Water vapour isotopic exchange by epiphytic bromeliads in tropical dry forests reflects niche differentiation and climatic signals



    Corresponding author
    1. Physiological Ecology Group, Department of Plant Sciences, University of Cambridge, Cambridge, CB2 3EA, UK and
    2. Unidad de Recursos Naturales, Centro de Investigación Científica de Yucatán, Calle 43 Num. 130 Churburná de Hidalgo, Mérida, 97200, México
    Search for more papers by this author

    1. Physiological Ecology Group, Department of Plant Sciences, University of Cambridge, Cambridge, CB2 3EA, UK and
    Search for more papers by this author

    1. Physiological Ecology Group, Department of Plant Sciences, University of Cambridge, Cambridge, CB2 3EA, UK and
    Search for more papers by this author

    1. Physiological Ecology Group, Department of Plant Sciences, University of Cambridge, Cambridge, CB2 3EA, UK and
    Search for more papers by this author

C. Reyes-García. Fax: +52 9999813900; e-mail: creyes@cicy.mx


The 18O signals in leaf water (δ18Olw) and organic material were dominated by atmospheric water vapour 18O signals (δ18Ovap) in tank and atmospheric life forms of epiphytic bromeliads with crassulacean acid metabolism (CAM), from a seasonally dry forest in Mexico. Under field conditions, the mean δ18Olw for all species was constant during the course of the day and systematically increased from wet to dry seasons (from 0 to +6‰), when relative water content (RWC) diminished from 70 to 30%. In the greenhouse, progressive enrichment from base to leaf tip was observed at low night-time humidity; under high humidity, the leaf tip equilibrated faster with δ18Ovap than the other leaf sections. Laboratory manipulations using an isotopically depleted water source showed that δ18Ovap was more rapidly incorporated than liquid water. Our data were consistent with a Craig–Gordon (C-G) model as modified by Helliker and Griffiths predicting that the influx and exchange of δ18Ovap control δ18Olw in certain epiphytic life forms, despite progressive tissue water loss. We use δ18Olw signals to define water-use strategies for the coexisting species which are consistent with habitat preference under natural conditions and life form. Bulk organic matter (δ18Oorg) is used to predict the δ18Ovap signal at the time of leaf expansion.


Understanding environmental and physiological constraints to the δ18O of bulk leaf water and bulk organic material (δ18Olw and δ18Oorg, respectively) offers important insights for plant ecophysiology, climatic reconstruction and the global hydrological cycle. Most studies have focused on C3 or C4 species, and few reports document plants showing crassulacean acid metabolism (CAM; but see Sternberg, DeNiro & Johnson 1986a; Tissue, Yakir & Nobel 1991), a group that can dominate arid environments. There are also few studies dealing with seasonal changes in habitats with pronounced drought stress (Cooper & DeNiro 1989; Pendall, Williams & Leavitt 2005). In this paper, we provide field and greenhouse data that support the implications of a recent model which demonstrated the importance of water vapour exchanges at high humidity, and suggested that organic material from CAM epiphytes acts as an indicator of the mean δ18O signal of water vapour during leaf expansion (δ18Ovap; Helliker & Griffiths 2007).

Leaf water isotopic enrichment, with respect to source (or stem) water, is influenced by evaporation, as described by Craig & Gordon (1965; for review, see Farquhar, Cernusak & Barnes 2007). Dependent on temperature and humidity, substantial 18O enrichment of leaf water occurs during transpiration (Dongmann et al. 1974). Determinants of the leaf water signal are, firstly, the isotopic composition of source water and the time taken for the leaf to reach isotopic steady state (e.g. when transpired water, δt, has the same δ18O as incoming stem water, δs: Yakir 1992; Harwood et al. 1998, 1999). Secondly, leaf temperature and humidity control the extent of evaporative enrichment, which in turn reflects the evaporation rate and the impact of evaporative cooling on δ18O fractionation factors (Barbour et al. 2000; Barbour & Farquhar 2004). Thirdly, leaf water δ18O may be modified by the Péclet effect, describing the retro-diffusion of enriched water away from the evaporative site back towards the veins (Farquhar & Lloyd 1993; Gan et al. 2003), which, in the case of a monocot with parallel leaf veins, results in a progressive enrichment from the base to the tip of the leaf (Helliker & Ehleringer 2000, 2002a; Barnes, Farquhar & Gan 2004; Ogée et al. 2007). Finally, the gross diffusive exchanges of water vapour may exceed net transpirational flux by an order of magnitude at high humidity (Farquhar & Cernusak 2005; Seibt et al. 2006; Helliker & Griffiths 2007). It has recently been hypothesized that such fluxes may lead to the dominance of the isotopic signal of water vapour on 18Olw signal, particularly for CAM plants, which primarily open their stomata at night (Helliker & Griffiths 2007).

Ultimately, one aim was to improve the understanding of the determinants of the 18O signal in cellulose (δ18Oc) and bulk organic material (δ18Oorg), in leaves or as stored in tree rings, for use in monitoring current climate change processes or for reconstructing past climatic conditions (Epstein, Thompson & Yapp 1977; Burk & Stuiver 1981; Yapp & Epstein 1982; Yakir & DeNiro 1990; Aucour, Hillaire-Marcel & Bonnefille 1996; Szczepanek et al. 2006). The δ18Oc and δ18Oorg reflect δ18Olw at the site of synthesis, adding a 27‰ enrichment during biosynthesis of cellulose, corresponding to a 24–20‰ enrichment when whole organic matter is measured (Gray & Thompson 1976; Borella et al. 1998; Barbour, Andrews & Farquhar 2001; Ferrio & Voltas 2005; Cullen & Grierson 2006; Sternberg et al. 2006). Many current studies reconstruct temperature, humidity and rain δ18O signatures from organic matter, assuming that atmospheric vapour δ18O (δ18Oa) is equilibrated with nearby oceanic or precipitation signals (Helliker & Griffiths 2007). It is difficult to sample atmospheric water vapour continuously at remote locations in the field (Helliker et al. 2002), and wind currents determining inland or oceanic signals are not easily predicted (Masson-Delmotte et al. 2005; Danis et al. 2006). Further errors in climatic reconstruction can be introduced by the physiology of the particular trees sampled (Ferrio & Voltas 2005).

Thus, there are considerable morphological and environmental determinants of δ18O at the leaf level, and in this paper, we explore the implications for vascular epiphytes (monocots), an intriguing model system wherein roots are primarily used as holdfasts, and precipitation inputs are often only transiently available (Griffiths 1989). Epiphytes display a variety of water-use strategies dependent on climatic gradients within the forest canopy and between forest formations (Griffiths & Smith 1983; Griffiths et al. 1986; Smith, Griffiths & Lüttge 1986). For bromeliads, these strategies can be related to life form, from ‘tank’ species, which have a reservoir of water stored between overlapping leaf bases, to the ‘atmospheric’ species, in drier habitats, which rehydrate by absorbing liquid water from precipitation (or dewfall and fog) events (Pittendrigh 1948; Martin 1994; Griffiths & Maxwell 1999; Reyes-García et al. 2008). Water use is also dictated by differences in the form, abundance and distribution of water-absorbing epidermal trichomes along the leaf (Martin 1994; Pierce et al. 2001; Benz and Martin 2006).

Helliker & Griffiths (2007) have recently developed a model to show that epiphytes with the CAM pathway may have a leaf water 18O isotopic signature dominated by atmospheric water vapour, based on work with the atmospheric bromeliad Tillandsia usneoides (Spanish moss). In the present study, we report seasonal and diurnal changes of 18O signals under field conditions for four epiphytic species, each with differences in water-use strategies and microclimatic requirements, from a coastal dry forest at Chamela, Mexico (Reyes-García et al. 2008). The extent of leaf water enrichment varied in these species, and we set out to analyse the contribution from changing seasonal precipitation, fog and dew formation, and water vapour exchange on the δ18Olw. Manipulations under greenhouse conditions, together with the use of natural abundance labelling (fogging) experiments, were used to explain results from the field. We then apply the model developed by Helliker & Griffiths (2007) to account for the differing patterns of enrichment for the four species in relation to their water-use strategies, niche differentiation and climatic influences in a seasonally dry forest.


The Craig–Gordon (C-G) model describes the effect of temperature and humidity on an evaporating lake or leaf, as the lighter 16O molecules tend to preferentially evaporate, leaving behind a body of water enriched in 18O relative to the inflowing water (Craig & Gordon 1965; Dongmann et al. 1974; Flanagan, Comstock & Ehleringer 1991):


where R is defined as 18O/16O; the subscripts L, s and a refer to the isotopic composition of leaf water, source water and atmospheric water vapour; wa/wi is the ratio of ambient to leaf intercellular vapour mole fraction; α* is the equilibrium fractionation factor given by the difference in Rliquid/Rvapour and is dependent on temperature; αK is the ratio of molecular diffusivity of H216O to H218O through air, boundary layer and stomata (Merlivat 1978; Farquhar et al. 1989; Flanagan et al. 1991; Farquhar & Lloyd 1993; Cappa et al. 2003; Farquhar & Cernusak 2005).

The model has then been modified by Helliker & Griffiths (2007) to account for epiphytes that lack an absorbent root system to replenish leaf water during evaporation. It is designed for an atmospheric epiphyte (most of these being succulent) and is equivalent to the evaporative changes that would occur from a glass of water:


where W represents leaf water volume (mol m−2), and g represents stomatal conductance. At high humidity, the first term in the equation is minimal and the second term dominates the isotopic signal of the leaf. Thus, water vapour exchanging across leaf surfaces increasingly dominates the isotopic signal of leaf water, when water vapour entering the leaf is quantitatively much greater than the net evaporative flux and RL = Ra · α*.

For testing Eqn 2, the initial δ18Olw at maximal relative water content (RWC) was derived for fully hydrated plants from measured data (in field or greenhouse), and the model was run under simulated conditions which allowed a constant, progressive change in RWC during dehydration (with a theoretical value of g, manipulated to match incremental changes in W with the actual RWC observed during the dehydration cycle). The equation was solved for RL iteratively, using the solver function of Excel.

The model was applied to δ18Olw data of the atmospheric Tillandsia intermedia, which showed a progressive enrichment under greenhouse conditions. All measurable parameters used in Eqn 2 were derived experimentally, and then were compared to extreme values of δ18Ovap and relative humidity (RH) to test the sensitivity of the model. When solving RL for the field data, the model was also used to test a range of potential values of δ18Ovap (−9 to −12‰), temperature (20–30C) and RH (75–100%).


Study site and plant material

Field studies were made in a tropical dry deciduous forest (Rzedowsky 1978) at the Biological Station of Chamela, near the Pacific Coast of Mexico at 19°30′N, 105°03′W. The locality has a marked rainy season from July to October when nearly 80% of the annual precipitation occurs (Bullock 1986), which we define in this paper as the ‘wet’ season. There is a subsequent dry season with a period of regular fog and dew formation (December–February: dry + dew), followed by a period of lower atmospheric water content (March–June: dry season). Average annual rainfall is 746 mm (1977–2006).

During the years of the field study (2002–2003), RH in Chamela remained high at dusk, fluctuating from 92 to 77% from the wet to the dry season. Daytime RH was 85% during the wet season, and dropped to 33% in the dry season. Day temperatures fluctuated from season to season, with maximum values of 29–45 °C in the wet and dry season, respectively. Contrastingly, temperature at night was stable throughout the year (26–27 °C).

Plant material was identified according to Espejo-Serna et al. (2004). Five of 10 species found at this habitat were analysed for seasonal changes in RWC, δ18Olw and δ18Oorg. The species were representative of the tank and atmospheric life forms found in this forest (Reyes-García et al. 2008). These life forms represent a gradation of water-use strategies, from Tillandsia makoyana (Baker), with more extensive water storage than in the shallower tank of Tillandsia rothii (Rauh), to the atmospheric life forms, Tillandsia eistetteri (Ehlers) and Tillandsia intermedia (Mez). For the greenhouse studies, the species had been acclimated for a year in a heated greenhouse at the Cambridge Botanic Garden; supplementary lighting was used to maintain the photoperiod in winter months. T. eistetteri had to be maintained in an incubator above liquid water in order to maintain viability.

Anatomical and RWC measurements

The density of water-absorbing trichomes and stomata were analysed by making leaf impressions of a fully expanded leaf from six different individuals per species. Imprints from abaxial and adaxial faces of the base, middle and tip of each leaf were made using the dental resin Exactoden (Casa IDEA S.A. de C.V., San Luis Potosí, SLP, Mexico), which maintains original cell size. A positive imprint was made using transparent nail polish, and this was examined with an optical microscope BH-2 (Olympus, Center Valley, PA, USA). After the trichome density was determined, the leaves were shaved in order to measure the underlying stomatal distribution.

The seasonal and greenhouse changes in RWC were monitored using the calculation RWC = (fresh weight − dry weight)/(turgid weight − dry weight). Turgid weight was determined after leaf sections were submerged for 24 h in distilled water. Dry weight was determined after 48 h of oven drying at 80 °C.

Seasonal and diurnal changes in δ18Olw in the field

Seasonal changes in δ18Olw were monitored in the field by sampling the base, middle and tip leaf sections of six fully expanded leaves of T. makoyana, T. rothii, T. eistetteri and T. intermedia during the wet and dry seasons and the transitional periods (early wet and dry + dew). Samples of leaf tissue collected in the field were placed in sealed exetainers in situ and then were transported to the UK for isotopic analysis as described further.

During each of the wet and dry + dew seasons, individuals of T. makoyana and T. intermedia were sampled for 1 d at 0800, 1300, 1900 and 2400 h in order to assess circadian changes in isotopic signals in the leaf (additional data for T. eistetteri were not presented). Averages of the six leaves in three leaf sections were presented. For the early wet and dry seasons, and in the greenhouse studies, all samples of bulk leaf water were taken at 1300 h.

Greenhouse manipulations to assess spatial changes in δ18Olw along the leaf

In order to assess the effect of RH and different water sources (rain and fog) on the δ18Olw signal along the leaf of epiphytic bromeliads, controlled experiments were carried out in a glasshouse in the Botanic Garden at Cambridge University. The experiment took place from May to July 2004 with the four species described earlier, using natural daylight conditions. The control treatment refers to the conditions under which the species were grown for a year previous to the experiments in the greenhouse, with daytime temperatures between 20 and 30 °C, night temperature controlled between 19 and 17 °C, and the RH at 75% over both day and night. The plants were sprayed three times a day with local rainwater [δ18O signature −6.43 ± 0.28 SE ‰ versus Standard Mean Ocean Water (SMOW), n = 8]. After the control samples were collected, the plants were subjected to drought until 30% RWC was obtained (after about 3 months), as observed in the field at the end of the dry season. The plants were then placed in a chamber with a humidifier for 15 d (fogging treatment). For each of the treatments, four samples of base, middle and tip leaf sections were collected for bulk water isotope analysis by storage in exetainers. The humidifier (Conair Cool Mist Face Sauna model 3704, Conair, Wigam, Lancashire, UK) used an ultrasonic mechanism and fan to distribute saturated water vapour at ambient temperature. The humidifier was charged daily with 250 mL of local rainwater, and a timer allowed vapour to be generated over a 15 min on/off cycle to reduce condensation and to prevent liquid dropping on to the plants. The cycles were run for 4 h at dawn to simulate natural dew formation in Chamela, as reported by Barradas & Glez-Medellin (1999). Liquid water which had condensed within the system was collected every 15 min for an hour and then was equilibrated for mass spectrometric analysis (see further). Values showed a very small enrichment (0.32 ± 0.36 SE ‰ with respect to water source, n = 4). The system was also tested using snowmelt (−14.22 ± 0.37 SE ‰, n = 10); it showed a slight isotopic enrichment (~1‰) in the water left in the humidifier the morning after the fogging (−12.86 ± 0.48 SE ‰, n = 3). From this, we infer that the humidifier, when used in such discrete periods and when regularly recharged, did not cause a significant fractionation in the isotopic composition of water vapour.

Greenhouse experiment to label δ18Olw signals

The mixing and turnover rates of leaf water pools were examined in the epiphyte T. intermedia, by applying liquid water and water vapour as a label. The source water was derived from melted snow water from northern Sweden (with a δ18O signal of −14‰). Plants that had been maintained under the greenhouse conditions described earlier, watered using rainwater, were then divided into two treatments: one batch was immersed daily for 2–3 min in a container of snowmelt (δ18O = −14‰) for 14 d, and then was drought stressed for 22 d; another batch was placed in the humidifier chamber which generated vapourized source water (δ18O = −14‰) at 15 min intervals throughout the night to maintain an RH close to 100%. No liquid water was applied to individuals under this treatment. Five samples of base, middle and tip leaf sections were collected on days 0, 1, 3, 5, 7, 10 and 14 of the watering or fogging treatments, and on days 0, 3, 5, 7, 10, 14 and 22 of the drought treatment.

A cryogenic trap (Helliker et al. 2002) was set up in the greenhouse and humidification chamber in order to trap atmospheric water vapour for subsequent isotopic analysis. Three samples of vapour were taken throughout the night at the end of each treatment, as described earlier.

Extraction and purification of leaf water for δ18Olw analysis

All the isotopic data were expressed as R relative to the international standard SMOW (ocean water, set arbitrarily to zero), and was given by the notation of δ = [(Rsample/Rstandard) − 1] × 1000, with R being the ratio of 18O/16O of the sample or standard, respectively. Leaf water was extracted for isotopic analysis using vacuum cryogenic distillation (Ehleringer & Osmond 1989). A 200 µL aliquot of leaf water was equilibrated in sealed glass tubes (field samples) or exetainers (greenhouse samples) with 1 mL of pure CO2 at room temperature for a minimum of 3 d (Compston & Epstein 1958). Corrections were made for gas volume and temperature. The sample was then purified on a vacuum line to remove any water vapour by passing the gas through two cryogenic cold traps. The gas sample was then collected in a stopcock vial and was connected to the duel inlet mass spectrometer (VG SIRA 10; modified by ProVac Ltd, Crewe, UK) for analysis.

δ18Oorg analysis

The δ18Oorg was quantified for the species using the dried samples collected for bulk leaf water. Samples of 3 leaves per species were ground and sent for analysis by Dr C. Keitel, Environmental Physiology Group, Research School of Biological Sciences, Australian National University, Canberra, Australia, following the protocol of Farquhar, Henry & Styles (1997).

Statistical analyses

Statistical analysis was performed using the statistical package STATISTICA (StatSoft, Incorporated, Tulsa, OK, USA) for two-way analysis of variance (anova) to assess the differences in isotopic signals under the different species and seasons. The same analysis was used for changes in isotopic signature or RWC under greenhouse conditions to test the effect of days of treatment and leaf sections. Individual differences among species, seasons or days were tested using Tukey tests. Significant differences were reported when P ≤ 0.05. Regressions were used to assess the relationship between distance from leaf base and trichome and stomata densities.


Seasonal and diurnal changes in δ18Olw in the field

The seasonal progression at Chamela represents the transition from rainy period (early wet: July, wet: August–October), a dry season when dewfall predominates (dry + dew: November–February) extended by an additional period when low atmospheric water content and drought stress predominates (dry season: March–June). RWC was initially 70% in the early wet and wet seasons, decreasing in the atmospherics during the dry + dew period (50–62%). All species were close to 30% RWC in the dry season (Table 1, season effect, two-way anova, P ≤ 0.001).

Table 1.  Seasonal changes in relative water content (RWC) (%) for bromeliad species under field and greenhouse conditions
SpeciesRWC (%)
Chamela, MexicoGreenhouse
WetDry + dewDryEarly wetControlFogging
  1. Samples were taken in the field for the dry, wet season and the transitional periods (early wet and dry + dew). Greenhouse control conditions correspond to well-watered plants, and fogging conditions represent droughted plants with a night-time fogging device and no additional water. Means of three leaves are shown with their SEs.

Tillandsia eistetteri65 ± 350 ± 829 ± 268 ± 577 ± 435 ± 7
Tillandsia intermedia71 ± 562 ± 224 ± 271 ± 355 ± 639 ± 4
Tillandsia rothii71 ± 670 ± 329 ± 575 ± 280 ± 332 ± 4
Tillandsia makoyana74 ± 567 ± 437 ± 666 ± 385 ± 341 ± 9

Diurnal patterns in δ18Olw under field conditions were analysed during two contrasting seasons (wet and dry + dew), with data for two representative species, T. intermedia and T. makoyana, presented in Fig. 1. There were no significant changes in δ18Olw during the day for either species, although values were generally more enriched in the dry season.

Figure 1.

Field measurements of diel leaf water δ18O (δ18Olw) (‰ versus SMOW) in the atmospheric species Tillandsia intermedia and the tank species Tillandsia makoyana. Samples from six leaves were taken during the wet and dry + dew season at 0800, 1300, 1900 and 2400 h. Means (including base, middle and tip) and SEs are shown.

Bulk leaf water of entire leaves from four representative epiphytic species under field conditions (Fig. 2) showed a seasonal progression in the δ18Olw enrichment in all species (season effect, two-way anova, P ≤ 0.001), with depleted values in the early wet and wet season (a mean of +0.14 and +0.72‰, respectively) and enriched values in the dry + dew and dry seasons (a mean of +3.34 and +4.17‰, respectively). The seasonal fluctuations were higher in the atmospheric species (T. eistetteri and T. intermedia) and in the shallow tank species (T. rothii), and were less marked in the deep tank species (T. makoyana).

Figure 2.

Seasonal changes in leaf water δ18O (δ18Olw) (‰ versus SMOW) of the atmospheric species Tillandsia eistetteri and Tillandsia intermedia and the tank species Tillandsia rothii and Tillandsia makoyana. Samples were taken in the field for the dry and wet season and the transitional periods (early wet and dry + dew). Means of six leaves (including base, middle and tip measurements of each leaf) are shown with their SEs.

Spatial changes in δ18Olw and anatomical traits along the leaf

Under field conditions, no differences in δ18Olw were found between the base, middle and tip leaf sections of the bromeliad leaves in any of the seasons, as shown for the two representative species, T. intermedia and T. makoyana, even though a trend towards a reduction in δ18Olw towards the tip was observed in the dry + dew season (Fig. 3a,c). Under greenhouse conditions with night-time RH below saturation (75%, control treatment), the tank species, T. makoyana, showed the typical enrichment of 8‰ from the base to leaf tip of a monocot (Fig. 3d), which was not shown in the atmospheric species T. intermedia (Fig. 3b). When the plants were drought stressed, but were supplied with vapourized local rainwater at night for 15 d (fogging treatment), both species showed a decrease in δ18Olw from the leaf base to the leaf tip, which was statistically significant in T. intermedia.

Figure 3.

Base, middle and tip of the leaf leaf water δ18O (δ18Olw) of two representative atmospheric and tank bromeliads in field and greenhouse conditions. Graphs (a) and (c) show data for the atmospheric Tillandsia intermedia and the tank Tillandsia makoyana, respectively, during the wet and dry + dew seasons. Graphs (c) and (d) show the same species under greenhouse conditions, with low relative humidity (RH) conditions at night but under well-watered conditions (control) or high night-time RH conditions and no liquid water available (fogging). Means of n = 6 for field and n = 4 for greenhouse measurements are shown with their SEs; the asterisks denote significant changes from base to tip (P ≤ 0.05).

Anatomical traits were analysed along the length of leaves in order to compare the interplay between source water inputs and evaporative effects. All species showed a positive correlation between stomatal density and distance from the leaf base to the leaf tip, for example, from 11.8 to 36.7 stomata per square millimetres at the base and tip, respectively, as found for the deep tank species T. makoyana (Table 2, R2 = 0.36, P ≤ 0.01). In contrast, for T. makoyana, there was a significant reduction of trichomes from base to tip (Table 2).

Table 2.  Trichome and stomatal density in leaf sections of bromeliad species from the dry forest of Chamela
SpeciesTrichome density (mm−2)Stomatal density (mm−2)
  1. Means of six leaves are shown with their SEs. Significant regressions of distance from base versus density are marked by asterisks: P ≤ 0.05 (*), P ≤ 0.01 (**).

Tillandsia eistetteri55.9 ± 8.851.3 ± 4.448.1 ± 2.711.2 ± 3.815.4 ± 0.917.7 ± 0.3**
Tillandsia intermedia56.9 ± 5.446.0 ± 7.252.1 ± 2.54.1 ± 1.28.5 ± 1.09.1 ± 2.1**
Tillandsia rothii61.5 ± 0.163.6 ± 0.163.6 ± 2.214.2 ± 0.113.6 ± 0.118.3 ± 1.8*
Tillandsia makoyana47.1 ± 0.847.1 ± 1.839.6 ± 7.4*11.8 ± 7.623.8 ± 3.736.7 ± 7.3**

Greenhouse experiment to label δ18Olw signals

Finally, we investigated the dynamics of water uptake by the atmospheric T. intermedia, by experimentally manipulating water sources: having been maintained under local rainwater, with a δ18O signature of −6‰, the plants were transferred to a water source with a signature of −14‰, which was provided either as liquid water or under high humidity by fogging (Fig. 4). For plants immersed for 2–3 min on a daily basis in the isotopically depleted water, the initial leaf water signal was maintained for 5–7 d before the new source liquid signal led to a reduced isotope ratio in leaf water (Fig. 4a). After 2 weeks of the liquid water treatment, the bulk leaf signal was only depleted by 4.5 and 2.0‰ for base and tip, respectively. There was no difference in δ18Olw or RWC between leaf sections (base, middle and tip: Fig. 4a, Table 3), and on average, RWC was 79% throughout this phase of the experiment (Table 3). Following the change in water source, which was tracked for 2 weeks, a 22 d drought was imposed on these plants in the greenhouse. There was a steep increase in δ18Olw of T. intermedia in response to this drought period (Fig. 4b), from −0.76‰ on day 0 to +8.79‰ on day 22 (two-way anova, day effect, P ≤ 0.001). This change was concomitant with a reduction in RWC from 84 to 70% (Table 3, two-way anova, day effect, P ≤ 0.05). The leaf tip increased δ18Olw more than the base (two-way anova, leaf section effect, P ≤ 0.05), although the extent of water deficit, measured as RWC, was similar across all leaf sections (Table 3).

Figure 4.

Leaf water δ18O (δ18Olw) in the leaf sections (base, middle and tip) of the epiphyte Tillandsia intermedia, in response to a change from −6 to −14‰ in (a) liquid water, (b) subsequent drought and (c) vapourized water generated in a fogging regime at night. The atmospheric vapour δ18O (δ18Ovap) in the greenhouse was −10.7‰ (for first two treatments: a,b) and −19.8‰ in fogging chamber (treatment c). The x-axis is categorical; means of five samples and SEs are shown; asterisks denote statistical differences between base and tip (P ≤ 0.05).

Table 3.  Relative water content (RWC) of Tillandsia intermedia during labelling and drought conditions in the greenhouse
DaysLabel – liquid waterLabel – foggingDrought
  1. Plants were subjected to three different treatments, two wherein source water δ18O was changed from –6 to –14‰ either as liquid or vapourized form, and a third wherein the plants were kept under drought conditions for 22 d.

06477 ± 68573 ± 49187 ± 26477 ± 685
380 ± 480 ± 387 ± 269 ± 665 ± 675 ± 2
580 ± 192 ± 178 ± 1685 ± 988 ± 1080 ± 782 ± 677 ± 286 ± 4
770 ± 476 ± 475 ± 752 ± 549 ± 478 ± 1
1075 ± 670 ± 282 ± 755 ± 877 ± 1774 ± 355 ± 252 ± 378 ± 3
1474 ± 491 ± 188 ± 273 ± 767 ± 374 ± 559 ± 564 ± 373 ± 10
2277 ± 175 ± 1165 ± 3

Under the fogging treatment, in contrast to the liquid water immersion regime, all leaf sections responded immediately to the vapour signal: within 3 d, there had been a depletion of 5–8‰ in the mid and tip sections (Fig. 4c). By day 5, the leaf tips showed the biggest shift, a depletion of 12‰ compared to day 0 (absolute value of −10‰ versus SMOW). The gradient from base to tip was largely maintained for the next 5 d (P ≤ 0.001, two-way anova, effect of leaf section on days 5, 7 and 10). A small change in overall RWC, from 76 to 62%, was found later in the fogging treatment from days 7 to 14 (two-way anova, effect of leaf section on days 7–14, P ≤ 0.001) when there was a higher depletion and higher RWC in the tip as compared with the base of the fogged plants (P ≤ 0.001, effects of day and leaf section).

Model data comparison

We applied Eqn 2 to the data for progressively droughted T. intermedia, shown in Fig. 4b, to compare the predicted and observed δ18Olw under these greenhouse conditions wherein all variables were measured (RH, temperature, source water δ18O, δ18Ovap; Fig. 5a, solid line). The initial δ18Olw (−1.69‰) and RWC (90%, converted to moles of water to replace W in Eqn 2) were those data measured for well-hydrated T. intermedia in the greenhouse. The ambient conditions (RH, 75% and temperature, 17 °C) used to parameterize the model were those measured in the greenhouse during the night-time (for wa/wi in Eqn 2). The measured value of δ18Ovap in the greenhouse was −11‰ (Ra). The value of g was manipulated to allow for gradual tissue water loss (values within the range of those reported for CAM bromeliads, between 0.02 and 0.0015 mol m−2 s−1; Martin 1994), with 12 successive iterations then used to derive values of δ18Olw under decreasing RWC. The modelled data, showed in a solid line, fits well with the measured data, represented by symbols. From the starting δ18Olw values in fully hydrated plants, progressive enrichment occurred as the plants lost water, until an equilibrium point was reached at around 10.5‰, after which progressive decreases in RWC were associated with little change in δ18Olw.

Figure 5.

Greenhouse leaf water δ18O (δ18Olw) values for the epiphyte Tillandsia intermedia during a drought experiment (symbols) and sensitivity tests of the non-steady-state Craig–Gordon (C-G) model (lines) as modified by Helliker & Griffiths (2007). Data points are shown previously in Fig. 4b; see Materials and Methods. Symbols represent means and SEs of five plants, classified by leaf sections (base, middle and tip). Data from the 30% relative water content (RWC) are part of a long-term drought experiment of the same species that was carried out in the same greenhouse. The modelled data were derived from iterations of Eqn 2 under decreasing RWC conditions from a starting point determined by field data. The lines represent models ran under different atmospheric vapour δ18O (δ18Ovap) −14, −11 (actual measured value in the greenhouse) and −8‰; (a) shows models ran using a 75% relative humidity (RH) input and (b) models ran using 90% RH. Asterisks denote the actual greenhouse conditions and also the best fitting model for the measured data.

A sensitivity analysis was performed to compare the effect of varying RH and δ18Ovap on the δ18Olw of a progressively droughted atmospheric bromeliad, and is included in Fig. 5. The model was run under different simulated δ18Ovap signals (−14 and −8‰), keeping all other variables as described earlier. The curves showed similar shapes, but the equilibrium point was reached at lower δ18Olw values (8‰) when the δ18Ovap was −14‰, compared to the δ18Olw (13.5‰) when δ18Ovap was −8‰ (Fig. 5a).

The sensitivity analyses showed that the model was highly responsive to changes in RH and δ18Ovap (Fig. 5). The same parameters (δ18Ovap−14, −11 and −8‰) were run under higher RH conditions (90%, Fig. 5b); the model then generated lower equilibrium point values of 1, 4 and 7‰δ18Olw as δ18Ovap was varied from −14, −11 and −8‰, respectively (Fig. 5b). RH affected not only the value at which δ18Olw reached equilibrium with δ18Ovap, but also the RWC at which this happened. At higher humidity, the equilibrium point δ18Olw values were lower, and occurred at higher RWC. Because starting δ18Olw was shown not to influence the equilibrium point with δ18Ovap by Helliker & Griffiths (2007), it was not tested here. Changes in temperature did not have a significant effect on the predicted response (data not shown).

Predicted values from Eqn 1 (C-G model) and Eqn 2[Helliker and Griffiths model (H-G)] were compared to field data of δ18Olw, with different theoretical inputs of δ18Ovap and nocturnal RH used as variables, using values representative of the contrasting microclimates preferred by each species under field conditions (Fig. 6). The H-G model was highly sensitive to drivers represented by δ18Ovap, RH and RWC. For three of the four species, measured δ18Olw correlated well with the model when the input of δ18Ovap was set to −10‰ (dotted and solid lines). When a value of −12‰ was used (broken line), predicted values were highly depleted compared to the observed δ18Olw. In contrast, the δ18Olw in the deep tank species, T. makoyana, was not affected by changes in RWC, with this behavior described well by the C-G model (shown as the horizontal dash–dot line, Fig. 6d). For the C-G model, the 0.9‰δ18O value of rain at Chamela and the leaf temperature of 26 °C for tank species and 26.4 °C for atmospheric species (Reyes-García et al. 2008) determined the equilibrium values predicting δ18Olw shown in each panel of Fig. 6.

Figure 6.

Field data of leaf water δ18O (δ18Olw) as a function of relative water content (RWC) compared to different predictions of the steady-state Craig–Gordon model (C-G) and the Helliker & Griffiths (H-G) model (Helliker & Griffiths 2007), for four epiphytic species growing in the seasonally dry forest of Chamela. The data points represent the averages of field data collected at different times of the year (data from Fig. 2 subdivided into sun and shade populations); lines represent four different predictions of the H-G and C-G models using different humidity and δ18O of atmospheric vapour inputs. The mean value of bulk organic matter δ18O (δ18Oorg) and SE for each species are shown (n = 3). RH, relative humidity.

The overall response of atmospheric and shallow tank species, measured as δ18Olw data from the field (Fig. 6a–c), was similar to T. intermedia in the greenhouse (Fig. 5), and the response was more pronounced for the high-humidity dependant, atmospheric species T. eistetteri (Fig. 6a). T. eistetteri data had the best fit to the H-G model when RH was set at a constant 90% (solid line), while T. intermedia (Fig. 6b) and T. rothii (Fig. 6c) had the best fit when humidity was adjusted (dotted line), from 96% (when the plants were well hydrated) to 85% (as the plants became drought stressed and RWC was less than 55%). This shift in RH was consistent with data measured in the field (Reyes-García & Griffiths 2008). The δ18Oorg values, included in Fig. 6, were higher in the two tank species, with values of 23.0–24.1‰, as opposed to values of 20.9–21.7‰ in the two atmospheric species (P ≤ 0.001). Organic values were very consistent, with no statistical differences found in the organic matter of leaf sections for exposed or shaded individuals (data not shown).


In an earlier laboratory study, it had been shown that water vapour influx at high external RH was the major determinant of the δ18Olw signal for T. usneoides, a rootless, extreme atmospheric bromeliad (Helliker & Griffiths 2007). By integrating measurements under natural conditions in the field with those undertaken in a temperate greenhouse, we have now been able to explore the extent of variations in isotopic enrichment in the 18O signal of leaf water for CAM bromeliads from a seasonally dry forest in Mexico. We anticipated that the δ18Olw would be damped to some extent by the capacity to store water in the bromeliad tank formed by overlapping leaf bases, or by more intermittent precipitation or dewfall available to the atmospheric life forms. However, the isotopic signal of leaf water observed in the studied species was also largely controlled by water vapour inputs as was the case for T. usneoides, because stomata of CAM plants open primarily at night when RH is normally high in the field (Helliker & Griffiths 2007).

Water use by epiphytes is equivalent to a model represented by an evaporating glass of water, tempered at high humidity by extremely high rates of water vapour diffusive exchange. Thus, the influx of water vapour changes the isotope composition of the residual pool of water: even though the overall volume diminishes, evaporative enrichment is constrained by the water vapour input signal by a predictable extent as RH increases over 50%. Thus, the H-G model (Helliker & Griffiths 2007) showed a remarkably good fit to measured responses on atmospheric epiphytes, both under natural conditions, and when the isotopic signal of source liquid or vapourized water was manipulated artificially. We now explore the implications of these findings for the interpretation of leaf water enrichment, and derivation of bulk organic 18O signals, for epiphytic life forms with contrasting habitat preferences under seasonally changeable conditions.

δ18Ovap as a driver of δ18Olw in CAM plants

In C3 species at low humidity, transpiration causes a daytime enrichment of about 20‰ in δ18Olw; the enrichment is gradually lost in the evening as depleted soil water continues to enter the leaf, but transpiration is reduced under higher RH (Cernusak, Pate & Farquhar 2002; Farquhar & Cernusak 2005; Seibt et al. 2006; Farquhar et al. 2007). For CAM plants, the diel cycle of δ18Olw enrichment is not always present as gas exchange occurs at night, when RH is high, limiting the extent of evaporative enrichment (Sternberg et al. 1986a), with vapour exchange actually leading to isotopic depletion (Helliker & Griffiths 2007). This is consistent with the lack of night-time enrichment observed in the field data of the present study (Fig. 1), and there is a tendency towards depletion in δ18Olw in the atmospheric species at dawn when humidity is higher, and dew and fog are frequent (Fig. 1; Barradas & Glez-Medellin 1999; Reyes-García et al. 2008). These forms of precipitation could be absorbed through the water-absorbing trichomes present in the epiphytic bromeliads or through the stomata, as has been suggested for some non-epiphytic plants from semi-arid environments (Noy-Meir 1974; Alessio et al. 2004; Kidron 2005). Additionally, lower fluctuations in δ18Olw may be expected in plants showing high specific leaf water content (Pendall et al. 2005). Nevertheless, a small night-time δ18Olw enrichment can occur, as was observed in three CAM species growing under 15% night-time RH (Tissue et al. 1991).

The extent of δ18Olw enrichment for CAM plants was also limited on a seasonal basis (Fig. 2). In semi-arid lands, Pendall et al. (2005) found a shift of about 15‰δ18Olw in two C3Pinus species between dry and wet seasons. The increase in δ18Olw during the drought periods corresponded to lower RH and was inversely related to stomatal conductance. In the present study, higher changes in RWC (from 70 to 30%, Table 1), indicative of the water lost through transpiration during a 5 month drought, in the four CAM bromeliad species, led to an enrichment in δ18Olw of only 3–6‰ (Fig. 2). These differences in enrichment between C3 and CAM species derive from the lower daytime RH (9%) during gas exchange in the Pinus species in comparison with the bromeliads, and the likely sensitivity of bromeliad stomata to low RH which would also reduce water vapour loss and exchange (Lange & Medina 1979; Griffiths et al. 1986).

Because of the dominant effect of δ18Ovap on δ18Olw, the typical enrichment from base to leaf tip which has been observed in C4 monocots (up to 49‰ difference in a grass leaf blade; Helliker & Ehleringer 2000) was not generally observed in these bromeliads (Fig. 3a,c). In contrast, under field conditions, all leaf sections were in equilibrium with δ18Ovap. In C4 monocots, the progressive enrichment is caused by the back diffusion of enriched water within the leaf blade, which mixes with inflowing vein water, and is swept on towards the leaf tip (Helliker & Ehleringer 2000, 2002a; Farquhar & Gan 2003; Barnes et al. 2004; Ogée et al. 2007). Any such enrichment in CAM epiphytes in the field was obscured by water vapour exchanging with leaf water, so no Péclet effect was detectable.

Under greenhouse conditions, with night-time RH below saturation (75%), enrichment from leaf base to leaf tip was observed (Fig. 3b,d). The tank species, which had fewer water-absorbing trichomes and more stomata in the leaf tip (Table 2), showed the typical monocot pattern of enrichment. When the night-time RH was increased, the leaf tips, with higher stomata density, presumably allowed more water vapour to exchange with leaf water lowering the δ18Olw values more rapidly, compared with the leaf base. This effect was also observed in the labelling experiment, when δ18Olw in leaf tips became depleted more immediately than leaf bases under high RH (fogging), and contrasted with the negligible impact when the labelled water was fed by submerging the entire plant in liquid water (Fig. 4).

The isotopic labelling data also supported the model developed from C-G by Helliker & Griffiths (2007), which suggested the use of atmospheric epiphytes as a marker of δ18Ovap. The influence of δ18Ovap has even been observed in C3 species, resulting in night-time depletion of δ18Olw (Seibt et al. 2006; Seibt, Wingate & Berry 2007). But for CAM plants, when translated into gross water vapour exchange, the influx of water vapour is significant, with the entire pool of leaf water in the atmospheric bromeliad T. usneoides effectively turning over within two to three nights at 95% RH (Helliker & Griffiths 2007). Thus, the δ18Olw is highly dependent on δ18Ovap, explaining the rapid exchange with the depleted (‘labelled’) source water (Fig. 4), which did not increase the RWC of T. intermedia (Table 3).

Observed and modelled δ18Olw: niche differentiation in epiphytes

The sensitivity analyses based on the H-G model (Helliker & Griffiths 2007) confirmed that when gas exchange occurs at high nocturnal RH and high RWC, the equilibrium points for δ18Olw more clearly predict δ18Ovap. This model could be parameterized to match leaf water content and isotopic composition as drought stress developed in the seasonal forest for three of four species tested (Fig. 6). The species coexist in the seasonally dry forest of Chamela, and were exposed to the same δ18Ovap signal. Differences in δ18Olw, as each species lost water during the dry months, were presumably the result of specific micro-environmental humidity conditions experienced by each species. Thus, we can use these signals to define the niche occupied by each species.

The species coexist by occupying different strata of the canopy, with T. intermedia found lower in the canopy (4.5 m) and the other three species found in the upper canopy (6.0 m, Reyes-García et al. 2008). The upper canopy is more humid at night than the lower canopy, with increased dew formation, even throughout the dry season (Barradas & Glez-Medellin 1999; Graham & Andrade 2004; Reyes-García & Griffiths 2008).

The species T. eistetteri, which inhabits the outer branches of the host trees and required a high-humidity regime in the greenhouse, showed a better fit between field and modelled data when RH was set to a constant 90%, consistent with the stable, humid upper canopy environment (solid line, Fig. 6). In contrast, the lower canopy species T. intermedia, showed a better fit for the model with RH changing from 96 to 85% as the plant was progressively droughted (dotted line). More dynamic seasonal changes in RH are found in the lower canopy, as the canopy cover in the wet season maintains high RH, which decreases when trees shed leaves during the dry season because of increasing radiation and temperature in the lower strata (Reyes-García et al. 2008; Reyes-García & Griffiths 2008). The habitat preference of T. intermedia could also account for the higher seasonal changes in δ18Olw compared with coexisting species (Fig. 2).

The differences in the modelled point of inflexion are indicative of the RH during gas exchange, and could be the result of two factors: (1) differences in RH in the microclimate, and (2) stomata reacting to changes in RH and closing under low RH, as has been described for some bromeliads (Lange & Medina 1979; Griffiths et al. 1986). Further studies would be needed to discern the interplay between the two factors mentioned. Reyes-García et al. (2008) found that the magnitude of CAM activity is extended by fog interception for the atmospherics into the dry season, and is also related to niche (Reyes-García & Griffiths 2008).

For the tank species, the extent of leaf water enrichment and responsiveness to seasonal climatic changes were also reflected in growth form, with T. makoyana forming a deeper tank than the more open rosette of T. rothii. Thus, for T. makoyana, the water stored in the tank could be the reason for the lack of correlation between RWC and δ18Olw (Fig. 6d), with the more continuous supply of water allowing the stomata to open at times of lower humidity. In this case, the δ18Olw data fit better the C-G model, and was dependent on the leaf temperature and the 0.9‰δ18O value of precipitation. Additionally, we have also found that bromeliad growth form and tank structure alter the degree of isotopic enrichment in water stored in the tank (Mejia-Chang and Griffiths, unpublished data).

Bromeliads as markers of atmospheric water vapour δ18Ovap

The H-G model (Helliker & Griffiths 2007) is very sensitive to δ18Ovap, with a change of 2‰ in the input value of δ18Ovap altering substantially the value at which δ18Olw reached an equilibrium point (which represents the convergence towards a single value determined by water vapour exchange at high humidity). Thus, field data for three of the species had the best fit under −10‰δ18Ovap (solid and dotted lines), which would imply δ18Ovap to be equilibrated with source water with a signal of 0‰: this value is expected considering that the site is adjacent to the Pacific Ocean and as fog inputs would be locally derived and therefore in equilibrium with SMOW.

The δ18Ovap signal can also be derived from the bulk organic matter of atmospheric bromeliads, providing an integrated signal of the vapour at the time of leaf growth (Helliker & Griffiths 2007). However, enrichment with respect to leaf water during cellulose synthesis should be taken into account (Sternberg & DeNiro 1983; Sternberg, DeNiro & Savidge 1986b; Yakir & DeNiro 1990; Helliker & Ehleringer 2002a,b; Sternberg et al. 2006). If we subtract the expected enrichment values of 20–24‰ for bulk organic matter (Gray & Thompson 1976; Borella et al. 1998; Barbour et al. 2000; Ferrio & Voltas 2005; Cullen & Grierson 2006) from the δ18Oorg values obtained for our four species (21–24‰, Fig. 6), we are left with values close to 0‰, our expected liquid water which also corresponds with the −10‰δ18Ovap signal over the Pacific Ocean (considering a −10‰ depletion as a result of phase change). Uncertainties of 4‰ in the enrichment at bulk organic matter synthesis can be reduced by using purified cellulose, or by calibrating each species, according to anatomical traits (i.e. lignin versus cellulose content, which can have different δ18O values) that can lead to differences in enrichment of bulk organic matter (Barbour et al. 2001; Sternberg et al. 2006). Additionally, other climatic, nutritional and physiological variables could result in changes in the δ18Oorg signal. The similarity in δ18Oorg between exposed and shaded populations and between base, middle and tip leaf sections suggest that the signals were temporally and spatially stable. This contrasts with studies of mistletoes and their hosts (Cernusak, Pate & Farquhar 2004), but the slow growth rate in bromeliads allows the δ18Oorg to be integrated across several seasons, and δ18Olw signal of the basal meristems was more stable than other leaf sections (data not shown).


By collecting water in a tank or having localized water absorption by leaf trichomes, the unique water use and physiological adaptations in the epiphytic habitat make bromeliads an interesting system for study. With the CAM pathway restricted to night-time stomatal opening under high humidity, there was little change in δ18Olw diurnally, or on a seasonal basis, considering the low RWC values attained in the dry season. The progressive enrichment from base to tip typical for a monocot, caused by increasing enrichment of water in leaf veins, is not present in these species under field conditions mainly because of large atmospheric water vapour exchanges at night. The H-G model (Helliker & Griffiths 2007) effectively described δ18Olw values under different experimental and natural conditions. The model allowed RH at the time of gas exchange to be predicted for non-tank species, which lacked a constant water supply and had a more direct relation between vapour and leaf water 18O. This allowed for prediction of RH in relation to the species-specific microhabitats. The use of a water source depleted in 18O proved the quantitative extent of coupling between atmospheric water vapour and leaf water 18O signals from an experimental perspective. In the field, the degree of coupling between 18O signals was dependent on life form, microclimate and vertical zonation within the canopy, tempered by the responses of individual species – reflecting dependence on trichomes for water uptake, and stomatal opening during periods of fog or high atmospheric water content.


We would like to thank the staff of the Biological Station of Chamela, Mexico for assistance and climate data. We are grateful to Dr B. Helliker for his insight and support, while Prof. G. Farquhar, Dr U. Seibt and three anonymous reviewers made helpful comments on the manuscript. We acknowledge M. Palomino and E. Arellano-Torres for field assistance, and B. Sachman for help in exporting the plants to the UK, while A. Goodall and P. Michna cared for the plants at Cambridge. Dr C. Keitel, Environmental Physiology Group, RSBS, Australia National University, undertook organic material isotopic analysis, and Dr N. Betson kindly collected the snowmelt water from Umeå, Sweden. C. Reyes-García received a PhD scholarship from CONACyT Mexico (169748), and additional funding from Cambridge Overseas Trust, Churchill College, and SRE de México.