Water-use efficiency and stable isotope composition were studied in three tropical tree species. Seedlings of Tectona grandis, Swietenia macrophylla and Platymiscium pinnatum were grown at either high or low water supply, and with or without added fertilizer. These three species previously exhibited low, intermediate and high whole-plant water-use efficiency (TE) when grown at high water supply in unfertilized soil. Responses of TE to water and nutrient availability varied among species. The TE was calculated as experiment-long dry matter production divided by cumulative water use. Species-specific offsets were observed in relationships between TE and whole-plant 13C discrimination (Δ13Cp). These offsets could be attributed to a breakdown in the relationship between Δ13Cp and the ratio of intercellular to ambient CO2 partial pressures (ci/ca) in P. pinnatum, and to variation among species in the leaf-to-air vapour pressure difference (v). Thus, a plot of v·TE against ci/ca showed a general relationship among species. Relationships between δ18O of stem dry matter and stomatal conductance ranged from strongly negative for S. macrophylla to no relationship for T. grandis. Results suggest inter-specific variation among tropical tree species in relationships between stable isotope ratios (δ13C and δ18O) and the gas exchange processes thought to affect them.
Ever since photosynthetic organisms began to colonize the land surface nearly 500 million years ago, they have faced an inevitable dilemma: terrestrial plants must expose moist tissues to the atmosphere, risking desiccation and death, to absorb CO2 for photosynthesis. Most terrestrial environments are subject to a limiting supply of water for plant transpiration in at least some part of the annual cycle. Thus, the efficiency with which terrestrial plants exchange water for CO2 could have important implications for plant performance and distribution.
Plant physiologists have applied the term water-use efficiency to describe the rate of CO2 uptake or plant dry matter production for a given rate of plant water loss (Bacon 2004). Water-use efficiency at the leaf level (A/E) can be expressed as the rate of diffusion of CO2 into the leaf during photosynthesis for a given rate of diffusion of water vapour out of the leaf (Farquhar & Richards 1984):
where A is photosynthesis (µmol CO2 m−2 s−1), E is transpiration (mmol H2O m−2 s−1), gc and gs are stomatal conductances to CO2 and water vapour, respectively (mol m−2 s−1), ca and ci are CO2 partial pressures in ambient air and leaf intercellular air spaces, respectively (Pa), ei and ea are water vapour partial pressures in the intercellular air spaces and ambient air, respectively (kPa), v is the leaf-to-air vapour pressure difference (kPa), defined as ei − ea, and 1.6 is the ratio of the diffusivity of CO2 to that of water vapour in air.
Equation (1) can be scaled from the leaf to the whole plant by taking into account respiratory C use and water loss not associated with photosynthesis. Respiratory C use results in a net efflux of CO2 from non-photosynthetic organs during the day and from the whole plant at night. Water loss not associated with photosynthesis can result from soil evaporation and transpiration at night. The whole-plant water-use efficiency, or transpiration efficiency of C gain (TE), can thus be defined as (Farquhar & Richards 1984; Hubick & Farquhar 1989):
where φc is the proportion of C taken up by photosynthesis that is subsequently lost to the atmosphere by respiration, and φw is unproductive water loss as a proportion of water loss associated with photosynthetic C uptake. If soil evaporation is factored out independently, the φw can be approximated as En/Ed, where En is nighttime transpiration and Ed is daytime transpiration. The second form of equation (2) is written as a function of ci/ca, because this term also relates to carbon isotope discrimination (Δ13C). The TE has units of mmol C mol−1 H2O.
For some applications, it is convenient to describe the leaf-to-air vapour pressure difference, v, as a product of the air vapour pressure deficit (D), and a second term, fv (Cernusak et al. 2008b). The fv is a scaling factor relating the magnitude of v to that of D, such that v = Dfv. If leaf temperature is equal to air temperature, fv will be unity, whereas if leaf temperature exceeds air temperature, fv will exceed unity. Separating v into these two components allows equation (2) to be written as
Factoring D from the right side of equation (3) allows comparison of the transpiration efficiency of plants grown under different evaporative conditions (Tanner & Sinclair 1983; Hubick & Farquhar 1989). That is to say, weighting TE by D adjusts for variation in TE that is purely environmental. The term D·TE has units of Pa mol C mol−1 H2O. For averaging purposes, the v in equation (2) and the D in equation (3) should be weighted by gs integrated over the period during which the leaf is illuminated (Farquhar et al. 1989b). However, this would require that the diurnal course of gs be known. In the absence of such information, we have calculated v and D in equations (2) and (3) as averages for daytime hours.
where a is discrimination against 13C during diffusion of CO2 through stomata (4.4‰), b is discrimination against 13C by carboxylating enzymes (∼29‰ for Rubisco), and d is a composite term that summarizes collectively discriminations associated with dissolution of CO2, liquid phase diffusion, photorespiration and dark respiration (Farquhar, Ehleringer & Hubick 1989a). The d was estimated to have a mean value near 3‰ for 15 tropical tree and liana species (Cernusak et al. 2008b), and other estimates in the literature range from approximately 0 to 4‰ (Hubick et al. 1986; Hubick & Farquhar 1989; Hubick 1990; Cernusak et al. 2007b). The Δ13C is defined with respect to CO2 in air as Δ13C = Ra/Rp − 1, where Ra is 13C/12C of CO2 in air and Rp is 13C/12C of plant C.
where δ18Os is δ18O of water taken up by roots from the soil, ε+ and εk are equilibrium and kinetic fractionation factors, respectively, and δ18Ov is δ18O of atmospheric water vapour. The ε+ can be modeled as a function of leaf temperature (Bottinga & Craig 1969), and εk as a function of diffusion resistance partitioning between stomata and boundary layer, with weighting by appropriate fractionation factors (Farquhar et al. 1989b; Cappa et al. 2003). The δ18O of average lamina leaf water (δ18OL) can then be described as (Farquhar & Lloyd 1993; Farquhar & Gan 2003),
The ℘ is a Péclet number, defined as EL/(CD18), where E is transpiration rate (mol m−2 s−1), L is a scaled effective path length (m), C is the molar concentration of water (mol m−3), and D18 is the diffusivity of H218O in water, which can be predicted as a function of leaf temperature (Cuntz et al. 2007). The δ18O of plant organic material (δ18Op) is expected to vary as a function of δ18OL (Barbour & Farquhar 2000):
where pex is the proportion of oxygen atoms that exchange with local water during cellulose synthesis, px is the proportion of un-enriched water at the site of tissue synthesis, εwc is the fractionation between organic oxygen and medium water, and εcp is the δ18O difference between tissue dry matter and the cellulose component. The pexpx, εwc and εcp have been observed to be reasonably constant for stem wood in trees (Cernusak, Farquhar & Pate 2005).
Although tropical trees grow in forests with relatively high annual precipitation, they are generally exposed to intermittent drought, especially as seedlings and saplings, which can influence the distribution and abundance of species in the environment (Engelbrecht & Kursar 2003; Engelbrecht, Kursar & Tyree 2005; Engelbrecht et al. 2007). Water-use efficiency may therefore be an important functional trait for tropical trees, and large variation in whole-plant water-use efficiency has been observed among seedlings of several tropical tree species (Cernusak et al. 2007a, 2008b). These observations were carried out under conditions of approximately uniform water and nutrient availability. However, water-use efficiency is known to vary in response to the availability of both water (Hubick et al. 1986; Zhang & Marshall 1994; Sun et al. 1996) and nutrients (Toft, Anderson & Nowak 1989; Livingston et al. 1999; Cernusak et al. 2007b). A decreased supply of soil water would be expected to cause ci/ca to decrease due to lower gs, whereas a higher nitrogen concentration in leaves would be expected to cause ci/ca to decrease due to higher A. Thus, either situation could lead to an increase in A/E and TE, as shown in equations (1) and (2). Our first objective in the present study was to test whether the ranking of TE previously established for three tropical tree species would be maintained under conditions of variable water and nutrient supply.
In our previous studies, we observed species-specific offsets in the relationship between whole-plant Δ13C (Δ13Cp) and TE (Cernusak et al. 2007a, 2008b), suggesting a possible uncoupling between the two parameters at the species level. As indicated in equations (2) and (4), variation among species in the relationship between TE and Δ13C could result from variable dependence among species of Δ13C on ci/ca, or from variable dependence among species of TE on ci/ca. Separating these two possibilities would provide a valuable insight into the use of Δ13C for inferring variation among species in TE. Our second objective in the present study was to investigate the physiological bases for offsets among species in the relationship between TE and Δ13C.
Equations (5) to (7) suggest that for plants grown in a common environment, δ18Op should decrease as a function of increasing gs. This would be caused by a decrease in ei associated with evaporative cooling of the leaf, a decrease in εk associated with a decrease in the proportion of total diffusive resistance accounted for by stomata, and an increase in the Pèclet number associated with an increase in E. However, recent reports have indicated that, in at least some instances, the relationship between δ18Op and gs or transpiration rate did not match this expectation (Sheshshayee et al. 2005; Cernusak et al. 2007b, 2008a). Thus, our third objective in the present study was to examine relationships between organic material δ18O and gs in three tropical tree species.
Plant material and study site
The three species employed were Platymiscium pinnatum (Jacq.) Dugand (Fabaceae), Swietenia macrophylla King (Meliaceae) and Tectona grandis Linn. f. (Verbenaceae). These three species exhibited highest, intermediate and lowest TE, respectively, of seven tropical tree species examined in a previous study (Cernusak et al. 2007a). Plants were grown at the Santa Cruz Experimental Field Facility of the Smithsonian Tropical Research Institute in Gamboa, Panama (9°07′N, 79°42′W). The altitude at the site is approximately 28 m above sea level. T. grandis and S. macrophylla were grown from seed collected in the Panama Canal watershed. Recently germinated seedlings of P. pinnatum, approximately two to four weeks old, were collected from the Azuero Peninsula, Panama and transplanted to the study site, due to difficulty in locating viable seed for this species. Meteorological conditions observed at the study site over the course of the experiment are presented in Table 1.
Table 1. Mean monthly meteorological conditions for daytime hours through the course of the experiment. Daytime hours were defined as between 0700 and 1730 local time. We focused on daytime hours to characterize the environmental conditions under which photosynthesis took place
Air temperature (°C)
Relative humidity (%)
Photosynthetically active radiation (µmol m−2 s−1)
Vapour pressure deficit (kPa)
Twenty seedlings of each species were planted individually in 19 L pots, for a total of 60 pots. The pots were placed underneath a rain shelter with a glass roof on tables approximately 0.8 m above a concrete surface. The rain shelter reduced incoming photosynthetically active radiation (PAR) by about 20% compared to that observed outside the shelter under sunny conditions. The shelter had no sidewalls, such that air temperature, wind speed and relative humidity were similar to ambient conditions. Each pot contained 17.5 kg dry homogenized soil mixture. The soil mixture comprised 80% by volume dark, air-dried topsoil and 20% by volume air-dried rice husks. The rice husks were added to improve soil structure and drainage. The pots were saturated with water and drained overnight to establish the pot water content at field capacity, which was determined to be 5.5 kg. The soil surface of each pot was then covered with 1.5 kg gravel to reduce soil evaporation. Twenty control pots with no plants were deployed along side the 60 pots containing plants to estimate pot water loss due to evaporation from the soil surface.
At the beginning the experiment, ten of the 20 pots for each species were randomly chosen to receive approximately 12 g Osmocote-Plus controlled-release fertilizer (Scotts-Sierra, Maryville, OH, USA). The fertilizer contained by weight 15% N, 9% P and 12% K, and had an estimated release time of five to six months. Five fertilized and five unfertilized pots from each species were then randomly allocated to receive reduced water supply. All pots started the experiment watered to field capacity. Those receiving the full water allocation were weighed each week and re-watered to near field capacity. Later in the experiment, pots were weighed and re-watered at shorter intervals, depending on water loss rates. Those receiving the reduced water allocation were allowed to dry down to pot water contents of less than 2.5 kg, or approximately 40% of field capacity, over several weeks. Thereafter they were weighed and re-watered to this pot water content each week, or at shorter intervals, as necessary. Pots were weighed to the nearest 5 g with a 64 kg capacity balance (Sartorius QS64B, Thomas, Swedesboro, NJ, USA). Drain holes at the bases of the pots were sealed for the duration of the experiment.
The experiment began on 27 October 2006. Initial plant dry weights were estimated by harvesting five representative individuals of each species. Mean values were 0.6, 0.4 and 0.7 g for T. grandis, S. macrophylla and P. pinnatum, respectively. The T. grandis plants grew considerably faster than those of S. macrophylla or P. pinnatum, and it was therefore necessary to harvest the former earlier than the latter. The T. grandis plants were harvested on 20 December 2006, and S. macrophylla and P. pinnatum were harvested on 1 March 2007.
Growth and transpiration measurements
Cumulative plant water use over the course of the experiment was calculated as the sum of pot water loss minus the average sum of water loss from the control pots. This calculation assumes that soil evaporation from pots without plants was the same as that from pots with plants. Half the control pots were allowed to dry down to pot water contents of 2.5 kg to estimate soil evaporation for the reduced water treatment. Prior to harvest, the pots were weighed at dawn and dusk for 2 d to partition diel transpiration into nighttime and daytime components. This allowed calculation of En/Ed. Immediately following harvest, total leaf area of each plant was measured with a LI-3100 Leaf Area Meter (Li-Cor, Lincoln, NE, USA). Harvested plants were dried to constant mass at 70 °C, and dry mass of leaves, stems and roots was determined separately for each plant to the nearest 0.02 g.
The mean relative growth rate (RGR) of each plant was calculated as RGR = [ln(m2) − ln(m1)]/t, where m1 and m2 are plant dry mass at the beginning and end of the experiment, respectively, and t is duration of the experiment (Blackman 1919). The mean transpiration rate (MTR) of each plant over the course of the experiment was calculated as MTR = Et/[(LA1 + LA2)0.5t], where Et is cumulative transpiration, and LA1 and LA2 are leaf area at the beginning and end of the experiment (Sheshshayee et al. 2005). The TE of each plant was calculated as TE = (mC2 − mC1 + lC)/Et, where mC1 and mC2 are the plant C mass at the beginning and end of the experiment, and lC is the C mass of leaves abscised over the course of the experiment.
To aid in comparison of plants that grew at different rates and over different time periods, we calculated growth-weighted averages of D, v and gravimetric soil water content (SWC) for each plant. Weekly dry matter increments were predicted from estimates of RGR (Cernusak et al. 2008b). Weekly averages of D were calculated from measurements of air temperature and relative humidity recorded by the weather station adjacent to the rain shelter (Winter et al. 2001; Winter, Aranda & Holtum 2005). We used data for the hours between 0700 and 1730 local time, the period during which most photosynthesis would have taken place. A leaf energy budget model developed by DGG de Pury and GD Farquhar (unpublished) and described by Barbour et al. (2000) was used to make predictions of average daytime leaf temperature. The model was parameterized with weekly averages of air temperature, relative humidity, wind speed and irradiance. Representative leaf area and stomatal conductance were additional input parameters. Weekly averages for v were calculated as the difference between predicted daytime ei and daytime ea. Weekly averages for SWC for each plant were calculated from the recorded pot weights. Average fresh mass for each plant for each week was calculated from the predicted dry mass at the beginning and end of the week and fresh mass to dry mass ratios measured at the end of the experiment. The weekly average SWC was calculated as the average water content of the pot for the week divided by the mass of air-dried soil mixture. Growth-weighted averages of D, v and SWC, over the course of the experiment were calculated for each plant as
where xg is the growth-weighted average of either D, v, or SWC, xi is the average value of the same parameter for week i, and mi is the dry mass increment for week i.
Leaf gas exchange measurements
Photosynthesis (A), stomatal conductance (gs) and ci/ca were measured on three to five fully expanded leaves per plant using an Li-6400 portable photosynthesis system (Li-Cor Inc., Lincoln, NE, USA). Measurements were made at PAR of 1200 µmol m−2 s−1, which was supplied by an artificial light source (Li-Cor). Mean v during measurements was 1.6 ± 0.3 kPa (mean ± 1 SD) and mean leaf temperature was 33 ± 1 °C (mean ± 1 SD). Measurements of T. grandis took place on 12 December 2006. These measurements were made approximately in the middle of a watering cycle. Measurements of S. macrophylla and P. pinnatum took place on 26 and 28 February 2007. Each plant of S. macrophylla and P. pinnatum was measured twice, once at the end of the watering cycle and once at the beginning. The two sets of measurements were averaged for each plant.
Stable isotope and elemental analyses
Leaves, stems and roots of each plant were ground separately to a fine, homogenous powder. Samples of approximately 3 mg were analysed for δ13C and C and N concentrations with an elemental analyser (ECS 4010, Costech Analytical Technologies, Valencia, CA, USA) coupled to an isotope ratio mass spectrometer (Delta XP, Finnigan MAT, Bremen, Germany). Approximately 1 mg of stem dry matter of each plant was analysed for δ18O on an isotope ratio mass spectrometer (Delta XP, Finnigan MAT) following pyrolysis in a high temperature furnace (Thermoquest TC/EA, Finnigan MAT). These analyses were carried out at the Stable Isotope Core Laboratory, Washington State University, Pullman, WA, USA. The δ13C and δ18O values have been expressed relative to standards of PeeDee Belemnite and Vienna Standard Mean Ocean Water, respectively. The Δ13C of plant material was calculated as Δ13C = (δa − δp)/(1 + δp), where δa is δ13C of CO2 in air and δp is δ13C of plant C. The δa was assumed to be –8‰. In a previous investigation at the same study site, the δ18O of irrigation water was determined to be –4.0‰ (Cernusak et al. 2007b), although possible deviations from that value cannot be precluded for the present study.
Relationships between continuous variables were analysed by least-squares linear regression. Variation among species in relationships between continuous variables was analysed with analysis of covariance. Variation among species and treatments in physiological and isotopic parameters was assessed by analysis of variance. In these analyses, the number of observations was 60, the degrees of freedom for species was two, the degrees of freedom for nutrient and water treatments were one each, and the degrees of freedom error was 48. Pair-wise comparisons among species following analyses of variance were made according to Tukey's method. Results were considered statistically significant at P < 0.05. Statistical analyses were carried out in Systat 11.0 (SPSS, Chicago, IL, USA).
Growth, biomass allocation and elemental composition
Growth-weighted SWC as a proportion of field capacity varied between high and low water supply regimes, as intended in the experimental design. Mean values were 0.82 and 0.37 for T. grandis, 0.85 and 0.34 for S. macrophylla, and 0.92 and 0.41 for P. pinnatum, for high and low water supply, respectively. A value of one implies field capacity and zero implies air-dried soil.
A summary of growth parameter variation among species and treatments is given in Table 2. Mean RGR varied significantly with species (P < 0.0001), nutrient treatment (P = 0.002) and water treatment (P = 0.0008), and with the interaction between water treatment and species (P = 0.0001). Root/shoot ratio varied with species (P < 0.0001) and nutrient treatment (P < 0.002). Leaf area ratio (LAR) varied with species (P < 0.0001), nutrient treatment (P < 0.0001), and the interactions between water treatment and species (P = 0.0006) and nutrient and water treatments (P = 0.02). Specific leaf area (SLA) varied with species (P < 0.0001), and interactions between water treatment and species (P = 0.02) and nutrient and water treatments (P = 0.02). The mean RGR across all three species correlated significantly with LAR (R2 = 0.63, P < 0.0001, n = 60), and with SLA (R2 = 0.38, P < 0.0001, n = 60).
Table 2. Morphological and physiological parameters for experimental plants. Tectona grandis plants were harvested on 20 December 2006, and those of Swietenia macrophylla and Platymiscium pinnatum on 1 March 2007. Values are presented as mean, with 1 SD in parentheses. For each treatment, n = 5. Treatments are as follows: −N−W (minus fertilizer, minus water), −N+W (minus fertilizer, plus water), +N−W (plus fertilizer, minus water) and +N+W (plus fertilizer, plus water)
Final dry mass (g)
Final leaf area (m2)
Relative growth rate (mg g−1 d−1)
Root/shoot ratio (g g−1)
Leaf area ratio (m2 kg−1)
Specific leaf area (m2 kg−1)
Leaf [N] (mg g−1)
Whole-plant [N] (mg g−1)
Whole-plant C/N (g g−1)
TE (mmol C mol−1 H2O)
Leaf δ13C (‰)
Stem δ13C (‰)
Root δ13C (‰)
Photosynthesis (µmol m−2 s−1)
Conductance (mmol m−2 s−1)
Whole-plant N concentration (Np) varied significantly by species (P < 0.0001), by nutrient treatment (P < 0.0001) and by water treatment (P = 0.001). It was highest in P. pinnatum, lowest in S. macrophylla and intermediate in T. grandis (Table 2). The Np was higher in fertilized than in unfertilized plants, and higher in plants at low than at high water supply (Table 2). There was a significant interaction between water treatment and species (P = 0.02). Patterns for leaf N concentration, both on a mass and an area basis, generally mirrored those for Np (Table 2). Whole-plant C/N ratio varied among species and treatments in an inverse pattern to that for Np (Table 2). Mean whole-plant C concentrations were 44.4, 45.4 and 45.1% for T. grandis, S. macrophylla and P. pinnatum, respectively.
Leaf gas exchange
Stomatal conductance to water vapour (gs) varied significantly among species (P < 0.0001) and between water treatments (P < 0.0001), but not between nutrient treatments. Interaction terms between nutrient treatment and species (P = 0.02) and water treatment and species (P = 0.0002) were significant, indicating variation among species in the response of gs to nutrient and water supply. T. grandis displayed the highest mean gs, S. macrophylla the lowest and P. pinnatum an intermediate value (Table 2). The gs of both S. macrophylla and T. grandis showed marked declines at low compared to high water supply, whereas that of P. pinnatum showed a slight increase at low compared to high water supply (Table 2).
Photosynthesis (A) varied significantly among species (P < 0.0001), and between nutrient treatments (P = 0.006), but not between water treatments. Interaction terms were significant between water treatment and species (P < 0.0001) and between nutrient and water treatments (P = 0.0001). Mean A was highest in T. grandis, lowest in S. macrophylla and intermediate in P. pinnatum, and it was higher in fertilized than in unfertilized plants (Table 2). The response of A to SWC varied among species, as evidenced by the significant water treatment by species interaction. The A of S. macrophylla showed a marked decline at low compared to high water supply, whereas that of T. grandis showed a smaller decline. In contrast, the A of P. pinnatum was higher at low than at high water supply (Table 2). The significant interaction between nutrient and water treatments resulted from a much stronger response of A to fertilizer addition at high compared to low water supply (Table 2).
Instantaneous measurements of ci/ca varied significantly among species (P < 0.0001), between nutrient treatments (P = 0.0007) and between water treatments (P < 0.0001). Additionally, interaction terms between nutrient treatment and species (P = 0.03) and between water treatment and species were significant (P < 0.0001). Mean values for ci/ca were significantly lower in S. macrophylla than in T. grandis or P. pinnatum, whereas they were similar between the latter two species (Table 2). The ci/ca was lower in fertilized than in unfertilized plants, and lower at low than at high water supply. The decrease in ci/ca in S. macrophylla at low water supply was much sharper than that in T. grandis or P. pinnatum, which lead to the strong water treatment by species interaction.
Variation in measurements of whole-plant transpiration is shown in Fig. 1a–f. The Ed was based on pot weights taken shortly before harvest at dawn and dusk, and therefore represents an average transpiration rate over the day. The MTR represents an average daily transpiration rate over the full course of the experiment, calculated by employing the assumption that leaf area increased linearly during that period. Patterns among species and treatments for Ed and MTR were generally in good agreement with each other (Fig. 1), and the two independent estimates of whole-plant transpiration were closely correlated across the full data set (r = 0.82, P < 0.0001, n = 59).
The Ed varied significantly by species (P = 0.03) and by nutrient treatment (P = 0.004), but not by water treatment. There was a strong interaction between water treatment and species (P < 0.0001), as can be clearly seen in Fig. 1a–c. The Ed decreased at low compared to high SWC in S. macrophylla (Fig. 1a), did not change depending on SWC in T. grandis (Fig. 1b), and increased at low compared to high SWC in P. pinnatum (Fig. 1c). Results for MTR were similar, with significant variation by species (P = 0.0001) and by nutrient treatment (P < 0.0001), but not by water treatment. Again, there was a significant interaction between water treatment and species (P < 0.0001), with patterns among species in response to decreasing SWC similar to those observed for Ed (Fig. 1d–f). For MTR there was also a moderately significant interaction between nutrient and water treatments (P = 0.05). Both Ed and MTR were significantly higher in unfertilized than in fertilized plants (Fig. 1).
The ratio of nighttime to daytime transpiration rates, En/Ed, varied significantly by species (P < 0.0001), but not by nutrient or water treatments. There was a significant interaction between water treatment and species (P = 0.0008). The mean En/Ed for T. grandis was 0.052, that for S. macrophylla was 0.032 and that for P. pinnatum was 0.012. The En/Ed was higher at low than at high water supply for S. macrophylla and P. pinnatum, but the converse was true for T. grandis.
The TE, calculated as experiment-long dry matter production divided by total water use for each plant, varied significantly by species (P < 0.0001), by nutrient treatment (P < 0.0001) and by water treatment (P < 0.0001). There were significant interactions between nutrient treatment and species (P = 0.04) and between water treatment and species (P = 0.0009), indicating variation among species in responses of TE to nutrient and water supply. Mean TE was significantly higher in T. grandis than in S. macrophylla or P. pinnatum, and similar in the latter two species (Table 2). The TE was higher in fertilized than in unfertilized plants, and higher at low than at high water supply. The TE of S. macrophylla and P. pinnatum increased in response to fertilizer addition, whereas that of T. grandis showed no response to fertilizer addition. On the other hand, TE of S. macrophylla and T. grandis increased with decreasing SWC, whereas that of P. pinnatum showed no response to SWC (Table 2).
Estimation of the growth-weighted vapour pressure deficit, D, showed that the higher TE of T. grandis resulted from a reduced air vapour pressure deficit for the period during which this species grew. The T. grandis plants were harvested in late December, whereas those of S. macrophylla and P. pinnatum were harvested at the beginning of March, due to variation among species in RGR. The vapour pressure deficit in January and February was about twice that in November and December (Table 1), as would be expected with the progression of the dry season at the study site. The D·TE showed a similar pattern of variation to TE within species (Fig. 2a–c); however, the ranking among the three species was different than for TE. The mean values of D·TE were 3.00, 2.66 and 2.07 Pa mol C mol−1 H2O for S. macrophylla, P. pinnatum and T. grandis, respectively.
Stable isotope composition
Mean δ13C for leaves, stems and roots for each treatment within each species is given in Table 2. Leaf δ13C was consistently more negative than that of stems or roots. Mean values across the full data set were –28.3, –26.7 and –26.2‰, for leaves, stems and roots, respectively.
Whole-plant Δ13C (Δ13Cp) varied by species (P < 0.0001), by nutrient treatment (P < 0.0001) and by water treatment (P < 0.0001). Interactions between nutrient treatment and species (P < 0.0001) and between water treatment and species (P < 0.0001) were significant, indicating variation among species in responses of Δ13Cp to nutrient and water availability. The Δ13Cp of S. macrophylla showed the strongest response to both fertilizer addition and decreasing SWC (Fig. 2d). That of T. grandis showed no response to fertilizer addition, and an intermediate response to decreasing SWC (Fig. 2e). Finally, Δ13Cp of P. pinnatum showed only weak responses to either fertilizer addition or decreasing SWC (Fig. 2f).
Slopes of the relationships between Δ13Cp and instantaneous ci/ca varied among species (P < 0.0001), as shown in Fig. 3. Regression equations indicated slope estimates of 18.0, 10.6 and 2.7‰ for S. macrophylla, T. grandis and P. pinnatum, respectively. The instantaneous ci/ca explained 96% of variation in Δ13Cp for S. macrophylla (P < 0.0001, n = 20), 67% for T. grandis (P < 0.0001, n = 20) and 10% for P. pinnatum (P = 0.18, n = 20).
Relationships between D·TE and Δ13Cp varied among species (Fig. 4a). The slopes of the relationships were not significantly different; however, there was significant variation in the intercepts (P < 0.0001). Thus, we observed species-specific offsets in the relationship between D·TE and Δ13Cp. Variation in the relationships for S. macrophylla and T. grandis could be reconciled by taking into account modeled differences between leaf temperature and air temperature. Thus, relationships between v·TE and Δ13Cp for S. macrophylla and T. grandis were very similar (Fig. 4c). However, the intercept of the relationship between v·TE and Δ13Cp for P. pinnatum still differed significantly from those for S. macrophylla and T. grandis (P < 0.0001). On the other hand, relationships for S. macrophylla and P. pinnatum could be reconciled by replacing Δ13Cp with instantaneous ci/ca (Fig. 4b & d). Accordingly, a plot of v·TE against instantaneous ci/ca (Fig. 4d) produced the greatest homogeneity among species for the four sets of relationships shown in Fig. 4. There still existed a small but significant variation in slopes among species in the relationships between v·TE and instantaneous ci/ca (P = 0.02). However, the vast majority of variation in v·TE across the full data set could be accounted for by considering only instantaneous ci/ca. Thus, a general linear model, taking as independent variables ci/ca, species, and the species by ci/ca interaction, explained 89% of variation in v·TE. On the other hand, ci/ca on its own explained 82% of variation in v·TE.
Relationships between the δ18O of stem dry matter (δ18Ost) and gs varied among species (P < 0.0001), as shown in Fig. 5. The δ18Ost showed a strong dependence on gs for S. macrophylla (R2 = 0.80, P < 0.0001, n = 20), a weak dependence for P. pinnatum (R2 = 0.19, P = 0.05, n = 20), and no dependence for T. grandis (R2 = 0.05, P = 0.37, n = 20). Analyses of the dependence of δ18Ost on three independent estimates of transpiration rate are shown in Table 3. The δ18Ost was strongly correlated with transpiration for S. macrophylla, weakly correlated for P. pinnatum, and not correlated for T. grandis (Table 3).
Table 3. Results of linear regression analyses with stem dry matter δ18O as the dependent variable and three different measures of transpiration rate alternately taken as independent variables. The Ei is the instantaneous transpiration rate (mmol m−2 s−1) measured with the portable photosynthesis system; Ed is the average transpiration rate over a day determined gravimetrically (mmol m−2 s−1); and MTR is the mean transpiration rate over the course of the experiment (mol m−2 d−1). Asterisks indicate statistical significance as follows: *(P < 0.05),**(P < 0.01) and***(P < 0.001);n = 20 for all analyses
Regression coefficient with stem δ18O as dependent variable
We used D·TE in our analysis to account for differences in TE that resulted from variation in D, the growth-weighted air vapour pressure deficit. Variation in D resulted primarily from the need to harvest the T. grandis plants at an earlier date than the other two species, due to faster RGR in the former (Table 2). The TE was calculated as dry matter production over the full course of the experiment divided by cumulative plant water use. Results of the present study add to a growing body of evidence showing that D·TE of S. macrophylla is higher than that of T. grandis. Both S. macrophylla (mahogany) and T. grandis (teak) are commercial timber species that are widely planted in the tropics. The mean D·TE of S. macrophylla across all treatments in the present study was 3.00 Pa mol C mol−1 H2O, whereas that of T. grandis was 2.07 Pa mol C mol−1 H2O; thus that of S. macrophylla was 45% higher compared to that of T. grandis. The ranking between these two species did not change among water and nutrient treatments. Differences between TE of S. macrophylla and T. grandis of a similar magnitude have been observed in previous experiments (Winter et al. 2005; Cernusak et al. 2007a, 2008b).
In contrast to the stability in the ranking of D·TE among treatments between S. macrophylla and T. grandis, that of P. pinnatum was variable. P. pinnatum was previously observed to have the highest D·TE among the three study species under high water supply in unfertilized soil (Cernusak et al. 2007a, 2008b). Under similar conditions in the present experiment, P. pinnatum also had the highest D·TE, consistent with the previous results. However, D·TE of P. pinnatum showed no response to decreasing SWC and only a weak response to fertilizer addition (Fig. 3c). Thus, under conditions of low water supply, or in fertilized soil, the D·TE of P. pinnatum was intermediate between that of S. macrophylla and T. grandis. P. pinnatum is a leguminous tree capable of forming N2-fixing root nodules. The leaf N concentration was higher in P. pinnatum than in S. macrophylla or T. grandis in all treatments; however, P. pinnatum enjoyed its greatest advantage in leaf N concentration over the other two species under conditions of high water supply in unfertilized soil (Table 2). Thus, the ability to symbiotically fix N2 from the atmosphere may have contributed toward higher water-use efficiency under conditions of relatively low soil N availability and high water supply.
Relationships between TE, Δ13Cp and ci/ca
We observed significant offsets among the three species in relationships between D·TE and Δ13Cp (Fig. 4a), consistent with previous results (Cernusak et al. 2007a, 2008b). In the present study, it could be demonstrated that the variation among species in the relationship between D·TE and Δ13Cp resulted from both an uncoupling of Δ13Cp from ci/ca in P. pinnatum, and from larger fv in T. grandis than in the other two species, where fv is defined as v/D. Thus, plotting v·TE against instantaneous ci/ca resulted in a generally uniform relationship among the three species (Fig. 4d). Data collected previously for these species (Cernusak et al. 2008b) can be added to the analysis. For the combined dataset, instantaneous ci/ca explained 87% of variation in v·TE (R2 = 0.87, P < 0.0001, n = 79). Adding species and a species by ci/ca interaction term resulted in only a very slight increase in the proportion of variation in v·TE that was explained (R2 = 0.89, P < 0.0001, n = 79), providing further evidence of a common relationship between v·TE and ci/ca among the three species.
These results suggest that in addition to ci/ca, fv can be an important source of variation in D·TE. Mean values of fv among the three species were 2.0, 1.4 and 1.3 for T. grandis, S. macrophylla and P. pinnatum, respectively. These values of fv correspond to mean predicted differences between leaf and air temperature of 2.9, 2.1 and 1.3 °C for the three species, respectively. The larger fv for T. grandis resulted from a combination of a larger representative leaf area employed in the leaf energy budget model (325 cm2 compared to 50 cm2 for S. macrophylla and P. pinnatum), and lower wind speeds during the growth period of T. grandis compared to that of S. macrophylla and P. pinnatum. The v·TE did not differ significantly between T. grandis and S. macrophylla in the present experiment (P = 0.25), supporting the suggestion that variation in TE between these two species, when grown under similar environmental conditions, results at least partly from differences in v, caused by differences in leaf temperature (Winter et al. 2005).
Equation (2) suggests that φc, the fraction of C taken up by photosynthesis that is subsequently lost to the atmosphere by respiration, can be estimated by plotting v·TE against ca(1 − ci/ca)/1.6. Figure 6 shows such a plot, and includes data from previous measurements on the same three species (Cernusak et al. 2008b). The slope of a relationship between v·TE and ca(1 − ci/ca)/1.6, with the intercept forced through the origin, is thus equal to (1 − φc)/(1 + φw). Assuming ca of 38 Pa, slope estimates for T. grandis, S. macrophylla and P. pinnatum are 0.705, 0.657 and 0.642, respectively. Combining these with estimates of φw, calculated as En/Ed (0.052, 0.032 and 0.012, respectively), results in estimates for φc of 0.258, 0.322 and 0.350 for T. grandis, S. macrophylla and P. pinnatum. This suggests a modest variation among species in φc. This method of estimating φc is rather indirect, and it would be helpful to confirm the results with direct measurements of whole-plant gas exchange. Nonetheless, it suggests that φc and φw are unlikely to cause large variations in TE, whereas ci/ca and v are primary controls. This agrees with some previous analyses (Cernusak et al. 2007b, 2008b), but contradicts others (Guehl, Fort & Ferhi 1995; Hobbie & Colpaert 2004).
We observed strong variations among the three species in the dependence of Δ13Cp on ci/ca (Fig. 3). In the case of P. pinnatum, the slope of the relationship between the two parameters did not differ significantly from zero. On the one hand, the timescales of measurement for Δ13Cp and ci/ca were very different, with ci/ca integrating over minutes and Δ13Cp integrating over months. On the other hand, the strong correlation between v·TE and ci/ca (Fig. 5d) suggests that the measured ci/ca values were generally representative of those experienced throughout the experiment. Unfortunately, our dataset provides no insight into whether the uncoupling between Δ13Cp and ci/ca in P. pinnatum occurred during photosynthesis, or as the result of a post-photosynthetic process (Hobbie & Werner 2004; Badeck et al. 2005; Bowling, Pataki & Randerson 2008; Cernusak et al. 2009). Simultaneous measurements of Δ13C and ci/ca during photosynthetic gas exchange would be helpful in this respect.
Dependence of δ18Ost on gs
There is growing interest in using measurements of δ18O in plant organic material as a proxy indicator for variation in gs (Farquhar & Lloyd 1993; Barbour et al. 2000; Barbour 2007). Such an application would be particularly useful in water-use efficiency research, as it would allow variation in Δ13C to be attributed to variation in either A or gs. The expectation of a negative relationship between δ18O of plant organic material and gs is theoretically based in steady-state leaf water 18O enrichment (Farquhar & Lloyd 1993; Farquhar, Cernusak & Barnes 2007; Ripullone et al. 2008), and the expected relationship has been observed in a number of species (Barbour & Farquhar 2000; Barbour et al. 2000, 2004; Grams et al. 2007; Sullivan & Welker 2007). In the present study, we observed a strong negative relationship between δ18Ost and gs in S. macrophylla, a weak negative relationship in P. pinnatum, and no relationship in T. grandis (Fig. 5). Correlations between δ18Ost and transpiration rate followed a similar pattern (Table 3). All species covered a similarly large range of gs, and we are therefore unable to explain the presence of a negative relationship in one species and not another.
The mean δ18Ost of T. grandis was less than that of S. macrophylla and P. pinnatum by approximately 3‰ (Fig. 5). This partly reflected the higher relative humidity during November and December, compared to that during January and February (Table 1). T. grandis was harvested on 20 December, whereas S. macrophylla and P. pinnatum were harvested on 1 March. The growth-weighted estimate of δ18Oe, calculated as described previously (Cernusak et al. 2008b), was lower by 3.0‰ for T. grandis compared to the average for S. macrophylla and P. pinnatum. Across the full data set, the δ18Oe explained 67% of variation in δ18Ost (R2 = 0.67, P < 0.0001, n = 60), suggesting a reasonable agreement between predicted variation in leaf water 18O enrichment and observed variation in stem dry matter δ18O.
Whole-plant transpiration was higher in unfertilized than in fertilized plants when expressed on a leaf area basis, and this pattern was consistent across the dataset (Fig. 1). Assuming that fertilization generally functioned to increase the N concentration of the soil solution, this result is consistent with previous results indicating increased plant transpiration in response to low N availability (Guehl et al. 1995; Livingston et al. 1999; Cernusak et al. 2007b; Cramer, Hoffmann & Verboom 2008). An increased transpiration rate in this case would serve to increase concentrations of nitrogenous solutes at the root surface due to convective transport (mass flow) and thus increase N absorption compared to that at a lower transpiration rate (Barber 1995). Interestingly, in our dataset, gs did not vary by nutrient treatment, suggesting that modulation of whole-plant transpiration in response to N availability did not occur through variation in gs. Canopy architecture may have played a more important role in coordinating whole-plant transpiration with N availability through increased self-shading and canopy boundary layer development in fertilized plants, caused by larger leaf areas (Table 2).
The response of whole-plant transpiration to decreased SWC varied among the three species (Fig. 1). The response of P. pinnatum was difficult to explain, in that it increased at low compared to high SWC (Fig. 1c & f). Measurements of gs were consistent with this response. For P. pinnatum, gs tended to increase with declining SWC from near field capacity to about one-third field capacity. At SWC less than about one-third field capacity, gs of P. pinnatum decreased linearly. In general, variation in growth and water use was more variable within treatments for P. pinnatum than for the other two species (Table 2). This could have related to the fact that the P. pinnatum seedlings were collected from the field and transplanted to the study site, whereas the other two species were germinated from seed at the study site under controlled conditions. Thus, it is possible that the higher transpiration rates at low compared to high SWC for P. pinnatum were coincidental, resulting from large background variation in physiological performance among the study population for this species. Nevertheless, a diverse range of responses of gs and whole-plant transpiration to declining SWC has been observed in other tropical tree species (Bonal et al. 2000; Bonal & Guehl 2001), and the apparent increase in transpiration in P. pinnatum in response to low water supply that we observed is therefore worthy of further investigation.
The response of D·TE to variable water and nutrient supply did indeed differ among the three species studied (Fig. 2a–c). Thus, their relative ranking was not maintained across treatments. Offsets were observed among species in relationships between D·TE and Δ13Cp (Fig. 4a). For P. pinnatum, this offset resulted from a breakdown in the dependence of Δ13Cp on ci/ca (Fig. 3). In the case of T. grandis, the offset resulted from variation in fv, such that it could be reconciled by substituting v·TE for D·TE (Fig. 4c). Thus, species-specific offsets in relationships between D·TE and Δ13Cp resulted from both variable dependence of Δ13C on ci/ca and variable dependence of D·TE on ci/ca. Additionally, we observed a clear variation among species in the dependence of stem dry matter δ18O on stomatal conductance (Fig. 5).
We thank Jorge Aranda, Milton Garcia, Aurelio Virgo, Lisa Petheram, Carlos Martinez, Tania Romero and Aneth Sarmiento for technical assistance in carrying out the experiment. Lucas A. Cernusak was supported by a Tupper Postdoctoral Fellowship from the Smithsonian Tropical Research Institute and by an Australian Postdoctoral Fellowship from the Australian Research Council.