Volume 225, Issue 6 p. 2314-2330
Full paper
Free Access

Ray fractions and carbohydrate dynamics of tree species along a 2750 m elevation gradient indicate climate response, not spatial storage limitation

Jessie M. Godfrey

Corresponding Author

Jessie M. Godfrey

Plant Sciences Department, University of California, Davis, CA, 95616 USA

Author for correspondence:

Jessie M. Godfrey

Tel: +1 510 847 7664

Email:[email protected]

Search for more papers by this author
Jason Riggio

Jason Riggio

Department of Wildlife, Fish, & Conservation Biology, University of California, Davis, CA, 95616 USA

Search for more papers by this author
Jessica Orozco

Jessica Orozco

Plant Sciences Department, University of California, Davis, CA, 95616 USA

Search for more papers by this author
Paula Guzmán-Delgado

Paula Guzmán-Delgado

Plant Sciences Department, University of California, Davis, CA, 95616 USA

Search for more papers by this author
Alana R. O. Chin

Alana R. O. Chin

Plant Sciences Department, University of California, Davis, CA, 95616 USA

Search for more papers by this author
Maciej A. Zwieniecki

Maciej A. Zwieniecki

Plant Sciences Department, University of California, Davis, CA, 95616 USA

Search for more papers by this author
First published: 06 December 2019
Citations: 15

Summary

  • Parenchyma cells in the xylem store nonstructural carbohydrates (NSC), providing reserves of energy that fuel woody perennials through periods of stress and/or limitations to photosynthesis. If the capacity for storage is subject to selection, then the fraction of wood occupied by living parenchyma should increase towards stressful environments.
  • Ray parenchyma fraction (RPF) and seasonal NSC dynamics were quantified for 12 conifers and three oaks along a transect spanning warm dry foothills (500 m above sea level) to cold wet treeline (3250 m asl) in California's central Sierra Nevada.
  • Mean RPF was higher for both conifer and oak species with warmer dryer ranges. [Correction added after first publication 30 January 2020: In the preceding sentence ‘Mean RPF was lower’ was changed to ‘Mean RPF was higher’.] RPF variability increased with elevation or in relation to associated climatic variables in conifers – treeline-dominant Pinus albicaulis had the lowest mean RPF measured (c. 3.7%), but the highest environmentally standardized variability index. Conifer RPF variability was explained by environment, increasing predominantly towards cooler wetter range edges. In oaks, NSC was explained by environment – values increasing for evergreen and decreasing for deciduous oaks with elevation. Lastly, all species surveyed appear to prioritize filling available RPF with sugar to achieve molarities that balance reasonable tensions over starch to maximize stored carbon.
  • RPF responds to environment but is unlikely to spatially constrain NSC storage.

Introduction

Most of a tree’s biomass (72–75%) is located in the xylem of its trunk and large-diameter branches (Reich et al., 2014). These woody tissues provide mechanical strength and a path for the transport of water, nutrients and nonstructural carbohydrates (NSC) between roots and leaves. Additionally, although much of the xylem is composed of dead cells (vessels/fibers or tracheae), a variable fraction of wood adjacent to the cambium is alive and able to persist for several years (Spicer & Holbrook, 2007), dynamically managing a tree’s resources. Referred to as xylem parenchyma, this living component of the wood is credited with the timely accumulation, transformation and release of water (Jupa et al., 2016), nutrients (Helmisaari & Siltala, 1989) and NSC (Plavcová & Jansen, 2015) as some buffer from environmental fluctuation. It is reasonable to expect that the efficiency and/or size of a buffer has evolved to accommodate environmental conditions across a species’ range and may be constrained by a species’ morphophysiological limits – some interplay between competing forms (cell walls, vessels/ tracheae, parenchyma) and/or functions (structural integrity, water transport, NSC storage or mobilization). Climate change already is testing the intrinsic constraints of trees, modifying precipitation and temperature regimes across the globe (Breshears et al., 2005; Allen et al., 2010). Where precipitation is decreasing and temperature is increasing, the effects of drought on tree mortality and forest composition are compounded everywhere by associated fires (Curtis et al., 2018) and, in montane ecosystems in particular, by emerging pests (Kurz et al., 2008; Anderegg et al., 2015) and restrictions to range expansion (Germino et al., 2002; Elsen & Tingley, 2015). The frequency, severity and duration of stress events are likely to continue along current trajectories, shifting seasonal climatic patterns (IPCC, 2013) and phenologies (Parmesan & Yohe, 2003) even further from historical norms. Future forest health and composition may thus be governed, in part, by a tree’s capacity to respond to these stresses with reserves of NSC, particularly those stored in its largest storage organ, the trunk.

Distinguished by orientation along a stem’s axis (axial parenchyma) or perpendicular to a stem’s axis (ray parenchyma), fractions of the wood taken up by living cells typically vary between 5% and 10% in conifers, and between 10% and 40% in angiosperm trees (Spicer, 2014; Morris et al., 2016a). Although several recent studies have investigated variation in parenchyma fractions across angiosperm species, concluding that variation is explained by a combination of climate (Morris et al., 2016a) and phylogeny (Zheng et al., 2019), no explanations have been offered for the variability among conifers. Furthermore, there are no studies, to our knowledge, which investigate the variability of parenchyma fraction along the environmental gradients experienced by a single species. In light of the increasing severity of stress events, it may be of relevance to know whether parenchyma fraction in both conifers and angiosperms increases towards extremes of elevation or climate not only across species, but also within species, providing adequate space for the energetic (NSC) buffering that might fend off heightened stress at range edges. Of course, this hinges upon greater storage capacity following greater parenchyma fractions and greater parenchyma fractions accommodating more NSC.

Flexible in form and concentration, NSC stored in parenchyma are credited with facilitating the maintenance or recovery of xylem functionality through periods of stress. Soluble sugars help to osmo-protect attendant parenchyma cells from freeze damage (Kasuga et al., 2007; Takata et al., 2007) and balance tension in the xylem with organic solutes (Hellkvist et al., 1974; Nobel, 2009). In preserving themselves, parenchyma cells also may preserve their ability to restore hydraulic conductivity lost to embolism diurnally (Zwieniecki & Holbrook, 2002), artificially (Secchi & Zwieniecki, 2011), with freezing (Sperry & Sullivan, 1992; Sauter et al., 1996; Améglio et al., 2002; Mayr et al., 2006) or with drought (Secchi et al., 2017; Klein et al., 2018). Followed by delivery into cavitated conduits, starch to sugar conversion may be a partially self-fulfilling signal of the stem’s need to restore osmotic balance by shifting solutes from the symplast to the apoplast (Secchi & Zwieniecki, 2016).The conversion of NSC from starch to soluble sugars in the xylem parenchyma may even be part of the signal pathway that calls additional sugars from the phloem into embolized conduits (Salleo et al., 1996; 2009; Nardini et al., 2011). However, reserves of NSC in the xylem are not only temporarily engaged as osmoticum; they also may be consumed by respiration or incorporated into the carbon scaffolds of new growth. NSC pools are often depleted when carbon transport is disrupted by pests or pathogens (Wiley et al., 2015) or when photosynthesis is reduced in cold seasons (Gruber et al., 2013) or dry seasons (Klein & Hoch, 2015), as well as with historically unseasonal cold stress (Sperling et al., 2015) or drought stress (Earles et al., 2018). It follows that depletions of NSC pools also are observed when photosynthesis is absent in the night (Gibon et al., 2004; Tixier et al., 2018) and that seasonal depletions are exaggerated in species that experience dormancy, for which photosynthesis is not just reduced, but absent for a period in every year (Martínez-Vilalta et al., 2016; Furze et al., 2019). The dependency of new growth on some remnant reserves following dormancy (Piispanen & Saranpää, 2001; Tixier et al., 2017) means that depletions that begin with senescence and extend through flush must be moderate enough to, at least, allow for the development of new photosynthetic organs. The balance of NSC, climate, stress and phenology is certainly delicate and important. However, evidence that this balance is or is not constrained by space is just beginning to emerge (Plavcová et al., 2016; Kiorapostolou et al., 2019; Trifilò et al., 2019).

Trees can move only generationally. An individual tree’s likelihood of survival is thus governed by the degree to which its location is suited for the base morphophysiological adaptations of its species, among them some potential for acclimation. It follows that if an individual grows near the edges of its species’ range, it is likely to experience levels of stress that push acclimation to species limits. Assuming that, among other services (see above paragraph), parenchyma fractions provide a major space for energy storage and that the energetic costs (maintenance respiration) of parenchyma maintenance are low relative to the energetic benefits of NSC storage, we hypothesized that the increased energy demands that follow increased environmental stress would result in higher parenchyma fractions for species (both angiosperms and conifers) occupying high (cold) and/or low (dry) elevations. In a greater departure from previous work, we also hypothesized that species with narrower ranges would be forced to explore the full scope of their acclimation potential, increasing the variability indices of their parenchyma fractions and leading to observations of higher parenchyma fractions at range edges. To test our hypotheses, we measured the ray parenchyma fractions (RPF) and seasonal NSC dynamics of 12 conifers and three oaks along a continuous gradient from the warm dry foothills of the central Sierra Nevada (500 m above sea level) to cold wet treeline (3250 m asl).

Materials and Methods

Sampling design

The western slope of California’s central Sierra Nevada provided an ideal system in which to examine morphophysiological responses to climatic gradients of moisture and temperature. The range’s low elevations are warm, but dry; its high elevations are wet, but cold (Fig. 1). Furthermore, a cross-section of its ecosystem has been relatively well-preserved. Now delimited as a 180 × 35 km slice across a puzzle of privately owned ranches, BLM land, national forests and national parks, the Yosemite Grinnell transect was established to quantify vertebrate biodiversity up and over the mountain range’s central latitudes (Grinnell & Storer, 1924). The area also is home to > 40 tree and woody shrub species with overlapping vertical ranges, often extending more than 1500 m in altitude (Paruk & Morales, 1997). We selected sampling sites using the Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) and R/raster (Hijmans, 2019) such that, within the Grinnell Transect (Koo & Moritz, 2007), six points were randomly distributed every 250 m from 500 m to 3250 m above sea level (72 sites in total) and the following additional parameters were met. The transect’s eastern border was confined by the Sierra Crest. The western border was confined by access to publicly owned land; sampling did not continue below 500 m asl because of this constraint’s limitations. Half of the samples at each sampling elevation were set to fall on declination-corrected E–SE aspects and half on W–SW aspects. We chose generally southern aspects to keep our observations within the same forest communities for a given elevation along our gradient, but we also were curious about potential differences resulting from dominant morning (E–SE) or afternoon (W–SW) exposures. As the DEM (with a coarseness of 90 m2) did not always generate site placements at the target elevation or aspect, in the field we adjusted sampling site locations to within 20 m of the target elevation and the desired aspect using a hand-held global positioning system (GPSMAP 64s; Garmin Ltd, Olathe, KS, USA). For access, the last parameter was that sampling sites be ≤ 500 m from a trail or road. The final latitudinal spread of points within the transect was 37.58–37.94°N.

Details are in the caption following the image
Maps of elevation (ELEV, m above sea level (asl)) and climate overlaid with randomized sampling sites (points). The area of Yosemite National Park is filled in or outlined by light gray. The boundaries of the Yosemite Grinnell Transect are outlined with a black rectangle, this polygon representing an area 35 km wide by 180 km long. Rasters of climate variables are averages of 30-yr normal PRISM data (2019) at 800 m resolution and include mean annual precipitation (MAP), mean annual temperature (Tmean), mean temperature at dewpoint (TDmean), maximum annual temperature (Tmax), maximum vapor pressure deficit (VPDmax), minimum annual temperature (Tmin) and minimum vapor pressure deficit (VPDmin).

All species with a trunk > 10 cm diameter at breast height and present within 30 m of a site’s center point were sampled (one tree per species per site). Conifer species included Tsuga mertensiana (Bong.) Carr., Abies concolor (Gordon) Lindley, Abies magnifica A.Murray bis, Pseudotsuga menziesii var. menziesii (Mayr) Franco, Pinus lambertiana Dougl, Pinus albicaulis Engelm, Pinus monticola Dougl., Pinus contorta subsp. murrayana (Balf.) Englm., Pinus sabiniana Dougl., Pinus jeffreyi Balf., Pinus ponderosa var. benthaminana Vasey and Calocedrus decurrens (Torr.) Florin. Oak species included Quercus douglasii Hook. & Arn., Q. chrysolepsis Liebm. and Q. kelloggii Newb. As sampling sites were randomized, we did not sample the highest elevation tree-form angiosperm present within the transect, Populus tremuloides Michx. We expect that sampling would have to be targeted for this species given its sparse distribution. Also because sampling sites were selected randomly, some species, such as P. contorta and P. ponderosa (RPF replications of 24 and 29, respectively), were very well represented. Others, particularly P. menziesii and Q. douglasii (RPF replications of 6 and 4, respectively), were not as well represented. All data for RPF (sampled once) and NSC (sampled several times between Autumn 2016 and Autumn 2017) for each species included in the study appear in Supporting Information Dataset S1.

Often, RPF numbers are higher than NSC numbers. The first reason for this is that we sampled additional species present at a site for RPF if we found these species after the initial sampling date (mostly in the case that our initial sampling occurred predawn when visibility was limited). The second reason for this is that disease, drought and/or mountain pine beetles killed eight individuals over the course of sampling and we removed these individuals from NSC analysis, but not RPF analysis. The third reason is that Mariposa County’s Detwiler Fire occurred between our Summer 2017 and Autumn 2017 sampling trips – this only had impacts on data for the species present at our lowest elevation sites, P. sabiniana and Q. douglasii. These species already were only sampled partially due to the low ends of their ranges falling below our sampling cut-off at 500 m asl. As data for these species were particularly limited, we did not include P. sabiniana or Q. douglasii in calculations of variability indices (see methods below) or in any examination of within-species variation along elevation or climatic gradients.

Sampling, morphological analysis, phylogeny and variability indices of RPF

In September 2016, one 5 mm (diameter) × 6–10 cm (length) tree core was extracted at breast height from a single individual of each species present at a sampling point making efforts to avoid tension or compression wood by sampling perpendicular to any slope. Cores were then placed immediately into 15-ml centrifuge tubes filled with a 3 : 1 ethanol : acetic acid solution. At the end of the sampling day, these falcon tubes were perfused individually for 10 min using a portable vacuum pump constructed of a rubber cork fitting the falcon tube, a male Luer lock and a female 100 ml syringe. Once in the laboratory, 20–30-µm tangential sections from the first centimeter of wood adjacent to the cambium were made with a sliding microtome. Tangential sections are thought to most accurately estimate ray parenchyma volumes (von Arx et al., 2015). Sections were then washed of solution and stained using safranin or left unstained (when UV illumination was used, e.g. for all oaks), and photographed with a ×10 objective. Six 760 µm × 428 µm images per core were taken (totaling 1.95 mm2) for all but P. albicaulis, the first species measured. Ten images per core (totaling 3.25 mm2) were taken for P. albicaulis. RPF was then quantified for each image by manually highlighting ray cell lumina using ImageJ software and dividing the fraction of highlighted pixels by the fraction of total pixels. An average of the 2D RPF value for an individual tree was made by averaging all assessed images and assumed to be consistent with 3D RPF values (volume). Although we did not measure axial parenchyma fractions, rays are the dominant (or only) component of parenchyma fractions in conifers (Esau, 1977; Carlquist, 2001) and often the primary component of parenchyma fractions in temperate angiosperms, including oaks (Spicer, 2014). Sticking to one parenchyma orientation made comparisons between species or individuals possible. Also in the interest of making a balanced assessment given a feasible slide number (the total slides manually highlighted was > 1200), whereas photographs were taken randomly within conifer slides (Fig. 2a–f), photographs were taken in sequence from a random starting location within a single slide for oaks (Fig. 2g–i). Coupled with the magnification required to identify parenchyma lumina, the structural diversity of oaks meant that consecutive images evenly captured the true value of an individual slide which might include the low RPF values generated by vessels (Fig. 2g) or the high RPF values generated by multiserrate rays tens of cells across (Fig. 2i).

Details are in the caption following the image
Representative images of low (a, d, g), moderate (b, e, h), and high (c, f, i) ray parenchyma fractions (RPF) for tangential slides of conifers Pinus albicaulis in the strobus pine subgenus (a–c) and Pinus contorta subsp. murrayana in the pinus pine subgenus (d–f), as well as of Quercus chrysolepsis, an evergreen oak (g–i). All images are at the same scale (see scale bar in g). All lower panels demonstrate characteristic highlighting of parenchyma lumina for pixel counts. In pines, images demonstrate variability within species and between related species due to changes in cell size and number as well as the presence of radial resin ducts flanked in parenchyma cells (b, e). Images also demonstrate lower RPF where vessels were present in oaks (g), but very high RPF where multiserrate rays were up to tens of cells across (i). Note the exclusion of axial parenchyma wherever present in both conifers (as epithelial cells adjacent to axial resin ducts in some members of the Pinaceae, e) and in oaks (present in all images but indicated only in h). The RPF of a given slide (RPFslide) followed by the fraction of slides with an RPF lower than or equal to the given slide appear across the top of related images.

In order to add value to the RPF data, phylogenetic relationships and variability indices also were assessed. The phylogenetic tree used in phylogenetic comparisons of RPF was generated from the National Center for Biotechnology Information (NCBI) database with R/rotl (Michonneau et al., 2016) and R/ape (Paradis & Schliep, 2018). Groups included the Cupressaceae, or cypress family, as well as four groups within the Pinaceae, or pine family – the abietoid subfamily, the Pseudotsuga genus, the Pinus subgenus pinus and the Pinus subgenus strobus. RPF variability was quantified using a modification of the Environmentally Standardized Plasticity Indices used to measure phenotypic variability relative to environmental variability when the genetic contribution to phenotypic variability is known (Valladares et al., 2006). As we do not account for genetic variability, an Environmentally Standardized Variability Index (ESVI) was generated for each species by first generating climate values for each site using all available 30-yr normal climate data compiled at 800 m2 resolution by the Parameter-elevation Regressions on Independent Slopes Model (PRISM Climate Group, 2019). We then pulled out the maximum and minimum values of elevation (actual) or climate (generated for sites from PRISM data) experienced by a given species as well as the maximum and minimum RPF (measured) of a given species. Subtracting maxima from minima and dividing the resulting range of RPF by the resulting range of elevation or climate produced our ESVI value.

Sampling and biochemical analysis of NSCs

At the same time that tree cores for morphological analysis were collected in September 2016, a second set of cores also was collected in the same manner and from the same individuals, but placed in a paper art straw instead of into solution. Cores were collected to maximize trunk starch concentrations (Salleo et al., 2009; Tixier et al., 2018) between 05:00 h and 10:00 h. By the end of the day these samples were dried for 1 h at c. 90°C using a backpacking oven (Bemco Inc., Simi Valley, CA, USA) or camp oven (Coleman Company Inc., Chicago, IL, USA), both with thermometers. Although all sites were resampled in June/July 2017 and September 2017, due to the difficulties of accessing high elevation sites through the winter and early spring, NSC samples were intentionally sought in Winter 2016 (December) and Spring 2017 (March) for only four species. We chose species to represent the study’s two conifer families, Pinaceae (P. ponderosa) and Cupressaceae (C. decurrens), as well as the study’s only angiosperm family, Fagaceae – both its evergreen (Q. chrysolepsis) and deciduous (Q. kelloggii) leaf habits. NSC samples for any additional species (A. magnifica, P. menziesii, P. lambertiana and P. sabiniana) were collected only incidentally when these species co-occurred with the aforementioned four. After transport back to the laboratory, all samples were further dried at c. 90°C for 48 h, and the first centimeter of wood adjacent to the cambium was separated from each core, homogenized using a minibead beater (Biospec Products Inc., Bartlesville, OK, USA), and analyzed using hydrolytic enzyme reactions coupled with a sulfuric acid-anthrone assay in a method modified from Leyva et al. (2008).

Estimating storage capacity

In order to estimate the maximum and minimum starch (Starchmax and Starchmin) or sugar (Sugarmax and Sugarmin) storage capacities of a given parenchyma fraction we made four separate calculations. RPF was used as a predictive variable (RPF = X cm3 parenchyma cm−3 wood). Constants for starch calculations included the specific weight of hydrated (wheat) starch (Dstarch = 1250 mg starch cm−3 parenchyma; Dengate et al., 1978), an estimate of the fraction of RPF which might be available for starch storage (AF = 0.4), and an estimate of wood density (WDlow = 0.4 g wood cm−3 wood, WDhigh = 1.0 g wood cm−3 wood). The starch available fraction was approximated from Nakaba et al. (2016), which found a maximum c. 40% starch fraction in the rays of Cryptomeria japonica, a conifer, and Salleo et al. (2009), which found that a maximum of c. 79% of parenchyma lumina demonstrated a starch fraction > 50% in the rays of Laurus nobilis, an angiosperm (79% × 50% = c. 40%). Low (applying to conifers) and high (applying to oaks) wood density values were approximated from Miles & Smith (2009). Constants for sugar calculations included the same low and high wood densities as well as the molar mass of sucrose (Mu = 342 g mol−1) and the approximate molarity of sucrose required to balance tensions between 10 bars (M10 = 0.4 mmol cm−3 parenchyma) and 40 bars (M40 = 1.6 mmol cm−3 parenchyma), approximate values reviewed by Michel (1972). All starch calculations assume the absence of sugar (but not the absence of organelles); all sugar calculations assume the absence of starch.
urn:x-wiley:0028646X:media:nph16361:nph16361-math-0001(Eqn 1)
urn:x-wiley:0028646X:media:nph16361:nph16361-math-0002(Eqn 2)
urn:x-wiley:0028646X:media:nph16361:nph16361-math-0003(Eqn 3)
urn:x-wiley:0028646X:media:nph16361:nph16361-math-0004(Eqn 4)

Statistical analysis

All climate variables which appear in maps, indices or models are 30-yr normal climate data compiled at 800 m2 resolution by the PRISM Climate Group (2019). Across-species variation in RPF was assessed along elevation and climate gradients using a combination of linear models and ANOVA, and among oak or conifer species and conifer phylogenetic groups using ANOVA and Tukey’s Honestly Significant Difference (Tukey’s HSD, P < 0.05). Across-species variation in ESVI values for all conifers sampled along the entirety of their transect (all, but P. sabiniana) was assessed using the centers of species range for each related elevation or climate variable as predictors in linear models. First- or second-order models were then fitted to each comparison and the model with the lowest Akaike Information Criterion (AIC) was selected for ANOVA. The potential environmental drivers of RPF variability within species were examined between W–SW and E–SE aspects for all species sampled as well as along gradients of elevation or climate for all species whose full ranges were sampled. Due to uneven sample numbers between compared aspects, we used ANOVA with Type II sums of squares to determine the significance of aspect’s effect on RPF for all 15 species. To assess the effects of elevation or climate on RPF, first- or second-order models were fitted to each comparison and the model with the lowest sample size-corrected AIC (AICc) was selected for ANOVA as well as input into multivariate models. As several climate variables were highly correlated, we used the step function in R to simplify these multivariate models. We ran the models for each species excepting P. sabiniana and Q. douglasii (both of which were not sampled at their lowest elevations) individually, removing variables with variable inflation factors > 3.0 (generated with the ‘vif’ function in R/car; Fox & Weisberg, 2011). We then compared all uni- and multivariate models to a null model which set the measured RPF of each individual tree equal to the species mean (i.e. the intercept given no slope). We determined significance for any model with an AIC < −2 relative to the null AIC (ΔAIC < −2) and whose input variables did not have a Pearson’s correlation coefficient > 0.6; we then checked ANOVA. Elevation was explored as a synthesizing predictor variable for NSC (starch, sugar and total NSC) within species. Linear mixed effect models using individual trees as a random effect (R/nlme; Pinheiro et al., 2018) combined with ANOVA were used to assess the effects of elevation, season and their interaction on NSC concentrations for all species whose full ranges were sampled (again excluding P. sabniana and Q. douglassi). All comparisons between RPF and starch, sugar or total NSC, as well as between sugar and starch include all measurements over the course of sampling and, again, use linear models combined with ANOVA.

Results

RPF is significantly correlated with elevation (−), precipitation (−) and mean annual temperature (+) across oaks, temperature (+) and minimum vapor pressure deficit (+) across conifers, but little variance is explained

Ray parenchyma fraction (RPF) decreased significantly towards higher elevations for oaks (Fig. 3a; P = 0.026) and decreased, but not significantly, towards higher elevations for conifers (Fig. 3b; P = 0.218). For mean annual precipitation we observed the same negative correlations, significant in oaks (Fig. 3c; P = 0.011) and not statistically significant in conifers (Fig. 3d; P = 0.423). For mean annual temperature we observed positive correlations, significant in both oaks (Fig. 3e; P = 0.012) and conifers (Fig. 3f; P = 0.019), with RPF increasing towards warmer elevations or decreasing towards colder elevations. Although most of the additional climate variables tested also demonstrated significant relationships in oaks, the only additional variable that demonstrated a significant relationship with conifer RPF was minimum vapor pressure deficit (VPDmin) (Fig. 3h; P = 0.024), RPF increasing towards ranges which stay relatively warm and dry year-round. Correlations for additional climate variables appear in Fig. S1. The explanatory power of all correlations was very low, all urn:x-wiley:0028646X:media:nph16361:nph16361-math-0005 values falling below 0.2. For conifers, species with a mean RPF consistently falling outside the confidence intervals of the linear model were A. magnifica on the high end (6.40 ± 0.26%, n = 12), and P. lambertiana (3.95 ± 0.18%, n = 14), P. monticola (3.96 ± 0.20%, n = 7) and P. albicaulis (3.69 ± 0.26%, n = 10) on the low end.

Details are in the caption following the image
Mean ray parenchyma fractions (RPF) and elevation (a and b, ELEV) or climate range centers of each species studied fit by linear regression (bold black lines). Climate range centers were calculated from 30-yr norms (PRISM Climate Group, 2019) of mean annual precipitation (c and d, MAP), mean annual temperature (e and f, Tmean), and minimum vapor pressure deficit (g and h, VPDmin) at sampled sites within the Yosemite Grinnell Transect. Oak species (a, c, e, g) included Quercus douglasii, Quercus chrysolepsis and Quercus kelloggii. Conifer species (b, d, f, h) included Tsuga mertensiana, Abies concolor, Abies magnifica, Pseudotsuga menziesii var. menziesii, Pinus lambertiana, Pinus albicaulis, Pinus monticola, Pinus contorta subsp. murrayana, Pinus sabiniana, Pinus jeffreyi, Pinus ponderosa and Calocedrus decurrens. Vertical bars represent SE around the mean RPF. Horizontal bars extend between the leading and trailing range edges of each species (we did not sample below 500 m). Black asterisks indicate that the slope of the related regression was significant (*, P < 0.05; ■, P < 0.1). Confidence intervals (95%, shaded areas) also are provided.

Splitting the axes of Fig. 3 into two separate graphs, one of observed elevation range (Fig. 4a) and one of species mean RPF (Fig. 4b), permits us to arrange species by their phylogenetic relationships (Fig. 4c), and display significant differences among species and phylogenetic groups (Fig. 4b). Observations of low RPF in P. lambertiana, P. albicaulis and P. monticola, described above, now lead to a more conclusive finding of significantly lower RPF across pines in their strobus subgenus relative to the other conifer groups studied. The RPF values of P. lambertiana, P. albicaulis and P. monticola, all members of this group, were significantly lower than five, seven and two of the nine of the other conifer species studied, respectively. Additionally, the elevation trend across oaks was confirmed despite the low sample size of Q. douglasii, this lowest elevation oak having a significantly higher mean RPF (18.29 ± 2.30%, n = 4) than its upslope counterparts, both the (also deciduous) Q. kelloggii (12.99 ± 0.71%, n = 22) and the evergreen Q. chrysolepsis (13.37 ± 0.79%, n = 11).

Details are in the caption following the image
Observed ranges (a), ray parenchyma fractions (RPF) (b) and phylogenetic groupings (c) of all species sampled along the Yosemite Grinnell Transect. Conifer species include Tsuga mertensiana, Abies concolor, Abies magnifica, Pseudotsuga menziesii var menziesii, Pinus lambertiana, Pinus albicaulis, Pinus monticola, Pinus contorta subsp. murrayana, Pinus sabiniana, Pinus jeffreyi, Pinus ponderosa and Calocedrus decurrens. Oak species include Quercus douglasii, Quercus chrysolepsis and Quercus kelloggii. Boxes in (a) encompass a species’ sampled range and symbols indicate the range center. Symbols in (b) indicate the mean RPF of a species and bars indicate SE around this mean; lower case letters indicate significance groupings of species assessed within conifers or oaks (Tukey’s Honestly Significant Difference (HSD), P < 0.05); upper case letters indicate significance groupings of phylogenetic groups (like symbol shapes in (a) and (b)) within conifers (Tukey’s HSD, P < 0.05). As a single genus was represented in the oaks, we made no similar comparison for angiosperms and symbol shapes indicate deciduous (circle) or evergreen (square) habits.

Variability of RPF increases with elevation across conifer species; replacing elevation with some climate variables increases significance and explained variance

Transforming conifer RPF values into an index of species RPF variability demonstrated that higher elevation conifers may have a greater capacity for RPF acclimation relative to range size than lower elevation conifers (urn:x-wiley:0028646X:media:nph16361:nph16361-math-0006 = 0.207, P = 0.090; Fig. 5a). Variance explained increased and probability values decreased when we replaced elevation with two climatic variables available in the PRISM database: mean annual temperature (urn:x-wiley:0028646X:media:nph16361:nph16361-math-0007 = 0.386, P = 0.058; Fig. 5b) and maximum VPD (VPDmax) (urn:x-wiley:0028646X:media:nph16361:nph16361-math-0008 = 0.422, P = 0.018; Fig. 5c). Variables that did not improve both the significance and explained variance of our model are included in Fig. S2.

Details are in the caption following the image
Best-fit linear models (bold black lines) comparing the response of conifer ray parenchyma fraction (RPF) variability to elevation or the 30-yr normals of climate variables (PRISM Climate Group, 2019). RPF variability was approximated by a conifer species’ Environmentally Standardized Variability Index (ESVI). A species’ ESVI value was calculated for elevation (ELEV, a), mean annual temperature (Tmean, b) or maximum vapor pressure deficit (VPDmax, c) as the difference between the species’ maximum and minimum RPF divided by the difference between the species’ maximum and minimum elevation/climate value (absolute). Whatever elevation or climate variable was included in the denominator of the index, the center of that variable’s range for each species was used as the model’s predictor. Conifers randomly sampled along the entirety of their elevation gradients and so included in this analysis were Tsuga mertensiana, Abies concolor, Abies magnifica, Pseudotsuga menziesii var. menziesii, Pinus lambertiana, Pinus albicaulis, Pinus monticola, Pinus contorta subsp. murrayana, Pinus jeffreyi, Pinus ponderosa and Calocedrus decurrens. Summary statistics and confidence intervals (shaded areas, 95%) are provided. asl, above sea level.

Elevation, aspect or climate variables predict RPF within conifers, but not within evaluated oaks

We were able to reject the null hypothesis that RPF was unrelated to elevation or at least one environmental variable for just over half of the evaluated conifer species (Table 1; Fig. 6). Aspect had a significant effect on RPF for one species, P. contorta (P = 0.003), individuals on E–SE-facing slopes demonstrating significantly higher RPF (5.64 ± 0.23%, n = 11) than those on W–SW-facing slopes (4.64 ± 0.20%, n = 13). The only climate variable shared among the five additional species for which environment (elevation/climate) significantly predicted RPF was VPDmax, RPF increasing as deficit decreased. All univariate best-fit (first or second order polynomial) models comparing RPF with elevation and climate variables for each species appear in Fig. 6. Summary statistics for univariate or multivariate models that improved on the null hypothesis (ΔAIC ≤ −2.0) appear in Table 1. The tested conifers for which neither aspect nor a combination of elevation and/or climate variables significantly predicted RPF better than the null model were T. mertensiana, P. lambertiana, P. monticola, P. ponderosa and C. decurrens. We also were unable to reject the null hypothesis for the tested evergreen Q. chrysolepsis or the deciduous Q. kelloggii (ΔAIC > −2, > 0.05).

Table 1. Summary statistics of comparisons between the mean ray parenchyma fractions (RPF) of individual trees and the elevation or climate associated with their locations
  Best models ΔAIC urn:x-wiley:0028646X:media:nph16361:nph16361-math-0009 P Sig
Tsuga mertensiana NULL na na na na
Abies concolor VPD max −4.5 0.306 0.031 *
Abies magnifica ELEV + MAP −4.8 0.439 0.056 .
ELEV −4.7 0.404 0.039 *
TD mean −3.2 0.287 0.042 *
T max −3.2 0.287 0.042 *
T mean −3.2 0.283 0.044 *
T min −2.7 0.256 0.054 .
VPD max −2.6 0.248 0.057 .
Pseudotsuga menziesii TDmean + VPDmax + VPDmin −11.3 0.861 0.082 .
VPDmax + Tmin −7.5 0.757 0.056 .
VPD max −5.5 0.661 0.031 *
T max −5.1 0.636 0.036 *
T mean −4.6 0.610 0.041 *
MAP −2.4 0.402 0.100 .
Pinus lambertiana NULL na na na na
Pinus albicaulis VPD max −4.9 0.472 0.045 *
T max −3.7 0.406 0.067 .
Pinus monticola NULL na na 0.027 *
Pinus contorta ASPECT na na 0.003 **
Pinus jeffreyi T max −3.5 0.305 0.036 *
VPD max −3.2 0.285 0.049 *
ELEV −2.3 0.233 0.064 .
T mean −2.2 0.223 0.069 .
Pinus ponderosa NULL na na na na
Calocedrus decurrens NULL na na na na
Quercus chrysolepsis NULL na na na na
Quercus kelloggii NULL na na na na
  • When aspect is listed as a species’ best model, probability values and significance codes are the results of an ANOVA comparing RPF on W–SW and E–SE-facing slopes. When elevation or climate variables appear in the best models list, statistics are the output of one or more first- or second-order polynomial models. When NULL appears as the best model for a species, neither aspect nor an elevation or climate model lowered the AIC (uncorrected) by ≥ 2 relative to the NULL model (the RPF of an individual tree is a function of the mean RPF of its species). Significance: ∗∗, P < 0.01; ∗, P < 0.05; P < 0.1; na, not applicable.
Details are in the caption following the image
Best-fit linear models (bold black lines) comparing the mean ray parenchyma fractions (RPF) of individual trees and the 30-yr normals of climate variables (PRISM Climate Group 2019) at tree sites for species randomly sampled within the Yosemite Grinnell Transect along their entire elevation gradients. Conifer species were Tsuga mertensiana, Abies concolor, Abies magnifica, Pseudotsuga menziesii var. menziesii, Pinus lambertiana, Pinus albicaulis, Pinus monticola, Pinus contorta subsp. murrayana, Pinus. jeffreyi, Pinus ponderosa and Calocedrus decurrens. Oak species were Quercus chrysolepsis and Quercus kelloggii. Climate variables used as predictors included elevation (ELEV), mean annual precipitation (MAP), mean dewpoint temperature (TDmean), mean annual temperature (Tmean), maximum annual temperature (Tmax), minimum annual temperature (Tmin), maximum vapor pressure deficit (VPDmax), and minimum vapor pressure deficit (VPDmin). Confidence intervals (95%) are shaded. Red asterisks indicate that a model improved on the null model (the mean RPF of an individual tree is a function of the mean RPF of its species) such that the Akaike Information Criterion decreased by ≥ 2. Black asterisks indicate that a model was significant (P < 0.05).

For incense cedar and oaks, the NSC content of parenchyma cells varies with environment

Although RPF in the only studied non-Pinaceae conifer (C. decurrens) and angiosperms (Q. chrysolepsis and Q. kelloggii) failed to demonstrate significant relationships with aspect, elevation or climate, we did observe highly significant relationships between NSC and elevation in these species (Fig. 7; Table 2). In a mixed effects model of repeated measures over several seasons using individual trees as a random effect, total NSC in the evergreen Q. chrysolepsis and the evergreen C. decurrens increased significantly with elevation (P < 0.001 and P = 0.001, respectively). When the same model was run for the deciduous Q. kelloggii, total NSC decreased significantly with elevation (P = 0.007). The relative contributions of starch (Fig. S3; Table S1) and sugar (Fig. S4; Table S2) grouped in a similar way. For Q. chrysolepsis and C. decurrens, sugar (P < 0.001 and P = 0.001, respectively) demonstrated a more significant relationship with elevation than starch (P = 0.002 and P = 0.029, respectively). For Q. kellogii, starch (P = 0.001) demonstrated a more significant relationship with elevation than sugar (P = 0.059).

Details are in the caption following the image
Seasonal trends (a) and elevation trends in each season (b) for all total nonstructural carbohydrate (NSC) measurements (combined sugar and starch). Samples were taken from the trunk wood (0–1 cm from cambium) of Tsuga mertensiana, Abies concolor, Abies magnifica, Pseudotsuga menziesii var menziesii, Pinus lambertiana, Pinus albicaulis, Pinus monticola, Pinus contorta subsp. murrayana, (Pinus sabiniana), Pinus jeffreyi, Pinus ponderosa and Calocedrus decurrens, and (Quercus douglasii), Quercus chrysolepsis and Quercus kelloggii. Species listed in parentheses were not included in elevation analysis as the low ends of their ranges (below 500 m above sea level (asl)) were not sampled. In both (a) and (b) the Pinus genus is split into its strobus (Pinus (s)) and pinus (Pinus (p)) subgenera and the Quercus genus is split by evergreen (Quercus (e)) or deciduous (Quercus (d)) habit. In (a) error bars indicate SE at a single time point and dashed lines connect points of a single species (with breaks at time points when no measurements for that species were made. In (b), solid black lines indicate the linear regression for a given season and confidence intervals (95%, shaded areas) are also provided; asterisks in the same color as the species legend indicate that the effect of elevation over time (repeated measures) was significant (***, P < 0.001; **, P < 0.01). Seasons include September 2016 (S16), December 2016 (D), March 2017 (M), June–July 2017 (J) and September 2017 (S17). Summary statistics are provided in Table 2.
Table 2. Probability values and significance codes for mixed effect models using elevation and season as fixed effects and tree ID as a random effect to predict total nonstructural carbohydrates (NSC) in the trunk wood (0–1 cm below cambium)
  Elevation Season Elevation : Season
P Significance P Significance P Significance
T. mertensiana 0.878 ns 0.726 ns 0.922 ns
A. concolor 0.901 ns 0.027 * 0.526 ns
A. magnifica 0.710 ns 0.195 ns 0.265 ns
P. menziesii 0.637 ns 0.596 ns 0.755 ns
P. lambertiana 0.308 ns 0.003 ** 0.341 ns
P. albicaulis 0.216 ns 0.002 ** 0.156 ns
P. monticola 0.604 ns 0.277 ns 0.774 ns
P. contorta 0.465 ns 0.170 ns 0.999 ns
P. jeffreyi 0.724 ns 0.088 . 0.958 ns
P. ponderosa 0.095 . 0.120 ns 0.587 ns
C. decurrens 0.023 * 0.085 . 0.387 ns
Q. chrysolepsis <0.001 *** 0.912 ns 0.572 ns
Q. kelloggii 0.010 ** 0.015 * 0.155 ns
  • Species randomly sampled along the entirety of their elevation range within the Yosemite Grinnell Transect and so included in this analysis were Tsuga mertensiana, Abies concolor, Abies magnifica, Pseudotsuga menziesii var menziesii, Pinus lambertiana, Pinus albicaulis, Pinus monticola, Pinus contorta subsp. murrayana, P. jeffreyi, P. ponderosa and Calocedrus decurrens, and Quercus chrysolepsis and Quercus kelloggii. Significance: ∗∗∗, P < 0.001; ∗∗, P < 0.01; ∗, P < 0.05; ., P < 0.1; ns, not significant.

Both oaks and conifers appear spatially unconstrained in their capacity to store NSC

Highly significant correlations were observed between the sugar and starch concentrations of all samples, the two NSC forms increasing relative to each other (P < 0.001, urn:x-wiley:0028646X:media:nph16361:nph16361-math-0010 = 0.259 and 0.149 for conifers and oaks, respectively; Fig. 8a). Comparisons between RPF and starch, sugar or total NSC led to two additional observations. First, starch concentrations were never maximized for either conifers or oaks; in fact, even the minimum estimation for starch storage (assuming high wood density and an availability fraction of just 40%) was barely reached (Fig. 8b). Second, all surveyed species appear to maintain sugar concentrations capable of balancing osmotic potentials between −1 and −4 MPa, maximizing sugar storage potential at c. 20 mg g−1 DW below an RPF of c. 5% for conifers and at c. 75 mg g−1 DW between an RPF of 7.5% and 15% for oaks (Fig. 8c). We observed significant positive relationships in oaks between RPF and total NSC (P = 0.036; Fig. 8d) driven by starch (P = 0.003; Fig. 8b), although both urn:x-wiley:0028646X:media:nph16361:nph16361-math-0011 values were < 0.1.

Details are in the caption following the image
Linear regressions (bold black lines) fitting all measurements (all individuals in all seasons) of starch to complementary measurements of sugar (a) as well as all measurements of ray parenchyma fraction (RPF) to complementary measurements of starch (b), sugar (c) and total nonstructural carbohydrates (NSC, d). Samples were taken from trunk wood (0–1 cm from cambium). Regressions fitting oaks (Quercus douglasii, Quercus chrysolepsi, and Quercus kelloggii) pass through warm-colored symbols; regressions fitting conifers (Tsuga mertensiana, Abies concolor, Abies magnifica, Pseudotsuga menziesii var menziesii, Pinus lambertiana, Pinus albicaulis, Pinus monticola, Pinus contorta subsp. murrayana, Pinus sabiniana, Pinus jeffreyi, Pinus ponderosa and Calocedrus decurrens) pass through cool-colored symbols. In (b) and (c), dashed lines indicate the maximum values of sugar or starch estimated for a given RPF and dotted lines indicate the minimum values of sugar or starch estimated for a given RPF. Black asterisks in line with the end of a regression indicate that the slope of that regression was significant (∗∗∗, P < 0.001, **, P < 0.01, *, P < 0.05). Confidence intervals (95%, shaded areas) also are provided.

Discussion

Higher RPF values were observed for oak species occupying ranges at low elevation associated with low precipitation and warm temperatures and conifers occupying ranges demonstrating some combination of dry and warm conditions year-round (high minimum vapor pressure deficits (VPDmin); Fig. 3). These trends lend support to the hypothesis that higher drought stress environments lead to species with higher parenchyma fractions. Although our sampling design restricted studied angiosperms to a single genus, there may still be value in comparing our findings with those of recent global (Morris et al., 2016a) and regional (China; Zheng et al., 2019) meta-analyses investigating relationships between climate and angiosperm RPF. Both meta-analyses concluded that parenchyma fraction in angiosperms increased with temperature (towards the equator) and the global analysis also found that parenchyma fraction decreased with precipitation. Our data corroborate these temperature and precipitation findings in angiosperms. Our data also directly extends the temperature findings to conifers and indirectly extends the combined temperature and precipitation findings, as VPDmin, to conifers. However, in studying this trait at a single latitude our findings differ from the global analysis’ positive correlations between parenchyma fractions and altitude. Our own study was limited in that the few species present along a single transect either reduce any potential for high explanatory power (oaks) or introduce the potential for species on the margins to reduce significance and explanatory power (conifers). Meta-analyses may fail to separate the confounding influences of altitude and latitude. Yet, comparisons of RPF with climatic features shared by latitude and altitude suggest complementary morphophysiological responses to environmental cues – species demonstrate increasing parenchyma fractions towards warmer and/or dryer environments.

Phylogeny also appears to be strongly linked to RPF – although this broad fact is well-established in foundational anatomy texts (Esau, 1977), it is important to revisit phylogenetic relationships now, in detail, because of their possible conservation implications. Our comparison of all studied conifer RPF values demonstrated that pines in the strobus (white/soft pine) subgenus are distinctly low (Fig. 4b). The low parenchyma fractions that we observed for the strobus subgenus are consistent with recent measurements of its only additional North American member, the type species Pinus strobus L. (eastern white pine), which also has an RPF of just under 4% (Spicer & Holbrook, 2007). Low RPF may indicate a less robust lattice of resin ducts, as has been suggested in the literature (Morris et al., 2016b). If true, this perhaps offers some explanation for the relatively high susceptibility of the strobus subgenus to emerging pests and diseases (Fig. 2; Gibson et al., 2008). Resin production and delivery is required for successful isolation of biotic perturbation at the stem surface (Trapp & Crouteau, 2001). The ducts that carry resin run both axially (up/down the stem) as well as radially and are often associated with parenchyma cells (Esau, 1977; Franceschi et al., 2005). North American white pines may have evolved a life cycle strategy in the historic absence of fungal pathogens such as white pine blister rust (Cronartium ribicola, Mielke, 1943) or the historic confinement to lower elevations of pests like the mountain pine beetle (Dendroctonus ponderosae, Amman, 1973). Although only tangentially related to our initial hypotheses, this begs immediate experimentation, particularly given the endangered status of Pinus albicaulis (Mahalovich & Stritch, 2013).

An in-depth look at several conifers provides evidence for two additional RPF observations, connected to each other, but surprisingly in contrast with findings across species. First, although the selective pressures of drought may lead to relatively high RPF for species occupying relatively warm and dry ranges, several conifers demonstrate that, relative to their ranges, they are most variable towards colder wetter elevations (Fig. 5). The higher variability indices of wetter colder elevation species might reflect some greater margin for acclimation that accommodates the stresses of relatively short cold growing seasons at high elevations (Bowling et al., 2018). We did not separate the contributions of genes and environment and so do not quantify plasticity sensu stricto (Valladares et al., 2006), but rather environmentally standardized trait variability, our Environmentally Standardized Variability Index (ESVI). This is perhaps unconventional, but we argue that it is reasonable, particularly as it helps to frame our second finding. Also in contrast to the finding across conifer species, that warmer dryer ranges appear to drive a species’ mean RPF up, within the range of any given species, warmer dryer conditions appear to drive RPF down (Fig. 6; Table 1). The low-elevation Pseudotsuga menziesii significantly increased its RPF towards high-moisture/low-temperature sites, despite the low sample size for this species. We observe the same trend in the mid-high elevation Abies magnifica. The low-mid-elevation Abies concolor departs from its closest relatives not in failing to increase RPF at moist sites, but also increasing RPF at dry sites. The other species for which we observed significant increases towards high moisture/low temperature sites were the mid-high elevation Pinus jeffreyi and the treeline-dominant P. albicaulis. The consistency of trends in additional conifers further suggests that high RPF is predominantly a feature of individuals experiencing low temperature/high moisture for their species. With more targeted sampling, trends lacking significance might be clarified for Tsuga mertensiana, found at just two elevations, and P. monticola, with a low sample size (n = 7). However, three relatively well-represented species within the transect also demonstrate no relationship with any of our input variables. For the high-elevation P. contorta, differences between W–SW and E–SE aspects suggest that this species may be particularly sensitive to the warmth of the afternoon sun, colder SE-facing slopes leading to significantly higher RPF as they play the same part as colder elevations in shortening growing seasons and perhaps driving acclimation. For P. lambertiana and P. ponderosa, there may be little need for individual acclimation. More likely, as these species experienced disproportionately high mortality in association with California’s severe 2012–2015 drought and mountain pine beetle outbreaks, (Fettig et al., 2019) we may not have been measuring the true breadth of variability in their populations.

For the rest of the species definitively demonstrating no relationships between RPF and environment, NSC were significantly correlated with elevation (Fig. 7). This was true for the only conifer species measured which was not in the Pinaceae family (Calocedrus decurrens) and for oaks. The evergreen species in this group (C. decurrens and Quercus chrysolepsis) demonstrate consistently lower NSC concentrations at lower elevations, suggesting that these individuals are operating with a smaller NSC safety margin than their upslope counterparts, perhaps due to the impacts of many years of climatic water deficit which skew towards lower elevations (Young et al., 2017). We are left wondering if stomatal sensitivity to water stress drives this relationship. Lower NSC reserves at lower (dryer) elevations may indicate anisohydricity. If growth stops before stomata close, this introduces a period in which reserves may be accumulated, a period that should shrink towards lower elevations as the time between cessation of growth and stomatal closure shortens along moisture gradients. Alternatively, if an evergreen species demonstrates no relationship between NSC and elevation (like the studied pines) this may be because it is relatively isohydric. Stomatal closure with the cessation of growth at some consistent water potential may effectively decouple tree carbon budgets from the full range of moisture regimes a species might otherwise experience. This reasoning would be consistent with divergent paths (and thresholds) for stomatal closure in related Cupressaceae and Pinaceae species (Brodribb et al., 2014). It would also be consistent with the relatively anisohydric behavior shared among oaks and members of the Cupressaceae, but not necessarily among pines (Meinzer et al., 2014; Martínez-Vilalta et al., 2017; Kannenberg et al., 2019). Again, immediate experimentation is warranted as an/isohydricity is a key ecophysiological distinction among species likely to explain plant survival and mortality in drought (McDowell et al., 2008). In contrast with the observed evergreen species, the deciduous Q. kelloggii shows lower NSC concentrations at higher elevations, its smaller margin upslope. It is likely that this species, the only species in our analysis that experiences true dormancy, has a shorter period between flushing and senescence (Vitasse et al., 2009) towards higher elevations, the NSC accumulation period shortening along gradients of temperature.

Finally, we provide evidence that trunk-stored NSC is unconstrained by physical space (Fig. 8). Concentrations of starch and sugar increase together (Fig. 8a), presumably as a tree requires them or is able to accumulate them, but without any apparent impediment. Although RPF may be driven by the phenotypic constraints of a species (Figs 3-5) or in some cases the constraints specific to an individual tree’s location (Fig. 6), if accommodation of NSC was an issue we believe we would have observed much higher levels of the most efficient carbon storage form, starch (Fig. 8b). This is despite a potential limitation of the study, that we did not include axial parenchyma in our analysis of oaks. Inclusion of axial parenchyma would have driven up the overall fraction of living cells in the wood and so the volumes in which starch might be stored, but not starch concentrations, just exaggerating the noted disparity. Instead, we observe immutable concentrations of the osmotically active form of NSC, sugar, at levels likely to balance reasonable tensions in the water column not levels that increase or decrease with RPF (Fig. 8c). Given the innate costs which accompany the maintenance of a living network – maintenance respiration (Ryan, 1990) and the exchange of space which might otherwise be used for water transport or mechanical strength (Carlquist, 2001; Ziemińska et al., 2015; Plavcová et al., 2019) – it is unlikely that the story is as simple as storage capacity far exceeds storage needs. Promising future experiments in this field may investigate the limits or prospects of additional parenchyma services. As proposed previously (Morris et al., 2016b) and discussed above in relation to P. albicaulis, parenchyma may play a critical role in combatting pests and pathogens, and there are myriad other timely but underexplored RPF hypotheses surrounding their role in recovery post-fire (Michaletz et al., 2012; Bär et al., 2019), even salinity management (Godfrey et al., 2019).

In view of the presented findings, we conclude that a tree’s capacity for storage within parenchyma fractions does not limit the accumulation of energy reserves. Instead, we suggest that some combination of adaptation and acclimation provides redundant storage volumes, which is perhaps unsurprising given that trees, as a rule, successfully accommodate enough reserves to meet the demands of the adverse conditions imposed (Barbaroux & Bréda, 2002; Hoch & Körner 2011; Piper et al., 2015; Piper et al., 2017). This is not to say that species will be unchallenged energetically or otherwise within their historical ranges. Unequivocally, some species or populations will be challenged and perhaps extirpated with climate change (Aitken et al., 2008). Indeed, we may already be observing climate change-related shifts in NSC dynamics (Bréda et al., 2006; Wiley et al., 2015; Earles et al., 2018). However, storage space is simply unlikely to be the morphophysiological feature that limits a species or individual’s response to these challenges.

Acknowledgements

The authors would like to thank Athena Demetry with Yosemite National Park, Steven Hollis with Sierra National Forest, Dave Horak with Stanislaus National Forest, and Andrea Kachulis with Inyo National Forest for permitting our fieldwork. We also would like to thank the Jastro Graduate Research Award (University of California, Davis) for funding our fieldwork, our colleague Aude Tixier for her guidance, undergraduate research assistant Henry Calhoun for his help processing carbohydrate samples, and volunteer field assistants Evelyn Williamson and Keith Bailey for their company and for keeping the pace. Last, we are grateful to the Emilio Gonzáles Esparcia Fellowship (Universidad Politécnica de Madrid, Spain) and the Katherine Esau Fellowship (University of California, Davis) for supporting our colleague PG-D.

    Author contributions

    JMG and MAZ developed the research idea through discussions and wrote the first draft; JMG and JR planned and designed the fieldwork; JMG, JR, JO and PG-D conducted the fieldwork; JMG conducted labwork and data analysis; AROC advised on anatomy and data analysis; and all authors contributed to writing of the manuscript.