Leaf Elemental Concentrations, Stoichiometry, and Resorption in Guam’s Coastal Karst Forests

: Greater knowledge concerning the interspeciﬁc diversity of the plant leaf ionome is required to effectively understand the spatiotemporal dynamics of biogeochemistry, but Micronesia has been ignored in this literature. The objectives of this study were to quantify the leaf ionome, resorption efﬁciency, and stoichiometry of leaves from 25 plant species representing Guam’s coastal karst forests. Carbon and nitrogen were quantiﬁed by dry combustion, and other minerals and metals were quantiﬁed by spectrometry. Nitrogen and calcium concentrations in Guam’s green leaves exceeded the published global means, but manganese and copper concentrations were less than the global means. The remainder of the elements were within the expected ranges. Nutrient resorption rates exhibited a decreasing order of potassium > phosphorus > nitrogen > zinc > copper. The term “accretion efﬁciency” is introduced to describe the accumulation of an element throughout leaf aging and senescence, and calcium and iron exhibited substantial accretion efﬁciency in this study. Stoichiometry relations indicated that Guam’s karst forest is most limited by phosphorus and then secondarily limited by nitrogen, although several individual taxa exhibited co-limitation by potassium. Five of the species are ofﬁcially listed on extinction threat lists. Of these, the Malvaceae tree Heriteria longipetiolata exhibited leaf traits depicting the most recalcitrant litter characteristics, and the Fabaceae tree Serianthes nelsonii exhibited leaf traits depicting the most labile litter characteristics. The contributions of these two tree species to spatiotemporal diversity in biogeochemistry appear to be profound, indicating species recovery efforts are of paramount importance for maintaining ecosystem function and soil heterotroph biodiversity in northern Guam. end of the spectrum for senesced leaf N and near the minimum end of the spectrum for senesced leaf lignin, cellulose, C:N, lignin:N, and cellulose:N. The unique contributions of


Introduction
The uptake of minerals and metals from soils is essential for sustaining plant health and growth. Ionomics is the study of elemental mineral and metal concentrations and relationships in plant tissues [1][2][3]. By comparing relative elemental concentration within and among plants, much can be learned about the physical location of these elements in a forest and the biogeochemical cycling that sustains the health of a forest. Therefore, the study of the mineral nutrition of plants has long been a central component of ecology [4], and is important in several aspects of plant physiology [5].
An adequate understanding of the ionome of co-existing plant species enables the calculation of response variables that increase our understanding of forest community ecology. One example is the comparison of concentrations of pairs of elements among organisms which enables various interpretations in the field of ecological stoichiometry [6][7][8][9]. These methods increase knowledge of how resource availability influences organismal function in complex ecosystems [10]. By directly comparing the relative concentrations of nitrogen, phosphorus, and potassium, an estimation of which of these macronutrients are most limiting of plant growth in a given space and time is achievable [11][12][13][14]. A second example is the calculation of the relative re-mobilization of elements in leaves as they senesce. These elements are returned to stem tissues prior to litterfall in a process that conserves resources

Materials and Methods
The Guam study site included the northeast coastal zone, characterized by a single coralline soil series (clayey-skeletal, gibbsitic, nonacid, isohyperthermic Lithic Ustorthents). All of the selected species were sympatric within this karst habitat, which ranged in latitude from 13.4306 • to 13.6259 • N, ranged in longitude from 144.8092 • to 144.9364 • E, and ranged in elevation from 50 to 160 m above sea level. Five of the species in the dataset were limited in population and are found on published lists of endangered species. The occurrence of Typhoon Dolphin on 15 May 2015 defoliated most of the northern forest trees, and this defoliation event was exploited as a unique opportunity to obtain green leaf samples from the listed trees without taking samples directly from the protected trees. Green and senesced leaf samples were collected from the detritus layer by exploiting the litterfall zone for the various taxa within 48 h of the tropical cyclone defoliation event. Fully expanded green leaves and senesced leaves with partial chlorosis and partial necrosis were selected for the samples. The assignment of dislodged leaves to specific trees was enabled by the biodiversity and leeward litterfall zone. In most cases, this was achieved by collecting leaves from small stems that were dislodged, not from individual leaves that were dislodged. This approach ensured that the senesced leaf tissue was fresh and the two leaf ages were appropriately paired.
There were 23 spermatophyte and two pteridophyte species selected for inclusion in this dataset to provide a wide array of functional groups (Table A1). Each spermatophyte species was represented by six single-tree replications, with the exception of Serianthes nelsonii Merr. This species is critically endangered and the Guam population is represented by a single mature tree. For this large-canopied tree, paired green and senesced litter samples were collected from six circumscribed zones in the litterfall footprint created by the tropical cyclone. Therefore, a total of 133 plants were included in the dataset for the spermatophyte species. The two terrestrial pteridophyte species Asplenium polyodon G. Forst. and Nephrolepis hirsutula (G. Forst.) C. Presl were not defoliated by the tropical cyclone. Green and senesced leaves were collected from the same plants to ensure fidelity in the pairing. Each of six fern replications was represented by up to five individual adjacent plants to ensure that enough tissue was available for the analyses.
All selected leaf samples were devoid of observable insect damage. All selected plants were located at least 50 m inside intact forest communities to ensure that edge trees were avoided. All selected plants were adjacent to other native plants to ensure non-native plants did not influence the microsite. All plants were located on sites with no observable historical anthropogenic disturbance, with the exception of C. micronesica. At the time of this study, Guam's in situ populations of this tree had been threatened by ubiquitous infestations of the invasive armored scale Aulacaspis yasumatsui Takagi for 10 years. There were several permanent plots in northern Guam where the trees had been protected with systemic insecticides. All C. micronesica samples were obtained from a protected plot that was located within the footprint of this study. This managed plot had been used for various other research objectives [22][23][24].
The six green and six senesced leaf samples for each of the 25 species were dried for 48 h at 75 • C in a forced draft oven, and then milled to pass through a 20-mesh screen. Total carbon (C) and nitrogen (N) content was determined by dry combustion (FLASH EA1112 CHN Analyzer, Thermo Fisher, Waltham, MA, USA) [25]. Milled leaflet samples were also digested by a microwave system with nitric acid and peroxide, then 16 elements were quantified by inductively coupled plasma optical emission spectrometry (ICP-OES; Spectro Genesis; SPECTRO Analytical Instruments, Kleve, Germany). These were the macronutrients K, phosphorus (P), calcium (Ca), and magnesium (Mg); the micronutrients manganese (Mn), iron (Fe), zinc (Zn), boron (B), copper (Cu), and nickel (Ni); and the metals/metalloids selenium (Se), lead (Pb), arsenic (As), cadmium (Cd), cobalt (Co), and chromium (Cr).
Green and senesced leaf samples were further analyzed to determine cellulose concentration according to AOAC International [26]. Moreover, leaf tissue lignin concentration was determined according to the acetyl-bromide method [27].
The stoichiometric relationships between consequential nutrient pairs were calculated. For green leaf tissue, the quotients N:P, N:K, and K:P were calculated. For green and senesced leaves, the quotients C:N, lignin:N, and cellulose:N were calculated.
The chemical characteristics of this karst soil series are well-documented [28], but individual trees of Guam's native species may modify the localized edaphic traits via plant-soil feedback [29,30]. In order to characterize the soil traits directly associated with the sampled trees, one 15 cm soil core was obtained beneath each sampled plant halfway between the main stem and the canopy dripline. For some species which preferentially colonized karst outcrops, an intact soil core was not achievable. For these few plants, the mineral soils within the karst pockets were collected. The study footprint was divided into four equal latitudinal zones, and all cores within each zone were combined and homogenized. This enabled four replications for the analysis and evaluation of the soil chemistry that directly influenced the sampled plants. Total C and N content was determined by dry combustion. Available P was determined by the Olsen method [31]. Other macro-and micro-nutrients were quantified following digestion with diethylenetriaminepentaacetic acid [32]. Analysis was performed using ICP-OES. The chemical traits of these soils are shown in Table A2.
The analysis of variance methods was similar for all concentrations for each of the 18 quantified elements for green and senesced tissue, the cellulose and lignin concentrations for senesced leaf tissue, and the stoichiometry quotients. For elements which were quantified in both green and senesced leaf tissue, the analysis was conducted separately for each leaf age. Levene's test for homogeneity of variances was applied for each element. Green leaf C, K, Mg, and N; senesced leaf C, K, N, and P; and cellulose exhibited homogeneous variances. Therefore, the PROC GLM procedure in SAS (SAS Institute, Cary, NC, USA) was employed for analysis of variance. Green leaf As, B, Ca, Cd, Co, Cr, Cu, Fe, Mn, Ni, P, Pb, Se, and Zn; senesced leaf As, B, Ca, Cd, Co, Cr, Cu, Fe, Mg, Mn, Ni, Pb, Se, and Zn; lignin; and all calculated stoichiometry variables exhibited heterogeneous variances. Therefore, the PROC MIXED procedure in SAS was employed for each analysis of variance, which allowed a designation of unequal variance for each analysis. For variables which were significant, means separation was accomplished with the Tukey-Kramer post hoc test of each pairwise comparison.
Nutrient remobilization or accumulation during leaf aging and senescence was determined for each of the 18 elements by the formula ((green-senesced)/green) × 100. When this calculation is positive, the metric is defined as "resorption efficiency" for non-structural elements that are mobilized and returned to the stem tissue during leaf senescence [15,16]. When the calculation is negative, the metric defines the relative accumulation which occurs for individual elements from the age of full leaf expansion until leaf senescence. This metric has been called "null resorption" [33]; here, I introduce use of the term "accretion efficiency" for this metric. A statistical test of the differences among the species was difficult due to the heterogeneous variances and different direction of resorption versus accretion. Therefore, the data are presented as means ± standard error.

Green Leaf Element Concentrations
The macronutrient concentrations of green leaves revealed considerable variation among the 25 taxa ( Figure 1). The mean leaf N was 23.3 mg·g −1 and exhibited a 2.6-fold range. The taxa separated into eight overlapping groups. The leaf P was greatest for Artocarpus mariannensis Trécul and Ochrosia oppositifolia (Lam.) K. Schum., and there were five overlapping groups among the remaining 23 taxa. The mean leaf P was 1.38 mg·g −1 . The mean leaf K was 10.9 mg·g −1 , exhibiting a 3.3-fold range, and the means separated into seven overlapping groups. The mean leaf Ca was 15.8 mg·g −1 , and the taxa separated into six overlapping groups. The mean leaf Mg was greatest for N. hirsutula and Tabernaemontana rotensis (Kaneh.) P.T. Li, and the overall mean was 4.6 mg·g −1 . The relative ranges in Ca and Mg values were greater than those for other macronutrients, with a 4.5-fold difference among the taxa. The mean leaf C was 454.2 mg·g −1 , and the leaf C was not different among the taxa.
The micronutrient concentrations of green leaves revealed more relative range in values among the taxa than the macronutrient concentrations did ( Figure 2). The differences found were from a 8.2-fold range for Mn to a 120-fold range for Ni. The concentrations ranked in magnitude as Mn > Fe > B > Zn > Cu > Ni. Leaf B was greatest for Ficus prolixa G. Forst. and Meiogyne cylindrocarpa (Burck) Heusden, and there were six overlapping groups for the other 23 species. All other micronutrients exhibited considerable overlap among the values, with four significant groups for Zn to seven groups for Cu.
Non-essential metals and metalloids were present in green leaves with concentrations up to only 4 µg·g −1 for Se in Pisonia grandis R. Br. leaves ( Figure 3). All six measured elements exhibited values below 0.5 µg·g −1 for many of the taxa. The concentrations ranked in magnitude as Se > Pb > As > Cr > Co > Cd. No species emerged as a universal metal accumulator for all six elements, with the ranking of species in the upper end of the range in concentration shifting among the species for each of the elements.
There were several outlier groups where one or two taxa were significantly different from the remainder of the taxa for As, B, Co, Cr, and P. This always occurred at the upper end of the range of concentrations, and never at the lower end of the range. Therefore, similarities in values of green leaf element concentrations among these Guam karst forest plants were greater for the plant species that exhibited values below the overall means.

Green Leaf Stoichiometry and Nutrient Limitations
The N:P mean for all 25 taxa was 18.2, and a 3.8-fold difference occurred among the taxa ( Figure 4). Five of the taxa exhibited N:P that indicated co-limitation of N and P, 12 of the taxa exhibited N:P that indicated P limitation, and eight of the taxa exhibited N:P

Green Leaf Stoichiometry and Nutrient Limitations
The N:P mean for all 25 taxa was 18.2, and a 3.8-fold difference occurred among the taxa ( Figure 4). Five of the taxa exhibited N:P that indicated co-limitation of N and P, 12 of the taxa exhibited N:P that indicated P limitation, and eight of the taxa exhibited N:P that indicated N limitation. The N:K mean for all 25 taxa was 2.2, which was slightly above the global N:K value that signifies co-limitation of N and K. Fourteen of these Guam forest species had concentrations which indicated N limitation relative to K, and 11 of the species were K-limited relative to N. The K:P mean for all 25 taxa was 8.4, which greatly exceeded the global K:P value that signifies co-limitation. The relative range for K:P exceeded that for the other two quotients, with a 4.1-fold difference among the data. the global N:K value that signifies co-limitation of N and K. Fourteen of these Guam forest species had concentrations which indicated N limitation relative to K, and 11 of the species were K-limited relative to N. The K:P mean for all 25 taxa was 8.4, which greatly exceeded the global K:P value that signifies co-limitation. The relative range for K:P exceeded that for the other two quotients, with a 4.1-fold difference among the data.
These data collectively indicate that P limitation is profound in the majority of Guam's karst forest plant species. For example, all of the taxa were P-limited relative to K, and most of the taxa were limited or co-limited by P relative to N. Some of the species exhibited overt single element limitations. For example, N. hirsutula was near the upper end of the range for N:P and K:P, but was near the lower end of the range for N:K. These relationships illuminate acute P limitation and secondary K limitation for this abundant fern. In contrast, E. joga was near the upper end of the N:K range and near the lower end of the K:P range. These relationships illuminate acute K limitation for this large canopy tree. Aglaia mariannensis, Eugenia bryanii Kaneh., and Eugenia reinwardtiana (Blume) A.Cunn. ex DC. emerged as severely N-limited plants, ranking at or near the lowest quotient for N:P and N:K. Pisonia grandis emerged as acutely P-limited compared to N, and acutely K-limited compared to N, and based on the K:P quotient, this large tree was more limited by P than by K.  These data collectively indicate that P limitation is profound in the majority of Guam's karst forest plant species. For example, all of the taxa were P-limited relative to K, and most of the taxa were limited or co-limited by P relative to N. Some of the species exhibited overt single element limitations. For example, N. hirsutula was near the upper end of the range for N:P and K:P, but was near the lower end of the range for N:K. These relationships illuminate acute P limitation and secondary K limitation for this abundant fern. In contrast, E. joga was near the upper end of the N:K range and near the lower end of the K:P range. These relationships illuminate acute K limitation for this large canopy tree. Aglaia mariannensis, Eugenia bryanii Kaneh., and Eugenia reinwardtiana (Blume) A.Cunn. ex DC. emerged as severely N-limited plants, ranking at or near the lowest quotient for N:P and N:K. Pisonia grandis emerged as acutely P-limited compared to N, and acutely K-limited compared to N, and based on the K:P quotient, this large tree was more limited by P than by K.

Leaf Stoichiometry Predicts Litter Quality
The range of green leaf lignin concentration ranged from about 100 to about 300 mg·g −1 ( Figure 5). The means separated into four overlapping groups. Cellulose concentrations were similar in range to those of lignin, and the means separated into five overlapping groups. The C:N of green leaves exhibited a three-fold range and the means separated into five overlapping groups. The lignin:N values of green leaves exhibited a 8.8-fold range, and the lignin:N value of Heritiera longipetiolata Kaneh. green leaves greatly exceeded that of the other species. The cellulose:N values exhibited a 5.7-fold difference among the species, and the means separated into six overlapping groups.

Senesced Leaf Element Concentrations
The macronutrient concentrations for senesced leaves were much lower than for green leaves and revealed considerable variation among the 25 taxa ( Figure 6). Mean leaf N was 13.0 mg·g −1 and exhibited a 3.3-fold range. The taxa separated into seven overlapping groups. The mean leaf P was 0.6 mg·g −1 , with a substantial range in the data, exhibiting a 7.9-fold difference among the taxa. Leaf P was greatest for A. mariannensis. Mean leaf K was 3.7 mg·g −1 , exhibited an 8.3-fold range, and the means separated into four overlapping groups. Leaf Ca exhibited a mean of 30.9 mg·g −1 and there was a 4.0-fold difference among the taxa. Leaf Mg mean was 4.4 mg·g −1 . The relative range in Mg values was 4.5-fold among the taxa. Senesced leaf C mean was 446.8 mg·g −1 and the taxa separated into two large overlapping groups.
The differences in micronutrient concentrations of senesced leaves did not align with those of green leaves, with some elements revealing resorption efficiency and others revealing accretion efficiency (Figure 7). The differences observed were from an 8.0-fold range for Mn to a 122-fold range for Ni. The senesced leaf concentrations ranked in magnitude as Fe > Mn > B > Zn > Cu > Ni, with Fe replacing Mn as the micronutrient present at greatest concentration when compared to green leaves. Within the data for each element, the general rankings among the taxa were similar for green and senesced leaves.
Non-essential metals and metalloids were present in senesced leaves with concentrations up to 10 µg·g −1 for Se in Intsia bijuga (Colebr.) Kuntze leaves ( Figure 8). All six measured elements exhibited values below 0.5 µg·g −1 for some of the taxa. The concentrations ranked in magnitude as Se > Pb > As > Cr > Cd > Co, with Co replacing Cd as the least abundant metal/metalloid when compared with green leaves. The relative order of senesced leaf metal and metalloid concentration changed among the taxa when compared with green leaves. The senesced leaf metals and metalloids tended to exhibit one to three taxa with much greater concentrations than the remainder of taxa, then more similar concentrations for the remaining taxa.

Senesced Leaf Stoichiometry and Litter Quality
The range in senesced leaf lignin concentration was similar to that of cellulose, with these two macromolecules ranging from about 100 to about 300 mg·g −1 (Figure 9). Lignin concentrations of senesced leaves of C. micronesica and H. longipetiolata were greater than those of the other 23 species, which separated into five overlapping groups. Cellulose concentrations of senesced leaves of H. longipetiolata, Pandanus dubius Spreng., and Pandanus tectorius Parkinson ex Du Roi were greater than those of the other 22 species, which separated into five overlapping groups. Elaeocarpus joga and S. nelsonii were the only two species which were positioned near the low end of the range of concentrations for both of these macromolecules. The C:N of senesced leaves ranged in value up to ≈70, and exhibited a three-fold range among the species. The lignin:N of senesced leaves exhibited values up to about 43, with a 8.6-fold difference among the species. The cellulose:N values ranged up to about 41, with a 6.8-fold difference among the species.

Resorption and Accretion
The macronutrients N, P, and K decreased considerably in concentration as leaves aged from the mature green stage to the senesced stage (see Appendix Table A3). Almost 70% of leaf K and 60% of leaf P were removed and translocated back to stems prior to leaf senescence ( Figure 10). Leaf N resorption was also considerable, with a mean of about

Resorption and Accretion
The macronutrients N, P, and K decreased considerably in concentration as leaves aged from the mature green stage to the senesced stage (see Table A3). Almost 70% of leaf K and 60% of leaf P were removed and translocated back to stems prior to leaf senescence ( Figure 10). Leaf N resorption was also considerable, with a mean of about 40%. In contrast, leaf Ca concentration increased dramatically as leaves aged and senesced (see Table A4), with the Ca concentration more than doubling throughout the lifespan of leaves. The macronutrients C and Mg exhibited minimal changes in concentration throughout the lifespan of the leaves in this study.
40%. In contrast, leaf Ca concentration increased dramatically as leaves aged and senesced (see Appendix Table A4), with the Ca concentration more than doubling throughout the lifespan of leaves. The macronutrients C and Mg exhibited minimal changes in concentration throughout the lifespan of the leaves in this study.
The micronutrients Cu and Zn were also reabsorbed by stems prior to leaf senescence, but to a much lesser degree than for the macronutrients (Figure 10). The micronutrient Fe increased in concentration as leaves aged, as revealed by a 60% accretion efficiency. In contrast, B, Mn, and Ni concentrations exhibited minimal change throughout the lifespan of the leaves in this study.
Five of the quantified metals and metalloids increased in concentration as leaves aged, with accretion efficiency ranging in value from ≈40% for Co to ≈75% for Cd ( Figure  10). In contrast, Pb exhibited minimal change in concentration throughout the lifespan of the leaves. Figure 10. The resorption efficiency (positive means) and accretion efficiency (negative means) of leaves from 25 plant species from Guam's northern karst forests. Mean ± standard error, n = 6. The standard error is smaller than the marker size for markers with no bars. Vertical dashed lines separate macronutrients, micronutrients, and metals/metalloids.

Discussion
Exploitation of 300 leaf samples collected over a two-day time span within a constrained habitat range representing 25 native species from 18 families determined the concentrations of 12 plant nutrients and six metals/metalloids for northern Guam karst forest habitat. Inclusion of several qualitative functional types was enabled by the selection of eudicot angiosperms, gymnosperms, monocot angiosperms, pteridophytes, and trees associated with N-fixing root symbionts. These published Guam data may now inform several global research agendas. First, compilations of large datasets to understand leaf traits that represent global canons [34][35][36][37] have not included data from Micronesian islands. Cooccurrence of at least four plant species in the same habitat is desired for the addition of new data to the global dataset [35], so these Guam data are available to add a Micronesian island to this global agenda. Second, karst habitats provide unique features that plants must cope with to thrive. In addition to the physical outcrops which limit the amounts of mineral soil available for root exploration, the soil reaction and calcium carbonate reduce phyto-availability of the plant nutrients Cu, Fe, Mn, P, and Zn [38,39]. More dedicated research into karst plants such as the taxa discussed herein may provide globally relevant answers to plant adaptation [40]. Third, our future ability to understand how native biodiversity responds to human-induced global change factors [41][42][43][44] will require benchmarks. For example, global warming may lead to taxon-specific changes in leaf nutrient Figure 10. The resorption efficiency (positive means) and accretion efficiency (negative means) of leaves from 25 plant species from Guam's northern karst forests. Mean ± standard error, n = 6. The standard error is smaller than the marker size for markers with no bars. Vertical dashed lines separate macronutrients, micronutrients, and metals/metalloids. The micronutrients Cu and Zn were also reabsorbed by stems prior to leaf senescence, but to a much lesser degree than for the macronutrients (Figure 10). The micronutrient Fe increased in concentration as leaves aged, as revealed by a 60% accretion efficiency. In contrast, B, Mn, and Ni concentrations exhibited minimal change throughout the lifespan of the leaves in this study.
Five of the quantified metals and metalloids increased in concentration as leaves aged, with accretion efficiency ranging in value from ≈40% for Co to ≈75% for Cd ( Figure 10). In contrast, Pb exhibited minimal change in concentration throughout the lifespan of the leaves.

Discussion
Exploitation of 300 leaf samples collected over a two-day time span within a constrained habitat range representing 25 native species from 18 families determined the concentrations of 12 plant nutrients and six metals/metalloids for northern Guam karst forest habitat. Inclusion of several qualitative functional types was enabled by the selection of eudicot angiosperms, gymnosperms, monocot angiosperms, pteridophytes, and trees associated with N-fixing root symbionts. These published Guam data may now inform several global research agendas. First, compilations of large datasets to understand leaf traits that represent global canons [34][35][36][37] have not included data from Micronesian islands. Co-occurrence of at least four plant species in the same habitat is desired for the addition of new data to the global dataset [35], so these Guam data are available to add a Micronesian island to this global agenda. Second, karst habitats provide unique features that plants must cope with to thrive. In addition to the physical outcrops which limit the amounts of mineral soil available for root exploration, the soil reaction and calcium carbonate reduce phyto-availability of the plant nutrients Cu, Fe, Mn, P, and Zn [38,39]. More dedicated research into karst plants such as the taxa discussed herein may provide globally relevant answers to plant adaptation [40]. Third, our future ability to understand how native biodiversity responds to human-induced global change factors [41][42][43][44] will require benchmarks. For example, global warming may lead to taxon-specific changes in leaf nutrient concentrations [45]. The Guam leaf traits reported herein may provide that benchmark for future studies as Guam and other Micronesian islands respond to the influence of global change factors.
The portion of the leaf ionome comprising essential nutrients is the functional ionome [39]. Comparisons of these Guam leaf traits with the global traits [34,36,37,[46][47][48] may enable a greater understanding of the native plant communities. The ranges of mass-based green leaf elemental concentrations were within published global datasets, but the means deviated for some of the elements. For example, the mean Mn and Cu concentrations in Guam's green leaves were less than the global means. In contrast, the mean N and Ca concentrations in Guam's green leaves exceeded the global means. The greater Ca in Guam's karst forest green leaves may be explained by the excessive pH and calcium carbonate in the soils [37], and more research is needed for Guam habitats from volcanic soils. Although Ca has not been studied as extensively as N and P, its direct influence on nutrient cycling, C sequestration, and pedogenesis is recognized [47,49].
Tian et al. [48] reported N and P means for Oceania that were less than half of those reported herein for Guam. However, their data representing Oceania were restricted to Australia and New Zealand. Guam is located in Oceania, and these findings indicate that the inclusion of more insular forest habitats in the data used to represent Oceania in global leaf ionome comparisons is warranted.
Eudicot angiosperm species represented the largest plant group in this study. There were two pteridophyte species, and these ferns exhibited greater Cu, K, Mg, Ni, and Zn concentrations than the remainder of the species, and less Cr, lignin, and Mn than the remainder of the species. There were two monocot species, and these Pandanus trees exhibited greater cellulose, Fe, and Mn than the remainder of the species, and less B, Mg, and P than the remainder of the species. There were three species associated with N-fixing root symbionts. These three tree species exhibited green leaf N that was 23% greater and senesced leaf N that was 48% greater than the remainder of the species. The differences in relative N concentration, as influenced by leaf age, were a result of a N resorption efficiency of only 32% for the trees with N-fixing endosymbionts, as compared to a 44% resorption efficiency for the remainder of the species.
These results illuminate the importance of phylogenetic and functional group diversity, and the profound influence that this diversity has on nutrient cycling in this Guam forest. Indeed, tree species characteristics and litter mixtures exert many direct and indirect controls over ecosystem properties, such as the heterotrophic soil microbial community [50,51].

Stoichiometry
The mean of the 25 taxa in this northern Guam karst forest exhibited a green leaf C:N:P relationship of 329:17:1. The reported mean of widespread forest foliage is 469:13:1 [52], and the reported global mean of all terrestrial plants is 375:13:1 [53]. These relationships indicate that less C may be sequestered in Guam's arboreal layer in relation to N and P than in the global forest mean.
Guam's N:P mean of 18.2 was greater than the global N:P mean of ≈16 [35,37]. Largescale comparisons indicate that a N:P of about 10 is optimal for plant growth-a N:P above ≈16 signifies P limitation, and a N:P below ≈14 signifies N limitation [4,[11][12][13][14]. Although the Guam population mean for N:P indicated collective P limitation, the N:P for eight of the species indicated N limitation. A leaf N:K below 2.1 signifies N limitation, and a N:K above 2.1 signifies K limitation [13]. The Guam population mean for N:K was remarkably close to 2.1, indicating co-limitation of N and K at the population level. The majority of the individual species exhibited N:K that signified N limitation. The two Eugenia species were positioned at the extreme end of the spectrum, indicating N limitation based on N:P and N:K. A leaf K:P above 3.4 signifies P limitation, and a K:P below 3.4 signifies K limitation [13]. The Guam population mean for K:P was 8.4, indicating extreme population-level P limitation in relation to K availability. Collectively, these stoichiometry relations indicate Guam's karst forest is most limited by P and secondarily limited by N, although several taxa exhibited co-limitation by K. Based on the K, N:K, and K:P relations, E. joga appears to be the species that is most limited by K.
Senesced leaf C, lignin, cellulose, and N stoichiometry are important for predicting litter decomposition dynamics [4,[54][55][56] and the influences on rhizodeposition and microbial activity [57]. The two Pandanus species and H. longipetiolata consistently appeared on the left, signifying recalcitrant litter, and S. nelsonii consistently appeared on the right, signifying labile litter. Litter incubation studies will be required to confirm decomposition dynamics among the species and the influence of litter identity mixing.
Broad interpretations of nutrient limitations based on N:P may lead to interpretation errors [58,59]. However, my comparisons among individual taxa add to important interpretations of functional diversity in Guam's forests. Moreover, the comprehensive assessment of all N:P:K interactions point to P as being the most limiting nutrient in northern Guam, and this is the first report to convey these findings.

Threatened Species Traits
Five of the taxa in this study are listed on the United States Endangered Species Act [60,61]. Policy-makers and conservation practitioners require evidence-informed answers to critical questions in order to adequately conserve these plants. The paucity of peer-reviewed publications limits the accuracy of most conservation and restoration agendas throughout Micronesia. For example, only one of these species is supported by a federal recovery plan, and this plan identifies the lack of research as a constraint to species recovery [62]. These are the first in situ biology data published for E. bryanii, H. longipetiolata, and T. rotensis. Although the results of some in situ studies have been published for S. nelsonii [30,[63][64][65], the data herein are the first published data on leaf ionome relations. Cycas micronesica has been the subject of numerous publications containing in situ and ex situ data on leaf nutrient relations [19,20]. However, the inclusion of C. micronesica in this dataset represents the first time that leaf nutrient relations have been determined within the same space and time with numerous other sympatric taxa.
Plant species diversity and the variation in the functional traits that accompany this diversity are critical for sustaining ecosystem services such as carbon storage, nutrient cycling, and spatial heterogeneity of plant and microbe biodiversity [65][66][67][68]. These issues illuminate the damage to ecosystem function that may accompany localized extirpations of a threatened plant species. Indeed, when an insular population of a tree species is endangered, the concept of functional redundancy becomes important in the prediction of what may happen if local extirpations occur. Functional redundancy occurs when two more taxa contribute similar ecosystem services, such that loss of one of the organisms does not lead to functional loss [69][70][71][72]. The coexistence of functionally similar taxa has been called the portfolio effect [73]. Within this context, a focus on species which perform unique functions should become an integral component of conservation and restoration efforts to ensure resilience in the face of a changing world [74]. In the context of the data reported herein, the species that exhibit means that are mid-range in Figures 1-9 do not exhibit particularly unique traits, and loss of one of these organisms would be theoretically buffered against the loss of the biogeochemistry functions they provide. In contrast, loss of one of the species with means that are the least or greatest in Figures 1-9 may risk more damage to the ecosystem services of a habitat.
For example, two of these species emerge as outliers in the ranking of species for several of the influential litter quality traits. The Malvaceae tree H. longipetiolata was positioned hear the minimum end of the spectrum for senesced leaf N and near the maximum end of the spectrum for senesced leaf lignin, cellulose, C:N, lignin:N, and cellulose:N. In contrast, the Fabaceae tree S. nelsonii was positioned near the maximum end of the spectrum for senesced leaf N and near the minimum end of the spectrum for senesced leaf lignin, cellulose, C:N, lignin:N, and cellulose:N. The unique contributions of these two tree species to biogeochemical cycling dynamics in northern Guam appear to be profound, indicating that the extirpation of these taxa from the palette of Guam's trees will directly damage the biodiversity of the detritus and soil food web due to a habitat-level constraint of litter quality diversity of the remaining litterfall.
Two of these threatened tree species associate with N-fixing root endosymbionts, revealing their importance in contributing N into the biosphere and pedosphere. The historical abundance of the legume tree S. nelsonii is not known, because pre-threat data on abundance were not recorded. The in situ Guam population has been reduced to a single mature tree [75]. The gymnosperm C. micronesica was the most abundant tree on Guam in 2002 [76], immediately prior to the 2003 invasion of the armored scale A. yasumatsui [77]. In situ mortality following infestations of this cycad-specific scale began in 2005, and as of 2020, the mortality had reached 96% [78]. The addition of atmospheric N to terrestrial systems is a crucial ecosystem service provided by these trees, and the ongoing declines in plant abundance carry a concomitant decline in the volume of N additions. Other risks associated with loss of C. micronesica have been discussed previously [79].

Leaf Aging and Senenence Dynamics
Leaf senescence is a highly choreographed process whereby some non-structural components are disassembled, mobilized, and absorbed by phloem tissue in subtending stems, thereby making them available for newly developing vegetative and reproductive organs [80,81]. Leaf aging and senescence dynamics in this northern Guam karst forest caused the species differences in leaf element concentrations to become less different with age. This assertion is validated by the number of groups of that were significantly different when comparing the green leaf data with the senesced leaf data. The green leaf element concentrations (Figures 1-3) separated into a mean of 6.1 different groups, but the senesced leaf element concentrations (Figures 6-8) separated into a mean of 4.3 different groups. The amplitude of absolute leaf nutrient concentration differences among the Guam forest species was influenced by resorption efficiency dynamics. For elements that were efficiently absorbed during leaf senescence (see Table A3), the range in differences among the species was greatly reduced in senesced leaves. For elements that exhibited accretion during leaf aging and senescence (see Table A4), the range in differences among the species was greatly increased in senesced leaves.
Global N resorption means have been compiled for numerous macronutrients and micronutrients [15][16][17][18]82,83]. The Guam resorption efficiency means reported herein were lower than global means for C, N, Ca, Mg, Cu, and B. Resorption efficiencies for K, P, and Zn in this study were similar or within the range of global means. Accretion efficiency for Fe in Guam's karst forests was more than double that for the global mean. These relationships collectively indicate that the species representing Guam's karst forests appear to be less efficient at nutrient recycling by way of resorption prior to leaf senescence than the average plant species represented in the global data.
The macromolecule lignin exhibited minimal changes in concentration during leaf aging. The macromolecule cellulose increased slightly with leaf age for the species at the high end of the senesced concentration range but did not change with leaf age for the species at the low end of the senesced concentration range. These patterns have been reported for other species [84]. Carbon investments into leaf cellulose and lignin represent investments into leaf toughness, which contributes directly to leaf lifespan [85].

Tropical Cyclones and Nutrient Cycling
Guam is positioned within the most active tropical cyclone zone worldwide [86]. The influence of these stochastic large-scale abiotic stressors is so strong that Guam's forests have been called "typhoon forests" [87].
The unique opportunity provided by the 15 May 2015 passage of Typhoon Dolphin provided an empirical look at the influence of tropical cyclone defoliation on litter quality in Guam's northern karst forests. For example, senesced leaves which exemplify natural litterfall exhibited a mean N concentration of 13 mg·g −1 , with a maximum of about 20 mg·g −1 (Figure 6). In contrast, green leaves exhibited a mean N concentration of 23 mg·g −1 , with a maximum of about 35 mg·g −1 (Figure 1). These data indicate a 57% increase in the influence of N on litter quality as a result of a tropical cyclone. By comparing the stoichiometry traits for green leaves ( Figure 5) with senesced leaves (Figure 9), similar conclusions can be drawn. The mean C:N was 38 for senesced leaves and only 21 for green leaves, the mean lignin:N was 15 for senesced leaves and only 8 for green leaves, and the mean cellulose:N was 16 for senesced leaves and only 8 for green leaves. These litter quality traits predict an acceleration of litter decomposition and nutrient turnover as a direct result of tropical cyclone behavior in this northern Guam karst forest. Clearly, Guam's frequent tropical cyclones not only influence physiognomy [87], they also profoundly influence the spatiotemporal dynamics of the island's biogeochemistry.
The primary drivers of increased litter quality following tropical cyclone defoliation in Guam's forests appeared to be the macronutrient concentrations of the litter, not the C or the C-based macromolecules. Indeed, the differences in Ca, N, P, and K between green and senesced leaves were substantial, but the differences in C, lignin, and cellulose were minimal.

Future Directions
The findings reported herein begin to apply the leaf economics spectrum ethos to Guam's forests, but non-nutrient leaf traits have not been reported for Guam's forest species. In order to fully integrate Guam's forest species into the global leaf economics spectrum research agenda, addition of the influence of species on maximum net photosynthesis, leaf longevity, and specific leaf area [34,35] may be some of the more important leaf traits to add to the Guam ecology research agenda.
The differences in leaf chemistry for C. micronesica, I. bijuga, and S. nelsonii leaf litter predicted the influence of species on the release of C and N from incubated litter [88]. The litter quality traits reported herein are useful for predicting the rates of nutrient turnover, but litter decomposition studies replicated in situ will be required to move beyond estimates and obtain the required empirical evidence.
Infestations of three insect herbivores elicited changes in leaf litter quality for C. micronesica [89]. Leaves exhibiting leaf herbivory were avoided during the sampling procedures in the present study. The influence of herbivory on leaf ionomics of the remaining species in this study are not known.
We are beginning to learn more about the various factors that influence how plants control leaf nutrient relations [90]. However, the manner in which organismal and habitat characteristics affect the leaf ionome, stoichiometry, and resorption remains unanswered for Guam's forests. Long-term nutrient manipulation studies provide valuable information for interpreting nutrient limitations and cycling in forest productivity. These types of studies require commitments that are decadal, not annual. Guam ecology research suffers from a history of being stifled by the mindset of short-term funding cycles, and the net result is that few long-term studies have been enacted [91]. Applying leaf ionome generalizations developed from non-tropical studies does not appear to satisfactorily explain tropical forest behaviors [92]. The initiation of long-term N and P nutrient addition studies within Micronesian forest plots would greatly expand our understanding of nutrient cycling and limitations on productivity in the region.

Data Availability Statement:
The data presented in this study are available in the Appendix A.

Conflicts of Interest:
The funder had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.