Data from The impact of human disturbances on the regeneration layer of tropical rainforests

We assessed the impact of fire, exotic species, logging and defaunation on regenerating tropical plants through meta-analysis. <b>Literature survey and data inclusion criteria</b>We searched for the available literature on forest regeneration responses to five major human disturbances in SCOPUS (January 1976 - September 2023). We included the following search terms: TITLE-ABS-KEY (logging OR "non-timber forest product*" OR harvesting OR defaunation OR hunting OR poaching OR “chronic disturb*” OR "human disturbance*" OR "anthropogenic disturbance*" OR "invasive species" OR "exotic plant*" OR "exotic species" OR “alien species” OR "forest fire*" OR wildfire*) AND TITLE-ABS-KEY ("tropical moist forest" OR "tropical forest" OR "tropical wet forest" OR rainforest OR "tropical humid forest" OR "rain forest") AND TITLE-ABS-KEY (seedling* OR sapling* OR "juvenile* tree*" OR "plant* recruitment" OR "tree* recruitment" OR "regenerating tree*" OR “forest regeneration”) AND NOT TITLE-ABS-KEY ("dry forest" OR "seasonal forest" OR subtropical OR "dry tropical forest" OR "temperate rainforest" OR savanna* OR scrubland* OR grassland*). We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework to ensure that our systematic review complies with scientific quality standards (Moher et al., 2010; O’Dea et al., 2021). We restricted our search to double-refereed articles or book chapters in English or Spanish, excluding grey literature and reviews (which were reviewed only to find articles not detected by our systematic search). The research was executed from Mexico. This search resulted in 591 studies, and additionally we found several relevant studies while reviewing literature (figure 1(a)). After filtering out irrelevant results (figure 1(b)), the number of studies was reduced to 260 (figure 1(c)). We then examined these articles in detail to determine whether they met the following five criteria. (1) The studies had to be conducted in regions with Af or Am climates according to the Köppen-Geiger Climate Classification (Beck et al., 2018); we assume that only these two categories unambiguously encompass the world’s humid tropical forests. We excluded studies in regions hosting tropical dry forests because they are subjected to very different drivers and limiting factors (Murphy &amp; Lugo, 1986; Poorter et al., 2019) and may have specific responses to human disturbances. (2) The study design should be appropriate for testing human disturbance effects on the regeneration community, which required comparing undisturbed (i.e. control) and disturbed (e.g., logged or defaunated) sites; we included one study that focused on a degradation gradient given the low number of studies for chronic disturbances. (3) The studies should separately assess the effects of disturbance on seedlings, saplings, or juveniles; studies that failed to distinguish these categories were used only if their results were separately presented for individuals with different diameter classes. We excluded studies reporting effects on the regeneration community but that used too wide ranges of size (e.g., diameter) classes (e.g., studies that examined together all plants &lt; 10 cm DBH), because such grouping of information could prevent the proper identification of specific disturbance effects on each life cycle category. (4) Studies focused on vines and/or herbs were excluded because these life forms have different environmental requirements than trees and palms. Finally, (5) the studies were required to report the mean, sample size, and a dispersion measure of the dependent variables (i.e. standard error, standard deviation) or to present figures from which these metrics could be extracted. Additionally, we excluded chronic disturbances, as the number of studies were too low (n = 5) to have reliable results. After this filtering, 99 studies encompassing 773 comparisons were included in the analyses (figure 1(d)).<b>Partitioning the regenerating community in the seedling, sapling, and juvenile components </b>Forest regeneration encompasses plant individuals in the early stages of their life cycles (i.e. seedlings, saplings, and juveniles), as these young plants represent the potential to replace dying adult canopy individuals. A major challenge when analysing the effects of human disturbances on these regeneration categories is the lack of consistency in their definition among studies (Winkler et al., 2024). For example, seedlings have been defined as individuals that retain the cotyledons (e.g., Farwig et al., 2009), but also as individuals &lt; 5 cm in DBH (e.g., Swaine &amp; Agyeman, 2008). Therefore, we needed to standardize the database by homogenizing the definition of each regeneration category. First, we defined seedlings as those individuals that still bear cotyledons (and/or a few true leaves) that are &lt; 50 cm tall. We also included in the seedling category relatively young individuals (&lt; 6 months old) established from germinated seeds. This definition implies that seedlings are individuals that still depend, totally or partially, on seed resources, making of this phase the most critical for plants with the highest mortality among stages of the plant life cycle (Eriksson &amp; Ehrlén, 2008). Next, we defined saplings as those individuals ≥ 50 cm tall but &lt; 2 cm in DBH. The sapling category also included: (i) individuals that were left to grow up to one year, because at that age the plant is likely capable of obtaining resources without seed dependence; (ii) individuals reported to measure &lt; 1 m in height without specifying the minimum height; and (iii) individuals reported to measure 10 to 100 cm in height, as we consider that most of them could fall in our sapling category. Finally, we defined juveniles as individuals &gt; 2 but ≤ 10 cm DBH. This stage of the plant life cycle represents an intermediate phase between very young individuals and adult trees (usually defined as individuals &gt; 10 cm DBH; e.g. Arroyo-Rodríguez et al., 2012). Survival at this stage is usually much higher than in the two previous regeneration categories, but they may still be more susceptible to some human disturbances (e.g., fire) than their conspecific adults (Slik et al., 2008).<b>Effect size calculation and moderators</b>From each study, we extracted the information for undisturbed (control) and disturbed treatments. Specifically, we recorded the mean values, a measure of data dispersion (e.g., standard error, standard deviation), and the sample sizes for the control and the treatment. When the study did not provide the data in text or tables, we obtained them from figures using WebPlotDigitizer version 4.6 (https://automeris.io/WebPlotDigitizer/). As we used several moderators (see below), several studies reported more than one comparison. We used Hedges’ g index (Borenstein et al., 2009) to calculate the effect size and confidence interval for each comparison. The Hedges’ g index estimates the standardized mean difference and variance between a control and treatment sites and can be used in studies with a small sample size (Gurevitch &amp; Hedges, 1999). In particular, negative values of Hedges’ g indicate a decrease in the response variable (e.g., richness or abundance) in disturbed sites, whereas positive values indicate the opposite. Effect sizes were considered statistically significant when their confidence intervals did not include the zero value. When possible, confounding factors were avoided to have only the pure effect of the disturbance of interest.We estimated the overall effects of four major human disturbances on tropical regenerating community. Then, we defined several moderators, some of which were shared among different human disturbances (e.g., response variables) whilst others were specific to each of them. The two shared moderator variables were: (i) response variable (e.g., abundance, species richness), and (ii) regeneration stage (i.e. seedlings, saplings or juveniles). The specific moderator variables were: (i) ecological guild (for defaunation, exotic/invasive species, and logging); (ii) time since disturbance (for fires and logging); (iii) recurrence (for fires); (iv) defaunated group (for defaunation); (v) animal size (for defaunation; each category was assigned based on the category assigned by the authors of the study); (vi) dispersal syndrome (for defaunation; (vii) logging method (for logging); and (viii) study type (for all disturbances except fires). Given the large number and variety of response variables, we classified them into eight major groups based on their affinities (Table S1).