Mapping bark bacteria: initial insights of stemflow-induced changes in bark surface phyla

ABSTRACT Life on and within bark surfaces is likely affected by many factors, including bark moisture, bark microrelief, bark pH, and the climatology of forests. The complex set of mutually interacting biotic and abiotic factors operating on bark surfaces presents a challenge in disentangling the linkages and connections between bark-dwelling organisms and biosphere-atmosphere interactions that modulate life in the canopy. Microbial communities on bark surfaces are under-represented in the literature. Given our knowledge of microbial diversity in the phyllosphere and rhizosphere and the impacts of climatology on that diversity, we hypothesized that tree bark would create a microenvironment that selects for specific microbes and that climatological and atmospheric influences from different land uses may further influence such microbial communities. Unfortunately, to our knowledge, no such study to test this hypothesis has been conducted. Recognizing this knowledge gap, samples were collected from bark and stemflow from two northern red oak trees in exurban and suburban forest fragments in the mid-Atlantic region before and after a rain event and subjected to DNA extraction and amplicon sequencing. Major phyla present on bark at both sites include Acidobacteria, Actinobacteria, Proteobacteria, and Bacteroidetes. Significant differences in amplicon sequence variants were observed between the two trees and between the northerly and southerly aspects of the sampled trees. The results of this pilot study highlight the critical need for future work examining the interplay among stemflow and microbial community composition and function (as a part of the larger ecosystem) in relation to varying land use. IMPORTANCE Compared with the phyllosphere, bacteria inhabiting bark surfaces are inadequately understood. Based on a preliminary pilot study, our work suggests that microbial populations vary across tree bark surfaces and may differ in relation to surrounding land use. Initial results suggest that stemflow, the water that flows along the bark surface, actively moves bacterial communities across a tree. These preliminary findings underscore the need for further study of niche microbial populations to determine whether there are connections between the biodiversity of microbiomes inhabiting corticular surfaces, land use, and hydrology.

been examined (17), and a more recent publication on corticular communities of Acer palmatum explored the bark surface microbiome (18).
Rainfall partitioning and elemental cycling studies have also gained additional momentum in temperate regions over the past several decades.However, there are still substantial gaps in our understanding of these processes, including those pertaining to the communities of microorganisms living on and within the bark surfaces of plants (4,5,19).Questions regarding the bark microbiome include how species composition of microorganisms varies by ecoregion and phenoseason, how these communities affect stemflow chemistry and elemental cycling, and how bark characteristics that are unique to individual woody plant species might impact the composition of microorganisms present on the bark (5).A limited number of studies have attempted to investigate the inverse of the aforementioned knowledge gaps, the impacts of stemflow on the soil microbial and epiphytic communities (20)(21)(22).In these studies, precipitation passing through a tree canopy (throughfall) and precipitation funneled down branches and stems (stemflow) both transfer nutrients and water to the base of the plant creating a "fertile island" in soils surrounding the trunk.Tree species may play a large part in establishing the soil microbial community, which may vary across an urban-to-rural gradient (22).
The bark surface of every individual tree provides a unique microenvironment in which the air turbulence, temperature, moisture, and pH will be altered from its macroenvironment due to bark texture, aspect, and position within the stand.Trees serve as a source of nutrient enrichment in forested watersheds through rainfall partitioning (23,24).Usually, less than ten percent, but more commonly less than five percent, of intercepted precipitation per projected canopy area gets funneled onto the bark surface and down stems and branches as stemflow.Stemflow represents a highly enriched, point-source input of water and nutrients at the base of trees in a forested ecosystem (5,(25)(26)(27)(28), presumably due to its long residence times on tree surfaces.
The longer residence time of water on bark not only allows for the chemical and biological (rotifers and nematodes are also shown to be transferred to the soil by stemflow) enrichment of stemflow (29)(30)(31)(32)(33)(34) but also provides for the existence of bacteria on the bark surface.Previous studies have shown contrasting effects on bacterial communities associated with the phyllosphere when observing the influence of the distance between trees of the same species (8,15,(35)(36)(37)(38).Extending this distance effect to the cortisphere, we hypothesize that the bark characteristics of individual trees, sometimes even within the same species (e.g., from age or individual tree life history), will induce changes in bacterial community composition present on tree bark due to the unique microclimate generated by bark surface morphology.Due to differences in temperature, non-point pollution, precipitation, nutrient inputs at the base of the tree, biodiversity, and air conditions between an exurban forested site and a suburban forest fragment, we additionally hypothesize that bacterial community composition will differ between sites.As such, we sampled the same tree species, northern red oak (Quercus rubra L.), in both suburban and exurban settings to determine the microbial populations on the bark surface before and after a rain event as well as in the stemflow induced by the rain event.Additionally, we visualized these data with geospatial interpretation to show how microbes may be transported on tree bark and how this transport may vary with hydrology and land use.

Storm event and stemflow
The stratiform rain event took place on 18-19 November 2017.Bark samples were collected before the event, on the afternoon of November 17, and after the event, on the afternoon of November 19, along with stemflow samples.Samples from each site were collected within 2 hours of each other.In the 2 weeks leading up to the storm event, the average air temperature was 7°C.The study trees, and most surrounding canopy trees, still had their foliage.The storm produced 8 mm of precipitation with winds averaging 1.4 m s −1 out of the west-northwest (average 276°).From the two northern red oak trees, this event produced 150 mL of stemflow at Fair Hill (exurban) and 23.91 L of stemflow at Banning Park (suburban).The differences in stemflow production between these two trees may be attributable to the individual branching structure of each tree, the variation in aspect (SE vs. SW) and potential rain shadows from surrounding trees as a result thereof, and the existence of some preferential flow path for stemflow on the tree at Banning Park.

Sequencing results
A total of 68 bark samples (32 samples from each tree's bark), two stemflow samples (one from each site for the event), and one sample from open precipitation were collected for microbial analysis.The experimental procedures used in this study produced 259,284 sequences from bark and stemflow.
An unpaired, two-tailed t-test of the Shannon Diversity Indices (P = 0.9722) suggests that, by conventional criteria, the two sites have similar amounts of bark community diversity (H = 1.96 for Fair Hill; H = 1.93 for Banning Park, when H max = 3.43).The major phyla of the bark samples at both sites include Acidobacteria, Actinobacteria, Proteobacteria (especially Alphaproteobacteria), and Bacteroidetes, with smaller relative abundances of Verrucomicrobia (Fig. 1; sample descriptions can be found in Table S1).Although community differences were primarily driven by the site (Fig. 2C and 3), some of the major taxa also varied significantly with the aspect from which the sample was taken (Bacteroidetes, P = 0.0242; Proteobacteria, P = 0.0372; and Verrucomicrobia, P = 0.0257) and with height (Actinobacteria, P = 0.0475).The NMDS plot shows the suburban Banning Park bark community samples had less variation among samples than the exurban Fair Hill bark community samples, which had more ecological variation among samples and several samples that were very different from the others (ex. 10 and 26; preand post-event, respectively, both east-facing, at 50 cm height; Fig. 3).
To determine the role of stemflow in microbial translocation, i.e., phyla being washed down the woody surface, the pre-event relative abundance data were subtracted from the post-event values (Δabundance).Locations with positive Δ abundance values have higher relative abundance data for a given phylum after the precipitation event relative to the pre-event conditions, values around zero (±20) show no difference in abundance after the precipitation event, and negative values indicate higher abundance numbers before the precipitation event.Only at the 3.96-m sampling height at Banning Park, the north and east sampling points were all Δabundance positive for all selected phyla (Tables 1 and 2; Fig. 4).Results for selected phyla are shown in Tables 1 and 2 and Fig. 4. For some bacteria {e.g., Chloroflexi, Deferribacteres, Fibrobacteres, Fusobac teria, Gracilibacteria, Hydrogenedentes, Microgenomates, Parcubacteria, Spirochaetae, unassigned bacteria, Tenericutes, Chlamydiae, Elusimicrobia, FBP [former candidate phylum FBP, now Abditibacteriota, as per Tahon et al. (39)], Gemmatimonadetes, and Saccharibacteria}, initial relative abundance is little to none.There is little to no change with the event (median abundance of less than ~15), so they are not included in this analysis.
Odds ratios indicate that measured taxa at the two sites differ in their response to exposure to rainfall events, such that the type of microbe and where it was located on the tree impacted the likelihood of there being more or less of that microbe after it rained during this particular event.At the Fair Hill site, Proteobacteria and Cyanobacteria have odds ratios above 1, while all other taxa are below 1. Comparing the samples by aspect, those collected on the north side of the tree had odds ratios above 1 for all taxa except Cyanobacteria and Verrucomicrobia.In contrast, Cyanobacteria had values above 1 for every other aspect.When the samples are compared by height, Proteobacteria had values above 1 for all sample heights, and Cyanobacteria had values above 1 for all heights except 2.44 m.The 3.96-m height had the most taxa with values above 1, and the 1.5-m and 2.44-m heights have the fewest values over 1 (Table 3).
At Banning Park, all selected taxa except Acidobacteria and Bacteroidetes had odds ratios above 1.Comparing the selected taxa by aspect, certain species favor the N-E aspect (Cyanobacteria, Verrucomicrobia, and Proteobacteria), and others favor S-W aspects (Planctomycetes, Actinobacteria, and Firmicutes).At the same time, Acidobacte ria only had an odds ratio above 1 for the western aspect.At all heights, Proteobacteria have odds ratios higher than 1, and the 1.5-m sampling height has the most odds ratios above 1, while the fewest odds ratios over 1 were found at 0.5 m.The odds ratio calculations show similar patterns to the contour plots (Fig. 4).More phyla tend to decrease in abundance with rain events on the north aspect of the tree at Fair Hill, except for Cyanobacteria and Verrucomicrobia (Table 1).
Changes in abundance values illustrate the complexity of the deposition and wash-off of the selected taxa above the stemflow collar.At Fair Hill, Acidobacteria, Verrucomi crobia, Bacteroidetes, and Proteobacteria all have positive Δabundance values for the southern aspect at all heights.In contrast, all other taxa have mixed Δabundance values with aspect and height.At Banning Park, Cyanobacteria, Actinobacteria, Bacteroidetes, and Proteobacteria had positive Δabundance values down the tree bole for the northern aspect.There are slightly greater odds for decreasing phyla Δabundance with rain events on the north aspect of the tree at Fair Hill, except for Cyanobacteria and Verrucomicrobia (Table 3).
The collar fitting for stemflow creates an internal control on rainfall impacts, as it physically prevents stemflow from reaching below it.On the Fair Hill Natural Resource Management Area (NRMA) tree bark above the stemflow collar, Cyanobacteria and Planctomycetes populations decreased after the rain event, while Proteobacteria, Verrucomicrobia, Acidobacteria, Actinobacteria, and Bacteroidetes all showed increased populations after the rain event (Table 2; Fig. 5B).Below the stemflow collar on the tree at Fair Hill, Proteobacteria, Bacteroidetes, and Cyanobacteria corticular populations decreased after the rain event, while Verrucomicrobia, Acidobacteria, Planctomycetes, and Actinobacteria populations all increased after the rain event (Table 2; Fig. 5A).
On the Banning Park tree above the stemflow collar, Firmicutes and Planctomycetes corticular populations decreased after the rain event, while Cyanobacteria, Verrucomi crobia, Actinobacteria, Bacteroidetes, Acidobacteria, and Proteobacteria populations on bark increased after the rain event (Table 2; Fig. 5B).Below the stemflow collar, Actinobacteria and Proteobacteria populations were both less after the rain event, and Acidobacteria, Bacteroidetes, Cyanobacteria, Verrucomicrobia, and Planctomycetes all showed larger corticular populations after the rain event (Table 2; Fig. 5A).
Banning Park stemflow samples were more diverse than the stemflow collected at Fair Hill NRMA (Shannon Diversity Indices of H = 1.52 and H = 0.88, respectively).Fair Hill NRMA stemflow and precipitation samples were dominated mainly by Proteo bacteria (particularly Gammaproteobacteria), Bacteroidetes, and Firmicutes (Fig. 1 and  2).On average, the abundance of Cyanobacteria and Planctomycetes was greater in stemflow at Fair Hill than in the open precipitation, while Proteobacteria, Verrucomi crobia, Acidobacteria, Actinobacteria, and Bacteroidetes show greater populations in precipitation than in stemflow (Table 2; Fig. 5C).Banning Park stemflow was dominated by Bacteroidetes and Firmicutes, and on average, Firmicutes and Planctomycetes show greater populations in stemflow than in precipitation, while Cyanobacteria, Verrucomi crobia, Actinobacteria, Bacteroidetes, Acidobacteria, and Proteobacteria populations are greater in the open precipitation (Table 2; Fig. 5C).

DISCUSSION
Our results are similar to the study by Leff et al. (17) using G. biloba and the study by Kobayashi and Aoyagi (18) on A. palmatum, which found large numbers of Acidobacte ria, Actinobacteria, Bacteroidetes, and Proteobacteria on trunk and bark tissues, some differences with height, and greatest phylotype richness on trunk and bark tissues.The Shannon Diversity Indices reported are on par with other aqueous and environmental samples (40).Bacteria on the bark surface and some deposited bacteria can be relocated down the trunk via precipitation partitioning, similar to nutrients and water (20)(21)(22).The stemflow collar, placed 1.3 m above the ground, created differences in abundance above and below the collar.Differences between the populations above and below the stemflow collar suggest that stemflow could be a mechanism of "repopulation" for these phyla.The stemflow collar effectively captures and diverts a majority of stemflow that would otherwise flow to the soil; if the stemflow is blocked, there is no increase in some phyla below the collar for those events, which suggests that stemflow may be a source for the deposition of bacteria to the trunk bark surface.The Δabundance values below the stemflow collar represent the change in abundance with stemflow excluded, meaning there would be no downward flux of microbes and any changes in values are due to direct precipitation interception by the trunk, soil splash, or biotic response to lessened atmospheric evaporative demand.But the Fair Hill stemflow collar was placed just before the event, while the Banning Park stemflow collar had been in place for 2 years.It is possible that the Banning Park tree could have developed a "stemflow-less" microbial ecosystem, while the Fair Hill tree below the stemflow collar may still represent a more stemflow-dependent ecosystem.Regardless, the phyla with the highest levels of depletion post-event are the same phyla that appear most commonly in stemflow: Proteobacteria, Cyanobacteria, and Bacteroidetes.
The sites show distinct clustering by site, suggesting suburban and exurban locations are a primary factor in discriminating the corticular microbiome (Fig. 2).The separa tion of exurban versus urban (and presumably dense suburban) microbiome has been established in numerous biomes, including the built environment (41), air (42), and soil (41).Considering its more rural, exurban location, Fair Hill NRMA may have an external source of Proteobacteria, P. agglomerans, possibly from hay baling in this semi-agricul tural area (43).
We expected that the most important controlling factor of differences in populations pre-and post-precipitation would be the rain itself and that we would see greater separations of the populations before and after the event due to moisture availability.Event occurrence was the least likely predictor in population differences after site, height, and aspect.The third most important factor, aspect, is primarily related to insolation.South-facing bark will experience more solar radiation throughout the day and therefore be warmer and drier than north-facing bark, which will be cooler and wetter.The higher moisture availability and likely existence of shade and damp-loving mosses and fungi with which bacteria are commonly associated lead us to believe this factor would also be more important.Height was an important factor in differentiating phyla, which we expected would be due to soil bacteria nearer the roots and leaf bacteria at the higher sampling locations.Still, even with this factor's importance, there seems to be a little discernible pattern as to why.Using the odds ratios (Table 3), we can directly compare the odds of a phylum increasing with aspect or height.It is possible to see that, beneath the stemflow collar, odds ratios favor increases in most phyla after the rain event, while Proteobacteria are most likely to decrease or stay nearly the same at all sampling locations.No aspect shows consistent Δabundance at all heights, as expected.But we propose that in addition to the disruption of stemflow (and therefore the transport of microbiota) by the stemflow collar, changing furrow networks of the bark may be responsible for the apparent lack of sequence or order to samples by either aspect or height on the trunk.The results of this pilot study indicate that an ecologically diverse community of bacteria is present on the bark surface of northern red oak at these two sites.We were able to show that site [Fair Hill NRMA (exurban) or Banning Park (suburban, Fig. 2)] is the most influential factor driving the differences in ecology between samples.Higher stemflow enrichment ratios at the Banning Park site (44) could be a key factor in the differing relative abundances, as the wealth of nutrients lowers the need for competition and specialization.Since this is a pilot study, more sampling and true replication would yield more robust results.Future work will require replicated sampling for multiple events and on multiple trees and simultaneous testing of solutes in stemflow for those events.The corticular microbiome may influence tree and soil health while being heavily influenced by micro-environmental niche space.In summary, our preliminary pilot study suggests that microbial populations vary across tree bark surfaces and may possibly differ in relation to surrounding land use but future work examining the interplay among stemflow and microbial community composition and function (as a part of the larger ecosystem) in relation to varying land use is needed.

Study site details
The sample collection for this study took place in the mid-Atlantic region of the United States, in one exurban forested area (Fair Hill Natural Resource Management Area) and one suburban forest fragment (Banning Regional Park).The University of Delaware maintains a long-term use agreement with Fair Hill NRMA.Fair Hill NRMA is in Cecil County, MD (elevation 73 m), in a humid subtropical region of the Köppen climate classification.The study tree is located on a south-west-facing slope in a 12-hec tare watershed within the NRMA near the Maryland-Pennsylvania border (39°43′15″N 75°49′51″ W).Historically, the NRMA was a mix of farms and woodland, a patchwork of open fields, and mixed-deciduous forest.The mean leaf area index (LAI) for the site is 5.3 m 2 m −2 , the mean diameter at breast height (DBH) is 40.8 cm, and the mean tree height is 27.8 m (45).
The Banning Park suburban forest fragment is in New Castle County, DE.The forest fragment is approximately 44 hectares (elevation 10 m, gentle south-east slope) located less than 33 km east of the NRMA in a small, unincorporated area on the outskirts of metropolitan Wilmington, DE (39°43′03″N 75°35′45″ W).Primary tree species in Banning Park are similar to those in Fair Hill NRMA, but the Banning Park forest fragment is bordered by residential neighborhoods, regional rail lines, and major roadways (I-95 and I-495 interchange, route US 202, DE-141, and DE-4).Banning Park is located within a mile of the GM Boxwood Industrial Brownfield site (operating 1947-2009) and its accompanying railyard (suggesting the potential for legacy pollutants, especially within the soils) (46) and an active campus occupied by BASF, the second-largest chemical manufacturing company in North America.The mean LAI for the suburban site is 3.83 m 2 m −2 , the mean DBH is 31.8cm, and the mean tree height is 18.8 m (44).
Northern red oak was selected for this study due to its presence and abundance at both sites.At the time of collection, other tree species within both stands were at varying stages of senescence and leaf drop, but the study trees were early in senescence and still maintained green leaves.Morphological advantages of red oak for this study include the deeply furrowed bark texture (ideal for bark-surface-scale microclimatology studies) and a round canopy with branch architecture that favors stemflow production (47).Finally, the frequent use of red oak in suburban and urban areas as "street trees" across most of the country (47) will expand the applicability of the results of this study.

Collection: bark tissue and stemflow
Using a ladder to access higher sampling points, uniform outer-bark tissue cores were collected at four heights (0.5 m, 1.5 m, 2.44 m, and 3.96 m) and in four cardinal directions (N, S, E, and W) from a single red oak tree at each site before and after one rain event using a 0.25" hollow punch.This resulted in two samples from each direction at each height, for a total of eight.The location of the initial core was marked with a white permanent marker, and the second core was collected at the same height, as close to the first core as possible, without overlap after the storm event.The punch was sterilized with 70% isopropyl alcohol followed by flame and rinsed in deionized water after each sample was collected, and the sample was placed into a labeled Nasco Whirl-pak bag.Pre-storm samples were collected on 17 November, before one forecasted frontal storm system, and post-storm samples were collected within 24 hours after the end of the event on 19 November.Collections from both trees took place within 4 hours on each collection day.Between lab procedures, samples were stored in a freezer to prevent changes to bacterial populations and the degradation of bacterial DNA.True replicates for this study were cost-prohibitive (48).Bacterial DNA was extracted from the bark tissue using a Qiagen DNeasy PowerSoil Kit (Hilden, Germany).
Rainfall depth and intensity in the open (i.e., not under the forest canopy) were measured using two meteorological stations closest to the study sites, equipped with tipping-bucket gauges (Fair Hill, MD, and Wilmington, DE-Talleyville) which are part of the Delaware Environmental Observing System, and an open precipitation sample was collected using an amber glass vial with a funnel attached and glass wool filter, posted at the Fair Hill, MD meteorological station.Stemflow collectors were attached to the two trees being sampled for microbial diversity.Stemflow collectors consisted of a polyethy lene stemflow collar stapled and then sealed around the tree's bole, which directed generated stemflow into a polyethylene storage container (45).Stemflow volume was measured, and a stemflow sample was collected into a sterile 200 mL high-density polyethylene (HDPE) container for DNA extraction using a Qiagen DNeasy PowerSoil Kit (Hilden, Germany).Stemflow samples were collected after the same events, at the same time as post-storm bark samples, from the same trees for which microbial samples from bark were collected.

DNA extraction and 16S rRNA gene sequencing and analysis
Aqueous stemflow samples were vacuum filtered onto a 0.22-µm mixed cellulose GSWP filter (Millipore, Burlington, MA) to collect microbial cells and immediately frozen at −20°C before DNA extraction with Milli-Q DI water as a negative control.One hundred milliliters of the material was filtered for each sample, except for Fair Hill NRMA, which had a 60-mL filter.The outermost layer of the tree bark samples was separated using a razor blade (flame sterilized after each use) and saved for subsequent DNA extraction.DNA was extracted using the Qiagen DNeasy PowerSoil Kit (Hilden, Germany).Each filter or tree bark scraping was placed directly into a provided bead-beating tube.The manufacturer's protocol was followed.DNA yield was determined using the Quant-iT PicoGreen dsDNA Assay Kit on a Tecan Infinite M200 Plate Reader.A blank of deionized (DI) water was run as a negative control.
Paired-end 2 × 250 bp Illumina MiSeq (Illumina, San Diego, CA) sequencing of the 16S rRNA gene was performed at the Environmental Sample Preparation and Sequencing Facility at Argonne National Laboratory using the universal V4 region primer set [515f/ 806r; (49)] using Environmental Sample Preparation and Sequencing Facility protocols.QIIME2 [version 2018.4.0; https://qiime2.org;(49,50)] was used to analyze the sequences, which included denoising using DADA2, calculating alpha and beta diversity metrics, and taxonomic assignments using the SILVA132 database.In the denoising process, both forward and reverse reads were trimmed to 146 bp to eliminate regions with low overall quality scores as determined by visual inspection in QIIME2.Before diversity measurements and taxonomic assignments, unassigned sequences, chloroplasts, and mitochondria were removed from the data set, which was subsequently rarefied to 3,813 sequences per sample for downstream analysis.

Additional visualization and analysis
Phylum Δabundance was visualized as a contour plot with the y-axis representing the height of the sampling location (m) and the x-axis representing the aspect using ggplot in Python.The north direction is shown at both ends to represent the closing of the cylinder, which is the tree bole.Data were smoothed using a third-order polynomial spline to aid in the interpretation of bark surface patterns of abundance.The blue vertical line represents the predominant wind direction for the storm event sampled, while the horizontal dashed line represents the height of the stemflow collar.Green areas show where a given phylum is more abundant after the storm event, white shows areas of no difference, and the brown end of the color map represents areas where phylum is less abundant after the storm event.
Odds ratios, a measurement of the association between exposure and outcome, were used to further understand the presence of phyla at the site.In this case, "exposure" is the rain event, resulting in either an increase or decrease in the given phylum.The closer the value is to 1, the more likely it is that the selected factor has little effect on the odds of that outcome (e.g., rain events have hardly any effect on the odds of finding an increase or decrease in Proteobacteria on south-facing bark in Banning Park and Fair Hill).A result of >1 means the exposure to the rain event is associated with higher odds of decreasing numbers of that phylum, while results <1 mean exposure to the rain event is associated with lower odds of decreasing numbers of that phylum (or an increase).
Significance and predictor screening analyses were conducted in SAS JMP 16.0.0.Specifically, P values are the results of one-way ANOVA unless otherwise stated.The suitability of this test was determined using Q-Q plots.Simpson's diversity index was calculated in Excel.

FIG 2
FIG 2 Heatmap of abundance for both bark tissue and stemflow and rainfall.(A) Abundance in stemflow from Banning Park and Fair Hill and the open precipitation sample from Fair Hill.Primary abundance in FH SF and Open Pg is P. agglomerans, followed by the Gammaproteobacteria genus and Bacteroidetes.(B) Unassigned bacteria are most abundant in bark tissue samples, as well as Acidobacteria and Proteobacteria.(C) Bark sample group primarily by site more than by height, direction, or time, although these factors are critical for individual phyla.Red lines indicate a lack of statistical significance.Heights labeled 1 through 4 represent 0.5 m, 1.52 m, 2.44 m, and 3.96 m, respectively.

FIG 3 TABLE 1 a
FIG 3 Non-multidimensional scaling plot of samples.Fair Hill samples are red, and Banning Park samples are blue.Stemflow collected from Fair Hill is brown, and stemflow collected from Banning Park is grey.Lab blanks are black.A clear separation is seen by site, also by sample type.The low biomass stemflows are close to the lab blank but not identical, with greater variation in Fair Hill stemflow.

FIG 4
FIG4 Contour plot of phylum Δabundance.Phylum Δabundance was visualized as a contour plot with the y-axis representing the height of the sampling location (m) and the x-axis representing the aspect.The north direction is shown at both ends to represent the closing of the cylinder, which is the tree bole.Data were smoothed using a third-order polynomial spline to aid in the interpretation of bark surface patterns of abundance.The blue vertical line represents the predominant wind direction for the storm event sampled, while the horizontal dashed line represents the height of the stemflow collar.Green areas show where a given phylum is more abundant after the storm event, white shows areas of no difference, and the brown end of the color map represents areas where phylum is less abundant after the storm event.Note the differences in color ramp values for different phyla.

FIG 5
FIG 5 Difference plots show the changes in the populations above the stemflow collar before and after rain events: (A) Δabundance on bark surface below the stemflow collar after a rain event.Values below the zero line indicate depletion of the population on bark after a rain event; values above the zero line indicate an increase in the population on bark after a rain event.(B) Δ abundance on the bark surface above the stemflow collar after a rain event.Values below the zero line indicate depletion of the population on bark after a rain event; values above the zero line indicate an increase in the population on bark after a rain event.(C) Most abundant phyla remain in stemflow after the subtraction of bulk precipitation.Values above the zero line indicate greater abundances of respective phyla in stemflow than rainfall.

TABLE 2
For clarity, the following tables provide the mean, median, and upper and lower percentile values of the Δabundances for dominant phyla at the two sites, both above and below the stemflow collar (39)P is former candidate phylum FBP, now Abditibacteriota(39).

TABLE 3
Odds ratios for the major phyla at the two sites for both (a) north, east, south, and west aspects and (a) 0.5-m, 1.52-m, 2.44-m, and 3.96-m heights a a Bold values indicate odds ratios greater than 1.