The seagrass study sites were located near the Redfish Bay State Scientific Area in Aransas Pass, Texas, USA ( Figure 1 ). The study location is bounded by the City of Aransas Pass Wastewater Treatment Plant on the northwest, the Intracoastal Waterway (ICWW) to the southeast, and centered around the ICW RV Park. Three study areas were selected based on perceived differences in epiphyte communities on the monotypic Thalassia testudinum beds in areas with different exposures to nutrient sources ( Figure 1 ; Table 1 ). The WWTP area received direct discharge of treated wastewater effluent (WWTP; Shallow: N 97.14861, W -27.88485; Medium: N 97.14790, W -27.88467; Deep: N 97.14761, W -27.88435). There were additional possible impacts from a fish cleaning station and associated feeding and roosting of pelicans. The “Control” area received indirect discharge of treated wastewater effluent that was additionally polished by wetlands (“Control”; Shallow: N 97.15044, W -27.88053; Deep: N 97.15002, W -27.88021). Both WWTP and “Control” areas were near mangrove stands as well. A third site called CI was near the “Control” area, but distal to the wetland. It had additional potential impacts from an adjacent subtidal oyster bed and the ICWW channel (CI; N 097.15044, W -27.88485) Site CI is impacted by boat and barge wake that washes over the shallow oyster reef. It is the most shallow site, with visibly higher epiphyte accumulations. All sampling stations are primarily sandy mud sediment, but the CI site additionally has an abundance of embedded oyster shell. Each site was expected to differ primarily by nutrient conditions, and preliminary observations suggested differences in the epiphyte communities.

Field conditions (depth, salinity, temperature, and general observations of wind, weather, etc.) were recorded at each visit to each sampling site to provide contextual information on factors that may affect seagrass condition and epiphyte. Depth was measured with a calibrated PVC pole. Salinity was measured using a refractometer (VEE GEE, STX-3), under standard conditions in the laboratory, on 50 mL water samples collected at seagrass canopy depth, without headspace. The temperature of the water column near the seagrass canopy height was measured with a calibrated thermometer.

Sediment for porewater analysis was sampled in May 2020 in triplicate from the six sampling sites ( Figure 1 ) using a 10 cm inner diameter PVC corer inserted to a depth of 20 cm. Sediments from a zone of 5-15 cm depth, which represents the root zone of the seagrass, was collected from the cores and centrifuged at 5000g for 20 minutes at 4°C for porewater extraction. After centrifugation, supernatants were filtered through VWR glass microfiber filters, grade 696 (particle retention: 1.2μm). The filtrates were stored at -25°C and shipped frozen for nutrient analyses at the University of California-Davis (UC Davis) Analytical Lab (Davis, CA, USA). Nutrient analyses included ammonium ( N H 4 + ), Nitrate ( N O 3 - ), total dissolved inorganic nitrogen (DIN), and phosphate ( P O 4 3- ). The nutrient levels of sediment porewater measured at this single time are considered to be an indicator of long-term relative nutrient exposure at each site.

The epiphyte-seagrass dynamics were quantified by biomass measurements and image-based metrics (Huang, 2020). Shoots of Thalassia testudinum were collected monthly or bimonthly at the six sampling locations from 2019 summer to 2020 spring. Regular sampling (at least bimonthly) across different seasons was necessary to capture the morphological response of seagrass in relation to accumulation patterns of epiphytes. The volume and time requirements of seagrass processing work necessitated that sampling in the “Control” and WWTP areas be staggered at approximately two- to three-week intervals. Three rings made of 7.6 cm diameter PVC pipe were placed haphazardly for triplicate replication of seagrass samples. Seagrass shoots located inside the ring (an area of approximately 45 cm²) were harvested with all leaves intact to equitably represent the morphological variation in seagrass growth and epiphyte accumulation. Collected seagrass shoots. placed into bottles without water to minimize disruption of the epiphytes, were stored dark and cold (4°C) at the lab, and imaged within 72 h.

For analyses, seagrass shoots were gently rinsed with deionized water, and the green portion of each individual leaf was measured for length and width and arranged on a fluid mount tray (Epson, Carson, CA) for imaging. As needed, microscope slides were used to weigh down twisted seagrass leaves to lie flat. Leaf images were captured by the Epson Perfection V-750 Pro color flatbed scanner (Epson, Carson, CA) using 24-bit color scanning at 1200 dpi resolution. After scanning, epiphytes were removed from each seagrass blade by scraping with a microscope slide. Epiphytes were quantitatively transferred to a pre-weighed empty aluminum dish. The epiphytes and epiphyte-free seagrass were dried to constant weight at 60°C to measure the biomass of each individual leaf.

Scanned seagrass leaf images (TIFF files) were analyzed to characterize the total leaf area, the epiphyte covered area and the uncovered leaf area (Huang, 2020). The pixels of seagrass and epiphytes on the images were classified using the spectral angle mapper (SAM) algorithm in ENVI 5.0 (L3 Harris Technologies, Niles, Ohio). Each image pixel was interpreted by a vector consisting of three hues of the light spectrum (Red band, Green band, and Blue band). The SAM algorithm distinguished the uncolonized seagrass leaf and epiphyte-covered areas based on a spectral angle between the vectors of targeted pixels and reference pixels in three-dimensional coordinates. The assignment of each pixel as seagrass or epiphyte was made using a threshold of spectral angle (< 2.3°). 482 reference pixels of seagrass, which had visually different colors due to variable growing conditions, were selected for establishing the seagrass spectral library. The epiphyte spectral library contained 843 reference pixels of epiphytes from variously colored groups of epiphytes. Finally, 2061 scanned images of each seagrass blade were analyzed via the seagrass and epiphyte spectral libraries.

Normalized and un-normalized biomass and image-derived metrics were evaluated. Correlations between epiphyte biomass vs. epiphyte covered area (number of epiphyte pixels), and between epiphyte load (epiphyte biomass/seagrass biomass) vs. leaf coverage % by epiphyte (epiphyte pixels/pixels of whole blade) were quantified by several regression models, including the linear, quadratic, power law, and the exponential models (Fong and Harwell, 1994). The best model was then selected, given the goodness of fits, such as linearity, R² value, the Akaike Information Criterion correction (AICc), and the intrinsic biological meaning (Johnson and Omland, 2004). The regression analyses were drawn from samples within different environmental contexts or sampling times to demonstrate their site- and seasonal- patterns via R version 3.6.2.

Stable isotope analyses were performed (separately) for seagrass and epiphytes to inform the physiology and status of T. testudinum and its epiphytes under the different environmental conditions at each sampling site. Four shoots of T. testudinum were collected from each of the 6 sampling sites in June 2020. The seagrass and epiphytes were processed separately as described above to obtain dried samples, which were then ground using a mortar and pestle. Leaf and epiphyte tissue content (total carbon and total nitrogen) and stable isotope ratios (δ¹³C, δ¹⁵N) were analyzed at the UC Davis Stable Isotope Facility (Davis, CA, USA). The mean standard deviation for reference material replicates was ±0.03 ‰ for δ¹⁵N and ±0.04 ‰ for δ¹³C, and the mean absolute accuracy for calibrated reference materials was within ±0.05 ‰ for δ¹⁵N and ±0.04 ‰ for δ¹³C.

The spatiotemporal variance of salinity, nutrient level of sediment porewater, and water temperature were tested using analysis of variance (ANOVA). Comparisons of the epiphyte-seagrass dynamics across environmental contexts from July 2019 to April 2020 were analyzed using two-way ANOVA followed by Tukey’s honest significant difference. The different sampling times in the “Control” area, “WWTP” area and CI site precluded the use of ANOVA for statistical comparisons between sampling locations across six sampling times. Instead, data were reallocated to four seasons based on the sampling date and average water temperature ( Table 1 ) to facilitate environmental comparisons. Seagrass growth response was inferred from the mean seagrass leaf biomass and mean seagrass leaf area (number of covered and uncovered leaf pixels), and the epiphyte accumulation was determined by epiphyte biomass per leaf and epiphyte covered leaf area (number of epiphyte pixels). The normalized expressions of epiphyte metrics relative to the host seagrass leaves, such as epiphyte load (epiphyte biomass/seagrass biomass) and image-derived % cover (epiphyte pixels/pixels of whole blade), were used to further understand the seagrass-epiphyte dynamic relationship spatiotemporally. A mixed model (Fixed × Random) nested ANOVA was performed to evaluate the significance of differences in epiphyte-seagrass dynamics among sites of different depths in the “Control” and “WWTP” areas separately. Variables were grouped by sampling times (random factor) nested in different ‘depth’ sites. Differences in the C and N content and the C and N isotope composition were compared between the sampling locations. All comparisons were analyzed using ANVOA, with significant differences between specific groups assessed using Tukey post-hoc test separately. All analyses were conducted via R version 3.6.2. with a false discovery probability of p = 0.1 through the “Westfall” and “Shaffer” procedures for multiple comparisons of balanced and unbalanced data, respectively (Shaffer, 1986; Westfall, 1997).

A true color image of an epiphyte-covered seagrass leaf and the classified outputs derived from the seagrass and the epiphyte spectral libraries are shown in Figure 2 . It takes around 15 seconds to classify a scanning image of an individual seagrass leaf. The seagrass-classified images showed 482 uncovered leaf spectral classes with different colors, including growing areas and senescing or injured areas, which presented different levels of green and yellow or brown, respectively. The epiphyte-classified images contained 843 spectral classes of variable epiphytic groups distinguished by different colors. Due to natural variation and errors of visually estimated differences during spectral library establishment, numerous spectral classes might characterize identical features.

Comparison of mean values of biomass- and imaging-based metrics across all study sites for the entire study period ( Figure 3 ) revealed similar site to site patterns (ordination) of variation. Both epiphyte metrics, biomass load and % coverage, were normalized to the amount of seagrass substrate. Although there were no significant differences in epiphyte accumulation between the WWTP and the “Control” areas, site CI exhibited significantly higher values for both biomass and imaging metrics (p < 0.05). Thus, both biomass- and imaging-based metrics revealed greater relative epiphyte accumulation at CI site. Leaf biomass and blade areas did not significantly differ among the CI, the CS, and the CD site across sampling times. Conversely, there was significantly lower leaf biomass at the WS site than at the WM and the WD site, respectively ( Figures 3A, B ; p < 0.05).

Water temperatures ranged from 16 to 32 °C and were stable in summer and early autumn (29.17~32.00 °C from July to October) ( Table 1 ). Temperatures dropped in November and fluctuated through winter, with an average temperature of 20.75 °C. In April, the mean temperature warmed to 24.00 °C ( Table 1 ).

Salinity levels ranged from 25 – 37 and differed significantly with seasonal change (highest in summer) (df = 3, F = 31.68, P < 0.05), but there was no significant difference in the mean salinity among the “Control” area, CI site, and “WWTP” area ( Table S1 ). However, the transition from Autumn to Winter ( Table 1 ) was accompanied by a decrease in salinity by about 8, as well as a decrease in temperature by 9-13 °C.

The depths of the study sites varied from 29 cm to 125 cm over the course of the study, with mean values ranging from 47.2 cm to 93.2 cm. The relationship of the seagrass and epiphyte metrics to variation in average depth of study sites showed only weak correlations ( Figure S1 ; R² = 0.31, 0.22). Likewise, with the exception of pixels of leaf area, only weak correlations were observed for seagrass and epiphyte metrics with mean temperature and mean salinity ( Figures S2 and S3 respectively). Divergence of the correlations of leaf area and leaf biomass metrics, with temperature and salinity indicated greater variability in the biomass data. Thus, mean depth, temperature or salinity differences do not appear to play a major role in the biomass observations.

The different potential environmental influences on each site from direct vs polished wastewater effluent, a shipping channel or oyster beds prompted an examination of relative nutrient levels. Sediment porewater nutrient levels were taken to be indicators of long-term nutrient exposure history at each site. Table 2 shows the results obtained from a measurement in May 2020. The average concentrations of nitrate ( N O 3 - ) and ammonium ( N H 4 + ) from the CI site sediment porewater were nearly 1.5 times higher than from the “Control” area. However, these differences were not significant among the three locations. The average concentration of phosphate ( P O 4 3- ) at the “WWTP” area was more than two times higher than at “Control” and CI sites (df = 2, F = 8.369, P < 0.05), which was significant ( Table 2 ). The total dissolved inorganic nitrogen (DIN = N H 4 + + N O 3 - ) (45.01~68.55 μmole/L) differed among the three sampling locations but was not significant. However, ratios of DIN:P ( N H 4 + + N O 3 - : P O 4 3- ), varied significantly from 7.82 in the “WWTP” area to 154.55 at the CI site (df = 2, F = 11.93, P < 0.05) ( Table 2 ).

The tissue C and N content and stable isotope ratios were examined ( Figure 4 ) to describe the status of epiphytes and seagrass at the environmentally unique sites. The carbon and nitrogen concentrations of seagrass were significantly higher than in corresponding epiphyte samples ( Figure 4 , ANOVA, p < 0.05). The C concentration of seagrass for the CI site was significantly higher than at the WWTP or “Control” areas. In contrast, the epiphyte C concentration was highest in the “Control” area. The N concentration of seagrass was not significantly different among the 3 locations, but the CI site had the lowest N concentration for the epiphytes. Epiphytes from the WWTP area had the lowest C/N ratio ( Figure 4 , ANOVA, p < 0.05), while seagrass did not show a significant difference in the C/N ratio among the three sampling locations.

The isotopic composition analysis of carbon (δ¹³C) demonstrated the greatest depletion of ¹³C for epiphytes at the WWTP area, whereas for seagrass, the depletion of ¹³C was greatest at the CI site ( Figure 4 , ANOVA, p < 0.05). For δ¹⁵N analysis, the δ¹⁵N values of epiphytes were significantly enriched compared to their seagrass hosts. Although there was no significant difference in the δ¹⁵N values of epiphytes among the three sampling locations, seagrass at the “Control” area was lighter than seagrass from the CI and WWTP sites ( Figure 4 ANOVA, p < 0.1).

Comparisons of the different measures of epiphyte accumulation were made across different seasonal environmental influences. Linear regression between dried biomass of epiphyte per seagrass leaf and the numbers of identified epiphyte pixels per leaf revealed unique accumulation patterns by different sampling locations and seasons. The proportion of variation explained by linear regression ranged from 0.67 to 0.94, and there was a marked and consistent increase in the slope of the regression line, from all sampling areas (environmental contexts), with seasonal progression from summer through winter ( Table 3 ). The different slopes demonstrate a seasonally changing relationship between epiphyte biomass and epiphyte coverage of the leaf. As an example, linear regression of the data from the “Control” area, July 31, 2019, had an R² of 0.81 and a slope of 0.017, representing that there was 0.017 µg epiphyte biomass per identified epiphyte pixel. Epiphyte accumulation gradually increased to 0.0559 µg per epiphyte pixel through winter into spring ( Table 3 ). Similarly, epiphyte accumulation in the “WWTP” area increased from 0.014 µg to 0.033 µg over the same period (but with a peak at 0.0675 µg in February). At the CI site, epiphytes increased from 0.024 µg to 0.0434 µg over the same period (but with a peak at 0.0709 µg in February).

Expression of epiphyte metrics relative to the host seagrass leaves is another approach to represent the epiphyte-seagrass dynamic relationship. The epiphyte load (epiphyte biomass/seagrass biomass) and the image-derived % epiphyte coverage were correlated and revealed seasonal changes in the seagrass-epiphyte relationship ( Figures 5 – 7 ). Potential regression models were applied across six sampling times for the three locations. Two scenarios emerged to describe the relationship between epiphyte load (Y) and epiphyte coverage of the leaf (X) under variable environmental conditions. Analyses of this relationship in the “WWTP” area ( Figure 5 ) showed a strong linear regression (Scenario 1) from August through December 2019. However, the exponential-rise model (Scenario 2) was the best-fitted model in February and April 2020 based on the AICc values ( Figure 5 ). In the “Control” area and CI site, the exponential model was generated during August ( Figures 6 , 7 ) and was also presented at the CI site in April 2020 ( Figure 6 ). The linear relationship between epiphyte load and epiphyte coverage held under most environmental circumstances, except August samplings from “Control” and CI; and April sampling from “WWTP” and CI ( Figures 5 – 7 ). Different slope values revealed different accumulation rates of epiphyte biomass relative to epiphyte coverage. The slopes of this relationship varied from 0.75 to 3.43, with a progressive increase from August to December and with the highest values observed during winter or spring, when the coolest observed water temperatures prevailed ( Table 1 ).

The observed changes in seagrass-normalized epiphyte accumulation could be due to changes in epiphyte growth, seagrass growth, or a combination of both. Figures 8 – 11 present different aspects of each possibility. Because sampling events at different sites did not occur simultaneously, measures were grouped by season for comparisons ( Table 1 ). Seagrass leaf biomass and imaged seagrass leaf area presented similar seasonal patterns in general, where the high levels in Summer or Autumn declined in the transition through Winter and Spring ( Figures 8 – 11 ). There were no significant differences in the seagrass leaf biomass and blade area between the “Control” area and the CI site from summer to spring ( Figure 8 , Table 4 ). The leaf biomass in the WWTP area was significantly higher than at the “Control” area and the CI site in the summer (df = 2, F = 8.10, p < 0.05) and autumn (df = 2, F = 10.06, p < 0.05), but it decreased strikingly and exhibited significantly lower value than the other two locations in the winter (df = 2, F = 3.40, p < 0.05) ( Figure 8 ). In the spring, there was no significant difference in the leaf biomass among the three locations ( Table 4 ).

The blade area also did not exhibit a significant difference between the “Control” area and the CI site from summer to spring ( Figure 9 ; Table S2 ). In the “Control” area (df = 3, F = 5.95, p < 0.05) and the CI site (df = 3, F = 16.50, p < 0.05), the blade areas were both significantly higher in the summer and autumn than in the winter and spring ( Figure 9 ). The blade area from the “WWTP” area was significantly higher than from the “Control” area and the CI site in the autumn (df = 3, F = 11.60, p < 0.05). A sharp decrease in average blade area from the “WWTP” area was observed in the transition to winter when the blade area was significantly lower than at the other two locations. Generally, the seasonal changes of leaf biomass and blade area among three sites with different environmental conditions were similar, with high values in summer and autumn which significantly decreased in winter and stayed low through April ( Table S2 ).

Seasonally aggregated observations of epiphyte accumulations expressed as epiphyte load and epiphyte coverage of leaves showed different patterns by season and site ( Figures 10 , 11 ). Epiphyte loads (epiphyte biomass relative to seagrass biomass) presented an inverse temporal pattern compared to seagrass growth. There were significant differences in epiphyte load among the three sampling locations ( Figure 10 ; Table 4 ). Epiphyte loads at the CI site were significantly higher than at the “Control” area and “WWTP” area in all seasons (df = 2, F = 70.64, p < 0.05). Opposite to the seagrass growth pattern, the epiphyte load pattern increased significantly (df = 2, F = 25.89, P < 0.05) in the winter ( Figure 10 ), except for the “Control” area, where epiphyte load did not change significantly in winter but was instead highest in the spring. On the contrary, the epiphyte load in the “WWTP” area decreased from winter to spring and was significantly lower than in the other two locations in the spring (df = 2, F = 8.03, p < 0.05). Overall, epiphyte loads exhibited seasonal changes inverse to the seagrass growth. The CI site had the greatest epiphyte/seagrass biomass ratio, whereas the “WWTP” area had the lowest values except in the winter.

Epiphyte coverage ( Figure 11 ) did not show a significant seasonal change but was affected by the environments of the three sampling locations. Across sampling periods, epiphyte coverage was highest at the CI site. This difference was significant in all seasons ( Table 4 ). The epiphyte coverage in the “Control” area was significantly higher than in the WWTP area in the summer (df = 2, F = 10.21, p < 0.05) and then there were no significant differences in epiphyte coverage from autumn to spring. Even though the seasonal pattern of epiphyte coverage was not similar to that for epiphyte load, these two indicators both ordinated the CI site as highest across all seasons ( Figure 11 , Table S2 ). The relative consistency in epiphyte coverage across seasons (CI varied from ~40 – 45%; WWTP varied from ~25 – 30%; Control varied from ~25 – 35%) highlights a unique aspect of the complicated epiphyte-seagrass relationship that is not readily apparent from biomass measurements.

The seasonal shift in seagrass-epiphyte dynamics evaluated by image analysis in this study is consistent with previous studies on multiple seagrass species, (Bulthuis and Woelkerling, 1983; Lee and Dunton, 1996; Moore and Wetzel, 2000; Hasegawa et al., 2007). Image analysis has the potential to unlock a wealth of knowledge regarding the seagrass-epiphyte dynamic relationship if it can be applied to detect changing epiphyte community composition with environmental conditions.

For all three sampling locations, the slopes of linear regressions of epiphyte biomass vs. epiphyte pixels changed in a consistent seasonal pattern where the slope increased with progression from autumn through winter. The observation of increased epiphyte biomass per unit of epiphyte-covered area is consistent with continued epiphyte growth despite observed decreases in seagrass leaf growth and production during winter and early spring ( Figures 5 – 7 ). Cooler water temperatures (and perhaps shorter days/lower light levels) may be critical environmental factors diminishing leaf growth (Zieman, 1975; Marbà et al., 1994; Zieman et al., 1999; Koch and Erskine, 2001; Kim et al., 2020). However, these observations could also be partly explained by shifts in the epiphyte community composition that increase biofilm “thickness” and/or density of the epiphyte biofilm per unit area of leaf coverage as previously observed (Corlett and Jones, 2007; Giovannetti et al., 2010). The epiphyte biofilms in this study varied from 14 – 70 ng biomass per pixel. The highest epiphyte biomass per epiphyte area (pixels) was observed at the CI site ( Table S2 ), which had significantly highest DIN/Phosphate ratio as well as unique influences from proximity to oysters and hydrodynamic disturbance from boat traffic in the ICWW.

Changes in the relationship between epiphyte load and epiphyte coverage are attributed to two types of epiphyte-seagrass dynamics under different seasonal conditions ( Figures 5 – 7 ). In the first accumulation scenario (typical of most samples from August – December), the strong linear relationship between epiphyte load and epiphyte coverage implied that primary epiphyte colonization on the available seagrass leaf surface was dominant due to the growth of new leaf surface ( Table 3 ).

In the second epiphyte accumulation phase, indicated by fitting exponential regressions of epiphyte load with epiphyte coverage for several specific sites and times, the model clearly demonstrates a strong tendency for accumulation of epiphyte biomass more rapidly than the expansion of epiphyte coverage. Epiphyte coverage of leaves saturates if leaf growth slows relative to the primary colonization rate. Existing epiphytes continue to grow in size, thickness and/or density (e.g., calcareous) in an orderly succession with three or more layers (Willcocks, 1982; Kitting et al., 1984; Corlett and Jones, 2007; Saha et al., 2019). On older leaves where growth has slowed, stopped, or even reversed (senescence), epiphyte biomass accumulates via further secondary colonization and growth contributed by filamentous algae and invertebrates (Novak, 1982; Armitage et al., 2006; Whalen et al., 2013), while the underlying primary coralline algae eventually die (Borum, 1987). It is not uncommon for epiphyte biomass to exceed seagrass biomass ( Figures 5 – 7 ; Kitting et al., 1984; Moncreiff et al., 1992). A community shift to a greater abundance of filamentous algae and/or heavier calcareous epiphytes, such as serpulid worms, would potentially increase both biomass and biological diversity.

While previous reports suggested a higher biomass accumulation rate of epiphytes on seagrass leaves in the summer (Heijs, 1984; Hasegawa et al., 2007), our observations noted a reverse seasonal pattern of maximum epiphyte accumulation, relative to seagrass growth, in the winter and spring seasons, coinciding with most observations of exponential correlations. The exceptions of the August exponential correlations at the “Control” and CI sites suggests that there are multiple environmental scenarios that can produce a similar physiological outcome. Shading is not the only impact of epiphyte biofilms. Availability of O₂, CO₂, or other gasses might be linked to localized necrosis or the initiation of leaf senescence to avoid excessive consumption of stored plant reserves. The recent works of Brodersen (Brodersen et al., 2015; Brodersen et al., 2020) have expanded on old ideas (Sand-Jensen, 1977) about epiphyte biofilms altering gas exchange and leaf physiology. Such measurements, alongside transcriptomic profiling (Crump et al., 2018) are likely to reveal underlying physiological regulatory mechanisms unique to seagrasses.

For the shift from the first to second accumulation scenarios of epiphytes, it is unclear whether the leaf stops growing due to environmental factors, allowing epiphytes to accumulate exponentially, or if the epiphytes outgrow the seagrass and reduce light to the point of reducing seagrass leaf growth. The two patterns may also occur simultaneously on different parts of the same long leaf, which could be further investigated with the imaging approach. However, even with the heaviest epiphyte accumulations, coverage only rarely reaches 100% (on senescing leaves), and the basal 5 cm of Thalassia testudinum leaves (near the leaf sheath) is usually not visibly covered by epiphytes. This two-phase epiphyte accumulation may imply an existing threshold or trigger point for the balance between leaf growth response to epiphyte shading and leaf senescence. Accordingly, environmental effects on seagrasses and epiphytes are considered separately below.

The seasonal growth pattern of T. testudinum at all locations, maximum in the summer and early autumn followed by a decline through winter or early spring ( Figures 8 , 9 ), is widely observed (Herzka and Dunton, 1997; Fourqurean et al., 2001) and may be initiated by low temperature and/or salinity (Fernández-Torquemada and Sánchez-Lizaso, 2005), with the former most likely in this study ( Tables 1 , 3 ; Figures S2 , S3 ). Slow recovery of leaf area in April is consistent with photosynthesis rates at sub-optimal water temperature (23° C) (Herzka and Dunton, 1997; Campbell et al., 2006; Collier and Waycott, 2014; Rasmusson et al., 2020). Decreasing temperatures and salinity may have initiated the significant decline of leaf biomass and leaf area in winter and early spring ( Tables 1 , 3 ; Figures S2 and S3 ). Seagrass biomass was only weakly correlated with depth and salinity, but seagrass leaf area did show correlation with both temperature and salinity. The variability in thickness of seagrass blades might cause the inconsistence between the leaf biomass and areas.

Sediment porewater nutrients were measured during May under the assumption that the porewater nutrient differences would reflect the long-term exposure history to water column nutrients. Seagrass biomass and leaf area were significantly higher at two of three WWTP sites during warmer water summer and autumn periods ( Figures 8 , 9 ), which correlated with the significantly greatest porewater phosphate levels at WWTP. There were no corresponding significant increases in DIN, or DIN/phosphate ratio at the WWTP site ( Table 2 ). Likewise N content and C:N ratios did not differ significantly in seagrasses from any of the study sites ( Figure 4 ), suggesting that all three sampling locations are replete with N (Duarte et al., 2018).

Observed differences in seagrass stable isotope ratios showed ¹³C significantly more depleted at CI, and ¹⁵N significantly less enriched at “Control”. The most negative δ¹³C values of seagrass leaves at the CI site ( Figure 4 ), could be explained by the seagrass leaves having low light and CO₂ availability due to coverage by the greatest accumulation of epiphytes (Grice et al., 1996; Lapointe et al., 2020; Premarathne et al., 2021). Carbon-limited photosynthesis would increase discrimination against ¹³C compared to “Control” and WWTP seagrasses. It suggests that the influence of light and/or CO₂ availability caused by high epiphyte accumulations might be an important factor controlling seagrass productivity in this study. However, conditions that seem inconsistent with differential light and/or CO₂ limitation are CI having the shallowest depth, as well as hydrodynamic turbulence from nearby boat traffic. The lower δ¹⁵N value of T. testudinum in the “Control” area compared to the WWTP area (and CI) might reflect the influence of the wetland used to polish the treated wastewater there. The WWTP site receives treated wastewater unpolished by wetlands, but also potential influence from a fish cleaning station and associated bird roosting.

Unlike the seasonal pattern of seagrass growth, the epiphyte load, relative to the seagrass host, presented an opposite seasonal pattern. The epiphyte load in the “WWTP” area and the CI site followed a similar seasonal pattern, with the maximum in winter and low relative epiphyte accumulation in summer ( Figure 10 ). The increased epiphyte load from summer to winter is consistent with previous reports of epiphyte accumulation on Cymodocea nodosa and Zostera marina (Nelson and Waaland, 1997; Reyes and Sansón, 2001). Seasonal seagrass growth condition has been suggested as the primary factor controlling the temporal epiphyte accumulation pattern (Momota and Nakaoka, 2018; Bračun et al., 2021). A faster leaf turnover rate may drive a lower epiphyte load in the summer than in the winter (Duarte and Sand-Jensen, 1990; Peterson et al., 2007), which causes a shorter period for epiphyte accumulation on the available leaf surface. However, our observation of exponential epiphyte accumulation in August suggests that the lifetime of the leaf is still sufficiently long enough to be overtaken by excessive epiphyte growth. Here, “excessive growth” might be somewhat arbitrarily defined as an epiphyte/seagrass biomass ratio > 1, but this value approximates the point of deviation from a linear accumulation response (See Figures 5 – 7 ).

Epiphyte colonization may be secondarily controlled by environmental conditions such as the DIN/phosphate ratio in this study (Biber et al., 2004). Despite the acute decline of seagrass leaf area in the winter and spring, epiphytes appear to secondarily colonize on top of the basal layers of diatoms and coralline red algae to initiate an exponential accumulation phase (Armitage et al., 2006; Corlett and Jones, 2007; Whalen et al., 2013). Community succession patterns and compositional changes would impact the per-pixel biomass values. Notably, the average % epiphyte coverage metric ( Figure 11 ) showed relatively little seasonal fluctuation compared to the epiphyte load metric ( Figure 10 ). It would be informative to distinguish and quantify changes in epiphyte community composition by combining image analysis, pigment analysis and molecular taxonomy.

Site to site differences in biotic and abiotic factors, especially temperature, nutrients, and grazer densities, also play essential roles in epiphyte accumulation. (Thom et al., 1995; Nelson and Waaland, 1997; Wear et al., 1999; Frankovich and Zieman, 2005; Whalen et al., 2013; Ruesink, 2016). Some epiphytic algae on T. testudinum tolerate low temperature but decline rapidly over 30 (Biber et al., 2004). Summer water temperatures possibly drive seasonal epiphyte community composition changes (Stanca and Parsons, 2021). Our observation that epiphyte coverage ordinated higher at CI site across all seasons indicates that environmental factors playing a major role in these site-to-site differences are consistent.

Oyster impacts on porewater nutrients (Booth and Heck, 2009; Wagner et al., 2015), hydrodynamics (Booth and Heck, 2009; Smith et al., 2009) and epiphyte community composition and relative abundances (Frankovich and Fourqurean, 1997; Smith et al., 2018), may provide the consistent forcing observed at CI. The significantly highest DIN/phosphate ratio of the sediment porewater suggests long-term greater nutrient exposure, and nutrient effects on epiphytes have been explored extensively (Frankovich and Zieman, 2005; Johnson et al., 2006; Frankovich et al., 2009; Nelson, 2017). But temporal shifts in top-down grazer control have been noted (Whalen et al., 2013), and 18S sequence abundance actually suggests a greater epiphytic animal abundance at CI (unpublished observations), Thus relatively less effective grazer control of epiphyte accumulation at CI seems incongruous as the major explanation for this site’s consistently greater epiphyte accumulation. On the other hand, hydrodynamic wave effects from proximity to the ICWW (Schanz et al., 2002) could disrupt grazing pressure across seasons.

Although the nitrogen sources for seagrass differ among the three locations, the similarity of δ¹⁵N of epiphytes did not reveal any such differences. Porewater nutrient differences may be more indicative of past nutrient history (e.g., untreated wastewater spill) as opposed to recent conditions affecting epiphytes. Thus increased epiphyte load and coverage at CI seems likely due to the influence of nutrient levels and hydrodynamics, on the epiphyte community composition and abundance (Armitage et al., 2005). The greater degree of carbon isotope discrimination by epiphytes (-12.46 < δ¹³C < -10.82) at the WWTP area than at other locations (-7.35 < δ¹³C < -7.82) might derive from different epiphytic communities (algae and animals) among the three locations (Smit et al., 2005), especially the presence of invertebrates. The seagrass host can also produce secondary compounds that influence microbial, algal and animal epibiota (Harrison, 1982; Harrison and Durance, 1985; Crump et al., 2018).

While nutrient enrichment at the “WWTP” area was expected to promote seagrass and epiphyte growth (Borum, 1985; Lee et al., 2007; Baggett et al., 2010), such effects were not observed on any consistent basis in comparison to nearby sites receiving only polished wastewater or under influence from oyster habitat. Future studies would benefit from monitoring grazers, epiphyte community change, and seasonal water column nutrient levels to improve understanding of the seagrass-epiphyte relationship.

This study demonstrated the utility of the image analysis approach to understand the seagrass-epiphyte dynamic relationship and the impact of environmental stressors on this relationship. The emerging interpretation from these results is that epiphyte community change from environmental factors impacts both biomass and imaging measures in sometimes different ways that provide complementary insights. Specifically, the relationship between epiphyte biomass accumulation and epiphyte coverage of the leaf provides insight into the morphology of the epiphyte biofilm in ways that impact the seagrass host such as light attenuation and gas exchange (Sand-Jensen, 1977; Brodersen et al., 2015). It is worth examining more closely to understand the significance of the epiphyte composition and colonization pattern in conjunction with seagrass physiology. However, evaluation of the image analysis method shows that classification discrepancies still exist (Huang, 2020), and a major challenge is accurate classification of diverse epiphyte constituents through refinement of epiphyte and seagrass reference spectra. An advanced image-analysis tool should be able to define the spatiotemporal changes in epiphyte communities and improve efficiency through deep learning with neural networks (Mehrubeoglu et al., 2021).