1. Introduction
The Mediterranean Basin is a climate warming and biodiversity hotspot, where aridification trends have been observed, which are expected to be magnified by warmer conditions during the late 21st century, negatively impacting its diverse woody flora [
1]. In this region, droughts during the late 20th century and early 21st century have been among the most intense of the past millennium [
2]. Such warmer and drier conditions are particularly affecting forests dominated by conifers (pines, firs, cedars, and junipers, among others), triggering dieback episodes, reducing productivity, and increasing mortality rates [
3,
4]. The long-term effects of climate and drought on the growth of these gymnosperm tree species have been identified through dendroecological and ecophysiological analyses [
5,
6,
7,
8], but we lack comparative analyses on coexisting gymnosperm trees and shrubs (however, see [
9]). This gap of research is very relevant because in the driest regions of the Mediterranean Basin and other semi-arid regions, treeless steppe-like landscapes are dominated by shrub species, often forming shallow roots with poor access to deep soil water, which may be sensitive to a severe winter-to-spring water deficit if they show an early growth onset [
8]. Previous studies have compared radial growth in coexisting tree and shrub species inhabiting Mediterranean, dry areas [
9,
10,
11,
12], but most of them were angiosperms, except for some juniper species (
Juniperus phoenicea L.), which showed widespread drought-induced dieback and mortality [
9,
12,
13]. Additional research is required to compare how climate, and particularly low soil moisture water availability, influence the growth of co-occurring tree and shrub gymnosperms.
Gymnosperms from the Mediterranean Basin and other dry and semi-arid regions comprise diverse growth habits from trees (e.g., pines) to small shrubs (e.g., some junipers) [
14]. This taxonomic group includes conifers (Pinophyta; e.g., Pinaceae and Cupressaceae families), but also other morphologically varied antique groups, such as Gnetophyta, where the Ephedraceae family is included. This family consists of the
Ephedra genus, which comprises about 60 vessel-bearing species, occurring as shrubs or vines and inhabiting dry sites in tropical, subtropical, and temperate areas from the northern and southern hemispheres [
15]. Furthermore,
Ephedra have been particularly successful in colonizing dry habitats on rocky and sandy substrates thanks to adaptive trends, such as high water-use efficiency, elevated conductivity through their wide vessels, and high photosynthesis rates [
16,
17]. They also present plastic hydraulic traits, including shifts in wood anatomy (e.g., shifting from vessel-bearing to nearly vessel-less rings) [
17,
18]. Other adaptive wood-anatomical responses to prevent xylem cavitation include changes in the number of vessel groupings and the degree of helical thickening. For instance,
Ephedra species growing in cold regions form vessels and tracheids with more pronounced helical thickenings [
17,
18]. Therefore, it could be expected that in
Ephedra shrubs from dry and semi-arid regions, which form annual rings [
19], radial growth is constrained by drought because wide earlywood conduits are more prone to cavitation and would lose hydraulic conductivity, thus reducing the cambial activity. To the best of our knowledge, few dendrochronological studies have been carried out with
Ephedra species from arid and semi-arid regions (however, see [
20,
21]).
Regarding juniper species, they are very successful pioneer species in dry habitats because of their small lumen area, bimodal radial growth pattern, and shallow roots, which make them able to rapidly exploit superficial water pools [
8,
9,
12]. These traits make junipers potentially better adapted to withstand drought stress than taller trees forming wider tracheids, such as pines, since junipers may experience a lower chance of a xylem embolism, whereas pines may be better able to grow during dry periods by exploiting deeper water pools.
Here we aim to compare the intra- and interannual radial growth patterns and the year-to-year growth and climate variability in four coexisting gymnosperms showing different growth habits that inhabit semi-arid Mediterranean regions: two trees (Pinus halepensis, Junipers thurifera) and two shrubs (Juniperus phoenicea, Ephedra nebrodensis). We evaluate whether the species’ growth responses to climate vary as a function of site dryness by comparing dry vs. very dry sites.
2. Materials and Methods
2.1. Study Sites and the Tree and Shrub Species
The study sites were situated in the semi-arid Middle Ebro Basin, near the piedmonts of the “Sierra de Alcubierre” range, Aragón, northeastern Spain. Two mid-elevation dry sites were located in the northern (Lanaja) and southern (Monegrillo) sides of this small range situated within a steppe landscape (see
Figure 1), whilst the other low-elevation, very dry site (Peñaflor) was located near the inner Middle Ebro Basin (
Table 1). The Monegrillo site was located on the south-oriented steep slopes of the “Sierra de Alcubierre” range, whereas the Lanaja site was located on gentle slopes of the north-oriented, wet side of this range. Both sites present slightly different climatic conditions and vegetation due to differences in elevation and orientation. All sites have basic soils formed by marls and gypsum, which were more abundant in the very dry site.
The study sites experienced continental Mediterranean climate conditions characterized by cold winters, dry summers, and relatively wet conditions in spring and autumn (
Figure 1). The vegetation was dominated by
Pinus halepensis Mill. (
Figure 1a) stands with scattered junipers (
Juniperus thurifera L. (
Figure 1b),
Juniperus phoenicea L. (
Figure 1c), and
Juniperus oxycedrus L.), some evergreen oaks (
Quercus coccifera L.,
Quercus ilex L.), and shrubs (
Rhamnus lycioides L.,
Salvia rosmarinus (L.) Sheild.,
Genista scorpius (L.) DC.,
Globularia alypum L.,
Lynum sufruticosum L., and
Thymus spp.). These shrub species are angiosperms, except for
J. phoenicea, but one major gymnosperm shrub was also present in this dry maquis-type shrubland,
Ephedra nebrodensis Tineo ex Guss. (hereafter referred to as
Ephedra (
Figure 1d)).
According to data from nearby climatic stations (Lanaja, 0°19′44″ W, 41°46′19″ N, 369 m a.s.l.; Monegrillo, 0°24′57″ W, 41°38′19″ N, 432 m a.s.l.; Zaragoza-Aula Dei, 0°48′37″ W, 41°43′54″ N, 231 m a.s.l.), the annual precipitation ranged between 463 (Lanaja dry site) to 353 mm (Peñaflor very dry site). The mean annual temperature ranged between 8.4 °C (Monegrillo dry site) to 13.8 °C (Peñaflor very dry site) (
Table 1). The mean minimum and maximum temperatures ranged from −9.2 to 38 °C. The annual climatic water balance (the difference between the precipitation and potential evapotranspiration) ranged from −410 mm (Lanaja dry site) to −620 mm (Peñaflor very dry site). The water balance was negative from April to September and reached the minimum values in July and August. The relative air humidity varied from 33 to 97%. Frosts can occur from November until March with a mean frequency of 13–26 days per year. Radiation fogs linked to high pressures are also frequent in the Ebro Basin from November to February (during December and January, the mean frequency of foggy days is 24% per month), and the fog layer is usually 300 to 350 m thick [
22].
2.2. Field Sampling
Field sampling was carried out during January (Ephedra) and October (rest of the species) in 2020. We sampled dominant, apparently healthy individuals (15–35 individuals per species) of the four study species. We measured their basal diameter and total height using tapes and a laser rangefinder (Nikon Forestry Pro II, Tokyo, Japan), respectively. In the case of trees, the diameter was measured at 1.3 m. In Ephedra, the height was measured with tapes.
We took two cores per individual at 1.3 m using Pressler increment borers in all species, except for Ephedra, in which case, we took basal cross-sections using a handsaw. In J. phoenicea, we took cross-sections at 0.5–1.0 m of 5–10 individuals per site to facilitate the cross-dating of cores.
2.3. Xylem Formation: Intra-Annual Growth Rates
In the very dry site (Peñaflor) we characterized the intra-annual growth rates during 2010 (Ephedra) and 2020 (the other three species) by periodically taking wood samples of five individuals per species. The 2010 and 2020 study years were relatively dry (2010, mean annual temperature 15.3 °C, annual precipitation 267 mm) and wet (2020, mean annual temperature 16.5 °C, annual precipitation 391 mm), respectively. We took small shoots of diameter 2–6 mm in Ephedra and microcores in the other species (extracted using a Trephor® microcorer (Belluno, Italy) every 14–30 days. Then, we obtained transversal sections (15–20 μm thick) using a sliding microtome (Leica SM2010 R, Leica Biosystems, Nussloch, Germany). Sections were mounted on glass slides, stained with 0.05% cresyl violet acetate, and fixed with Eukitt® (Orsatec, Bobingen, Germany) to quantify the amount of newly formed xylems (not lignified cells) which were stained as blue tissue. Images of sections were taken at 40–100x magnification with a digital camera mounted on a light microscope (Olympus BH2, Tokyo, Japan), and then we measured the width of the new xylems to obtain radial growth rates using the ImageJ analysis software (ver. 1.5i, NIH, Bethesda, MD, USA).
2.4. Climate Data
We obtained long (period 1950–2020), homogeneous series of monthly climate data (mean temperature, total precipitation) from the Climate Explorer webpage [
23] by considering the 0.1° grid located over each study site. These monthly climate data were transformed into seasonal data by averaging (temperature) or summing (precipitation) the monthly data.
To compare the climate–growth relationships at different time resolutions, we also obtained weekly climate data for the period 1961–2019 from a high-resolution (1.1 km
2) Spanish dataset [
24]. Specifically, for each site, we obtained a series of mean maximum (TMx) and minimum (TMn) temperatures, climatic water balance (P-PET, difference between precipitation and reference evapotranspiration calculated using the FAO56 Penman–Monteith equation), relative air humidity (RH), and vapor pressure deficit (VPD).
Finally, we obtained estimates of the soil moisture (corresponding to the upper 10 cm), gridded at 1.00–1.25° for the period 1979–2016, corresponding to remote-sensing products of the Climate Change Initiative (CCI) dataset of the European Space Agency (ESA) [
25].
2.5. Dendrochronological Data
The wood samples were processed to calculate the year-to-year radial growth variability, quantified via the ring width, by using dendrochronological methods [
26]. Samples were air-dried and sanded with sandpapers of progressively finer grain until the annual rings were clearly visible. Then, samples were visually cross-dated by annotating narrow rings and the cross-dating was verified using COFECHA software (ver. 6.06P, Laboratory of Tree-Ring Research, The Univ. of Arizona, AZ, USA) [
27]. In all cases, two radii per individual tree or shrub were dated, and the ring widths were measured along them to the nearest 0.01 mm using a Lintab-TSAP measuring device (Rinntech
TM, Heidelberg, Germany). In the cross-sections, we measured the whole radius from the youngest ring near the bark to the oldest ring near the pith.
The resulting tree ring width data were detrended and standardized using the ARSTAN ver.44 software (Tree-Ring Lab, Columbia University, NY, USA) [
28]. Detrending allowed for removing growth trends due to changes in size, age, and stand dynamics. The detrending was done by fitting a 67% cubic smoothing spline with a 50% cutoff frequency [
26]. Then, the resulting detrended series were pre-whitened with low-order autoregressive models to remove the growth persistence. Individual, pre-whitened series of ring width indices (RWI) were combined into site mean residual series or chronologies using bi-weight robust means [
26].
Several tree ring statistics were calculated for each chronology (see
Table 2): mean ring width and its standard deviation (SD); the first-order autocorrelation (AC1) of the ring width data, which quantified the serial dependence between rings; the mean sensitivity (MS) of standard ring width indices, which measured the relative change in width between consecutive rings; the mean correlation (Rbar) between individual indexed series, which accounted for the coherence in growth variability within each species in each site [
29]. Lastly, we defined the best-replicated period of each chronology by calculating its expressed population signal (EPS), which measures how well replicated the chronology was, considering a minimum threshold of EPS ≥ 0.85 [
30].
2.6. Statistical Analyses
We checked the normality of the variables by using Shapiro–Wilk tests. To compare variables between sites we used Student’s t-tests. Pearson correlations were calculated between seasonal (mean temperatures, summed precipitation), monthly, and weekly climate variables and the species’ chronologies for the common period of 1961–2019, except in the case of the shorter Ephedra series, which encompassed the period of 1989–2019. In the case of seasonal and monthly climate variables, correlations were calculated from the prior to the current September. In all cases, we considered the 0.05 and 0.01 significance levels.
Linear mixed effect models [
31] were used to test for the differences in the ring width characteristics between sites (i.e., dry vs. very dry) and between growth forms (shrub vs. tree). Particularly, we compared the ring width, its autocorrelation between series of shrubs and trees, and between series in dry and very dry sites. To compare ring width characteristics between shrubs and trees, the chronology identity was used as a random factor to account for the fact that series were gathered in different sites. To compare the ring width characteristics between dry and wet places, species identity was used as a random factor to account for the fact that four different species were studied at the two sites. In the two cases, a constant variance structure was included to account for the fact that the number of series (and thus, the variance) varied between chronologies. Models were fitted using the nlme package [
32] in the R statistical environment ve. 4.0.3 (The R Foundation, Vienna, Austria) [
33].
We also used linear mixed effect models to test for the relationship between ring width and drought severity. The 12-month-long June Standardized Precipitation Evapotranspiration Index (SPEI) was used to assess the drought severity following previous studies [
7]. We compared the responses of residual, pre-whitened series of the ring width indices to drought severity and how it varied between sites (i.e., dry vs. very dry sites) and growth form (shrub vs. tree). The same random structure was used as explained before. We proposed models including the SPEI and its interaction with site type or growth form for the period 1989–2019. Fitted models were ranked according to their Akaike information criterion (AIC) and the most parsimonious model showing the lowest AIC was selected.
2.7. VS-Lite Growth Model
We used the VS-Lite forward growth model to assess differences in the climatic controls of tree growth between sites for the four species. The VS-Lite forward model was used to characterize the climatic drivers of radial growth [
34,
35,
36]. The model formulation contains several parameters: a growth–temperature parameter (gT) and its two subparameters determining the minimum (
T1) and optimal (
T2) growth temperatures, and the growth–soil moisture parameter (gM) and its two subparameters determining the minimum (
M1) and optimal (
M2) soil moisture [
33]. The VS-Lite simulates nonlinear growth response of the mean series of ring width indices as a function of monthly temperature and precipitation, based on the principle of limiting factors [
34]. To estimate the model parameters, we followed a Bayesian framework [
35]. The
T1 and
M1 subparameters determine when growth will occur, and the
T2 and
M2 subparameters determine when growth is not limited anymore by temperature and soil moisture, respectively. We estimated the solar radiation (gE parameter) from the site latitude by considering no interannual variability. We used these parameters to simulate the ring width indices for the 1960–2019 calibration period; then, we divided this period in two subperiods to evaluate the temporal stability of the growth responses, except in the case of the short
Ephedra series. We assumed uniform priors for the growth function parameters, and independent, normally distributed errors for the ring width indices. Then, 10,000 iterations were run using three parallel chains and a white Gaussian noise model error [
35,
36]. Snow dynamics were not considered since snowfall is rare at the study sites. To estimate monthly soil moisture from temperature and total precipitation, the model used the empirical leaky bucket model of hydrology, whilst other parameters (e.g., runoff, root depth, and growing season length) were taken from previous studies on similar sites and species [
13,
37].
4. Discussion
The intra-annual growth pattern was characterized by a major spring peak for all species and a minor autumn peak for some species (pines and junipers), which is consistent with previous xylogenesis studies [
7,
38,
39]. The different phenological patterns of growth were not associated with growth habit, but with phylogeny, since both trees and shrubs showed this bimodal pattern that has already been described in Mediterranean pines and junipers from very dry sites [
7,
38,
39]. The use of basal samples in
Ephedra could also explain its lower MS and Rbar values and affect the correlations between climate and growth, but assessing the importance of this factor would require taking basal cores in trees, which is out of the scope of this study. The unimodal growth pattern of
Ephedra at the very dry site and its early growth start (April) suggested a great dependence on prior winter precipitation and early spring soil moisture. This may be explained by the reliance on shallow soil water pools (depth < 50 cm) in xerophytic shrubs with narrow leaves, whilst tree species may be able to explore deeper soil water sources, particularly in the dry summer [
40,
41]. In the case of the pine and juniper tree and shrub species coexisting in semi-arid regions, junipers tend to use deep soil water sources during wet seasons (spring, autumn), whilst pines are more able to use deeper water pools during the dry summer [
42,
43], albeit similar short-term responses of the two species have been also observed in response to dry conditions [
44,
45]. Such varied responses agree with the existence of a dimorphic root system, which allows evergreen trees and shrubs to use shallow soil water sources after rain events and deep sources during dry seasons [
46].
The differential use of water resources and ecophysiological strategies (isohydric pine vs. anisohydric junipers) could explain the different climate–growth responses observed in coexisting junipers and pines, regardless of their growth habit. At monthly scales, the growth of the trees
P. halepensis and
J. thurifera depended on wet and cool conditions and high soil moisture in winter and spring, whereas the growth of the shrub
J. phoenicea was improved by cool June–July conditions and elevated soil moisture in July. This was also observed in the analyses at weekly scales, which uncovered the sensitivity of
J. phoenicea growth to warm and dry (low RH, elevated VPD) June–July conditions, particularly at the very dry site. The positive association of growth with July soil moisture was also observed in
J. thurifera from the dry site. The dependency of the junipers’ growth on summer soil moisture indicates a lower capacity to keep some slow growth in summer as compared with the drought-tolerant and water-saving
P. halepensis [
47]. This would explain the more bimodal growth patterns of junipers as compared with coexisting pines [
39] and a high potential to grow in wet autumns, particularly for
J. phoenicea, which showed the most delayed autumn growth peak. The
Ephedra positive response to autumn temperatures at the dry site could suggest a bimodal behavior in less stressful sites but the intra-annual growth data do not confirm this. Further research could follow the xylogenesis of this species at sites with different water availability to test that idea.
The sensitivity to warm conditions leading to a high evaporative demand make junipers prone to drought-induced dieback in response to elevated spring–summer temperatures and at sites with soils showing a low water-holding capacity [
12,
13]. The climate–growth associations of
Ephedra also make it a candidate species to show dieback and mortality in response to warm and dry summer conditions, given its sensitivity to summer soil moisture, albeit they may also form coarse roots and access moderately deep (10–50 cm) soil water sources [
48]. The response to summer climate conditions was also found for
Ephedra procera growing in Iranian deserts with ≈250 mm of annual precipitation [
20], but not in other species growing in cold mountain sites [
21]. In contrast, warm and dry winter-to-spring conditions could trigger growth decline and dieback for
P. halepensis by inducing xylem embolism, particularly in the fine roots [
4,
49,
50].
The VS-Lite results (
Figure 11 and
Figure 12) agreed with the measured growth rates (
Table 2) and, importantly, with the climate correlations (
Figure 4,
Figure 5,
Figure 6,
Figure 7,
Figure 8,
Figure 9 and
Figure 10) but allowed for better identifying the main constraints of growth, such as low soil moisture. The VS-Lite model has been applied to
P. halepensis and
J. phoenicea growth data and showed the species’ dependency on winter-to-spring and spring-to-summer cool and wet conditions, respectively [
12,
37,
51]. Our simulations illustrated the pronounced growth sensitivity of the two tree species to dry and warm conditions during the growing season, and were consistent with the higher site-to-site growth variability observed in shrubs. By comparing wet and dry years, the model forecasted a higher sensitivity of
P. halepensis to increasingly warmer and drier conditions at the very dry site, but also of
J. phoenicea and
Ephedra at the dry site (see
Figure 12). Moreover, the model detected a rapid shift toward warmer and drier climate conditions and growth constraints due to reduced soil moisture in the 1980s after the wet and cool 1970s. The different responses of trees and shrubs could be interpreted as phenotypic variability and local adaptation of the shrub species to the harsh conditions of the very dry site [
39]. A similar explanation could be applied to
J. thurifera from the very dry site, which showed lower soil moisture limitations than in the dry site. The study
J. thurifera relict stands could be locally adapted to the harsh climate (dryness) and soil (gypsum) site conditions [
39].
We followed a correlative approach by studying the climate–growth associations but it must be considered that the actual drought stress depends on the climate, tree, and stand features, including the genetic composition, site conditions, and soil water dynamics [
52,
53]. For instance, great intraspecific variability in the soil water uptake [
41] and in wood density [
54] have been observed in provenance trials of
P. halepensis, with populations from more arid regions taking up more water from deep soil layers, albeit this was not translated into improved growth. In the field, local factors, such as a higher surface rock cover, may increase the soil water concentration and mitigate the negative impact of warm and dry conditions on the growth and survival of
P. halepensis at semi-arid sites [
55]. Therefore, considering the local site factors (e.g., soil water sources and competition) and individual tree features (e.g., tree size and functional traits, such as wood density) could improve our understanding of the climate–growth associations in dry regions and refine the simulations produced by forward models, such as VS-Lite, which have been mainly applied to trees [
37,
51]. In the case of shrubs, microsite conditions should be explicitly accounted for [
10]. For instance,
Ephedra species may form longer roots and have higher soil nitrogen concentrations under the canopy than in the interspaces between plants [
48].
We argue that a better and more mechanistic approach toward understanding climate–growth responses in tree and shrub species from dry regions, such as semi-arid Mediterranean areas, should integrate functional knowledge with growth sensitivity, including indirect proxies of gas exchange, hydraulics, and soil water uptake, such as wood anatomy and isotopes [
12,
40,
56].