We collected 105 terrestrial plant samples for plant wax analysis across three lake catchments spanning 14 degrees of latitude in the ECA during summer field seasons in 2019 and 2021 (Fig. 1b). Lake Africa (AFR; Informal name; 72.42° N, 77.44° W; 895 m a.s.l.) is located in northern Baffin Island and represents our northernmost sampling location in this study. Due to its high elevation, AFR resides in bioclimate subzone A and its catchment is dominated by mats of liverworts, mosses, and lichen and contains only one vascular plant species, a grass, Phippsia algida. Lake CF8 (Informal name; 70.56° N, 68.95° W; 195 m a.s.l.) is located in the Clyde Foreland of northeastern Baffin Island and is characterized as bioclimate subzone C. While lichens and cryptogam mats are still the most common vegetation types in the catchment, prostrate dwarf shrubs (e.g., Cassiope tetragona) and graminoids (e.g., Luzula confusa) are also present . Lake 3LN (Informal name; 58.10° N, 68.46° W; 61 m a.s.l.) is on the Canadian mainland in the Nunavik region of northern Quebec. In contrast to the previous two lakes, Lake 3LN is south of the tree line and is classified as a sub-Arctic bioclimate. The catchment contains a wide variety of vascular plant vegetation, including trees (Larix laricinia, Picea mariana), shrubs (e.g., Betula glandulosa, Rhododendron sp., Salix sp., Alnus viridis), and graminoids (e.g., Carex sp., Eriophorum sp.). Bryophytes are also present in wetter areas of the catchment and are dominated by Pleurozium schreberi and Sphagnum sp. We also include previously published terrestrial plant wax data from Lake Qaupat (QPT; 63.68° N, 68.20° W; 33 m a.s.l.) in southern Baffin Island for this ECA transect . Positioned in bioclimate subzone D, the Lake QPT catchment is dominated by Betula glandulosa and Salix sp. shrubs, similar to Lake 3LN, but does not contain any trees. The ECA transects included 139 plant samples across seven major plant growth forms; trees (n = 6), shrubs (n = 43), forbs (n = 1), graminoids (n = 17), mosses (n = 36), liverworts (n = 18), and lichens (n = 18).

To expand the sample size and range of environmental conditions, we compiled published plant wax data from terrestrial plants from sampling sites across the entire Arctic; within the latitude range spanned by the ECA transect (Fig. 1a). Regions added in this compilation include: west Greenland , northern Norway , northern Russia , and Alaska/Yukon/Northwest Territories . This compilation increased the number of unique sampling/environmental locations to 36, increased the total plant wax sample count to 386, and added ferns to the list of plant growth forms in this study; trees (n = 18), shrubs (n = 183), forbs (n = 17), ferns (n = 15), graminoids (n = 52), mosses (n = 63), liverworts (n = 18), lichens (n = 20). Different studies reported different types of plant wax data, and most studies only reported one plant wax compound class and/or one stable isotope, which limits the sample size for some analyses of individual plant wax data types (Table S1). Compiled stable isotope data had to be accompanied by the concentration or relative abundance of each chain-length in order to calculate amount-weighted average values for each sample (Sect. 2.3).

We used ERA5 reanalysis as our data source for climate parameters at each sampling location and sampling year, including monthly temperature (Fig. 2a), precipitation amount (Fig. 2b), and relative humidity (Fig. 2c). Plant waxes are produced yearly by individual plants, primarily during peak leaf flush in the early part of the growing season , and therefore closely reflect the environmental conditions of the year they were sampled. The Arctic growing season, during which plant waxes are produced, is limited by the duration of temperatures above freezing . Therefore, we express the growing season environmental conditions of each plant sample as the mean of the months above freezing (MAF) for individual sampling years. Because ERA5 data can be extracted for the grid cell containing each study site, they are preferable to station data, which in the ECA study area are few in number and all coastal, leading to discrepancies due to differences in distance, elevation, and other features . Since relative humidity (RH) is not directly provided by ERA5, we calculated it for each month using ERA5 temperature (T) and dew point temperature (Dp) in Eq. (1) . We also compared plant wax indices to sample site latitude and elevation (Bakkelund et al., 2018; O'Connor et al., 2020), which was reported in all compiled publications (Fig. 2d).

We calculated MAF average amount-weighted precipitation isotope δ2H (Fig. 2e) using monthly average δ2H from the Online Isotopes in Precipitation Calculator OIPC; and ERA5 precipitation amount. We justify our use of MAF precipitation δ2H based on several studies which show that shallow Arctic soil water, from which most terrestrial plants obtain their water, generally reflects growing season precipitation without the influence of ²H-depleted snowmelt . This is attributed to the soil still being frozen and impermeable during the snowmelt period, which causes the meltwater to be lost from the system as surface runoff . Additionally, variations in Arctic precipitation stable isotope ratios are primarily driven by seasonal changes in temperature and moisture sources, rather than amount-driven fractionation during individual precipitation events . 1RH=100×e17.625×Dp/243.04+Dp/e17.625×T/(243.04+T)

We extracted and quantified plant wax n-alkanoic acids and n-alkanes from 105 ECA plant samples using the standard methods of the University at Buffalo Organic and Stable Isotope Biogeochemistry Laboratory . We measured n-alkanoic acid (C₂₀ to C₃₂) and n-alkane (C₂₁ to C₃₃) chain-length concentrations (µg g⁻¹ dried plant) using a Trace 1310 Gas Chromatograph-Flame Ionized Detector (GC-FID), with dual AI 1310 autosamplers and 30 m HP-1MS fused silica columns. We derived chain-length concentrations from GC-FID peak areas using external calibration curves of C₂₈ n-alkanoic acids and C₂₉ n-alkanes. We also measured for sample recovery during the full extraction and instrumental analysis processes using internal monounsaturated cis-eicosenoic n-alkanoic acid (C_20,1) and C₃₆ n-alkane standards. We examined the overall distributions of plant wax chain-lengths by calculating the Average Chain-Length (ACL; Eqs. 2, 3) for each sample . We calculated the Carbon Preference Index (CPI; Eqs. 4, 5) to describe the difference in production of even vs. odd and odd vs. even chain-lengths in n-alkanoic acids and n-alkanes, respectively . CPI was not calculated for samples where the total abundance of chain-lengths in the denominator was equal to zero, either due to compounds being too small to quantify or not being reported in other publications (i.e., no odd-chain n-alkanoic acid data available). 2ACLn-acid=ΣCneven×neven/ΣCneven3ACLn-alkane=ΣCnodd×nodd/ΣCnodd4CPIn-acid=ΣC20-30even+ΣC22-32even/2×ΣC21-31odd5CPIn-alkane=ΣC21-31odd+ΣC23-33odd/2×ΣC22-32even

We measured the compound-specific stable carbon (δ13C) and hydrogen (δ2H) isotope ratios of C₂₂ through C₂₈ even-chain n-alkanoic acid chain-lengths in a subset of the ECA plant samples using the methods described in . Stable isotope analysis was conducted in the University at Buffalo Organic and Stable Isotope Biogeochemistry Laboratory using a Thermo Delta V+ Isotope Ratio Mass Spectrometer (IRMS) with a split/splitless injector and a TriPlus RSH autosampler, connected to the IRMS via IsoLink II and Conflo IV with all samples and standards run in triplicate. Within each IRMS sequence, we ran standards of C₂₀ and C₂₈ n-alkanoic acids to calibrate sample δ13C results to the Vienna Pee Dee Belemnite (VPDB) scale and calibrate δ2H results to the Vienna Standard Mean Ocean Water (VSMOW) scale. These C₂₀ and C₂₈ standards were also used to correct for chromatogram peak size linearity. Additionaly, we used standards of C₁₈ and C₂₄ n-alkanoic acids to correct for instrument drift in each sequence. For δ2H analysis, we measured the H3+ factor at the beginning of each IRMS sequence. Sequences run in the Fall of 2022 had H3+ factors ranging from 4.873 ± 0.025 to 5.005 ± 0.038 (mean ± 1σ) and H3+ factors in sequences run during the Summer of 2023 ranged from 3.400 ± 0.055 to 3.509 ± 0.016. We reported δ13C and δ2H uncertainty as the Standard Error of the Mean (SEM), accounting for uncertainty from instrument analysis and triplicate measurements of each sample. SEM uncertainty ranged from 0.1 ‰ to 0.6 ‰ with an average of 0.2 ± 0.1 ‰ for δ13C and 3.1 ‰ to 9.4 ‰ with an average of 4.2 ± 1.3 ‰ for δ2H. We calculated the net apparent δ2H fractionation (ϵapp) between amount-weighted precipitation δ2H values for the months above freezing and plant wax chain-length δ2H (Eq. 6). In Sect. 3, we report plant wax δ13C, δ2H, and ϵapp as the chain-length abundance-weighted averages of C₂₂ through C₂₈ for n-alkanoic acids and C₂₃ through C₂₉ for n-alkanes. Doing so allows us to robustly compare stable isotope values between samples by accounting for the variability in isotope values and concentrations of individual chain-lengths. Average values, hereafter, are expressed as the mean ± 1σ standard deviation. 6ϵapp=[(1000+δ2Hplantwax/1000+δ2HMAFprecipitaion)-1]×1000

We used Shapiro-Wilk tests to determine whether the values of plant wax indices within individual plant growth forms were normally distributed (Shapiro-Wilk test p-value ≥ 0.05). We then employed Mann-Whitney U tests to evaluate whether chain-length or stable isotope values of different plant groups were significantly different (Mann-Whitney U test p-value < 0.05). We also used Pearson correlations to evaluate linear relationships between plant wax chain-length and stable isotope indices vs. scalar environmental parameters (latitude, temperature, relative humidity, precipitation amount, elevation). To examine the potential influence of individual plant growth forms in this analysis, we perform multiple Pearson correlations for the same plant wax data type and set of environmental parameters while removing one growth form each time. We compared the resulting Pearson r-values to determine which growth form had the greatest impact on the correlation strength of the entire dataset when removed. For all statistical analyses, we used the averages of replicate species from the same location and sampling year (i.e., three Alnus viridis from Lake 3LN in 2021) to avoid overrepresentation of those samples Table S2;. We required each correlation to have a minimum of three observations per pair of variables. Shapiro-Wilk tests, Mann-Whitney U tests, and Pearson correlations were performed using the associated functions in the SciPy Python package v1.15.2 . These three statistical analyses were only performed on the pan-Arctic dataset due to the larger number of measurements per plant growth form and number of sampling sites for environmental comparisons.

We used Principal Component Analysis (PCA) to determine the primary modes of variability across all plant wax chain-length distributions (i.e., even-chain C₂₀ to C₃₂ n-alkanoic acids and odd-chain C₂₁ to C₃₃ n-alkanes). Plant wax chain-length relative abundance data naturally has a correlation bias because the relative abundances of all chain-lengths are constrained to a constant sum . To address this, we apply a centered log-ratio (clr) transformation on the plant wax chain-length data prior to PCA. Chain-length concentrations of 0 µg g⁻¹ dry plant, either because chromatogram peaks were below the limit of detection or were not reported in other publications, were not compatible when calculating the sample’s geometric mean for clr . We replaced concentration values of zero with a small number equal to 1/N2 where N is the number of plant wax chain-lengths used for PCA . PCA was conducted using the decomposition.PCA() function from the Scikit-learn project . We performed PCA on plant wax data from all ECA lakes (AFR, CF8, QPT, 3LN) and the pan-Arctic dataset.

We quantified plant wax n-alkanoic acid chain-length concentrations in 103 plant samples and n-alkane chain-length concentrations in 101 samples from Lake AFR, CF8, and 3LN in the ECA. Two n-alkanoic acid samples and four n-alkane samples had concentrations that were below GC-FID detection. ECA plant wax chain-length concentrations, per gram of dried plant, were highly variable. Total n-alkanoic acid concentrations ranged from 1.3 to 13 839 µg g⁻¹ with an average of 745 ± 1618 µg g⁻¹ (Fig. 3a), and n-alkane concentrations ranged from 0.5 to 6553 µg g⁻¹ with an average of 625 ± 1100 µg g⁻¹ (Fig. 3b). Shrubs and mosses produced the greatest total concentrations of the two plant wax compounds among vascular and non-vascular growth forms, respectively. Liverworts produced the lowest total plant wax concentrations out of all growth forms. These patterns held true for the pan-Arctic data compilation, though forbs (only n = 1 for the ECA dataset) produced very high concentrations of n-alkanes. Average pan-Arctic total n-alkanoic acid and n-alkane concentrations were 549 ± 1139 and 854 ± 2072 µg g⁻¹, respectively. Some plant growth forms had substantial differences between the total concentrations of each plant wax compound. For example, ECA trees produced an average of 1297 ± 934 µg g⁻¹ n-alkanoic acids but only 47.9 ± 18.5 µg g⁻¹ n-alkanes, while lichens produced 46.6 ± 54.4 µg g⁻¹ n-alkanoic acids but 546 ± 888 µg g⁻¹ n-alkanes. Shapiro-Wilk tests showed that all but one plant growth form in each compound class, shrub n-alkanoic acids and fern n-alkanes, were log-normally distributed (Table S3). Mann-Whitney U tests showed that total plant wax n-alkanoic acid and n-alkane concentrations of shrubs and lichens were significantly different from most other plant growth forms (Fig. 4a, b).

We calculated ACL and CPI using the full range of plant wax chain-lengths measured; C₂₀–C₃₂ for n-alkanoic acids, C₂₁–C₃₃ for n-alkanes. ACL in ECA terrestrial plant wax n-alkanoic acid samples ranged from 21.5 to 29.3 with an average of 24.2 ± 1.5 (Fig. 3c), and n-alkane ACL ranged from 21.4 to 31.0 with an average of 26.9 ± 1.9 (Fig. 3d). Trees and shrubs produced the greatest proportion of longer chain-lengths, overall, for both compound classes while liverworts and lichens produced the most mid-chain waxes. However, some trees and shrubs, such as Picea mariana, Cassiope tetragona, and Betula glandulosa, did produce high relative abundances of mid-chain waxes. Average pan-Arctic n-alkanoic acid and n-alkane ACL were 24.9 ± 1.7 and 27.2 ± 1.9, respectively. Shrubs were the only plant growth form with a non-normal distribution of n-alkanoic acid ACL values, whereas only half of the plant growth forms had normally-distributed n-alkane ACL values (Fig. 3c, d). Liverwort and lichen ACL values were the most significantly different among plant growth forms (Fig. 4c, d). Distributions of n-alkanoic acid ACL in all non-vascular growth forms were significantly different from trees, shrubs, and forbs (Fig. 4c). CPI in ECA terrestrial plants was highly variable, ranging from 1.5 to 135 with an average of 12.7 ± 20.4 for n-alkanoic acids (Fig. S1a) and from 0.8 to 44.8 with an average of 5.0 ± 6.6 for n-alkanes (Fig. S1b). Trees and lichens had both the highest mean CPI values and variability in both the ECA and pan-Arctic datasets, while liverworts had the lowest CPI values overall. Within the pan-Arctic dataset, all n-alkanoic acid samples had CPI values ≥ 1 and 98 % were ≥ 2, while 93 % of n-alkane CPI values were ≥ 1 and 74 % were ≥ 2.

We measured even-chain n-alkanoic acid (C₂₂ through C₂₈) δ13C in 25 ECA plants. Hereafter, we refer to δ13C results as the concentration-weighted average of those chain-lengths (C₂₃ through C₂₉ for n-alkanes in other datasets). ECA n-alkanoic acid δ13C ranged from -38.1 ‰ to -29.3 ‰ with an average of -33.5 ± 2.5 ‰ (Fig. 3e). Pan-Arctic n-alkanoic acid δ13C ranged from -40.1 ‰ to -27.1 ‰ with an average of -33.0 ± 2.6 ‰ and n-alkane δ13C ranged from -35.4 ‰ to -26.7 ‰ with an average of -32.2 ± 2.3 ‰. The range in ECA n-alkanoic acid δ13C between chain-lengths within an individual sample was an average of 3.2 ± 2.3 ‰, which was greater than the average ranges in pan-Arctic n-alkanoic acids (2.6 ± 1.9 ‰) and n-alkanes (1.3 ± 0.6 ‰). In both the ECA and pan-Arctic n-alkanoic acid datasets, trees were the most ¹³C-enriched and mosses were the most ¹³C-depleted. The pan-Arctic data compilation included n-alkane δ13C only for shrubs, forbs, ferns, and graminoids. Among these four growth forms, shrubs were the most ¹³C-enriched and ferns were the most ¹³C-depleted (Fig. 3f). Shapiro-Wilk tests were limited by many growth forms not having enough measurements (n<3) in each compound class. All growth forms that did have enough measurements were normally distributed (Table S1). Shrub and moss n-alkanoic acid δ13C were the most significantly different from other plant growth forms (Fig. 4e). No growth forms had significantly different distributions of n-alkane δ13C (Fig. 4f).

We measured even-chain n-alkanoic acid δ2H, across the same chain-length range as δ13C, in 24 ECA plants. One sample that was first analyzed for δ13C did not have sufficient plant wax material remaining to also measure δ2H. Again, we refer to the δ2H and ϵapp results as the chain-length concentration-weighted average value in each plant sample. The individual sample range in ECA n-alkanoic acid δ2H between chain-lengths was an average of 31.5 ± 18.3 ‰, which was greater than the average range in pan-Arctic n-alkanoic acids (25.3 ± 14.3 ‰) and n-alkanes (17.0 ± 13.4 ‰). In the pan-Arctic dataset using all plant growth forms, we found that n-alkanoic acid δ2H had a weak, positive correlation (r = 0.25; p = 0.02) with MAF precipitation δ2H (Fig. S2a), while n-alkane δ2H had a moderate, positive correlation (r = 0.64; p<0.01) with MAF precipitation δ2H (Fig. S2b). When removing the data from Hollabåttjønnen Bog in northern Norway , however, the correlation between n-alkane δ2H and precipitation δ2H became weakly negative and not statistically significant (r = -0.18; p = 0.13). Additionally, the two plant wax compound-specific δ2H datasets do not represent the same samples/sampling locations, with the lack of n-alkane δ2H from ECA plants as an example.

Since δ2H values of individual plants are partially determined by the δ2H of precipitation at each site, we normalize that large amount of variability in our datasets by comparing the ϵapp values between plant growth forms, instead. ECA n-alkanoic acid ϵapp ranged from -202.7 ‰ to -92.2 ‰ with an average of -135.7 ± 25.9 ‰ (Fig. 3g). Pan-Arctic n-alkanoic acid ϵapp ranged from -202.7 ‰ to -47.4 ‰ with an average of -130.2 ± 27.2 ‰, and n-alkane ϵapp ranged from -221.4 ‰ to -37.6 ‰ with an average of -117.0 ± 31.1 ‰. Graminoids had the most negative n-alkanoic acid ϵapp values in both the ECA and pan-Arctic datasets. Liverworts and lichens had the least negative n-alkanoic acid ϵapp values in the ECA dataset, which were also the only such samples of those growth forms in the pan-Arctic. This pattern of graminoids and lichens (no liverwort data available) having the most and least negative ϵapp values, respectively, was also true for pan-Arctic n-alkanes (Fig. 3h). Similar to δ13C, we were unable to perform Shapiro-Wilk and Mann-Whitney U tests on the δ2H of some growth forms due an insufficient number of measurements (Table S1, Fig. 4g, h). Most growth forms with enough ϵapp measurements were found to be normally distributed except for shrubs, despite having the most measurements available among all growth forms (Fig. 3g, h). Moss and lichen ϵapp distributions were the most significantly different from other growth forms in the n-alkanoic acid (Fig. 4g) and n-alkane (Fig. 4h) datasets, respectively. However, Mann-Whitney U tests with stable isotope data more often showed that measurement distributions within individual plant growth forms were not significantly different compared to tests with total plant wax concentration and ACL data.

We performed PCA on plant wax n-alkanoic acid and n-alkane chain-length abundances using both the ECA lakes dataset and pan-Arctic data (Fig. 5). The first two principal components (PCs) in all analyses explained the majority of overall variance, ranging from 69 % to 77 % (Fig. S3a–d). The explained variance successively decreased substantially for the remaining PCs, 3–7, with each explaining no more than 14 % of the total variance. Eigenvectors for the four shortest chain-length plant waxes (C₂₀–C₂₆ / C₂₁–C₂₇; n-alkanoic acids / n-alkanes) always had positive loadings on PC1 (x-axis), and eigenvectors for the three longest chain-length plant waxes (C₂₈–C₃₂ / C₂₉–C₃₃) had negative loadings. Generally, eigenvectors for the one to two shortest and longest chain-length had negative loadings on PC2 (y-axis) and eigenvectors for the remaining chain-length plant waxes had positive loadings. For interpreting the PCA scores of individual samples, PC1 scores reflect a greater abundance of mid-chain (positive) vs. long-chain (negative) waxes, and negative PC2 scores represent a greater abundance of the longest (C₃₂ / C₃₃) and/or shortest (C₂₀ / C₂₁) chain-lengths.

ECA vascular plants tended to produce more longer-chain n-alkanoic acids than non-vascular plants, resulting in them plotting more negatively on PC1 (Fig. 5a). This relationship partly covaries with sampling location, since non-vascular plants are more prevalent in the northern ECA sites (AFR, CF8) while vascular plants dominate southern ECA sites (QPT, 3LN). Pan-Arctic vascular plant n-alkanoic acid distributions were more variable, but were mostly absent from Quadrant IV containing the C₂₀ and C₂₂ loading factors (Fig. 5c). The majority of non-vascular pan-Arctic n-alkanoic acids had positive PC1 scores associated with shorter-chain length production and negative PC2 scores. ECA vascular plant n-alkanes also tended to have negative PC1 scores, while moss PC scores were highly variable (Fig. 5b). ECA liverwort and lichen n-alkane distributions were more tightly clustered than n-alkanoic acids, mostly plotting along the C₂₁ loading factor. This pattern of high PC score variability among n-alkane distributions for all plant growth forms, except for liverworts and lichens, was also true for pan-Arctic n-alkane distributions (Fig. 5d).

The sampling locations in the ECA and pan-Arctic datasets spanned a substantial range in MAF (growing season) environmental parameters across 14° latitude. Temperatures ranged from 3.6 to 12.4 °C (Fig. 2a), total precipitation ranged from 130.3 to 803.5 mm (Fig. 2b), relative humidity ranged from 56.3 % to 80.1 % (Fig. 2c), elevation ranged from 0 to 950 m a.s.l. (Fig. 2d), and OIPC-derived precipitation δ2H ranged from -145.5 ‰ to -86.6 ‰ (Fig. 2e). The transect of ECA sites (AFR, CF8, QPT, 3LN) also contained a wide range of precipitation δ2H values, but are generally characterized as being colder and drier compared to other sampling sites in the pan-Arctic dataset. We used sample site latitude, temperature, precipitation amount, relative humidity, and elevation in Pearson correlation tests with plant wax data. Correlations between individual environmental parameters (i.e., temperature vs. precipitation) were mostly weak (|r| ≤ 0.4), with the only correlation between temperature and relative humidity (r = -0.50) being of moderate (0.4<|r| ≤ 0.8) strength (Fig. 6a, b; outside of the black boxes).

Pearson correlations between plant wax n-alkanoic acid / n-alkane data from all plant growth forms and environmental parameters yielded weak to moderate positive and negative linear relationships (Fig. 6a, b; black boxes), but with many of them not being statistically significant (p ≥ 0.05). The only significant relationships with n-alkanoic acid data were weak positive (r = 0.16, 0.19) correlations between total (log) concentration and precipitation amount and relative humidity, weak to moderate positive (r = 0.36, 0.42) correlations between ACL and temperature and precipitation amount, and a weak negative correlation (r = -0.23) between total concentration and latitude (Fig. 6a). There were no significant correlations between n-alkanoic stable isotope data and environmental parameters. Terrestrial plant n-alkane had a weak positive (r = 0.17) and weak negative (r = -0.26, -0.18) correlations between ACL and precipitation amount, latitude, and elevation, respectively (Fig. 6b). n-alkane ϵapp also had a weak positive (r = 0.37) correlation with precipitation amount. The δ13C values of n-alkanes were only reported in plants from a single study in west Greenland sampled during a single year (2015), which prevented correlations with environmental parameters.

We repeated these Pearson correlations with the plant wax data of a single growth form removed each time (e.g., performed correlations with all trees removed, then all shrubs removed, etc.). This allowed us to assess the influence of different growth forms on each correlation via the change in magnitude and direction of the r-value each time (Fig. 6c, d) while tagging the growth form that caused the greatest change when removed (Fig. 6e, f). Pearson r-value changes ranged from -0.37 to 0.20 for the n-alkanoic acid dataset (Fig. 6c), with the removal of mosses and shrubs being responsible for each extreme, respectively (Fig. 6e). These two growth forms were most often responsible for the greatest change in r-value for plant wax vs. environmental correlations. Changes in n-alkane correlation r-values ranged from -0.21 to 0.15 (Fig. 6d) with lichens and shrubs being responsible for the extreme values and the most frequent changes in r-value (Fig. 6f). Liverwort and lichen removal also caused some of the largest r-value changes in the n-alkanoic acid dataset, while graminoid removal did the same for the n-alkane dataset.

To further dissect potential empirical relationships between plant wax data and environmental parameters, we investigate linear relationships between several subsets of the pan-Arctic data, including vascular and non-vascular plants (Fig. 7), individual plant growth forms (Fig. 8), and individual plant genera (Fig. 9). For all data subsets listed, the majority of correlations were not statistically significant, due to either high variability (very low r-value) in the plant wax data, low sample sizes, or a combination of the two.

Vascular plants reflected similar correlation patterns to the full dataset: Weak to moderate positive correlations with n-alkanoic acid concentration and ACL (Fig. 7a), and a moderate correlation between n-alkane ϵapp and precipitation amount (Fig. 7b). Non-vascular plant waxes revealed different relationships with environmental parameters. Both n-alkanoic acid and n-alkane ACL were moderately positively (r = 0.51, 0.60) correlated with precipitation amount, whereas n-alkanoic acid δ13C and ϵapp were both moderately to strongly negatively (r = -0.67, -0.86) correlated with precipitation amount (Fig. 7c, d). Additionally, non-vascular plant n-alkanoic acid ACL (r = -0.51), δ13C (r = 0.76), and ϵapp (r = 0.63) were all moderately correlated with latitude (Fig. 7c).

Shrubs, graminoids, and mosses contained the most plant wax measurements among plant growth forms, and spanned the greatest number of sampling locations/unique environmental data points. The shrub-only data subset, containing the most measurements among growth forms, maintained the correlation patterns present in the full pan-Arctic and vascular plant groupings but with a now-weak positive correlation between n-alkane ϵapp and precipitation amount (Fig. 8a, b). Graminoids displayed a number of moderate correlations not seen in other data subsets, including relationships between ϵapp and precipitation amount (r = -0.60, -0.55; n-alkanoic acids, n-alkanes), relative humidity (r = 0.63; n-alkanoic acids), and elevation (r = -0.66; n-alkanes). Graminoids also had a negative (r = -0.53) correlation between total n-alkanoic acid total concentration and temperature plus a positive correlation (r = 0.42) between total n-alkane concentration and latitude (Fig. 8c, d). Mosses had weakly negative correlations between n-alkanoic acid ACL and latitude (r = -0.37) and total concentration and temperature (r = -0.39), along with weakly positive correlations between total concentration and relative humidity (r = 0.31) and ACL and precipitation amount (r = 0.37) (Fig. 8e). The only significant correlation present in moss n-alkanes was a moderately positive (r = 0.57) correlation between ACL and precipitation amount (Fig. 8f).

Betula sp. contained the greatest number of measurements and unique sampling locations among all Arctic plant genera, and all but one (56 of 57) samples were from the Betula nana/glandulosa complex. We excluded the Betula pubescens sample from Hollabåttjønnen Bog in these correlations due to it being classified as a tree, not a shrub like Betula nana and Betula glandulosa. Salix sp. was the second most sampled genus, but contained a more diverse set of species (e.g., Salix glauca, Salix arctica, Salix pulchra, etc.). Salix sp. n-alkanoic acids showed several significant correlations with environmental parameters, including a notably strong positive (r = 0.96) relationship between δ13C and latitude (Fig. 9a). The only significant correlation for Salix sp. n-alkanes was a moderately negative (r = -0.47) relationship between ACL and elevation (Fig. 9b). Similarly, Betula sp. n-alkanoic acids only had one significant correlation between ACL and latitude (Fig. 9c). Betula sp. n-alkane ACL and ϵapp were moderately negatively (r = -0.73) and moderately positively (r = 0.70) correlated with precipitation amount (Fig. 9d).

Plant wax production is often a response mechanism by individual plants to prevent water loss under stress : more water stress causes a higher total concentration of plant waxes. demonstrated that total n-alkane concentrations in Eucalyptus sp. were negatively correlated with annual precipitation amount and relative humidity. However, we generally do not observe this relationship between terrestrial Arctic plants and MAF total precipitation amount or relative humidity. Cold summer temperatures and restricted soil drainage due to the presence of bedrock and permafrost minimize water limitations during the growing season in the Arctic , the period during which plant wax synthesis occurs . The only exception was non-vascular plant n-alkanes, which had a weak negative correlation with relative humidity (Fig. 7d). In fact, the environmental correlations of moderate strength with total concentration were actually positive, as seen in vascular plant (r = 0.41) and graminoid (r = 0.49) n-alkanoic acid data subsets with relative humidity (Figs. 7a, 8c). also suggested that total plant wax concentrations increased with elevation in tropical forest vegetation in response to decreasing temperatures and lower nutrient availability. However, this correlation for terrestrial Arctic plants was, at best, of weak strength in non-vascular (r = 0.26) and graminoid (r = 0.40) n-alkane data subsets (Figs. 7d, 8d). It is possible that any trends in total plant wax production were masked by the high measurement variability present in all plant growth forms: the average standard deviation for growth form-specific n-alkanoic acid concentrations was 572 and 891 µg g⁻¹ for n-alkanes.

Shapiro-Wilk tests showing that total plant wax concentrations within individual plant growth forms were often log-normally distributed (Table S3) agreed with other large compilations of plant wax data from western Africa and the Tibetan Plateau . Some studies have demonstrated significant differences in total plant wax concentrations between plant types, notably that angiosperms produce more n-alkanes than gymnosperms . However, fewer such comparisons have been performed between the groupings of plant growth forms analyzed in this study. Shrubs have distributions of total n-alkanoic acid and n-alkane concentrations that are significantly different (Fig. 4a, b) and greater (Fig. 3a, b) than most other plant growth forms. This supports the hypothesis that waxes from these plants have greater potential to be strongly represented in Arctic sedimentary records . The opposite may be the case for growth forms that produce relatively low plant wax concentration, including liverworts and lichens.

The mechanisms behind a plant’s production of different plant wax carbon chain-lengths in response to environmental conditions is much less understood compared to the total production of these compounds. It has been suggested that plants produce longer chain-lengths, resulting in a greater ACL value, to combat water stress in warmer and/or drier growing conditions . These trends have been observed in plant n-alkane ACL across substantial temperature and aridity gradients in the eastern U.S. , central U.S. , and Europe . Therefore, we would expect terrestrial Arctic plant wax ACL to be positively correlated with temperature and negatively correlated with precipitation amount and relative humidity.

We do observe weak to moderate positive correlations between ACL and temperature in all plant n-alkanoic acids (r = 0.36; Fig. 6a), vascular plants (r = 0.36; Fig. 7a), shrubs (r = 0.38; Fig. 8a), and graminoids (r = 0.49; Fig. 8c), but not in any n-alkane data subsets. Nearly all significant correlations between ACL and precipitation amount and relative humidity were weakly to moderately positive, with the exception of Betula sp. n-alkane ACL (r = -0.73; Fig. 9d). Apart from that moderate correlation in Betula sp., the lack of significant correlations, otherwise, supports previous findings on Betula sp. that their plant wax distributions are highly variable but not governed by environmental variables . While these results disagree with the assumed mechanisms behind plant wax chain-length production, and also do not find ACL to have a significant relationship with water availability. also showed that the ACL for different genera can have opposing positive and negative correlations to precipitation and relative humidity within the same environmental gradient. Such an effect may obfuscate plant wax relationships in individual taxa to environmental data when data are grouped at the growth form level and above.

Mann-Whitney U tests revealed that ACL values did not differ between vascular plant growth forms, while non-vascular plants, particularly liverworts and lichens, had unique ACL distributions relative to all other growth forms (Fig. 4c, d). These patterns were also seen in the PCA biplot space where non-vascular plants tended to cluster in Quadrant IV with the two shortest n-alkanoic acid (C₂₀, C₂₂) and n-alkane (C₂₁, C₂₃) chain-lengths (Fig. 5c, d). Trees, which had some of the highest average ACL values, mostly plotted in Quadrant III of the PCA biplots containing the two longest chain-lengths. This is in contrast to other sampled temperate and subtropical/arid biomes where tree ACL was shown to be lower than many shrubs and graminoids . found similar ACL patterns in n-alkanes from temperate plant species where vascular plant species had similar, highly variable ACL distributions while mosses had distinctly lower values. Interestingly, pan-Arctic moss n-alkane ACL values were not significantly different from vascular plants, whereas the trend held true for pan-Arctic n-alkanoic acids. These patterns in Arctic vegetation echo warnings from previous compilations of plant ACL values that chain-length distributions are not sufficient for fingerprinting different types of vascular plants , but discerning changes between vascular and non-vascular plant wax sources may still be possible.

The fractionation of stable carbon isotopes between the atmosphere and C₃ plant waxes is strongly influenced by a plant's stomatal conductance in its leaves responding to water availability . In global compilations of plant samples, this mechanism is illustrated as negative correlations between plant wax n-alkane δ13C and temperature and precipitation amount . Arctic plant wax n-alkanoic acid and n-alkane δ13C values all fell within the global range of variability (∼ -45 ‰ to -20 ‰) for C₃ photosynthesizers . However, the only significant correlation between δ13C and any environmental variable was a strongly negative (r = -0.86) relationship between non-vascular n-alkanoic acids and precipitation amount (Fig. 7c). Interestingly, mosses are the only growth form of the non-vascular plants that can have stomata within their structure, while liverworts and lichens do not . Therefore, non-vascular plant ¹³C fractionation sensitivity to water availability in these growth forms is likely driven by other physiological factors.

Arctic plant wax δ13C data was limited to this study, Lake QPT in the ECA , and west Greenland , so it is possible that significant correlations in vascular plants were not resolved due to a low measurement count across a smaller gradient of temperature and precipitation amount. This also applies to investigating differences in δ13C between plant growth forms. Shrubs had the most significantly difference distribution of n-alkanoic acid δ13C values based on Mann-Whitney U tests (Fig. 4e), but many growth forms did not have any measurements (Fig. 3e, f). The high degree of overlap in δ13C values between plant growth forms suggests that caution should be used when interpreting past terrestrial vegetation changes using sedimentary plant wax δ13C in the Arctic. This large overlap impacts the ability to construct unique δ13C-based vegetation endmembers for numerical modeling of plant wax sources .

Plant wax δ2H values and their associated ϵapp relative to their source water δ2H are controlled by multiple factors, including evaporative ²H-enrichment that occurs in the soil and plant's xylem prior to plant wax synthesis, as well as the biosynthetic fractionation factor (ϵbio) which heavily discriminates against the heavier ²H isotope . found moderate positive correlations between ϵbio and temperature and precipitation amount along a north-south transect of Alaskan shrubs and forbs. However, disentangling the environmental controls on ²H fractionation pre- (evaporation) and post- (ϵbio) plant wax synthesis would require more measurements of plant leaf water, which are currently limited to only two other Arctic datasets/sampling locations in addition to the data from .

If we assume ϵbio is constant for individual plants, increased evaporation results in a smaller absolute (less negative) ϵapp value . In this framework, environmental correlations with ϵapp values are expected to relate to evaporation potential: positive correlations with increasing temperature and negative correlations with increasing precipitation and relative humidity. However, the only cases in which these expected relationships were present in the pan-Arctic dataset were a moderately negative correlation between non-vascular n-alkanoic acid ϵapp and precipitation amount (r = -0.67; Fig. 7c) as well as moderately negative correlations between graminoid n-alkanoic acid ϵapp and precipitation amount (r = -0.60) and relative humidity (r = -0.63; Fig. 8c). A number of data subsets expressed opposite relationships than expected, including moderately positive correlations between n-alkane ϵapp in vascular plants (n = 0.47; Fig. 7c), graminoids (r = 0.55; Fig. 8d), and Betula sp. (r = 0.70; Fig. 9d). found weakly negative correlations between mean annual precipitation amount and terrestrial monocot (r = -0.36) and dicot (r = -0.38) n-alkane ϵapp across China. However, mean annual precipitation amounts in their sampling locations ranged from 30 to 1720 mm, with much more scatter in ϵapp values in the range of 30 to ∼ 600 mm of mean annual precipitation . It is possible that the ranges of MAF total precipitation amount, temperature, and relative humidity in the pan-Arctic dataset are not enough to drive consistent, significant changes in ϵapp . Though, it is also possible that some of the total ϵapp variability is lost by using spatially interpolated OIPC precipitation isotope δ2H values instead of in situ measurements .

Previous studies have also found positive correlations between ϵapp and latitude in Arctic vegetation, where latitude serves as a proxy for the daily length of daylight during the growing season . These correlations were supported by experimental results where increased light exposure promoted more plant evapotranspiration, causing ϵapp values to be more enriched . In the pan-Arctic dataset, the only data subset that showed a similar, positive correlation (r = 0.63) between ϵapp and latitude was non-vascular n-alkanoic acids (Fig. 7c). Based on their n-alkanoic acid ϵapp correlations between precipitation amount and latitude, it could be possible that non-vascular terrestrial Arctic plants are more sensitive to evaporation than vascular ones. However, this pattern is likely not robust since these correlations are not present in non-vascular n-alkanes (Fig. 7d) and also do not manifest in mosses (Fig. 8e, f), despite them being the most abundant non-vascular plant growth form in the pan-Arctic dataset.

Weak correlations between plant wax and precipitation δ2H illustrates the high variability in ϵapp at each sampling location (Fig. S2). The anomalously ²H-enriched values from Hollabåttjønnen Bog in northern Norway were likely caused by those plants using bog water (-60 ‰) that was much more ²H-enriched than our calculated amount-weighted MAF precipitation δ2H (-87 ‰) based on ERA5 reanalysis and the Online Isotopes in Precipitation calculator . While such issues can occur when using this approach, it serves as a consistent data source and methodology for datasets where site-specific meteorological and precipitation isotope data were not collected nor available elsewhere. For example, this information was not collected at each ECA lake and there are large discrepancies in distance to the nearest weather and precipitation isotope monitoring stations . Additionally, found high ϵapp variance (standard deviation > 20 ‰) in deciduous and evergreen trees growing in the same climate and irrigated with δ2H-controlled water. Therefore, high ϵapp variability within individual growth forms may also be a function of plant physiology, not just from discrepancies between measured and inferred plant source water. and other studies also note the substantial variability we observe in plant wax δ2H and δ13C values between chain-lengths from the same sample and have attributed it to systemic changes in biosynthetic fractionation as plants produce different wax chain-lengths.

The lack of distinct distributions of ϵapp values between plant growth forms may be beneficial for reconstructing past Arctic hydrology. The only pattern consistent with other, globally-distributed plant wax ϵapp compilations was Arctic graminoids having the most negative values, on average, among growth forms . Some studies report vegetation-corrected source water δ2H records based on ϵapp values of different plant types and their estimated contributions to the sediment . Based on our results, calculating MAF precipitation δ2H using pan-Arctic ϵapp averages of -130.2 ± 27.2 ‰ for n-alkanoic acids and -117.0 ± 31.1 ‰ for n-alkanes may be sufficient in the absence of a paired vegetation record from the same site.