Various plant species and communities exhibit distinct patterns of n-alkane homologues, tipically characterised by a pronounced odd-over-even predominance, while most bacteria and few plant species preferentially produce even carbon numbered homologues (Dembicki et al., 1976 and refs. therein).

The shortest homologues (<C₂₁, typically C₁₇) are commonly associated with bacteria and algae (Han and Calvin, 1969; Arp et al., 1999; Sachse and Sachs, 2008; Chatterjee et al., 2023). Sphagnum spp. (C₂₃; Nott et al., 2000) and aquatic plant chain lengths tend to peak at C_21–25, in contrast to terrestrial plants, which often that peak at C_27–33 (Cranwell et al., 1987; Ficken et al., 2000; Gao et al., 2011; Bush and McInerney, 2013; Liu H. et al., 2019), although recent research challenges this dichotomy (e.g., Stefanescu et al., 2023).

Among terrestrial plants, trees/shrubs typically exhibit maximum abundances at C_27–29 while grasses show a predominance of C₃₁ (Cranwell, 1984; Meyers, 2003; Struck et al., 2020). While broad observations suggest that the average chain length (ACL) correlates with temperature/aridity-driven changes in production preferences along environmental gradients, different plant species may exhibit opposing behaviours (Hoffmann et al., 2013; Tipple and Pagani, 2013; Bush and McInerney, 2015; Teunissen van Manen et al., 2019).

Although ACL differences have been employed to discriminate among plant groups in particular environments (e.g., distinguishing gymnosperms/angiosperms and monocots/dicots on the Chinese Loess plateau; Liu et al., 2018), no definitive universally applicable correlation of ACL to plant groups has been established to date (e.g., Bush and McInerney, 2013). Recent studies emphasize the need to complement ACL and other indices (e.g., aquatic plant index) with additional other indicators (e.g., ¹³C-composition of n-alkanes) to effectively distinguish sources among submersed, floating, and emergent plants (e.g., Yu et al., 2021).

The hydrogen isotopic composition of leaf wax n-alkanes (δ²H_wax) from modern plants has been found to generally correlate with the δ²H value of the plant source water, which often reflects precipitation δ²H (δ²H_p; e.g., Sessions et al., 1999; Sauer et al., 2001; Sachse et al., 2004b). In fact, the δ²H_wax values from modern sedimentary wax compounds globally also robustly record local δ²H_p values (Ladd et al., 2018; Liu and An, 2019; McFarlin et al., 2019). On this basis, the δ²H of sedimentary n-alkanes, particularly of their C₂₉ and C₃₁ homologues, recovered both in marine and continental cores as well as in biologically reworked deposits (e.g., mammals’ middens; Chase et al., 2019), has been used as a tracer for paleo-precipitation isotopic composition (e.g., Sachse et al., 2004a; Niedermeyer et al., 2014, 2016b; Tierney et al., 2017; Ardenghi et al., 2019). The accuracy of δ²H_wax and chain length distribution as paleo-climate indicators relies on the understanding of n-alkane synthesis and of the influence and interaction of several internal (plant physiology) and external (environmental) variables.

As n-alkanes retain the δ²H values established at the time of biosynthesis (Sachse et al., 2012), the average δ²H_wax of single leaves can be biased toward growth water values available at the early stages of leaf growth (Sachse et al., 2010, 2015; Tipple et al., 2013; Tipple and Pagani, 2013; Gamarra and Kahmen, 2017). This may be caused by the timing of leaf synthesis (e.g., early growth often relies on synthates stored at the end of the previous year; Gao et al., 2015; Newberry et al., 2015; Freimuth et al., 2017) and, more generally, changes in the biosynthetic water pool linked to external factors such as temperature and relative humidity (Hou et al., 2008; Kahmen et al., 2008; Zhou et al., 2011; Douglas et al., 2012; Gao et al., 2014b; Bai et al., 2019; Jacob et al., 2021; Eensalu et al., 2023), light intensity (Yang et al., 2009), salinity (Ladd and Sachs, 2012; He et al., 2017; Ceccopieri et al., 2021; Wang et al., 2022), seasonality (Sessions, 2006; Pedentchouk et al., 2008; Kahmen et al., 2011; Newberry et al., 2015; Liu et al., 2017), elevation (e.g., Pérez-Angel et al., 2022) and precipitation δ²H (δ²H_p; Sachse et al., 2006, 2010, 2012; Liu and Yang, 2008; Rao et al., 2009; Feakins and Sessions, 2010).

Internal factors, linked to plant physiology (e.g., biosynthetic pathway, rooting depth) can also contribute to the variation in δ²H_wax (Smith and Freeman, 2006; Gao et al., 2015; Cormier et al., 2018, 2019; Eley et al., 2018) and modulate the correlation between δ²H_p and δ²H_wax, normally expressed as the net (or apparent) ²H-fractionation between n-alkanes and mean annual precipitation (e.g., ε_29/MAP) (Liu et al., 2006, 2016; Hou et al., 2007b; Sachse et al., 2012; Gao et al., 2014a; Gamarra et al., 2016; Eley et al., 2018). Despite the resulting high inter- and intra-specific variability characterising δ²H_wax (e.g., 40‰ between leaves of the same plant; Sachse et al., 2009; Newberry et al., 2015), larger datasets reveal a generally stable correlation between δ²H_wax and δ²H_p in plant communities and derived sediments (Chikaraishi and Naraoka, 2003; Sachse et al., 2006; Hou et al., 2008; Rao et al., 2009; Feakins and Sessions, 2010; Chen et al., 2022). Recent interpretive structural modelling has grouped and hierarchically classified multiple controlling factors such as δ²H_p, evapotranspiration, and plant types, according to their relative impact on δ²H_wax variability (Liu and An, 2018).

For this work we selected two sampling sites (Figure 1): one, the Tenaghi Philippon (TP) peatland in NE Greece (Figures 1, 2A), is the site of a historical permanent large wetland, recently (1930s) converted to farmland, as well as the site of important sedimentary paleoclimate archive (e.g., Wijmstra, 1969; Tzedakis et al., 2006; Pross et al., 2015; Schemmel et al., 2016; Ardenghi et al., 2019), and one, the Nisí fen, is a relatively undisturbed permanent-seasonal marshland at the Edessaios river’s source in NW Greece (Figures 1, 2B) and a good ecological analogue for the original TP environment (Kalaitzidis, 2007). Both the original TP marshland and Nisí fen are examples of topogenous mesotrophic mires (i.e., mires that are predominantly fed by local karstic aquifers) situated in the lowest areas of intra-mountainous basins. The two sites share similar basic climatic parameters (temperature, precipitation; Table 1) and the same Köppen-Geiger climate classification (Csa, i.e., warm temperate with arid, hot summers; Kottek et al., 2006) with localised humid conditions during most of the year, and occasional frost episodes, due to thermal inversion and cold bursts from the northern mountain chains, punctuating the relatively mild and wet winters (Christanis, 1994; Pross et al., 2015; Hellenic National Meteorological Service, 2018).

The TP peatland is the result of millions of years of accumulation of Cyperaceae and other peat-forming helophytes in what once was a fen environment characterised by alkaline conditions (Kalaitzidis, 2007). The original wetland was almost completely drained in the 1930s, resulting also in the alteration of its natural water regime. However, stable residual pockets of helophytic communities survive in areas along the draining canals. Circa 16% of the samples were collected from these areas, in an effort to provide a basis for direct data comparison between TP and the mostly undisturbed Nisí fen site.

We selected six species of widespread perennial monocots: the C₃ graminoids Carex riparia (greater pond sedge), Cladium mariscus (saw-sedge), Phragmites australis (common reed), Typha angustifolia (narrowleaf cattail) and Scirpus lacustris (common club-rush), and the C₄ graminoid Cyperus longus (galingale). We focused on helophytes, firstly graminoids, as they have a preeminent role in the formation of peat deposits in these types of marshlands, which often produce important paleoclimatic archives (e.g., Miola et al., 2006; Borromei et al., 2010; Chawchai et al., 2015; Pross et al., 2015) and are not Sphagnum dominated such as acidic (e.g., Carex and Cladium are phylo-calcareous genera; Buczek, 2005; Theocharopoulos et al., 2006) peat-forming oligotrophic bogs (Anderson et al., 2013).

To cover a broader range of wetland species and plant forms, we sampled a group of five perennial eudicot species, the C₃ forbs Cirsium palustre (swamp thistle, biennial/perennial), Stachys palustris (marsh woundwort), Mentha aquatica (water mint), Lythrum salicaria (purple loosestrife), Galium uliginosum (fen bedstraw), as well as the C₃ pteridophyte Equisetum fluviatile (water horsetail) and the C₃ the perennial basal angiosperm Nymphaea alba (white water lily, a rooted hydrophyte with subaerial floating leaves). The latter four species have been sampled in smaller numbers (Supplementary Table S1) due to difficulties in locating/reaching them (L. salicaria, G. uliginosum, N. alba) and different morphology (E. fluviatile). All the plants listed above are common Eurasian wetland species, which are also widespread worldwide at the genus level (e.g., McNaughton, 1966; Mulligan et al., 1983; Theocharopoulos et al., 2006; Bryson and Carter, 2008; Trin et al., 2014).

The concentration of n-alkanes (n-C_14–37) in our plant samples (Supplementary Figure S2; Table 2) exhibited a wide range, spanning from 1.3 to 2,147.3 μg/g. In contrast, soil samples displayed relatively lower and less variable concentrations, ranging from 0.4 to 45.5 μg/g).

Concentrations were higher in eudicots/forbs and lower in monocots/graminoids. Leaves consistently showed approximately the same to ∼80 times more n-alkanes than their corresponding lower stem samples (Table 2). This distinction between leaves and stems was notably pronounced among forbs in spring, while graminoids, such as C. mariscus and S. lacustris, exhibited less overall leaf/stem differentiation. Our data revealed no consistent difference in n-alkane concentrations between spring and summer at either site. Nevertheless, the leaves of four species (S. Palustris, C. longus, S. lacustris in Nisí, and P. australis in TP) exhibited a spring-to-summer increase in mean concentration surpassing their respective standard deviation (Table 2). Similarly, while no distinct difference emerged between samples of the same species from Nisí and TP (only four species in common; Supplementary Table S1), leaves of C. longus and P. australis showed higher n-alkane concentration in TP, albeit often close to the respective standard deviation values; Table 2).

The carbon preference index (CPI) exhibited a wide range across samples: from 0.4 to 86.2 in leaves (with a mean of specific averages at 16.1 ± 7.6), 1.8 to 54.4 in lower stems (mean of specific averages at 14.4 ± 7.1), and 0.9 to 10.4 in soil samples (mean of specific averages at 5.2 ± 1.7). Overall, CPI displayed considerable inter- and intra-specific variability, with all samples recording values well above 1, except for three C. longus leaf samples, characterised by low n-alkane concentrations (Table 2). This suggests a prevalent odd-over-even predominance, notably stronger in plant samples (with a mean of 15.2 ± 7.3) than in soil samples. The sole discernible pattern was seasonal, with CPI generally decreasing from spring to summer, especially in graminoids (Supplementary Table S3) in both leaf and lower stem samples.

The average chain length index (ACL) exhibited a diverse range across samples: from 25.5 to 32.0 in leaves (with a mean of specific averages at 29.0 ± 0.5), 21.4 to 31.7 in stems (mean of specific averages at 27.7 ± 0.6), and 26.4 to 29.4 in soil samples (mean of specific averages at 28.0 ± 0.7). Most samples fell within the typical ACL range for terrestrial plants (27–31), including the floating hydrophyte N. alba. The exception was E. fluviatile, the only pteridophyte (i.e., seedless/flowerless vascular plants, mostly ferns) analysed in this study, with an ACL range of 25–26. Eudicots/forbs and leaves showed ACL higher than, respectively, monocots/graminoids and stems (Supplementary Table S3), as well as soil samples (ca. 27–29). ACL also showed a general increase from spring to summer in eudicot/forb leaves (average increase of 0.4–2.2), although stability was observed in monocot/graminoid leaves. Most stem samples showed an average increase of 0.6–5.8 during this seasonal transition.

Plant wax n-alkane δ²H values in the C_20–35 range exhibited a broad spectrum, ranging from −281‰ to −48‰ (mean ≈ −180‰, σ ≈ 41‰; detailed descriptive statistics for each homologue are reported in Supplementary Table S4). Notably, even chain lengths displayed generally less depleted values (mean ≈ −140‰) compared to odd chain lengths (mean ≈ −171‰; Figure 3). For both odd and even homologues, the most depleted values were observed in the C_27–31 range (odd ≈ −179‰, even ≈ −147‰), while progressively less depleted values characterised shorter (odd ≈ −171‰, even ≈ −154‰) and longer (odd ≈ −176‰, even ≈ −129‰) homologues.

The δ²H weighted mean (δ²H_CWMA) strongly reflected the values of C₂₇-C₂₉ and C₃₁, the three most abundant homologues in all leaf samples and most lower stem samples (Supplementary Table S4). In leaf samples, δ²H_CWMA ranged from −273‰ to −115‰ (average −198‰ ± 28‰), while in stem samples, it ranged from −253‰ to −152‰ (average −196‰ ± 27‰). Among leaf samples, species with the least depleted δ²H values were G. uliginosum among forbs, C. longus among graminoids, E. fluviatile, and N. alba, while M. aquatica and C. mariscus exhibited the most depleted values. Lower stem samples displayed higher interspecific variability.

Summer δ²H values of concentration-weighted mean n-alkanes from leaves (CWMA δ²H_lwax) were generally less depleted than spring values (Supplementary Table S3; Figure 4), with exception noted for S. lacustris and E. fluviatile (likely influenced by sampling bias; see Section 4.2.3), as well as S. palustris and C. riparia. In Nisí, focusing on C₃ species, graminoids exhibited the most depleted δ²H_lwax (−214‰, −215‰), while generally less depleted values (Supplementary Table S3) characterised forbs (−205‰, −183‰) and E. fluviatile (−171‰).

Surface and soil water δ²H values (Figure 5; Supplementary Table S2) varied between −72‰ and +37‰. Across both sites and seasons, the δ²H values for surface and soil water were well clustered around their means, with a maximum standard deviation of 1.8‰. Notably, surface water maintained a consistent mean δ²H value (ca −54‰) in both seasons at both sites. In contrast, soil water δ²H values consistently trended lower in Nisí compared to TP, demonstrating an increase from spring (−49.1‰ TP; −60.2‰ Nisí) to summer (−46.5‰ TP; −49.9‰ Nisí; Supplementary Table S2). Both surface water and soil water showed more ²H-depleted values compared to water from leaves and lower stems.

Leaf water δ²H values for ranged from 37‰ to −41‰, with two outliers at −71‰ (mean ≈ −8‰, median ≈ −8‰, σ ≈ 14‰). In contrast, lower stem water δ²H values varied between −18‰ and −65‰ (mean ≈ −49‰, median ≈ −49‰, σ ≈ 10‰).

Seasonal, species-specific mean δ²H values for lower stem water ranged from −62‰ to −24‰, while leaf water values ranged from −25‰ to +29‰. Except for those of M. aquatica, S. palustris, and S. lacustris, spring values typically exhibit a greater ²H-depletion compared to summer values (approximately 5‰–15‰). However, the variability closely aligned with the standard deviation for each species, as outlined in Table 3B and Figure 5. Similarly, for both lower stem and leaf water, forbs displayed higher δ²H values than graminoids (ca 5‰–10‰ on average) and TP showed higher values than Nisí (Table 3B; Supplementary Table S3). Stem to leaf water ²H-fractionation (ε_stw-lw) ranged between +13‰ and +59‰ (Table 3B).

Leaf samples show consistently higher n-alkane concentrations than stem samples (up to 80 times and ca. 15 times on average; 3.1.1, Table 2). This can be attributed to the necessity for a thicker cuticle to mitigate desiccation in the upper organs, more exposed to sunlight (i.e., leaves and flowers; Gamarra and Kahmen, 2015; Speckert et al., 2023). More in detail, while forbs show on average higher leaf n-alkane concentrations than graminoids, they also show greater difference between leaves and stems (40 times on average) than graminoids (7 times on average). This is common (e.g., He et al., 2020; Corcoran et al., 2022) and likely due to morphological differences between these two groups.

The CPI (Table 2) shows no consistent differences between species nor between leaves and stems, with several species (e.g., P. australis, C. palustris) exhibiting contrasting leaf-to-stem CPI ratio values in different sampling instances. The CPI values decrease from leaf/stem samples to soil, likely due to preferential degradation of longer homologues (Yan et al., 2021) and the production of microbial derived shorter/even homologues (Thomas et al., 2021; Corcoran et al., 2022 and refs therein), making it impossible to use CPI as a proxy to trace specific plant sources in sediments (Chen et al., 2022).

In terms of ACL, leaves show similar, although generally higher values then the corresponding stem samples (Table 2). Relative to leaves, stem samples often display a broader distribution of odd chain lengths centred around the same C_max (Supplementary Figure S2), along with analogous but generally lower ACL values, with a few exceptions (e.g., C. riparia in Nisí in spring; Table 2). These observations align with previously reported differences between leaves and roots in C₄ grasses such as Enneapogon avenaceus and Astrebla pectinata (Kuhn et al., 2010), but are in contrast with the findings of He et al. (2020) for similar Poaceae and Cyperaceae emergent species, reporting much lower ACL values in roots relative to leaves. The similarity in ACL and the higher concentrations of n-alkanes in leaves compared to stems indicate that, as observed in alpine and temperate grassland species (Gamarra and Kahmen, 2015), also in these fen environments the distinctive signal deposited in sediments is predominantly characterized by the chain length distribution of the leaves (and not the stems) of helophytic plants.

In terms of plant forms, ACL values in our data generally differ between forbs and graminoids. Graminoids leaves exhibit lower ACL (27–29) than forbs leaves (29–32), with the exception of the C₄ graminoid C. longus (29–31), as highlighted by the PCA in Figure 6; stem samples show an analogous pattern. Very slight, although opposite, differences between forbs and graminoids ACL have been detected before on plants from a similar fen environment on the Chinese Loess Plateau (Liu et al., 2018). Overall, graminoids show higher values on the Chinese Loess Plateau (∼30.0) than in Nisí and TP (∼28.6), while forbs have a similar/lower ACL range than graminoids (Supplementary Table S3). On one hand, this opposite behaviour of ACL could be a response to the differences between the climates of the Chinese Loess Plateau (Köppen BS) and of the Mediterranean borderlands. In fact, even if both are characterised by seasonal aridity, the timing is different between these two regions: on the Chinese Loess Plateau winter (and not summer) is the most arid season (Kong et al., 2018). Alternatively, the different behaviour of ACL in our study could simply be the result of different species associations (Diefendorf et al., 2021).

Also, the other two plant forms present in our data show ACL values lower than forbs/eudicots. The floating hydrophyte N. alba shows an ACL (∼29) more characteristic of emerged helophytes (probably indicating the subaerial condition of its leaf blades; Yu et al., 2021). Interestingly, the only pteridophyta (E. fluviatile) shows an ACL (25–26, the lowest in our results) closer to submerged hydrophytes than to terrestrial plants (Figure 6; Yu et al., 2021).

In our soil samples, ACL is relatively stable (27.3–28.7; Table 2) and mean values tend to be similar to mean ACL values of monocots/graminoids (leaves and stems) in both locations and seasons, likely reflecting a major contribution of this plant group to sediment/peat deposition (Liu et al., 2018), as also highlighted in the PCA (Figure 6).

We are aware that caution must be applied to the interpretation of ACL environmental significance, as (1) it has already been shown that relative humidity gradients can have opposite effects on the ACL of different species (Hoffmann et al., 2013; Chen et al., 2022) and (2) plant contributions to sediments in any environment is subject to multiple taphonomic variables (e.g., Corcoran et al., 2022 and refs therein). However, our results indicate ACL as a potential tool to differentiate between plant form (forbs vs. graminoids) relative contribution in this particular environment. Based on our data, the n-alkane signal stored in the sediments of this fen environment with this type of helophytic plant community seems to be mostly representative of local graminoids, and particularly of their leaves.

Our samples show no consistent change in n-alkane concentration between seasons, at both sites (Table 2). While S. palustris, C. longus, and S. lacustris leaves show clear increases from spring to summer, the opposite is true for M. aquatica, C. mariscus, and E. fluviatile, while values for all the other species remain within a standard deviation and thus substantially unvaried. Concentrations are expected to positively correlate to the summer temperature increase, to enhance the cuticular transpiration barrier, as shown, for example, in Juniperus monosperma leaves, (Diefendorf et al., 2021; Shi et al., 2021 and refs therein). The reason behind this inconsistency is unclear, but it likely results from the effect of other factors controlling n-alkane production and preservation, such as, for example, leaf abrasion, aridity, and variability in the leaf sampling procedure (Eglinton and Eglinton, 2008).

Similarly, ACL values do not show a clear seasonal pattern. Forbs such as C. palustris, M. aquatica, S. palustris, as well as S. lacustris and E. fluviatile show increasing ACL values from spring to summer, but all other species show no clear seasonal change. While seasonal ACL increases have been detected before (e.g., from April to September in Salix viminalis leaves, Newberry et al., 2015, in other eudicot/monocot species, Cui et al., 2008, and even in a 15 years study of a subalpine meadow, Shi et al., 2021), Diefendorf et al. (2021) found ACL to be unrelated to temperature increases and to be instead relatively constant within a species. Based on our data, neither n-alkane concentration nor ACL do appear to provide valuable information on seasonality in these type of fen vegetational communities.

In contrast, CPI shows a consistent seasonal decrease from spring to summer. The decrease in CPI seems generally related to an increase of even homologues, while odd homologues concentrations appear relatively stable (Supplementary Figure S2). An analogous CPI seasonal pattern has been detected in maple leaves (Chikaraishi and Naraoka, 2006) and other eudicot/monocot species (Cui et al., 2008) and linked to leaf senescence. However, as none of our leaves/stems showed signs of fading when collected in July, this is an unlikely explanation for the seasonal CPI shift in our samples. Instead, the CPI increase could be related to the enhanced summer aridity. Drought conditions have been shown to speed up lipid metabolism while affecting enzymatic chain elongation, leading to an increased production of mid-short n-alkanes homologues and their relative degradation by-products, including even homologues (Post-Beittenmiller, 1996; Shepherd and Griffiths, 2006; Srivastava and Wiesenberg, 2018; Speckert et al., 2023).

In general, our data indicate that CPI values in plant samples differ between the growing (higher values) and the dry (lower values) seasons. However, due to preferential degradation and microbial activity (Thomas et al., 2021; Corcoran et al., 2022 and refs therein; see Section 4.1.1), soil/sediment CPI is likely an unreliable proxy for tracing seasonality in sediments, at least in this kind of Mediterranean fen environments.

Surface water and soil water at a depth of 20–30 cm exhibited very similar hydrogen isotopic values (Figure 5), pointing to a clear link between the two water pools. δ²H values for surface and soil water show very similar ranges at both sites in each season (except for 3 unusually high values in TP spring soil water). This reinforces the hypothesis that the Nisí site can be considered as an analogue for the ancient TP site in terms of δ²H of source water based on the opinion of researchers that have described the environmental, climatic, and geological characteristics of both sites (e.g., Christanis, 1994; Kalaitzidis, 2007; see Table 1) as well as on the data of spring/source water δ²H reported here and throughout Greece (Dotsika et al., 2010).

The sole contemporary dataset for annual precipitation δ²H (δ²H_p) in northern Greece is derived from the Thessaloniki GNIP station, covering a period of 2.5 years from 2001 to 2003). The station is situated approximately 105 km W-SW of the Tenaghi Philippon peatland and 90 km E-SE of Nisí fen (IAEA/WMO, 2017; Figure 1).

The δ²H_p values range from −99‰ to −3‰ (with an annual precipitation weighted average of −48‰) and show a distinct seasonal pattern. The most negative values characterise the abundant precipitation during the wet winter-spring period reflecting a weighted mean of approximately −52‰. Conversely, the least negative are observed during the dry Mediterranean summer, hovering around −31‰. As a result, surface and soil water δ²H at the end of spring serve as a reliable approximation of the δ²H_p annual weighted mean (ca. −48‰).

Our data affirms this pattern, as soil (and surface) water consistently exhibit δ²H values similar to this annual mean not only in spring (−49.1‰ in TP; −60.2‰ in Nisí), but also in summer (−46.5‰ in TP; −49.9‰ in Nisí). This alignment is likely attributed to the elevated amounts of winter-spring precipitation and, most probably, to the smearing effect of the karstic aquifer feeding the mire(s) (Kalaitzidis, 2007; Dotsika et al., 2010). Remarkably, the OIPC provides comparable annual δ²H values for precipitation (−51‰ ± 2‰ in Nisí; −48‰ ± 1‰ in TP); similarly, local sampling in NW (−60.4‰ ± 0.8‰) and NE (−50.6‰ ± 0.7‰) Greece, where Nisí fen and TP are located, respectively, supports similar values for spring water (Dotsika et al., 2010). This suggests that these helophytic plant communities in these Mediterranean regions do indeed utilise water that faithfully represents local annual δ²H_p values.

Most species showed less depleted stem water δ²H (δ²H_stw) values in the summer, and, within the same species, δ²H_stw values were less depleted in TP than in Nisí (Table 3B). The ²H-fractionation between the soil water and the lower stem water (ε_soilw-stw) of each species (Table 4) was 1) consistently positive (except for P. australis), 2) often within the respective standard deviations, and 3) lacking a clear seasonal pattern. Additionally, as ²H-fractionation during water absorption in roots is considered unlikely (Ehleringer and Dawson, 1992; White et al., 1994; Jacob et al., 2021), the tendency to less depleted δ²H_stw values appears to reflect an occasional initial minor evapotranspiration effect occurring at the crown root and/or at the base of the stem (Eensalu et al., 2023). This difference is more pronounced for forbs (δ²H_stw up to approximately 22‰ less depleted than δ²H_soilw) than for graminids (generally in the 0‰–5‰ range) and mirrors the δ¹⁸O differences between the soil and stem water of two grasses (Dactylis glomerata, Lolium perenne) and a forb (Trifolium pratense; Barnard et al., 2006). We speculate that the overall low/absent enrichment of stem water relative to soil water in monocots/graminoids might be attributed to a protective outer sheath present at the base of most grasses (which was removed during sampling; Barnard et al., 2006). The occasional negative difference between the δ²H_stw and the δ²H_soilw values for P. australis (down to ca. −10‰ in summer), could be explained by its greater rooting depth (Burdick et al., 2001; Moore et al., 2012; Huang et al., 2018; Zhao et al., 2018; Huang and Meyers, 2019), a phenomenon already documented for some species growing in arid environments (Feakins and Sessions, 2010; Corcoran et al., 2022).

Overall, the data indicates an absence of evaporative enrichment between soil and stem water or at the very least, a very limited occurrence of it. Despite specific variations in rooting depth, leading to a decoupling between δ²H_stw and δ²H_soilw, when considering the plant community as a whole, the average δ²H_stw is representative of the δ²H_soilw for both seasons and sites (Figure 5). Moreover, δ²H_soilw has been previously linked to the local mean annual δ²H_p (see Section 4.2.1). This demonstrates that the initial source of biosynthetic hydrogen within the plants (δ²H_stw) at both sites isotopically reflects mean annual precipitation, with an at least annual integration or smoothening of the seasonal differences of the δ²H_p signal, a conclusion comparable to findings reported in similar studies (e.g., Eensalu et al., 2023). Consequently, any intra- and inter-specific δ²H_wax variability will primarily result from differences in physiology (e.g., growth form) and/or environmental conditions (e.g., aridity) during the growing season, affecting the internal processes of the plants (e.g., Griepentrog et al., 2019).

Generally, δ²H_lw values exhibited greater enrichment in summer compared to spring (Table 3). In most species, the difference between spring and summer δ²H_lw closely mirrored the seasonal variation in δ²H_stw (ca. 5‰–10‰). M. aquatica also conformed to this pattern when a single clear outlier (−72‰) was excluded from the summer mean. The seemingly distinct behaviour of S. lacustris (ca 10‰ more depleted in summer) can be attributed to a sampling bias: its morphology posed challenges in clearly separating stem and leaf sections in the summer specimens. Similarly, the significant seasonal difference in E. fluviatile δ²H_lw (ca 36‰ more negative in summer) is likely linked to the different physiology of this genus, which produces green photosynthesising stems and leaves in spring while new, low photosynthesising stems, crowned by a spore-bearing cone, emerge in summer (the latter were, in fact, sampled as “whole”).

The ²H-fractionation between lower stem and leaf water (ε_stw-lw; +27‰ to +60‰; Table 4) tends to be greater in species with greater height, such as P. australis, T. angustifolia, and C. mariscus (notably, Cladium jamaicense in the Florida Everglades also showed very depleted δ²H_lwax values, −231‰; He et al., 2020), and relatively broader, more exposed leaves (as in C. palustris), potentially indicating increased evapo-transpiration.

It is important to note that our dataset offers a discrete rather than continuous representation of internal water δ²H, a parameter that has been found to exhibit considerable variability, with seasonal to hourly fluctuations (Sachse et al., 2009, 2010). Our sampling approach captures only selected “snapshots” of stem/leaf water δ²H distributed over a few days in spring and summer. Nevertheless, our ε_stw-lw values demonstrate consistency between spring (+44.7‰ ± 12‰) and summer (+45.2‰ ± 10‰) species-specific averages. This stability aligns with findings in studies involving a limited number of tundra vascular plant species in western Greenland (σ=12‰; Berke et al., 2019) and suggests a levelling out of the considerable variability in stem/leaf evapo-transpiration over the extended time frame of the entire vegetative season.

The ²H-fractionation from leaf water to leaf n-alkanes (ε_lw-lwax) is known to present high specific variability. While interspecific variability is predominantly associated with differences in plant physiology, intraspecific variability (up to ca. 70‰) is attributed to diurnal/seasonal variability of leaf water δ²H values (Sachse et al., 2012). Given these considerations, the study of ε_lw-lwax (as well as ε_stw-stwax) tipically requires high frequency discrete sampling, ideally approximating continuous sampling, of leaf/stem water and wax n-alkanes throughout the entire vegetative period (e.g., Kahmen et al., 2013; Tipple et al., 2013; Newberry et al., 2015; Sachse et al., 2015).

As previously mentioned (see Section 4.2.3), our study does not rely on such continuous sampling, but rather adopts a more sporadic approach. Nevertheless, our sampling occurred during the main arc of the day, over an average of 2-3 consecutive days per season and site (see Section 2.2). To mitigate the uncertainty inherent in the sporadic sampling of individual specimens, we aggregated the leaf and lower stem water δ²H values of all samples (395), without regard to species, computing season and site averages for δ²H_lw and δ²H_stw (Table 3). We regard these averages as robust estimates of community-wide δ²H_lw and δ²H_stw during spring and summer (Feakins and Sessions, 2010; Sachse et al., 2010). Utilising these consolidated values (Table 3B) and species-specific averages of δ²H_lwax and δ²H_stwax (CWMA), we derived species-specific ²H-fractionation values from lower stem water to leaf water (ε_stw-lwax), leaf water to leaf wax (ε_lw-lwax), stem water to stem wax (ε_stw-stwax), and leaf water to stem wax (ε_lw-stwax), during growth and maturity seasons (Table 4).

At Nisí, especially in summer (Supplementary Table S3), C₃ graminoids showed significantly greater ε_lw-lwax (average of species-specific mean values: spring −202‰ ± 11‰, summer −213‰ ± 10‰) than forbs (spr. −195‰ ± 8‰, sum. −176‰ ± 11‰; Table 4). The resulting ε_lw-lwax difference between C₃ graminoids and forbs in summer (i.e., at the end of the growing season; 36‰ ± 15‰) closely mirrors the reported difference in net fractionation between the same groups (37‰ ± 42‰; Sachse et al., 2012). The larger associated error with net fractionation (±42‰) suggests that, while biosynthetic fractionation may tend to a stable mean value within plant groups, much of uncertainty in net fractionation likely stems from variations in environmental conditions affecting the water pool (δ²H_p/δ²H_soilw/δ²H_stw/δ²H_lw).

The δ²H values of n-alkanes in stem samples did not exhibit a discernible seasonal pattern and our δ²H_stwax values were relatively similar to the δ²H_lwax values of their leaf counterpart (although often slightly less depleted; Figure 4). ε_stw-stwax values tended to be lower than ε_lw-lwax and displayed significant variability (Figure 7; Table 4). However, when δ²H_lw is used instead of δ²H_stw, the resulting ε_lw-stwax values are much closer to ε_lw-lwax than ε_stw-stwax (Figure 7).

Our proposed explanation for the higher similarity between ε_lw-stwax and ε_lw-lwax (than between ε_stw-stwax and ε_lw-lwax) involves a shared substrate for the n-alkanes synthesis between lower stems and leaves. Due the low photosynthetic activity at the base of the stem (Brazel and Ó’Maoileídigh, 2019 and refs. therein), n-alkanes produced there must derive their hydrogen (at least partially) from a substrate previously synthetised in leaves (Sessions, 2006; Shepherd and Griffiths, 2006). The δ²H of this substrate would be marked by δ²H_lw, used locally and then transported down to other parts of the plant, influencing the δ²H value of locally produced n-alkanes, especially in organs less exposed to light (as, in our case, the base of the stem). This would explain both the more stable relationship of δ²H_stwax to δ²H_lw (ε_lw-stwax) compared to δ²H_stw (ε_stw-stwax), as well as δ²H_stwax slightly less depleted values compared to δ²H_lwax, as the substrate delivered to the stem would be ²H-enriched after part of ¹H was preferentially removed by biosynthetic ²H-fractionation during the synthesis of leaf wax compounds.

This proposed mechanism closely parallels to the one employed to explain less depleted δ²H_lwax at the base of some monocot and eudicot leaves (Gao et al., 2015). It potentially aligns with other findings indicating that the δ²H values of n-alkanes in leaves are more positive than in stems and more negative than in roots (Gamarra and Kahmen, 2015; He et al., 2020). The δ²H_wax of those stems, exposed to light, would be influenced by the more negative δ²H_stw, unlike our lower stem samples (confined to the section close to the roots), relying more on leaf-synthesised substrate. However, another study (Liu J. et al., 2019) found δ²H_wax to be more negative in roots than leaves in two C₃ species (Artemisia vestita–monocot, and Stipa bungeana–eudicot), and vice versa in Bothriochloa ischaemum, a C₄ monocot species. This underscores the necessity of gathering additional data on plant organs other than leaves to elucidate the variability in δ²H_wax between different plant sections.