stunted individuals embedded in a fire-dominated landscape. Payette et al. [2001] review a number of studies showing that white spruce in Canada have increased in density without any significant displacement of the arctic tree line. In the Kenai Mountains above 450 m asl, the forest-tundra ecotone is an alpine one, where fire plays a minor role, and white spruce and mountain hemlock are both increasing in density and germinating as new seedlings far above krummholz. The rising alpine forest-tundra ecotone on cool slopes is consistent with predictions of Payette et al. and Holtmeier and Broll [2005] under a warmer and more humid climate change scenario. Under those conditions, as would be seen in the Kenai Mountains at lower elevations on cooler aspects, soil moisture and proximity to lower elevation seed sources leads to an advancing local tree line. There are multiple line of evidence from this study supporting an advance of tree line as a response to a warmer, more humid climate on cool aspects due to biophysical factors: (1) Cooler aspect (N, NW, NE) soils are more mesic, because of less evaporation than warmer aspects (S, SE, SW, W, S). (2) The Kenai lacks permafrost, increasing soil moisture availability during the growing season. (3) Lower elevations increased in woodiness more than higher elevations, due in part to closer proximity to seed sources. (4) Seedling cohorts found at ∼100 m above krummholz were most numerous in the third decade of regional warming. They were least numerous in the most recent, hottest, and driest decade, perhaps too dry for recruitment [Gedalof and Smith, 2001], as hypothesized for white spruce's recent failure to recruit higher in the warming interior of Alaska [Barber et al., 2000; Wilmking et al., 2004]. However, this latest cohort was also the shortest in height, and so most difficult to thoroughly sample. (5) Within the intensively studied local sites, we found that the greatest increases in woodiness there were consistent with patterns found across the region. Namely, the cooler, lower elevation aspects nearest seed sources changed the most, explaining much of the within and among site variation in the documented changes. (6) Wherever trees grew, their density increased, whereas shrubs actually decreased on cooler slopes below 650 m and increased elsewhere up to ∼1000 m asl. (7) Tree line rose only on cooler aspects, ∼50 m vertically in less than 50 a.

[35] In summary, we interpret the changes observed in the Kenai Mountains forest-tundra ecotone to behave similarly to the conceptual model advanced by Holtmeier and Broll [2005] in their review of tree lines and environmental change. Holtmeier and Broll advocate that detecting tree line changes in response to climate change at the regional and smaller scale requires a comprehensive approach. In this study we have documented the climatic character with seasonal and interannual variability in moisture and temperature; investigated altitudinal gradients; modeled insolation using sunshine hours, inclination, aspect and exposure to Sun; identified seed sources at lower elevations; discounted anthropogenic local effects, insect pests and browsing on conifers at tree line; and determined that soil temperature is above freezing for part of the growing season. As described by Holtmeier and Broll, a climatic tree line within a continental climate will rise under increasing warmth and humidity, but will remain static, or even retreat, under an increasingly warmer and drier climate. At the local and regional scale, the effects of topography and site conditions can negate the effects of global warming, as seen in the differences between cool and warm aspects here. Moreover, we observed changing tree physiognomy and a half-century of seed-based recruitment and successful establishment of trees above krummholz, recruitment that is coincident with region-wide warming. Whether these seedlings ultimately advance forest is difficult to assess, as further warming and drying may lead to mortality and reduction in growth [e.g., Wilmking et al., 2004] or shift the fire regime. Elsewhere in Alaska Lloyd and Fastie [2002] document tree line change but identify that rising temperatures do not necessitate rising tree line.

[36] The results from the Kenai strengthen dendrochronology that shows a significant and fundamental change in the climate of Alaska and the Yukon since the 1970s [Barber et al., 2004; Lloyd and Fastie, 2002; Danby and Hik, 2007]. Increasingly woody landscapes in Alaska and nearby Yukon have been observed in low elevation settings across the north slope of the Brooks Range [Sturm et al., 2001; Tape et al., 2006], the Seward Peninsula [Suarez et al., 1999; Silapaswan et al., 2001; Lloyd et al., 2003], the Kenai lowlands [Klein et al., 2005], and elsewhere in Alaska [Riordan et al., 2006]. Changes in alpine settings have been documented across the interior of the state [Lloyd, 2005] and the Yukon [Danby and Hik, 2007]. The results we present here fill an important geographic gap in the understanding of vegetative change in Alaska.

[37] The data also support the idea that tree line rises as a nonlinear, infilling phenomenon [Barber et al., 2000; Payette et al., 2001; Grace et al., 2002; Sveinbjornsson et al., 2002; Lloyd and Fastie, 2002; Holtmeier and Broll, 2005; Zeng and Malanson, 2006]. In a series of recent papers, Malanson and coworkers have explored the details of positive feedback loops at tree line in northern Montana, both empirically [Alftine and Malanson, 2004; Alftine et al., 2004; Resler et al., 2005; Butler et al., 2004] and theoretically [Zeng and Malanson, 2006]. They propose that an endogeneous feedback process alone can lead to infilling and spread of trees in the forest-tundra ecotone, making separation of that process from climate response difficult. However, in the current study, because of the clear difference between cool and warm aspects, the difference in the pace of infilling and the rapidity of tree line rise on cool slopes, an endogeneous patch-filling process is unlikely to account for what we observe as a tree line rise of 10 m/decade. More likely, facilitating feedback [Zeng and Malanson, 2006] accounts for the timberline rise of ∼1 m/decade and the infilling of closed canopy patches that were formerly open woodland. As for shrub expansion, Tape et al. [2006] present a simple yet robust, logistic patch model that shows two facets of a landscape-wide expansion: (1) because of habitat heterogeneity there should be a wide range of changes during expansion, such that at no time will patches be expanding everywhere at detectable levels and (2) there should be no place where patches are declining. Tape et al. [2006] used the model to argue for a pan-arctic shrub expansion, because both facets were satisfied. In the case of the Kenai forest-tundra ecotone, there were elevations and aspects where shrubs were declining; however, their model may apply to tree expansion. Deciding among patch models is beyond the scope of this report on forest tundra ecotone changes, but it is quite likely that both models apply to tree patches but not shrub patches below 700 m asl and both apply to shrub patches 700–1000 m asl.

6. Conclusions and Implications

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Site Description
  5. 3. Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions and Implications
  9. Acknowledgments
  10. References
  11. Supporting Information

[38] The variable changes seen in the Kenai Mountains illustrate how forest ecotone change is contingent on biophysical processes that vary both by region and by landscape, such that rapid rise in tree line can occur near to slopes with no tree line rise. While warming alone appears to be sufficient to increase shrub cover and tree density, mesic slopes and warming are likely necessary for rising tree line on the Kenai Peninsula. Together with evidence for vegetation change concurrent with climatic change in the Kenai lowlands [Klein et al., 2005], a warming and drying climate is leading to a woodier landscape all across the western Kenai Peninsula. Such changes will likely lead to concurrent changes for animals using these landscapes, such as moose and wetland birds in the lowlands, and Dall sheep, caribou, and wolverine in the mountains. If the next half century shows climate change in the same direction and magnitude as the last half century, there will likely be even more substantial biotic change than documented here.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Site Description
  5. 3. Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions and Implications
  9. Acknowledgments
  10. References
  11. Supporting Information

[39] Funding for this study was provided by the U.S. Fish and Wildlife Service at the Kenai National Wildlife Refuge and by Alaska Pacific University. C. Tobin and M. Loso provided manuscript advice, and the research owes much to the orthorectification and scanning of air photos by A. DeVolder, formerly with the Kenai Peninsula Borough Spruce Bark Beetle Task Force. Two anonymous reviewers contributed critically to the improvement of the paper. Much of the data were collected by K. Timm and A. McMahon during thesis projects at Alaska Pacific University.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Site Description
  5. 3. Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions and Implications
  9. Acknowledgments
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Site Description
  5. 3. Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions and Implications
  9. Acknowledgments
  10. References
  11. Supporting Information
FilenameFormatSizeDescription
jgrg254-sup-0001-t01.txtplain text document1KTab-delimited Table 1.
jgrg254-sup-0002-t02.txtplain text document0KTab-delimited Table 2.
jgrg254-sup-0003-t03.txtplain text document1KTab-delimited Table 3.
jgrg254-sup-0004-t04.txtplain text document1KTab-delimited Table 4.
jgrg254-sup-0005-t05.txtplain text document1KTab-delimited Table 5.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.