Parsing Long-Term Tree Recruitment, Growth, and Mortality to Identify Hurricane Effects on Structural and Compositional Change in a Tropical Forest

After hurricane disturbances in tropical forests, the size structure and species composition are affected by immediate mortality, and subsequent recruitment and individual growth. Often, immediate post-disturbance stand-level data are presented but understanding of the components that affect changes in growth and longer-term responses to forest structure and composition are lacking. To answer questions about how mortality, recruitment, and growth change among successional Plant Functional Types (PFT) through time after a hurricane disturbance, we use long-term census data (1989–2014) collected in the Luquillo Experimental Forest, Puerto Rico. We developed an algorithm to fill missing diameter data from the long-term data set that was collected three months after Hurricane Hugo; and subsequently at five-year intervals. Both the immediate hurricane-induced mortality and subsequent mortality were lower in stems with larger diameters, but varied among successional PFTs Early, Mid, Late, and Palm. Tree growth rates were observed to decrease with time since the hurricane disturbance. Five years after the hurricane, mortality was minimal but then increased gradually with time. In contrast, recruitment was highest five years after the hurricane and then decreased with time. The palm Prestoea montana became the most abundant species in the forest after the hurricane, as it had the lowest immediate hurricane-induced and subsequent mortality, and the highest recruitment. Twenty-five years after the hurricane, the palm and the Late PFT dominate the forest after shifting species composition from pre-hurricane conditions.


Introduction
Hurricanes have major effects on tropical forests [1][2][3][4][5][6]. Hurricane-force winds (compounded and aided by heavy rains, saturated soils, and loosened roots [4,7]) uproot and snap trees, break branches, and defoliate surrounding vegetation. The results are gaps in the forest and hence increasing light levels within the forest canopy [8]. Defoliation may recover in months to years [9][10][11][12][13][14], but forest structure and composition may shift over decades following three stages of species succession [15][16][17][18][19]. Early successional (pioneer) species establish and recruit in open gaps formed after hurricane disturbances, growing rapidly in the high light environment. Mid successional species, which have intermediate growth rate and are somewhat shade tolerant, gradually substitute early successional species as the gaps close and the canopy recovers. Late successional species, which have low growth rates and are shade tolerant, reach the canopy, and become dominant in the plant community as the forest matures until the next disturbance.
Forest composition and structure are also critical to hurricane-induced mortalityless wind-resistant structure and composition would lead to higher mortality, especially recovered its stem density and basal area after selective logging. Hence, the recovery of the forest from Hurricane Hugo is defined as secondary succession (a natural disturbance in mature native secondary forest) [36]. After Hugo in 1989, nine other named storms and hurricanes have passed near the area [30,37], but only Hurricane Georges (category 3), in 1998, resulted in localized defoliation and minimal mortality in the study area [14,38]. The most recent hurricanes at Bisley-Irma and Maria (September 2017)-are addressed in separate studies [20].

Diameter and Species Observations
Eighty-five permanent plots were established at Bisley in 1989 and censuses were conducted three months before Hurricane Hugo (pre-Hugo 1989), three months after the hurricane (post-Hugo 1989), and every five years since then (1994, 1999, 2004, 2009, and 2014). The pre-Hugo census was started in late 1988 and was completed in the summer of 1989 (three months before the hurricane), the post-Hugo census was started three months after Hugo and completed within five months. Other censuses were started in April and completed between June and July of their respective year. In addition, six more plots were established in 1994-one in 2009, and three in 2014-making 95 plots in total. The plots are 10-m diameter circles and 40 m apart in a grid extending over 13 hectares. The topographic location of each plot is recorded (i.e., valley, slope, and ridge). Stems in the plots with diameter at breast height (1.3 m above the ground) (DBH) equal to or greater than 2.5 cm were tagged, the species were identified, DBH measured (with a precision of 0.1 cm), and heights (H) of the two tallest stems per plot were measured. The tag number of each stem is unique at the plot level and provides the year the stem was first recorded in the census. The pre-Hugo 1989 census was completed three months before the hurricane, and the post-Hugo 1989 census was started three months after the hurricane, thus they directly measure the immediate mortality caused by the hurricane-mortality meaning that no recovery was found in later censuses.
The plots at Bisley were initially established to explore watershed scale biomass and forest standing stocks and protocols appropriate for that purpose were used. Previous studies at this site investigated the relative changes in standing stocks, estimates on stem density, basal area, aboveground biomass, and species richness and diversity in relation to succession after Hurricane Hugo [14,29,39,40]. Those studies, however, lacked long-term analyses of mortality, recruitment, and growth rates in response to hurricane disturbance.

Plant Functional Types (PFT) of Species
Sixty-six species were identified, including a species of herb (a banana species Musa x paradisiaca), a palm species Prestoea montana, two species of tree-ferns (Cyathea arborea and C. portoricensis), and 62 species of trees. The banana species disappeared from the forest after Hurricane Hugo in 1989. We categorized species into successional plant functional types (PFT) based on their documented response to disturbances [40][41][42][43][44]. The classification into PFT is a complementary approach to understand forest recovery and succession, and to contribute to on-going modeling efforts of vegetation dynamics [45][46][47]. The classifications of PFT have important applications for predictive ecology and modeling, serving to reduce the large numbers of plant species into a smaller quantity of functional classes based on ecological concepts [48,49]. Building upon a previous work [44], in a study of this same forest (Luquillo Experimental Forest) and same forest type (tabonuco), the species of trees and tree-ferns are categorized into one of three tropical successional PFTs: Early, Mid, or Late (Table S1 for the PFT of each species). Early PFT includes species that establish and recruit in open canopy and gaps formed after disturbances and grow rapidly in the high light environment. Mid PFT includes species that have intermediate growth rates and are somewhat shade tolerant. Late PFT includes species that have low growth rates and are shade tolerant. Very few species were not classified by previous studies [41][42][43][44], and in the rare case that a species was still not classified after the above two steps, we assigned a PFT to the species based on descriptions of successional traits such as seed characteristics, recruitment, and growth [50,51].
Previous studies have either excluded palms from analyses [52] or considered palms as a mid or late successional species [41][42][43][44]. However, the palm Prestoea montana possesses some early successional traits, such as low "wood" density and high fecundity under open canopy [53,54], allowing them to recruit quickly when the canopy opens; and some late successional traits, such as tolerance to shade [26], which favor their growth and survival when the canopy closes. Therefore, following a previous study [20], we separate the palm species from the three broadly used successional PFTs and categorize it as Palm PFT.

Data Analyses
We estimated the missing values in the dataset (Supplementary Information S1) and calculated lower bounds for the measurement error (Supplementary Information S2). To study the mortality and recruitment changes with time since disturbance, we calculated the recruitment rate of each PFT in each five-year census interval, and the mortality rate of each PFT and of each DBH class. The recruitment rate was calculated as the percentage of the number of newly recruited stems to the number of stems that were alive in the previous census. Similarly, the mortality was calculated as the percentage of the number of stems that existed in the previous census but not in the current census to the number of stems that existed in the previous census. The immediate hurricane-induced mortality was calculated as the percentage of the number of stems that existed in pre-Hugo census but not in post-Hugo census to the number of stems that existed in pre-Hugo census.
Diameter growth was calculated as the diameter measurement difference between two consecutive five-year-interval censuses. Given the sampling rate of the census, our estimates of growth were cm per 5-years but expressed as cm per year (cm yr −1 ) of each standing stem. The relative stem diameter growth rate (%) was calculated as the ratio of the absolute diameter growth rate to the previous-census diameter, expressed as a percentage. Similarly, we calculated the absolute basal area growth rates (cm 2 yr −1 ) of each stem. Aboveground biomass (AGB) of each tree was calculated as AGB = exp(2.475 ln(DBH) − 2.399) following the equations from Scatena et al. [29]. The growth rates of each species and PFT during each census interval were analyzed to identify the differences among species and among PFTs, and their changes with time since disturbance. Mortality and growth rates were calculated for each stem separately, and then grouped according to PFT. Note that height growth is more significant than diameter growth for the palm species [53,54]. However, since we do not have height information for all palms, we did not calculate the height growth for palms.
Lastly, to explore forest community structure and composition over 25 years, we selected the most abundant species to quantify the proportional changes in species composition. The most abundant species were defined as those whose individual abundance accounts for more than 5% of the total abundance in the forest in one or more censuses. Seven species were selected as abundant: two Early species (Cecropia schreberiana (CECSCH) and Psychotria berteriana (PSYBER)), one Mid species (Ocotea leucoxylon (OCOLEU)), three Late species (Sloanea berteriana (SLOBER), Dacryodes excelsa (DACEXC), and Cyathea portoricensis (CYAPOR)), and the Palm species (Prestoea montana, PREMON). The seven species in total account for 62%, 60%, 74%, 67%, 65%, 64%, and 66% of the total abundance in each of the seven censuses, respectively.
To identify the effect of topography on species distribution, we calculated the averages of the proportion of each species across plots with the same topographic setting; slopes, valleys, and ridges [38]. The species proportion in a plot is defined as the ratio between the number of stems of the species to the total number of stems in the plot. To identify the effect of topography on demographic rates, we calculated the distribution of growth rate for stems within each topographic setting and tested the difference of the distribution between any two settings using two-sample t test. For mortality and recruitment, we calculated the rates using plots with the same topographic settings and tested the difference of the rates between any two settings using a z-score test [55] following Zhang et al. [20].

Mortality and Recruitment over the Course of 25 Years
The overall immediate stem mortality from Hurricane Hugo (1989) was 58%. Mortality varied among PFTs, with Early PFT having the highest (82%), followed by Mid (55%), then Late (54%), and with Palm (42%) having the lowest mortality value (Figure 1a) overall. Immediate hurricane-induced mortality and subsequent mortality varied by diameter classes (Figure 1b). Smaller stems had higher mortality than larger stems for all censuses except censuses 1994 and 1999, where intermediate diameter stems (DBH between 5 and 10 cm) had the highest mortality among the four diameter classes. except censuses 1994 and 1999, where intermediate diameter stems (DBH between 5 and 10 cm) had the highest mortality among the four diameter classes.

Diameter Growth Rates
For all stems during the 25-year study period, average diameter growth rate decreased as time after the disturbance increased: 0. 33 Figure 2). However, in the cases with a statistically significant regression coefficient, the coefficient of determination and the regression coefficient itself are very small, denoting a very weak relationship.
For all stems with diameter less than 20 cm, the average growth rate was the highest in census 1994, five years after the hurricane. Average growth rate decreased with time since the hurricane disturbance for all PFTs, especially for the Early and Palm PFTs (Figure 3). For stems with diameter between 20 and 40 cm, the growth rate of the Early PFT decreased with time since disturbance (Figure 3a), while the growth rate of the Mid and Late PFTs first decreased and then increased with time since disturbance (Figure 3b,c). For stems with DBH larger than 40 cm, the growth rate of Late increased with time since disturbance. For all stems with diameter less than 20 cm, the average growth rate was the highest in census 1994, five years after the hurricane. Average growth rate decreased with time since the hurricane disturbance for all PFTs, especially for the Early and Palm PFTs ( Figure  3). For stems with diameter between 20 and 40 cm, the growth rate of the Early PFT decreased with time since disturbance (Figure 3a), while the growth rate of the Mid and Late PFTs first decreased and then increased with time since disturbance (Figure 3b,c). For stems with DBH larger than 40 cm, the growth rate of Late increased with time since disturbance.   For all stems with diameter less than 20 cm, the average growth rate was the highest in census 1994, five years after the hurricane. Average growth rate decreased with time since the hurricane disturbance for all PFTs, especially for the Early and Palm PFTs ( Figure  3). For stems with diameter between 20 and 40 cm, the growth rate of the Early PFT decreased with time since disturbance (Figure 3a), while the growth rate of the Mid and Late PFTs first decreased and then increased with time since disturbance (Figure 3b,c). For stems with DBH larger than 40 cm, the growth rate of Late increased with time since disturbance.  Larger stems have lower relative growth rates as observed in Late PFT stems ( Figure S6c). Like the absolute growth rate, the relative growth rate also decreased with time since disturbance, especially for Early and Palm PFTs, and Mid and Late PFTs with small diameters (Figure S6a-d). Basal area growth is derived from diameter and is highly correlated with diameter growth (Figure S6e-h).

Shifts in Species Composition
Shifts in species composition of the forest were quantified by focusing on the proportions of the seven most abundant species through time (Figure 4). In terms of abundance, the forest was originally dominated by two Late successional species, Sloanea berteriana (SLOBER) and Dacryodes excelsa (DACEXC) (Figure 4a). For the ten years following Hurricane Hugo (Figure 4c,d), Early successional species Cecropia schreberiana (CECSCH) and Psychotria berteriana (PSYBER) recruited and increased and then started to decrease 20 and 25 years later (Figure 4f,g). Mid successional species Ocotea leucoxylon (OCOLEU), which was less abundant after the hurricane (Figure 4b), recruited and increased in proportion ten years after the hurricane (Figure 4d) and then maintained its abundance (Figure 4e-g). The Late successional species SLOBER and DACXEC also began to increase their abundance 15 years after the hurricane, surpassing the Early successional CECSCH and PSYBER, then maintaining their abundance 25 years later (Figure 4e-g). Another Late successional species, the tree-fern Cyathea portoricensis (CYAPOR), surpassed its pre-disturbance abundance and became a dominant species in the forest in 2014 (Figure 4g). The Palm species PREMON increased its dominance in the forest after the disturbance (Figure 4b) and continued to increase throughout the 25 years after the disturbance and became the most abundant species in the forest by 2014 (Figure 4c-g).
disturbance, especially for Early and Palm PFTs, and Mid and Late PFTs with small diameters ( Figure S6a-d). Basal area growth is derived from diameter and is highly correlated with diameter growth (Figure S6e-h).

Shifts in Species Composition
Shifts in species composition of the forest were quantified by focusing on the proportions of the seven most abundant species through time (Figure 4). In terms of abundance, the forest was originally dominated by two Late successional species, Sloanea berteriana (SLOBER) and Dacryodes excelsa (DACEXC) (Figure 4a). For the ten years following Hurricane Hugo (Figure 4c,d), Early successional species Cecropia schreberiana (CECSCH) and Psychotria berteriana (PSYBER) recruited and increased and then started to decrease 20 and 25 years later (Figure 4f,g). Mid successional species Ocotea leucoxylon (OCOLEU), which was less abundant after the hurricane (Figure 4b), recruited and increased in proportion ten years after the hurricane (Figure 4d) and then maintained its abundance (Figure 4eg). The Late successional species SLOBER and DACXEC also began to increase their abundance 15 years after the hurricane, surpassing the Early successional CECSCH and PSYBER, then maintaining their abundance 25 years later (Figure 4e-g). Another Late successional species, the tree-fern Cyathea portoricensis (CYAPOR), surpassed its pre-disturbance abundance and became a dominant species in the forest in 2014 (Figure 4g). The Palm species PREMON increased its dominance in the forest after the disturbance (Figure 4b) and continued to increase throughout the 25 years after the disturbance and became the most abundant species in the forest by 2014 (Figure 4c-g).

Impact of Topography
The demographic rates vary among the topographic settings (i.e., valley, ridge, and slope) (Tables S2-S4). Specifically, Mid PFT has significantly lower growth rate on ridges (0.18 cm yr −1 ) than on slopes (0.29 cm yr −1 ) (p < 0.0001), Late PFT has significantly lower growth rate on valleys (0.16 cm yr −1 ) than on ridges (0.25 cm yr −1 ) (p = 0.0053), and Palm PFT has significantly higher growth rates on valley plots (0.16 cm yr −1 ) than on slope plots (p = 0.0015). Mortality was generally lower on ridges than on slopes or valleys, and the difference was significant in 1989 (p = 0.0001), 2004 (p < 0.0001), and 2009 (p = 0.0022). Recruitment was the highest in valleys in the first five years after the hurricane (1994; p < 0.0001). These variations are linked with the shifts of species distribution among the topographic settings after the hurricane disturbance ( Figure 5). difference was significant in 1989 (p = 0.0001), 2004 (p < 0.0001), and 2009 (p = 0.0022). Recruitment was the highest in valleys in the first five years after the hurricane (1994; p < 0.0001). These variations are linked with the shifts of species distribution among the topographic settings after the hurricane disturbance ( Figure 5).
The Early successional species CECSCH increased its proportional abundance in valleys and slopes, but never dominated on ridges (Figure 5a). The proportion of CECSCH in ridge plots is significantly lower than that in valley or slope plots (p < 0.0001). The tree fern CYAPOR (Figure 5d) increased the proportion of its abundance in valleys and slopes after the hurricane. The proportion of CAYPOR in valley/slope plots is significantly higher than in ridge plots (p < 0.0001). This is in contrast to the Late successional species DACEXC, which had significantly lower proportion in plots in valleys and slopes than in plots on ridges (p < 0.0001). This dominant topographical distribution of the Late species DACEXC remained unchanged during the 25-year post-hurricane study (Figure 5e). The palm PREMON was more abundant in valley plots (Figure 5g) before the hurricane and increased its proportional abundance in valley plots as time after the hurricane increased. The proportion of PREMON in valley plots is significantly higher than slope plots (p = 0.0011) and ridge plots (p = 0.0091). Further, palms recruited and maintained a significant abundance in all the topographical settings and an increase in proportion of palms relative to all stems was recorded in every census up to 25 years after the hurricane (Figure 5g).   Table S1.
The Early successional species CECSCH increased its proportional abundance in valleys and slopes, but never dominated on ridges (Figure 5a). The proportion of CECSCH in ridge plots is significantly lower than that in valley or slope plots (p < 0.0001). The tree fern CYAPOR (Figure 5d) increased the proportion of its abundance in valleys and slopes after the hurricane. The proportion of CAYPOR in valley/slope plots is significantly higher than in ridge plots (p < 0.0001). This is in contrast to the Late successional species DACEXC, which had significantly lower proportion in plots in valleys and slopes than in plots on ridges (p < 0.0001). This dominant topographical distribution of the Late species DACEXC remained unchanged during the 25-year post-hurricane study (Figure 5e). The palm PRE-MON was more abundant in valley plots (Figure 5g) before the hurricane and increased its proportional abundance in valley plots as time after the hurricane increased. The proportion of PREMON in valley plots is significantly higher than slope plots (p = 0.0011) and ridge plots (p = 0.0091). Further, palms recruited and maintained a significant abundance in all the topographical settings and an increase in proportion of palms relative to all stems was recorded in every census up to 25 years after the hurricane (Figure 5g).

Patterns of Forest Structure following a Hurricane Disturbance
Changes in recruitment, growth, and mortality through time explain Hurricane Hugo's long-term effects on the forest structure and composition. Variation in the immediate mortality among PFTs and among stem sizes resulted in an immediate change of forest size structure and species composition. The variation in diameter growth and recruitment with time resulted in a shift in size structure as the forest recovered. Before Hurricane Hugo, the forest at Bisley was dominated by Late PFT with diameters less than 10 cm ( Figure 6). Because Early had the highest mortality and Palm had the lowest mortality among the four PFTs during the hurricane disturbance, Early PFT decreased in abundance right after the hurricane, while Palm increased its proportion in the forest. As smaller stems had higher immediate mortality, the size structure of the forest shifted from being dominated by small stems (2.5-5 cm) to a uniformly distributed size structure immediately post hurricane ( Figure 6). In an experimental manipulative study of hurricane effects, Shiels et al. [3] found that forest responses to hurricane effects were driven by changes in the canopy structure and openness of the canopy, more than from fallen hurricane debris. In our study, when the canopy opened due to the disturbance, both Early and Palm had higher recruitment than mortality, which resulted in increases in their abundance five years after the disturbance. Twenty years after the hurricane disturbance, the forest was again dominated by Late PFT with diameters less than 10 cm, which may be the beginning of the self-thinning stage until it is interrupted by the next disturbance [56,57]. These changes in PFT abundances have been documented in other tropical forests in response to changes in climate [58]. In our study system, the response to changes in the post-disturbance abiotic environment, inherently altered forest composition which in turn lead to concomitant changes in forest structure. Despite these shifts, the proportion of larger stems from the Late PFT did not change significantly throughout the 25-year study period. The maintenance of "non-pioneer" or Late PFT is associated to their documented post-hurricane resistance responses including resprouting or direct regeneration [57,59,60].  Table S1.

Patterns of Forest Structure Following a Hurricane Disturbance
Changes in recruitment, growth, and mortality through time explain Hurricane Hugo's long-term effects on the forest structure and composition. Variation in the immediate mortality among PFTs and among stem sizes resulted in an immediate change of forest size structure and species composition. The variation in diameter growth and recruitment with time resulted in a shift in size structure as the forest recovered. Before Hurricane Hugo, the forest at Bisley was dominated by Late PFT with diameters less than 10 cm (Figure 6). Because Early had the highest mortality and Palm had the lowest mortality among the four PFTs during the hurricane disturbance, Early PFT decreased in abundance right after the hurricane, while Palm increased its proportion in the forest. As smaller stems had higher immediate mortality, the size structure of the forest shifted from being dominated by small stems (2.5-5 cm) to a uniformly distributed size structure immediately post hurricane ( Figure 6). In an experimental manipulative study of hurricane effects, Shiels et al. [3] found that forest responses to hurricane effects were driven by changes in the canopy structure and openness of the canopy, more than from fallen hurricane debris. In our study, when the canopy opened due to the disturbance, both Early and Palm had higher recruitment than mortality, which resulted in increases in their abundance five years after the disturbance. Twenty years after the hurricane disturbance, the forest was again dominated by Late PFT with diameters less than 10 cm, which may be the beginning of the self-thinning stage until it is interrupted by the next disturbance [56,57]. These changes in PFT abundances have been documented in other tropical forests in response to changes in climate [58]. In our study system, the response to changes in the postdisturbance abiotic environment, inherently altered forest composition which in turn lead to concomitant changes in forest structure. Despite these shifts, the proportion of larger stems from the Late PFT did not change significantly throughout the 25-year study period. The maintenance of "non-pioneer" or Late PFT is associated to their documented posthurricane resistance responses including resprouting or direct regeneration [57,59,60].

Composition Dynamics of Dominant Species and Plant Functional Types
Changes in subsequent mortality and recruitment among the PFTs resulted in further shifts of species composition in the forest. As the canopy closed with progression of successional trajectories after the hurricane disturbance, the mortality rate of all PFTs increased. This brought compositional change, as mortality rates of Early and Mid PFTs exceeded their recruitment rates. Meanwhile the consistently lower mortality and higher recruitment of Palm resulted in making it the most dominant species during the posthurricane successional trajectory, as palms increased in all topographic settings (ridge, slope, valley) in the forest, and tolerated shade under a closing canopy which contributed to their low subsequent/background mortality. The resistant and resilient characteristics of palms favored their dominance in the successional post-hurricane forest, and this is reflected in their distribution in the steep, sloped and rugged montane terrain [4,54,61]. Twenty years after the disturbance (in 2009), mortality rates of Late and Palm began to exceed their recruitment rates, which decreased their relative total abundance in the forest.
Although there was an increase in Palm during post-hurricane succession, it occurred within the established hierarchical dominance of the same species that consistently remained the most abundant in the forest during the 25-year period of documented succession. The Late PFT species-D. excelsa and S. berteriana-are characteristic species that define the vegetation association of this forest type [24], and they maintained proportional dominance along with the increase in palm abundance. Burslem et al. [17] also found that over a 30-year period, there was a constancy of the most abundant species after a hurricane disturbance. The observed heterogeneity of species composition and shift in abundances of the dominant species at the study site may respond to changes in the forest environment during succession. Changes in forest climate and abiotic environment, similar to those observed in post-hurricane canopy and associated understory light conditions, have been identified as drivers of change in PFT successional patterns [62]. Local topographical variation in tropical forests has been known to mediate structure and composition [63,64], and also play a role in the effects from disturbances such as hurricanes and climatic conditions [14,38,65,66].

Concluding Remarks
In this study, we linked the changes in structure and composition of the forest with the changes in demographic components (i.e., recruitment, growth, and mortality). We showed that the structural and compositional change in the first 5-10 years after the disturbance were due to the high recruitment of Early successional species, which leads to compositional change where Early PFT dominates, and to structural change where small stems dominate the forest. During succession, mortality increases with time due to canopy closing, which is mostly from new recruits, leading to the forest thinning and structural change to a median-size dominated forest. The growth rate decreases with time, making the basal area and aboveground biomass accumulation slow. Even though the biomass and basal area are reaching steady state, the structure and composition are still changing with different recruitment and mortality rates among different PFTs. Throughout the 25 years of this study, Early PFT consistently maintained the highest mortality and Palm had the lowest. The low hurricane-induced mortality, high recruitment rate, and low subsequent mortality of palms explain their increased abundance in the forest after the hurricane disturbance and during the following decades. Palms have demonstrated resistance and resilience to hurricane disturbance [4,61,67], and are likely to maintain their high abundance in the forest with frequent hurricane disturbances in the future, unless density dependent mortality sets in as with the effect of pathogens [68].
A warming climate will probably lead to increased intensity and frequency of hurricanes that make landfall [21,22]. In fact, before the forest could fully recover after Hurricane Hugo, the successional trajectory of the forest was reset by two subsequent hurricanes, Irma and María, in 2017. The way forests respond to future disturbances will be affected by the size structure and species composition [20], which are continually changing in response to past disturbances [6,38,69]. Tropical forests worldwide continue to be important carbon sinks and contribute to climate change mitigation [70]. Therefore, enhancing knowledge about tropical forest structural and compositional responses to hurricane disturbances under a changing climate is fundamental to understanding their future outcomes and their complex role in the global carbon balance [71][72][73][74][75].
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/f13050796/s1, Supplementary Information S1: Diameter-Filling Algorithm for Missing Data; Supplementary Information S2: Diameter Measurement Error; Figure S1: Boxplot of diameter (DBH) of recruited stems (a) and diameter growth rates (cm 5 years −1 ) of standing stems (b) for the four plant functional types (PFT; Early, Mid, Late, and Palm) using the Bisley Experimental Watersheds data set with unadjusted data negative growth rates. The boxplots show outliers (three standard deviations from the mean), and 0th, 25th, 50th, 75th, 100th quantiles of the data without outliers; Figure S2: Flow chart decision diagram of the algorithm for filling missing diameter (DBH) of individual stems. Reference stems are from a template set, where DBH data of individual stems has no missing values in all census years (1989 to 2014). d i represents DBH at i th census year, and x represents missing-DBH census year(s), and m and n represent two none-missing-DBH census years that are closest to year x; Figure S3: Time series of Basal Area (m 2 ha −1 ), Stem Density (# ha −1 ), and Growth Rate (cm yr −1 ) per each Plant Functional Type (PFT). Adjusted/filled data set in circle symbols with hatched lines and the unadjusted/with missing values data set in square symbols and solid line. The X-axis shows the census years. For growth rates, the year in the x-axis indicates the first census used to calculate five-year-interval growth rate (converted to cm yr −1 by dividing 5). For example, 1989 means the growth rate (cm yr −1 ) between 1989 and 1994, averaged for all trees in the same PFT.; Figure S4: Measurement error (cm) of diameter. (a) Histogram of errors and the mean value of the measurement error. (b) Scatter plot of errors and the corresponding diameter (DBH) values. The Spearman's correlation coefficient (r) between the measurement error and DBH and the corresponding p-value (p) are shown for routine error (<2 cm) and processing error (≥2 cm); Figure S5: The scatter plots of distribution of measurement error and diameter (DBH) for 29 out of 65 species. Species shown are only those that had more than three samples of measurement error. The Spearman's correlation coefficient (r) between measurement error and DBH and the corresponding p-value (p) are calculated for each species. The species that has a significant correlation SLOBER (p < 0.05) is boxed in red; Figure S6: Relative diameter growth rate (a-d) and basal area (e-h), in five censuses for each diameter (DBH) size of (a) Early, (b) Mid, (c) Late, and (d) Palm PFTs. The markers and the bars indicate the mean and standard error, respectively; only values with more than 10 samples are shown. Lines in between square symbols are used for ease of visual connection of sample points and do not represent data; Table S1: Code, Genus and species, Family, and Plant Function Type (PFT) of each species, listed in the order of descending abundance in the forest for each PFT. The scientific name of species follows the Integrated Taxonomic Information System (https://www.itis.gov/ accessed on 17 April 2022); Table S2: Growth rates (mean and standard error; cm yr −1 ) of each PFT in each topographic settings: valley, slope, and ridge. If the difference between two settings for a PFT is significant at 99% confidence level (p < 0.01), then the growth rate is marked with a letter: "v" means significantly different from valley plots, "s" means significantly different from slope plots, and "r" means significantly different from the ridge; Table S3: Same as Table S2, but for the mortality rate (%); Table S4: Same as Table S2, but for the recruitment rate (%). References [76][77][78][79][80][81][82][83][84][85][86][87][88][89] are cited in the supplementary materials.