Leaf water δ¹⁸O reflects water vapour exchange and uptake by C₃ and CAM epiphytic bromeliads in Panama

Keywords: C₃, CAM, Crassulacean acid metabolism, Tillandsia fasciculata, Werauhia sanguinolenta, photosynthetic pathway, gas exchange, epiphyte, oxygen isotopes, altitudinal gradient, mesocosm.

Epiphytes provide sensitive climatic indicators, with their distribution reflecting both the microclimate within a particular forest canopy, as well as altitudinal zonation between forest formations (Gómez González et al. 2017; Horwath et al. 2019). Vascular epiphytes can comprise a significant component of diversity in lower montane forest canopies, and the distribution of C₃ and Crassulacean acid metabolism (CAM) epiphytic bromeliads within neotropical forests has long provided a model system to integrate physiological ecology, life-form and habitat preference (Osmond 1978; Griffiths and Smith 1983; Crayn et al. 2015; Males and Griffiths 2018). Here, we analyse the distribution of C₃ and CAM bromeliads along an altitudinal gradient in western Panama using stable isotopes to evaluate contrasting strategies for water use and exchange of water vapour.

Epiphytic bromeliad species exhibit great diversity in functional forms through an array of morphological and physiological traits. The vegetative body of epiphytic bromeliads is mostly comprised of leaves, whereas roots and stems are highly reduced (Males 2016; Leroy et al. 2019). Leaves are displayed in a rosette that often forms a water reservoir or ‘tank’ between overlapping leaf bases. The water-impounding tank constitutes a stable water source between rain events (North et al. 2013; Males 2016) but may be subject to direct evaporation (Zotz and Thomas 1999; Males and Griffiths 2018). With extensive leaf trichome cover (Benzing 1976), leaf wetting characteristics (Pierce et al. 2001; Leroy et al. 2019) also allow bromeliads to inform current studies on direct water uptake by leaves (Dawson and Goldsmith 2018; Berry et al. 2019).

Water availability and storage are a major determinant of bromeliad morphology and physiology (North et al. 2013; Males 2016; Males and Griffiths 2018). The C₃ pathway tends to predominate in areas where rainfall is more frequent, whereas nocturnal stomatal opening (when transpiration is reduced) allows CAM bromeliads to predominate more in semiarid tropical forests or exposed portions of the canopy (Griffiths and Smith 1983). Bromeliad stomata are particularly sensitive to ambient vapour pressure (Lange and Medina 1979). The stable isotope signals in leaf carbon (δ¹³C) distinguish those species with CAM (Osmond 1978), whereas oxygen isotopes (δ¹⁸O) reflect water use characteristics (Seibt et al. 2008; Cernusak et al. 2016; Dubbert et al. 2017). Precipitation and source water inputs are usually depleted in ¹⁸O by around –5 to –10‰ relative to the Vienna standard mean ocean water (VSMOW) mass spectrometric standard but varies on a seasonal basis (Cernusak et al. 2016). For C₃ plants, leaf water is normally evaporatively enriched in ¹⁸O during transpiration, and then subject to an additional biochemical enrichment of some +27‰ when transferred into organic material (Sternberg et al. 2006; Cernusak et al. 2016; Lehmann et al. 2020).

The δ¹⁸O signal in leaf water and organic material in bromeliads reflects the relative inputs from contrasting water sources such as precipitation, tank water (which may become evaporatively enriched between rain events) and atmospheric water vapour (Farquhar and Cernusak 2005; Seibt et al. 2008; Cernusak et al. 2016; Lehmann et al. 2020; Benettin et al. 2021). The net efflux of water vapour during gas exchange normally leads to evaporative enrichment of ¹⁸O in leaf water (δ¹⁸O_lw) (Harwood et al. 1998), whereas water vapour ingress under high ambient humidity can lead to more depleted δ¹⁸O_lw (Farquhar and Cernusak 2005), allowing niche specialisation to be defined for C₃ and CAM epiphytic bromeliads (Helliker and Griffiths 2007; Reyes-García et al. 2008; Helliker 2011; Lehmann et al. 2020). Provided that the interplay between water sources, water use and degree of steady-state equilibration for isotopic signals in both C₃ and CAM tissues can be well defined, ¹⁸O signals in both leaf water (δ¹⁸O_lw) and organic material (δ¹⁸O_OM) of epiphytes can be used to model climatic conditions along altitudinal and latitudinal gradients for the epiphytic bromeliad Tillandsia usneoides Sw. (Helliker 2014).

The objectives of this study were, first, to define the distribution of C₃ and CAM epiphytic bromeliads along an altitudinal gradient in western Panama, using δ¹³C analyses to identify the metabolic pathway; and second, to use a combination of field sampling and laboratory experimentation to identify the determinants of δ¹⁸O_lw and δ¹⁸O_OM signals in the two photosynthetic pathways. We expected that CAM species would be more prevalent in the drier lowlands, with a higher abundance of C₃ species in higher elevation forest systems. The contrasting climatic conditions and water source inputs along the altitudinal gradient allowed us to test the hypothesis that net water vapour uptake under high humidity would re-set the δ¹⁸O_lw for epiphytes (Farquhar and Cernusak 2005; Helliker and Griffiths 2007; Reyes-García et al. 2008; Seibt et al. 2008; Lehmann et al. 2020). For C₃ bromeliads, we hypothesised that that δ¹⁸O_lw would show evaporative enrichment in lowland habitats, but would be more depleted at high altitude cloudforest habitats due to net water vapour inputs. For CAM bromeliads, we predicted that δ¹⁸O_lw would also be depleted, due to water vapour inputs associated with stomatal opening at night under high atmospheric humidity (Helliker and Griffiths 2007; Reyes-García et al. 2008). Finally, we also set out to test whether water source inputs, from precipitation trapped in the bromeliad tanks, would be affected by growing season and leaf morphology, with higher evaporative ¹⁸O enrichment associated with more open leaf rosettes.

To test these hypotheses, and summarised in Fig. 1, we compared δ¹⁸O_lw, δ¹⁸O_OM and source water inputs for epiphytic CAM and C₃ bromeliads along a 1300 m gradient over the Central Cordillera in western Panama, from the Pacific Coast in Chiriquí province, north to the Caribbean coast in Bocas del Toro. Measurements were taken in both rainy and dry seasons to capture the variations in source water δ¹⁸O inputs and likely evaporative demand. A more detailed analysis was undertaken for two sympatric species, Tillandsia fasciculata (CAM) and Werauhia sanguinolenta (Linden ex Cogniaux & Marchal) J.R.Grant (C₃), that dominate the lower part of the altitudinal transect (between sea level and 600 m; Pierce et al. 2002a; Zotz et al. 2005). These species were also investigated experimentally, using an enclosed mesocosm, to measure the direct contribution from water vapour uptake under well-watered and drought-stressed conditions for C₃ and CAM systems (Helliker and Griffiths 2007). This study confirmed the field work observations, showing that water vapour uptake at high humidities masked evaporative enrichment during transpiration, and was consistent with a mechanistic model of ¹⁸O enrichment, explaining depleted δ¹⁸O_lw for CAM and upland C₃ bromeliad epiphytes.

Fig. 1.  Summary of the fieldwork and experimental set up showing the scope of sampling and progression from fieldwork to laboratory mesocosm experimentation.

Fieldwork was conducted in 2006 along an altitudinal and precipitation gradient across the Central Cordillera of western Panama, where the dry season extends from January to April (Cavelier et al. 1996). The gradient encompassed 20 sites, incorporating lowland coastal sites located within 1 km of the Pacific Coast, up to pristine montane forest sites at 1300 m above sea level and back down to pastureland within 500 m of the Caribbean Coast (see Supplementary material Table S1). Monthly rainfall during the wet season is over 700 mm in the montane forest, but around 400 mm in the lowlands. During the dry season, the lowlands receive minimal rainfall, with montane sites receiving around 150 mm (data provided by Mr Ambrosio Morales, ETESA (Empresa de Transmisión Eléctrica, Panama)).

In order to determine the natural range of spatial and seasonal variation in δ¹⁸O_lw and δ¹⁸O_OM for C₃ and CAM species, samples of leaf tissue from 36 bromeliad species (Table S2) were collected during the dry (February) and rainy (June) seasons in 2006 along the altitudinal gradient (Fig. 2). Leaf material was taken from the middle portion of fully expanded leaves to reduce isotopic variation associated to different leaf portions (Helliker and Ehleringer 2000; Ogée et al. 2007). Leaf samples were stored in sealed glass tubes. To avoid differential effects of transpiration all collections were made between 1100 and 1400 hours. If available, samples of tank water and precipitation were also collected with leaf tissue, for subsequent isotopic analysis.

Fig. 2.  Map of the study area (a) showing monthly precipitation (mm) for selected sites across the Central Cordillera in western Panama. The diagram (b) illustrates the relationship between site elevation and distance to the Caribbean (km), the best indicator of rainfall isotopic signature in the region. Each site is classified as Fragmented low-land (Pacific, closed circle), Fragmented mid-elevation (closed square), Pristine forest (closed triangle) and Fragmented low-land (Caribbean, open circle), with the site at 31 km representing the montane cloudforest Fortuna Nature Reserve. Total number of epiphytic bromeliads is shown for each site (pie charts) with C₃ species in black and CAM species in grey, with identification based upon carbon isotope ratio analysis (δ¹³C) for each specimen. Note that distances between sites are not scaled.

For three of the sites, representing contrasting environmental conditions, an additional water source, atmospheric vapour, was sampled in order to understand the relationship with δ¹⁸O_lw. The sites represented lowland (90 m elevation, 8°15′839′′N, 81°51′243′′W) and mid elevation (650 m elevation, 8°35′134′′N, 82°14′125′′W) pasture lands with dispersed trees, as well as a cloudforest (978 m elevation, 8°35′125′′N, 82°14′151′′W). Because δ¹⁸O of atmospheric water vapour can be very variable across the day due to atmospheric conditions, the samples were taken at five times during the day: 0600, 1000, 1400, 1800, and 2200 hours. Values shown here represent the mean of those measurements. Atmospheric vapour samples were collected using a dry ice-ethanol trap (Helliker et al. 2002); and were processed at the University of Cambridge, UK.

Individuals of Werauhia sanguinolenta (Linden ex Cogniaux & Marchal) J.R.Grant (C₃) and Tillandsia fasciculata (Sw. var. fasciculata; CAM) adapted to greenhouse conditions in Cambridge, UK were used to investigate the relationship between liquid water and atmospheric water vapour as determinants of C₃ and CAM leaf water isotopic composition. Following Helliker and Griffiths (2007), plants were placed inside 50 L sealed mesocosm within a growth chamber with a 12 h photoperiod (300 μmol m^–2 s^–1 at plant height; day/night temperatures within the mesocosm held at 25 and 20°C respectively, monitored with thermocouple thermometer). The epiphytes rested on a mesh above a water-filled tray, with three 5 cm electric fans facing across the reservoir used to generate a high humidity environment; for the low humidity regime, the mesocosm water reservoir was emptied. Two successive 16 day treatments were undertaken with either ‘fully hydrated’ plants (water in tanks, n = 14) or ‘water-stressed’ treatments (no water in tanks, n = 15). Plants were kept at continuously high RH regime (High/High, day: 90%, night: 90%) for 8 days, and for a further 8 days at diurnally variable RH (with water removed from mesocosm reservoir, leading to a Low/High RH regime, day: 50%, night: 90%). The tanks of fully hydrated plants were emptied and refilled each day with local tap water with a δ¹⁸O of –7.0 ‰; for water stressed plants with empty tanks atmospheric vapour was the only water source. The replicate plants (14 or 15 in the two experiments) had rosettes with many leaves, allowing triplicate leaf samples to be collected at random from different plants on days 0, 3, 7, 8, 9, 10, 11, 15, and 16 for isotopic analyses. Leaf relative water content (RWC) was measured at day 0 and at the end of each 8-day period using the relationship RWC = (fresh weigh – dry weight)/(saturated weight – dry weight) × 100. Over the 16-day experimental period, tissue RWC decreased respectively from 91.6 to 86.6 ± 0.8% (CAM) and 68 to 65.4 ± 2.1% (C₃) for the well-watered treatments (s.e. ranging from 0.8 to 2.1%), with a further decline to 62.3 ± 2.6% (CAM) and 31.3 ± 1.0% (C₃) following the drought treatment. The mean δ¹⁸O of the water used in the reservoir to generate water vapour was –6.2‰ (s.e. ± 0.2‰) (n = 14), and the atmospheric water vapour of air passing through the experimental chamber was –19.3‰ (s.e. ± 0.4‰) and –17.8‰ (s.e. ± 0.3‰), for the water-stressed (n = 15) and fully hydrated treatments (n = 14), respectively.

Leaf tissues sampled in the field and from the mesocosm experiment were initially stored in 12 mL glass tubes sealed with a screw cap and Viton rubber seal. Bulk leaf water was extracted from leaf tissue using a cryogenic vacuum distillation system and analysed isotopically after equilibration with CO₂ on a mass spectrometer (VG SIRA 10, Modified by ProVac Systems, Crewe, UK), as described by Reyes-García et al. (2008). All δ¹⁸O values are reported with respect to VSMOW.

Leaf water δ¹⁸O values were corrected for atmospheric water vapour adsorbed to trichomes using the method of Helliker and Griffiths (2007). The adsorbed water was estimated to be 4 and 11% of the total collected water volume for 50 and 90% relative humidity, increasing δ¹⁸O_lw by 0.1 and 0.5‰, respectively.

A sub-sample of leaves from the most common species collected in the field was oven-dried at 70°C for 3 days and ground to a fine powder and subdivided for isotopic analysis of organic matter. δ¹⁸O_OM analysis was completed by Dr C. Keitel, Australian National University, Canberra, following (Farquhar et al. 1997). δ¹³C_OM was determined at the Godwin Laboratory, University of Cambridge, by an elemental analyser (Thermo Finnigan TC/EA) attached to a Thermo Delta V mass spectrometer (for details, see Horwath et al. 2019). Organic material was not sampled from the mesocosm experiment as insufficient new growth would have occurred during the short experimental period.

Patterns of δ¹⁸O_lw during the chamber experiments were simulated using a non-steady-state (NSS) model of leaf water enrichment (Seibt et al. 2008), taking into account the isotopic composition of source water and water vapour, the equilibrium liquid-vapour fractionation (9.8‰ at 20°C), the kinetic fractionation during diffusion of vapour (26.5‰, based on the ratio of stomatal and leaf boundary layer conductances; Farquhar et al. 1989; Farquhar and Cernusak 2005; Barbour 2007; Cernusak et al. 2016), RH of air, leaf temperature, stomatal conductance, and changes in leaf water content. Calculations started from observed δ¹⁸O_lw values. For the well-watered plants, the isotopic composition of tank water was used as the source water term. Stomatal conductance (g_s) was parameterised from gas exchange measurements (using a Li-Cor 6400 system, Li-Cor Biosciences) for W. sanguinolenta (C₃) during the day (0.016 mol m^–2 s^–1) and set to a small value at night (0.001 mol m^–2 s^–1). For T. fasciculata (CAM), g_s was set to the opposite diurnal pattern (0.016 mol m^–2 s^–1 at night, 0.001 mol m^–2 s^–1 during the day). The difference between evaporating site and bulk leaf water due to the Péclet effect was also taken into account (eqns 17, 21, and 22 in Farquhar and Cernusak 2005). As the bromeliad leaves had a high leaf water content (22–34 mol m^–2 for well-watered plants), an effective path length of 30 cm was used to calculate the Péclet number.

For plants under water stressed conditions (no source water input), the model becomes equivalent to that of Helliker and Griffiths (2007). Stomatal conductance was calculated from the observed leaf water loss. Assuming again small g_s at night (C₃) or day (CAM), g_s during the day (C₃) or night (CAM) was adjusted so that the cumulative transpiration matched the observed water loss from the leaves. The resulting g_s values were 0.005 mol m^–2 s^–1 and 0.001 mol m^–2 s^–1 (C₃), and 0.004 mol m^–2 s^–1 and 0.003 mol m^–2 s^–1 (CAM), for the two experimental periods (High/High, Low/High RH), respectively. In the absence of advection from source water towards the sites of evaporation, there is no Péclet effect. In addition, the Péclet number would be small (0.02) due to the small transpiration rates, so that the difference between evaporating site and bulk leaf water would only be in the order of 1% (Farquhar and Cernusak 2005).

Statistical analysis was carried out in StatView version 5.0 (SAS Institute Inc.) and R (ver. 3.6.3; R Core Team 2018). All analyses were considered significant at P < 0.05 and data presented as mean ± s.e. Depending upon the normality of the data distribution, t-test or Wilcoxon analyses were used to compare dry vs rainy season variation in water sources, as well as overall seasonal variation in isotopic composition of leaf water for C₃ and CAM species. Where appropriate, ANOVA’s were used and if significant differences were found, Tukey´s method for pair-wise comparison was applied to discriminate differences within variables. Paired t-test analyses were used to compare the isotopic variation in the humidity chamber experiments. Variation in δ¹⁸O_lw in relation to δ¹⁸O tank water along the altitudinal gradient was evaluated with a regression analysis. All analyses were considered significant at P < 0.05 and data presented as mean ± s.e.

In Fig. 2a, the seasonality of precipitation is depicted across the continental divide in Panama, running south to north from the Province of Chiriqui (Pacific) to Bocas del Toro (Caribbean). A more distinct seasonal pattern of rainfall is found for the low altitude Pacific Coast, as compared with the high altitude cloudforest site at the Fortuna Nature Reserve (1239 m). Although the focus of this study was primarily on the southern aspect of this altitudinal transect, it is intriguing that the distribution of CAM bromeliads predominates in the lowland habitats on both sides of the continental divide (Fig. 2b), consistent with the higher temperatures and evaporative demand likely to be experienced in these habitats. At the higher altitude sites and those facing the prevailing rainfall inputs on the northern slope, the number of bromeliad epiphytes increases and become dominated by those expressing the C₃ pathway (Fig. 2b).

Water vapour was significantly depleted in δ¹⁸O, by around 7‰ relative to rainfall (t-test, t = –16.47, P < 0.001, Fig. 3a,b). Seasonal differences in precipitation, tank water and water vapour were significant (factorial ANOVA, seasonal effect, P < 0.01). Dry season water vapour was significantly enriched compared with the wet season (t-test, t = 2.34, P = 0.037). Water vapour from the lowland sites was significantly enriched in δ¹⁸O, in relation to the cloudforest site (t-test, t = –4.63, P < 0.001). Rainfall δ¹⁸O was also significantly more enriched during the dry season than during the wet season (W = 1262, P < 0.0001). In the wet season there was no significant difference between the tank water and rainfall δ¹⁸O (W = 3310, P = 0.9, Fig. 3a). During the dry season, evaporative enrichment increased the tank water δ¹⁸O above that of rainfall at low and mid-altitude sites (t = –4.47, P = 0.0005), whereas in the cloudforest sites, the two water sources were not significantly different (W = 258.5, P = 0.06) so tank water represented precipitation inputs (Fig. 3b).

Fig. 3.  Mean isotopic components for epiphytic Bromeliads summarised on a seasonal basis along an altitudinal and precipitation gradient in Panama. Data collated as mean values for specimens collected for low-, mid- and high-altitude (cloudforest) sites along a transect from the Pacific Coast to Fortuna; (a) Wet season δ¹⁸O (±s.e.) in water sources (Legend: rainfall = circles, water vapour = triangles, tank water = grey bars). (b) dry season δ¹⁸O in water sources for C₃ and CAM species (symbols as in Fig. 2a). (c) Boxplot (median and 1.5 × IQR) of δ¹⁸O in leaf water for CAM and C₃ species in the wet season. (d) Boxplot (median and 1.5 × IQR) of δ¹⁸O in leaf water for CAM and C₃ species in the dry season. (e) Boxplot (median and 1.5 × IQR) in annual leaf organic matter for CAM and C₃ species. (f) Boxplot (median and 1.5 × IQR) in annual leaf organic matter for CAM and C₃ species. Data derived from total number of C₃ + CAM samples across sites for Lowland (n = 35); Mid altitude (n = 40) and (cloud) Forest (n = 20), with species representation for each group: lowland (C3 = 2; CAM = 5); Mid altitude (C3 = 6; CAM = 7); (cloud) Forest (C₃ = 20).

Within all CAM and C₃ species, δ¹⁸O_lw values were significantly isotopically depleted during the wet season (C₃: –2.6 ± 1.2‰, CAM: –7.1 ± 0.7‰; Fig. 3c) than in the dry season (C₃: +3.7 ± 0.9‰, CAM: –0.6 ± 0.8‰), with CAM species having significantly more depleted δ¹⁸O_lw than C₃ species (factorial ANOVA seasonal and photosynthetic pathway effects P < 0.05; Fig. 3d). The high altitude, montane forest species had the most isotopically depleted C₃ δ¹⁸O_lw values in the wet and dry season (factorial ANOVA, significant season, elevation and photosynthetic pathway effects P < 0.05 and Fig. 3c, d). In comparison, for organic material, which integrates wet and dry season precipitation inputs, there was no significant variation in δ¹⁸O_OM between elevation for either CAM species (W = 223.5, P = 0.06) or C₃ species (ANOVA P = 0.1318, F = 2.095, DF = 2). CAM species growing at mid-elevation sites had significantly less negative values of δ¹³C (–13.2‰ ± 0.2) than plants at the lowlands (–15.0‰ ± 0.2) (Fig. 3f) (W = 6688, P < 0.001), and in contrast, no significant variation was detected for C₃ plants across habitats (ANOVA P > 0.05). As would be expected due to the different metabolic pathways, an offset in δ¹³C_OM of ~10‰ was measured between CAM and C₃ bromeliads.

The overlapping leaf bases ‘tanks’ provide a reservoir of water to sustain hydraulic supply between rain events, although tank water also tends to evaporate directly during the dry season. Fig. 4 shows the systematic variations in C₃ and CAM δ¹⁸O_lw in relation to tank water δ¹⁸, for all samples across both wet and dry seasons. For both C₃ and CAM species, when calculated independently of elevation and season, δ¹⁸O_lw was positively correlated with the δ¹⁸O of source tank water (Fig. 4, P < 0.0001, r² = 0.57 and P < 0.0001, r² = 0.75 for C₃ and CAM species, respectively). Across the range of tank water δ¹⁸O_lw values measured, there is a tendency for C₃ leaf water δ¹⁸O_lw to be more enriched than CAM (t-test, t = –3.57, P = 0.0007) with the offset reduced in the more negative values seen for the wet season (Fig. 4). This data is consistent with leaf water values for the two individual C₃ (W. sanguinolenta) and CAM (T. fasciculata) species shown in Table 1. The δ¹⁸O_lw signals of C₃ leaves was enriched in ¹⁸O relative to CAM, and were consistent with the overall pattern for all other sympatric C₃ and CAM species (Fig. 3c, d). There was also a shift within δ¹⁸O_lw for the CAM T. fasciculata, which was more pronounced from dry to wet season (depleted by around 9‰ at both altitudes) than in the C₃ W. sanguinolenta (Table 1).

Fig. 4.  Variation of δ¹⁸O in bulk leaf water as a function of tank water. Samples from epiphytic bromeliads growing along an altitudinal gradient in western Panama in the wet and dry seasons (data from Fig. 3). Lines represent linear models applied to the C₃ (solid) and CAM (dashed) data. Data include 37 combined paired samples for leaf and tank water, representing C₃ (23 dry; 14 wet season) and 22 CAM (14 dry; 8 wet season).

Table 1.  Comparison of seasonal variations in isotopic composition of mean leaf water and combined tank water for T. fasciculata and W. sanguinolenta
Data from field sites representing altitudinal extremes of co-occurrence along the altitudinal gradient. Values in parentheses for leaf water (δ¹⁸O_lw ‰) are s.e. (n = 12–15), with different letters indicating significant differences for a given column

We explored the systematic differences in δ¹⁸O_lw between two CAM and C₃ plants under controlled conditions in a mesocosm experiment, and analysed the role of water sources, photosynthetic pathways and evaporative control using a non-steady-state (NSS) model of leaf water enrichment (Farquhar and Cernusak 2005). For plants without source water input (drought stressed), the model is equivalent to that of Helliker and Griffiths (2007).

Although the tank water was replaced on a daily basis in the well-watered treatment (with local tap water with δ¹⁸O –7.0‰), over each 24 h period tank water became enriched for T. fasciculata (–4.8‰ ± 0.1) (n = 32) relative to W. sanguinolenta (–5.4‰ ± 0.1) (n = 40) (t-test, P < 0.0001), probably reflecting the more open leaf structure and shallower tank in T. fasciculata.

For well-watered plants, δ¹⁸O_lw decreased over the first 8 days of high RH (12‰ for T. fasciculata, 8‰ for W. sanguinolenta, P < 0.003) (Fig. 5a). After the switch to low daytime RH, isotopic enrichment occurred progressively for W. sanguinolenta, and after 8 days the δ¹⁸O_lw was enriched more in the C₃ (12‰, P = 0.0007) than the CAM plant (3‰, P = 0.04), relative to the respective values following the high humidity regime (Fig. 5a). The model simulations compare well to the observed δ¹⁸O_lw values (Fig. 5a). The model reproduces the gradual decrease of δ¹⁸O_lw in both species during the initial period of constant high RH (High/High), and the subsequent evaporative increase of δ¹⁸O_lw for the C₃ W. sanguinolenta but constant with depleted δ¹⁸O_lw for the CAM T. fasciculata during the final period of low daytime RH (Low/High).

Fig. 5.  Fluctuation of leaf water δ¹⁸O for T. fasciculata and W. sanguinolenta exposed to controlled relative humidity. Plants were subjected to 8 days of high day- and night-time relative humidity (day: 90%, night: 90%), and subsequently for a further 8 days to low daytime only relative humidity (day: 50%, night: 90%). (a) Fully hydrated plants; (b) water-stressed plants (n = 3 ± s.e.). Also shown are results from a non-steady-state (NSS) model of leaf water enrichment for each of the species (Werauhia NSS, continuous line; Tillandsia NSS, dashed line).

Water-stressed plants also experienced a depletion in δ¹⁸O_lw over the first 8 days at high RH (4‰ for T. fasciculata, 10‰ for W. sanguinolenta, P < 0.008) (Fig. 5b). No changes in isotopic values were found in either species during the low daytime RH treatment (Fig. 5b). For hydrated plants the model simulations reproduce the general patterns in δ¹⁸O_lw, but under droughted conditions the model under-predicted δ¹⁸O_lw for both species towards the end of the high-RH period and for the first few days of the low-RH period (Fig. 5b). The model captures reduced level of enrichment seen under droughted conditions in the C₃ W. sanguinolenta, with minimal gas exchange is occurring under low daytime humidity (Fig. 5b) consistent with directly measured stomatal conductance (data not shown).

The distribution of CAM and C₃ epiphytic bromeliads across the continental divide in western Panama was used as the basis for a comparison of the coupling in water budgets for the two photosynthetic pathways. The study set out to define whether leaf water oxygen stable isotopes (δ¹⁸O_lw) reflect the seasonal variability in precipitation, changes in δ¹⁸O of tank water reserves, and contrasting daily gas exchange patterns, along an altitudinal transect running north from the Pacific to montane cloud forest at Fortuna (Cavelier et al. 1996; Zotz et al. 2005; Gómez González et al. 2017; Benettin et al. 2021). The overall aim was to investigate whether water vapour exchange across leaf surfaces, when atmospheric water vapour concentrations are close to saturation, leads to significant inward fluxes of water which ‘reset’ the δ¹⁸O_lw under natural conditions in the field (Farquhar and Cernusak 2005; Seibt et al. 2006; Helliker and Griffiths 2007; Helliker 2011; Lehmann et al. 2020) in addition to potential liquid water uptake (Pierce et al. 2001; Dawson and Goldsmith 2018; Berry et al. 2019).

The distribution of C₃ and CAM bromeliad epiphytes did conform to the hypothesis that C₃ species would dominate in higher, cloudforest habitats (Fig. 2), and we explore the integration of bromeliad hydraulic properties and δ¹⁸O signals in ecological terms in the final section below. We also set out to test the hypothesis that the gas exchange of C₃ and CAM bromeliads would result in contrasting temporal and spatial drivers for δ¹⁸O_lw, when compared with source water δ¹⁸O inputs measured in terms of precipitation, tank water and water vapour (Dubbert et al. 2017; Benettin et al. 2021). Depending on seasonal conditions (dry or wet season), we thought that daytime gas exchange associated with C₃ species should be subject to evaporative enrichment in δ¹⁸O_lw at lower altitudes in the dry season. The impact of net water vapour uptake under high humidities was predicted to result in lower δ¹⁸O_lw at high altitude cloudforest sites in the dry season, and for low altitude sites in the wet season, as seen in experimental studies (Helliker and Griffiths 2007; Reyes-García et al. 2008; Lehmann et al. 2020).

The data shown in Fig. 3 confirm these predictions, where the significant shifts in δ¹⁸O source water inputs as enriched precipitation, tank water and water vapour in the dry season, relative to the wet season (Fig. 3a, b). The associated δ¹⁸O_lw signals in C₃ bromeliads were evaporatively enriched at lower altitudes, and showed a progressive depletion with altitude in the rainy season, consistent with an increasing role for water vapour uptake, particularly in humid cloudforest habitats (Fig. 3b, c). These shifts are consistent with those seen experimentally with a range of C₃ species (Lehmann et al. 2020), as well as cloudforest bryophyte communities, where water vapour was also shown to dominate hydrological budgets (Horwath et al. 2019).

The night-time stomatal opening associated with the CAM pathway should allow for net water vapour uptake to dominate δ¹⁸O_lw signals, although we expected to see contrasting responses between rainy and dry seasons. The data for the range of CAM species found at lower altitude sites were again consistent with these predictions, with significantly lower δ¹⁸O_lw values when compared with sympatric C₃ species, and the general predictions of water vapour uptake seen previously for CAM plants (Helliker and Griffiths 2007; Reyes-Garcia et al. 2008; Helliker 2011, 2014; Lehmann et al. 2020).

The δ¹⁸O_lw signal was highly correlated with the tank water δ¹⁸O (Fig. 4) and isotopic signatures of tank water reflected evaporative enrichment relative to precipitation under more arid, lowland conditions (Fig. 3a–d). At high altitude, the δ¹⁸O of precipitation inputs tended to be similar to regularly replenished tank water ¹⁸O in both wet and dry seasons (Fig. 3a, b). Integration of leaf water signal into organic matter occurs over time (Barbour 2007; Cernusak et al. 2016), and thus δ¹⁸O_lw represents a more instantaneous marker for coupling of water budgets relative to the organic matter ¹⁸O signal (Fig. 3e), although the sampling in this study did not show systematic shifts in the δ¹⁸O_OM for bromeliad tissues, relative to those seen in other studies (Cernusak et al. 2016; Horwath et al. 2019).

The study then focussed on two sympatric species, T. fasciculata (CAM) and W. sanguinolenta (C₃), which are widely distributed across the lower altitudes of the transect (from sea level to 600 m) (Pierce et al. 2002a; Zotz et al. 2005). Overall, the variation in δ¹⁸O_lw was highly responsive to seasonality, precipitation inputs and ambient humidity (Table 1; Figs 3a–d, 4) and, by inference, stomatal sensitivity during gas exchange (Lange and Medina 1979; Griffiths et al. 1986). The experimental manipulation of both species in a humidity chamber mesocosm provided additional evidence for the interplay between water vapour inputs, recharge by precipitation, the holding capacity of the bromeliad ‘tank’, and the regulation of net leaf water use.

The experimental component of our study validated the notion that under high humidity, the gross diffusive exchanges of water vapour can dominate leaf water isotopic signatures (Farquhar and Cernusak 2005; Seibt et al. 2006; Helliker and Griffiths 2007; Seibt et al. 2007; Reyes-García et al. 2008; Helliker 2011, 2014; Lehmann et al. 2020). Understanding NSS controls on the isotopic signature of plant water and tissue organic material is important during gas exchange by certain species (Farquhar and Cernusak 2005; Cernusak et al. 2008; Lai et al. 2008; Cernusak et al. 2016; Benettin et al. 2021). Our simulations with a NSS leaf water model illustrate how the contrasting diurnal patterns of gas exchange affect δ¹⁸O_lw for the two species during the humidity chamber experiments.

For the initial experimental period, despite the distinct stomatal opening patterns of C₃ and CAM, the δ¹⁸O_lw became progressively depleted under saturating atmospheric water vapour by day and night. In contrast, for the final period of low RH during the day, which more closely mimics field conditions at lower altitudes, the daytime rates of C₃ gas exchange coincided with low RH, resulting in evaporative leaf water enrichment for the well-watered plants (Fig. 5a). As a consequence of the high leaf water content, low stomatal conductance and low leaf hydraulic conductance representative of tank bromeliads (North et al. 2013; Males 2016; Males and Griffiths 2018), both species in our study had very low rates of leaf water turnover and thus their δ¹⁸O_lw values changed relatively slowly over a time course of days (Helliker and Griffiths 2007; Cernusak et al. 2008; Lehmann et al. 2020).

Alternatively, under droughted conditions, the sensitivity of bromeliad stomata to humidity (Lange and Medina 1979; Griffiths et al. 1986) resulted in reduced water vapour exchange and reduced evaporative enrichment in the C₃ δ¹⁸O_lw measured empirically and modelled (Fig. 5b). The NSS model also utilised bromeliad tank water as an external water source for leaves similar to soil water in rooted plants. Thus, the gradual evaporative enrichment of tank water over time, seen in both the laboratory mesocosm on a daily basis, and under seasonal conditions in the field (Figs 3, 4) represents an important control on δ¹⁸O_lw and organic material in tank bromeliads.

The manipulative component of our study showed that at high RH, leaf water δ¹⁸O signals of both species were dominated by water vapour exchange, even as the water status of the plants, measured as relative water content (RWC), declined (Fig. 4; see ‘Materials and methods’ for experimental details). Our result is consistent with previous observations demonstrating that δ¹⁸O_lw signals in atmospheric CAM bromeliads were dominated by water vapour when exposed to high RH (Helliker and Griffiths 2007; Reyes-García et al. 2008; Helliker 2011, 2014; Lehmann et al. 2020).

The control of water vapour on δ¹⁸O_lw under high humidity conditions did not lead to significant shifts in organic material isotopic signatures in the samples analysed under field conditions. The lower δ¹⁸O_lw values for C₃ bromeliads in the high altitude cloudforest sites, relative to lowland sites, were consistent with the more negative δ¹³C values found in organic material of those species (Fig. 3f), also consistent with higher stomatal conductances (Seibt et al. 2008). The comparison of sympatric C₃ and CAM species across the lower altitudinal gradient (Table 1) showed that T. fasciculata (CAM) performing gas exchange under high RH at night consistently showed depleted δ¹⁸O_lw. In contrast, when W. sanguinolenta (C₃) was exposed to lower RH during gas exchange and δ¹⁸O_lw was evaporatively enriched. The mesocosm manipulations showed that C₃ species are as susceptible to the influx of water vapour as CAM species, which was supported by the field data for bromeliads from the high elevation cloud forest sites, which has been confirmed by the recent manipulations of C₃ and CAM species under saturating water vapour (Lehmann et al. 2020). Evidence of water vapour as a main control on δ¹⁸O_lw for Inga sp., a C₃ understory tree, has been reported for the Amazon rainforest (Lai et al. 2008). Furthermore, demonstration of nocturnal transpiration and stomatal conductance across different biomes (Dawson et al. 2007; Seibt et al. 2007) indicate that the contribution of water vapour might be significant to leaf water at a global scale, particularly when RH is high, in addition to liquid water uptake by wet leaves (Pierce et al. 2001; Dawson and Goldsmith 2018; Berry et al. 2019; Leroy et al. 2019).

Conversely, the relatively small fluctuations observed in δ¹⁸O_lw during water stress and low RH in the mesocosm (Fig. 4) reflects the overriding control of stomatal conductance, as stomatal sensitivity to low RH and drought has been consistently shown for bromeliads (Fig. 5, see also Lange and Medina 1979; Griffiths et al. 1986; Zotz and Andrade 1998). The temporal separation of gas exchange for C₃ and CAM species is also reflected in the differential signatures for the bromeliads growing along the altitudinal gradient (Fig. 3), where C₃ species consistently showed higher δ¹⁸O_lw in comparison with sympatric CAM species. Our findings suggest that δ¹⁸O_lw in C₃ bromeliads provide clear markers of local climatic conditions of humidity and water sources.

The results of this study highlight the sensitivity of epiphytic bromeliads to drought, water stress and changes in local relative humidity, and offers information on the life form and metabolic pathway variation in δ¹⁸O_lw for plants exposed to common environmental conditions. The δ¹⁸O_lw signal in bromeliad epiphytes can be used as a proxy to understand shifts in the water use of the species which may result from deforestation and its effect on microclimate, as it diminishes rain and fog inputs (Gómez González et al. 2017; Horwath et al. 2019) and increases canopy openness that in turn increases light incidence and evaporative demand (Williams et al. 2020). This type of monitoring is also relevant as temperatures increase and precipitation patterns are affected by global warming (Horwath et al. 2019). Shifts in δ¹⁸O_lw and δ¹⁸O_OM were found to reflect water vapour inputs for the atmospheric CAM T. usneoides, as precipitation δ¹⁸O varied along a latitudinal gradient (Helliker 2014).

We explored an altitudinal gradient and did not find a clear trend of higher enrichment organic material at the lower, more open, drier forests, compared with the more moist, upper montane forests. This lack of a correlation may be explained because most of the sites were secondary forests and pastures, with reduced canopy cover below 1000 m (Fig. 2) or differing growing seasons for C₃ and CAM bromeliads. Significant within-site variation was possibly introduced by the different species sampled, which have different tank capacities, plant size, succulence, trichome cover and other ecophysiological traits that influence water use (Pierce et al. 2002a, 2002b; North et al. 2013; Males 2016; Males and Griffiths 2018).

Finally, we also note that the overall distribution of CAM bromeliads across the altitudinal transect is consistent with the predicted capacity to cope with semiarid habitats (Osmond 1978) as seen in lowland habitats on both sides of the continental divide in Panama (Fig. 2) and other tropical coastal locations (Griffiths and Smith 1983; Griffiths et al. 1986). Although sadly there were no CAM bromeliads found in the cloud forest habitat at Fortuna (in contrast to that seen in eastern Panama, Pierce et al. 2002b), the combined use of stable isotopes of oxygen and carbon have helped to de-mistify bromeliads, which continue to set CAM as a curiosity in context (Osmond 1978).

The authors declare no conflicts of interest.

Monica Mejía-Chang was a recipient of a British Council Chevening Fellowship, a Short-term Fellowship from the Smithsonian Tropical Research Institute and a PhD scholarship from the government of Panama through the Senacyt-IFARHU 2005–2010 program. Casandra Reyes-García received a traveling grant from Fondo Sectorial SEP-CONACYT 221490. Jessica Royles was supported by NERC NE/M001946/1.