Ecotypic differences in drought‐coping ability in an endemic California oak

Abstract

The frequency and severity of drought in the Western United States have significantly increased. California endemic blue oaks (Quercus douglasii) are predicted to be negatively impacted by extreme drought and are already experiencing dieback in the driest areas of their distribution. To explore whether there is drought‐adaptive variation among blue oak populations, we conducted a greenhouse common garden drought experiment with seedlings from five sites along a range‐wide precipitation gradient. We investigated seedling performance under experimental drought (fluorescence/maximum fluorescence [Fv/Fm], stomatal conductance, and leaf desiccation). We measured physiological and morphological traits, including average leaf area, specific leaf area (SLA), leaf margin morphology, C:N, and carbon isotope discrimination (δ¹³C, a proxy for water use efficiency). We used generalized linear mixed models to understand the relationship between seedling performance and traits, and the mean annual precipitation (MAP) of the seedling source site. We found significant relationships between source site MAP and Fv/Fm, percent of green leaves, and plant stomatal conductance, with plants from drier source sites performing better under the experimental drought. Seedlings from drier sites also had a lower C:N ratio, consistent with adaptation to more arid environments. Our study points to population‐level variation in seedling drought adaptation. Climate‐forward conservation strategies that conserve or leverage drought‐adaptive genetic resources from the species' threatened dry range edge could support climate change resilience in a future drier environment.

INTRODUCTION

Local adaptation in plants has long been a focal point of study for ecologists and is of particular interest today because of its relevance to climate-adaptive conservation. Early common garden studies reported that plant traits can differ across populations of the same species in varying habitat types (Turesson, 1922). Early reciprocal transplant experiments also found that climatic selective pressures lead to distinct populations often referred to as “ecotypes” (Clausen et al., 1941; Turesson, 1922), and more recent studies have identified genetic bases for ecotypic differences (e.g., Milano et al., 2016; Todesco et al., 2020). Meta-analyses of plant local adaptation show that local adaptation occurs regularly in plants but is not persistently found across all species studied (Leimu & Fischer, 2008; Oduor et al., 2016). Local adaptation to climate in plants can often center around trade-offs in resource-conservative versus resource-acquisitive strategies that can confer contrasting advantages in resource-rich versus resource-poor environments. Plants with the resource-conservative strategies of slower growth, higher investment in leaves, lower specific leaf area (SLA), and higher drought tolerance tend to thrive in resource-scarce environments, including drier climates. In contrast, resource-acquisitive strategies, such as faster growth, higher SLA, and higher competitive ability, confer advantages in more resource-rich conditions, including wetter climates (Chapin et al., 1993; Díaz et al., 2016; Pérez-Ramos et al., 2013; Reich, 2014; Wright et al., 2004).

While local adaptation has historically conferred an advantage to a local population, rapid, directional climate change may create mismatches between new climate conditions and preexisting local adaptations, causing local maladaptation to emerge (Browne et al., 2019). Trees are often locally adapted, showing moderate to strong adaptation leading to population differentiation across climatic gradients (Howe et al., 2003; Savolainen et al., 2007). Therefore, with rapid climate change, they are at higher risk of emergent maladaptation than many other taxa (Aitken et al., 2008; Alberto et al., 2013; Leimu & Fischer, 2008; Savolainen et al., 2007). Additionally, trees' typically long generation times and dispersal limitations limit their capacity to adapt to rapid climate change through natural gene flow (Aitken et al., 2008; Anderegg et al., 2012; Hamrick, 2004; Parmesan & Yohe, 2003; Vessella et al., 2017). Therefore, certain tree species may need help tracking their climatic niche (Moran, 2020), particularly where limited gene flow and/or habitat fragmentation constrain their natural evolutionary response (Savolainen et al., 2007). Although the concept of local adaptation has historically indicated the use of local seed sources in restoration (McKay et al., 2005), rapid climate change could require strategies such as genetic rescue of climate-threatened lineages and the use of more diverse or climate-adaptive seed sources for effective restoration and conservation.

The Western United States increasingly faces severe drought, negatively impacting numerous tree species and ecosystems (Allen et al., 2010; Miao et al., 2009), including California oaks (Das et al., 2020; McLaughlin et al., 2020; Park Williams et al., 2013). We chose the endemic California blue oak (Quercus douglasii) as a study species because of its foundational role in woodland ecosystems (Pavlick et al., 1991), vulnerability to climate change (Brown et al., 2018; Kueppers et al., 2005), with projections that it will lose over half of its current range by the end of the century, and recent extensive drought-related die-offs concentrated at the xeric range edge (Brown et al., 2018; Das et al., 2020; McLaughlin et al., 2020). To better understand how and whether local adaptation will affect blue oak's response to climate change and to inform climate-adaptive management, we need information on how and whether populations are differentially adapted to climate across the species' range.

Trees often show population-level differences consistent with resource-acquisitive versus resource-conservative trade-offs (Chapin et al., 1993), responsive to local resource constraints. These differences can be due to local adaptation or phenotypic plasticity. Valley oak (Quercus lobata) (MacDonald, 2017), Cork oak (Quercus suber) (Ramírez-Valiente et al., 2010), and Holm oak (Quercus ilex) (Gratani et al., 2003; Limousin et al., 2012) all have shown substantial plasticity in leaf traits. Ecotypic and genetic adaptation to climate also has been reported in oaks. The genetic structure of valley oak adult populations is associated with climate (Sork et al., 2010). Furthermore, populations of valley oaks across California showed a general, species-wide response to water stress but differed in regional gene module expressions (Mead et al., 2019). Cork oak displayed high population-level differentiation in leaf traits in a common garden, in which populations from dry sites had lower SLA, greater elastic adjustment in response to drought stress and leaf nitrogen concentration, indicating higher drought tolerance (Ghouil et al., 2020; Ramírez-Valiente et al., 2010, 2014). Furthermore, in common garden studies of northern and southern edge cork oak populations, southern ecotypes invested more in deeper root systems (potentially facilitating water uptake from deeper soil layers), while northern edge populations had higher aboveground biomass allocation, showing faster resource acquisition (Matías et al., 2019).

Current evidence for local adaptation in blue oak, our study species, is mixed. Planting experiments found that local first-year seedlings had higher survival and growth than nonlocal seedlings, indicating local adaptation (McLaughlin et al., 2022; Rice et al., 1997). However, other studies on adults in a long-term common garden found high within-population heterogeneity but limited evidence for local adaptation in drought tolerance traits (Anderegg et al., 2023; Skelton et al., 2019) or phenology (Papper & Ackerly, 2021). Based on this, Papper (unpublished doctoral dissertation) concluded that blue oaks likely have historically had high gene flow among populations.

Because the seedling stage is often a crucial and sensitive life history stage for oaks (Davis et al., 1991; Mclaughlin & Zavaleta, 2012; Tyler et al., 2008) and where local adaptation may manifest, we studied seedling responses to experimental drought. We focused on drought responses because of the frequent water limitation in Mediterranean oak woodlands and the projected increase in drought frequency and severity in our study region (Diffenbaugh et al., 2015; Flint & Flint, 2012; Williams et al., 2015). To evaluate whether there are ecotypic differences in blue oak drought response that align with source site precipitation, we conducted a greenhouse experiment with an experimental drought on plants grown from seed collected from five sites across a distribution-scale climate gradient (Figure 1). This controlled greenhouse experiment can complement existing field common garden studies by minimizing environmental variation and forcing drought stress to reveal population-level characteristics that might be expressed only in very dry conditions. We predicted that after controlling for maternal effects and plant size, seedlings from drier source sites would perform better under the experimental drought than those from the wetter sites. We predicted that seedlings from drier source sites would sustain green leaves longer throughout the experimental drought. We predicted that as drought conditions increased, seedlings from drier source sites would have higher stomatal conductance (gs) (the rate at which carbon dioxide moves from the atmosphere into the leaf interior, which plays a critical role in plant water regulation; Damour et al., 2010). We predicted that as the experimental drought progressed, seedlings from drier source sites would maintain higher Fv/Fm (maximum variable fluorescence/maximum fluorescence) values. Fv/Fm is a potential indicator of plant water stress that can be used to interpret plant response to drought (Li et al., 2008; Valladares & Sánchez-Gómez, 2006; Zhuang et al., 2020). We also hypothesized that seedlings from drier sites will have a higher percentage of survival than those from the wetter sites. We also hypothesized that plants from drier sites would show a suite of traits typically associated with resource-conservative strategies and higher drought tolerance, including lower SLA, higher leaf δ¹³C, higher leaf carbon to nitrogen ratios, lower leaf area, lower growth, and higher physical leaf defenses. We have summarized these traits and response measures, their expected correlation with source site precipitation, rationale for including each measure, and relevant citations in Table 1. Finally, we hypothesized that overall, seedlings from drier sites would have higher survival than those from wetter sites.

[IMAGE OMITTED. SEE PDF]

TABLE 1 Description of traits or response measures expected to correlate with source site precipitation, rationale, and citations.

Response measure/trait	Expected correlation with source site precipitation	Rationale	Citations
Fv/Fm	Higher Fv/Fm values in source sites with lower MAP	Low Fv/Fm values can be an indicator of higher plant stress; seedlings from drier sites may be more drought adapted and less stressed under experimental drought	Valladares and Sánchez-Gómez (2006), Li et al. (2008), Zhuang et al. (2020)
Percent of green leaves	Higher % of green leaves in source sites with lower MAP	Plants in drier environments can have multiple adaptations to drier conditions that lead to lower leaf loss	Del Arco, Escudero and Garrido (1991), Puigdefábregas and Pugnaire (1999)
% Survival	Higher % survival in source sites with lower MAP	Plants in drier environments can have multiple adaptations to drier conditions that lead to higher survival	Lombardini and Rossi (2019)
Stomatal conductance	Higher stomatal conductance in source sites with lower MAP	Oaks in drier environments can be adapted to maintain photosynthesis under dry conditions	Dickson and Tomlinson (1996), Fotelli et al. (2000)
Carbon:nitrogen ratio	Higher C:N ratio in source sites with lower MAP	High C:N ratio is an indicator of greater resource investment in individual leaves, thereby increasing leaf longevity, a trait often associated with drier environments	Nobel (2009)
Leaf area	Lower leaf area in source sites with lower MAP	Reducing leaf area is a common drought tolerance mechanism as it can limit the transpiration surface area of a leaf and thereby reduce water loss	Clauw et al. (2015), Carlson et al. (2016), Cruz et al. (2019)
Specific leaf area	Lower SLA in source sites with lower MAP	Plants from resource-poor environments often invest more in individual leaves and have lower SLA than those in resource-rich, water-abundant environments	Ramírez-Valiente et al. (2014), Pérez-Harguindeguy et al. (2016)
Leaf margin morphology	Higher leaf spikes in source sites with lower MAP	Plants may invest more in leaf physical defenses in dry conditions	Coley et al. (1985)
Leaf δ13C	Higher leaf δ13C in source sites with lower MAP	Plants in drier sites may have higher leaf δ13C, a proxy for higher water use efficiency	Luong et al. (2021)
Plant height	Lower plant height in source sites with lower MAP	Drier site oaks may invest less in aboveground growth and more in root systems to reach groundwater	Dickson and Tomlinson (1996), Grünzweig et al. (2008); Matías et al. (2019)

METHODS

Study system

Blue oak is a foundational species in blue oak woodlands and is endemic to California. It is widespread within its narrow range and is winter-deciduous. It covers approximately 1.2 million hectares of mostly rangeland across the foothills of California's Coast Ranges and the western Sierra Nevada (Pavlick et al., 1991). It has a high biodiversity value because of its capacity to enhance soil quality, increase fertility across various habitats, and sequester relatively large amounts of carbon (Dahlgren et al., 2003; Pavlick et al., 1991). It provides a habitat for important reptiles and amphibians (Block & Morrison, 1998) and mammals (Pavlick et al., 1991). It is also an important cultural species for many Native American tribes (Anderson, 2005).

Seed source sites and collection

In October 2020, we collected acorns from five sites in California across the blue oak distribution's precipitation gradient (Table 2; Figure 1). Our study sites ranged from protected areas to private ranches (Table 2). We collected acorns from within blue oak-dominated woodlands. California oak woodlands are known to have extensive microclimatic variation (McLaughlin et al., 2017), which may locally influence the moisture availability to individual trees. We extracted average climate data from 1981 to 2010 (Table 2) for each seed source site from the California Basin Characterization Model (Flint et al., 2013, 2021).

TABLE 2 Seed source sites with mean annual precipitation (MAP), climate water deficit, maximum summer and winter temperatures, and minimum temperatures averaged over a 30-year period (1981–2010).

Site	Site abbreviation	MAP (cm)	Climate water deficit (mm)	Maximum temperature (°C)	Minimum temperature (°C)
Hopland Research Extension Center	HREC	100	739.8	31.2	1.8
Pepperwood Preserve	PWR	86	794.2	28.4	1.7
Hastings Natural History Reservation	HR	53	878.9	27	1.7
Sedgwick Reserve	SR	38	939.1	31.4	1.4
Toll House Ranch	TH	25	1213.7	34.3	1.6

We collected 25–30 ripe acorns from multiple branches from 14 to 16 randomly selected trees at each seed source site (Figure 1; Table 2) based on availability (total acorns collected: Hopland Research Extension Center [HREC], 494; Pepperwood Preserve [PWR], 397; Hastings Natural History Reservation [HR], 520; Sedgwick Reserve [SR], 499; and Toll House Ranch [TH], 382). We stored them in a refrigerator at approximately 4°C for 3 months before planting them in the greenhouse. We conducted a floating viability test of acorns similar to Pons and Pausas (2007) and discarded nonviable samples. We weighed each acorn before planting and standardized weights by selecting acorns of similar sizes across sites and populations (Appendix S1: Figure S1). We harvested the acorns directly from the trees and not from the ground and did not sterilize them.

Greenhouse experimental design and planting protocols

In February 2021, we began an experiment in UC Santa Cruz's Coastal Campus greenhouses. We implemented a randomized block design that comprised 25 blocks, with 10 seedlings in each block. We planted 50 acorns from each of our five seed source sites (250 acorns total, two acorns from two maternal trees per site in each block) in autoclaved potting soil (Sunshine Mix). We planted the acorns in plastic pots: D40 size (25.4 cm length and 6.35 cm width), selected for their capacity to accommodate the root growth of oak seedlings. We watered the seedlings weekly to field capacity for 6 months. Subsequently, we initiated an experimental drought in September. We stopped watering half of the blocks (125 seedlings) (“experimental drought”) and continued to water the other half at field capacity (“well-watered”) for 6 months.

Seedling performance measurements—Experimental drought seedlings

We measured Fv/Fm, degree of leaf desiccation, and stomatal conductance on the experimental drought seedlings only. We also measured aboveground biomass to include in models to account for the effects of plant size on performance measures indirectly (Appendix S1: Figure S2). We measured variable fluorescence over maximum fluorescence (Fv/Fm) for 4 weeks (Days 7, 15, 23, and 28 [Appendix S1: Figure S3]) using a fluorometer (OS1p Modulated Fluorometer, Opti-Sciences) until the plants were so desiccated that the instrument did not record any values. For consistency, plants were dark-adapted overnight before pre-dawn Fv/Fm assessments to minimize the impact of diurnal variations (Murchie & Lawson, 2013). We measured around the same time (6:30–7:30 a.m.) before sunrise on all days of measurements.

We measured stomatal conductance (gs) and soil moisture between 11:00 a.m. and 1:00 p.m. on Days 9 (n = 25) and 32 (n = 20) of the experimental drought. We measured one leaf per plant using a porometer (SC-1, Decagon Devices, Pullman, WA, USA). We kept the porometer in the greenhouse 2 h before measurements, and it was calibrated before every use. To contextualize stomatal conductance measurements, which can be highly dependent on soil water potential, we measured soil moisture in each pot directly prior to measuring stomatal conductance (Appendix S4: Figure S4) using time domain reflectometry (TDR). We used a 20-cm Hydrosense II probe (Campbell Scientific, Logan, Utah, USA). We inserted the probe pair vertically at two opposite sides of each pot and averaged the two measurements. Then, to calculate soil water potential, we conducted a calibration for the greenhouse soil samples using methods described in Jupa et al. (2021). We started by completely drying the soil in a drying oven and then mixed the dry soil samples with an exact amount of water to create a gradient in gravimetric water content. We used 10 soil samples between 0.1 and 1 g g⁻¹. Then, we mixed the soil samples thoroughly with water added and left them closed in separate boxes for 48 h to equilibrate the soil moisture. We made sure we had a sufficient amount of soil to be able to completely stick the probe of TDR into it. We used conical pots covered with a wrap. After 48 h of soil equilibration, we measured soil conductivity with TDR and water potential using a dewpoint meter (WP4C, Meter Group, USA) in each container. Then, we plotted the water potential or gravimetric water content against TDR responses (in microvolts) and fit the data with a power function to generate an equation. Then, we used that equation to calculate soil water potential for all the TDR values that we took during the experimental drought in the greenhouse.

To represent the amount of leaf desiccation and record the timing of senescence of each plant, we counted the total number of leaves and the number of fully desiccated leaves per plant on Days 29 (n = 68) and 33 (n = 68) of the experimental drought. We considered plants “functionally dead” when they had no photosynthetically active leaves and loss of stem flexibility in the upper one-third of the stem (Valladares & Sánchez-Gómez, 2006). Then, we collected the aboveground biomass, dried it in the oven for 48 h at 60°C, and measured the dry mass. A small number of leaves fell in the greenhouse before these measurements took place and, therefore, were not accounted for in the direct aboveground biomass measurements. To correct this, we estimated fallen leaf biomass by multiplying the site-specific average dry leaf mass from the well-watered plants by the number of unaccounted-for leaves for each experimental drought plant (based on the total number of leaves before desiccation began). We then added this value to the directly measured aboveground biomass for an estimated value of total aboveground biomass.

Seedling trait measurements—Well-watered seedlings

To avoid destructive sampling of the experimental drought seedlings, we measured traits, including physiological traits (leaf C:N and leaf δ¹³C) and morphological or structural traits (SLA, average leaf area, plant height, and leaf margin morphology) only on the well-watered seedlings. For SLA and average leaf area, we sampled three leaves from each plant in November 2021. We photographed all the leaves and then measured the leaf area for each sampled leaf by analyzing digital images in the ImageJ software (Ferreira & Rasmand, 2011). We then dried the leaves in an oven set at 60°C for 48 h. Post-drying, we weighed the leaves to obtain their dry mass (in grams). We calculated each leaf's SLA by taking the leaf area to dry mass ratio. Finally, we computed an average SLA value for each plant.

For leaf δ¹³C and C:N ratio measurements, we collected a single leaf from each of six individuals from the driest (TH) and eight individuals from the wettest (HREC) source sites in October 2022 and freeze-dried them for two days. After the leaves were completely dry, we ground them with a mortar and pestle. We weighed and encapsulated the samples in tin and analyzed them for carbon stable isotopes and C:N ratio at the University of California Santa Cruz Stable Isotope Laboratory using a CE Instruments NC2500 elemental analyzer coupled to a Thermo Scientific DELTA plus XP isotope ratio mass spectrometer via a Thermo Scientific Conflo III. Measurements were corrected to VPDB (Vienna PeeDee Belemnite) for δ¹³C against an in-house gelatin standard reference material (PUGel), which is extensively calibrated against international standard reference materials. Measurements were also corrected for size effects, blank-mixing effects, and drift effects.

To measure plant height, we selected 11–17 plants from each seed source site (1–3 plants from each maternal tree) in October 2022. We measured the plant height from the soil to the highest leaf using a measuring tape. To measure leaf margin morphology, we selected two leaves from 5 to 6 randomly selected plants from each seed source site. Each plant was from a different maternal tree. Using a hand lens, we carefully inspected the corners of each leaf for any visible spikes and counted the number of spikes on the leaf margin. We also took into consideration the possibility of spike fractures and counted them as spikes as well. Our spike count did not include pointy leaf segments without a noticeable outward spike. We measured leaf area using ImageJ (Ferreira & Rasmand, 2011) and then calculated an average leaf spike per area. Finally, we calculated an average spike/area for each plant.

Model selection and statistical analysis

Model selection: Because we were primarily interested in whether seedling performance and traits were related to seed source site precipitation, in a series of models, we first analyzed performance and trait measurements with seed source site mean annual precipitation (MAP) as a predictor variable. If MAP was not a significant predictor, we used the seedling source site as a predictor variable to determine whether there were differences among sites unrelated to precipitation. If the results of these analyses were significant, we ran post hoc Tukey pairwise comparisons to test which sites differed using the package estimated marginal means (R package, emmeans, Length, 2023). We also tested whether acorn mass or aboveground biomass was a better predictor for our models and picked aboveground biomass as the covariate for the models based on the model performance. For response variables for which there were multiple replicates per maternal tree and site, to account for the lack of independence of samples, we used generalized linear mixed-effects models to incorporate the maternal tree and site as random effects. For the response variables where the maternal trees were all from different trees, we utilized a linear regression model.

Seedling performance measurements—Experimental drought seedlings

We chose to analyze performance measures on experimental drought seedlings as close to the third week of the experimental drought as possible (between 23 and 32 days of experimental drought) to avoid the early periods when all plants still had ample water and the later periods when most plants had more fully desiccated and were unresponsive to the instruments. We refer to this period as a “mid-stage drought.”

For the response variable Fv/Fm, we ran a generalized linear mixed-effects model with MAP and seedling aboveground biomass as fixed effects, maternal tree and site as random effects for Day 23 of the experimental drought. We used a zero-inflated model with a beta distribution and a logit link function (TH, n = 13; SR, n = 11; HR, n = 15; PWR, n = 13; and HREC, n = 8) (R package, glmmTMB; Brooks et al., 2017).

We evaluated survival and leaf desiccation with two models based on measurements taken on Day 29 of the experimental drought. First, we ran a generalized linear mixed model with a binomial distribution with a logit link function (R package, glmmTMB; Brooks et al., 2017), with the predictor variables MAP and aboveground biomass as fixed effects, maternal tree and site as random effects, and whether the plant was functionally dead or alive (plants with at least one green leaf vs. plants with no green leaves) as the response variable. To explore whether differences in the extent of leaf desiccation were related to source site climate for plants that were functionally alive, we used data from the subset of plants that had at least one green leaf for the second model. For this analysis, we ran a generalized linear mixed model with a Gaussian family with the predictor variables site, MAP, and aboveground biomass as covariates, maternal tree and site as random effects, with the percentage of remaining green leaves as the response variable (R package, glmmTMB; Brooks et al., 2017) (TH, n = 15; SR, n = 12; HR, n = 18; PWR, n = 13; and HREC, n = 10).

To determine whether stomatal conductance varied with source site MAP, and whether stomatal conductance at a given water potential varied across source sites, we constructed a generalized linear mixed model with gamma distribution and a log link function. The model consisted of stomatal conductance values on Day 32 of the experimental drought as the response variable, utilizing an interaction between site MAP and soil water potential as the predictor variables, including maternal tree and site as random variables (R package, glmmTMB; Brooks et al., 2017) (TH, n = 4; SR, n = 4; HR, n = 5; PWR, n = 4; and HREC, n = 3).

To assess whether the relationship between soil water potential and aboveground biomass differed with source site MAP, we conducted linear models with soil water potential as the response variable and an interaction between MAP and aboveground biomass as predictors for Days 9 and 32 (Day 9: TH, n = 4; SR, n = 5; HR, n = 8; PWR, n = 5; and HREC, n = 3; Day 32: TH, n = 4; SR, n = 4; HR, n = 5; PWR, n = 4; and HREC, n = 3).

Seedling trait measurements—Well-watered seedlings

To assess differences in leaf δ¹³C between the wettest and driest sites, we conducted a linear regression model using the seedling source site as the predictor variable and leaf δ¹³C as the response variable (TH, n = 6; HREC, n = 8). All plants in this analysis were from different maternal trees. To evaluate C:N ratio differences between the driest and wettest sites, we ran a linear model with leaf C:N as the response variable and MAP as the predictor variable (TH, n = 6; HREC, n = 8).

We utilized a linear mixed-effects model (R package, lme4; Bates et al., 2014) to test for differences in average leaf area among all seed source sites, with seedling source site as a predictor variable and maternal tree as a random variable (TH, n = 20; SR, n = 14; HR, n = 11; PWR, n = 13; and HREC, n = 10). We utilized a linear model to test for differences in SLA among all seed source sites, with SLA as a response variable and seedling source site as a predictor variable (TH, n = 20; SR, n = 14; HR, n = 11; PWR, n = 13; and HREC, n = 10). To evaluate whether there were differences in leaf margin morphology among sites, we ran a linear model with leaf margin morphology as the response variable and seedling source site as the predictor and leaf margin morphology per area as the response variable (TH, n = 5; SR, n = 5; HR, n = 6; PWR, n = 5; and HREC, n = 5).

We utilized linear mixed-effects models (R package, lme4; Bates et al., 2014) to test for differences in plant height among all seed source sites, with site as a predictor variable and maternal tree as a random variable (TH, n = 17; SR, n = 12; HR, n = 15; PWR, n = 11; and HREC, n = 12).

We standardized all the continuous predictor variables before running the models. For all the models, we checked model fit for underdispersion or overdispersion with the “DHARMa” package in R. We also performed residual diagnostic checks by plotting QQ plots for residuals (Hartig, 2022). We conducted all our analyses in R v4.2.0 (R Core Team, 2022).

RESULTS

Better performance of dry source site seedlings under experimental drought (experimental drought seedlings)

Seedlings from drier sites outperformed those from wetter sites during the experimental drought, as evidenced by multiple performance measures. On Day 23, seedlings from drier source sites showed lower stress levels, as evidenced by higher Fv/Fm values (z = −1.94, Pr(|z|) = 0.05; Appendix S1: Table S1). Notably, the average Fv/Fm values of seedlings from the driest site were more than fourfold higher than those from the most mesic site (Figure 2; Appendix S1: Table S1). Aboveground biomass was also found to be a highly significant predictor for Fv/Fm values (z = −3.18, Pr(|z|) = 0.00149; Appendix S1: Table S1).

[IMAGE OMITTED. SEE PDF]

Seedlings from drier sites retained their green leaves longer than those from wetter sites, as measured on Day 29 of the experimental drought (Figure 3). Of the plants that were not totally desiccated, individuals from drier sites maintained a higher percentage of green leaves than those from wetter sites (z = −1.98, Pr(|z|) = 0.05; Appendix S1: Table S3).

[IMAGE OMITTED. SEE PDF]

Seedlings from wetter sites were more likely to be fully desiccated than those from drier sites, with more than 80% of the plants from the driest source site (TH) but only 30% of those from the wettest source site (HREC) remaining functionally alive on Day 29 after cessation of watering (z = −2.6, Pr(|z|) = 0.009; Figure 4; Appendix S1: Table S2).

[IMAGE OMITTED. SEE PDF]

On Day 32 of the experimental drought, seedling stomatal behavior was different among source sites. Seedlings from drier sites had higher stomatal conductance than those from wetter sites (z = −2.24, Pr(|z|) = 0.02; Figure 5; Appendix S1: Table S4). We also saw a marginally significant interaction between MAP and soil water potential (z = −1.7, Pr(|z|) = 0.08; Figure 5; Appendix S1: Table S4 and Figure S5).

[IMAGE OMITTED. SEE PDF]

Seedlings from drier sites did not retain higher soil moisture for a given biomass on Day 9 (t = −1.004, Pr(|t|) = 0.32; Appendix S1: Figure S6, Table S5) or Day 32 (t = −0.84, Pr(|t|) = 0.41; Appendix S1: Figure S7, Table S6) indicated by our finding of no significant interaction between MAP and aboveground biomass effects on soil water potential on either day.

Seedling trait differences with seed source site climate (well-watered seedlings)

Leaf C:N ratio in well-watered plants differed between seedlings from the driest and wettest sites, with the mean leaf C:N ratio at the driest site less than half of that at the wettest site (t = 2.25, Pr(|t|) = 0.044; Figure 6; Appendix S1: Table S7).

[IMAGE OMITTED. SEE PDF]

Neither average leaf area, SLA, nor leaf margin morphology in well-watered plants varied significantly with source site MAP or among seedling source sites (Appendix S1: Figures S8–S10, Tables S8–S10). Leaf δ¹³C was not significantly different between seedlings from the driest and wettest sites (Appendix S1: Figure S11, Table S11). Plant height in well-watered plants varied among sites, but the post hoc test with pairwise comparison did not show any differences among the site pairs (Appendix S1: Figure S12, Tables S12 and S13).

DISCUSSION

We found ecotypic differences in various performance measures and physiological and morphological characteristics in blue oak seedlings from populations across a wetter to drier climatic gradient, consistent with an argument that blue oaks are locally adapted to precipitation. Overall, we observed better performance under drought in seedlings from drier sites, and many, though not all, of the plant traits we measured varied as expected with source site MAP. We found a significant relationship between source site MAP and Fv/Fm, percent of green leaves, and stomatal conductance in drought-exposed plants during mid-stage drought. We also found a significant relationship between source site MAP and leaf C:N ratio in seedlings grown under well-watered conditions. However, we did not find a significant relationship between source site MAP and average leaf area, SLA, carbon isotope discrimination (δ¹³C, a proxy for water use efficiency), leaf margin morphology, or aboveground biomass. We found significant differences in aboveground biomass (plant height) among sites, but the differences did not correspond to source site MAP.

Performance measures—Experimental drought seedlings

Seedlings from drier sites exhibited significantly elevated Fv/Fm ratios during mid-stage drought, indicating robustness against drought-induced stress. Different-sized plants may have faced different levels of drought at a given date during the experimental drought, and, as such, the time course of Fv/Fm values reflects both precipitation and size effects. Nevertheless, a strong signal of source site MAP emerged, indicating that when accounting for size, Fv/Fm is still influenced by source site conditions (Appendix S1: Figure S13). The variability in Fv/Fm values observed within our source sites may be partially explained by the high variation in microenvironmental conditions in blue oak woodlands (McLaughlin et al., 2017).

Fv/Fm represents the maximum fluorescence/variable fluorescence (Krause & Weis, 1991; van Kooten & Snel, 1990). Fv/Fm represents a plant's quantum yield and can be used as one potential indicator of plant water stress (Lichtenthaler & Miehé, 1997; Urban et al., 2017) and physiological state (Krause & Weis, 1991; Maxwell & Johnson, 2000; Ogaya et al., 2011; Urban & Alphonsout, 2007). Stressors such as drought, high light, and heat stress impact PS II, reflected by low Fv/Fm values (Demmig & Björkman, 1987; Demmig-Adams & Adams, 2018; Long et al., 1994). Fv/Fm has been shown to reflect impaired photosynthetic efficiency or photoinhibitory damage as drought exposure progresses (Epron et al., 1992; Lichtenthaler & Rinderle, 1988; van Kooten & Snel, 1990; Zhuang et al., 2020). Numerous studies have observed significantly reduced Fv/Fm values under drought stress conditions (Epron et al., 1992; Li et al., 2008; Tribulato et al., 2019; Valladares & Sánchez-Gómez, 2006; Zhuang et al., 2020). Under optimal well-watered conditions, most plants exhibit an Fv/Fm ratio value between 0.78 and 0.85 (Demmig & Björkman, 1987). Fv/Fm below approximately 0.75 can represent damage to a plant's photosynthetic apparatus, which tends to correspond to other forms of physiological damage, such as xylem embolism and cell necrosis (Johnson et al., 2022; Tonet et al., 2023).

Leaf desiccation is another performance indicator, and differential timing among populations in leaf drying in response to water stress can indicate differences in seedling drought-coping capacity. During mid-stage drought, seedlings from drier source sites maintained a significantly higher percentage of green leaves than those from wetter sites, indicating robustness against drought-induced stress. Further, there was a large difference in apparent seedling survival at this time: only 20% of seedlings from the driest source site, but 70% of those from the wettest source site were functionally dead. Prolonged drought may limit photosynthesis as plants close their stomata to reduce water loss, which could lead to leaf loss. Although, leaf desiccation and abscission can be a drought avoidance strategy in some situations (Kaproth et al., 2023; Skelton et al., 2021).

Seedlings from the driest site (TH) showed distinct stomatal behavior by maintaining higher stomatal conductance at lower soil water potentials during mid-stage drought and at lower soil water potentials (marginal significance), while the populations from wetter sites showed lower stomatal conductance during mid-stage drought at lower soil water potentials (marginal significance). All populations had overlapping soil water potential levels, but only the driest TH site showed some upregulation of stomatal conductance under low soil moisture values. However, we interpret this result with caution because of the small number of plant samples and marginal significance.

Under water stress, plants often reduce stomatal conductance to conserve water and mitigate drought-induced damage (Hessini et al., 2008). However, this reduces their capacity to photosynthesize and grow and exposes them to photooxidative stress. Stomatal closure can also increase leaf or canopy temperature as latent heat transfer associated with transpiration is reduced (Urban et al., 2017). Maintaining stomatal conductance despite drought allows a plant to continue photosynthesis and can be characteristic of drought-adapted oak species (Fotelli et al., 2000). Interspecific studies on oaks show that under drought-stressed conditions, more drought-adapted species kept stomata open, maintaining carbon fixation, as opposed to less drought-adapted oaks, which closed their stomata sooner or at higher levels of soil moisture (Dickson & Tomlinson, 1996). Stomatal behavior under drought stress can also differ among oak populations. Our results align with the work of Kubiske and Abrams (1992), which found ecotypic differences in Northern red oak (Quercus rubra), where drier source site seedlings experienced a lesser reduction in stomatal conductance than wetter source site seedlings under drought conditions. Plants from drier sites, accustomed to surviving in drier conditions, may have photosynthetic apparatus better equipped to tolerate higher levels of tissue water deficit than those accustomed to wetter conditions. As expected, all seedlings in our experiment reduced stomatal conductance over time. However, the delay in shutting down stomatal conductance in the seedlings from drier sites allowed for continued metabolism and could explain the higher percentage of green leaves and the higher Fv/Fm values in the drier site seedlings many weeks into the experimental drought.

Physiological traits—Well-watered seedlings

We measured a suite of leaf traits on the well-watered seedlings. Leaf C:N ratio (the ratio of nonstructural carbohydrate C metabolites to total nitrogen metabolites in the cell) can relate to plant drought tolerance in various ways (Martin et al., 2002). We expected that plants from drier source sites would have a higher leaf C:N ratio based on the assumption that a high C:N ratio is an indicator of greater resource investment in individual leaves, thereby increasing leaf longevity (Nobel, 2009), likely more important in low water resource environments. For example, Hu et al. (2013) found that more drought-tolerant oak species have higher leaf C:N ratios than less drought-tolerant species. However, contrary to our expectation, we found substantially lower leaf C:N ratios in seedlings from the driest source site than from the wettest source site. These differences could be explained by the various roles that high N can play in plant physiology and life history strategies. C:N balance is an important factor in the process of leaf senescence (Chen et al., 2015) and may be related to the later senescence observed in our driest source site seedlings. Alternatively, leaf C:N ratios can relate to site-specific herbivore pressure. Leaf toughness frequently correlates with higher C:N ratios (Agrawal & Fishbein, 2006), and higher C:N ratios may decrease leaf palatability (Pérez-Harguindeguy et al., 2003). Previous studies on California oaks have shown that more mesic sites had higher herbivore pressure (McLaughlin & Zavaleta, 2013). Thus, the higher C:N ratio we observed in seedlings from wetter source sites could relate to higher herbivore pressure in these sites.

We expected to find high leaf carbon isotope discrimination (δ¹³C, a proxy for water use efficiency) in drier site seedlings, indicative of higher water use efficiency (Luong et al., 2021). However, we did not find differences in δ¹³C among populations. A study of cork oak that exposed northern and southern provenance seedlings to drought found δ¹³C to be a plastic metric, as the southern provenance seedlings shifted their δ¹³C to be more water efficient when exposed to drought conditions (Matías et al., 2019). We might not have found source site differences in our study because we evaluated δ¹³C only under well-watered conditions, whereas differences in δ¹³C may emerge only under drought stress (Matías et al., 2019). However, it is not straightforward to interpret as there are many factors to consider. We need to understand the difference in efficiency of photosynthetic machinery among the driest and wettest site seedlings to completely interpret these findings as water use efficiency is also a function of photosynthetic capacity, which could certainly differ between these two populations.

Morphological traits—Well-watered seedlings

Reducing leaf area is a common drought tolerance mechanism for plants as it can limit the transpiration surface area of a leaf and thereby reduce water loss (Carlson et al., 2016; Clauw et al., 2015; Cruz et al., 2019). This pattern has also been shown in Northern red oak (Q. rubra), where xeric ecotypes had significantly smaller leaves than mesic ecotypes (Kubiske & Abrams, 1992). Anderegg et al. (2023) also found a relationship between MAP and leaf size, where sites with higher precipitation had a higher leaf area of blue oaks when compared with those in the lower precipitation sites. However, in our study, we did not see any differences in leaf size among seedlings from different seed source site precipitation. While population differences were not detected in our well-watered seedlings, they might emerge under drier conditions.

Seedlings with faster growth may be able to compete better with grasses or avoid shading for optimal growth (Jensen et al., 2011). Further, dry-adaptive oaks may invest less in aboveground growth and more in root systems to reach groundwater, particularly in dry sites or drought conditions (Dickson & Tomlinson, 1996; Grünzweig et al., 2008; Matías et al., 2019). However, contrary to our hypothesis that aboveground growth would be higher in seedlings from wetter source sites, the plant height in our seedling populations was unrelated to source site precipitation. Because our seedlings were grown in small pots with likely root restriction, we could not effectively evaluate root biomass allocations and assess differences in root-related traits.

Also, contrary to our hypothesis that SLA (a measure that indicates leaf thickness and density) would be lower in seedlings from drier source sites, we did not see differences in SLA among populations. SLA is considered an important leaf trait that determines the investment in structural leaf defenses and higher leaf life span (Cunningham et al., 1999; Pérez-Harguindeguy et al., 2016). Plants from resource-poor environments often invest more in individual leaves and have lower SLA than those in resource-rich, water-abundant environments (Pérez-Harguindeguy et al., 2016; Ramírez-Valiente et al., 2014). Multiple oak species from xeric environments had lower SLA than oaks from mesic environments (Kaproth et al., 2023). However, for blue oaks, SLA might be a trait fixed across populations. Alternatively, it may be that SLA is responsive to drought (Luong et al., 2021) but did not differ among populations under the well-watered conditions in which we took our SLA measurements. Finally, leaf margin morphology did not differ by site and was unrelated to the source site MAP. Local influences of herbivory pressure, which could influence the development of leaf spikes, could not be studied in the greenhouse setting.

CONCLUSIONS

Overall, our results support the hypothesis that ecotypic variation and local adaptation to precipitation occur among blue oak populations, consistent with previous findings (Mclaughlin et al., 2020). These results differ from the conclusions of other studies that showed little drought-adaptive local adaptation in adult blue oaks based on findings from a common garden experiment in the mesic part of the blue oak distribution (Anderegg et al., 2023; Papper, unpublished doctoral dissertation; Papper & Ackerly, 2021; Skelton et al., 2019). These differences likely stem from the fact that our study focused on the relatively highly drought-sensitive seedling stage (Mahall et al., 2009; Matzner et al., 2003) rather than the adult stage, at which point selective pressures may be lower. Our study also isolated the effects of drought conditions through experimental drought in a way that would not have been possible at a relatively wet field site. Although we attempted to limit maternal effects by controlling for acorn mass, completely ruling out their influence would require a second generation, which is prohibitive in long-lived species like blue oaks. To address this issue, we recommend that future research relate differences in phenotypic traits among populations to genetic differences, using techniques such as generation and landscape sequencing, gene expression profiles, and epigenetic analysis (Sork, 2015).

The consistently higher drought tolerance in seedlings from the driest source sites indicates that dry range-edge populations could be important for climate adaptation in a more drought-prone future. The drier site seedlings exhibited a suite of performance measures and traits under drought conditions, including maintaining higher Fv/Fm values, higher conductance, higher leaf C:N ratios, and a higher percentage of green leaves than the wetter site seedlings. However, we did not see expected differences in traits in the well-watered seedlings. Together, these findings suggest that genetic-by-environment interactions may be important to understanding drought management strategies in blue oaks.

Blue oaks, like many other Western tree species, are already experiencing increased dieback at the driest sites of their distribution (Das et al., 2020; McLaughlin et al., 2020; Park Williams et al., 2013). This is a serious concern given that drought-adaptive genetic resources, which could be important for climate-adaptive restoration across the species distribution (Hampe & Petit, 2005; Razgour et al., 2013), appear to be concentrated in these declining regions. In the near term, threatened populations of dry-adapted ecotypes should be targeted for genetic rescue before in situ populations are lost.

AUTHOR CONTRIBUTIONS

Sushmita Poudel was involved in conceptualization, writing the original draft, data collection, visualization, data analysis, and funding acquisition. Erika Zavaleta was involved in conceptualization, writing—review and editing, project administration, methodology, and funding acquisition. Blair Mclaughlin was involved in writing—reviewing and editing the original draft, methodology, conceptualization, and project administration.

ACKNOWLEDGMENTS

For funding, we thank the Langenheim fellowship and Plant Science Fund from the Ecology and Evolutionary Biology Department UCSC, Howard-Kohn Scholarship, The Marin Chapter, California Native Plant Society, and and the PEO fellowship. We thank the staff at Hastings Reserve, Blue Oak Ranch Reserve, Sedgwick Reserve, Pepperwood Preserve, the Hopland Research and Extension Center; the UCSC greenhouse staff including Sylvie Childress and Laura Pamer; The Nature Conservancy and Zac Principe; The Bureau of Land Management and Ryan E O'Dell. We thank research assistants Wendy Olvera, Jessica Gallardo, Olivia Wilms, Juniper Allen-Cantu, Razi Lederman-Beach, Sonali Bhandari, and Pramod Pandey; and the Conservation Science and Solutions Lab, especially Kelly Zilliacus, Tim Brown, Kathryn Bernier, Abe Borker, and Reza Goljani Amirkhiz. We also thank Jarmila Pitterman for helpful advice and comments on the ecophysiological parts of the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT

Data (Poudel et al., 2025) are available from Dryad: .

References

Agrawal, A. A., and M. Fishbein. 2006. “Plant Defense Syndromes.” Ecology 87: S132–S149. [DOI: https://dx.doi.org/10.1890/0012-9658(2006)87%5B132:pds%5D2.0.co;2].