Introduction

Human population density along coastlines is roughly three times the overall global average (Small and Nicholls 2003), and as human population growth continues, urbanization along coasts is accelerating (Seto et al. 2011; Bloom 2011). The intense concentration of humans in coastal areas disrupts adjacent ecosystems through three main pathways: resource exploitation, pollution, and ocean sprawl (Todd et al. 2019). Resource exploitation includes extraction of living (e.g., fishing, aquaculture) and non-living (e.g., dredging, mining, oil and gas extraction, water use for cooling or desalination) resources (Todd et al. 2019). Pollution near urban centers includes sediments, nutrients, plastic debris, chemicals, and pathogens, as well as increased light and noise (Todd et al. 2019; Carlson et al. 2019). Ocean sprawl includes the development of reclaimed land and artificial islands, artificial coastal defenses, ports, docks, marinas, oil infrastructure, and submarine cables and pipelines (Todd et al. 2019). Each pathway has its own impacts on coastal ecosystems, but they can also act together synergistically, leading to profoundly different ecological conditions near urban centers compared to pre-urbanization conditions (e.g., habitat and keystone species loss).

It is largely recognized that coastal urbanization negatively impacts coastal ecosystem health. However, many coastal cities have been human population centers since before ecosystem monitoring was initiated (Röthig et al. 2023), making it difficult to fully isolate and characterize the impact of urbanization on coastal ecosystems due to missing or shifting baselines (Roberts et al. 2017). Additional methodological issues, including the difficulty in identifying suitable whole-habitat replicates and lacking well-defined indices of urbanization across regions, further complicate measuring urbanization effects on ecosystems (Lee et al. 2006). The recent urbanization of many tropical and subtropical islands compared to continental coastlines allows for more direct measurement of the ecosystem effects of urbanization since more modern monitoring technologies were developed before or during urbanization and because island chains provide naturally replicated habitats in close geographic vicinity. For example, the relatively recent urbanization of the Caribbean Island of Roatan allowed for satellite data spanning the entire urbanization period to be analyzed to directly correlate urbanization with mangrove forest removal (Tuholske et al. 2017). In another study, biodiversity on urbanized atolls in the South Pacific was directly compared to nearby uninhabited atolls to demonstrate that urbanization significantly reduced biodiversity (Steibl et al. 2021).

While tropical and subtropical islands present unique opportunities to study urbanization effects on coastal ecosystems, they are also uniquely vulnerable to urbanization (Russell and Kueffer 2019). Oceanic islands are biodiversity hotspots but are simultaneously experiencing some of the highest rates of species loss due to anthropogenic pressure (Kier et al. 2009). Smaller overall area on islands means that more urban development is closer to the coast (Cao et al. 2017) and there is a higher likelihood of land reclamation projects being undertaken (Heery et al. 2018; Masucci and Reimer 2019). Islands also attract large volumes of tourists that can overwhelm local facilities and lead to more pollution and habitat degradation (DeGeorges et al. 2010; Baum et al. 2015). However, since urbanization is ongoing in many island systems, it may still be possible to minimize impacts on ecosystem functioning through research-informed sustainable development (Ghina 2003).

Like many tropical and subtropical islands, Okinawa Island (Japan) has undergone urbanization and population growth relatively recently (Kobayashi 2022; Uehara et al. 2019), with rapid urbanization commencing after 1972 when administrative control of the island was returned to Japan from the U.S.A. (Li et al. 2018). Sprawling high-density population centers now fill the southern third of the island, while the northern third of the island remains rural (i.e., composed of forests, agricultural operations, and small villages). Alongside urbanization, Okinawa’s nearshore marine ecosystems have been subjected to resource exploitation (Kobayashi 2022; Uehara et al. 2019), pollution (Ares et al. 2020; Li et al. 2018), and ocean sprawl (Masucci and Reimer 2019)—the three main pathways through which urbanization impacts coastal ecosystems (Todd et al. 2019). Okinawa has a strong land-sea connection due to frequent and intense rains associated with spring monsoons and summer/fall typhoons (Singh et al. 2022). Consequently, ongoing environmental challenges related to urbanization—particularly stormwater runoff and associated pollutants, such as wastewater and sediment pollution (red soil)—may be especially relevant in Okinawa (Ares et al. 2020). In contrast, the oceanic setting of Okinawa may dampen urbanization effects on nearshore ecosystems due to strong coastal currents and dilution effects (Ares et al. 2020; Rintoul et al. 2022).

In this study, we leveraged the natural laboratory set up by Okinawa’s urbanization pattern to investigate the impact of intense urbanization on the nearshore ecosystem of a subtropical oceanic island. We assessed microbial community composition (i.e., 16S rRNA gene metabarcode analysis) and physicochemical environmental parameters every 2 weeks for 1 year at four sites with varying levels of urbanization based on land cover classification from satellite imagery. We focused on microbial community composition since microbes respond rapidly to changing physical and chemical conditions, leading to shifts in microbial community dynamics being detectable before responses in economically important macro-organisms—such as corals, mollusks, fish, or aquacultured seaweeds (McLellan et al. 2015). Moreover, specific microbes are indicators of pollution (e.g., fecal indicator bacteria) and some microbes are considered pollutants themselves (e.g., the coral pathogen Serratia marcescens; Sutherland et al. 2011). Thus, shifting microbial communities and the presence of specific microbes can indicate or predict degradation of nearshore marine ecosystems (Becker et al. 2023). By analyzing the microbial community at urban and rural sites across a relatively high-resolution time series, this study will disentangle seasonal variability from anthropogenic influences, allowing for better understanding of how urbanization changes coastal microbial community dynamics. Persistent changes in microbial community dynamics at urban sites would suggest that increased coastal management is needed in Okinawa, as well as in other urbanizing island systems.

Methods

Sampling Area Description

Nearshore seawater samples were collected along the west coast of Okinawa Island in the Ryukyu archipelago, South Japan. The island shows well-defined dominant land uses, with concentrated urbanized watersheds primarily situated in the southern third of the island (Fig. 1). As an exception, Nago City, a major city on the island, lies in the northern part of the island. Four primary sampling sites were selected based on watershed size and the percent of the watershed with land cover classified as urban. Watersheds were delineated and their land areas were calculated using a digital elevation model (DEM) interpolated at 30 m based on a 2008 10-m LiDAR survey provided by the Geospatial Information Authority of Japan (https://fgd.gsi.go.jp/download/menu.php). Land cover classification was based on a Jan 4, 2015, Landsat 8 Operational Land Imager image obtained from the US Geological Survey (Scene: LC81130422015004LGN00) following methods described in Ross et al. (2018). Land cover classes were defined as agriculture, dominated by sugarcane and other crops at various stages; forest, dense tree stands with closed canopy; scrub, short woody vegetation without tree canopy; grass, short trimmed vegetation found in golf courses and airfields; rock/dirt; sand; urban, a complex mix of man-made surface materials such as concrete and asphalt with limited vegetation; water, including fresh and marine bodies of water; and unclassified, pixels that could not be classified (Fig. 1).

Fig. 1
figure 1

Geographic setting of sampling sites and subsites on the west coast of Okinawa Island, Japan. Left panel: Map of Okinawa Island with land cover classifications based on satellite data. Thin black lines indicate watershed boundaries based on a LiDAR digital elevation model. Primary sampling sites are highlighted by color coded boxes (purple, urban sites; blue, rural sites) with points demarcating the subsites where samples were collected. Right panels: Detailed maps of the four primary sampling sites with the subsites demarcated as points. Point color indicates the subsites—blue, south (S); red, central (C); yellow, north (N)—where water samples and measurements were taken. Satellite true-color imagery is from Google Earth

The four primary sampling sites (Urban 1 (U1) - Ginowan, Urban 2 (U2) - Nago, Rural 1 (R1) - Tancha, and Rural 2 (R2) - Ogimi) are located in watersheds of similar size (407,002–993,281 m2) and are classified as urban or rural based on the percent of the watershed with land cover classified as urban. Specifically, urban primary sites U1 and U2 are in watersheds with > 70% of land cover classified as urban and rural primary sites R1 and R2 are in watersheds with > 70% land cover classified as agriculture, forest, scrub, grass, rock/dirt, and sand. In addition, both urban sites have artificial shorelines created from land reclamation projects that filled in the coral lagoon (Masucci and Reimer 2019). Primary sampling sites were centered on the freshwater outflow point into the ocean from the study watershed. Three subsites at each primary site were designated for sample collection based on their proximity to the point source of freshwater outflow: central (C) subsites are perpendicularly offshore to the outflow and North (N) and South (S) subsites were ~ 200 m North and South of the outflows, respectively. Subsites were sampled synoptically every other week for 1 year, from September 17, 2020, to September 2, 2021, leading to 25 sampling events in total.

Seawater Sampling

Nearshore surface seawater was collected for DNA metabarcoding by submerging acid-cleaned 500-mL Nalgene bottles below the sea surface in the uppermost 20–50 cm of the water column. Seawater for dissolved macronutrient analysis was collected in acid-cleaned 50-mL Falcon tubes. A total of 300 of each sample type was collected: four primary sampling sites, three subsites per primary site, and 25 collection events. Samples were transported to the lab on ice and in the dark. After transport, seawater samples for metabarcoding were immediately filtered through 0.2-µm pore-size Polytetrafluoroethylene filters (Millipore) under gentle vacuum pressure, and filters were stored at − 80 °C for later DNA extraction. Samples for dissolved macronutrient analysis were syringe-filtered and stored at − 20 °C until chemical analysis. Physicochemical properties—dissolved oxygen (DO), salinity, sea surface temperature (SST), turbidity, and chlorophyll a fluorescence (Chla a)—were measured at each subsite with a RINKO conductivity, temperature, and depth (CTD) probe (JFE Advantech, Japan).

Nutrient Analyses

Nutrient concentrations—including nitrate (NO3), nitrite (NO2), ammonium (NH4+), phosphate (PO43−), and silica (SiO2)—were determined on a QuAAtro39 Continuous Segmented Flow Analyzer (SEAL Analytical) following manufacturer guidelines. Final concentrations were calculated through AACE software (SEAL Analytical). Nutrient analysis was carried out at the Okinawa Prefecture Fisheries and Ocean Technology Center.

DNA Extraction and Metabarcode Sequencing

DNA was extracted from frozen filters following the manufacturer’s protocol for the DNeasy Power Water Kit (Qiagen), including the optional heating step. Metabarcode sequencing libraries were prepared for the V3–V4 region of the bacterial 16S ribosomal RNA gene following Illumina’s 16S Metagenomic Sequencing Library Preparation manual without any modifications. The V3–V4 region of the 16S rRNA gene provides good taxonomic resolution for environmental bacterial communities and is widely sequenced for such studies, increasing the intercomparability of study results (Klindworth et al. 2013). Sequencing was performed by the Okinawa Institute of Science and Technology Sequencing Center using 2 × 300-bp v3 chemistry on the Illumina MiSeq platform. Out of 300 samples, 291 produced sequencing results that passed all quality filters. Overall, 53 million sequencing reads were generated, with 62,656–612,998 sequencing reads per sample (mean = 182,814). Sequencing data are available from the NCBI Sequencing Read Archive (SRA) under the accession PRJNA1044524.

Bioinformatic and Statistical Analyses

Sequencing reads were denoised using the Divisive Amplicon Denoising Algorithm (Callahan et al. 2016) with the DADA2 plug-in for QIIME 2 (Bolyen et al. 2019). Taxonomy was assigned to representative amplicon sequence variants (ASVs) using a naive Bayes classifier trained on the SILVA 99% consensus taxonomy (version 132; Quast et al. 2013) with the QIIME 2 feature-classifier plug-in (Bokulich et al. 2018). The results were imported into the R statistical environment (R Core Team 2018) for further analysis with the phyloseq (Mcmurdie and Holmes 2013) and vegan (Oksanen et al. 2019) R packages. Rarefaction sampling was performed and plotted with the ggrare function, and all samples reached richness saturation within their total sample size (Fig. S1). Alpha diversity estimates (richness and Shannon index) were determined with the breakaway R package (Willis et al. 2017), and the statistical significance of differences in mean alpha diversity between urban and rural sites each month was tested with pairwise Wilcox tests. To minimize compositional bias inherent in metabarcoding data, we used the Aitchison distance between samples, which includes a centered log-ratio transformation to normalize data (Gloor et al. 2017), for principal coordinates analyses (PCoA). Permutational multivariate analyses of variance (PERMANOVA) on Aitchison distances were performed with the adonis2 function (999 permutations) in the vegan R package to test whether shifts in community composition between sample types were statistically significant (Oksanen et al. 2019). The influence of environmental parameters in shaping bacterial community composition was investigated through redundancy analysis (RDA) and variance partitioning using functions from the vegan R package. Lastly, the SPIEC-EASI R package was used for network analysis and to infer co-occurrence patterns between taxa (Kurtz et al. 2015). Intermediate data files and the code necessary to replicate analyses are available in a GitHub repository (https://github.com/maggimars/UrbanOki) where an interactive HTML document (https://maggimars.github.io/UrbanOki/Amplicons.html) can also be found.

Results

Land Cover Effects on Physicochemical Parameters in Nearshore Ecosystems

The annual pattern in SST was nearly identical across sites and subsites (Fig. 2). In contrast, salinity was lower and more variable at the urban sites compared to rural sites, especially at the central and north subsites for both urban sampling sites (Fig. 2). Turbidity was variable at all locations and has multiple causes–including phytoplankton growth, sediment resuspension, and soil pollution (Fig. 2). The percent saturation of dissolved oxygen (DO) was highest at the northern rural site (R2 - Ogimi; all subsites) compared to the other three study sites (Fig. 2), potentially caused by more wave action in that region. Nitrate + nitrite, ammonium, and phosphate concentrations were low at rural sites, whereas these macronutrients were elevated at urban sites—particularly at the central and north subsites of the U1 - Ginowan sampling area (Fig. 3). Silica concentrations were variable at all sites (Fig. 3). Decreases in salinity were strongly correlated with high nitrate + nitrite concentrations at urban sites (R = − 0.65, p = 0) and were also correlated with increased phosphate (R = − 0.454, p = 0) and Chl a (R = − 0.29, p = 0; Fig. 4). Nitrate + nitrite concentrations were also strongly correlated with phosphate concentrations at urban sites (R = 0.76, p = 0) and chlorophyll a was positively correlated with nitrate + nitrite (R = 0.34, p = 0), ammonium (R = 0.27, p = 0.001), and phosphate (R = 0.26, p = 0.001) at urban sites but not rural sites (Fig. 4). Overall, chlorophyll a fluorescence was higher at urban sites (U1 = 0.47 ± 0.03 RFU, U2 = 0.43 ± 0.05 RFU) compared to rural sites (R1 = 0.34 ± 0.03 RFU, R2 = 0.18 ± 0.01 RFU), particularly at central and north subsites (Fig. S2). At rural sites, salinity was significantly negatively correlated with silica concentrations (R = − 0.36, p = 0; Fig. 3).

Fig. 2
figure 2

Time series of physical parameters measured at nearshore urban and rural sites along the west coast of Okinawa Island, Japan. Sea surface temperature (SST), salinity, turbidity, and dissolved oxygen (DO) were measured at each subsite (south, central, and north) within each primary sampling site (U1, U2, R1, and R2) using a RINKO CTD probe. Plots are faceted by primary sampling site (vertically) and parameter (horizontally). Point and line color represent subsites—blue, south (S); red, central (C); yellow, north (N)—where water samples and measurements were taken. The annual pattern in SST was nearly identical across sites and subsites. Salinity was lower and more variable at the urban sites, especially at central and north subsites. Turbidity was variable at all locations. The percent saturation of DO was highest at the northern rural site (R2 - Ogimi)

Fig. 3
figure 3

Time series of chemical parameters measured at nearshore urban and rural areas along the west coast of Okinawa Island, Japan. Nitrate + nitrite (NO2 + NO3), ammonium (NH4), phosphate (PO4), and silica (SiO2) were measured in water samples collected from subsites (south, central, and north) within each primary sampling site (U1, U2, R1, and R2) with a QuAAtro39 Continuous Segmented Flow Analyzer. Plots are faceted by primary sampling site (vertically) and parameter (horizontally). Point and line color represents subsites—blue, south (S); red, central (C); yellow, north (N)—where water samples and measurements were taken. Nitrate + nitrite, ammonium, and phosphate concentrations were rarely above the detection limit at rural sites, while these macronutrients were elevated at urban sites—particularly the central and north subsites of the U1 - Ginowan primary sampling site. Silica was variable at all sites

Fig. 4
figure 4

Correlation between environmental parameters at urban and rural sites on the west coast of Okinawa Island, Japan. Values for all parameters were z-scaled and the Pearson correlation coefficients were calculated for all parameters at the two urban primary sites (U1 - Ginowan, U2 - Nago) and the two rural primary sites (R1 - Tancha, R2 - Ogimi). Correlation coefficients are visualized for all comparisons by both point size and color: darker blue colors are more positively correlated, darker red colors are more negatively correlated, and point size reflects the absolute value of the correlation coefficient. The statistical significance of each correlation coefficient was evaluated with the cor.mtest function in the corrplot R package. Correlation coefficients were considered statistically significant if the p-value was ≤ 0.01 and significant correlations are marked with an asterisk (*) on the plot

Effects of Land Cover on Nearshore Bacterial Communities

We used metabarcode analysis with the 16S rRNA gene to investigate the effect of land cover (urban or rural) on nearshore bacterial community composition across a biweekly time series spanning a full year. Alpha diversity metrics (ASV richness and Shannon index) were lower in warmer months (May–August) compared to cooler months at both urban and rural sites (Fig. S3). However, the mean richness and Shannon indices were higher at urban sites compared to rural sites across all months (Fig. S3), with the difference in richness statistically significant in six out of twelve months (Jan, March, June, July, Oct, and Dec; Table S1) and the difference in Shannon index statistically significant in eight out of twelve months (Jan, March June, July, Aug, Sept, Oct, Dec; Table S1). The bacterial community compositions in nearshore waters of urban sites were significantly different from the communities in the nearshore waters of rural sites (PERMANOVA, 999 permutations, F = 8.5, p = 0.001; Fig. S4). PCoA plots for nearshore bacterial communities in urban and rural regions showed contrasting patterns (Fig. 5). The warmer months (May through the beginning of August) formed tight clusters in the PCoA plots for both rural (Fig. 5A) and urban sites (Fig. 5B). In the PCoA plot for rural sites, samples from seasons cluster so that they are separated roughly by quadrant, with winter samples in the top right, spring and early summer samples in the top left, samples from late summer in the bottom left, and autumn samples in the bottom right. In the PCoA plot for urban sites, samples separate based on which urban site they originated from, with more samples from U1 - Ginowan in the bottom half of the plot and more samples from U2 - Nago in the top half of the plot. These differences in clustering patterns suggest that the seasonal succession cycle is disrupted at urban sites compared to rural sites. However, results from PERMANOVA analyses on the season and primary site variables were significant for both rural and urban sites (Table 1). The subsite variable (i.e., the proximity to a freshwater outlet) was only significant for urban sites (F = 3.63, p = 0.001; Table 1).

Fig. 5
figure 5

Principal coordinates analysis (PCoA) plots of Aitchison distances between bacterial communities through time at two nearshore sites adjacent to rural areas (A) and two nearshore sites adjacent to urban areas (B). Shape indicates sampling sites, with circles representing the more southern site (site 1) of both the rural and urban sites, and triangles representing the more northern sites (site 2). The color indicates the month of the year, with blues representing winter months, purples representing spring months, greens representing summer, and yellow–red representing autumn. Samples from late spring and early summer form tight clusters in both plots. Samples from rural sites (A) separate into quadrants based on the season (winter; top right, autumn; bottom right, late summer; bottom left, spring and early summer; top left), whereas samples from the two urban sites (B) separate by site location on the secondary (y) axis and the seasonal cycle is not visible

Table 1 Permutational multivariate analyses of variance (PERMANOVA) results for tests run on rural and urban nearshore bacterial communities with month, primary site, and subsite variables

We performed a redundancy analysis (RDA) to visualize which environmental variables contributed to the clustering observed in PCoA ordination plots (Fig. S5). An analysis of variance (ANOVA) was run on the RDA to determine if model results were significant and an ANOVA by term was used to test which variables’ contributions were statistically significant (Table S2). Variance partitioning was calculated for all significant variables. SST significantly affected community composition clustering among samples from both urban (F = 3.6, p = 0.004, variance partition = 0.7%) and rural sites (F = 4.2, p = 0.012, variance partition = 2.2%). Dissolved oxygen (F = 13.4, p = 0.001, variance partition = 7.4%), phosphate (F = 6.9, p = 0.004, variance partition = 3.1%), and salinity (F = 2.83, p = 0.045, variance partition = 0.6%) were also significant determinants of community composition in rural samples. In contrast, nitrate + nitrite (F = 5.4, p = 0.001, variance partition = 1.5%) was the only significant determinant of community composition in urban samples after SST (Fig. S5, Table S2).

Plotting the relative abundance of ASVs grouped by bacterial order showed clear differentiation between samples collected at urban and rural sites (Figs. 6 and S6). Eight bacterial orders were present in three or more urban samples but absent in all rural samples (Fig. 6; highlighted in red) and nine orders were prevalent among urban samples but rare in rural samples (Fig. 6; highlighted in orange). The cumulative relative abundance of these orders ranged from 1.3 to 63.7% (mean = 19.6%) at subsites in Ginowan (U1) and 1.1–57.9% (mean = 15.4%) at subsites in Nago (U2), whereas the cumulative relative abundance of these orders never exceeded 9.8% at rural sites (Fig. 7). Members of these orders were more abundant at the central and north subsites of the two urban sampling sites than at the south subsites (Fig. 7), which are also the subsites that appear more impacted by runoff and freshwater input based on physicochemical parameters (Figs. 2 and 3). Interestingly, seven of these orders were shown to covary in urban samples through network analysis (Fig. 8a; Pseudomonadales, Clostridiales, Saccharimonadales, Campylobacterales, Bacteroidales, Betaproteobacteriales, Thiotrichales). The covariance of these orders suggests they may share a common source or that their growth is supported by shared resources associated with urbanization.

Fig. 6
figure 6

Relative abundance of major bacterial orders in each sample. Color represents the relative abundance of each order and was determined by summing the relative abundance of each ASV classified as belonging to the order. Samples are grouped by primary site and columns represent subsites (S; southern subsite, C; central subsite, and N; northern subsite). Order names are highlighted as follows: Orange; orders that are prevalent among samples from urban sites but rarely found in rural samples, Red; orders that are prevalent among or present in ≥ 3 samples from urban sites but not found in any rural samples, Blue; orders that are prevalent among or present in ≥ 3 samples from rural sites but not found in any urban samples

Fig. 7
figure 7

Relative abundance of bacterial orders identified as either prevalent among urban samples and rare in rural samples or prevalent among/present in ≥ 3 urban samples and absent in all rural samples. Samples are grouped by site and subsite (S; southern subsite, C; central subsite, and N; northern subsite) and stacked bar chart fill colors indicate bacterial order. All orders highlighted in red or orange in Fig. 6 are included. Orders that are rare at rural sites make up a large proportion of communities in urban samples, particularly at the central and northern subsites

Fig. 8
figure 8

Covariance networks for bacterial orders detected at urban and rural sites. Orange nodes represent orders that are prevalent among samples from urban sites but were rarely observed in rural samples. Red nodes represent orders that are prevalent among or present in ≥ 3 samples from urban sites but not found in any rural samples. Green edges connecting nodes indicate significantly positive covariance between node orders (no negative relationships could be detected). There is a cluster composed solely of “red” (Pseudomonadales, Clostridiales, Saccharimonadales) and “orange” (Campylobacterales, Bacteroidales, Betaproteobacteriales, Thiotrichales) orders in the urban site network, indicating that these orders co-occur across samples collected from the urban sites

Discussion

Coastlines around the world are rapidly being urbanized, with widespread consequences for adjacent marine ecosystems. Despite the apparent impacts of urbanization in many nearshore areas, it can be difficult to fully assess the effects due to shifting or absent baselines and methodological challenges. In this study, we investigated the impacts of urbanization on the physicochemical conditions and microbial communities in nearshore marine ecosystems. Studying microbial responses to urbanization is a good first step in assessing the impacts of urbanization because changes in microbial communities can indicate disruption to the ecosystem more broadly (Nogales et al. 2011). Okinawa Island has urbanized relatively recently compared to many other coastal urban centers and the pattern of urbanization—with high-density urban centers in the south and rural regions in the north—allows for ecological effects of urbanization to be investigated more directly. We detected an altered physicochemical setting in nearshore waters along urban coastlines compared to rural settings, with increased nutrient concentrations at urban sites. Microbial communities at urban sites were significantly different from communities at rural sites in all seasons. The differences largely reflected increased microbial diversity at urban sites due to the presence of bacterial taxa that were absent or rare at rural sites. Many of the additional taxa found at urban sites are associated with anthropogenic sources, such sewage treatment facilities or stormwater runoff. Our results demonstrate that urbanization is changing the surrounding marine ecosystem and that mitigation approaches that include terrestrial management—such as restoring natural habitats along coastlines or improving wastewater treatment—may be more effective than approaches solely focused on marine regulations.

Urbanization Alters the Physicochemical Characteristics of Nearshore Ecosystems

Urbanization had a dramatic effect on salinity and nutrient conditions in Okinawa’s nearshore waters. Urban sites exhibited highly variable salinity, with salinities nearing zero on several dates. Dips in salinity were most pronounced at the central and northern urban subsites, which were more likely to to be influenced by the freshwater outflow. Subsite, and thus proximity to freshwater outflows, had less effect on salinity at rural sites. While watershed size—which influences freshwater input volume—was similar across sites, differences in geomorphology at urban and rural sites may influence residence times for inflowing freshwater and suspended sediments. The rural sites are semi-enclosed coral lagoons, leading to longer residence times for inflowing freshwater and suspended sediments (Sakamaki et al. 2022). In contrast, urban sites underwent substantial land reclamation that completely filled in the coral lagoons (Masucci and Reimer 2019), which should decrease the residence times for terrigenous inputs (Sakamaki et al. 2022). While the residence times for the sites in this study are undetermined, the projected geomorphology-driven variability in residence time is incongruent with our observations. Consequently, the difference in salinity between urban and rural sites is most likely due instead to the variation in impervious land cover (concretization) between sites. Urban watersheds in this study have scarce natural or permeable ground cover, which increases flooding and runoff to coastal areas (Blum et al. 2020). Wide variations in salinity—like those seen at the urban sites in this study—can have severe biological implications. Decreased salinity from runoff reduces coral holobiont respiration and photosynthetic rates, heightens coral bleaching and mortality risk, disrupts symbiotic relationships between the coral, microbiome, and zooxanthellae, and reduces coral larval and gamete survival (Röthig et al. 2023). Moreover, salinity is a major driver for bacterioplankton community composition (Jurdzinski et al. 2023; Lozupone and Knight 2007), and reduced salinity in nearshore regions can promote the growth of pathogenic bacteria (Bordalo et al. 2002; Burge et al. 2014; Randa et al. 2004) and increase bacterial pathogenicity (Barca et al. 2023).

The major macronutrients—nitrogen and phosphorus—were enriched throughout the annual cycle at urban sites compared to rural sites. Both urban and rural land uses contribute to nutrient loading with nutrient pollution deriving from sewage, fertilizers, detergents, and pet waste in urban areas (Hobbie et al. 2017) and from agricultural fertilizers, manure application, and livestock waste in rural areas (Del Rossi et al. 2023). However, nutrient export is positively correlated with impervious land cover, making nutrient pollution more likely to reach the coast in urban areas (Duan et al. 2012). Nitrogen and phosphorus were consistently elevated at urban sites compared to rural sites, but were also highly variable. Inorganic nutrient concentrations are temporally variable due to rapid uptake by microorganisms and short-term hydrodynamic processes (Viana and Bode 2013) but sampling across a time series allowed the overall elevation at urban sites to be detected. Nutrient loading increases primary production (Howarth et al. 2021) and chlorophyll a was significantly positively correlated with nitrogen and phosphorus at urban sites. Excess nutrient-induced primary production increases heterotrophic bacterial growth and respiration that can lead to hypoxia when combined with stratification (Obenour et al. 2012), and dissolved oxygen was significantly negatively correlated with nitrate and nitrite at urban sites. Nutrient loading can harm coral reefs by fueling the growth of fleshy macroalgae that compete with corals for light and space and increase bioerosion (Silbiger et al. 2018), and by amplifying the negative effects of ocean acidification and rising temperatures (DeCarlo et al. 2015). Together the altered physicochemical conditions associated with urbanization have potential to degrade the coral reef ecosystems that are important to Okinawa for fisheries and tourism.

Urbanization Changes Nearshore Bacterioplankton Community Composition

There were significantly different bacterial communities in nearshore waters adjacent to urban areas compared to those adjacent to rural areas, which is reflected in elevated bacterial diversity at urban sites. In addition to a shared core bacterial community of ubiquitous marine bacteria, urban sites hosted additional taxa that were rare or absent at rural sites. Bacteria in the orders Cloacimonadales, Clostridiales, Lactobacillales, Pseudomonadales, Saccharimonadales, Selenomonadales, Sphingobacteriales, and Thiomicrospirales were present in at least three urban samples but absent in all rural samples. Bacteria in the orders Aeromonadales, Alteromonadales, Bacteroidales, Betaproteobacteriales, Campylobacterales, Chitinophagales, Sphingomonadales, Thiotrichales, and Vibrionales were prevalent among urban samples but were rare in rural samples. The majority of these orders contain bacteria known to be associated with anthropogenic sources, including some that are considered fecal indicator bacteria (FIB) and some that are potentially pathogenic to humans and marine life (Verburg et al. 2021). In contrast, Verrucomicrobiales was the only order detected in multiple rural samples but absent in all urban samples. Bacteria in the order Verrucomicrobiales are ubiquitous in soils (Bergmann et al. 2011), which may explain their absence in urbanized regions where there is little natural soil exposed.

Among bacterial groups present in urban regions but absent in rural areas, there are several of particular concern due to the inclusion of fecal indicator bacteria or known pathogens. Specifically, Clostridiales and Pseudomonadales were prevalent at both urban sites and contain known human pathogens. At the genus level, most of the Clostridiales were classified as Blautia spp., Clostridium sensu stricto 1, and several other genera in the Ruminococcaceae family. Blautia spp. and Ruminococcaceae bacteria are both common components of microbiomes in humans and other mammals, contributing upwards of 50% of mammalian intestinal microbiome communities (Liu et al. 2021). Clostridium sensu stricto 1 includes the “true” Clostridium species, such as Clostridium perfringens that is a reliable indicator of human fecal contamination due to its specificity to sewage and ability to form environmentally stable cysts (Stelma 2018). While Pseudomonadales includes several known pathogens in the genera Pseudomonas and Acinetobacter, the Pseudomonadales ASVs abundant at urban sites were Pseudomonas hussainii and Acinetobacter indicus, which are not known human or animal pathogens. However, A. indicus was originally isolated for its chitinase activity, which may allow it to act as an invertebrate pathogen in reef systems (Akram et al. 2022). Cloacimonadales, Lactobacillales, and Thiomocrospirales are often associated with wastewater treatment plants, sewage sludge, and wastewater effluent (Shakeri Yekta et al. 2019; Verburg et al. 2021; Meyer et al. 2016). High abundances of these microbes in urban ecosystems may be due to their presence in wastewater reaching the coastal ecosystem.

Bacteria in orders that were prevalent in urban samples, but rare in rural samples, also included groups of concern (i.e., Vibrionales, Campylobacterales, Betaproteobacteriales, and Bacteroidales). The increased prevalence of Vibrionales at urban sites is alarming due to a recent surge in Vibrio spp. infections leading to human fatalities and limb amputations in developed coastal areas (Archer et al. 2023). Severe vibriosis cases are most often caused by V. vulnificans, but multiple Vibrio species can cause serious illness in humans (Baker-Austin et al. 2018) and marine organisms (Grimes 2020). Vibrio fortis and Vibrio chagasii were abundant in Ginowan (Urban Site 1), Photobacterium leiognathi subsp. mandapamensis was the major Vibrionales species found in Nago (Urban Site 2), and Photobacterium damselae subsp. damselae was detected at both sites. V. fortis, V. chagasii, and P. damselae are marine pathogens—V. fortis causes coral bleaching (Sun et al. 2023) and enteritis in fishes (Wang et al. 2016), V. chagasii infects molluscs (Urtubia et al. 2023), and P. damselae causes necrotizing fasciitis in humans and fish (Rivas et al. 2013)while P. mandapamensis is not pathogenic (Urbanczyk et al. 2011). Urbanization-associated changes in physicochemical conditions could be responsible for increased abundance of Vibrionales at urban sites as several marine Vibrio species become more prevalent and infectious when salinity is reduced to 5–25 PSU (Sullivan and Neigel 2018) and salinity was often within this range at urban sites.

Campylobacterales, Betaproteobacteriales, and Bacteroidales were the most abundant groups prevalent among urban samples and rare in rural samples. Campylobacterales includes clinical pathogens (Man 2011) and environmental bacteria (Fera et al. 2004), and can be abundant in urban road-runoff samples (Liguori et al. 2021). Most Campylobacterales ASVs abundant at urban sites were classified as Arcobacter spp. Three of the five described Arcobacter species are associated with human disease and are found in human diarrhea and sewage samples, while the other two species are found in a variety of environmental settings—including rivers, salt marshes, and nearshore marine waters (Fera et al. 2004). Betaproteobacteriales is a large and diverse taxonomic group including many species found in wastewater treatment plants (Zhang et al. 2017). We detected 102 separate Betaproteobacteriales genera, but C39 (family: Rhodocyclaceae) was the most abundant and is associated with pollution and wastewater input in coastal ecosystems (Kopprio et al. 2020; Nascimento et al. 2018). Bacteroidales include dominant members of mammalian gut microbiomes (Magne et al. 2020) and sewage sludge communities (Shakeri Yekta et al. 2019), and have been found in road runoff (Liguori et al. 2021). The majority of Bacteroidales at urban sites belong to the genera Bacteroides and Prevotella, which are major mammalian gut bacteria and likely originate from sewage (Wu et al. 2011).

Overall, our results showcase a clear anthropogenic impact on microbial communities in nearshore environments adjacent to heavily urbanized watersheds in Okinawa. The urban ecosystems were consistently altered rather than exhibiting episodic or seasonal disturbances. Previous work focusing on the Tancha (R1) sampling site showed that extreme runoff events associated with typhoons caused short-lived perturbation of nearshore microbial communities, with the bacterial community composition returning to baseline only 48–72 h after storms passed (Ares et al. 2020). The quick return to pre-typhoon conditions highlights the quick-flushing of Okinawa’s coral lagoons, as well as the overall resiliency of nearshore ecosystems in more rural regions of Okinawa’s coast. In contrast, the consistently altered state observed at urban sites may indicate that these ecosystems are so altered that they have reached a tipping point and experienced a regime change. This regime change may interact with other components of the system (e.g., corals, algae, and other invertebrates) and create feedback loops that further degrade the ecosystem and its functioning (Becker et al. 2023; Qin et al. 2020). Alternatively, the consistently altered microbial communities observed at urban sites may reflect sustained, ongoing disturbances that would likely also have downstream ecological consequences (Ruprecht et al. 2021).

Conclusions and Future Directions

Biweekly observations made in nearshore waters adjacent to urban and rural watersheds revealed the profound influence of urbanization on salinity, macronutrient concentrations, and microbial communities in a recently urbanized subtropical island system. Expanding this work to include metatranscriptomics would allow for better characterization of the active metabolisms of microbes in urban and rural areas and could shed light on microbial nutrient uptake, as well as pathogenicity. Nonetheless, the observed changes in physicochemical parameters and microbial communities have likely contributed to degrading nearby coral reefs, as coral disease, death, and loss of diversity have been attributed to reduced salinity, nutrient loading, and introduced pathogenic bacteria in other systems. Coral reefs around Okinawa and other subtropical and tropical islands provide myriad ecosystem services, including supporting tourism and fisheries as well as protecting islands from waves and erosion, but the ability of reefs to perform these ecosystem services declines as reef health declines. To protect these key cultural, economic, and ecological resources, wholistic conservation measures that encompass both the reef ecosystem and the adjacent watershed are needed. Our results highlight the need for comprehensive conservation efforts that include land use management and coastal rehabilitation to ultimately protect important nearshore ecosystems, such as coral reefs. Moreover, more extensive wastewater treatment and moving outfalls further offshore can significantly reduce nutrient and pathogen loading on reefs. Such mitigation efforts may become more critical as cyclonic tropical storms increase in both frequency and intensity due to global climate change. Tropical cyclones deliver large amounts of precipitation that trigger extreme runoff events. Rural sites in Okinawa have proven to be resilient to extreme storms, but more study is needed to understand how urbanization influences ecosystem resiliency in the face of climate change.