Tree regeneration following drought- and insect-induced mortality in piñon–juniper woodlands

Authors


Summary

  • Widespread piñon (Pinus edulis) mortality occurred across the southwestern USA during 2002–2003 in response to drought and bark beetle infestations. Given the recent mortality and changes in regional climate over the past several decades, there is a keen interest in post-mortality regeneration dynamics in piñon–juniper woodlands.
  • Here, we examined piñon and Utah juniper (Juniperus osteosperma) recruitment at 30 sites across southwestern Colorado, USA that spanned a gradient of adult piñon mortality levels (10–100%) to understand current regeneration dynamics.
  • Piñon and juniper recruitment was greater at sites with more tree and shrub cover. Piñon recruitment was more strongly facilitated than juniper recruitment by trees and shrubs. New (post-mortality) piñon recruitment was negatively affected by recent mortality. However, mortality had no effect on piñon advanced regeneration (juveniles established pre-mortality) and did not shift juvenile piñon dominance.
  • Our results highlight the importance of shrubs and juniper trees for the facilitation of piñon establishment and survival. Regardless of adult piñon mortality levels, areas with low tree and shrub cover may become increasingly juniper dominated as a result of the few suitable microsites for piñon establishment and survival. In areas with high piñon mortality and high tree and shrub cover, our results suggest that piñon is regenerating via advanced regeneration.

Introduction

Global climate change has been predicted to increase the frequency and severity of drought events occurring under warmer temperatures (Meehl & Tebaldi, 2004; Seager et al., 2007), which may lead to substantial shifts in vegetation distribution (Choat et al., 2012). Extensive tree mortality has already occurred across the western USA (Breshears et al., 2005; Van Mantgem & Stephenson, 2007; Van Mantgem et al., 2009; Worrall et al., 2010) and globally (Allen et al., 2010; Carnicer et al., 2011) over the past decade as a result of a combination of drought, insect outbreaks and warmer temperatures. These regional mortality events have altered water and energy fluxes (Guardiola-Claramonte et al., 2011; Royer et al., 2011; Adams et al., 2012), carbon cycling (Kurz et al., 2008; Brown et al., 2010; Pfeifer et al., 2011; Hicke et al., 2012) and ecosystem services (Walton, 2012; Anderegg et al., 2013). Yet, at present, we lack a clear understanding of how these regional mortality events may affect future forest and woodland composition, which is critical in understanding the long-term effects of recent tree mortality on these ecosystem processes and functions (Kurz et al., 2008; Anderegg et al., 2013). Thus, research examining tree regeneration patterns following recent mortality under a warmer, drier climate will help to elucidate the long-term changes in water and energy fluxes and carbon stocks.

Several abiotic and biotic factors are important drivers of successful tree regeneration under warmer, drier conditions. With increasing aridity, tree regeneration may become increasingly dependent on the availability of cooler, wetter microsites to facilitate seedling establishment and survival (Kitzberger et al., 2000; Sthultz 2007). In arid and semi-arid ecosystems, tree seedling establishment and survival are greater beneath tree and shrub canopies than in the adjacent canopy interspaces (Chambers, 2001; Mueller et al., 2005). Tree recruitment beneath canopies is probably a result of reduced solar radiation, resulting in less plant transpiration loss, cooler soil temperatures and less soil water loss through evaporation, as well as increased soil organic matter and nutrient availability (Callaway et al., 1996; Kitzberger et al., 2000; Chambers, 2001). Thus, facilitation, or the positive interaction among plants, by trees and shrubs may become increasingly important for tree regeneration by reducing water stress. However, trees and shrubs may also compete for resources, such as sunlight and water, and may negatively affect seedling growth and survival (Callaway et al., 1996; Breshears et al., 1997), particularly among later life stages (Callaway & Walker, 1997). Soil physical and chemical properties, such as texture and organic matter, influence soil water availability and may also be important in promoting seedling establishment and survival under a warmer, drier climate.

In this study, we examine how vegetation structure and soil properties influence piñon pine (Pinus edulis Engelm., hereafter referred to as piñon) and Utah juniper (Juniperus osteosperma (Torr.) Little, hereafter referred to as juniper) regeneration dynamics following drought- and insect-induced adult piñon mortality. We focus our research on woodlands co-dominated by piñon and juniper, because they are widely distributed across the southwestern USA and provide a range of ecosystem services to humans and critical habitat for a variety of wildlife species (Brown et al., 2001). Importantly, during the most recent multi-year drought (2002–2003), piñon experienced extensive mortality across populations, with mortality reaching upwards of 90% in some stands (Breshears et al., 2005; Mueller et al., 2005). This widespread piñon mortality event was caused by a combination of high temperatures, low precipitation and high bark beetle (Ips confusus) infestation (Breshears et al., 2005; McDowell et al., 2008; Adams et al., 2009). Mortality was highest among reproductively mature piñon trees (Mueller et al., 2005; Floyd et al., 2009), which resulted in large increases in relative dominance by adult juniper (Mueller et al., 2005; Koepke et al., 2010). Although recent piñon mortality increased adult juniper dominance, future woodland composition will be largely determined by regeneration dynamics.

With increasingly arid conditions, several authors have hypothesized that juvenile juniper dominance may increase relative to piñon, because juniper is more drought tolerant and is also more abundant in warmer, drier sites (Linton et al., 1998; West et al., 2007; Breshears et al., 2008; McDowell et al., 2008). In addition, although both piñon and juniper have greater rates of seedling establishment and survival beneath the canopy of trees and shrubs (Miller & Rose, 1995; Mueller et al., 2005), juniper is better able to establish in open environments and is also better able to compete with herbaceous vegetation than piñon (Chambers et al., 1999). Consequently, juniper establishment is commonly greater than piñon establishment following overstory tree and shrub removal (Tausch & Tueller, 1977; Everett & Ward, 1984; Redmond et al., 2013). Following this, we predict that the high mortality of overstory piñon may lead to reduced suitable microsites for piñon establishment, and shift the juvenile tree community to become more juniper dominated.

In addition to changes in suitable microsite availability, seed availability may influence piñon and juniper recruitment. Piñon cone production has declined by over 40% in areas of New Mexico over the last several decades (Redmond et al., 2012). These declines have been associated with recent increases in temperature (Redmond et al., 2012) and may negatively impact recent recruitment. Piñon seed viability also declines rapidly after 1 yr of storage (Meeuwig & Bassett, 1983), unlike juniper (Johnsen, 1959). Thus, piñon recruitment following high adult mortality may be dependent on the survival of juveniles that established before the mortality (i.e. advanced regeneration) as a result of limited seed availability.

Through an observational study that examined piñon and juniper recruitment across a gradient of adult piñon mortality levels in southwestern Colorado, we addressed the following hypotheses: piñon and juniper recruitment is greater at sites with more tree and shrub cover and higher soil available water capacity; piñon recruitment is more strongly facilitated than juniper recruitment by overstory trees and shrubs; piñon recruitment following high adult mortality is dependent on advanced regeneration; juvenile piñon dominance decreases with increased adult piñon mortality.

Materials and Methods

Study area

From May to June 2012, we sampled 30 sites across southwestern Colorado that had experienced varying levels (10–100%) of adult piñon mortality during the 2002–2003 drought (Fig. 1). Sites were similar in elevation (2080 ± 70 m), climate (mean annual temperature, 9.4 ± 0.8°C; mean annual precipitation, 398 ± 28 mm; data from the PRISM Climate Database, http://www.prism.oregonstate.edu) and juniper mortality (< 12%). We selected sites that were in the mid to upper elevation limits of piñon–juniper woodlands, as these elevations had the greatest mortality (Breshears et al., 2005). Sites were located on either Gladel–Pulpit complex soils (ecological sites, Pinyon–Juniper (Gladel) or Loamy Foothill (Pulpit)) or Wetherill soils (Loamy Foothill) (Natural Resources Conservation Service, 2006), which are the two most spatially extensive soil complexes in the region dominated by piñon and juniper (Natural Resources Conservation Service, 2006). Both of these soil complexes are characterized as well-drained soils derived from sandstone with a depth ranging from 0.5 to > 1.5 m (Natural Resources Conservation Service, 2006; and see Supporting Information Table S1 for soil texture and soil organic matter data for each site). To select sites, we first used Geographic Information Systems to locate areas within southwestern Colorado that were of the two soil complexes, between 2000 and 2200 m in elevation, had a slope of < 10% and were located on federally owned land. We then used Google Earth (Version 7.0; Google Inc., Mountainview, CA, USA) imagery taken in 2011 to identify areas of high and low mortality that were within 1 km of a road. Following this, we selected 15 sites that were located on Wetherill soils and 15 sites that were located on Gladel–Pulpit soils, half of which were considered to be high-mortality (> 60% mortality) sites and half of which were considered to be low-mortality (< 40% mortality) sites. All sites were generally at least 2 km apart (Fig. 1). However, we included stands that were within 2 km of one another if they differed in percentage mortality or differed by the Natural Resources Conservation Service soil map unit and vegetation characteristics. At our sites, the only two tree species were piñon and juniper. The common shrub species were antelope bitterbrush (Purshia tridentata), mountain mahogany (Cercocarpus montanus) and sagebrush (Artemesia spp.).

Figure 1.

Map of the 30 sites sampled in the USA. Different colored circles illustrate the different levels of adult piñon mortality that occurred at each site based on adult stem density.

Data collection

At each site, we established three 50-m-long transects, spaced 25 m apart. We surveyed for all tree juveniles, which included seedlings (basal diameter < 2.5 cm) and saplings (basal diameter ≥ 2.5 and < 5 cm), along a 10-m belt at each transect (i.e. 1500 m2 surveyed for tree juveniles per site). For all tree juveniles encountered, we recorded the species, status (dead or alive), basal diameter (diameter at the root collar) and microsite (beneath canopy vs canopy interspace). In addition, for each piñon seedling, we estimated whether the seedling established before the mortality event (i.e. ≥ 10 yr, which we refer to as an advanced seedling) or after the mortality event (i.e. ≤ 8 yr, which we refer to as a new seedling) by counting the number of annual whorls along the main stem, a method that has been used to age piñon (Sthultz et al., 2007) and other pine species (Collins et al., 2011).

To quantify tree and shrub density, cover, basal area (BA) and mortality, at each site we established five circular plots of 7 m in radius. One circular plot was located in the center of the middle transect and the other four circular plots were located at each end of the outer transects. We measured the height, two perpendicular canopy widths (live only), basal diameter (trees only) and status (live vs dead) for each tree and shrub species located within each circular plot. For all dead trees, we also noted whether the tree mortality was recent (≤ 10 yr) by following the guidelines in Jacobi et al. (2005), which were based on piñon in southwestern Colorado. Trees were considered to be recently (≤ 10 yr) dead if 33% or less of the branches were broken, most small limbs were present and at least 50% of the bark was attached and not > 2 mm from the tree.

To examine whether piñon and juniper juveniles were randomly distributed within a site or were associated with specific microsites, we first determined the relative availability of microsites using the line point intercept method. Every 50 cm along each transect, we recorded the microsite (tree and/or shrub presence vs interspace), for a total of 100 points/transect and 300 points/site. We used the line point intercept method, rather than data from our circular plots, to determine the relative availability of microsites, because tree and shrub canopies often overlapped. Therefore, the circular plot data probably overestimated canopy cover.

At each site, soil cores were taken from 0–10 cm soil depth at each transect end (six per site) to estimate soil available water capacity in surface soils. Soil cores were 2.5 cm in diameter and taken 1 m out from the canopy edge of juniper and/or piñon. Within a month of collection, soil samples were taken back to the University of Colorado, where they were dried for 48 h at 60°C, weighed, sieved through a 2-mm sieve and re-weighed. The six soil samples taken at each site were then combined, and organic matter content (%) was estimated using the loss on ignition method (Ball, 1964; Ben-Dor & Banin, 1989) with a muffle furnace set at 580°C. Percentage sand, silt and clay were estimated following Kettler et al. (2001). Using our estimates of percentage sand, silt, clay and organic matter content (see Table S1 for these data), we calculated soil available water capacity by subtracting the wilting point (θ1500) from the field capacity (θ33), which we calculated using the equations in table 1 of Saxton & Rawls (2006).

Table 1. The predictor variables and their ecological function used in each stepwise multiple linear regression for each size class
Predictor variableEcological function
  1. Predictor variables only used to predict piñon juvenile density, juniper juvenile density or piñon juvenile dominance are denoted by ‘P’, ‘J’ and ‘%P’, respectively. Piñon mortality was calculated on the basis of basal area rather than stem density. All of the vegetation predictor variables were calculated using the pooled circular plot data.

Live piñon basal areaPiñon seed availability and microsites after mortality
Live + dead piñon basal areaPiñon seed availability and microsites before mortality
Percentage piñon mortalityMicrosite changes (dead piñon microsites)
Live shrub coverMicrosite
Live juniper cover (P only)Microsite
Soil available water capacityWater availability
Live juniper basal area (J and %P only)Juniper seed availability and microsite

Data analyses

We divided tree seedlings into two size classes on the basis of their basal diameter (new seedlings, basal diameter ≤ 0.5 cm; advanced seedlings, basal diameter > 0.5 and < 2.5 cm), because we were unable to age all of the tree seedlings. These size classes were chosen using piñon seedling size–age data across our 30 sites (see Fig. S1), which indicate that 78% of the seedlings of ≤ 0.5 cm were < 9 yr old (i.e. established post-mortality), whereas 98% of seedlings of > 0.5 cm were > 10 yr old (i.e. established before mortality). We were unable to date juniper seedlings using the annual whorl method. However, piñon and juniper trees in southern Utah have a similar relationship between basal diameter and tree age among juveniles with a basal diameter of ≥ 2.0 cm (Fig. S2; M. D. Redmond, N. S. Cobb, M. Miller & N. N. Barger, unpublished). This suggests that the size–age relationship among piñon and juniper seedlings may also be similar. Because we divided seedlings into age classes (new seedlings and advanced seedlings) on the basis of their size for the analyses, it is important to note that some seedlings may be incorrectly classified, which may reduce our ability to detect differences in vegetation associations between new seedlings and advanced seedlings.

We evaluated the vegetation and soil characteristics associated with piñon and juniper recruitment by performing stepwise multiple linear regression analyses for each juvenile size class (new seedlings, advanced seedlings and saplings). For these analyses, we used each site as a replicate by pooling and then averaging transect and circular plot data. In each linear regression model, juvenile density was used as a response variable and various vegetation and soil characteristics that were identified a priori were used as predictor variables (see Table 1 for a list of all predictor variables used and their hypothesized function). It should be noted that we used piñon BA rather than canopy cover as a predictor variable for all regression analyses, because we did not collect canopy cover data of dead piñon trees. Further, live piñon BA is a strong predictor of live piñon canopy cover (simple linear regression, R2 = 0.91, P < 0.0001). We also used juniper BA rather than canopy cover as a predictor variable for the regression analyses with juniper juveniles (Table 1), because juniper often did not grow vertically, and so canopy cover probably overestimated seed availability in these cases.

Each complete model was simplified using the stepAIC procedure as part of the package MASS (Venables & Ripley, 2002) in R (R Development Core Team, 2011) following the procedure outlined in Zuur et al. (2007). The stepAIC procedure examines all predictor variables using backward selection to produce a final model based on the minimization of the Akaike Information Criterion (AIC; Akaike, 1973). Piñon new seedlings and juniper saplings were log transformed to meet the assumptions of these tests.

Following the same procedure as above, we examined the vegetation and soil characteristics associated with piñon dominance for each juvenile size class to determine potential changes in future stand composition across our study sites (see Table 1 for the list of predictor variables). Piñon dominance was calculated as:

display math

To determine whether recent increases in aridity and recent mortality had shifted piñon dominance, we first examined whether juvenile piñon dominance had shifted relative to adult piñon dominance, regardless of piñon mortality levels. To do this, we performed paired Student's t-tests to compare juvenile piñon dominance at each size class (new seedling, advanced seedling and sapling) with adult piñon dominance before recent mortality (calculated using live and dead tree density). To determine whether recent mortality had shifted juvenile piñon dominance relative to adult piñon dominance, we performed regression analyses to examine whether the difference in dominance between juveniles and adults (response variable) was affected by piñon mortality (predictor variable). We calculated the difference in dominance between juveniles at each size class and adults by dividing juvenile piñon dominancesize class X by adult piñon dominance before recent mortality.

To determine whether overstory trees and shrubs had a facilitative or competitive effect on piñon and juniper recruitment, and how this effect may differ between the two species and between the three juvenile size classes, we used the relative interaction index proposed by Armas et al. (2004). The interaction index reveals whether competition or facilitation is occurring within each microsite by representing the relative difference in seedling density occurring beneath tree and shrub canopies and in the interspace, and is calculated as:

display math

Thus, the interaction index ranges from −1 to 1, with a positive interaction index representing a facilitative effect of canopy microsites on recruitment, and a negative interaction index representing a competitive effect of canopy microsites. For each species and each juvenile size class, we estimated the juvenile density at each microsite by dividing the total number of juveniles found underneath the canopy (or interspace) microsite by the total cover of live and dead shrubs and trees (or the total cover of area with no trees and shrubs), which was calculated using the line point intercept data. Because we wanted to compare how the interaction index varies between the two species and the three juvenile size classes, we only used sites that had a sufficient number of juveniles sampled in each size class and each species. To do this, we only used sites that, given a uniform distribution of juveniles, we would expect at least one individual of each size class and each species to be located beneath the canopy (i.e. total number of individualsspecies X, size class Y × proportion canopy cover ≥ 1). We omitted 14 sites from the analysis as these sites did not meet the criteria. Using the 16 sites that met our criteria, we calculated the interaction index at each site for each species and each juvenile size class. We then performed a series of non-parametric paired Wilcoxon signed rank tests (nine paired tests in total) to determine whether the interaction index differed between piñon and juniper within each juvenile size class (three paired tests) and whether the interaction index differed between each juvenile size class within each species (six paired tests). We used non-parametric paired Wilcoxon signed rank tests rather than a paired two-way ANOVA because of the non-normality of the data, even after transformations.

We also examined the effect of canopy presence on piñon sapling survival by comparing the interaction indices of live piñon saplings and dead piñon saplings beneath the canopy of trees and shrubs. Similar to above, we only used sites that had a sufficient number of live and dead piñon saplings, such that, given a uniform distribution, we would expect to find at least one live and dead sapling underneath the canopy. Using the 18 sites that met our criteria, we performed a paired Wilcoxon signed rank test to determine whether the interaction index differs between live and dead piñon saplings.

As multiple comparisons (10) were conducted using interaction indices, we applied the false discovery rate (FDR) correction using a standard step-up procedure to control the inflation of type I error derived from repeated testing (Benjamini & Hochberg, 1995; García, 2004). All analyses were performed in R (R Development Core Team, 2011), with α = 0.05.

Results

As intended, our study sites spanned a gradient of adult piñon mortality levels, with mortality ranging from 16% to 100% based on BA (9–100% based on stem density). In addition, there was high variability in shrub cover (0–14%), tree density (169–1105 stems ha−1) and tree BA (12–52 m2 ha−1) across our study sites (Tables 2, S2), as well as some variability in soil available water capacity (11–16%; Table S2). Piñon and juniper juvenile (i.e. seedlings and saplings) densities were also highly variable across our study sites, with juvenile densities ranging from 27 to 948 stems ha−1 for piñon and 40 to 827 stems ha−1 for juniper (Table 2). Among both piñon and juniper, the density of advanced seedlings (i.e. seedlings that probably established before adult mortality) was over 2.5-fold higher than the density of new seedlings (i.e. seedlings that probably established following adult mortality), on average (Table 2). However, the difference in density between new seedlings and advanced seedlings was highly variable across the 30 study sites (Table S2), such that, at some sites, new seedlings were over 1.4 times more abundant than advanced seedlings, whereas, at others, advanced seedlings were over 25 times more abundant than new seedlings (Table S2).

Table 2. Stem density and piñon dominance across our 30 sites
Size classPiñon density (stems ha−1)Juniper density (stems ha−1)Piñon dominance (%)
  1. Data are means ± 1SE. The range is included in parentheses.

New seedlings67 ± 16 (0–300)34 ± 5 (0–107)54 ± 5 (0–100)
Advanced seedlings194 ± 20 (13–480)193 ± 19 (6–333)51 ± 3 (22–93)
Sapling137 ± 16 (0–327)90 ± 16 (0–480)58 ± 5 (0–100)
Adult (pre-mortality)488 ± 54 (78–1039)447 ± 25 (169–753)49 ± 3 (13–77)
Adult (post-mortality)268 ± 34 (0–623)421 ± 24 (155–727)35 ± 4 (0–70)

Vegetation and soil characteristics associated with tree recruitment

The density of piñon new seedlings was positively associated with live shrub cover, live piñon BA and live juniper cover (R2 = 0.63, P < 0.0001; Table 3; Fig. 2a). Notably, there was no relationship between percentage piñon mortality and piñon new seedling density. Given that live piñon BA was strongly associated with piñon new seedling density, our results suggest that only current live piñon abundance influences new seedling establishment.

Table 3. Results of the best linear models based on stepAIC used to predict piñon and juniper densities and piñon dominance in each juvenile size class
 Size class R 2 P valuePredictor variablesβsP value
  1. The standardized regression coefficient (βs) and P value are given for each predictor variable used in the final model based on stepAIC. = 30 (except = 28 for piñon dominance of new seedlings because of a lack of new seedlings at two sites).

  2. P-values of statistically significant predictor variables are shown in bold (P <0.05).

PiñonNew seedling0.63<0.0001Live shrub cover0.50 0.001
Live piñon basal area0.47 0.002
Live juniper cover0.36 0.01
Adv. seedling0.64<0.0001Live shrub cover0.70 < 0.001
Live + dead piñon basal area0.55 0.004
Live juniper cover0.38 0.01
Soil available water capacity0.31 0.03
Sapling0.310.001Live shrub cover0.56 0.001
JuniperNew seedling0.260.02Live piñon basal area0.40 0.02
Live juniper basal area0.340.05
Adv. seedling0.090.11Soil available water capacity0.300.11
Sapling0.210.01Live juniper basal area−0.46 0.01
Piñon dominanceNew seedling0.320.002Live shrub cover0.56 0.002
Adv. seedling0.350.01Live shrub cover0.39 0.02
Live juniper basal area0.330.06
Live + dead piñon basal area0.300.08
Sapling0.59<0.0001Live juniper basal area0.56 0.0002
Live shrub cover0.46 0.001
Live + dead piñon basal area0.30 0.03
Figure 2.

(a) Plot of the logarithm of the density of piñon new seedlings (i.e. seedlings with a basal diameter ≤ 0.5 cm, which probably established after adult mortality) against percentage shrub cover and live adult piñon basal area (BA), the two strongest predictor variables. (b) Plot of the density of piñon advanced seedlings (i.e. seedlings with a basal diameter > 0.5 cm, which probably established before adult mortality) against percentage shrub cover and live and dead adult piñon basal area, the two strongest predictor variables. Dashed grid represents the multiple linear regression model using the two strongest predictor variables in the regression models presented in Table 3.

The density of piñon advanced seedlings was also positively associated with live shrub and juniper cover (Table 3; Fig. 2b). However, unlike piñon new seedlings, the density of piñon advanced seedlings was positively associated with piñon BA before adult mortality and soil available water capacity (final model R2 = 0.64, < 0.0001; Table 3; Fig. 2b). Given that piñon BA before adult mortality was a strong predictor of piñon advance seedling density, these results suggest that recent mortality did not affect the abundance of advanced seedlings.

Although there was a strong association between the density of piñon seedlings and vegetation structure (Table 3), piñon sapling density was only positively associated with live shrub cover (R2 = 0.31, = 0.001; Table 3).

Juniper juveniles were more weakly associated than piñon juveniles with vegetation and soil characteristics. Juniper new seedling density was positively associated with live piñon BA and live juniper BA, whereby 26% of the variability of juniper new seedlings could be explained by these two variables (Table 3). Unlike juniper new seedlings, there was no relationship between juniper advanced seedling density and the predictor variables (R2 = 0.09, = 0.11; Table 3). There was weak evidence that the presence of adult junipers may inhibit the advancement of junipers to the sapling size class: juniper sapling density was weakly and negatively related to juniper BA (R2 = 0.21, = 0.01; Table 3).

Vegetation and soil characteristics associated with juvenile piñon dominance

Juvenile piñon dominance across our study sites was highly variable, ranging from 19% to 86% (Table 2). Piñon dominance among new seedlings was positively influenced by live shrub cover (R2 = 0.32, = 0.002; Table 3). Dominance of both piñon advanced seedlings and piñon saplings was positively associated with live shrub cover, live juniper BA and live + dead piñon BA (R2 = 0.35 and 0.59 for advanced seedlings and saplings, respectively; Table 3). These results highlight how tree and shrub microsites are important for juvenile piñon dominance, and suggest that percentage piñon mortality did not influence juvenile piñon dominance.

There were no significant differences between adult piñon dominance before recent mortality and juvenile piñon dominance among all size classes (paired Student's t-tests; > 0.09; Fig. 3; Table 2), suggesting that there was no directional shift in juvenile piñon dominance relative to adult piñon dominance across our 30 sites. There was also no relationship between piñon mortality and the difference in piñon dominance among juveniles and adults (R2 < 0.1; > 0.1; Fig. 3). Although there was some variability in the relationship between juvenile piñon dominance and adult piñon dominance (Fig. 3), our results suggest that, overall, the juvenile tree community is similar to that of the adult tree community before the recent mortality (Fig. 3).

Figure 3.

The relationship between juvenile piñon dominance and adult piñon dominance across the 30 sampled sites in the USA. Different colored circles illustrate the different levels of adult piñon mortality that occurred at each site based on adult basal area. The line illustrates where juvenile piñon dominance is equal to adult piñon dominance.

Net effects of facilitation and competition by trees and shrubs

Consistent across all juvenile size classes, both piñon and juniper were more likely to be found beneath tree and shrub canopies than in the interspace, as evidenced by the high positive interaction indices (indicating facilitation) between shrub and tree cover and juvenile piñon and juniper trees (Fig. 4). In addition, although piñon and juniper new seedlings had high interaction indices (> 0.90), with over 90% of new seedlings occurring beneath trees and shrubs, piñon and juniper saplings had significantly lower interaction indices (0.76 and 0.40, respectively; Fig. 4). Unlike new seedlings, piñon advanced seedlings and saplings had significantly higher interaction indices than juniper (Fig. 4), suggesting that, at these larger juvenile size classes, juniper is less able to survive than piñon beneath tree and shrub canopies.

Figure 4.

The interaction indices between shrub and tree cover and live juvenile piñon (gray bars) and juniper (black bars) trees. Interaction indices between 0 and 1 represent a facilitative effect of shrub and tree cover on piñon and juniper recruitment, with increasingly positive values representing an increasingly facilitative effect. Asterisks denote significant differences between piñon and juniper juveniles within each size class, and different letters denote significant differences between juvenile size classes within each species. Error bars are ± 1SE.

Both live and dead piñon saplings were more likely to be found beneath the canopy of a tree or shrub than in the interspace, as evidenced by the high positive interaction indices between shrub and tree cover and piñon sapling densities (mean interaction index ± 1SE = 0.76 ± 0.06 and 0.83 ± 0.07 for live and dead piñon saplings, respectively). However, interaction indices did not differ between live and dead piñon saplings (paired Wilcox signed rank test, P = 0.3), indicating that piñon sapling survival was the same in both canopy and interspace microsites.

Discussion

Vegetation and soil characteristics associated with tree recruitment

Our results suggest that recovery following recent piñon mortality will be dependent on advanced regeneration, findings which are consistent with other studies examining tree regeneration following drought- and insect-induced tree mortality (Veblen et al., 1991; Axelson et al., 2009; Collins et al., 2011; Diskin et al., 2011; Kayes & Tinker, 2012). New (post-mortality) piñon recruitment was strongly positively associated with live adult piñon density (Table 3; Fig. 2a), and was thus negatively affected by recent piñon mortality. The low levels of new piñon recruitment following high mortality are probably a result of seed limitations, as piñon seed viability declines rapidly after the first year (Meeuwig & Bassett, 1983), and may also be caused by declines in canopy microsites. Notably, piñon advanced (pre-mortality) seedlings and saplings were not affected by overstory piñon mortality and, instead, were positively associated with piñon density at the time of establishment (i.e. live + dead piñon BA; Table 3). These results highlight the importance of advanced regeneration for piñon recovery, and support previous findings of low juvenile piñon mortality in areas with high adult piñon mortality (Negrón & Wilson, 2003; Mueller et al., 2005). The relatively high levels of advanced regeneration following drought- and beetle-induced mortality in this study and others (Veblen et al., 1991; Axelson et al., 2009; Collins et al., 2011; Diskin et al., 2011; Kayes & Tinker, 2012) suggest that there is a greater bottleneck in obtaining new recruitment, which is dependent on seed production, dispersal, germination and seedling establishment, as compared with advanced regeneration, which is only dependent on juvenile survival following recent mortality.

Shrub and juniper cover was strongly positively associated with piñon juvenile density (Table 3), which supports our hypothesis that trees and shrubs facilitate piñon establishment and is consistent with other field observations and experimental studies (Floyd, 1982; Chambers, 2001; Mueller et al., 2005; Sthultz et al., 2007). The shading provided by trees and shrubs reduces solar radiation, leading to lower soil temperatures and less evapotranspiration (Breshears et al., 1997; Chambers, 2001; Royer et al., 2010, 2012), and may result in higher soil water contents (Chambers, 2001). Higher nutrient availability and higher organic matter content have also been reported underneath trees and shrubs in piñon–juniper woodlands (Padien & Lajtha, 1992; Chambers, 2001), which may also positively influence piñon recruitment.

In addition to vegetation structure, our results suggest that high soil available water capacity may be important for the promotion of piñon juvenile survival. Regionally, there was greater adult piñon mortality in areas with lower soil available water capacity (Peterman et al., 2012). Although the range of soil available water capacity across our sites was relatively narrow compared with other studies (Peterman et al., 2012), there was a positive association between soil available water capacity and piñon advanced seedling density (Table 3). This positive association may be caused by greater piñon seedling survival during the 2002–2003 drought in areas with higher soil available water capacity. Because piñon recovery following high adult mortality will probably be dependent on advanced regeneration, these results suggest that areas with lower soil available water capacity may have less piñon regeneration following high adult mortality.

Juniper juveniles were not as strongly associated with vegetation structure as piñon, and were also more commonly found in the canopy interspaces than piñon (Table 3; Fig. 4). These results support our hypothesis that piñon is more dependent on overstory trees and shrubs to facilitate seedling establishment and survival, and are consistent with previous field studies (Chambers et al., 1999). However, juniper new seedlings were still strongly facilitated by overstory trees and shrubs, unlike juniper advanced seedlings and saplings (Table 3; Fig. 4). This suggests that competition for light and, potentially, water may be more important than the facilitative effects of overstory trees and shrubs once juniper seedlings become established.

Shifts in the juvenile tree community

Counter to our predictions, piñon mortality resulted in no shift in the juvenile tree community (Table 2; Fig. 3), suggesting that future woodland composition may be similar to the woodland composition in the past. The lack of a shift in the juvenile tree community is in contrast with other studies that have examined tree regeneration patterns following drought- and/or beetle-induced mortality (Veblen et al., 1991; Suarez & Kitzberger, 2008; Collins et al., 2011; Kayes & Tinker, 2012). However, all of these studies examined tree regeneration following the mortality of relatively shade-intolerant species, such as lodgepole pine (Pinus contorta). In contrast, piñon is more shade tolerant than juniper (Barney & Frischknecht, 1974; Chambers et al., 1999), which, we hypothesize, enabled piñon to have higher juvenile densities before mortality, leading to sufficient advanced regeneration. Recovery following drought- and insect-induced tree mortality will be largely dependent on juveniles present in the community before the mortality event (Veblen et al., 1991; Suarez & Kitzberger, 2008; Collins et al., 2011). Thus, there may be greater shifts in the juvenile tree community following overstory mortality of shade-intolerant tree species because of little advanced regeneration.

Juvenile piñon dominance was still highly variable across our study sites (Table 2; Fig. 3), and was positively related to tree and shrub cover (Table 3). This suggests that there may be a shift in the juvenile tree community in the future. Once more dead piñon snags fall, piñon dominance may become negatively impacted by the loss in canopy microsites. There has also been significant shrub mortality and juniper dieback in piñon–juniper woodlands in northern Arizona following recent droughts (Gitlin et al., 2006; Koepke et al., 2010), which could further reduce suitable microsites and affect future piñon dominance. In addition, drought- and insect-induced mortality events are predicted to increase in the southwestern USA (Williams et al., 2013). These predicted future mortality events may result in a shift in juvenile piñon dominance. In particular, as these older juveniles grow, they will probably become more susceptible to bark beetle attack and less likely to survive future droughts (Floyd et al., 2009), which may result in less advanced regeneration.

Net effects of facilitation and competition by trees and shrubs

Juveniles of both piñon and juniper across all size classes were more likely to be found underneath the canopy of trees and shrubs than in the adjacent interspace (Fig. 4), which supports the hypothesis that trees and shrubs facilitate piñon and juniper establishment (Miller & Rose, 1995; Chambers, 2001; Mueller et al., 2005; Sthultz et al., 2007). The high percentage of piñon and juniper seedlings underneath tree and shrub canopies may also be partly a result of the caching of seeds underneath trees and shrubs by birds and small mammals (Vander Wall & Balda, 1977; Vander Wall, 1997).

New seedlings of piñon and juniper were more likely than saplings to be found underneath the canopy of trees and shrubs (Fig. 4). There are two, non-mutually exclusive, hypotheses for why there is a decline in the positive association between shrub and tree canopies and tree juveniles with juvenile age. One hypothesis is that, as juveniles become older, increased competition with overstory trees and shrubs may outweigh the facilitative effects. In support of this hypothesis, both juniper and various shrub species have been reported to negatively impact adult piñon growth through below-ground competition (Haskins & Gehring, 2004; McHugh & Gehring, 2006), and shifts from facilitation to competition are known to occur among a variety of plant species with changes in life stages (Callaway & Walker, 1997). Another hypothesis is that these older juveniles established during cooler, wetter climate conditions, and thus were not as dependent on tree and shrub microsites at the time of establishment. In support of this hypothesis, previous research on piñon (Sthultz et al., 2007) and other tree species (Greenlee & Callaway, 1996; Kitzberger et al., 2000) has found that the net effects of competition and facilitation vary across spatial and temporal gradients of abiotic stress.

Interestingly, there was no difference in survival between piñon saplings located beneath the canopy of a tree or shrub and those located in the canopy interspace. This suggests that, at this life history stage, facilitation and competition were equal between canopy microsites and piñon saplings in our study area. However, piñon sapling density was positively associated with shrub cover (Table 3). This may be caused by the facilitative effects of shrubs on piñon seedling establishment (Fig. 4) and the aboveground competitive release that occurs once the sapling grows taller than the shrub.

Conclusion

Drought- and insect-induced piñon mortality events have dramatically altered woodland structure across the southwestern USA (Breshears et al., 2005; Mueller et al., 2005), affecting water and energy fluxes and carbon stocks (Royer et al., 2011; Adams et al., 2012; Hicke et al., 2012). Our results indicate that piñon mortality does not strongly affect successional trajectories in piñon–juniper woodlands because of advanced regeneration. Accordingly, this landscape appears to be resilient to recent drought- and beetle-induced mortality because of the high densities of juveniles that survived the drought. However, piñon and juniper have slow growth rates in these semi-arid ecosystems (Despain, 1989; Barger et al., 2009), which suggests that ecosystem processes and function may be altered in high-mortality stands for 50 yr or more. As hypothesized, trees and shrubs appear to facilitate both piñon and juniper seedling establishment, and piñon juveniles were more dependent than juniper on tree and shrub microsites. Piñon advanced seedlings were also positively associated with soil available water capacity, unlike juniper. These results highlight the greater drought tolerance of juniper relative to piñon (Linton et al., 1998; West et al., 2007; Breshears et al., 2008; McDowell et al., 2008), and suggest that tree and shrub microsites may become increasingly important for future piñon recruitment, given the predicted increases in aridity across the southwestern USA (Seager et al., 2007; Williams et al., 2013).

Acknowledgements

We thank N. Touchet, C. Luben and E. Golden for assistance in the field and/or laboratory. Additional thanks are due to the members of Barger's laboratory, members of the Climate and Forest Dynamics seminar taught by T. Veblen and three anonymous reviewers for helpful comments on previous versions of the manuscript. This project was funded by the University of Colorado, Department of Ecology and Evolutionary Biology, a Beverly Sears Graduate Student Grant, a John Marr Graduate Student Grant, a United States Department of Agriculture (USDA) National Institute of Food and Agriculture Award #2011-67003-30755 and a National Science Foundation (NSF) Graduate Research Fellowship to M.D.R. under Grant No. DGE 1144083.

Ancillary