Diversity in Burned Pinyon–Juniper Woodlands Across Fire and Soil Parent Material Gradients

Abstract

Co-varying disturbance and environmental gradients can shape vegetation dynamics and increase the diversity of plant communities and their features. Pinyon–juniper woodlands are widespread in semi-arid climates of western North America, encompassing extensive environmental gradients, and a knowledge gap is how the diversity in features of these communities changes across co-varying gradients in fire history and soil. In pinyon–juniper communities spanning soil parent materials (basalt, limestone) and recent fire histories (0–4 prescribed fires or managed wildfires and 5–43 years since fire) in Grand Canyon-Parashant National Monument (Arizona, USA), we examined variation at 25 sites in three categories of plant community features including fuels, tree structure, and understory vegetation. Based on ordinations, canonical correlation analysis, and permutation tests, plant community features varied primarily with the number of fires, soil coarseness and chemistry, and additionally with tree structure for understory vegetation. Fire and soil variables accounted for 33% of the variance in fuels and tree structure, and together with tree structure, 56% of the variance in understories. The cover of the non-native annual Bromus tectorum was higher where fires had occurred more recently. In turn, B. tectorum was positively associated with the percentage of dead trees and negatively associated with native forb species richness. Based on a dendroecological analysis of 127 Pinus monophylla and Juniperus osteosperma trees, only 18% of trees presently around our study sites originated before the 1870s (Euro-American settlement) and <2% originated before the 1820s. Increasing contemporary fire activity facilitated by the National Park Service since the 1980s corresponded with increasing tree mortality and open-structured stands, apparently more closely resembling pre-settlement conditions. Using physical geography, such as soil parent material, as a landscape template shows promise for (i) incorporating diversity in long-term community change serving as a baseline for vegetation management, (ii) customizing applying treatments to unique conditions on different soil types, and (iii) benchmarking monitoring metrics of vegetation management effectiveness to levels scaled to biophysical variation across the landscape.

1. Introduction

Co-varying environmental and disturbance gradients can be associated with a diversity of species and plant community features across landscapes [1,2,3]. For example, soils derived from different parent materials can vary in texture and chemistry, leading to differences in community structure and species assemblages associated with particular parent materials [4,5,6]. Disturbances occurring across environmental gradients can further increase diversity, such as through reducing competitive dominants to enable species coexistence [7]. Although disturbances can accentuate vegetational differences across environmental gradients, they could instead homogenize vegetative conditions [8]. Depending on pre-disturbance conditions, for example, fires either differentially or similarly affecting woodland tree structure across soil types could in turn enhance or reduce heterogeneity in understory plant communities across landscapes [9,10].

We examined the diversity of plant community features across gradients of soil parent materials and the history of fire activity within the last 43 years in a landscape of pinyon–juniper woodlands containing the small trees Pinus monophylla (hereafter Pinus) and Juniperus osteosperma (hereafter Juniperus). Pinyon–juniper communities containing various tree species within these genera occupy 20 million hectares across a range of soils in semi-arid climates in western North America [11]. One of the largest challenges for ecological research, ecological restoration, and fire and conservation management in pinyon–juniper communities is understanding and incorporating long-term community change and response to disturbances, including fires [12]. A commonly used structural classification includes three broad types of pinyon–juniper communities [13]. First, persistent pinyon–juniper woodlands are thought to occur on sites with infrequent stand-replacing fires (e.g., return intervals of 300–500+ years), where sparse fine fuel limited fire spread and where contemporary communities have many old trees or are in stages of re-establishment since the last severe disturbance. Second, wooded shrublands are thought to contain moderate densities of trees and shrubs and potentially mixed-severity fire regimes including surface fires or periodic crown fires. Third, pinyon–juniper savannas historically had low tree density and more frequent surface fires partly fueled by herbaceous plants. It is also thought that around many pinyon–juniper communities, portions of landscapes historically contained few or no trees, such as grasslands or sagebrush (Artemisia spp.)-dominated shrublands because of climatic, edaphic, or fire constraints on tree recruitment [14].

Since the mid-1800s to early 1900s, land use change associated with settlement occurred in pinyon–juniper communities, such as the introduction of non-native plants (e.g., the annual grass Bromus tectorum), woodcutting (for fuelwood or uses such as fence posts), livestock grazing that reduced fine fuels and potentially fire activity, and the intentional suppression of fires in some areas [15,16,17]. Decreased fire activity, livestock grazing potentially reducing herbaceous competition with tree seedlings, and possibly changing climatic conditions (warming after the mid-1800s) are thought to be associated with two general types of tree structural change in some formerly more open pinyon–juniper communities: infilling and expansion [18]. Infilling involves increased tree density in historically more open pinyon–juniper communities. Expansion involves the establishment of trees in areas that historically had few or no trees.

Although some wildlife benefit from dense pinyon–juniper stands, increasing tree density has been associated with degraded habitat quality for open habitat-affiliated wildlife and pollinators, lowered understory productivity, soil erosion, and increased risk of severe wildfires often followed by accelerated non-native plant invasion [19]. To address these negative effects, mechanical tree cutting, prescribed fires, or allowing wildfires to burn if the effects are considered desirable (managed wildfires) are being adopted as restoration or management strategies in pinyon–juniper communities thought to have experienced increased tree density [12].

Many authors have noted that these pinyon–juniper communities, as among the most widespread vegetation types in western North America, occupy a range of landforms and soils and may exhibit high variability in changes in tree structure in the last 100+ years and responses to contemporary fires, e.g., [20,21,22]. In fact, a recent review of pinyon–juniper research highlighted understanding diversity in these communities, including after fires across environmental gradients, as a major knowledge gap and research need [23]. Across the dual gradients of soil parent material and fire activity, we examined variation in pinyon–juniper communities in three categories of characteristics, which were fuels, tree structure, and understory vegetation. We asked the following question: which combinations of soil and fire history variables are most strongly associated with variation in post-fire pinyon–juniper community characteristics? Our results are intended to advance the understanding of diversity in post-disturbance vegetation across environmental gradients and to facilitate incorporating variation in physical geography when applying vegetation management treatments.

2. Materials and Methods

2.1. Study Area

We conducted this study in the 84,357 ha portion administered by the National Park Service of the 411,652 ha Grand Canyon-Parashant National Monument, Arizona, USA (Figure 1). Within the Colorado Plateau ecoregion, our 25 study plots (detailed below) were in pinyon–juniper communities at elevations of 1800–2000 m. The Köppen climate classification for the study area is cold semi-arid. In a pinyon–juniper community at an elevation of 1878 m in the study area, precipitation averaged 398 mm/year (165–664 mm/year range; Yellow John Mountain, Arizona; 1988 through 2023 records; Western Regional Climate Center, Reno, NV). Daily temperature ranges averaged −9–10 °C in January and 10–32 °C in July. In summarizing land use history, Ireland et al. [24] noted that the Southern Paiutes inhabited the area before Euro-American land use change primarily beginning in the 1870s with livestock grazing. Livestock likely reduced fine fuels and potentially fire activity. At least two sawmills operated in the area by the early and mid-1900s, but they are thought to have focused on Pinus ponderosa sawtimber rather than on the smaller Pinus and Juniperus trees at lower elevations. In P. ponderosa forests at higher elevations or moister sites around pinyon–juniper communities near our study area, surface fires were frequent (fire return intervals averaging 3–5 years for all fire-scarred trees and 7–23 years for fires scarring ≥ 25% of trees) in the 1600s until the 1880s [24]. After the 1880s, fire became uncommon through the 1900s, which Ireland et al. [24] hypothesized was primarily related to a combination of livestock grazing that reduced fine fuels and a management policy of fire exclusion. Although fire history in P. ponderosa forests suggests that spreading fires were frequent before Euro-American settlement and virtually ceased thereafter, pre-settlement fire regimes are less clear in the pinyon–juniper communities. However, pinyon–juniper communities similarly experienced little widespread fire since the late 1800s [25] until land managers began incorporating prescribed and managed wildfires in the decades preceding our study.

2.2. Data Collection and Sample Processing

Using paper and digital records provided by the National Park Service (Grand Canyon-Parashant National Monument, St. George, UT, USA) of prescribed fires and wildfires occurring from 1980 (when records were available) to 2018 (the most recent fire) in pinyon–juniper communities in the study area, we created polygons representing different combinations of time since the most recent fire (years) and number of fires incurred. Polygons of fires ranged from ~0.1–1600 ha. The size range resulted from factors such as physical barriers to fire spread (e.g., changes in topography or fuels [26]), changes in weather during fires, and anthropogenic fire management activities (ignitions and suppression depending on fire locations and behavior). We did not separate prescribed fires and wildfires for two reasons, namely that most wildfires became managed fires under a policy of allowing wildfires to continue burning if meeting fire management goals, termed resource objective wildfires [27], and it was not possible to distinguish between their influences in our study. For example, some plots had been burned by one or more prescribed fires as well as wildfires, or one or the other but with varying years since fire. Although this did not enable distinguishing between fire types, the variable fire histories among plots enabled evaluating communities across a diverse gradient of fire history (Table S1, Figure 2).

Within fire history polygons, we randomly selected polygons and one point (using Universal Transverse Mercator coordinates, zone 12) within each selected polygon at which to establish a plot. We sampled a total of 25 plots (5 in 2022 and 20 in 2023 in mid--to-late summer from the end of June to early October). During years of data collection in our study, precipitation was 116% of the average in 2022 (462 mm) and 2023 (464 mm). Sampling included 5 unburned plots (assigned a 50-year time since fire to reflect no fire since at least the 1970s) and 20 plots burned 1–4 times between 1980 and 2022–2023 (Figure 1, Table S1).

To identify soil parent material for each plot, we used a soil survey in which soil taxonomic units (typically series or two closely associated series with similar parent materials) were mapped at a 1:24,000 scale [26]. We classified parent material as either basalt or limestone. Soils derived from basalt were mapped as Argiustolls and Haplusterts at the great group level in U.S. soil taxonomy, while limestone soils were Paleustolls [26].

To sample fuels, trees, and understory vegetation on each of the 25 plots, we followed the National Park Service fire monitoring handbook [28]. In summary, in each 20 m × 30 m (600 m²) plot, we measured the number of intercepts of fine (herbaceous) and woody fuels and depth of the Oi (undecomposed litter) and Oea (duff) organic soil horizons along four 15 m long transects. From the intercepts and diameters measured for sound and rotten pieces of woody fuel, we calculated average woody fuel weight as kg/m² for each plot. Along two 30 m point–line intercept transects running across the bottom and top of each plot, we counted the number of points (out of 100 evenly spaced points) that intersected live and dead fine and woody fuels (overlapping fuels, if present at a point, were tabulated separately). From these transects, we calculated the frequency of fine and woody fuels.

In the entire 600 m² of each plot for each live tree or snag (≥1 cm in diameter), we recorded the species and trunk diameter at 1.4 m, or at the root collar where the trunk arose from the root system near ground level, for trees that branched below a height of 1.4 m (mostly Juniperus). We also counted the number of seedling-size live individuals (<1 cm in diameter at 1.4 m). We randomly selected trees of each conifer taxa (Pinus and Juniperus) by size class (≥ 1 < 10 cm, ≥ 10 < 20 cm, and ≥20 cm in diameter at a coring height of 0.4 m) from which to collect an increment core using 5 mm diameter increment borers. We sought to collect at least two cores per size class for each species on each plot. If there were insufficient trees within a plot, we collected cores from trees within 20 m of plots or accepted fewer trees from a plot. Cores were mounted, sanded increasingly fine using up to 2000-grit sandpaper, and years to pith was estimated using a pith locator if cores missed the pith. We used the imaging software CooRecorder 9.4 (Cybis Elektronik & Data AB, Saltsjöbaden, Sweden) to scan cores at a resolution ≥94 dots/mm to estimate the age of each tree. Cores are assumed to represent the minimum ages of trees, not including the time required to reach coring height, which averaged 9 years in one study [20].

To sample understory communities, we counted the number of shrubs by species in a 5 m × 30 m (150 m²) transect across the center of each plot [28]. We used cover classes (0.1, 0.5, 1% intervals to 5% cover, thereafter 5% intervals to 100%) to categorize the aerial cover of each vascular plant species rooted in 1 m × 1 m quadrats (10 per plot, with 5 evenly spaced along the 30 m centerline and 5 along the northern plot line). We sampled in mid-to-late summer to correspond with peak plant cover, and to avoid excluding earlier flowering species (especially annuals), we included senesced annual–biennial plants in cover categorizations. Nomenclature and the classification of species by growth form (e.g., annual forb) and native or non-native status to the U.S. follow the Plants Database [29].

We sampled site factors on each plot including topography and soils. From the center of each plot, we used a clinometer to measure slope gradient and aspect. As the maximum slope gradient was only 25% and most (84%) plots had slope gradients ≤ 10% (hence minimal or no slope aspect), we did not include these variables further in analyses. On the western and eastern edges of each plot (1 m outside of plots), we collected three subsamples (each 1 m apart) of the 0–5 cm mineral soil using a 4 cm diameter metal corer. Each subsample was 63 cm³, totaling 378 cm³ for the 6 subsamples/plot which were composited for each plot. We sieved air-dried samples through a 2 mm sieve and measured the weight of coarse fragments (gravel > 2 mm in diameter) and bulk density including volume occupied by gravel. On the < 2 mm fraction, we analyzed samples for texture (hydrometer method) and pH and electrical conductivity (both using 1:1 soil/H₂O extracts [30]). Compared with soil derived from basalt, limestone soils were coarser (sandier with more gravel), had lower bulk density, and had higher electrical conductivity and pH (Table S2).

2.3. Statistical Analysis

We prepared a database including eight habitat variables (time since fire and number of fires and soil variables) and 18 univariate response variables in three categories including fuels, tree structure, and understory vegetation (Table S1). We derived further habitat variables using cluster analysis and principal component analysis (PCA). In addition to the binary classification of soil parent material (basalt, limestone), we derived a three-category fire activity variable (high, low, unburned) by applying cluster analysis (Euclidean distance and Ward’s linkage method, 7.8% chaining, and 6006 sum of squares) to the fire variables of time since fire and number of fires a plot had received. We then applied PCA (cross-products matrix computed using correlation) separately to the fire and soil variables to generate PCs for use in the analyses described below (Figures S1 and S2).

We conducted complementary multivariate analyses on the fuel, tree, and understory response variables. For each of the categories of response variables, we ordinated them using PCA and included habitat variables (including the PCs described above) as potential correlates with ordination axes. Next, we performed canonical correlation to examine associations between each group of response variables with habitat variables. To both reduce multicollinearity and exceed sample sizes (N > 3× total variables) relative to the number of variables recommended by McGarigal et al. [31], we performed analyses on subsets of the variables by removing strongly correlated (Pearson r > |0.70|) variables within the respective response and habitat variable sets, retaining the habitat and response variables most strongly correlated with each other (Table S3). We analyzed understory species composition data (relative cover as cover of species_i/Σ cover of all species on each plot) using non-metric multidimensional scaling ordination (Sørensen distance, random starting coordinates, and 250 runs with real and randomized data). We then used multi-response permutation procedures (Sørensen distance, n_i/Ʃ n group weighting) to compare species composition across fire activity and soil parent material categories. To identify individual species associated with these categories, we used indicator species analysis [32]. We performed ordinations, cluster analysis, multi-response permutation procedures, and indicator species analysis in PC-ORD 7.10 [33]. We performed canonical correlation in SAS 9.4 [34].

3. Results

3.1. Tree Age Distribution

At the coring height of 0.4 m, 82% of trees sampled at study sites were younger than 150 years old and had reached coring height after the 1870s, approximately the time of settlement (Figure 3). Over half the sampled trees reached coring height after the 1920s. Only seven (5%) of the 127 sampled trees were in the oldest age class of ≥200 years, corresponding to reaching coring height before 1822.

3.2. Ordinations

3.2.1. Fuels

In the PCA ordination of fuels, variation occurred along a primary gradient of fuel depth (Oi and Oea organic horizon thickness with loadings of 0.57 and 0.62, respectively) and woody fuel weight (0.51 loading) on PC1 that accounted for 38% of the total variance in fuels (Figure 4). The secondary gradient was litter frequency (0.77 loading for woody and 0.56 for fine litter frequency) on PC2 representing 24% of variance. Increasing fuel amounts and frequencies were correlated with coarser soils, lower soil bulk density, and higher soil electrical conductivity, all generally associated with limestone soil. Fire variables were less closely associated with variation in fuels, with an exception being that the number of fires and woody litter frequency were positively associated, as also reflected in bivariate correlation (Pearson r = 0.47; Table S3).

3.2.2. Tree Structure

A primary gradient of variation in tree structure in the PCA ordination included the percent of dead Juniperus trees (0.59 loading) and live Juniperus trees/ha (−0.49) along PC1 (35% of total variance; Figure 5). A secondary gradient portrayed variation in the percent of dead Pinus trees (−0.56 loading) and densities of Pinus (−0.49) and Juniperus (−0.49) seedlings along PC2 (27% of variance). The percentage of dead trees of both tree genera increased with increasing fire activity, which was also associated with limestone soil. Higher live tree densities were associated with less or no fire activity. Tree seedling density was negatively associated with higher soil pH, also indicated in bivariate correlations (Pearson r = −0.44 for Juniperus and −0.71 for Pinus; Table S3).

3.2.3. Understory Plant Communities

For univariate understory variables in the PCA ordination, PC1 extracted 36% of the total variance and portrayed a gradient of native forb species richness (−0.54 loading), shrub density (0.52) and cover (0.49), and native perennial grass cover (−0.28; Figure 6). Habitat variables most strongly associated with PC1 included time since fire, the number of fires, soil electrical conductivity, and Juniperus live tree density and % dead. Supporting the PC1 gradient, bivariate correlations showed that native forb species richness declined as the number of fires increased (and with recency of fire) and as Juniperus % dead trees increased (Table S3). Conversely, shrub density increased as the number of fires increased and as the Juniperus % dead increased (Table S3). Native forb species richness and shrub density were themselves negatively correlated (Pearson r = −0.55, Table S3).

The second gradient (PC2: 23% of total variance) in the understory PCA ordination reflected variation in non-native Bromus tectorum cover (−0.68 loading) and native forb cover (−0.57). In bivariate correlation analysis, B. tectorum cover increased with more recent fire activity, a higher percentage of dead trees, and higher pH generally associated with subsets of limestone soil (Table S3).

Non-metric multidimensional scaling ordination revealed the separation of species composition of understory communities according to fire activity and soil parent material (Figure 7). The two-axis ordination had a stress of 23.7 and represented 55% of the variance in the original data (30% for Axis 1 and 25% for Axis 2). A variety of understory species (e.g., Purshia mexicana, Bouteloua gracilis) and univariate understory metrics (e.g., forb species richness) were correlated with understory species compositional variation. For example, communities with higher relative cover of the major native perennial grasses B. gracilis and Koeleria macrantha were associated with higher native forb species richness. Numerous fuel, tree, soil, and fire variables were also associated with community compositional variation. For example, sandy, gravelly soil, an increasing number of fires, and higher percentage of dead Juniperus trees were associated with shrub-dominated communities (e.g., Artemisia tridentata).

The multi-response permutation procedure tests supported the separation of community compositional groupings portrayed in the ordination according to gradients of fire activity and soil parent material. Each fire activity category (high, low, unburned) and soil parent material (basalt, limestone) displayed unique community composition (Figure 7 inset).

3.3. Indicator Species Analysis of Understory Communities

Several understory species were significantly associated with fire activity categories and soil parent material types (

Table 1. Indicator species analysis assessing species associated with fire activity groups and types of soil parent material in Grand Canyon-Parashant National Monument, Arizona, USA. Separately for the density of shrubs and cover of all vascular plant growth forms, the group that a species was significantly (indicator value ≥ 50 with p < 0.05) associated with is shown. Indicator values range from 0 (no association of a species with any group) to 100 (maximum association). All indicator species were native, except for the non-native Bromus tectorum (*).

Species	Group	Indicator Value (p)
––––––––––––––––––––––––––––––––– Shrub density ––––––––––––––––––––––––––––––––
Fire activity
Gutierrezia sarothrae	High	54 (0.041)
Purshia mexicana	High	78 (<0.001)
None	Low
None	Unburned
Soil parent material
Artemisia tridentata	Limestone	64 (0.039)
Purshia mexicana	Limestone	89 (<0.001)
None	Basalt
–––––––––––––––––––––––––––––––––––– Cover ––––––––––––––––––––––––––––––––––––
Fire activity
Artemisia tridentata (S) 1	High	61 (0.016)
Bromus tectorum * (AG)	High	85 (0.001)
Juniperus osteosperma (T)	Low	69 (0.008)
Bouteloua gracilis (PG)	Unburned	57 (0.017)
Erigeron divergens (BF)	Unburned	55 (0.025)
Hymenoxys cooperi (PF)	Unburned	73 (0.002)
Menodora scabra (PF)	Unburned	68 (0.002)
Phlox spp.	Unburned	63 (0.007)
Soil parent material
Artemisia tridentata (S)	Limestone	57 (0.008)
Bromus tectorum * (AG)	Limestone	80 (0.002)
Allium acuminatum (PF)	Basalt	55 (0.003)
Bouteloua gracilis (PG)	Basalt	66 (0.003)
Hymenoxys cooperi (PF)	Basalt	100 (<0.001)
Hymenopappus filifolius (PF)	Basalt	52 (0.029)
Koeleria macrantha (PG)	Basalt	62 (0.010)
Lupinus brevicaulis (AF)	Basalt	70 (0.003)
Menodora scabra (PF)	Basalt	64 (<0.001)
Penstemon linarioides (PF)	Basalt	65 (0.010)
Phlox spp. (PF)	Basalt	65 (0.014)

¹ AF, annual forb; AG, annual grass; BF, biennial forb; PF, perennial forb; PG, perennial grass; S, shrub; and T, tree.

). For shrub density, Gutierrezia sarothrae and Purshia mexicana were associated with high fire activity, with P. mexicana also associated with limestone soil. No shrubs were associated with low or no fire activity. For cover, no native herbaceous species were associated with any fires but rather were associated with unburned plots (e.g., the perennial grass Bouteloua gracilis and perennial forb Menodora scabra). In contrast, the non-native annual grass Bromus tectorum was associated with high fire activity. Results for fire activity were qualitatively similar when only two groups at a time were compared. For example, when only high fire activity and unburned plots were compared with each other, the only significant indicator species among native herbaceous plants were for unburned plots.

3.4. Canonical Correlation Between Sets of Variables

Canonical correlation extracted dominant gradients within sets of habitat and response variables and revealed that combinations of habitat variables accounted for a third to over half of the variance in sets of response variables (

Table 2. Canonical correlation analysis of habitat variables with sets (fuel, trees, understory) of response variables in pinyon–juniper communities, Grand Canyon-Parashant National Monument, Arizona, USA.

Variate	Habitat 1	Response 1	Canonical Correlation
	–––––– Fire-soil ––––––	––––––– Fuel –––––––
1	EC 0.90	Fine litter 0.75	0.82 ± 0.07 2
	No. fires 0.63	Woody litter 0.73
	26% 3	23%	15%
2	Gravel 0.85	Oea 0.79	0.79 ± 0.08
	Sand 0.74	Fuel weight 0.59
	22%	29%	18%
	–––––– Fire-soil ––––––	––––––– Trees –––––––
1	pH 0.83	Juniperus % dead 0.87	0.93 ± 0.03
	No. fires 0.74	Pinus seedlings/ha −0.57
	33%	28%	24%
2	Sand 0.60	Pinus seedlings/ha 0.71	0.71 ± 0.10
	No. fires 0.56	Juniperus % dead 0.42
	19%	17%	9%
	–––– Fire-soil-trees ––––	–––––– Understory ––––––
1	No. fires 0.86	SR native forb −0.83	0.92 ± 0.03
	EC 0.62	Shrubs/ha 0.63
	38%	38%	33%
2	Gravel −0.64	Bromus tectorum cover 0.81	0.71 ± 0.10
	Juniperus trees/ha −0.49	SR shrub −0.66
	20%	35%	23%

¹ The habitat and response columns show their variables followed by their correlations with their own canonical variates. The two variables exhibiting the largest |r| are shown for each variate in the respective habitat or response variates. Oea is duff thickness, EC is electrical conductivity, and SR is species richness. ² Mean ± standard error of the canonical correlation. ³ Rows with percents display the standardized percent variance portrayed by their own canonical variates for their own variables (habitat and response variables separately) and the standardized percent variance in the response canonical variates accounted for by the habitat canonical variates (canonical correlation column). Totaling up the % variance for variates 1 and 2 represents the cumulative variance portrayed for the respective habitat or response variates and for the response variates by the habitat variates in the canonical correlation column.

). Within the set of soil–fire habitat variables, joint gradients in soil chemistry (electrical conductivity and pH) or coarseness (sand and gravel content) and the number of fires represented about half of the variance in habitat variables within the first two variates.

Within fuel response variables, the first variate represented a gradient in litter frequency and the second represented fuel depth and weight, cumulatively portraying 52% of the variance in fuels. Fire and soil variables accounted for 33% of the variance in fuels.

For the tree structure response variables, their variates portrayed combinations of gradients in the percent of dead Juniperus and the density of live Juniperus trees and Pinus seedlings. Soil and fire variables also accounted for 33% of the variance in tree structure.

Within the set of understory variables, variate 1 portrayed a gradient of forb species richness and shrub density and variate 2 non-native Bromus tectorum cover and shrub species richness. Native forb richness and B. tectorum cover showed opposing relationships with the variates, consistent with a negative bivariate correlation (Pearson r = −0.55) between B. tectorum cover and native forb richness (Table S3). The understory variates cumulatively represented 73% of the variance in the understory variables. In turn, a set of habitat variables including the number of fires, soil electrical conductivity and coarseness, and Juniperus tree density portrayed 56% of the variance in understory variables.

4. Discussion

4.1. Study Limitations

Four of the main limitations of our study included its retrospective nature, the comparison of only two soil parent material types, challenges with cross-dating trees of the study species, and the one-time sampling limiting inference of potential inter-annual fluctuations in community characteristics. We sought to accommodate these limitations in several ways. First, the retrospective nature of our study and limited information on the behaviors of many fires (e.g., wildfires burning before observations) precluded us from differentiating among different types of fires. We sought to address this by grouping plots into broader fire activity categories, focusing on the degree of fire activity a site experienced within the last 43 years rather than on responses to specific types of fires. Future research could further differentiate the effects of particular types of fires, such as fire seasonality [35], including experimentally applying fire treatments across soil parent materials or environmental gradients to assess potential interactions [36,37].

The second limitation was that we only compared two soil parent material types, and there are many more in American Southwest conifer forests [38]. However, the two parent materials (basalt and limestone) we compared predominated in our study area [39] and are broadly distributed in southwestern conifer forests, e.g., [40,41,42]. Moreover, by collecting and analyzing continuous soil variables, we were able to represent variability in soil properties within and across the broad parent material types that could be further explored. For example, increasing amounts of chert within a matrix of limestone can lower pH and modify other properties of soils derived primarily from limestone [26].

The third limitation was challenges also extensively noted in the literature with cross-dating Juniperus trees, resulting in many studies including ours to focus on ring counts or minimum ages for likely partially dated trees [20,43,44]. There can also be variation among sites in the time trees require to grow to coring height [20]. This can be estimated by sampling growth rate to coring height in contemporary seedlings, but this may not match the time required in previous centuries for the oldest trees. Retrospectively determining this typically necessitates cutting cross-sections at the root collar (and comparing with the age at coring height) via the destructive sampling of old trees, which was not appropriate in our study in a national park setting. We sought to minimize limitations with cross-dating by grouping trees into 50-year age classes to enable approximate comparisons between broad time periods, such as circa pre- or post-1870s representing general land use change associated with settlement in the study region [24].

The fourth limitation of our study was that it was a “snapshot” of plant community features at one time and was not intended to portray potential inter-annual fluctuations in plant community conditions. Annual plants may be among the community components we examined most sensitive to inter-annual fluctuations, and we sought to address this by at least including senesced annuals in our community assessments. Beatley [45] noted that senesced annuals in a semi-arid climate can remain standing for 1–2 years. Precipitation in our 2022–2023 sampling was near average, within 16% of the 36-year average.

4.2. Long-Term Changes in Tree Structure

The analysis of tree age structure, combined with prior research [13,14,19], suggests that the study area was more open with fewer trees in the 1800s and experienced a marked increase in tree density in the 1900s. Compared with Pinus ponderosa trees readily recording surface fire scars, the difficulty of reconstructing fire history in pinyon–juniper communities complicates interpreting whether the tree density increase may represent ongoing recovery from a large historical wildfire or climatic or anthropogenic influences associated with settlement [46]. While Ireland et al. [24] found that surface fires were frequent before the 1870s in P. ponderosa forests adjacent to our study area, these fires did not necessarily burn through pinyon–juniper sites and further work would be required to attempt fire history reconstruction there. Recovery of a mature pinyon–juniper community from stand-replacing fire can be a multi-century process because of lags (often 30+ years) in tree recruitment due to factors such as a lack of seed sources, few nurse plants to facilitate seedlings, and the semi-arid climate rarely conducive to tree establishment [44,47,48].

Another scenario of change is that at least moderately frequent historical fires (e.g., multiple fires/century) limited tree establishment to only portions of landscapes and that mosaics of pinyon–juniper communities were interspersed with shrublands or herbaceous-dominated patches [18]. After settlement with both livestock grazing and fire suppression potentially reducing fire activity, perhaps coupled with reduced herbaceous competition with tree seedlings or climatic changes, tree establishment increased [12]. The processes of infilling and expansion of trees in historically more open habitats have been observed across millions of hectares of western North America [49]. Further research to reconstruct historical reference conditions and ranges of variability across environmental gradients [4] would be required to understand long-term change in pinyon–juniper communities in our study area. Our results do, however, support the notion that a major increase in tree density occurred over the last 150 years, forming the basis for the National Park Service beginning to use prescribed fires and managed wildfires to reduce tree density under a management goal of reestablishing more open habitats (Figure 2).

4.3. Variation in Community Characteristics Along the Fire Activity Gradient

The major variation in plant community features along the fire activity gradient was that as fire activity increased, the percentage of dead trees, total shrub density and cover, and non-native Bromus tectorum cover all increased, while the density of live trees and native forb species richness decreased. Although modeling individual tree survival was beyond the scope of our study, we found that the percentage of dead trees was higher on plots burned the most times. Effects of repeated fires on tree mortality in western conifer forests have varied substantially with factors such as the amount of fuel accumulation between fires affecting fire severity, tree population density-dependent effects hinging on how much initial fires reduce tree density, and additional disturbances such as insects or droughts, e.g., [50,51]. Modeling fire-related individual tree mortality is a research need in pinyon–juniper communities and could help refine predicting the effects of single or multiple fires varying in characteristics [52].

Shrub density and cover increased as fire activity and the percentage of dead trees increased. A common, though not universal, pattern reported in the literature for pinyon–juniper communities after stand-replacing fires is that shrub abundance can initially decline within the first post-fire decade, then increase to or exceed pre-fire levels at 50+ years after a fire before declining again as trees predominate [48,53,54]. After multiple fires, shrub response can vary with factors such as fire severity, tree canopy cover or the duration that trees have been at increased density on a site if long-term tree dominance has increased, and traits of the constituent shrub species [21]. In our study, indicator species analysis revealed that no shrub species were associated with low or no fire activity. In contrast, Gutierrezia sarothrae was associated with high fire activity, consistent with post-fire increases in this species previously observed in pinyon–juniper communities by Huffman et al. [15,44]. Gutierrezia sarothrae is often a relatively short-lived (4–7 years) prolific seed producer that can form dense stands [55,56]. Another shrub species we found associated with high fire activity, the evergreen Purshia mexicana, was previously described as having markedly lower density and frequency 13–15 years post-fire in burned areas relative to unburned pinyon–juniper communities in Arizona [57]. When using cover, we found that Artemisia tridentata was also significantly associated with high fire activity. One of the most extensively studied shrub species in pinyon–juniper communities, A. tridentata, has generally declined in the first 1–12 years after fires, then often increased to or exceeded pre-fire abundance, at least if not competitively reduced by trees, e.g., [58,59,60]. Artemisia tridentata regenerates only via seed and can produce copious but short-lived seed crops, making the species vulnerable to declines if fires are frequent or sufficiently large to limit seed dispersal [61]. These observations exemplify the utility of further autecological investigations of species’ natural histories for planning and potentially better predicting the influences of burning in different seasons, frequencies, and severities.

The negative relationship of native forb species richness with fire activity and no forb species associated with fire in indicator species analysis was surprising to us, given that forbs were fire-stimulated in several earlier studies in pinyon–juniper communities [53,54,55,62] and in other western North American forests, e.g., [63,64,65]. There are several possibilities as to why forbs were limited at sites with a history of fires in our study. It is possible that the minimum time since fire of five years in our study did not detect an earlier post-fire increase if one had occurred but dissipated. The post-fire increase in forbs can be quite transient in just the early years post-fire in pinyon–juniper communities [53,54,55]. However, Bates et al. [17] found that forb abundance remained elevated for 4–8 years following prescribed fire. Similarly, Everett and Ward [62] found that post-fire forb increases were sustained for at least five years. It is possible that weather during our 2022–2023 study period was not favorable for forbs, although precipitation was slightly above average at 116% both years. Another possibility is that forbs were competitively suppressed by shrubs and non-native annual grasses, which were both associated with high fire activity. Prior correlational studies have found negative associations between the non-native annual Bromus tectorum and forbs in burned pinyon–juniper communities, e.g., [66,67]. A competition experiment revealed that B. tectorum competitively suppresses many native forbs [68].

We found that increasing cover of Bromus tectorum was associated with high fire activity, recent fire occurrences, and higher percentages of dead trees. Post-fire chronosequence studies in pinyon–juniper communities have reported that B. tectorum cover was highest in younger burns and declined by six [53] or eight years post-fire [54]. Experimentally imposing prescribed fire also revealed that B. tectorum abundance increased within five years relative to pre-fire or unburned areas, sometimes following an initial lag [15,16,49]. Additionally, after wildfires in the Great Basin Desert in Utah, Ott et al. [66] found that native herbaceous species (especially forbs) increased the first post-fire year but sharply declined in years 2–3 concomitant with rising B. tectorum cover. Collectively, these studies and our results suggest that further assessing the response of non-native annual grasses to variations in fire management strategies, as well as how control treatments (e.g., herbicides) may modify the responses, is warranted to determine if reducing non-native annual grasses can enable native herbaceous species to respond more positively to fires.

Other than litter frequency, variation in ground fuels (O horizon depth and woody fuel weight) was more closely associated with soils than with fire variables. A minimal relationship between fuel depth and fire activity could relate to the balance of fires consuming the O horizon while also transferring material to the ground in our study’s 5+-year post-fire period, such as needle drops from damaged or dead trees [69]. This may also account for the observed greater litter frequency in our study where fire activity had been greatest. Total fuel loads are primarily driven by heavy pieces of coarse woody debris [70]. The creation of coarse woody debris is influenced by the degree of tree mortality and how long snags remain standing [44]. In California, Reed et al. [69] found that 33% of Pinus monophylla snags fell within five years of tree death, though remaining snags could persist much longer. Owing to slow wood decomposition in the semi-arid climate, woody ground fuel loads can primarily vary with the degree of consumption in future fires [70].

4.4. Variation Along the Soil Parent Material Gradient

On coarse soil with higher pH or electrical conductivity (generally limestone soil), the percentage of dead Juniperus trees, non-native grass cover, and shrub density increased. In contrast, the density of Pinus and Juniperus seedlings and species richness of native forbs decreased. Finer-textured soils supported more native herbaceous plants, which generally occurred on basalt soil with less fire activity.

Based on our soil analysis coupled with prior research, primary differences between soils derived from the two broad parent material types include moisture availability, mechanical properties such as shrink–swell activity, and nutrient availability [36,41,71]. With an average of 398 mm/year of precipitation, our study area was just above a commonly cited threshold of 370 mm below which the inverse soil texture effect can occur [72]. In arid regions receiving less than 370 mm of annual precipitation, sandy soil can promote infiltration into deeper layers to reduce evaporative loss from surface soil, resulting in greater water availability and plant growth relative to finer-textured soil [72]. With increasing precipitation and humidity, these benefits on sandy soil can be outweighed by the lower water-holding capacity of sandy soil and loss of water to deep drainage, resulting in coarser soil generally being driest in moister climates [72].

Soil parent material and fire could have interacted in several ways in the co-varying soil–fire gradient context of our study. Fire activity was more prevalent on limestone soil, a function of greater fuel loads likely resulting in greater wildfire spread and prescribed fires under a National Park Service vegetation management goal emphasizing re-establishing open stands. As an example of soil parent material potentially modifying fire outcomes for plant growth, limestone soil can be less susceptible (compared with basalt soil with more clay) to the frost heaving of tree seedlings [42,73]. However, the greater mortality of fire-susceptible tree seedlings [50] from more prevalent fire activity on limestone soil could account for the lower density of tree seedlings inhabiting limestone soil.

In our study, we were able to examine variation across co-varying soil–fire gradients. Partitioning the relative causal influences of soil and fire would require experimental approaches we encourage in future research, given the diversity of soil types experiencing fire in pinyon–juniper communities. In prior experimental research in higher elevation Pinus ponderosa forests in Arizona, we found that effects of treatments uniformly applied across soil parent materials varied among soils derived from limestone, basalt, and benmorite [36]. Similar experimentation could be a next research step to explore the predictability of fire responses across diverse soil types in pinyon–juniper communities.

4.5. Considerations for General Patterns of Vegetation Dynamics

Our finding of associations of heterogeneity in community features with joint soil–fire gradients adds to the global literature focused on variability across dual environmental and disturbance gradients. As one of the major types of environmental gradients, soil parent materials can lead to the genesis of a diversity of soils [74]. Understanding how soil parent materials can interact with disturbances and shape vegetation responses to them continues to be an active area of research globally, including as climates and disturbance regimes change [1,3,6]. For example, in Belgium, Verstraeten et al. [8] found that species composition and the diversity of plant communities recovering from forest cutting varied across soil parent materials differing in pH. In Japan, rates of forest recovery following landslide disturbances were higher on soil derived from colluvium [75]. Considering fire disturbances specifically, Staver et al. [3] found that fire activity was greater on moister finer-textured soil, where herbaceous biomass was also higher than on drier sandier soil in South African savannas. In Canadian boreal forests, pathways of post-wildfire community change varied across gradients in soil drainage and organic matter content [9,76,77]. This sampling of studies highlights the variety of ways different soil gradients can influence vegetation dynamics and post-disturbance community conditions among global habitats. These observations support linking environment and disturbance as jointly or independently varying gradients for studying the diversity of plant community dynamics, e.g., [2,5,7].

4.6. Applications for Vegetation Management

Variation in plant community features across gradients in soil and fire could be applied several ways to vegetation management. For example, knowledge of this variation could be used to prioritize management actions on sites responding most positively to management activities or that are most deviated from reference or desired ecological conditions [6,36]. In our study, for example, the high cover of the non-native annual Bromus tectorum was associated with limestone soil and high fire activity. This annual is one of the highest-priority non-native plants for management because it can alter fuels and fire behavior, compete with native plants, and reduce overall carrying capacity for wildlife [19]. Control treatments for B. tectorum in our study could be prioritized on limestone soil and where fires have recently occurred or are planned. On the other hand, early detection monitoring and the treatment of small incipient populations could be prioritized on basalt soil and where fire activity was less prevalent. Variation in the longevity of fire effects on fuels and community features across soils is another key consideration. The persistence of fire effects can vary with factors such as fire severity and rates of tree recruitment [78]. For example, we found that tree seedling density was negatively correlated with coarse-textured soil and higher pH, suggesting that soil-driven seedling density could affect how long post-fire stands remain open.

More generally, our results illustrate the potential for incorporating heterogeneity in plant communities across the physical geography template of landscapes to improve vegetation conservation and management. We provide three examples. First, reference conditions, or baseline conditions against which to set or evaluate management effectiveness, can vary across soil types and geomorphic features [4,5,20,43]. A next step in this research in our study could be reconstructing pre-settlement forest structure and multi-century woodland change across soil types. Second, management treatments could be customized to the unique ecological communities and anticipated responses spanning spatial biophysical gradients across the landscape [2,6,7,74]. Different soil types could have variable management needs and may respond heterogeneously to treatments [1,22,36]. Third, given differences in baseline conditions and the anticipated magnitude of change, metrics for monitoring vegetation management effectiveness could be tailored to quantitative levels scaled to communities differing in productivity, structure, and diversity across soil types [8,9,75]. Each of these three examples highlight how further research augmenting the existing literature could advance incorporating variation in physical geography across the landscape as a framework for understanding and managing vegetation diversity.