Cities with older infrastructure that combine both sanitary, industrial, and storm sewers are at risk of overwhelming the capacity of these sewer systems during wet weather events, which trigger combined sewer overflows (CSOs) into nearby waterways (Burian et al., 2000). These CSOs can be triggered by a single precipitation event, and are related to significant economic, human health, and environmental problems (Lee and Bang, 2000; Marsalek and Rochfort, 2004; Gasperi et al., 2010). The frequency of CSO events is expected to increase substantially in coming years, as models predict the proportion of the world urban population to reach 66% by 2050, which is nearly double the urban population from 1950, putting great strain on the quality of water resources (Vorosmarty et al., 2000; Grimm et al., 2008). One common strategy for mitigating the frequency and velocity of CSO events is the construction of green infrastructure, which are pervious, vegetated areas in cities that capture precipitation and reduce the loading in combined sewer systems from runoff derived from impervious surfaces (Carson et al., 2013). Bioswales, which are permeable surfaces containing trees and other vegetation that are engineered to divert stormwater runoff from the street and into the soils, have become particularly prominent in urban development initiatives to reduce CSO events (Anderson et al., 2016; Woznicki et al., 2018). However, few studies have evaluated the ecology of these systems, so whether or not particular plant palettes are more or less effective for the multi-functionality of their benefits remains uncertain.

While plants are the prominent biological design feature of bioswales, microbes that reside within and around the plants are a relatively unexplored aspect of bioswales that have the potential to significantly increase green infrastructure functionality (McGuire et al., 2015; Fulthorpe et al., 2018). Plants and soil play critical roles in storm-water absorption and overall bioswale effectiveness (Xiao and McPherson, 2011), and plants rely on soil microbes to gain access to nutrients and water that otherwise would not be available (Auge, 2001; van der Heijden et al., 2008). As an extended component of a plant’s phenotype, microbes have the potential to facilitate plant resilience to frequent drying and wetting events that occur in bioswale soils, which would enhance plant survival and longevity in these systems. In addition, pollutants and excess nitrogen can be considerably high in bioswale soils due to the influx of stormwater (Shetty et al., 2018), and there is evidence that high abundances of microbial genes involved in contaminant degradation are expressed in bioswale soils (Gill et al., 2017). Previous studies have found that bioswale plants vary in their efficiency at pollutant removal (Read et al., 2008), and these effects are likely microbially mediated (Friesen et al., 2011). Plant-associated microbes may also influence plant physiological responses such as transpiration rates (Li et al., 2018), which have the potential to directly affect the quantity of water movement in bioswales.

In this study, we evaluated plant-associated microbial communities in bioswales in New York City, and measured relative rates of transpiration in the plants they were associated with. We chose five herbaceous target plant species that are common in NYC bioswale plant palettes and hypothesized that (1) each plant species would host unique soil microbial assemblages; (2) water conductance of the five-plant species both within and across bioswales would reveal species-specific patterns; and (3) variation in plant physiology would correlate with variation in soil microbial composition. These results present potential improvements for the design of bioswales for optimal hydrology and other desired functions such as pollutant removal.

Bioswales play an integral role in mitigating the incidences of storm water overflow events, and this study is among the first of its kind to detect relationships between the physiological traits of plants, the abiotic properties of soils, and the community composition of soil microbes that occupy these engineered ecosystems. In non-engineered ecosystems, microbial communities tend to be influenced first by climate, then by soils, and finally by biotic factors such as plant species (Fierer, 2017). We found strong separation of microbes across bioswales that was related to soil physical and chemical properties, and possible microclimatic effects that were not measured in the current study. At the local scale, microbial communities segregated according to which plant species they were collected near, likely in response to the specific aboveground and belowground traits of those plant species (Barberán et al., 2015; Leff et al., 2018; Zhalnina et al., 2018). The functional consequences of these specific plant-microbial assembly patterns were not assessed in our study, but may influence the capacity of bioswale soils to cycle nutrients and metabolize stormwater-derived pollutants such as hydrocarbons. In another study of bioswales in New York City, RNA sequencing revealed high expression levels of microbial genes involved in the decontamination and degradation of a suite of organic pollutants associated with stormwater (Gill et al., 2017). Together, the results from the current study and those of the Gill et al. (2017) study suggest that the location of the bioswale and the particular plant palette selected for planting, can ultimate influence the assembly and function of the soil microbial communities in these engineered ecosystems.

The significant interspecific variation among our five target plant species in two traits related to water movement and transpiration rates, indicates that species selection in bioswale designs will play an important role in determining the quantity of water that is captured and diverted from combined sewer systems. Specifically, we found that Nepeta (Lamiaceae), a widely used herbaceous plant in NYC green infrastructure, maintained the highest rates of water conductance across our bioswale study sites and may be beneficial to include and maintain in bioswale plant palettes. By contrast, Panicum (Poaceae) consistently had the lowest transpiration rates, probably due to its C4 physiology traits adapted for water conservation. While C4 plants may be beneficial for climates that experience frequent drought, they may not be the best choice for plant palettes in cities such as New York, where precipitation occurs frequently. Our study focused on the herbaceous understory plants of the bioswales, but other life forms such as trees and shrubs are often planted in these types of green infrastructure installations, and would be beneficial to examine in future studies to get integrated estimates of plant water movement from soils. One study of bioswale trees near Chicago found similar results to ours in that tree species varied widely in their transpiration and growth rates, and had dramatic differences on estimated bioswale water budgets (Scharenbroch et al., 2016). Bioswale plant choice can also influence the quantity of pollutants filtered from diverted stormwater (Read et al., 2008), although plant-associated microbes likely mediate those effects. Accounting for context-dependent plant trait variation across different urban climates and soils will be important to identify which plant palette configurations will ultimately contribute to more dimensions of the multifunctional ecosystem services of bioswales.

We found significant relationships between plant transpiration rates and the abundance of several groups of fungi and bacteria in the soils, which could result from two alternative explanations: (1) transpiration rates and a subset of the soil microbial taxa are independently affected by the same abiotic factors or (2) microbes residing in and around the rhizosphere of plants influence plant physiology, or (3) the plant transpiration is influencing the microbial communities in the soil in some way. While we could not distinguish between these hypotheses in the current study, there is accumulating evidence on the associations of microbes with plastic plant phenotypes and physiological processes (Friesen et al., 2011). Some studies find that microbes can influence physiological responses of plants, as microbes can make a unique structural and biochemical pathways available to plants. In a controlled experiment to examine plant soil relations, it is speculated that plants can select for a subset of microbes at different stages of development for different functional purposes (Chaparro et al., 2013). Similarly, studies find that physiological traits of plants can influence the surrounding microbial communities. For example, in a common garden test of spruce seedlings, several groups of fungi and bacteria were significant correlated with a suite of plant physiological traits including transpiration rates (Li et al., 2018), suggesting that these physiological traits can influence associated microbial communities. It is likely that some of the relationships detected were a result of abiotic factors similarly structuring microbial communities and plant transpiration rates and that other relationships were the result of direct interactions between the plants and microbes. Stochastic assembly processes also likely occur, making some groups of microbes weakly related to either plant or abiotic factors. Future experimental manipulations to explore these patterns would be useful in further understanding the functional consequences of plant species selection on water movement in green infrastructure, and how the belowground community matters. Interestingly, the fungi taxa Comamonadaceae and bacteria taxa Chryososporium (Medina et al., 2010) had Pearson correlation values close to 1, across the top three plant water conductors within and across bioswales (Figures 5, 6). Controlled experiments found that species in the Comamondaceae (Li et al., 2015) and Chryososporium host physiological benefits to plants in low quality soils.

Further research is necessary to examine the particular taxa that influence plant transpiration rates in controlled and in situ experiments.

Surprisingly, we found low abundance and diversity of AM fungi in soils collected near bioswale plants. This result may be due to the high inputs of nutrient-rich stormwater runoff that is intentionally shunted into bioswale inlets. High nutrient loads are known to decrease AM fungal abundance, as plants allocate less carbon to these belowground symbionts when nutrients are readily available to plant roots (Treseder, 2004). The vast majority of AM fungal OTUs in our dataset were classified as unidentified, highlighting the need for continued research in the taxonomy and identification of AM fungi, even in urban ecosystems. Of those that were identified, the most abundant genera detected in this were Glomus and Rhizophagus, both of which are cosmopolitan in their distributions and are widely regarded as having generalist life history strategies due to their large genomes and frequent occurrences in disturbed environments (Börstler et al., 2008; Tisserant et al., 2013). Specifically, Glomus species are considered ruderal taxa due to their dominance in agricultural and other frequently disturbed soils, fast growth rate and rhizophilic nature (Chagnon et al., 2013; Weber et al., 2019). In past studies, Glomus innoculations significantly improved plant transpiration abilities and increased resistance to drought in controlled experiments (Augé, 2004; Siddiqui and Singh, 2004). The high relative abundances of these taxa are consistent with previous findings of AM fungal communities in New York City green roof soils, where Glomus and Rhizophagus were the most abundant AM fungal taxa detected.

However, the green roof soils harbored significantly higher relative abundances of AM fungi than the present study, which is likely due to the relatively nutrient-limited and stressful environment of green roof communities (McGuire et al., 2013). Of the five plant species we focused on in this study, we detected the highest abundance of AM fungal OTUs associating with Panicum species. Panicum is a large genus of perennial and annual grasses that are highly dependent on their symbioses with AM fungi for nutrient acquisition, pathogen suppression and tolerance to environmental stressors (Clark et al., 1999). We hypothesized that variation in plant physiology would result in unique assemblages of microbial communities; shifts in the microclimate of individual plants’ rhizospheres due to physical and chemical differences in root structure likely influence assembly of different AM fungi, and plants vary in their dependency on mycorrhizal fungi. The highest abundance of AM fungi detected in the Panicum rhizosphere soils may be due to grass species’ tendency to harbor higher loads of AM fungal symbionts; however, we only sampled the soil nearby plants and not the roots themselves and therefore likely only detected AM extra-radical mycelia and spores. Further research investigating the AM fungi colonizing the different plant species roots in addition to the rhizosphere would help unpack the factors contributing to the shifts in AM fungal abundance observed in this study.

The metagenomic data has been uploaded to the Sequence Read Archive (SRA). Our reference number is PRJNA549763.

OB, KM, and MP conceived the ideas and designed the methodology. OB, JH, and SB collected the data. OB, DD, AG, KS, KM, SB, and MP interpreted the results. OB, KM, DD, and KS wrote the manuscript. All authors contributed to the drafts and gave final approval for publication.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.