Can Functional Traits Explain Plant Coexistence? A Case Study with Tropical Lianas and Trees

Organisms are adapted to their environment through a suite of anatomical, morphological, and physiological traits. These functional traits are commonly thought to determine an organism’s tolerance to environmental conditions. However, the differences in functional traits among co-occurring species, and whether trait differences mediate competition and coexistence is still poorly understood. Here we review studies comparing functional traits in two co-occurring tropical woody plant guilds, lianas and trees, to understand whether competing plant guilds differ in functional traits and how these differences may help to explain tropical woody plant coexistence. We examined 36 separate studies that compared a total of 140 different functional traits of co-occurring lianas and trees. We conducted a meta-analysis for ten of these functional traits, those that were present in at least five studies. We found that the mean trait value between lianas and trees differed significantly in four of the ten functional traits. Lianas differed from trees mainly in functional traits related to a faster resource acquisition life history strategy. However, the lack of difference in the remaining six functional traits indicates that lianas are not restricted to the fast end of the plant life–history continuum. Differences in functional traits between lianas and trees suggest these plant guilds may coexist in tropical forests by specializing in different life–history strategies, but there is still a significant overlap in the life–history strategies between these two competing guilds.


Introduction
Plant and animal species are uniquely adapted to both the abiotic and biotic components of their environment. Many of these adaptations are known as "functional traits" and are specific anatomical, morphological, and physiological characteristics that define species' performance and fitness, and may ultimately determine species' coexistence [1,2]. Functional traits may explain how species tolerate their environment and utilize resources, as well as compete with co-occurring species [2]. Over the past two decades, there has been a surge of research on functional traits, which has advanced our understanding of plant life-history strategies and the ecological differences among species [3][4][5][6].
One of the ultimate goals of the functional trait approach is to provide a mechanistic explanation for how species are adapted to their environment and for species coexistence in ecological communities [2,7]. Theoretically, differences in functional traits can explain species coexistence if these differences allow species to partition resources to limit competition and prevent competitive exclusion [7,8]. For example, species with functional traits related to efficient uptake and utilization of one set of resources may studies on Web of Science and Google Scholar using the key search terms "tree", "liana" and "functional trait". The Web of Science search retrieved 76 studies in total. The Google Scholar search retrieved 1210 results and included all of the search results that we found on Web of Science. Using the three key terms in combination with additional terms did not improve the total number of studies comparing liana and tree functional traits retrieved by the search engines. We also included two additional studies that compared functional traits between lianas and trees that were present in the LianaEcologyProject.com database, but that were missed by the other two search engines [25,26]. We restricted our selection to studies that measured functional traits in co-occurring species of both lianas and trees in tropical forests. We compiled all the traits measured in each study into our database (Table 1 and Table S1). We restricted our analyses to the functional traits that were measured (and thus served as replicates) in at least five separate studies, which was our minimum level for comparisons between the two growth forms (including single studies that measured functional traits in more than one environment). Thus, our database included a total of 36 studies that compared a total of 140 different functional traits of co-occurring lianas and trees (Table 1 and Table S1).

Data Collection and Analyses
We compiled the mean trait values for 10 functional traits that met our study selection criteria. (1) Specific leaf area (SLA); (2) maximum area-based net photosynthetic rate (A max ); (3) stomatal conductance (gs); (4) mass-based nitrogen concentration (N mass ); (5) mass-based phosphorus concentration (P mass ); (6) sapwood specific conductivity (K s ); (7) wood density (WD); (8) leaf turgor loss point (π tlp ); (9) carbon isotopic composition (δ 13 C); and (10) water potential at 50 percent loss of conductivity (P50). We recorded the mean trait value for different forest types (wet, dry, flood plain or karst forest), successional stage (secondary, mature) and season (wet and dry) analyzed in single studies [34,37,41,48]. In some cases (4 studies), we digitally measured the mean trait values directly from the published figures [32,37,42,52]. We could not retrieve stomatal conductance (gs) data from four studies [26,31,46,55], limiting our analysis of gs to seven studies. We were also unable to retrieve SLA data from one study [31], A max from two studies [31,46]; and K s from one study [53]. The functional trait K s from [39] was not included due to discrepancies in the value range with the other study values.
Some studies reported leaf mass per area (LMA) instead of specific leaf area (SLA) [29,34,39,[41][42][43]46,48,50] and one study [25] measured specific leaf weight (SLW). SLA, SLW and LMA are related traits that represent ratios of area per mass and mass per area, respectively. Since SLA is the inverse of LMA, we transformed leaf mass per area (LMA), including specific leaf weight (SLW), to specific leaf area (SLA) using the equation SLA=1/LMA and compared SLA between the plant guilds for all studies that measured such ratios. In the studies that we included in our analysis, wood density was calculated as the ratio of dry mass to the fresh volume of a stem segment. Even though other studies have used the term "wood specific gravity", which seems appropriate for the methods that were used [60], we chose to be consistent with the cited literature and thus we use the term "wood density" hereafter. Santiago and Wright [16] measured net photosynthetic rate on a mass basis (A mass ); we converted this measurement to a net photosynthetic rate per area basis (hereafter, A max ) by dividing SLA by A mass [47]. Finally, to determine if functional traits between lianas and trees differed significantly, we performed a Student's t-test for paired data for each functional trait (10 traits). We performed the statistical analysis using R [61] and created the figures using Python [62,63].

Results
We assembled a total of 140 different functional traits for tropical lianas and trees from 36 separate studies ( Table 1). The studies were conducted in eleven countries: thirteen of the studies were from forests of Panama, nine were from China, three from Brazil, two from French Guiana and two from Australia, and one study from each of the following: Colombia, Costa Rica, Gabon, India, Madagascar and Mexico. One of the studies, Rios et al. [47], included data from multiple studies, and we used the data that they compiled for four of the studies in China [32,33,58,59]. The 36 tree-liana functional trait studies in our database were conducted in forests that varied widely in annual precipitation and seasonality (Table 1). Chamela, in Jalisco, Mexico, was the driest site with 748 mm of annual precipitation [25] and Daintree rainforest observatory, Australia, was the wettest site, with 4900 mm of annual precipitation [27]. Most of the studies were conducted in seasonal lowland tropical forests, with dry season lengths varying from 3 months in the wet forests of San Lorenzo, Panama, to 6-8 months in the dry-evergreen forest of Coromandel Coast, India and the dry forest of Chamela, Mexico [16,25,48,53,54]. Nineteen studies measured functional traits during the wet season, six studies measured traits during the dry season [26,27,39,52,53,55], and five studies measured traits in both seasons [31,33,34,45,55]. Six studies omitted information on the season when the data were collected [25,36,40,41,44,47].
Of the 140 functional traits, ten traits were measured in five or more different studies. Specific leaf area (SLA) was the most commonly measured trait in the assembled literature, and it was present in 15 studies. Ten studies measured leaf mass per area (LMA), which we converted to SLA, resulting in a total of 25 studies that measured the ratios of leaf area/mass or mass/area. Fourteen studies measured maximum net photosynthetic rates (A max ); thirteen studies measured mass-based nitrogen concentration (N mass ); eleven studies measured stomatal conductance (gs); twelve studies measured wood density (WD); eight studies measured carbon isotopic composition (δ 13 C); seven studies measured sapwood specific conductivity (K s ) and water potential at 50 percent loss of conductivity (P50); and finally, six studies measured leaf turgor loss point (π tlp ); and mass-based phosphorus concentration (P mass ). Interestingly, most of the functional traits used in these studies were physiological traits or traits of considerable measurement complexity rather than the simple and easy to measure traits like SLA and WD [64].
From the 10 functional traits that we analyzed, four showed significant differences between lianas and trees. The mean SLA was significantly greater for lianas than trees (Table 2). Lianas had higher mean values of SLA than trees in twenty of the twenty-four comparisons ( Figure 1A). Both N mass and P mass were significantly greater for lianas than trees (Table 2). Lianas had higher mean values of N mass than trees in ten of the twelve comparisons ( Figure 1B), and higher values of P mass in all six comparisons ( Figure 1C). The functional trait K s was significantly greater for lianas than trees (Table 2) and all five studies supported this conclusion. The mean values of A max did not differ significantly between lianas and trees; values of A max were higher for lianas in 8 of the 13 comparisons (Figure 2A). The five remaining traits showed no significant difference between growth forms (Table 2, Figure 2). of Nmass than trees in ten of the twelve comparisons ( Figure 1B), and higher values of Pmass in all six comparisons ( Figure 1C). The functional trait Ks was significantly greater for lianas than trees ( Table  2) and all five studies supported this conclusion. The mean values of Amax did not differ significantly between lianas and trees; values of Amax were higher for lianas in 8 of the 13 comparisons (Figure 2A). The five remaining traits showed no significant difference between growth forms (Table 2, Figure 2). , Sánchez-Azofeifa et al. [48], and Zhu et al. [57], which measured functional traits in more than one environmental condition. Error bars represent standard errors.  [29] Marechaux et al. [45], and Zhu et al. [57], which measured functional traits in more than one environmental condition or in other parts of the plant.    [29] Marechaux et al. [45], and Zhu et al. [57], which measured functional traits in more than one environmental condition or in other parts of the plant.

Discussion
We found an impressive total number of functional traits (140) in the 36 published comparisons of lianas and trees, which illustrates that ecologists are now measuring a wide variety of liana functional traits. Liana traits have been measured at the leaf, stem, root, and whole-plant levels ( Table S1). The majority of these measured traits were at the leaf level, and included anatomical, morphological, and physiological measurements. There were also many functional trait measurements at the stem level, with the majority focused on stem anatomy and morphology. There were far fewer traits measured at the root level-attesting the difficulty in harvesting plants for roots measurements (e.g., [51]). Even with 140 total traits measured in our collection of 36 published studies, only 10 traits were replicated sufficiently for us to use them in our liana-tree comparisons.
Lianas had significantly higher mean trait values than trees for four of the 10 traits that we examined (SLA, N mass , P mass , K s ). Species with high values for these four traits often have a "fast" life-history strategy, characterized by light-demanding species with quick growth and poor survival, that can dominate recruitment sites in treefall gaps or in the early stages of succession after land abandonment [3][4][5][6]. For example, high values of SLA indicate a lower investment in leaf structural tissues and secondary compounds for defense in exchange for a larger leaf surface area with a higher photosynthetic capacity [4,65,66]. Moreover, high values of leaf nutrient traits such as N mass and P mass are also characteristic of species with higher photosynthesis and a "fast" life-history strategy [3,67]. Photosynthesis is often limited by the foliar concentrations of nitrogen and n-rich compounds (e.g., rubisco), thus nitrogen seems to be less limited for lianas than for trees. In turn, leaf phosphorus content can regulate photosynthetic rates indirectly for a given nitrogen concentration [67]. Functional traits such as SLA, N mass and P mass , all of which were higher for lianas, often show positive correlations with A max values [3]. Therefore, we had expected that lianas would have higher photosynthetic rates by acquiring and using greater quantities of both nitrogen and phosphorus compared to trees [65,67]. However, mean A max values were not significantly different between plant guilds, and lianas had higher A max than trees in only eight of the thirteen comparisons. Nonetheless, higher values of SLA, N mass , and P mass , suggest that lianas tend to reside more towards the "fast" life-history strategy compared to co-occurring trees; a conclusion that would have been stronger had A max also been higher in lianas.
Lianas have significantly higher sapwood specific conductivity (K s ) than trees, demonstrating that lianas can move more water per stem diameter. This striking difference in K s between lianas and trees was also observed in a recent global study where lianas had the highest K s values, on average, compared to other plant functional types [68]. The greater values of K s in lianas are attributed to the presence of remarkably large xylem vessels. Lianas throughout the tropics have a wide range of vessel diameters and lumen sizes, incorporating a combination of both large and small vessels in their stems [69,70]. The large vessels within a liana stem, however, have been shown to contribute the most to plant-level hydraulic conductivity [15,71]. In addition, lianas may have a suite of anatomical functional traits that increase their hydraulic efficiency and, ultimately their hydraulic conductivity, such as low wood density, very long vessel elements, and modified perforation plates on their vessel elements [15].
The higher capacity for lianas to move water would suggest a decrease in hydraulic safety, as suggested by the efficiency-safety trade-off that is evident in most tree species [33,72,73]. Paradoxically, lianas can grow particularly well during dry periods [22] and are also more abundant in drier and highly seasonal forests [23,74,75]. This unusual pattern of high liana abundance in highly seasonal forests, combined with wider vessels in lianas compared to trees [37], seems to contradict the efficiency-safety trade-off. That is, as K s increases, plants can move more water; however, plants also face a higher risk of embolism and cavitation in drier conditions, particularly for plants with large vessels that have wide lumens [14,22,74]. One way that lianas may resolve this paradox is by avoiding embolism and cavitation by carefully controlling the amount of time they open their stomata, and by conducting more photosynthesis early in the day when vapor pressure deficit is low (the fast and furious hypothesis [14]). At some level, all plants, including lianas, must adhere to the efficiency-safety trade-off. Nonetheless, these clear differences between lianas and trees in their capacity to move water may explain their coexistence and relative abundance if each group differs in their ability to exploit resources seasonally, as well as their vulnerability to low levels of these resources [8,22].
The alignment of functional traits along the slow-to-fast life-history continuum may also explain tree-liana coexistence in tropical forests. Both lianas and trees have a range of shade-tolerance and each plant guild includes species that span the fast-slow life-history strategy axis [76,77]. However, our meta-analysis suggests that lianas tend to have a "faster" strategy compared to trees. The relatively high values of K s , SLA, N mass , and P mass shared by lianas are consistent with a "fast" life-history strategy [26,32,37,38,51,53,59]. Indeed, in a comparison of lianas and trees on Barro Colorado Island Panama, the majority of the liana species were associated with disturbance, and treefall gaps appeared to maintain liana diversity, whereas only a small proportion of tree species were maintained by gaps and the vast majority of trees were shade-tolerant [77][78][79]. These differences between lianas and trees along the slow-fast life-history continuum, expressed by their mean trait values, could be enough, theoretically, to allow lianas and trees to coexist.
Plant functional traits such as δ 13 C, gs, P50, and π tlp , which are proxies for hydraulic efficiency and drought tolerance, did not differ between lianas and trees in our analyses. The lack of significant differences for these traits indicates that lianas and trees have considerable overlap in hydraulic life-history strategies. These trait similarities may reflect adaptations to the environmental constraints that co-occurring lianas and trees share. Alternatively, it is possible that the similarities between lianas and trees are biased by the season in which the trait was measured. For instance, Maréchaux et al. [45] observed a seasonal difference in the drought tolerance related trait turgor loss point (π tlp ) for lianas but not for trees, suggesting a seasonal adjustment of this functional trait in lianas. Moreover, Cai et al. [33] also observed a seasonal dependence of the differences in leaf N and P contents between lianas and trees, with meaningful differences only during the dry season but not during the wet season (see also [51]). Thus, measuring functional traits of lianas and trees within and across seasons may further elucidate underlying differences in life-history strategies between these two plant guilds. Additional studies are necessary to evaluate the importance of functional traits such as δ 13 C, gs, P50, and π tlp to explain the coexistence of tropical lianas and trees.
Another dimension in which lianas and trees differ in their life-history strategies, which may explain their coexistence, is the "high leaf efficiency hypothesis". This hypothesis proposes that lianas invest in leaves that are inexpensive and easy to replace but have high photosynthetic efficiency [43,46,50]. Thus, lianas may invest in leaves with an increased surface area for light interception relative to its mass (high SLA) and high leaf nitrogen (N) and phosphorus (P) content to increase photosynthetic rates (A). In fact, the high values of SLA, N mass , P mass , and K s for lianas support the high leaf efficiency hypothesis. By contrast, the mean trait values of tree leaves suggest more investment in structure and defense, which confers longer leaf lifespans [66,77]. Thus, trees appear to have less photosynthetically efficient leaves when compared to lianas, but their leaves are better defended and last much longer [77]. Thus, the tradeoff between leaf cost, leaf efficiency and leaf life span is another way by which lianas and trees differ along the fast-slow life-history continuum.
High leaf efficiency, while beneficial in terms of photosynthetic gain, may leave lianas at a disadvantage compared to trees if lianas are more vulnerable to herbivory [80]. Lianas tend to invest less than co-occurring trees in chemical defenses of their leaves [76]. However, lianas may overcome this increased vulnerability to herbivory by investing in non-chemical traits that are involved in herbivory defense [80]. For instance, Schupp and Feener [81] observed that lianas of Barro Colorado Island, in Panama, have more extra-floral nectaries and other ant-defense mechanisms compared to trees, which could provide an important non-chemical defense strategy. Herbivory may decrease the competitive advantage that lianas have over trees at the leaf level and may facilitate liana-tree coexistence [81]. Additional studies are necessary for a more complete understanding of how trade-offs and ecological interactions differ in liana and tree life-history strategies and how these differences mediate plant coexistence in tropical forests.
In summary, functional traits may explain liana-tree coexistence. Four functional traits (SLA, N mass , P mass , and K s ) were particularly important in defining differences between the two plant guilds. These differences suggest that lianas have a tendency for a "fast" life-history strategy compared to trees, with lianas likely realizing faster growth rates and the capacity to more rapidly capitalize on available resources. Additional studies that directly compare functional traits between lianas and trees in different seasons and environmental conditions may further our understanding of the differences between these two important plant guilds and how they coexist in tropical forests.