1Habitat specialization models predict that adaptations to environmental conditions explain species distributions. In tropical rainforests, the ability of the seedlings to survive during drought has been shown to be a key determinant of species distributions. We hypothesize that differences among species in their tolerance to low tissue water status is the mechanism underlying differences in performance during drought.
2To test this hypothesis we quantified tolerance to low leaf water status for over 20 species from central Panama in screenhouse experiments using two different experimental approaches. Results from both approaches were highly correlated with each other.
3We found that tolerance to low leaf water status correlated with species drought performance in the field and with their distribution across a gradient of dry season length, with the more desiccation-tolerant species having higher survival in dry relative to irrigated conditions, and occurring in drier areas. These results support the hypothesis that, in tropical forests, tolerance to low tissue water status governs seedling performance during drought, as well as being a determinant of species distribution patterns.
4Lower tolerance to low leaf water status was correlated with greater stem hydraulic conductance. In addition, all species tested, including both desiccation-sensitive and desiccation-resistant species, showed similar losses of xylem conductance, about 80%, when near death. These results suggest that a principal mechanism by which desiccation leads to plant mortality is the loss of xylem conductivity.
Understanding the mechanisms by which abiotic interactions determine the abundance and distribution of organisms is a central goal of ecology. Research in the last two decades in tropical rainforests has demonstrated that rainfall is a key factor. Rainfall varies up to 10-fold between dry, moist and wet tropical forests (Holdridge 1947; Murphy & Lugo 1986; Clinebell et al. 1995; Walsh 1996) and the distributions of many plants correlate with rainfall (Veenendaal & Swaine 1998; Bongers et al. 1999; Baltzer et al. 2008). As a result, a marked turnover of plant species is associated with this moisture gradient (beta diversity) and such turnover makes a large contribution to regional diversity (Condit et al. 2002; Davidar et al. 2007). Additionally, wetter forests harbour higher diversity (alpha diversity) than drier forests (Gentry 1988; Clinebell et al. 1995; ter Steege et al. 2003. Accordingly, understanding the role that rainfall and seasonality play in plant distributions is central to tropical ecology.
Recently we found that drought performance is an important determinant of distribution with respect to the length of the dry season, such that species with poor drought performance are excluded from dry forest (Engelbrecht et al. 2007a). We refer to the ability of a species to survive under field conditions during low water availability as drought performance, a term that encompasses the multiple stresses that may be present under natural drought including changes in pests or nutrients (Tyree et al. 2003). In the present study, we have excluded factors such as natural enemies and soil fertility from consideration and focus on the role of drought resistance in drought performance. We define drought resistance as the suite of traits that support survival under conditions of a single stress, low water availability.
Strategies of drought resistance include desiccation delay and tolerance to low tissue water status (Tyree et al. 2003). Desiccation delay involves traits that increase access to water and reduce water loss. Deep roots, early stomatal responses, low cuticular conductance, water storage in stem of other organs, osmotic adjustment, and leaf shedding all can contribute to desiccation delay, and can be directly measured. Analysis of plant water status over a drying cycle in a common garden can serve as an integrated, quantitative measure of species differences in desiccation delay. Tolerance to low tissue water status, or desiccation tolerance, is based upon continued plant function despite water loss. The ability of plants to tolerate water deficits is promoted by physiological traits such as greater resistance of the xylem to embolism, permitting continued water transport and gas exchange, or the ability of cells (especially meristems) to remain alive at low relative water content and low water potentials (Ψ). For most species, little is known of the relative importance of desiccation delay and desiccation tolerance in promoting survival during a period of reduced water availability (Tobin, Lopez & Kursar 1999; Tyree et al. 2003; Nepstad et al. 2007).
Our analysis of five species from Panama suggested that their drought performance depended predominantly on their desiccation tolerance (Tyree et al. 2003). Similar methods applied in an independent analysis of floristic changes across a climate gradient in Southeast Asia showed that desiccation tolerance correlated with distribution (Baltzer et al. 2008). In the present study we developed a second, independent method to assess desiccation tolerance in which we quantified the water status of plants at 50% mortality. Both the methods are related to the analysis of lethal water status or the degree of stress that causes incipient damage (Moore & Chapman 1986; Ludlow & Muchow 1990; Engelbrecht, Tyree & Kursar 2007b). We extended our data set to over 20 species to further investigate the role of desiccation tolerance in drought performance and distribution.
Additionally, we assessed the mechanism leading to mortality during desiccation. Xylem cavitation is generally believed to be the proximate cause of plant death (Davis et al. 2002). Nevertheless few studies correlate vulnerability to loss of hydraulic conductance with the water potential at which plants die. If cavitation causes mortality, the magnitude of the loss of xylem conductance in severely wilted plants is expected to be similar, regardless of whether a species became severely wilted at high or low water potential, as Tyree et al. (2003) found for four tropical species. In the present study we expanded the previous data set to 12 species, including both desiccation-sensitive and desiccation-resistant species.
site and plant material
Screenhouse experiments were carried out at the Smithsonian Tropical Research Institute facilities on Barro Colorado Island in the Republic of Panama. We studied 33 species; most are shade-tolerant trees that are common in central Panama (Table 1). Seeds or seedlings were collected from three or more trees in central Panama, mainly in the forests in the Barro Colorado Nature Monument. In this forest the average annual rainfall is 2600 mm, with a pronounced 4-month dry season (Leigh 1999).
Table 1. Study species. All species except Apeiba aspera and Ochroma pyramidale are shade-tolerant and can persist in the forest understorey
‘Symbol’ is used in the Figures to identify individual species. Dp is the drought performance in the field experiment. Species with ‘nd’ were not included in the field experiment. ‘Distribution’ indicates the occurrence across the Isthmus of Panama and ‘Classification’ indicates how species distributions were entered into the analyses (see Methods).†Pithecellobium rufescens in Croat (1978). ‡Pouteria unilocularis in Croat (1978). §Swartzia simplex var. grandiflora. For authorities consult the Tropicos data base (http://www.tropicos.org).
Desiccation tolerance was assessed in two ways: by measuring the leaf water status corresponding to 50% of plants having either shoot or plant death (experiment 1), and at the severely wilted stage (experiment 2). Leaf water status was evaluated as leaf water potential (Ψleaf) and as relative leaf water content (RWC). Leaf water content per leaf dry weight, wd, was determined as (fresh weight – dry weight)/dry weight. RWC was determined as the ratio of the wd of droughted plants divided by the wd of fully hydrated plants and converted to a percent.
Plants were 1–3 years old and about 10 to 40 cm tall and kept in a screenhouse to exclude herbivores. Shade cloth reduced the light conditions to 6–7% of that measured in a nearby clearing. Plants were grown in pots of 1·7–1·9 L volume in well-mixed soils collected from the forest and irrigated at least two times per week. For the drought treatment, plants were placed in roofed (rainproof) enclosures and dried to a range of visually assessed wilting stages by withholding watering. Drying times in the greenhouse agreed well with those that we had previously determined in the field for seedlings of shade-tolerant species, 2 weeks to 2 months (Engelbrecht & Kursar 2003; Engelbrecht, Kursar & M.T. Tyree 2005), as well as for seedlings of pioneer species, less than a week (Engelbrecht et al. 2006). Sample sizes were 20–47 or 6–31 individuals per species and for experiments 1 and 2, respectively. The only exceptions were Apeiba aspera and Ochroma pyramidale with 84 and 85 seedlings, respectively. Both species, which typically occur in tree-fall light gaps and have tiny seedlings (< 15 mm high), were grown under high light conditions (30–50% of full sun) in ‘row seed flats’, with 15–20 plants per flat (25 × 3·2 cm and 3 cm deep; Hummert International, Earth City, MO). Trials for the different species were staggered over a 10-month period.
water status at 50% mortality (experiment 1)
For 24 species we assessed plants at a range of desiccation stages for RWC, Ψleaf, and survival after rehydration. We used these data to calculate the RWC and Ψleaf at 50% mortality. We term this ‘lethal desiccation 50′ or LD50 by analogy with the terminology used in dose-response analysis, with LD50RWC and LD50Ψ , representing the RWC and the Ψleaf at 50% mortality, respectively. Plants were left unwatered for 10–140 days (4 days in Ochroma seedlings), giving plants that ranged from only very mildly desiccated to dead. Leaves were sampled for RWC and Ψleaf; plants were rewatered and maintained in a well-watered state. In order to maximize the statistical power for identifying the LD50, an excess of plants were desiccated to slightly less than, at the point of, and slightly beyond severely wilted. Three months after rewatering the plant's shoots were scored as alive or dead. For most species, this coincided with the death of the entire plant. Five species, Cojoba rufescens, Crossopetalum parviflorum, Hybanthus prunifolius, Piper cordulatum and Piper trigonum, had major shoot dieback but did eventually resprout from the roots or lower stem. The reason for using the shoot death criterion is that, because the LD50 analysis is based upon leaf water status, the measurements are only meaningful to the point of hydraulic failure. Once the upper stem has died (or leaf abscission is initiated), the remaining living tissues at the base of the shoot or in the root are not in hydraulic contact with the leaves. Consequently, because this experiment quantifies the leaf water status, it measures the water status of the shoot but not of the surviving tissues. Analyses with the above five species excluded, however, gave very similar results.
The lethal desiccation state was determined in a dose-response analysis of survival as a function of leaf RWC or Ψleaf. The data were fitted using probit regression (PROC PROBIT; SAS Institute 1999). The probability of survival was modelled as:
where Φ−1 (p) is the inverse of the cumulative normal probability function (Young & Young 1998). Untransformed values for RWC and Ψleaf were entered into the model. Implementation of the ‘INVERSECL’ option of SAS computed the water status at a range of survival probabilities and also implemented Fieller's theorem to give the 95% confidence limits. A worked example of the calculation of confidence intervals is provided in Collett (2003). To obtain RWC and Ψleaf at 50% mortality, we set P = 0·5 in eqn 1 and obtain Φ−1 (p) = 0. Values at 50% mortality then equal:
Appendices S1 and S2 contain values for the intercept (β0), beta (β1), LD50 and SE (from Fieller's theorem) (see Supplementary Material). As shown in Fig. 1a,b, we also calculated the probability of survival as a function of RWC or Ψleaf as follows:
where Φ is the cumulative normal probability function (see Young & Young 1998 for a worked example). For a given Φ (RWC) or Φ (Ψleaf), the value for P was obtained from a table of the cumulative normal probability function (as found in statistics texts, for example, Table A in Moore & McCabe 2003). In some species, no variation was observed: all plants above a specific water status lived and all of those below a specific water status died (Fig. 1c,d). Such data were inappropriate for statistical analysis and the LD50's were estimated by interpolation. We report the midpoint between the water status of the plant that died at the highest and that of the surviving plant with the lowest water status. All three values are reported in Appendices S1 and S2.
water status at the severely wilted stage (experiment 2)
For 28 species, we assessed the RWC and Ψleaf of severely wilted plants represented by, SWRWC and SWΨ, respectively. We also include data for five species from Tyree et al. 2003. The drying time to reach the severely wilted state varied from 15 to 52 days. We defined severely wilted plants as those with 25–50% leaf necrosis (modified from Tyree et al. 2003). Additionally we restricted the severely wilted state such that more than 40% of plants had to survive when irrigation was resumed. To that end, a subset of 6–20 severely wilted plants per species were rewatered and scored for survival 3 months later. Survivorship averaged 77% across species (range 43–100%, with eight species at 100%). When the visual definition proved inadequate (Lacmellea panamensis, Swartzia simplex and Posoqueria latifolia, all plants dead), we redefined severely wilted as any visual symptom that occurred just before plants die and repeated the trial if enough plants were available (P. latifolia).
measurement of leaf water potentials (Ψleaf) and relative water content (rwc)
Before the start of each species trial the total leaf area of each plant was assessed (Tyree et al. 2002; Engelbrecht & Kursar 2003). In the morning, leaf samples were collected for water status measurements and plants were photographed (Nikon Cool Pix 1000). The images were used for retrospective analyses of the wilting states. Leaf water potentials, Ψleaf, were determined using thermocouple psychrometers. Leaves were dried and cleaned with Kimwipes and, 5–15 min later, 5·6 mm2 leaf discs were taken using leaf-cutter psychrometers (Merrill Engineering, Logan, UT). For well-watered controls, on the day prior to the measurements, pots were immersed in water for 5 min, drained for 30 s and the shoot and pot were enclosed in a plastic bag until the measurement. In experiment 1 two leaf discs were collected per psychrometer and four measurements were made per plant, in experiment 2, one leaf disc was collected per psychrometer and ten measurements were made per plant. We collected tissue that appeared to be living, usually from the youngest leaves.
We used a CR7 data logger with A3769 psychrometric modules (Campbell Instruments, Logan, UT) in psychrometric mode after 3–5 h equilibration in a water bath at 23·5–25·5 °C using cooling times of 15 s (for 0·3–4 MPa) and 45 s (for 4–8 MPa). The psychrometers were calibrated with NaCl solutions. Ψleaf was determined by psychrometry down to −8 to −9 MPa. Water potentials between −9 and −12 MPa were estimated from the water content per leaf disc (Tyree et al. 2002). Following the Ψleaf determination, we measured the leaf disk fresh weight to the nearest 0·001 mg (Sartorius MC5 microbalance; Precision Weighing Machines, Bradford, Massachusetts). The disk was dried for 2·5 days at 60 °C and reweighed (dry weight). Low values of leaf water status that were over two SEs from the mean for the plant were removed on the assumption that such leaf segments were not in hydraulic contact with the stem.
drought-performance under field conditions and species distributions
Seedling drought performance in the forest understorey was assessed for 48 species in the field in the Barro Colorado Nature Monument during 2000–2001 and 2002–2003, with eight species used in both experiments. Drought performance in the field, Dp, was calculated based on the ratio of survival in dry and wet plots:
where Sd and Sw were relative survival in non-irrigated and irrigated plots, respectively. Results of the field experiments have been reported (Engelbrecht & Kursar 2003; Engelbrecht et al. 2007a). Eighteen of the species in the present study also had Dp values.
A pronounced rainfall gradient over a distance of only 65 km and at less than 100 m above sea level occurs across the Isthmus of Panama. Rainfall ranges from 1650 mm year−1 at the Pacific side to 3000 mm year−1 at the Caribbean side, and dry season length varies from 150–115 days (Engelbrecht et al. 2007a). To relate desiccation tolerance to distribution, we classified species as present or absent in the Pacific, middle, or Caribbean regions of the Isthmus (Table 1). Occurrence was determined from field observations, collection locales of herbarium specimens, the Missouri Botanical Garden's Tropicos data base (http://www.tropicos.org), the Centre for Tropical forest Science data sets (http://ctfs.si.edu/datasets/) and Engelbrecht et al. (2007a). Species found on the Pacific side, including those with distributions in the Pacific plus middle or all across the Isthmus, were classified as dry-side species. Species distributed in the Caribbean only or in the Caribbean plus middle were classified as wet-side species (Table 1). Species only found in the middle were excluded. Thus, 14 of our study species occurred in dry forests, and 14 were restricted to wetter forests.
stem hydraulic conductance
Hydraulic conductance was determined on the whole stem (with the leaves removed) using a vacuum method and a Sartorius CP2250 balance (with an accuracy of 0·05 mg, Precision Weighing Machines, Bradford MA, USA; Kolb, Sperry & Lamont 1996; Tyree et al. 2003). Flow was determined at the following sequence of vacuum pressures: 0, −24, −47, −71, −59, −36, −12 and 0 kPa and kws was obtained by linear regression of flow as a function of pressure (corrected to 25 °C). To account for differences in plant size, we scaled conductance to the total leaf area, giving leaf-specific hydraulic conductance or kws (kg s−1 m−2 MPa−1). We used the leaf area before plants had been drought-stressed so that stem hydraulic conductance was scaled to the same leaf area in watered and wilted plants. Thus, our calculation of stem conductance was not adjusted to the leaf area present at the time of the measurement of the severely wilted plants (having 25–50% loss of living leaf area). All kws data are in Appendix S3.
tolerance of low leaf water status
The capacity to tolerate low leaf water status varied widely among species. The lethal leaf water status assessed as LD50RWC varied from 58·5% for Calophyllum longifolium down to 7·0% for Hybanthus prunifolius, that is, some species showed 50% mortality after they had lost 40% of their leaf water content relative to the fully hydrated leaf, whereas others tolerated a loss of 93% of their leaf water. LD50Ψ ranged from −1·9 MPa for Virola surinamensis down to −12 MPa (or less) for Alseis blackiana and Chrysophyllum argenteum. The 95% confidence limits for LD50RWC and LD50Ψ were large, often 50–100% of LD50RWC and LD50Ψ (not shown). From a regression of confidence intervals as a function of sample sizes (not shown), we estimated that about 50 plants would be required to decrease the 95% confidence interval to < 25% of the modelled LD50.
Desiccation tolerance assessed as leaf water status of severely wilted plants, SWRWC, showed a similarly wide variation: from 61·1% in Virola surinamensis to 16·2% in Hybanthus prunifolius. Hence species were severely wilted after losing 40–84% of their leaf water. SWΨ varied from −2·1 MPa for V. surinamensis to −9·9 MPa for Alibertia edulis (Fig. 2).
Regression analysis indicated that the relationship between RWC and Ψleaf was linear in both experiments and not significantly different from each other (Fig. 3a,b). Our two independent assessments of desiccation tolerance were highly significantly correlated with each other. The slopes, 0·98 for RWC and 1·11 for Ψleaf, were not significantly different from 1·0 (Fig. 3c,d), indicating that the two experiments assessed similar physiological states. The LD50 experiment may give lower water status values than the severely wilted experiment if leaves were in the process of abscission (e.g., species with low leaf dry weight per leaf area). For three of four such species we noted that the critical water status at the LD50 was indeed lower than at the severely wilted stage (Fig. 3c,d filled circles, Ab Car, Hpr but not Cc).
the relationship of tolerance to low leaf water status with plant drought performance and distributions
Drought performance (Dp), assessed as the percent survivorship in unwatered relative to irrigated plots in the forest understorey (Engelbrecht & Kursar 2003; Engelbrecht et al. 2007a) varied considerably in our study species from very low, 0% for Beilschmedia pendula to very high, 100% for Dipteryx panamensis. The Dp values from the field experiments were significantly correlated with desiccation tolerance assessed as SWRWC and SWΨ (Fig. 4a,b) as well as LD50RWC and LD50Ψ (Fig. 4c,d).
If desiccation tolerance is a determinant of plant distributions, we predicted that species typical of drier forests should have higher desiccation tolerance, that is, lower SWRWC, SWΨ, LD50RWC and LD50Ψ, than species restricted to wetter forests. As expected we found lower SWRWC and SWΨ for the species from the drier side of the Isthmus (Fig. 5). The LD50's of species from the wet and dry sides did not differ, nor was there an obvious trend.
stem hydraulic conductance
The leaf-area-specific stem hydraulic conductance, kws, of well-watered control plants varied among species by a factor of 4·3, from 3·74 to 16·1 × 10−5 kg s−1 m−2 MPa−1. For severely wilted plants, kws, which we scaled to the total leaf area present before plants were desiccated (see Methods), strongly decreased and varied among species by a factor of 9·0, from 0·54 to 4·85 × 10−5 kg s−1 m−2 MPa−1 (Fig. 6, upper panel). In contrast to the large variation among species in kws, the variation in loss of conductance was relatively small: from the well-watered to the severely wilted stage species lost 79·8% (±3·1%, 1 SE) of their leaf-specific stem conductance, with a range of 61·1–95·3% (Fig. 6, lower panel).
Both drought performance (Dp) and SWΨ, measured under field and lab conditions respectively, were highly correlated with kws. Species with higher Dp and lower SWΨ had lower kws in both the well-watered and the severely wilted plants (Fig. 7). There was no relationship of kws with SWRWC, LD50RWC or LD50Ψ.
the contribution of tolerance of low leaf water status to drought performance
Our study species, all common in Central Panama, gave the full range of drought performance, Dp, from 0% to 100% (Engelbrecht & Kursar 2003; Engelbrecht et al. 2007a). Such large differences among species in their performance during drought could be due to a number of mechanisms. One strategy by which species may vary in drought performance, desiccation tolerance, is the focus of this study. All of our measures of the capacity to tolerate low leaf water status, LD50Ψ, LD50RWC, SWΨ and SWRWC, showed wide variation among species and were highly correlated with Dp, as determined in field experiments (Fig. 4). The significant correlation of desiccation tolerance with drought performance in the field has two implications. The results suggest that drought performance was indeed mainly driven by the lack of water and by differences among species in their ability to cope with low water availability, rather than by the multiple further stresses that may act during drought, such as changes in pests or soil nutrients. Second, the results strongly suggest that, in tropical forest seedlings and saplings, desiccation tolerance is a principle strategy by which species cope with low water availability.
A second strategy leading to differences in drought performance is desiccation delay, or avoidance. In other studies we found that several surrogates for desiccation delay, such as seedling size, rooting depth, root : shoot ratio and cuticular conductance, were unrelated to species drought performance (unpublished data). As well, we found that minimum dry-season water potentials in the field, an integrated measure of desiccation delay, did not correlate with either desiccation tolerance or drought performance (unpublished data). We can therefore rule out that a correlation between desiccation tolerance and delay led to a spurious correlation of desiccation tolerance with drought performance and that, in reality, desiccation delay promotes drought performance. Instead the study species cope with low water availability primarily by desiccation tolerance. Nevertheless, for saplings or established trees and for seedlings in habitats that experience very dry surface soil, desiccation delay through deeper roots could be an important survival strategy (Padilla & Pugnaire 2007).
tolerance to low leaf water status as a determinant of distribution in tropical rainforest
Many observations point to drought as a key determinant of tree abundance and distribution in tropical forests (Gentry 1986; Condit 1998; Veenendaal & Swaine 1998; Bongers et al. 1999). We recently showed that seedling drought performance is important in shaping species distributions in tropical forests: species occurring on the drier side of the Isthmus have higher drought performance, Dp, as measured in our field experiments than species restricted to wetter forests (Engelbrecht et al. 2007a). In the present study we hypothesized that desiccation tolerance of seedlings determines drought performance, and thus shapes species distributions. As predicted, we found higher desiccation tolerance for the species occurring on the drier side of the Isthmus than on the wetter side, in particular, lower SWRWC and SWΨ (Fig. 5). The fact that we find a strong relationship between desiccation tolerance and species distribution, even though the gradient in drought length across the Isthmus of Panama is only moderate (from 115 to 150 days, Engelbrecht et al. 2007a), indicates that species are very sensitive to differences in drought regimes. In addition, a recent study in Southeast Asia obtained similar results, with species found in drier forests able to tolerate lower Ψleaf and RWC (Baltzer et al. 2008).
the role of stem hydraulic conductance and vulnerability to cavitation for mortality during desiccation
We found that species with lower leaf-area-specific stem hydraulic conductance, kws, had greater desiccation tolerance and drought performance (Fig. 7). The mechanism behind such a relationship could be a trade-off between greater xylem conductance and lower resistance to cavitation. This mechanism has frequently been suggested although the relationship has not been consistently demonstrated (Brodribb & Hill 1999; Pockman & Sperry 2000; Hacke et al. 2006).
To test whether a threshold loss of about 80% of stem conductance leads to mortality in most species, as suggested by Tyree et al. (2003), we carried out a comparable experiment with seven more species and obtained similar results. The loss of conductance at the severely wilted stage was 79·8% ± 3·1%, with a range of 61·1–95·3% (Fig. 6, lower panel). For these species, a much larger variation was found in desiccation tolerance, with SWΨ values between −2·5 and −9·3 MPa, as well as in kws, with values between 3·74 and 16·1 × 10−5 kg s−1 m−2 MPa−1 (Fig. 7). Our results are consistent with the hypothesis that loss of xylem function may cause death during desiccation (Davis et al. 2002). Hence, the crucial mechanism that allows cell and plant survival despite water loss may be the prevention of xylem embolism. Our results show that, in tropical rainforest habitats, those species with higher leaf-specific shoot conductance (which may trade off with resistance to cavitation) had worse drought performance, were less desiccation tolerance and may be excluded from drier habitats. Contrasting results were obtained in a study of two Mediterranean species in which the species with higher resistance to cavitation had lower seedling survivorship during drought (Vilagrosa et al. 2003).
comparison of the two methods for assessing tolerance to low leaf water status
In this study we assessed desiccation tolerance in two ways: as the water status of plants in a severely wilted state (SWRWC and SWΨ) an approach that was based on an earlier study (Tyree et al. 2003). The second assesses the leaf water status associated with 50% mortality (LD50RWC and LD50Ψ). This approach was developed a priori and is based on well-established concepts of dose-response analysis from toxicology (Collett 2003).
We found more statistically significant relationships with distribution and kws using SWRWC and SWΨ than with LD50RWC or LD50Ψ (Figs 5 and 7), suggesting that the severely wilted method may better reflect the critical physiological processes. To address, in part, the issue that the severely wilted method depends on a subjective, visual scoring of stress, we rewatered severely wilted plants and scored these for survival (see the discussion of Lacmellea, Swartzia and Posoqueria in Methods). The LD50 approach was independent of subjective assessments of plant appearance, so that it might be expected to be the more reliable method. Nevertheless, a major drawback of the LD50 method is that it requires considerably larger samples sizes, 50 plants, to reach reasonably small confidence intervals (vs. 5–10 for the severely wilted approach). Also, for extremely stressed plants, some leaves or leaf segments that initiate abscission may not be in hydraulic contact with the stem, leading to an overestimation of desiccation tolerance (see Results).
While both methods have their drawbacks, the values that we independently obtained by the two methods were highly significantly linearly related with slopes close to 1·0 (Fig. 3c,d). This suggests that both methods assessed very similar physiological states of the plants, and augments our confidence in both methods. The relationships with ecological parameters (drought performance and distribution) and with physiological parameters (hydraulic conductance) furthermore suggest that they are useful for resolving ecological and physiological questions.
Ours and other studies indicate that drought significantly impacts the dynamics, distribution and habitat specialization of tropical rainforest species. Drought also may significantly impact productivity and carbon balance at the ecosystem level (Schuur 2003). At the same time, climate projections for tropical regions include increased evapotranspiration, changes in drought conditions and large changes in local rainfall resulting from displacements in tropical rain belts (Christensen et al. 2007). Our ability to predict how such changes will impact tropical forests will be of decisive ecological significance. Hence relating desiccation tolerance with distribution, such as described herein, will become increasingly necessary.
Authors thank N. Anaya, M. Rios, G. Rivas, R. Perez, S. Aguilar, and T. Brenes-Arguedas for assistance with determining species distributions, D. Feener for statistical advice, J. Sperry for comments on the text and the Autoridad Nacional del Ambiente of Panama for research and collection permits. The project was funded by the Andrew W. Mellon Foundation and the USDA Forest Service.