Lianas and Trees From a Seasonally Dry and a Wet Tropical Forest Did Not Differ in Embolism Resistance but Did Differ in Xylem Anatomical Traits in the Dry Forest

One of the most prominent changes in neotropical forests has been the increase in abundance and size of lianas. Studies suggest that lianas have more acquisitive strategies than trees, which could allow them to take advantage of water more effectively when it is available in water-limited forests, but few studies compared across growth form (i.e., lianas vs. trees) and forest type (i.e., wet vs. seasonally dry). We measured hydraulic and anatomical traits of co-occurring lianas and trees that convey drought resistance (xylem embolism resistance and intervessel pit membranes) and water transport capacity (xylem vessel diameter and density) in a seasonally dry and a wet evergreen tropical forest to address: (1) Are there differences between vulnerability to embolisms (P50—water potential at 50% loss of hydraulic conductivity) and hydraulic safety margins (HSM) across growth form and forest type? (2) How do vessel diameter and density vary across growth form and forest type? (3) Are there differences in xylem intervessel pit membrane thickness across growth form and forest type and does it predict xylem embolism vulnerability in trees and lianas? We examined hydraulic and xylem anatomical traits of 32 species—eight lianas and eight trees in each forest type. We found no difference in P50 and HSMs between lianas and trees and between the wetter and drier forest. Dry forest lianas had 81% greater maximum vessel diameter and 125% greater range in vessel diameter sizes than dry forest trees but, there was no significant difference between life forms in the wet forest. Dry forest species had 50% greater vessel density and 30% greater maximum pit membrane thickness than wet forest ones. Maximum pit membrane thickness was correlated to P50 and HSMs. The main difference between lianas and trees occurred in the dry forest, where lianas had larger maximum xylem vessel size than trees, implying that they have proportionally greater hydraulic conductive capacity than the trees in seasonal forests.

One of the most prominent changes in neotropical forests has been the increase in abundance and size of lianas. Studies suggest that lianas have more acquisitive strategies than trees, which could allow them to take advantage of water more effectively when it is available in water-limited forests, but few studies compared across growth form (i.e., lianas vs. trees) and forest type (i.e., wet vs. seasonally dry). We measured hydraulic and anatomical traits of co-occurring lianas and trees that convey drought resistance (xylem embolism resistance and intervessel pit membranes) and water transport capacity (xylem vessel diameter and density) in a seasonally dry and a wet evergreen tropical forest to address: (1) Are there differences between vulnerability to embolisms (P 50 -water potential at 50% loss of hydraulic conductivity) and hydraulic safety margins (HSM) across growth form and forest type? (2) How do vessel diameter and density vary across growth form and forest type? (3) Are there differences in xylem intervessel pit membrane thickness across growth form and forest type and does it predict xylem embolism vulnerability in trees and lianas? We examined hydraulic and xylem anatomical traits of 32 species-eight lianas and eight trees in each forest type. We found no difference in P 50 and HSMs between lianas and trees and between the wetter and drier forest. Dry forest lianas had 81% greater maximum vessel diameter and 125% greater range in vessel diameter sizes than dry forest trees but, there was no significant difference between life forms in the wet forest. Dry forest species had 50% greater vessel density and 30% greater maximum pit membrane thickness than

INTRODUCTION
Anthropogenic climate change has been causing shifts in rainfall patterns that are altering ecosystems worldwide and severe drought events have led to widespread forest mortality across many ecosystems, including tropical forests (Anderegg et al., 2012;Allen et al., 2015;Aleixo et al., 2019;Powers et al., 2020). Tropical ecosystems store the largest quantity of terrestrial aboveground biomass (Bonan, 2008;Pan et al., 2011) and rising tropical forest mortality could lead to large carbon losses to the atmosphere, turning these forests from a carbon sink to a source, and exacerbating climate change (Gatti et al., 2014). Because natural drought events are anticipated to become more severe and last longer (Duffy et al., 2015;Chadwick et al., 2016), high rates of forest mortality are likely to persist and even rise (Brodribb et al., 2020b). Increases in tree mortality and growth declines have been associated with severe drought across several neotropical forests (Chazdon et al., 2005;Uriarte et al., 2016;Leitold et al., 2018;Powers et al., 2020). Drought impacts, however, differ among species, and rates of mortality and declines in growth depend on a suite of species' non-independent functional traits, including hydraulic traits. Groups of species with the same growth form may also share traits leading them to perform similarly when water is limited, and this may result in, for example, lianas being less impacted from seasonality and drought compared to trees (Schnitzer, 2005;Schnitzer and Bongers, 2011;Schnitzer and van der Heijden, 2019;Smith-Martin et al., 2019). Yet, growth form-level responses to drought are still poorly understood . Accurately predicting how tropical forests respond to drought, including more frequent and severe drought events, requires a deeper understanding of not only species-level but also growth formlevel responses to water deficit.
Lianas are one of the most abundant life forms, second only to trees, in many tropical forests (Gentry, 1991(Gentry, , 1995 representing up to 35% of woody species (Schnitzer et al., 2012) and up to 40% of woody stems (Schnitzer and Bongers, 2011). Contrary to most vascular plant groups which are more abundant in wetter tropical forests compared to drier ones, lianas are particularly dominant in forests that experience seasonal drought (Gentry, 1995;Schnitzer, 2005;DeWalt et al., 2015;Schnitzer, 2018). One of the most notable changes in neotropical forests over the past several decades has been the increase in abundance and size of lianas (Phillips et al., 2002;Wright et al., 2004;Schnitzer and Bongers, 2011;Yorke et al., 2013). The leading explanation for this increase in lianas is the seasonal growth advantage hypothesis, which states that plant species that grow well during seasonal drought obtain a growth advantage in forests with increasing seasonality relative to co-occurring species that grow poorly during seasonal drought (Schnitzer, 2005(Schnitzer, , 2018Schnitzer and van der Heijden, 2019). Many lines of evidence support the seasonal growth advantage hypothesis; trees seem to suffer more from lack of water during seasonal drought than lianas and lianas may maintain better whole-plant water status (Smith-Martin et al., 2019) and grow more than trees during drought (Schnitzer and van der Heijden, 2019). Moreover, some studies have shown that lianas have more acquisitive strategies than trees (Asner and Martin, 2012;Werden et al., 2017;Smith-Martin et al., 2019;Medina-Vega et al., 2021a), which could allow them to take advantage of water more effectively when it is available in water-limited forests. However, liana resource acquisition strategies may differ depending on rainfall regimes. For example, lianas have been found to have "cheaper" and more efficient leaves in a drier forest compared to trees, whereas this difference did not exist in a wet forest (Medina-Vega et al., 2021a). Despite some evidence that differences between lianas and trees may be more pronounced in drier forests (Medina-Vega et al., 2021a), few studies have explored the physiological and anatomical traits that underly the differences in water use of lianas and trees in wetter and drier forests.
Functional traits related to water movement and drought resistance can shed light on performance under different water availability regimes. One of the mechanisms involved in droughtinduced plant mortality is the catastrophic failure of the plant hydraulic system caused by embolisms in xylem conduits that restrict the movement of water to the leaves (Sperry et al., , 2002Sperry and Saliendra, 1994;Barigah et al., 2013;Hochberg et al., 2017;Johnson et al., 2018;Rodriguez-Dominguez et al., 2018;Powers et al., 2020;Brodribb et al., 2021;Johnson et al., 2022). Vascular plants transport water under negative pressure from the roots to the leaves through xylem conduits (Dixon and Joly, 1895;Pockman et al., 1995;Angeles et al., 2004). The continuous column of water in the plant vascular system exists in a metastable state because of the negative pressure of the water column (Dixon and Joly, 1895;Pockman et al., 1995;Angeles et al., 2004). During drought, this pressure becomes more negative, increasing the probability of embolisms being formed and propagating through "air-seeding" from a gas-filled conduits to neighboring, sap-filled ones via bordered pits with mesoporous pit membranes that have pore sizes between 5 and 50 nm (Zimmermann, 1983;Lewis, 1988;Tyree and Sperry, 1989;Sperry and Saliendra, 1994;Jansen et al., 2018;Kaack et al., 2019). This results in the blockage of xylem conduits by gas emboli (Zimmermann, 1983;Lewis, 1988;Tyree and Sperry, 1989;Sperry and Saliendra, 1994;Jansen et al., 2018;Kaack et al., 2019) and can ultimately lead to hydraulic failure (Brodribb and Cochard, 2009;Brodribb et al., 2010Brodribb et al., , 2020bBarigah et al., 2013;Cochard and Delzon, 2013).
Xylem anatomical characteristics may vary depending on water availability (e.g., rainfall regimes) and on the resource use strategy of a plant. First, xylem vessel diameter is a key factor in water transport efficiency. The hydraulic conductance of a xylem conduit is theoretically equal to the fourth power of the diameter (Zimmermann, 1983;Tyree and Ewers, 1991), meaning that larger vessels have much higher potential conductivity than ones with smaller diameters. However, high water transport efficiency, and thus, high photosynthetic capacity is often thought to come at the cost of increased vulnerability to xylem embolism (Hargrave et al., 1994;Brodribb and Feild, 2000;Hubbard et al., 2001;Martínez-Vilalta et al., 2002;Hacke et al., 2006;Lens et al., 2011;Markesteijn et al., 2011). Vessels with larger diameters are expected to be more vulnerable to embolism than narrower vessels (Carlquist, 1988;Hargrave et al., 1994;Hacke et al., 2017;Jacobsen et al., 2019) although there are also papers suggesting that conduit diameter does not affect embolism resistance (Ryu et al., 2016;Guan et al., 2022). Larger vessels are thought to be at greater risk of implosion due to their larger lumen diameter (Hacke et al., 2001), and these vessels are speculated to be more prone to air-seeding because of their larger pitted wall areas (Hargrave et al., 1994;Jarbeau et al., 1995;Wheeler et al., 2005;Christman et al., 2009Christman et al., , 2012. Lianas have been shown to have large vessel diameter distributions within the same individual (Ewers et al., 1990;Angyalossy et al., 2012Angyalossy et al., , 2015Rosell and Olson, 2014;Meunier et al., 2020), but comparative data on vessel diameter distributions from co-occurring liana and tree species are lacking. Second, vessel density may also vary depending on the rainfall regime, as the density of xylem vessels is negatively related to groundwater availability (Schume et al., 2004). A third anatomical trait that may vary depending on water availability and hydraulic strategy are pit membranes in bordered pits. For water to move between two adjacent vessels it must pass through pit membranes (i.e., a mesh of cellulose microfibrils) in bordered pits that play an important role in restricting the spread of embolisms between conduits, while contributing to the hydraulic resistance of water transport (Choat et al., 2008;Kaack et al., 2019). It has long been suggested that pit membrane thickness is a key characteristic in determining embolism resistance (Jansen et al., 2009;Lens et al., 2011Lens et al., , 2013Li et al., 2016;Dória et al., 2018;Kaack et al., 2021) and some previous studies have found an association between pit membrane thickness and embolism resistance (Dória et al., 2018;Jansen et al., 2018;Trueba et al., 2019;Levionnois et al., 2020a;Kaack et al., 2021), although pit membrane thickness in lianas was not examined in these studies.
In summary, the distribution patterns of lianas relative to trees across rainfall gradients are typically explained by the hypothesis that lianas have different, "more acquisitive" trait values compared to trees in drier forests, but these differences are less pronounced in wetter or less seasonal forests. However, the physiological and anatomical traits that underlie these patterns remain unclear. To elucidate the underlying mechanisms that lead to lianas outperforming trees in drier but not wetter forests we used targeted hydraulic and anatomical trait measurements of co-occurring lianas and trees in two tropical forests with different levels of rainfall to address the following questions: (1) Are there differences between vulnerability to embolisms (P 50 -water potential at which a plant has lost 50% hydraulic conductivity) and risk of hydraulic failure (HSM-hydraulic safety margins) across growth form and forest type? (2) How do vessel diameter and density vary across growth form and forest type? (3) Are there differences in xylem inter-vessel pit membrane thickness across growth form and forest type and does it predict xylem embolism vulnerability in trees and lianas? We expected that within each forest lianas will have more acquisitive traits associated with greater hydraulic conductive capacity than trees at the expense of hydraulic safety-e.g., thinner pit membranes, larger xylem vessels, greater vulnerability to embolism, and narrower hydraulic safety margins.

Study Sites
We conducted this study in two lowland forests with distinct rainfall that were located on opposite sides of the Isthmus of Panama: Parque Natural Metropolitano (PNM) a seasonally dry tropical forest referred to as drier forest ( Figure 1A) and Parque Nacional San Lorenzo (PNSL) a wet evergreen tropical forest referred to as wetter forest ( Figure 1B). At each location, the Smithsonian Tropical Research Institute has a tower crane that is used to access the forest canopy. PNM (8 • 59 41.55 N, 79 • 32 35.22 W) is a secondary forest located on the Pacific side of the Isthmus and is 30 m above sea level. At PNM, the mean temperature is 26 • C and annual rainfall is 1,911 mm with a distinct dry season from January to April with over 90% of the rainfall occurring between May and December (Parolari et al., 2020). While we were conducting the study at PNM, the canopy crane was being repaired so we had to collect all the samples from the ground with a pruner attached to extension poles at PNM and nearby Parque Nacional Soberania near El Charco (PNS; 9 • 05 06.1 N, 79 • 39 57.9 W), which has a similar seasonal rainfall pattern compared to PNM with a slightly higher mean annual rainfall of 2,132 mm (Parolari et al., 2020) and also shares many of the same plant species with PNM. We collected exclusively sun-exposed samples using a heavy-duty pruner head (Jameson, Clover, SC, United States) mounted on three or four (depending on sample height) 6-feet (1.83 m) fiberglass poles allowing us to sample at heights over 5 m. Due to the seasonally dry tropical forest having a more open canopy this facilitated the collection of sun-exposed branches. PNSL (9 • 16 51.71 N, 79 • 58 28.27 W) is an old-growth forest on the Caribbean side of the Isthmus of Panama and is 130 m above sea level. The mean temperature at PNSL is 25.3 • C with a mean annual rainfall of 3,236 mm with a period of reduced rainfall from January to March (Parolari et al., 2020). The canopy crane at PNSL is 52 m tall and has a jib of 54 m (Slot and Winter, 2017). While we recognize that there could be sampling error due to the samples at PNM and PNS being collected from the ground and the ones from PNSL being collected from the canopy crane, we minimized this error as much as possible by exclusively sampling sun-exposed branches at all locations.

Species and Sample Collection
At each study site, we selected eight species of lianas and eight species of trees, for which we could access at least two individuals from the canopy cranes or from the ground with a pole pruner, for a total of 32 species from 24 plant families (Table 1). For the optical vulnerability curves, we collected sun-exposed branches that were >1.5 m long, immediately sealed the branches in large plastic bags with wet paper towels, and took them back to the Smithsonian Tropical Research Institute Gamboa Lab. For the transmission electron microscopy (TEM) imaging, we collected one ∼10 cm long and ∼1 cm in diameter branch section for each species (32 samples in total). As we collected each sample, we immediately wrapped them in moist paper towels and sealed them in Ziploc bags. All TEM samples were collected over a 48h period and shipped fresh with 3-day shipping service to Ulm University, Germany.

Leaf and Stem Optical Vulnerability Curves
We used the optical vulnerability technique to measure xylem embolism accumulation in leaves and stems as described by Brodribb et al. (2016Brodribb et al. ( , 2017. We chose to use this technique as it has been validated in many studies on many species (Brodribb et al., 2016(Brodribb et al., , 2017(Brodribb et al., , 2020aSkelton et al., 2018;Gauthey et al., 2020;Johnson et al., 2020;Pereira et al., 2020;Guan et al., 2021) with strong agreement in all but one case . Briefly, for each branch, we secured a leaf and a small distal stem (∼3-6 mm in diameter depending on the species) inside a custom-built 3D-printed clamp (OpenSourceOV-OSOV) fitted with a small 8-megapixel Raspberry Pi camera and six bright light-emitting diodes operated by a Raspberry Pi microcomputer. Details of materials and instructions for construction, image capturing, and post-image processing using OSOV clamps are explained in detail at http://www.opensourceov.org. For the stems, we carefully removed a small area of bark (∼2 cm 2 ) from the small distal branch with healthy foliage to expose the xylem. Then we applied adhesive hydrogel (Tensive, Fairfield, NJ, United States) to the exposed xylem to reduce surface reflection and aid light penetration into the xylem, then we covered the area with a round glass coverslip, and secured the branch in the OSOV clamp. Once the leaf and stem were secured in the clamps, we set them to take a picture every 2 min until no embolism events were recorded for at least 12 h (∼72-96 h depending on the species) in a laboratory with controlled temperature that was maintained at around 23 o C. As the branches dried, we used a Pressure Chamber Instrument (PMS Instrument Company, Albany, OR, United States) to measure leaf water potential ( ) of excised leaves from the same branch. We used ImageJ software to analyze the pictures of leaf and stem embolisms following Brodribb et al. (2016Brodribb et al. ( , 2017 and as described in great detail at http://www.opensourceov.org. Briefly, we stacked all the images and converted the images to 8-bit grayscale with pixel values ranging from black (0) to white (255). Each image was subtracted from the next image in the sequence to reveal embolisms that appear as changes in light intensity (differences in pixel values). We removed manually differences in pixels due to noise or artifacts (e.g., sample movements or shrinkage) using the remove outlier function in ImageJ. Embolism accumulation in each stem was quantified as a cumulative total of embolized pixels in each image divided by the total number of embolized pixels in the fully dried sample (cumulative percentage xylem embolism). To determine the water potential at the time of image capture ( x ), we fit a linear regression to the water potential measurements over time and extracted the values at 50% of the cumulative embolisms that had occurred (P 50 ). Full details of the procedure, including an overview of the technique, image processing, as well as ImageJ scripts, are available at http://www.opensourceov.org.

Vessel and Pit Membrane Thickness Imaging and Measuring
The samples were prepared following standard TEM techniques (Jansen et al., 2009Scholz et al., 2013;Li et al., 2016). Briefly, stem sections were debarked and small sections were cut from the outermost sapwood. To avoid dehydration, the sections were kept wet and the initial steps of sample preparation were with a JEOL 1400 TEM (JEOL, Tokyo, Japan). TEM images of a minimum of 20 bordered intervessel pits per species were taken with a digital camera. We measured the thickness of each pit membrane at five different locations including at the thinnest and at the thickest point using ImageJ software (National Institutes of Health, Bethesda, MD, United States). Using the images of the semi-thin sections, we measured vessel diameter and vessel density. We strived to measure at least 20 xylem vessels per species although for four species we measured fewer. The number of vessels measured per species ranged from 14-173 vessels.

Data Analysis
Of the 32 species that we measured, we obtained optical vulnerability curves (OVC) for between 1 and 4 individuals for 28 of the species (Supplementary Table 1). We obtained leaf OVCs for 27 of the species and stems OVCs for 21 species (Supplementary Table 1). For the additional samples we measured we were unable to detect embolisms in the image stacks, thus we were unable to construct OVCs for these samples and had to discard these measurements. We successfully measured vessel diameter, vessel density, and pit membrane thickness on 31 of the 32 target species-for one species (Stizophyllum riparium) we were unable to prepare semithin wood sections for the wood anatomy measurements because the wood sections kept disintegrating (Supplementary Table 1). Among our focal species, eight (Coccoloba excelsa, Miconia minutiflora, Pittoniotis trichantha, Stigmaphyllon hypagyreum, Tachigali versicolor, Terminalia amazonia, Tocoyena pittieri, and Vochysia ferruginea) had vestured pits (protuberances of lignified cell wall on the borders of the pits) and the other 23 species did not. Because Levionnois et al. (2020a) found that species with vestured pits had thinner pit membranes than species without vestured pits, we conducted Welch's t-tests to determine if there was a significant difference between the pit membrane thickness of vestured-pit species (eight species) and non-vestured-pit species (23 species). We found no difference between vestured and non-vestured species [median pit membrane thickness t (10) = 0.59, p = 0.569; minimum pit membrane thickness t (10) = 0.40, p = 0.694; maximum pit membrane thickness t (10) = 0.88, p = 0.400], thus we omitted this variable (vestured pits or not) from all further analyses. For each species, we calculated minimum, median, maximum pit membrane thickness-these values were obtained by averaging, for each species, the minimum, median, and maximum pit membrane thickness values obtained from each of the individually measured pit membranes (at least 20 measured pits per wood sample and one sample per species). We calculated hydraulic safety margins as the difference between leaf turgor loss point ( tlp ), which has been shown to be a good proxy for stomatal closure (Brodribb et al., 2003;Rodriguez-Dominguez et al., 2016;Martin-StPaul et al., 2017), and the water potential at which 50% of total embolisms had occurred: For these calculations, we used previously collected tlp that had been measured on the same species at the same two study Frontiers in Forests and Global Change | www.frontiersin.org sites (Medina-Vega et al., 2021a,b). For all analyses with P 50 and HSM, we used species-level means for leaves and for stems or if there was only one measurement for a particular species then we used that value. We used two-way ANOVAs with forest type, growth form, and the interaction between these two variables as predictors and the response variables used were median, minimum, and maximum xylem pit membrane thickness, maximum vessel diameter, range in vessel diameter, mean vessel diameter, minimum vessel diameter, mean vessel density, leaf P 50 , stem P 50 , leaf HSM, and stem HSM. All ANOVAs were fit with the Anova function from the R/car package. When necessary, to be able to interpret significant differences, post-hoc Tukey's HSD tests were conducted with R/EMMEANS package. To link anatomical characteristics to metrics of vulnerability to drought, we used linear regressions. We evaluated the association between median, minimum, and maximum pit membrane thickness and the drought resistance traits: leaf P 50 , stem P 50 , leaf HSM, stem HSM. We also fit linear regressions between leaf P 50 and stem P 50 and the xylem vessel anatomical measurements: mean, minimum, maximum, and range in vessel diameter and mean vessel density. All analyses were conducted in R (Version 4.0.5).

P 50 and Hydraulic Safety Margins
Interestingly, we found no significant difference in P 50 and HSM between lianas and trees within each forest nor across the two forest types (Figures 2A,B, Supplementary Figures 1A,B, and Supplementary Table 2). Overall, species fell along a gradient of P 50 (Supplementary Figure 2A) and HSM values (Supplementary Figure 2B)

Vessel Diameter and Density
Dry forest lianas had 81% greater mean maximum vessel diameter across species than co-occurring dry forest trees and 68% wider maximum vessel diameters than the wet forest trees (Figure 3A and Supplementary Table 2). There was no difference in maximum vessel diameters between wet forest lianas and the trees from both the wet and dry forest ( Figure 3A and Supplementary Table 2). The range in xylem vessel diameter followed a similar pattern with dry forest lianas having between 110-125% greater range in size of vessel diameters than trees independent of forest type, and again the wetter forest lianas fell in between the dry forest lianas and all the trees (Figure 3B and Supplementary Table 2). We found no significant difference between liana and tree mean and minimum vessel diameters within each forest type and across the two forests (Supplementary Figures 3A,B). Lianas and trees from the drier forest had over a 50% greater vessel density than species from the wetter forest ( Figure 3C and Supplementary Table 2). There was no significant association between leaf and stem P50 with xylem vessel diameter or vessel density (Supplementary Figures 4A-E).

Pit Membrane Thickness
Overall, drier forest species consistently had on average 30% greater maximum pit membrane thickness than wetter forest species, with similar results for median and minimum pit membrane thickness (Supplementary Table 2 and Figure 4). Within each forest type, there was no difference between lianas and trees (Supplementary Table 2). When comparing pit membrane thickness among growth form and forest type, lianas and trees from the drier forest had significantly thicker pit membranes than the trees in the wet forest but not greater than the wet forest lianas (Figures 5A-C and Supplementary Table 2). Species with thicker pit membranes were less vulnerable to embolism and less at risk of hydraulic failure. Species with more negative leaf and stem P 50 values also tended to have thicker pit membranes ( Figure 6A and Supplementary Table 3). Species with thicker pit membranes also tended to have wider hydraulic safety margins ( Figure 6B and Supplementary Table 3).

DISCUSSION
We examined hydraulic and xylem anatomical traits of 32 woody plant species-eight species of lianas and eight species of trees growing in a drier and more seasonal tropical forest on the dry end of the rainfall gradient across the Isthmus of Panama and eight liana and eight tree species from the wettest end of the rainfall gradient. We found no difference in embolism resistance and hydraulic safety between lianas and trees. The main difference between lianas and trees was in vessel diameter and not P 50 , HSM, and pit membrane thickness. Lianas tended to have larger maximum vessel diameter and a greater range in vessel diameter sizes than trees in the seasonally dry forest but not in the wet forest, suggesting that lianas are able to make larger vessels in the dry forest without an increase in the risk of embolism formation. Dry forest species in general had a higher vessel density than the wet forest species. We found thicker pit membranes among lianas and trees from the drier forest compared to trees in the wetter forest; however, there was no difference between lianas from the drier and wetter forest, nor between lianas and trees within each forest. Our results show an association between thicker pit membranes and greater resistance to embolisms (P 50 ) and hydraulic failure (HSM); but despite this association, species fell along a gradient of P 50 and HSMs with no significant difference between the wetter and drier forest species nor between lianas and trees. Overall, the main difference between lianas and trees occurred in the drier forest where lianas have larger maximum xylem vessel size and a greater range in vessel sizes than the trees, implying that they have greater hydraulic conductive capacity than the trees. Being able to move larger quantities of water when it is available in seasonally dry tropical forests without an apparent increase in vulnerability to embolisms could be contributing to lianas outperforming trees and contributing to the increase in lianas that has been documented in many Neotropical forests.

Gradient in Resistance to Embolisms and Hydraulic Failure Among the Wetter and Drier Forests
Lianas and trees from the two forest sites fell along a gradient of P 50 and hydraulic safety margins. Despite the finding that dry forest species tended to have thicker pit membranes and that there was an association between greater pit membrane thickness and greater embolism resistance, this did not translate into significant differences in P 50 and HSMs between species from the two forests. Consistent with our findings, several studies have observed a wide range in P 50 and HSM values within tropical forest communities elsewhere (Santiago et al., 2018;Barros et al., 2019;Oliveira et al., 2019Oliveira et al., , 2021van der Sande et al., 2019;Ziegler et al., 2019;Fontes et al., 2020;Powers et al., 2020;Vargas et al., 2021). The wide range in HSMs indicates that not all the species within our two studied forests function at the edge of their hydraulic capacity. This is at odds with the expectation that species should operate with narrow HSMs to sustain CO 2 assimilation for as long as possible before closing their stomata as water becomes more limited Choat et al., 2012;Barros et al., 2019;Fontes et al., 2020). A growing number of studies show that there is a lot more variation in terms of hydraulic safety within forests, even at locations where water is not limited (Santiago et al., 2018;Barros et al., 2019;Oliveira et al., 2019Oliveira et al., , 2021van der Sande et al., 2019;Ziegler et al., 2019;Fontes et al., 2020;Powers et al., 2020;Vargas et al., 2021), suggesting that within-community variation in embolism resistance and hydraulic safety is more widespread than previously thought. This is important also because Anderegg et al. (2016) showed that communities with greater hydraulic diversity are more resilient to drought.

Differences in Xylem Vessel Size Between Lianas and Trees
The most striking anatomical difference that we found between lianas and trees was in the vessel diameter distributions: dry forest lianas had much greater maximum vessel diameters and a greater range in vessel diameter size than the trees in both forests. The   dry forest lianas had 81% greater maximum vessel diameter than co-occurring dry forest trees, 68% higher than wet forest trees, and 125 and 110% respectively greater vessel diameter range than the dry forest and wet forest trees. Because lianas are structural parasites (Stevens, 1987) they tend to have high canopy to stem ratios compared to trees (Putz, 1983;Ewers and Fisher, 1991;Ewers et al., 1992;Schnitzer and Bongers, 2002). A previous study conducted at the same location and with the same species as our study found that lianas in the dry forest had a lower Huber Value (HV, branch-cross section/leaf area) than the cooccurring trees and the lowest values overall across both the dry and wet forest, whereas lianas and trees in the wet forest did not differ in their HV (Medina-Vega et al., 2021a). The larger vessel diameters we found in the dry forest lianas could explain why they also have the lowest HV because smaller stems diameters could potentially supply water to a greater leaf canopy. Other previous studies have also shown that smaller liana stems support propositionally larger canopies, for example, Ewers and Fisher (1991) showed that within the genus Bauhinia, species with liana growth form had smaller stems proportional to the distal leaf area than trees of the same genus. Similar to our speculations, other authors have also suggested that the proportionally larger canopies of lianas are thought to be made possible due to lianas having, at least a proportion of, longer and wider xylem vessels (Ewers and Fisher, 1991;Angyalossy et al., 2015). We found no difference in mean nor minimum vessel diameter between lianas and trees. Similarly, Werden et al. (2017), also found no difference in mean vessel diameter between dry forest lianas and trees, and in an extensive study of 424 species conducted on Mexico, Rosell and Olson (2014) also found no difference in mean vessel diameter of climbing plants compared to self-supporting ones. Consistent with our findings, Meunier et al. (2020) showed a large range in vessel sizes in French Guiana rainforest liana species, however, this study did not examine the vessel size of co-occurring trees. Rosell and Olson (2014) also found that the climbing plants in their study had, on average, a small number of wider vessels and narrower vessels than those found in self-supporting plants. Taken together these findings imply that the big difference between co-occurring lianas and trees may be attributed to the observation that lianas have some proportion of vessels that are much larger than the largest tree vessels and that the spread or range of vessel diameters from small to large is much greater in lianas than in cooccurring trees; differences that may be particularly accentuated in seasonally dry forests.
In theory, the largest vessels in lianas lead to higher hydraulic conductive capacity than co-occurring trees. Because flow rates increase with the fourth power of vessel diameter (Zimmermann, 1978), larger vessels can theoretically move exponentially more water than smaller ones, meaning that the population of larger vessels found in liana stems would make lianas much more effective at moving water than the co-occurring trees. Putz (1983) found that most of the xylem vessels in (sub)tropical liana stems are functional, whereas in trees, xylem vessels are only functional for a few years (sapwood) and then transition into nonconductive wood that provides structural support; by contrast, the structural needs of lianas are much lower as they rely on trees for support potentially allowing their vessels to be active for longer. Because lianas have some large vessels and most vessels are conductive, each increment in a liana's stem diameter can, in theory, supply water and nutrients to proportionally more leaves than an equivalent stem increment in a tree (Putz, 1983). Indeed, previous studies have found that lianas have greater stomatal conductance than trees growing under the same environmental conditions (Johnson et al., 2013;Smith-Martin et al., 2019) and higher stem hydraulic conductance (De Guzman et al., 2016). This proportionally greater hydraulic capacity of lianas than trees may be contributing to their competitive advantage in seasonal forests allowing them to take advantage of water when it is available, which is consistent with a more resource-acquisitive strategy. However, a recent study with grapevine (Vitis vinifera) using reconstructed xylem vessel network from X-ray microcomputed tomography and in vivo magnetic resonance velocity flow imaging found that the heterogeneity of the vessel network leads to sap flowing from wide to narrow vessels, such that the wide vessels only accounted for 15% of the total sap flow (Bouda et al., 2019).

Dry Tropical Forest Species Had Thicker Pit Membranes Than the Wet Forest
Overall dry forest species had on average 30% thicker pit membranes than wet forest species. We found that when growth form was taken into account, lianas and trees from the dry forest had thicker pit membranes than wet forest trees, but the wet forest lianas fell in between the co-occurring wet forest trees and dry forest lianas and trees. Our findings suggest that the trees from the dry forest may have thicker pit membranes as an adaptation to growing in drier environments that experience greater water deficits compared to wet forest trees. Interestingly, lianas do not seem to employ this strategy of thick pit membranes to prevent embolism as there was no difference between dry forest and wet forest lianas, despite the overall 30% higher values. While we did not find a significant difference between the pit membrane thickness of species that had vestured pits compared to the ones that did not, vestured pits is a common occurrence among tropical species (eight out of 31 of or species had them) and is an anatomical feature that could also be contributing to embolism resistance (Jansen et al., 2003;Medeiros et al., 2019). The pit membrane thicknesses that we found for the dry forest species fell on the low end of the values found by Levionnois et al. (2020a) in a wet forest in French Guiana with similar mean annual rainfall (3,102 mm) as our wet forest site (3,236 mm). Levionnois et al. (2020a) found maximum pit membrane thicknesses that ranged from 400 nm to over 1,000 nm, whereas the values that we record for our dry forest species fell in between 300 and 600 nm and our wet forest species have overall lower values than what was found in French Guiana. Pit membrane thickness seems to be sitespecific with some wetter forest species tending to have thicker pit membranes than species from other forests with similar rainfall. Similar to Levionnois et al. (2020a) we found that pit membrane thickness is species-specific with some species in the wet forest having thicker pit membranes than species in the seasonally dry forest. However, independent of species-specific and sitespecific differences, our findings do show that when comparing within relatively smaller geographic ranges (across the isthmus of Panama) there is an overall increase in pit membrane thickness with a decrease in rainfall.
Species with thicker pit membranes tended to be more resistant to drought-induced xylem embolisms, as measured by P 50 and HSM. Our findings reinforce the expected importance of pit membrane thickness for preventing air-seeding in tropical forest species both in wetter and drier forests. Our findings align with the findings from two previous studies, which also found an association between pit membrane thickness and resistance to embolisms (Trueba et al., 2019;Levionnois et al., 2020a). Studying species in the wet forest of French Guiana, Levionnois et al. (2020a) found a high association between P 50 and maximum pit membrane thickness (R 2 = 0.53) for 15 species with nonvestured pits as they had found that species with vestured pit had significantly thinner pit membranes. The association we found between maximum pit membrane thickness and P 50 values for leaves (R 2 = 0.36) and stems (R 2 = 0.43) was lower but still notable for our 31 species, although, because we found no significant difference between vestured and non-vestured species we did not remove vestured species. The range in P 50 values was also greater among the French Guiana species, with species with P 50 values as low as −8 MPa meaning that this forest has some highly drought-resistant species (Levionnois et al., 2020a,b), whereas only one of our focal species had a P 50 of −5 MPa and most of our studied species had higher values. Thus, the thinner pit membranes that we found among our focal species align with the recorded higher P 50 values, and the thicker pit membranes among the French Guiana species align with the more negative P 50 values found for those species (Levionnois et al., 2020a,b). One other study also examined the association between pit membrane thickness and P 50 in vessel-less and vesselbearing species from rainforests in New Caledonia (Trueba et al., 2019). In this study, Trueba et al. (2019) found a similar overall association between pit membrane thickness and P 50 (R 2 = 0.34), although this association was much higher for the seven vesselbearing species (R 2 = 0.95). Overall, pit membrane thickness seems to play an important role in preventing embolisms in tropical species as had also been shown for temperate species (Dória et al., 2018;Jansen et al., 2018;Kaack et al., 2021).

CONCLUSION
We conclude that: (1) there was no significant difference of P 50 and HSMs among forests and life forms, with species from both forests falling along a gradient of levels of resistance to embolisms and hydraulic failure. However, (2), lianas in the drier forest did have greater maximum vessel diameter and greater range in vessel diameter, which could allow them to be more efficient at moving water with their larger vessel. (3) Having thicker pit membranes is a potential adaptation of species in drier forests for withstanding greater levels of water deficit, which species in wetter forests would not experience. (4) Lianas did not have thinner pit membranes than trees within each forest, contrary to what we expected due to their predicted higher conductive capacity. (5) Thicker pit membranes are related to greater resistance to embolisms (P 50 ) and hydraulic failure (HSM), underscoring the role of this trait in resistance to embolisms and hydraulic failure. The overlapping hydraulic traits between forests and among life forms suggests a range of responses of woody plant species to changing rainfall conditions, which ultimately may feedback to alter community composition.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AUTHOR CONTRIBUTIONS
CS-M, SJ, TB, CL, and JP designed the research. CS-M and JM-V collected data in the field. CS-M, SJ, and AH performed TEM sample preparation, imaging, and processing. CS-M analyzed data and wrote a draft of the manuscript with substantial input from SJ, JM-V, TB, CL, and JP. All authors contributed to the article and approved the submitted version.