STUDY AREA

Moving from north to south along an ∼1000 km latitudinal gradient, our study area encompasses four approximately linear mountain ranges that form the easternmost extent of upper tree-line environments in the Rocky Mountains from ∼44°-35°N: (1) Bighorn, (2) Medicine Bow, (3) Front Range, and (4) Sangre de Cristo (Fig. 1). This latitudinal gradient bisects a precipitation dipole at approximately 40°N, where opposite moisture conditions often exist to the north and south of this parallel, especially during El Niño events that normally cause above-average precipitation below 40°N and dry conditions above (Dettinger et al., 1998). This boundary coincides with differences in synoptic climatology as well, with more frequent intrusions of maritime Pacific air masses to the north of 40° in the winter and a stronger summer monsoon influence to the south (Mitchell, 1976). For this reason, we classify mountain ranges north of 40° as the Central Rocky Mountains (Bighorn and Medicine Bow) and ranges to the south as the Southern Rocky Mountains (Front Range and Sangre de Cristo).

TABLE 1

Study site characteristics.

The floristic composition of upper treeline ecotones is codominated by Engelmann spruce (Picea engelmannii) and sub-alpine fir (Abies lasiocarpa) along the entire latitudinal gradient. Higher proportions of spruce exist in the south and fir becomes more dominant toward the north, particularly in the Bighorn Mountains. Other tree species include Rocky Mountain bristlecone pine (Pinus aristata) on south-facing slopes in the Sangre de Cristo and Front Range mountains and lodgepole pine (P. contorta) on more xeric sites in the Bighorn, Medicine Bow, and northern Front Range mountains.

HYDROCLIMATOLOGY

We used the Precipitation-elevation Regressions on Independent Slopes Model (PRISM) climate data website (PRISM Group, Oregon State University,  http://www.prismclimate.org, created 1 June 2010) to access monthly climate data (temperature and precipitation) for each mountain peak in the study area. These data were selected to compensate for the dearth of available high-elevation weather stations and because they take into account the dominant role of elevation and topography in dictating climatic conditions throughout mountain environments (Daly et al., 2008). Given our emphasis on ecological changes at upper treeline within the context of the climatic water budget, we focused on the main climate inputs for high-elevation areas of the Rocky Mountains, including mean annual air temperature (MAT) and cool season (November–April) precipitation (Table 1, Fig. 2). Broad geographic patterns of these two climate variables reveal more uniform conditions among mountain peaks in the Bighorn and Sangre de Cristo Mountains, whereas more pronounced variability exists within the Front Range and especially the Medicine Bow Mountains (Fig. 2). Not surprisingly, the same general spatial patterns emerge when assessing how these inputs combine to form the climatic water deficit, with the notable trend of increased aridity along a south-north gradient (excluding Kannaday Peak [KP], Fig. 2).

Cool season precipitation represents a conservative estimate of total snowfall accumulation and makes up a majority of moisture delivered to upper treeline environments in the Central Rocky Mountains (55.8%; Table 1). In contrast, it only accounts for 43.9% of annual precipitation in the Southern Rockies (Table 1), with the lowest percentages in the Sangre de Cristos where monsoon-derived rainfall (July–September) surpasses monthly snow totals accrued during the cool season (Elliott, 2011). This explains why, despite receiving roughly equal amounts of cool season precipitation, mountain peaks in the Bighorns typically experience more severe climatic water deficits during the summer growing season (Fig. 2). Furthermore, vegetation in the Bighorns rely almost exclusively on isolated convective storms during the growing season (e.g., rather than the Southwest Monsoon), which generally fail to replenish evaporative demand.

FIGURE 2.

Box-and-whisker plots of 20th-century (1900–1999) cool season precipitation (November–April), mean annual temperature, and climatic water deficit for each site arranged along a north-south (left-right) gradient. Diamonds represent values outside the interquartile range (e.g., outliers). All precipitation and temperature data are from PRISM Group, Oregon State University ( http://www.prism.oregonstate.edu).

Geographic patterns of MAT generally correspond to latitudinal position (excluding KP), with the warmest temperatures in the Sangre de Cristo Mountains (mean = 1.8 °C) at the southern end of the transect and lowest on the northern end in the Bighorns (mean = -0.01 °C; Table 1). Temperature trends during the 20th century indicate widespread warming throughout the study area with minimum temperatures consistently >0.5 °C above the century-long mean (1900–1999) from ca. 1940–2007 (e.g., Elliott, 2012a). Maximum temperature, on the other hand, peaked in the mid-1950s and subsequently decreased during the latter half of the 20th century (Elliott, 2012a). From a long-term perspective, recent temperature increases in the Bighorn and Medicine Bow Mountains equal the intense warmth experienced during the early- to mid-Holocene (11,000–6000 years before A.D. 1950) when summer insolation was greater than present (Shuman, 2012).

FIELD METHODS

Field data for this study were collected at the same time and followed the same systematic design described by Elliott (2011), which essentially focused on locating mountain peaks with climatic treeline boundaries (see Holtmeier and Broll, 2005; Butler et al., 2007) on both north- and south-facing slopes (azimuths 135° to 225° and 315° to 45°, respectively). In total, eleven mountain peaks were sampled and at each site (n = 22) nested-belt transects were placed through the entire eco tone boundary on contrasting north and south aspects, extending from the outpost free (term after Paulsen et al., 2000) downslope to below timberline. Here treeline is defined as the variable-length transition zone between the uppermost limit of individual trees with an upright growth form and timberline, the elevational limit of closed-canopy forest. The length of each transect varied depending on the upward position of the outpost tree and transect width was divided into two parts to ensure an adequate number of saplings to calculate age corrections and analyze regeneration patterns. Total transect width above timberline was 40 m. and below timberline, the width was reduced to 20 m to accommodate for the general increases in tree density. For illustrations of this transect design, please refer to Elliott and Kipfmueller (2010). Krummholz were not sampled because of the different microclimatic influences on establishment compared to upright trees (Holtmeier, 2009).

DENDROECOLOGICAL DATA

We collected age-structure data by extracting two increment cores at 30 cm above the ground from all living trees (≥5 cm diameter at breast height [dbh]) along the transect. Every sapling (<5 cm dbh, ≥1.2 cm diameter at ground level [dgl]) within the transect was harvested at ground level. Sections were taken from 10–15 saplings growing in both open and closed-canopy environments to determine a correction factor for age-at-coring height for trees above and below timberline, respectively (cf. Veblen, 1992). Seedlings (<1.2 cm dgl) were inventoried as alive or dead throughout the transect and excluded from the age-structure data.

Tree cores and sapling cross sections were prepared following standard procedures for dendrochronology (Stokes and Smiley, 1996). Pith estimators were used to geometrically determine the number of missed rings when the pith was not obtained during field sampling (Applequist, 1958). Dates of tree establishment were calculated based on tree age at the time of sampling (summer of 2007 or 2008) and by adding the appropriate age to coring-height correction. Age to coring-height corrections were calculated by taking the mean age at 30 cm of harvested saplings and adding it to the age of the cross-dated individual tree cores collected from the same height. Age corrections were further stratified by species, mountain range, and position relative to timberline to account for the different growth rates of saplings in open (e.g., above timberline; mean = 19.5 yr) versus more closed-canopy conditions (e.g., below timberline; mean = 32.3 yr). All cross-dating was achieved by using the visual list method proposed by Yamaguchi (1991). Tree establishment dates were grouped into five-year intervals for the period 1900–2000. Further details of age to coring-height calculations can be found in Elliott (2012a).

CLIMATIC WATER DEFICIT

To assess the role of temperature-precipitation interactions via the climatic water deficit on regeneration dynamics at upper treeline during the 20th century, we used a Thornthwaite-based method to model the climatic water budget (Thornthwaite and Mather, 1955; Hamon, 1963; Dingman, 2002). Following Stephenson (1998), climatic water deficit represents the limits placed on evapotranspiration by water availability and thus renders it a suitable proxy for the length and severity of absolute drought conditions. Inputs for this model include latitude, mean monthly temperature and precipitation, and soil water holding capacity. Monthly temperature and precipitation data inputs were obtained from the PRISM website ( http://www.prismclimate.org). Soil available water capacity (AWC) data were obtained for sites in the Southern Rocky Mountains through the Natural Resources Conservation Service (NRCS) Web Soil Survey ( http://websoilsurvey.nrcs.usda.gov/app/HomePage.htm). Data are incomplete or unavailable for our study sites in the Medicine Bow and Bighorn mountain ranges. Yet, given that water stress is common at upper treeline because of generally thin soils with a low moisture-holding capacity (Sveinbjömsson, 2000), we compensated for this by assigning an AWC of 75 mm for these sites. This value equals half the saturation level of 150 mm typical of more well-developed soils (cf. Cowell and Urban, 2010). In the case of different AWC values between soils above and below timberline, we used the mean. Precise climatic water deficit calculations for each site rather than each mountain peak were not possible given the spatial resolution of PRSIM data (4 km resolution), which typically meant that both aspects were contained within a single climate grid.

Climatic water deficit data are produced in the form of a continuous monthly time series (e.g., 1900–2000), along with a suite of other climatic water budget variables (e.g., potential evapotranspiration [PET] and actual evapotranspiration [AET]) that are influenced by both current and antecedent soil moisture conditions. PET is a measure of the water that could be lost to the atmosphere, assuming the available heat energy (temperature), and AET is the actual moisture lost to the atmosphere from soil and vegetation (Cowell and Urban, 2010). An advantage of the Thornthwaite-based approach is the limited input requirements in comparison to the robust output, whereas drawbacks stem from the exclusion of local-scale data such as wind, humidity, and land cover (Stephenson, 1998). These data, however, are typically unavailable for high-elevation treeline sites and this method is arguably the most appropriate for examining broad regions (cf. Vörösmarty et al., 1998).

DATA ANALYSIS

Of central importance to this study is classifying temporal variations in the climatic water deficit as wet or dry during the 20th century. To do this, we standardized annual deviations from the 20th-century mean (1900–1999) and averaged these to create z-score pentads in order to match the minimum temporal resolution of the age-structure data. The rationale for this was to create five-year periods of relatively wet (+ anomaly) or dry (- anomaly) soil moisture conditions to examine how given deficit regimes (e.g., wet or dry) align with regime shifts in tree establishment and associated upper treeline advance on contrasting slope aspects.

According to Williams et al. (2011), a “phenomenalogical” regime shift takes place when a variable of a system changes abruptly compared to rates of past change. This relatively broad definition permits the statistical verification of regime-shifts by identifying a change point in time when sudden changes happen to key ecological variables (Anderson et al., 2009; Williams et al., 2011), such as the rate of tree establishment. Thus, for the purposes of this research, we used regime-shift analysis as a tool to quantify statistically significant changes in the rate of tree establishment rather than for arguing about the potential unprecedented nature of population-level controls on tree establishment. This is largely due to the fact that our age-structure data are confined to the 20th century and therefore unable to justify the uniqueness of these changes at the centennial scale. To test for regime-shift changes in the rate of tree establishment, we used a sequential algorithm developed by Rodionov (2004). This is a data-driven method where an a priori hypothesis on the timing of regime shifts is not needed and sequential t-tests are conducted on time series data to detect regime-shift changes (Rodionov, 2004). A regime-shift change is identified when the cumulative sum of normalized deviations from the mean value of a potential new regime is significantly different from the mean of the current regime (Rodionov, 2004). We used a 0.05 significance level to quantify statistically significant deviations in the age-structure data and to guard against spurious regime shifts that may appear in the middle of a time series (Overland et al., 2008), we visually verified that the regime-shift analysis fit the clear visual trends present (Elliott, 2012a). We selected a 10-year interval to detect regime-shift changes in the age-structure data (Elliott, 2011).

To assess the total distance of upper treeline advance, we re-constructed the uppermost elevation of the outpost tree along the transect for each five-year period during the 20th century (1900–1995 age classes). We then stratified the values based on whether a given advance occurred during “wet” or “dry” pentads to quantitatively compare the influence of climatic water deficit conditions on this process using a Mann-Whitney U Test. This nonparametric procedure tests for statistically significant differences in the median value of a time series. We also employed this method to test for significant shifts in the elevational extent of treeline on contrasting north- and south-facing slopes during wet and dry periods. We evaluated overall treeline advance using a 10 m increase as the minimum criterion (Liang et al., 2011). The rationale for a dual analysis of ecological regime-shift changes with pulses of upper treeline advance is to evaluate two separate lines of evidence regarding the extent to which upper treeline ecotonal dynamics are driven by abrupt, threshold-type changes.

ECOLOGICAL REGIME SHIFTS

Regime-shift analysis measured a significant increase in the rate of tree regeneration (p < 0.05) at 77.2% of sites sampled (n = 17/22, Figs. 3 and 4). The primary spatiotemporal attributes of these threshold responses include the following: (1) regime shifts were concentrated during the latter half of the 20th century (1945–1985 age classes), and (2) the initial triggering was asynchronous between opposite north-and south-facing slopes on all but one peak (n = 10/11, Figs. 3 and 4).

FIGURE 3.

Data from the Central Rocky Mountains showing 5-year phases (wet or dry) of the climatic water deficit (z-scores), pulses of upper treeline advance (m), and north- and south-facing age-structure data (n). Phases of the climatic water deficit are plotted with respect to the 20th century mean (1900–1999). The dashed line on the treeline advance graphs represents the 10 m minimum criterion used and the bold lines superimposed on the age-structure data represent the results of regime-shift analysis (p = 0.05). Refer to Table 1 for specific site information. An asterisk refers to a non-uniform axis.

Drought conditions coincided with 64.7%- (n = 11/17) of regime-shift increases in tree regeneration and this was the most pronounced on north-facing slopes (n = 7/11, Figs. 3 and 4). On the other hand, 50% of regime shifts on south-facing slopes aligned with wet conditions (n = 4/8, Figs. 3 and 4), which were most prevalent in the Front Range (n = 3/3 sites; Fig. 4, parts A–C). The south-facing slope of Medicine Bow Peak Massif therefore provides an interesting comparison because, unlike xeric slopes on nearby peaks (e.g., Kannady and sites in the Front Range), dry conditions evidently favor regime-shift increases in tree establishment (Fig. 3, parts D–E, and Fig. 4, parts A–C). Collectively, spatiotemporal patterns along the latitudinal gradient suggest that demographic processes on north-facing slopes were most responsive to drought conditions, whereas regeneration dynamics varied more on south-facing slopes with respect to fluxes in the climatic water deficit.

PULSES OF UPPER TREELINE ADVANCE

Pulses of upper treeline advance (>10 m) occurred 28 times during the 20th century, and 85.7% (n = 24) were confined to the period 1945–1995. These pulses were spread among every mountain peak except Mount Evans Massif in the Colorado Front Range and were more frequent on south-facing slopes, especially in the Medicine Bow and elsewhere in the Front Range (Figs. 3 and 4). The overall mean distance (±1 SD) of treeline advance was 54.0 m (±44.2 m), and given the nearly ubiquitous nature of it, there was no significant difference between the distance of advance on north- and south-facing slopes (p = 0.212, Fig. 5). From a temporal perspective, the median distance of upslope advance was significantly greater (p = 0.001) during the latter half of the 20th century compared to the first (24.8 m vs. 0.3 m, respectively).

FIGURE 4.

Data from the Southern Rocky Mountains showing 5-year phases (wet or dry) of the climatic water deficit (z-scores), pulses of upper treeline advance (m), and north- and south-facing age-structure data (n). Phases of the climatic water deficit are plotted with respect to the 20th century mean (1900–1999). The dashed line on the treeline advance graph represents the 10 m minimum criterion used, and the bold lines superimposed on the age-structure data represent the results of regime-shift analysis (p = 0.05). Refer to Table 1 for specific site information. An asterisk refers to a non-uniform axis.

Upper treeline advance was more frequent during drought (57.1%, n = 16/28), and although not statistically significant (p = 0.146), elevational increases were nearly twice as large during drought versus wet periods (35.6 m vs. 18.5 m, respectively, Fig. 5). On more me sic north-facing slopes, however, treeline advance was significantly greater (p = 0.049) during dry periods with more than a fourfold difference in median distances of upslope migration (26.6 m vs. 6.4 m, respectively). South-facing slopes exhibited the widest range of variability (range of advance = 0-187.0 m), and similar to opposite north-facing slopes, this was greater during dry periods than wet (43.5 m vs. 24.8 m, respectively, Fig. 5). Alternatively, two key exceptions exist on south-facing slopes (sites MB and TP) where treeline advanced more than 60 m under wet conditions (Fig. 5). Results from reconstructing the history of upper treeline advance during the 20th century highlight the critical role drought plays in facilitating these processes, particularly on north-facing slopes.

FIGURE 5.

Stratified distance of upper treeline advance at each site along the latitudinal gradient based on wet or dry climatic water deficit conditions. See Table 1 for site abbreviations.

DROUGHT AND TREE REGENERATION AT UPPER TREELINE

Results from this study offer key insight into how effective slope aspect is in modulating the interactions of broad-scale temperature inputs with more fine-scale moisture conditions and, by extension, the susceptibility of demographic processes to abrupt switches during drought at upper treeline. It has been suggested that in the presence of drought, climatically sensitive ecosystems, such as upper treeline ecotones, may be predisposed to threshold-induced regime shifts that coincide with more steady increases in temperature that help undermine system stability (Foley et al., 2003; Scheffer and Carpenter, 2003). Indeed, this type of “droughttriggering” scenario explains the abrupt regime-shift increases in regeneration dynamics at upper treeline from a regional perspective along this same latitudinal gradient during the early 1950s (1950–1954 age class) when rising temperatures coincided with a sharp reduction in snowfall (Elliott. 2012a). The finer scale results derived from individual mountain peaks in this study, however, offer a notable refinement to the drought-triggering hypothesis; namely, that within the context of warmer temperatures, the likelihood of drought events triggering an ecological regime shift in tree establishment at upper treeline varies by slope aspect. Thus, the slope aspect mediation of ecological regime-shift changes in tree regeneration reinforces the critical control it exerts on both temperature-precipitation and climate-vegetation interactions in high-elevation mountain environments.

From an ecoclimatological standpoint, however, it is important to consider that “drought” in this case primarily stems from reduced snowpack rather than a persistent reduction in growing-season precipitation, which is more typical at lower elevations where snow constitutes a lower proportion of annual moisture. Correspondingly, a reduction in snowpack likely prolongs favorable growing season conditions, which has facilitated both tree regeneration and upper treeline advance in the Swedish Scandes (Kullman, 2002) and Canadian Rocky Mountains (Luckman and Kavanagh, 1998). Although growing season length has been previously identified as a chief constraint on demographic processes at upper treeline, findings from this study provide new evidence suggesting that the ecological impacts of these conditions on tree demography are not uniform and are, instead, often contingent on slope aspect.

The slope aspect mediation of abrupt changes in upper treeline ecotonal dynamics can be explained by examining how the interactions between soil thermal regimes and snow often vary across a single mountain peak. For example, Körner and Paulsen (2004) demonstrated the importance of soil temperature thresholds in governing the elevational extent of tree regeneration, and although they reported insignificant temperature differences between contrasting slope exposures, this was likely the result of placing measuring devices underneath shaded canopies. In addition to necessary temperature requirements, studies suggest that the amount of snowfall and subsequent snowpack accumulation can severely limit the length of the growing season and thus rates of successful tree establishment within upper treeline ecotones (Minnich, 1984; Hättenschwiler and Smith, 1999). particularly on windward (Rochefort and Peterson, 1996) and/or north-facing slopes (Elliott and Kipfmueller, 2011). In fact, an important linkage between persistent snowpack and critical soil temperatures was recently made in southeastern Tibet, where researchers identified a 20–30 day delay in the onset of necessary soil warming on more shaded north-facing slopes compared to opposite south-facing slopes during the spring (Liu and Luo, 2011). This finding likely corresponds to accelerated rates of soil formation on north-facing slopes compared to south-facing slopes where increased climatic water deficits coupled with higher solar insolation impede pedogenic processes (Shickoff, 2005). Thus, although soil is one of the most complex site factors at upper treeline (Holtmeier, 2009), we conclude that regeneration processes are evidently unencumbered by late-lying snowpack during periods of drought, especially on north-facing slopes, and therefore able to respond more abruptly to the spatiotemporal alignment of favorable temperature and moisture conditions.

ROLE OF LOCAL HYDROCLIMATE

Evidence for demographic processes responding rapidly to drought in normally snow-heavy environments exists on north-facing slopes in the Bighorns (sites BM, PR), Medicine Bows (site KP), and Front Range (sites EM, PP) along with south-facing slopes on the other peak in the Medicine Bows (site MB) and Crown Point (CP) in the Front Range (Figs. 2–5). In some cases, however, results from this research imply that even in the presence of 20th century warming and during periods of drought, regeneration dynamics on north-facing slopes continue to be impeded by snow accumulation. This is the most discernible on the north-facing slope of Medicine Bow Peak Massif and Crown Point where regime shifts and pulses of treeline advance were absent. Coincidentally, both of these mountain peaks typically receive a high proportion of annual precipitation from snow, and the ecological impacts of this are almost certainly amplified on north-facing slopes (Table 1), which bolsters the likelihood of snow limiting tree regeneration and related biotic responses to climate variability in these hydroclimate settings dominated by snow.

Despite the widespread evidence for snow-limited tree establishment at upper treeline during wet periods along the latitudinal gradient, some instances remain where this is evidently not the case; most notably on south-facing slopes in the Front Range and on Kannaday Peak in the Medicine Bow Mountains (KP) where regime-shift increases coincide with wet conditions. Ecophysiological studies from the Rocky Mountains suggest that seedlings growing within upper treeline ecotones are consistently subject to combinations of high sun exposure and related rises in soil temperature that ultimately limit rates of establishment (Broderson et al., 2006; Moyes et al., 2013), especially on more xeric south-facing slopes (Weisberg and Baker, 1995; Germino et al., 2002). The predominance of clustered spatial patterns of tree establishment above timberline on south-facing slopes in this area provide further evidence for frequent moisture stress (Elliott and Kipfmueller, 2010). As for Kannaday Peak, tree regeneration is clearly operating under the influence of a much warmer and drier climate than that of neighboring Medicine Bow Peak Massif (Table 1, Fig. 2). In contrast to the frequent snow limitations for tree establishment on north-facing slopes, these interactions highlight the importance of wet conditions in maintaining regeneration dynamics on xeric south-facing slopes, which supports previous work from the northern Front Range (Hessl and Baker, 1997). More generally, these results provide broad-scale support for findings on Mount Ranier in the Cascades (Rochefort and Peterson, 1996) and in the Andes Mountains of northern Patagonia (Daniels and Veblen, 2004) where the relative influence of temperature-precipitation interactions were contingent on both slope aspect and local hydroclimate regimes. More specifically, this spatial variability portends uneven ecological responses to drought across upper treeline ecotones.

DATA LIMITATIONS

As with any study that relies on static age-structure data and modeled climate-variable data (e.g., climatic water deficit), a brief discussion of potential limitations is appropriate in order to fully justify our interpretations of the results.

AGE-STRUCTURE DATA

Although the dendroecological results from this study illustrate explainable patterns along a latitudinal gradient in the Rocky Mountains, there are inherent limitations to the interpretation of static age-structure data (Veblen, 1992; Johnson et al., 1994). For instance, statistically significant pulses of upper treeline advance were most concentrated during the latter half of the 20th century, which might just reflect a loss of evidence from earlier advances through tree mortality. Furthermore, successful tree establishment is dependent on rates of survival and mortality, with abrupt regime-shift changes resulting from above-average establishment and subsequent survival and/or low mortality. Equally important, the expected frequency of living trees naturally decreases through time, therefore making it more likely for regime-shift changes to occur more recently. Yet, dead seedlings only accounted for 3% of the total recorded (n = 11/366) and evidence of previous tree mortality was lacking from every site, with the critically important caveats that it is nearly impossible to fully understand seedling regeneration dynamics through a single season of fieldwork at each site and because of how quickly dead ones fully decompose. This is noteworthy considering that evidence of previous mortality, in the form of standing dead trees/snags, can persist on the landscape for several decades in these high-elevation environments.

CLIMATIC WATER DEFICIT

We plotted standardized deviations of both annual temperature (MAT) and precipitation (Precip) relative to the 20th century mean to see if the resulting climatic water deficit (CWD) phases (wet or dry) were logical, particularly for the Central Rocky Mountains where data are unavailable and sites were assigned an AWC value of 75 mm (Fig. 6). Given that “wet” or “dry” phases are all produced by plausible combinations of temperature and precipitation, our results suggest that the assignment of this value does not fundamentally alter the calculation of the climatic water deficit.