1 Introduction

Stable isotope ratios of carbon (δ13C) and oxygen (δ18O) as measured in tree rings (cellulose or wood) are commonly used to understand eco-physiological processes and environmental conditions governing tree-growth (McCarroll and Loader 2004; Gessler et al. 2014). Because of the relationship between photosynthetic carbon isotopic fractionation (Δ13C) and the ratio of leaf internal and ambient CO2 concentrations (ci/ca), δ13C in tree rings has been used to characterize the relationship between the physiology involved in photosynthesis and the environment (Farquhar et al. 1989). This tool further provides a retrospective annual to intra-annual record of photosynthesis activity under varying atmospheric CO2 via δ13C-derived estimates of intrinsic water use efficiency (iWUE), the ratio of carbon gain to water loss (Saurer et al. 2014, see Chap. 17). The δ18O in tree rings is related to the δ18O of soil water, which reflects the δ18O of precipitation (see Chap. 18), and leaf water evaporative enrichment (Roden and Ehleringer 2000; Barbour et al. 2004; Treydte et al. 2014). Thus δ18O in tree rings has been used to reconstruct past eco-hydrological processes and atmospheric circulation patterns (Sidorova et al. 2010; Ballantyne et al. 2011; Brienen et al. 2012; Zhu et al. 2012; Xu et al. 2019, 2020; Nagavciuc et al. 2019), and provided a tool to evaluate the effects of relative humidity and air temperature on photosynthesis (Wright and Leavitt 2006b).

Sampling living trees for stable isotope analyses of tree rings involves many choices. Typically, the tree species and general locations selected are determined by the research questions. For example, in temperate or water limited environments, δ13C has been used to reconstruct drought and its impact on tree physiology (Lévesque et al. 2014). In boreal or temperature limited environment, δ13C has been used to reconstruct cloudiness (Young et al. 2010). Optimizing field sampling to have the best chance of addressing the research hypothesis and recovering the desired ‘signal’ requires choices regarding the site microenvironment, individual tree characteristics, and appropriate sampling equipment. Some of these choices can affect the stable isotope ratio(s) recorded in the wood (tree rings). However, other decisions related to practicalities of getting adequate sample mass and the methods employed during the subsequent processing must also be taken into consideration.

After choosing the research question and the tree species, the most important field consideration is site selection. The most common site selection criteria include proximity to compatible instrumented weather data, extra-site conditions, and stand characteristics (e.g., monoculture/mixed species, natural/planted, density, substrate, aspect, slope, and age). Ideally, instrumented weather data have been recorded close to the site of interest, at a similar elevation and aspect. Extra-site conditions refer to the surrounding topography and land use. For example, a hydrologic connection (with respect to lateral flow through soils) of the tree site with higher elevations likely indicates that the available water in deeper soils will have a mixture of local stable oxygen isotope values and values characteristic of higher elevations (see Chap. 18 for source water considerations in site selection). Subsurface flow and runoff can also cause differences across the area chosen for sampling, by providing water with different stable oxygen isotope values to some trees (Dawson and Ehleringer 1991). Stable carbon isotope values can also be affected through differences in water status driven by differences in microsite conditions, i.e., if some of the trees have access to the subsurface flow or runoff while others do not. In addition, major changes in land use can alter the timing of moisture input, e.g., from snowpack, and can even affect the local climate (Salati and Nobre 1991; D’Almeida et al. 2007; Perugini et al. 2017). Selecting sites where environmental data and historical land-use data is available maybe critical to the success of a project, and researchers should consider carefully what information they need for their specific research questions when selecting sites.

Stand characteristics can affect the growth rate of the trees and their sensitivity to changes in the local climate. For example, competition for light and water can reduce the ability of a tree to respond optimally to favorable environmental conditions (Moreno-Gutiérrez et al. 2012; Voelker et al. 2019). Mature trees within a closed canopy stand will be buffered to some extent from the direct effects of drought and heat by lower vapor pressure deficit caused by the transpiration of the neighboring trees (Talbott et al. 2003). Subcanopy trees within a closed-canopy forest may incorporate a larger proportion of respired CO2 during photosynthesis than canopy-dominant trees (see Chaps. 19 and 24) and may have lower carbon assimilation rates due to lower light input (e.g., Cerling et al. 2004). Stable isotopes contained within subcanopy trees may reflect more information on stand dynamics and competition than canopy-dominant trees, particularly if a tree has changed competitive status over time. The isotopic composition within canopy dominant trees is more likely to reflect climate variation over time (Barnard et al. 2012).

Achieving the desired temporal resolution from the sampled wood requires knowledge about the length of the growing season and the growth rate of the species at the chosen site. Obtaining an adequate amount of material (e.g., whole wood, cellulose) for the desired temporal resolution may require careful consideration of the diameter of the increment borer used for sampling and the number of samples taken from each tree. Many studies involving subannual sampling use the relatively time-stable earlywood/latewood boundary as a time marker to designate the subannual divisions, but subannual sampling can also be executed using successive equal segments within a ring (Xu et al. 2020). These two options have a technical but fundamental difference. On the one hand, utilization of the earlywood/latewood technique yields a predetermined number of slices per year, whereas the successive equal-segment cut technique yields a variable number of samples per year and among trees because it will depend on the ring width of individual trees and specific years. Independent of the sectioning method used, variable tree growth rates should play an important role in how the samples are prepared. It is necessary to avoid unbalanced mass among trees if the samples are combined. For example, if the trees are combined (pooled), it is preferable to select the same mass of material per tree to avoid biased isotopic signals from the trees that have more mass on the same section (Leavitt 2010; Dorado Liñán et al. 2011).

For tree-ring stable isotope analyses, several components can be analyzed: whole wood, holocellulose, α-cellulose, or lignin. The α-cellulose molecule is often the preferred material because it is more resistant to environmental degradation than all major wood-constituents except lignin. Additionally, α-cellulose is a single molecule formed by a specific series of biochemical steps and it does not undergo isotopic exchange after deposition, in contrast to hemicellulose, for example, which can exchange at the carbonyl bonds. Several methods can be used to extract α-cellulose from whole wood material (see Chap. 5 for cellulose extraction details). While there is a prevalence of α-cellulose analyses in tree-ring isotope studies based on the advantages described above, whole wood remains another component of choice because it requires less analytical time and cost. A number of studies have assessed the coherence of the climate signal recorded in both whole wood and α-cellulose (Borella et al. 1998; Barbour et al. 2001; Loader et al. 2003; Ferrio and Voltas 2005; Cullen and Grierson 2006; Sidorova et al. 2008; Szymczak et al. 2011; Bégin et al. 2015; Weigt et al. 2015), but there is no consensus as to whether both components record similar climate signals (i.e., strength and timing) for both δ13C and δ18O. Ultimately, the choice of α-cellulose versus whole wood can be informed by a preliminary analysis of the studied site and species (Guerrieri et al. 2017), particularly when aiming to conduct large-scale (e.g. network) or long-term (e.g. millennial scale) paleoclimate and physiology reconstructions (Leavitt et al. 2010).

This chapter provides guidelines for field sampling and sample processing in the context of the current major areas of research using stable isotope ratios from stem wood, i.e., tree physiology and climatology. Specifically, it provides an overview of how to account for various factors influencing the isotopic signal recorded in tree rings when designing and conducting a tree-ring stable isotope study for either paleoclimate or ecophysiology investigations. The overview will cover theoretical and practical aspects of site and tree selection, sample collection and tree ring sample preparation. Finally, an emphasis on high-resolution tree ring isotope analyses will be presented in order to highlight the potential ecophysiological and paleoclimatological insights that may be gained from highly resolved intra-annual measurements of stable isotopes in tree rings.

2 Sample Collection

2.1 Site and Tree Selection

Site selection should minimize the influence of any local micro-environment effects on the isotopic signal recorded by individual selected trees. A common approach is to select canopy-dominant trees with no direct competition (McCarroll and Loader 2004; Leavitt 2010). Usually, dominant trees will also be older than subdominant or subcanopy trees, and when trees of the same age are compared, the dominant trees are likely to have wider ring widths for the same years. The choice of older trees is primarily motivated by the goal of developing long multi-centennial paleoclimate and paleophysiological reconstructions or to avoid isotopic trends observed in early growth years, the so-called "juvenile or canopy effect” for δ13C (Monserud and Marshall 2001) or “age-related-effect” for δ18O (Treydte et al. 2006). Age-related isotope trends appear to be more consistent for δ13C than δ18O (Young et al. 2011; Xu et al. 2011; Labuhn et al. 2016; Lavergne et al. 2017; Friedman et al. 2019; Duffy et al. 2019). Overall, fewer studies have addressed these trends and their causes for δ18O compared to δ13C. Age-related trends are avoided by only including rings formed after the first (or inner) 20–50 years from the pith. The number of years included within the juvenile period is not constant, so a preliminary test on a single core may be advisable prior to expending money and time on unusable samples. An alternative to omitting the juvenile tree growth is to mathematically remove the juvenile isotopic trend using different detrending methods such as the regional curve standardization “RCS” detrending method (Gagen et al. 2008; Esper et al. 2010, 2015; Helama et al. 2015).

The majority of paleoclimate tree-ring isotope-based studies use an approach similar to classic dendrochronology studies where old, dominant and sensitive trees (“big trees”) are selected (Gagen et al. 2011; Loader et al. 2013a). Minimizing micro-environment effects is addressed ad hoc during field collection (through tree selection), while isotopic trends related to developmental stages are addressed post hoc during sample preparation by excluding the inner rings or by detrending. The “classic” dendroclimatology sampling approach can, however, be biased when estimating past and projected responses of forest growth (based on tree-ring width) to climate variability and trends (Klesse et al. 2018a). This bias in targeting climatically sensitive and dominant trees for δ13C chronologies has been investigated with respect to their use to infer physiological responses of trees to rising atmospheric CO2 concentrations, i.e., iWUE (Brienen et al. 2017). In this recent analysis of iWUE derived from tree-ring δ13C, Brienen et al. (2017) used a size-stratified sampling approach to infer how developmental effects, particularly the interaction between tree height and age, affect the derived trend of 13C discrimination and the corresponding iWUE magnitude over the twentieth century. Within a given site, the size-stratified approach involves sampling trees from all size classes at similar CO2 levels. Such stratification enables comparison of age or height effects independent of CO2 trends (Klesse et al. 2018b). This sampling approach has shown that when accounting for tree height, the magnitude of iWUE trends derived prior to height correction was significantly diminished or nonexistent in conifers (Monserud and Marshall 2001). There are several mechanisms, physical and physiological, by which tree-height can affect variations and trends in tree-ring δ13C. These include gradients of light, VPD and δ13C of atmospheric CO2 between canopy and subcanopy trees. These mechanisms are indirect because they result from tree status in the canopy (canopy position) and therefore are mediated by stand structure, particularly in closed canopy forests (e.g. shade tolerant tree species) (Vadeboncoeur et al. 2020).

In contrast, direct mechanisms of height effect result from hydraulic resistance linked to either reduced water potential or turgor when trees get taller. This leads to reduced stomatal conductance (McDowell et al. 2011), reduced CO2 concentration inside the leaves and consequently less isotopic discrimination, which explains artefactual increases in iWUE as being related to height rather than a physiological response of stomatal conductance and photosynthesis to higher atmospheric CO2. An important caveat when considering the height effect is the underlying assumption that the height of average photosynthetic carbon gain impacting stem growth at ~130 cm (height at which tree rings are commonly sampled) is linearly related to height growth. This in turn depends on species, light compensation points, stand density and relative dominance of the sampled individuals. Brienen et al. (2017) found that at constant atmospheric CO2, iWUE in young understory trees increased by a factor of two or three throughout a tree’s life span in several species. These developmental trends were attributed to an increase in tree height and were not restricted to the early part of tree growth as previously thought (Treydte et al. 2006; Esper et al. 2010, 2015; Helama et al. 2015).

Despite potential biases of height effects discussed above, the magnitude of iWUE increases over the twentieth century, as derived from tree-ring δ13C measurements from several tree species in temperate, boreal and tropical forests is, however, remarkably consistent amongst tree species and forest ecosystems (Saurer et al. 2004, 2014; van der Sleen et al. 2014; Frank et al. 2015). Further, it is consistent with estimates derived from δ13C measurements of atmospheric CO2 (Keeling et al. 2017). The iWUE estimates derived from tree rings using the “classic” sampling strategy are therefore robust and are further supported by results of process-based biogeochemical (Keller et al. 2017) and vegetation models (Frank et al. 2015). Recent findings corroborate that using δ13C for climate and physiological inferences is reliable for years when trees are in canopy dominant positions, particularly for shade-tolerant species (Klesse et al. 2018a) with the underlying assumption that stand density and therefore competition remained unchanged over time (Voelker et al. 2019).

The application of tree-ring isotope studies to constrain the carbon cycle and atmosphere-biosphere interactions at the ecosystem level offer exciting opportunities (Belmecheri et al. 2014; Babst et al. 2014; Guerrieri et al. 2019; Lavergne et al. 2019, 2020; Szejner et al. 2020a). However, the commonly used sampling strategies ignore spatial scaling issues (e.g., stem, to leaf, to stand, to forest, to region, to global) (Medlyn et al. 2017; Yi et al. 2019). One example illustrating this challenge is the divergence in iWUE magnitude and other physiological metrics between tree-ring isotopes, leaf-gas exchange measurements, and estimates from direct measurements of carbon and water exchange using Eddy Covariance flux towers (Medlyn et al. 2017; Keen 2019; Yi et al. 2019; Lavergne et al. 2019). These discrepancies persist at site level, with trees sampled in the vicinity (and footprint) of a flux tower (Belmecheri et al. 2014; Medlyn et al. 2017; Guerrieri et al. 2019). Methodological assessments to identify sampling strategies when reconstructing forest productivity from tree-ring widths have demonstrated the influence of sampling design on estimates of forest growth and productivity (Nehrbass-Ahles et al. 2014). Future research in tree-ring isotopes when applied as records of carbon–water dynamics may benefit from conducting similar sampling design assessments.

Ultimately, site choice and sampling design will depend on the scientific application (paleoclimate, ecophysiology), but also require prior knowledge of environmental and micro-climate conditions at the site(s), either through site reconnaissance, monitoring (Wright 2001; Wright and Leavitt 2006a; Treydte et al. 2014; Belmecheri et al. 2018), or preliminary and published studies.

2.2 Sample Replication

Typically, in a homogenous site, between four and six trees are sufficient to achieve a representative common variability at the site level. This is similar to the strong common signal measured by “Expressed Population Signal” when using tree-ring widths (Wigley et al. 1984; McCarroll and Loader 2004). However, the number of trees required to build a representative isotopic chronology that captures a significant proportion of the site variance depends on the tree species, site conditions and the isotope considered (C, O or H). One approach would be to assess the between-tree variability (or inter-tree differences) on a subset of years-rings (10–30 years) or on only every 5th or 10th ring over the time period of interest before analyzing all of the samples (Leavitt 2010). This exploratory step is particularly relevant when considering pooled versus individual measurements of tree-ring isotopes (see Sect. 4.3.2 of this chapter).

While the majority of isotopic studies report between 4–6 trees as a sufficient sample size to achieve a representative signal (McCarroll and Loader 2004; Leavitt 2010), this number should be considered a minimum to be informed by preliminary isotopic analyses as described above. The number of trees required can be influenced by attributes of individual trees, such as circumferential variability in tree vigor, stand status (e.g. dominant, subdominant, subcanopy), micro-climate and local hydrology, and strongly depends on the environmental signal recorded in the tree-ring isotopes. These factors contribute to circumferential isotope variability within a ring (intra-tree) and between trees within a site (inter-tree). The typical inter-tree isotopic variability is 1–3‰ for δ13C and 1–4‰ for δ18O, whilst the approximate intra-tree variability is 0.5–1.5‰ for δ13C and 0.5–2‰ for δ18O, though this aspect has not been widely studied. Because the goal for paleoclimate reconstructions is to estimate the site mean value, not individual tree values, and because inter-tree isotopic variability is considerably greater than intra-tree variability, sampling a higher number of trees in a given site has been prioritized in order to capture a representative isotopic record with high precision of the sample mean for paleoclimate reconstructions.

The strength of the environmental signal recorded in the tree-ring isotopes can be assessed through (1) inter-series correlations of tree-ring isotopic series and (2) correlations between tree-ring isotopes and environmental signals (e.g. climate data). It is therefore not unusual to use up to 8–10 trees per site (Daux et al. 2011). When the aim is to build robust paleoclimate and physiological reconstructions, the number of trees is determined by the confidence interval represented by the absolute difference between individual tree-ring isotopic time-series (Loader et al. 2013b). In this context, increasing the number of trees will enhance the dominant environmental signal and reduce uncertainties in reconstructions. The absolute difference in the stable isotope values between trees can vary through time, and consequently, the theoretical number of trees required for a representative isotopic chronology may vary, too. Changes in the strength of the common signal through time cannot be determined beforehand, so the number of trees included in developing the isotopic chronology should be if possible.

2.3 Choosing Field Sampling Equipment

Sampling wood from living trees is usually done with an increment borer. Standard diameters of the borer bit include 4.3, 5.15, 10 and 12 mm. The larger borer sizes are often chosen to ensure that an adequate amount of wood is recovered for stable isotope analyses, but these borers are relatively expensive and smaller diameter borers may already be available. It is worthwhile to consider the total mass of α-cellulose that can be recovered from wood based on the average ring width (or the minimum ring width of the sample), or the average width of the desired tree-ring subdivision.

Analytical masses used for isotopic analysis and reported in publications from 2009 to 2019 (n = 285) range from 0.04 to over 1 mg for carbon isotope analysis (median = 0.3 mg; n = 145) and from 0.05 to over 1 mg for oxygen isotope analysis (median = 0.2 mg, n = 140). The exact amount required depends on the analytical method and machine sensitivity. Therefore, determining the sample mass required for isotopic analyses should be confirmed with the analytical laboratory of interest.

Table 4.1 indicates the mass of α-cellulose recoverable from wood samples when using 4.3 mm diameter and 5.15 mm diameter increment borers, assuming 40% α-cellulose. The values in Table 4.1 assume that 10% of the volume is removed during surfacing (e.g., sanding, microtome). The cross-diameter distance equivalent to 10% of the volume, to account for surfacing, is 0.81 mm and 0.68 mm for the 5.15 mm and 4.3 mm borers, respectively. This table can be used to determine the amount of α-cellulose that can be recovered from a specific subdivision increment for the two standard diameters of increment borer. A better estimate for individual tree species can be calculated if the percentage of α-cellulose in the wood is known.

Table 4.1 Alpha-cellulose recovery: Theoretical amount of α-cellulose (mg) recoverable from wood cores using 5.15 and 4.3 mm increment borers and containing 40% α-cellulose. The value of 40% α-cellulose was chosen because it is less than the percentage present in most tree species (N = 83, Table 3, (Pettersen 1984)). These values assume that 10% of the wood was removed during surfacing. For the 5.15 and 4.3 mm borers this means removing wood to a depth of ~0.8 and ~0.7 mm, respectively. The “ring or subdivision width” indicates the width in mm of the whole ring or of a ring subdivision (e.g. earlywood, latewood, thin sections). If the percentage of α-cellulose in the wood of a chosen tree species is known, then a more accurate estimate of the maximum recoverable α-cellulose can be estimated (see text)

The following example describes the procedure to use Table 4.1 for a given tree species and the width of a ring or a ring-subdivision: Pseudotsuga menziesii (Douglas-fir) is reported to contain about 45% α-cellulose. The specific gravity of this species is ~0.47 (though the specific gravity of wood varies with moisture content, ring width, and with the part of the ring considered e.g. earlywood versus latewood). Let’s assume that the trees were sampled using a 5.15 mm borer, and that the cores are sampled at 0.5 mm wide subdivisions. According to Table 4.1, a core from a 5.15 mm borer, for a tree with a wood density of 0.47 and a subdivision width of 0.5 mm will yield ~1.8 mg of α-cellulose. These values are calculated assuming 40% α-cellulose. To convert the α-cellulose to the specific yield for Douglas-fir of 45% α-cellulose, simply multiply the yield from Table 4.1 by the ratio of the actual α-cellulose % to the % of α-cellulose used to make Table 4.1 (1.8 mg * 45%/40% = 2.0 mg α-cellulose).

Material losses during processing, other than during surfacing and chemical extraction, are not included in these calculations, so the table values should be considered the maximum recoverable amount of α-cellulose for a species with 40% α-cellulose. The values presented in the Table 4.1 can be applied to any desired linear distance to be subsampled (e.g., whole ring, earlywood, latewood, etc.), but when considering the density of the wood to be sampled, researchers should be aware that the density of latewood is somewhat greater than both the density of earlywood and the reported average wood density for the tree species.

Consider Pinus ponderosa as another example of the use of the information summarized in Table 4.1. Wood density within a single species can vary greatly depending on the tree age, the rate of growth, seasonal changes in climate, and other factors determining the wood composition. A typical value for the specific gravity (unitless; approximately equal to the density measured in g cm−3) of oven-dried Pinus ponderosa is 0.40 (Miles and Smith 2009), i.e., equal to ca. 0.40 g cm−3. Based on Table 4.1, recovery of 2.0 mg of α-cellulose using a 4.3 mm increment borer, requires a ring width or ring subdivision of at least 0.95 mm. For a 5.15 mm increment borer, the ring width or ring subdivision required to recover 2.0 mg of cellulose is at least 0.67 mm (Table 4.1). For a 12 mm increment borer (not shown), the ring width or ring subdivision must be at least 0.125 mm. For analysis of either δ13C or δ18O in Pinus ponderosa, with two replicates, using the median values typically reported for the analytical masses (δ13C median = 0.3 mg; δ18O median = 0.2 mg), a minimum linear distance (sampling resolution) of 0.3 mm (4.3 mm borer) or 0.2 mm (5.15 mm borer) is required for δ13C and 0.2 mm (4.3 mm borer) or 0.13 mm (5.15 mm borer) is required for δ18O. At least twice the mass is required if both isotopes are to be analyzed. Of course, an assessment that inadequate material will be recovered from a single core can be offset by pooling material from multiple cores (see Sect. 4.3.2 on pooling).

Typically, two cores are sampled per tree, from opposing radial directions across any slope (not upslope or downslope), for establishing tree-ring chronologies, irrespective of the diameter of the increment borer. The same two cores can be used for isotope analyses which will circumvent the potential problem of intra-tree variability (Leavitt and Long 1984) by pooling rings or subdivision of rings while ensuring that sufficient material will be recovered when using smaller diameter increment borers. Albeit, when not constrained by the amount of recovered material, e.g., access to- and use of a 12 mm increment borer or analyzing whole rings, it is not unusual to sample one core per tree.

3 Sample Preparation

3.1 Sampling Resolution

Most tree rings exhibit anatomical characteristics within annual rings that demarcate tissues with relatively large-diameter tracheids/vessels known as earlywood (EW) from adjacent tissues with dense, small-diameter tracheids/vessels, which are known as latewood (LW) (Pallardy 2008). In most trees, the EW forms in the spring and the LW in the summer. In some tree species, other wood anatomical characteristics can occasionally be identified, e.g., the so-called false ring, defined as an abrupt intra-annual density fluctuation (IADF) , consisting of layers of tracheids with small diameters within the EW and sometimes impinging on the LW (Szejner 2011; Battipaglia et al. 2016; Babst et al. 2016; Pacheco et al. 2018; Belmecheri et al. 2018). Depending on the species, these variations are the expression of phenological processes related to xylogenesis, driven by temperature and water availability (Budelsky 1969; Yoshimura and Suzuki 1975; Vaganov et al. 2006; Zalloni et al. 2016). For example, cell enlargement is driven by turgor pressure and accordingly by water availability (Lockhart 1965; Rathgeber et al. 2016), and extreme or abrupt reductions in water availability can result in the formation of IADFs. These anatomical variations can serve as temporal markers that can be used to investigate intra-seasonal variation in tree-climate interactions using stable isotopes (Castagneri et al. 2018; Belmecheri et al. 2018).

The majority of tree-ring isotope studies are based on the analyses of whole rings or latewood portion of the ring, for both Gymnosperms and Angiosperms (specifically ring-porous angiosperms). Tree-ring stable isotope publications surveyed between 2009 and 2019 report 18 studies based on blocks of rings, 237 studies based on whole rings, 30 studies based on latewood subdivision, 28 studies based on both earlywood/latewood subdivisions, and 35 studies based on subdivisions smaller than earlywood/latewood. For deciduous tree-species, the LW portion of the ring is preferred for isotopic analyses because it corresponds to recent photoassimilates as opposed to remobilized carbon from previous growing seasons used for EW (Kagawa et al. 2006a, b). Therefore, the isotopic composition of LW is imprinted by environmental conditions experienced by trees during the current growing season. In contrast, EW in most conifer species rely almost entirely on current photosynthates.

For Gymnosperms, separating LW and EW is not necessary in some cases (Kress et al. 2010; Daux et al. 2011). Analyses of pine species at the European tree line showed that the δ13C signal was coherent amongst EW and LW (Kress et al. 2010). In this case, the coherence of the δ13C isotopic signal amongst these two portions of growth is related to the short duration of the growing season, and smaller carbohydrate pools. In such cases, analyzing the whole ring offers an advantage in the presence of narrow rings and is more cost-effective for developing long isotopic chronologies.

Recently, analyses of δ13C in EW from Quercus species have been used to constrain seasonal dynamics and strategies of carbon storage as they relate the phenological phases to climate trends (e.g. warming) (Kimak and Leuenberger 2015). Using δ13C from EW and LW in mature Quercus trees, McCarroll et al. (2017) demonstrated the existence of two pools consisting of only non-structural carbohydrates (NSCs). For EW formation, trees preferably used younger reserves (NSCs from the previous year), but accessed older reserves (NSCs from earlier years) when poor environmental conditions prevailed during the previous growing season (McCarroll et al. 2017). The inter-seasonal δ13C variations described for Quercus species are consistent with the highly resolved intra-seasonal variations of δ13C described for several broad-leaf species (Fagus, Populus, and Morus). Isotopic measurements were conducted at 100 µm resolution for two consecutive years and showed a consistent tri-phasic δ13C pattern of enrichment-depletion-enrichment (Helle et al. 2004). At the beginning of the EW portion, δ13C values rise to reach the maximum observed values, followed by a decline (gradual or abrupt) to the minimum observed values within the LW portion. The end of the LW is marked by an increase of δ13C values and is similar to the next year (ring) EW δ13C. The increase marks the switch-over from current to stored carbohydrates use. This seasonal pattern is interpreted as reflecting isotopic fractionations related to post-photosynthetic processes and downstream metabolism (Gessler et al. 2009; Offermann et al. 2011) (see Chap. 13), which can lead to a decoupling between the signal recorded in the leaf and that stored in the wood or cellulose (Gessler et al. 2014).

In regions with distinct seasonality such as cool and wet spring and hot dry summers, the EW and LW portions will record different climate signals that can be averaged out if the whole ring is analyzed. Beyond EW and LW temporal resolution, a systematic investigation of intra-annual variations of δ13C and δ18O offers the advantage of identifying the seasonality of the climate signal recorded in various portions of the tree-ring throughout the growing season. This then allows reconstruction of the long-term variability of the seasonal climate signal (Roden et al. 2009; Johnstone et al. 2013). Intra-annual δ18O measured from coast redwoods revealed a consistent δ18O pattern across 10 years and from multiple trees/sites, with the most depleted δ18O recorded in the central portion of the ring. The low δ18O values were related to depleted source water δ18O and a reduction of evaporative enrichment related to fog during the summertime, and were interpreted as signaling a switch from stored to current summertime photosynthates. In this study, it was possible to identify the timing of the seasonal climate signal and therefore the portion of the ring suitable for longer-term reconstructions (Johnstone et al. 2013). This one example illustrates the value of considering what portion of the tree ring will be analyzed, and the relevance of such an approach to enhance the interpretation of the isotopic results but also optimize utilization of resources.

In temperate forests, most isotopic studies of tree rings with temporal resolution higher than earlywood and latewood focused on only few years (2–10 years). In contrast, high resolution intra-annual investigation is very common in tropical forests and particularly for tree-species lacking visible growth rings (Verheyden et al. 2004; Anchukaitis et al. 2008), and has been used to develop long and highly-resolved isotopic records (Poussart et al. 2004; Cintra et al. 2019). High-resolution intra-annual tree-ring studies have improved the knowledge and understanding of the underlying physiological processes of tree response to climate and environmental variability (Walcroft et al. 1997; Leavitt et al. 2002; Barbour et al. 2002; Helle et al. 2004; Roden et al. 2009; Kimak and Leuenberger 2015; McCarroll et al. 2017). This is further illustrated by the use of process-based models for interpretation of seasonal variation in δ13C and δ18O (Walcroft et al. 1997; Barbour et al. 2002; Ogée et al. 2009) (see Chap. 26). A detailed model-data comparison for pine trees over two consecutive growing seasons revealed that a single-substrate model (a simple model for carbohydrate reserve) represents a reasonable assumption for interpreting seasonal variations of tree-ring δ13C and δ18O driven by precipitation and soil moisture (Ogée et al. 2009). Interestingly, a sensitivity analysis of the single substrate model revealed that the modeled EW isotopic signal was sensitive to the onset of the cell-wall thickening phase, during which almost all cellulose is deposited.

Recent insights into xylogenesis have improved the understanding of the timing and duration of cell enlargement and cell-wall thickening phases throughout the tree-ring formation (Cuny et al. 2014, 2015). The duration and rates of cellular differentiation processes are highly variable throughout the growing season (Cuny et al. 2013; Ziaco et al. 2016). For conifers in France, the cell enlargement phase for EW had a longer duration, ~20 days, compared to the cell wall thickening phase of ~5 days. The opposite was observed later in the growing season where LW cells had a shorter enlargement phase and an extended thickening phase, ~50 days (Cuny et al. 2014, 2015). The timing and duration of cell-wall thickening imply that lags between the onset of cell formation and secondary cell-wall formation should be accounted for when aligning climate drivers with isotopic signals recorded in cellulose (Monson et al. 2018; Belmecheri et al. 2018). Until recently, the kinetics of xylogenesis processes were not integrated into the empirical or process-based interpretations of tree-ring isotopic studies (Ogée et al. 2009; Belmecheri et al. 2018). In the absence of such information, tree-ring isotopic studies based on annual or seasonal time resolution can lead to a smoothed or mixed signal of intra-seasonal xylogenesis processes (Chap. 15). For paleoclimatic and ecophysiological inferences, systematic isotopic investigations at high intra-annual resolution offer great potential to gain a better theoretical understanding of the coupling between tree-ring formation, isotopic fractionation and environmental conditions (Wright and Leavitt 2006b; Roden et al. 2009; Vaganov et al. 2011).

The whole ring, EW or LW portions can be accurately and manually separated using a scalpel or a razor blade under a binocular magnifier or stereoscope (e.g. Daux et al. 2011). For high intra-seasonal resolution, a sliding (Barbour et al. 2002; Helle et al. 2004; Ogée et al. 2009) or rotary (Anchukaitis et al. 2008; Szejner et al. 2020b) microtome is used to cut continuous and tangential slices at the desired resolution incrementally. When sampling thin slices with a microtome, it is important to ensure that the ring-boundaries are parallel to each other for an accurate representation of the radial growth progression and that the orientation of the blade is parallel to the fiber direction. Note that the pronounced arc of rings near the pith precludes the use of a microtome for accurate ring subdivision, because of mixing of wood produced during different time periods when cutting with a straight blade across the arc of the ring. Slices as thin as 10 µm can yield enough wood mass for cellulose extraction and isotopic analyses (Table 4.1). This can be achieved using 12 mm tree cores or wood segments (from stem discs).

3.2 Sample Pooling

It is advisable when resources (cost, time) are not limiting to analyze trees individually. The major advantage of analyzing individual trees is the assessment of statistical uncertainties associated with isotopic signal arising from the between-tree variability, but also to correct for age-related trends as discussed above. For physiological and ecological applications, analyzing individual trees enables statistical analysis where a single tree is used as a random or fixed factor when assessing drivers of physiological response of trees to environmental change. This is increasingly used in mixed-effect linear models (Guerrieri et al. 2019).

On the other hand, pooling dated rings from multiple trees reduces sample processing time and resources use. A compromise between these two procedures is a split-pool protocol where rings from multiple trees are pooled for each year but also analyzed separately at a fixed frequency of every 5 or 10 years allowing for inter-tree isotopic errors to be quantified (Wright and Leavitt 2006b; Dorado Liñán et al. 2011; Szejner et al. 2016) (see Chap. 6). Rings from multiple trees for any given year can be pooled independently of their mass differences (but see Dorado Liñán et al. 2011), though adequate sample homogeneity pre- or post-chemical extraction (see following Sect. 4.3.3) is crucial for obtaining a composite (pooled) isotopic time-series that is similar to the average of the individual tree measurements (Leavitt 2010; Dorado Liñán et al. 2011).

Beyond resources, the study aim will determine the sample preparation protocol. For instance, it is possible to pool consecutive years within a tree (blocks of several rings, typically 5 year blocks) to develop long isotope chronologies when the goal is to investigate paleoclimate trends and low-frequency variations (Mayr et al. 2003; Gagen et al. 2012). In such cases, pooling across years seems to be a practical choice that yields a robust isotope-based reconstruction with sufficient sample replication, but the method may not be suitable for some physiological investigations of long-term trends given that pooling adjacent years may involve mixing the isotopic signals of EW and LW sections and their sources, respectively. Nevertheless, multiple-year pooling with individual tree analysis offers the advantage of estimating the confidence interval of an isotope chronology mean and therefore of paleoclimate reconstructions (Boettger and Friedrich 2009). There are different strategies to develop isotope chronologies using multiple-year pooling. The first one consists simply of serial pooling of tree-ring blocks from an individual tree (Leavitt and Lara 1994). The second one consists of pooling tree-ring blocks from an individual tree, but shifted by one year between trees, which produces a “quasi-annual” pooled chronology that retains replication in each year (Boettger and Friedrich 2009). The third strategy consists of supplementing the second strategy by analyzing a larger number of trees where tree-cohorts join (e.g. ~10 trees for cohort-join points vs 5 trees for shifted tree-ring blocks). This technique is termed Offset-pool plus Join-Point and was developed by Gagen et al. (2012) to rapidly generate a robust millennial length δ13C chronology by combining living trees and sub-fossil wood. However, such reconstruction relied on prior work in the same site exploring the strength of the climate signal as well as age-related trends, and consequently can be applied in similar settings to the original study.

In summary a balance between the objectives of the study and available resources will determine the choice and the relevance of analyzing trees individually, pooling rings within individual trees, or pooling the same sampling units from multiple trees prior to chemical extraction of cellulose or isotopic measurements. Yet, the ability to provide uncertainty estimates should remain a central goal and the strategies described above offer a few options to achieve this goal.

3.3 Particle Size Requirements for Chemical Extraction and Analytical Repeatability

When α-cellulose is to be analyzed for stable isotopes of δ13C and δ18O, we recommend the following steps: (1) reduction of the wood sample prior to chemical extraction, and (2) further particle size reduction using an ultrasonic probe after chemical extraction (Laumer et al. 2009). Large differences in the stable isotope ratios can occur across a tree ring, within the same year, between different sides of the same tree and between neighboring trees of the same species (e.g., Leavitt and Long 1986; Saurer et al. 1997). The current analytical masses used are typically a few hundred micrograms, and usually a subsample from each sample is analyzed, not the entire sample (e.g. whole tree ring or a subdivision of a tree ring). When multiple samples are pooled as described above, homogeneity of the sample is critical for precision (unless the sample mass from which the subsample is drawn is not much larger than the size of the subsample). The subsample must be isotopically representative of the average value for the entire sample within the precision of mass spectrometry, whether pooled or not, or the accuracy and repeatability of the results of the analysis will not be maximized (see Chap. 6 for more details on accuracy and precision). The maximum acceptable particle size that will ensure homogeneity within a pooled or unpooled wood sample is somewhat dependent on the wood density of the tree species, and intra-annual variability in density (e.g. earlywood/latewood), but the most important consideration is the relationship between analytical mass of the sample and of the subsample chosen for stable isotope analysis.

In 1998, Borella et al., published results of a study that considered the number of particles required for analysis given (1) a typical wood density of ~600 kg/m3, (2) a deviation of 1 or 1.5‰ in the measurements of the isotopic ratios within the original sample, (3) the mass of the original sample (immediately prior to analysis), (4) the desired ‰ deviation from the actual value (precision), and (5) the probability that a given particle size will yield a value within 1σ or 2σ of the actual value. At that time, Borella et al. (1998) stated that “a grain mass of <6 μg should be achieved to ensure an accuracy of 0.15‰”, which they report as being achievable using a sieve size of 0.15 mm (~80 mesh) for cellulose and 0.10 mm (120 mesh) for whole wood.

Advances in instrumentation now allow analysis of much smaller masses than considered by Borella et al. (1998), where a mass of 1–1.5 mg out of an original post-processing mass of 10–50 mg was used for the online analyses, a range for the subsample of ~2–10% of the original sample mass. Over the last 10 years of publications reporting stable isotope results from wood (2009–2019), the average analytical mass reported for δ13C analyses was 0.67 ± 0.6 mg (N = 145; median value = 0.35 mg), and the average mass used for δ18O was 0.39 ± 0.3 mg (N = 140; median value = 0.25 mg). The current median values for analytical masses of 0.3 mg (δ13C) and 0.2 mg (δ18O) are less than ~20–30% of the analytical values used by Borella et al. (1998), suggesting that the required particle mass under current conditions should be ~20–30% of the 6 μg maximum particle mass recommended by Borella et al. (1998), or between 1 and 2 μg. Particle masses of this size are achievable using some mills, but the resulting particle size is likely to be smaller than the pore size of the filtering media used during cellulose extraction, resulting in sample loss during chemical processing (see below).

Note, however, that Borella et al. (1998) state that the results presented in their Table 1, from which their recommendation was drawn, are for an “extreme case” where two samples that differed by 2‰ were combined. Individual tree rings, or combinations of the same year using multiple cores from the same tree are unlikely to contain two pools that consistently differ by 2‰, at least for δ13C, for which their study was done. The variability of δ18O values across a tree ring and between the same years on different trees is larger than the spread for δ13C. Examples of large intra-annual variations in δ18O in tree-ring α-cellulose are shown in Fig. 4.3, with intra-annual ranges in the δ18O across thin sections from two years at 4 sites of >6 and >8‰. In this case the within-ring variability is much larger than the between-tree (i.e. between-site) variability.

What is the best method to use for obtaining the particle mass necessary to ensure homogeneity?

Many methods have been employed for reducing the sizes of wood particles prior to chemical extraction (Fig. 4.1), including manual cutting (scalpel, razor blade, etc.), mortar and pestle, and various kinds of milling. Many types of mills are available for wood sample reduction, including grinding/pulverizing mills (Yokoyama and Inoue 2007), impact mills (Nied 2007) and robotic micro-milling (Dodd et al. 2008), although the latter can be extremely time-consuming (Voelker et al. 2018). Milling has been the most common pre-chemical extraction method for particle size reduction prior to chemical extraction over the last 10 years (Fig. 4.1). Reduction methods such as microtome thin sectioning, laser ablation (Drew et al. 2009; Soudant et al. 2016) and the plate method (Li et al. 2011; Kagawa et al. 2015) can circumvent any questions about maximum acceptable particle size, unless the samples are to be pooled, in which case an additional particle reduction step may be required to ensure homogeneity. Yet minimizing the size of the particles prior to chemical extraction is limited by another factor. The resulting particle sizes from some mills are small enough to ensure homogeneity, but most methods for chemical extraction involve the use of filtering media (The Brendel Method is one exception—see Chap. 5 for cellulose extraction details). Filtering media, such as fritted glass filters (No. 2, porosity 40–100 μm; e.g., Loader et al. 1997; Rinne et al. 2005) and polymer pouching material (ANKOM F57; porosity 25 μm; e.g., Szejner et al. 2016, 2018; Belmecheri et al. 2018) have porosities much larger than the particle sizes obtainable through many milling methods. Milling to too fine a particle size will result in substantial sample loss during the subsequent chemical extraction but milling to a very coarse size can result in incomplete chemical extraction. Ideally the particle size after milling will be only slightly larger than the porosity of any filtering media. For example, pre-extraction milling to 20 mesh (~840 microns) can yield adequate particle size. Yet, the suitable particle size for the filtering media is too large to provide adequate homogenization of the sample when small samples masses are being analyzed, we therefore recommend a further reduction of the particle sizes after chemical extraction through the use of an ultrasonic probe (Laumer et al. 2009), which results in complete separation and mixing of the α-cellulose fibers.

Fig. 4.1
figure 1

Methods used to obtain the final particle size analyzed in research published between 2009 and 2019 (n = 342). No particle size reduction is involved in the “Plate Method” which is based on cellulose extraction of a whole tree-ring lath (cross-section). Other methods such as laser ablation and microtome thin sectioning are not plotted because they represented only a small percent of the papers. Note the increase in the percentage of papers using the ultrasonic probe method [Dark blue bars] since the method was first described by Laumer et al (2009) and the covariance of the use of this method in recent years with the use of various forms of the word ‘homogenous’ when referring to the analytical subsample [Red bars]

In summary, given the changes in instrumentation that now allow the analysis of δ13C and δ18O from wood-derived samples in the range of 100–200 μg, we recommend reduction of the wood to a particle size larger than the filtering media prior to chemical extraction, to ensure the completion of the chemical extraction and to reduce the length of the fibers, followed by further particle size reduction using an ultrasonic probe (Laumer et al. 2009). This combination yields samples with the greatest possible homogeneity for the typical methods employed prior to stable isotope analysis of α-cellulose.

4 Towards Subseasonal-Resolution Analyses of Tree-Ring Records

4.1 Important Considerations

The appropriate sampling resolution for stable isotope analysis in tree-rings depends on many factors, including the research question, the width of the rings of the chosen trees, and the climate variability and repeating patterns within that variability, factors that are best expressed in the following questions: What temporal resolution would best answer my research question? Is there enough material available in my samples to allow sub-tree-ring sampling? How much does the climate vary across each growing season at my tree site? Are there intra-annual climate changes at my site that are temporally consistent on an inter-annual basis?

Increasing the number of samples per tree-ring may provide additional data. But to be useful in any given research effort, the increase in the sampling resolution must provide additional information that is pertinent to answering the research question. If finer temporal gradations in the stable isotope data (as a proxy for either environmental or physiological changes) would better inform the research effort and/or increase the certainty of some aspect of the research, then an increase in the sampling resolution is worth considering.

Practical limitations such as narrow tree rings and/or smaller diameter core samples may confound efforts to increase the sampling resolution. The amount of material (wood, α-cellulose, etc.) required for stable isotope analysis will determine, in part, whether increased sampling resolution is possible. This consideration was addressed in detail earlier in this chapter. The amount of time required to mature the secondary cell wall of each cell (where almost all the “wood” resides) is another consideration that may potentially limit the value of increasing the sampling resolution. The secondary cell wall maturation timing can range from a few weeks to a few months (Cuny et al. 2015), with the maturation of neighboring cells to the inside or outside of the tree ring being slightly ahead or behind any particular cell. The practical result is a smoothing of the stable isotope signal relative to the stable isotope values that would have been present at any given point in time in the newly formed photoassimilates. The degree of overlap in maturation across a tree ring will limit the recoverable temporal resolution of the factor of interest, i.e. environmental changes or physiological changes. Finally, for high resolution sampling to be useful, there must be greater variability in the higher resolution data than in the lower resolution data, and ideally that variability will recur in a temporally consistent inter-annual pattern. Stable isotope values that vary only slightly across each tree ring will probably not provide enough additional information to justify the additional cost and efforts. Moreover, stable isotope values that vary across each tree ring in each year, but in a non-systematic way, may be difficult to interpret, especially for tree rings formed prior to the existence of local weather data.

Many climate regimes have consistent cool season to warm season shifts in temperature, and some regions also have consistent intra-annual patterns in the precipitation. Monsoonal and Mediterranean climates are good examples of climate regimes with consistent inter-annual and intra-annual climate changes. Such distinct climate variations can result in seasonal and sub-seasonal changes in the stable isotope ratios fixed in the tree-ring cellulose and can even produce visual evidence of the seasonal changes in the wood (i.e. false latewood bands-FLBs, also called inter-annual density fluctuations-IADFs). In these cases, stable isotope analysis at high-resolution within each tree ring (e.g., intra-annual thin sections, or similar) can provide useful seasonal and even sub-seasonal information, both during the period of instrumented weather data and before (Szymczak et al. 2019; Pacheco et al. 2020), with appropriate calibration/verification of the stable isotope time series with the instrumental data.

4.2 Sampling Resolution Comparison

As an example of the potential value of high-resolution (intra-annual) sampling of tree rings, we combined high-resolution stable isotope values (100 µm) sampled across two consecutive tree rings to represent a range of sampling resolutions from low resolution (whole tree ring) to high resolution (100 µm; Fig. 4.2). The stable isotope data used to produce Fig. 4.2 came from α-cellulose in a tree that grew in a monsoonal climate (i.e. the North American Monsoon in Southwestern US).

Fig. 4.2
figure 2

Subseasonal variation of tree-ring α-cellulose δ18O recorded across a range of sampling resolutions. Lower resolution δ18O data points were produced by averaging the δ18O values from the 100 µm thin sections (circles). The whole ring value is indicated by the dashed line. An earlywood/latewood tree-ring subdivision is indicated by the thick green line (earlywood) and the thick blue line (latewood). A 4-part tree-ring subdivision (quartiles) is indicated by the brown lines. The grey line is a 30 day smoothing spline of the maximum daily VPD (hPa) provided here to contextualize environmental drivers of subseasonal δ18O variations recorded in tree-ring α-cellulose. Note that the Y-axis for the VPD does not extend to the top of the graph. The position of the false ring within each year is indicated with a filled circle. DOY = Day of the Year

What value was added in this case by having higher resolution data?

  1. (1)

    Whole-ring—Note that the whole ring δ18O values (dashed lines) for the two years differ by about 2‰. This 2‰ difference between the two years would be the extent of the information available with whole ring analysis.

  2. (2)

    Earlywood/Latewood—A further subdivision into earlywood (springwood) and latewood (summerwood) reveals that the latewood δ18O values are enriched compared with the whole ring δ18O values, and that the earlywood δ18O values are slightly depleted compared with the whole ring δ18O values, but the overall pattern is similar between the two years. The Earlywood δ18O values between the two years are similar, but there is a 2‰ difference between the latewood in the two years. Information about differences between the seasons, with similarities between the annual earlywood/latewood pattern, is revealed by using two intra-ring sampling subdivisions.

  3. (3)

    Quartiles—Further subdivision of the tree-ring δ18O data into quartiles begins to reveal additional intra-annual isotope information when compared with the whole ring δ18O: (a) the first quartile is very enriched, (b) the second quartile is very depleted, (c) the third quartile reveals differences between the two years, though both are depleted relative to the whole ring value, and (d) the fourth quartile, roughly equivalent to the latewood, is slightly enriched. The variability of the δ18O above and below the tree-ring mean δ18O value would not be recognized with either whole ring or earlywood/latewood analysis.

  4. (4)

    100 µm thin sections—Values from the original data (circles) indicate transitions across the tree rings, with a very enriched first three slices, a 3–4 slice transition to the most depleted values, followed by a gradual increase in the δ18O values until the end of the growing seasons. Arguably, the results from the quartile analysis provide almost the same information recovered in the 100 µm thin sections, though information that aids the interpretation will be added in the following case study.

4.3 Case Study: Pinus Ponderosa Growing in Southwestern US [Southern Arizona]

In the following section, the high-resolution δ18O subseasonal variations presented above, is further replicated in three additional tree sites to investigate the consistency of the seasonal pattern discussed earlier and its spatial coherence across Southern Arizona.

The annual precipitation in Southern Arizona is consistently bimodal, with a cool-season component and a warm season (monsoon) component (Sheppard et al. 2002). A hot and extremely dry period from April to June always separates these two precipitation modes. During the dry period, tree species growing at high elevations (~2000–2600 m) experience very high transpiration rates caused initially by high vapor pressure deficit (VPD) which is later combined with depleted soil moisture. The combination of extremely dry air and low soil moisture causes the growth in these trees to slow, often resulting in a reduction in the cell sizes (Belmecheri et al. 2018). Following the onset of the North American Monsoon (NAM) precipitation (July 7th ±9 days; Higgins et al. 1999), many of the trees are able to resume their cambial activity and growth (Budelsky 1969). In these cases, the smaller lignified cells produced during the dry period are followed by larger cells, forming a visual growth transition. This feature, often called a “false ring”, “interannual density fluctuation (IADF)” , or “false latewood band (FLB)”, when combined with the consistent timing of the onset of the NAM, allows the FLBs to be used as a visual intra-annual time marker, separating spring growth from growth in the summer and autumn.

Examination of the α-cellulose δ18O extracted from 100 μm thin sections (using a rotary microtome) from high elevation trees growing in two proximate mountain ranges (The Santa Catalina and Pinaleño mountains), at four different elevations (between 2040–2680 m) and during two consecutive years (1998 and 1999) reveals consistent δ18O patterns across the growing seasons in all cases (Fig. 4.3). In general, the δ18O values in cool season precipitation in Arizona are more depleted than those in warm season precipitation, because of the effect of cooler temperatures on the condensation fractionation (Eastoe and Dettman 2016). Stable isotope analyses of precipitation collected at a location close to two of the studied sites (HEL and WCP) show the expected cool season/warm season pattern in precipitation δ18O (Wright 2001). The timing of budburst at one of the sites (HEL) indicated that the growing seasons began in mid-April in both years at a time when soil moisture was close to field capacity (Wright and Leavitt 2006a). The comparisons of the precipitation δ18O values with soil and xylem water δ18O showed strong covariance, indicating little or no groundwater access (Belmecheri et al. 2018). Yet the δ18O values in the first few slices from each year at all the sites were much higher than the δ18O values measured on slices produced during the summer season, the opposite of expectations if the cool season/warm season precipitation δ18O signal was dominating the δ18O values in the photoassimilates.

Fig. 4.3
figure 3

Tree-ring δ18O (W. E. Wright, unpublished data) from 100 μm thin sections for two consecutive years 1998 CE [Left] and 1999 CE [Right]. The data comes from samples taken at four sites located in two Arizona mountain ranges (U.S. Southwest) separated by about 80–90 km. False Latewood Bands (FLBs) occurred in both years at the lower elevation sites [UAC and HEL, circles]. In the 1998 graph, the δ18O values for the FLBs [filled circles] occurred at the same point in time, but their δ18O was different. In the 1999 graph, δ18O values for the FLBs from the two sites plot almost at the same point in time and have very similar δ18O, so the symbols are indistinguishable on the graph. The higher elevation sites [CPP and WCP, triangles] did not form FLBs in these years, though the intra-annual δ18O patterns are very similar. Note that the x-axis values for the δ18O from all the sites are plotted at constant time intervals relative to starting dates of Day 135 (May 15) and ending dates of Day 304 (October 31), based on observed leaf extension timing, which roughly correspond to the growing season (Wright and Leavitt 2006a). Consequently, the placement of the δ18O data points relative to the x-axis is based on estimates for the beginning and end of the growing seasons, and therefore does not include the shift forward in timing required to account for the maturation of the secondary cell walls (i.e. about 4–6 weeks on average)

Approximately 58% of the final δ18O value in xylem α-cellulose comes from the leaf water, while about 42% of the oxygen atoms in the photoassimilates produced using leaf water subsequently exchange with stem water during cellulose synthesis (Roden and Ehleringer 1999). It is the 58% of the oxygen atoms that retain values related to the leaf water enrichment that provide one possible explanation of the observed cellulose δ18O seasonal pattern. The degree of evaporative enrichment of leaf water δ18O differs greatly with VPD as mediated by water availability in the soil, such that leaf water δ18O values during transpiration may be the same as the source water δ18O if the atmospheric humidity is 100%, but will become very enriched if transpiration occurs into very dry air. The very enriched early growth in the data shown in Fig. 4.3 is consistent with stable isotope model outputs using local climate data, which suggests that extreme evaporative enrichment caused by high VPD may be the main driver of the high δ18O values (Belmecheri et al. 2018). An alternative interpretation is the use of enriched stored photoassimilates. But the important point here is that coarse resolution sampling would not have revealed the unusual early growing season values.

Another detail that would not have been revealed by coarser sampling is the temporal relationship of the FLBs (IADFs) to the stable isotope values. A FLB is formed during a period of extreme water stress, which causes partial or complete growth cessation resulting in the production of smaller cells. Yet in these examples the summer rains began soon after the formation of the FLB, so most of the maturation occurred under conditions of high humidity (lower vapor pressure deficit). The combined effects of reduced evaporative enrichment of the leaf water and increased diffusion of depleted water vapor 18O into the stomata result in leaf water δ18O that is less enriched than during the dry period. Photoassimilates formed using this summer leaf water are then used to mature the small cells in the FLB. Consequently, the δ18O value for the thin section containing the FLB is one of the most depleted in both tree rings (Belmecheri et al. 2018). This pattern can be explained by the kinetics (timing and duration) of cell-wall enlargement and thickening phases for the FLB cells. While cell-wall enlargement of the FLBs is initiated during the arid period, secondary cell-wall deposition begins and extends through the moist monsoon period (July to mid-September).

Evidence for the offset in the timing between the appearance of the small cells and the maturation of the secondary cell walls of the small cells would not be available at a coarser sampling resolution. In addition, thin section analysis allowed us to recognize that the intra-annual patterns in the δ18O of the same years were very similar, and that these δ18O pattern similarities were independent of the presence or absence of a FLB (Fig. 4.3).

4.4 Preliminary Assessments

High-resolution intra-annual sampling can also be used in preliminary assessments to inform subsequent sampling for long-term reconstructions. In some instances, answering a research question may require only information from a limited part of the growing season, so information provided by analysis of many tree-ring subdivisions may be superfluous. In one example, the research question involved determining the position of a summer fog drip signal within tree rings of Sequoia sempervirens (Roden et al. 2009). The researchers already knew that the stable isotope values in fog drip were always distinguishable from other water sources. Preliminary high-resolution sampling allowed the researchers to focus their subsequent sampling on only the portion of each tree ring that was incorporating fog drip. They were also able to account for some time variance in the fog drip by expanding the sampling increment. The sampling focus and the modification of the width of the sampling unit were facilitated by the preliminary high-resolution sampling.

The data presented in the Southwestern US case study described earlier was also conducted as a preliminary assessment. Initially, the relationships between the presence or absence of FLBs and seasonal transitions in precipitation occurrence were unknown. There was no FLB within the rings in many tree rings at some sites, so an assessment of the high resolution differences between rings with and without FLBs was necessary to be certain that the visual criteria we would use to make subseasonal subdivisions (5 per ring in the case of our studies Szejner et al., 2020) would capture adequate details about intra-annual shifts in the source water and in water stress whether or not there was a FLB present. Careful preliminary assessments can help researchers to optimize information gained from stable isotope analysis when research resources are limited.

5 Conclusion

As the application of stable isotope measurements in tree rings is becoming essential in the fields of paleoclimate and paleophysiology, its full potential is still limited by time and cost of sample preparation and processing. These constraints persist despite remarkable progress in analytical techniques and a better understanding of biases and uncertainties related to sampling of trees (number and status of trees) and rings (temporal resolution). When planning tree-ring isotope studies, balancing project goals and resources is often unavoidable; however, the goal should still be to produce robust measurements of the sample mean at any chosen temporal-resolution, and to quantify uncertainties of these measurements. Analyzing individual trees allows uncertainties in the paleoclimate or paleophysiology reconstructions to be assessed. Otherwise, choosing between individual trees vs pooling multiple trees, or choosing between whole wood and α-cellulose should at least be informed by preliminary analyses on a subset of years/trees. In all cases, sample homogeneity (wood or α-cellulose) is key to achieving precise and representative (within or between tree variability) measurements. Advances in xylogenesis (wood phenology), wood anatomy and mechanistic understanding of how stable isotopes are incorporated in tree-rings offer an opportunity to shift the temporal-scale perspective at which ecophysiological processes, forest-environment interactions, and paleoclimate reconstructions are investigated using tree-ring stable isotopes.

Tree-ring stable isotope studies in conjunction with process-based models, and xylogenesis studies can inform the next generation of vegetation modeling efforts (Zuidema et al. 2018; Friend et al. 2019). With scientific motivations and resource limitations in mind, sampling for isotopic analyses can be approached from the perspective of exploiting and leveraging the same data for both ecophysiology and paleoclimate studies, while reducing uncertainties in estimates and inferences of forest growth, ecophysiology, and climate sensitivity.