Sewage-derived nitrogen dispersal and N-fixation in Port Phillip Bay in south-eastern Australia

Study site

Port Phillip Bay is a 1930 km² marine embayment in south-eastern Australia. The Yarra River, which passes through Melbourne, enters the north of the bay, whereas the bay is connected to Bass Strait by a narrow (3 km) entrance to the south (Fig. 1). The Western Treatment Plant (WTP) treats sewage from 55% of Melbourne’s population of 5 million and discharges ~400 ML day⁻¹ of tertiary sewage at four outlets located along 10 km of the north-western shoreline of Port Phillip Bay in the Geelong Arm (Fig. 1). The WTP is the world’s largest lagoon sewage treatment plant (108 km²); however, between 2004 and 2012, approximately half the sewage was treated by activated sludge plants (ASPs). Most effluent is discharged from the northernmost outlet (15 East), which receives all the effluent from the ASPs and some effluent from the lagoons. The other three outlets (Lake Borrie, 145 West and Murtcaim) received effluent only from the extensive lagoon system, and the residence time of sewage in these lagoons is longer (up to 50 days).

Fig. 1.

Sampling sites in Port Phillip Bay. (a) Location of the STP inputs and of sampling sites on river inputs. The locations of all transplant sites are shown, including Williamstown, where only Hormosira was transplanted. Arrows indicate direction of transplantation. (b) Offshore sites (○) where Botryocladia and Mytilus were deployed, including offshore sites adjacent to pylons with Mytilus attached (●). Numbered sites (red) show where δ¹⁵N was measured for intertidal species. Intertidal sites: 1. Point Lonsdale; 2. Queenscliff pilots pier; 3. St Leonards pier; 4. Indented Head; 5. Portarlington; 6. Avila Road; 7. Eastern Beach; 8. Pivot; 9. Geelong Grammar; 10. West of Avalon Beach; 11. Point Lillias; 12. Point Wilson quarry; 13. Murtcaim outlet; 14. Kirk Point; 15. 145 West outlet; 16. 15 East outlet; 17. Point Cook; 18. Altona pier; 19. North of Altona pier; 20. Williamstown; 21. St Kilda pier; 22. Brighton; 23. Ricketts Point; 24. Davey Bay; 25. Snapper Point; 26. Mount Martha; 27. Sorrento (Collins Settlement). Five larger circles show the five moorings or pylons with the lowest δ¹⁵N values.

The outfalls discharge the treated effluent close to shore onto inter-tidal sand flats. The freshwater discharge is buoyant and thereby highly influenced by wind. The circulation in the bay is three-dimensional with up/downwelling around the shorelines (Black et al. 1993), but the WTP plume often travels along the north-western shoreline towards the Yarra River, especially during the winter when the winds are mostly from the westerly quadrant. Similar behaviour is seen in this paper because the experiments were conducted in winter from early July to early September (Table 1).

Nutrient, freshwater and δ¹⁵N inputs and bay nutrients

Between July and September 2012, while Botryocladia and Mytilus were deployed in Port Phillip Bay, salinity, nutrient concentrations (NO_X, NH₄⁺, PO₄⁻³) and δ¹⁵N of NO₃⁻ and NH₄+ were measured at nine major discharges (Table 1a), whereas salinity and nutrients were measured at 10 intertidal sites on the western shoreline and at the 42 offshore mooring sites (Fig. 1, Table 1b). The frequency of each of these measurements is shown in Table 1a, b. Flow rates of all discharges (measured by Melbourne Water, and at Altona STP by City West Water) and mean DIN concentrations were used to calculate the mean daily load of ammonium, nitrate and of δ¹⁵N (Table 2).

Salinity was measured in the laboratory with a YSI Pro2030, calibrated against 0.582% KCl solution. Nutrient concentrations, and δ¹⁵N of NO₃⁻ and NH₄+, were measured using water samples filtered in the field by using 0.45 μm syringe filters, then frozen until analysed. Nutrient samples were analysed by the Water Studies Centre, Monash University. δ¹⁵N of NO₃⁻ and NH₄+ were analysed at Colorado Plateau Stable Isotope Laboratory (CPSIL), Northern Arizona University (Parry 2013). Nutrient samples with salinities of >10‰ were excluded from the estimates to ensure only undiluted nutrient and δ¹⁵N inputs were measured. This threshold was chosen on the basis of plots of the relationships between nutrient concentrations versus salinity and salinity versus the difference between sampling time and low tide (Parry 2013).

During the field study, the nine major discharges to the Bay exhibited elevated mean concentrations of nutrients and δ¹⁵N of NH₄⁺ and NOx (Table 2). The largest dissolved inorganic nitrogen (DIN) inputs to Port Phillip Bay came from the 15 East outlet of the WTP (10.6 tonnes N day⁻¹), the Yarra River (4.1 tonnes N day⁻¹), 145 West outlet of the WTP (1.7 tonnes N day⁻¹) and the Werribee River (0.83 tonnes N day⁻¹). All other inputs were less than 100 kg N day⁻¹ (Table 2). Notably, WTP discharges more NH₄⁺ than NOx, while rivers and drains are dominated by NOx.

The values of δ¹⁵N in DIN were two to three times higher at the WTP discharges (17–32‰) than in the Yarra River (8‰) and Werribee River (13‰, Table 2). Discharge from Altona STP was mid-range, with δ¹⁵N = 15‰, but flows were small (Table 2).

Spatial distribution of δ¹⁵N and δ¹³C in biota

The red alga Botryocladia sonderi and the mussel Mytilus galloprovincialis were deployed in early July 2012 on moorings throughout Port Phillip Bay. Two months later in early September (Table 1C), δ¹⁵N and δ¹³C (algae only) were measured for these two offshore species and for four intertidal species.

The intertidal species were the algae, Hormosira banksia and Capreolia implexa, the filter-feeding serpulid worm, Galeolaria caespitosa, and the (intertidal or shallow subtidal) filter-feeding mussel, Mytilus galloprovincialis. In total, 27 sites around the perimeter of the bay were monitored (Fig. 1). Seasonal changes in δ¹⁵N were also measured for these four species at 15 sites on the west of the bay (Table 1C).

The offshore species Botryocladia and Mytilus were deployed on the same 42 moorings. Moorings were located throughout the bay and 15 were deployed 50 m from channel markers where Mytilus was also sampled in situ (Fig. 1). Four moorings, two in Corio Bay and two in the east of the bay, were lost.

Offshore moorings consisted of rope held on the seabed with a steel weight connected to a 200-mm diameter polystyrene float 1–1.5 m below the sea surface and a small surface buoy. Botryocladia and Mytilus were attached 0.5–1.0 m below the subsurface buoy. At each site, Botryocladia was deployed with two different methods. Botryocladia stems were inserted into 30 cm lengths of surgical rubber tubing and held there using small cable ties and it was also placed in an open-ended plastic cylinder enclosed by a mesh bag held tight by a cable tie (Parry 2013). Samples obtained from both deployment methods consisted of 4–12 newly grown (pink, shiny, clean and distal) vesicles that were dissected from each plant.

Mytilus was obtained from a mussel farm at Pinnace Channel near the open-ocean entrance to the Bay (Fig. 1) and 20 were deployed in mesh oyster bags on each mooring.

Following collection, Capreolia was shaken in seawater to remove as much sand as possible, then any epiphytes and any substrate still attached to its holdfasts were carefully removed under a dissecting microscope. Galeolaria and Mytilus were dissected in the laboratory with fine scissors. At each site, opercular adductor muscles and feeding crowns were removed from 30 Galeolaria and a composite sample of each tissue was sampled and analysed. Similarly, at each site the posterior adductor muscle and a sample of gill tissue (approximately 1 cm × 1 cm) was dissected from five Mytilus individuals and a composite sample of each tissue was analysed. At each site, a sample of 10 penultimate Hormosira vesicles, with smaller newly grown distal vesicles attached, were collected. Two samples were analysed from each site. One consisted of old growth in the 10 larger (5–10 mm diameter) penultimate vesicles and the other consisted of new growth from all the attached (<5 mm diameter) distal vesicles, typically 15–20 small vesicles per sample.

All samples for δ¹⁵N and δ¹³C analysis were dried at 60°C for 24–48 h, then ground in a mortar and pestle. Values of δ¹⁵N and δ¹³C for algae and invertebrates were measured at the Water Studies Centre, Monash University.

Transplant experiments

Transplant experiments were conducted to determine the period over which Capreolia, Hormosira, Galeolaria and Mytilus integrate the ¹⁵N signal by transferring them to higher-background ¹⁵N sites. Changes in their δ¹⁵N were monitored weekly initially, then at longer intervals, for 4–5 months. Between 9 and 12 June 2012, all four species were transplanted from areas of low ¹⁵N (Mount Martha and Pinnace Channel) to areas of high ¹⁵N (Point Cook and Point Wilson). The latter sites were the closest intertidal reefs north and south of the WTP. In addition, Hormosira was transplanted to Williamstown (slightly north of Point Cook) because there was no in situ Hormosira at Point Cook, so conditions there may not have been suitable for it. Mytilus was transplanted from a mussel farm at Pinnace Channel and deployed on short 1-m moorings in water 1–1.5 m deep. Other taxa were transplanted from Mount Martha on rocks that were placed adjacent to the same taxa as those growing in situ, except where there was no in situ Hormosira at Point Cook.

Turnover of isotopes in tissue is usually approximated by an exponential relationship (McIntyre and Flecker 2006), as adopted here for transplanted Galeolaria and Mytilus. These exponential curves are of the form y = a+be^kt, where a, b and k are constants and t = time. The half-life of tissue turnover is given by T_1/2 = log_e2/k, where e is the base for natural logarithms (Gustafson et al. 2007; Ankjærø et al. 2012). Exponential curves were not fitted to describe changes in δ¹⁵N of the algae Capreolia and Hormosira because most change in these species occurred in the week after their transplantation. For Galeolaria, changes after 3 months no longer fitted an exponential curve; hence, the curve was fitted excluding the last two sampling dates (Parry 2013).

Application of Bubbles NPZ Model

The three-dimensional nutrient–phytoplankton–zooplankton (NPZ) model Bubbles has been shown to provide close predictions of nutrients and planktonic productivity (Black et al. 2016; Jenkins and Black 2019, 2021). When calibrated against field data in Port Phillip Bay, the model predicted the concentration of the 12 most abundant phytoplankton, down to individual species and total diatoms and flagellates.

Most recently, Bubbles was calibrated over a 10-year period from 2010 to 2019 inclusive (Black et al. 2022). Calibration compared the model to monthly measurements from the EPA (Victoria) at six sites that encompass the main hydrodynamic and biological zones around Port Phillip Bay, from the lower currents of Corio Bay to the higher nutrient loads near WTP, the Yarra River, and the ocean-influenced Sands region at Dromana. The 10-year period experienced a range of wet and dry years. The model simulations are annual, starting from 1 July of each year. The model uses a cold start where key variables such as the initial N values are set from data at six sites interpolated onto all model cells. Consequently, the first 1–2 weeks of the simulation are still coming into balance with the actual baywide conditions, so the initial 2 weeks of model runs were excluded from the analyses in this paper.

The model was checked against 12 individual species of phytoplankton, total diatoms and total flagellates, nutrient concentrations, ratio of N species, seabed N and Si fluxes/concentrations, temperature, salinity and Chl-a for a total of 22 different parameters. The model output is written to a stand-alone file at 1 hourly intervals over the 10-year period. In this paper, the required data were extracted and analysed for the period of the ¹⁵N experiments.

Model Bubbles uses output from a three-dimensional, stratified ocean/atmosphere hydrodynamic model to specify the currents for advection and diffusion. Bubbles considers the N cycle via growth and decay of phytoplankton and zooplankton, immediate remineralisation, seabed denitrification and losses to the atmosphere.

The model was used to:

Nutrient measurements

Offshore water sampling was undertaken three times between July and September 2012. When combined with simultaneous intertidal sampling, these showed that NH₄⁺, NOx and PO₄⁻³ concentrations were high along the north-western coast adjacent to and north of the WTP. NOx was also high near the Patterson and Yarra rivers (Fig. 2), whereas the WTP was the dominant source of DIN. Concentrations of all nutrients declined offshore from the WTP, but NH₄⁺ and NOx declined more sharply than did PO₄⁻³ (Fig. 2).

Fig. 2.

Mean nutrient concentrations based on onshore (N = 8) and offshore (N = 3) measurements between June and September 2012 for (a) NH₄⁺, (b) NOx and (c) PO₄⁻³. (d) Output from Bubbles model for NH₄⁺ + NOx (μM) in the period 14 July 2012 to 4 September 2012. The arrows show the key similarities, namely (I) bulge in plume towards Bellarine Peninsula, (II) high concentration around WTP, (III) low concentrations to the west of the Yarra River near Altona, (IV) Yarra River and Patterson River plumes.

Nutrient concentrations at the 10 intertidal sites showed the anticipated high values north of the WTP (Sites 14, 17, 18 and 20, Table 3, Fig. 3c). A secondary DIN peak occurred on the southern side of the Geelong Arm at Site 6 (Fig. 3c).

Fig. 3.

Spatial differences at Sites 1–27 (see Fig. 1) in (a) δ¹⁵N of four intertidal species growing in situ, (b) δ¹³C of two intertidal algae growing in situ, and (c) DIN (NH₄ + NOx) (μM), on the basis of measurements between June and September 2012. All error bars are standard errors.

As a validation of the numerical model, Fig. 2d shows DIN (NOx + NH₄⁺) averaged over the experimental period from 14 July to 4 September 2012. DIN was measured on three occasions, always during calm conditions, when the wind was offshore from the north-western quadrant, whereas the model averages show all conditions at 1-hourly intervals over a 6-week period, when wind and plume dynamics differed widely (Fig. 4). Extreme plume fluctuations expose monitored species to WTP effluent where N enriched in ¹⁵N will be taken up, but they do not dominate the 2-month mean. Both the patterns and magnitudes in the model (Fig. 2d) are very similar to the NH₄⁺ and NO_x measurements (Fig. 2a, b). Key features include (I) high concentrations near the shoreline around the WTP outfalls; (II) bulge to the south-east towards Bellarine Peninsula (as seen in the NH₄⁺ measurements); (III) low concentrations between Altona and the Yarra River; and (IV) plumes associated with the Yarra and Patterson rivers.

Fig. 4.

Examples of NH₄⁺ + NOx (μM) plume dynamics during different wind conditions between 3 July and 4 September 2012.

Differences in δ¹⁵N and δ¹³C owing to season, tissue type and deployment method

Temporal differences in δ¹⁵N were small in all species compared with spatial differences, and showed no clear seasonal pattern, except that δ¹⁵N for Galeolaria was consistently 0.5–1.5‰ lower at all sites in March 2012 (Supplementary Fig. S1). Similarly, differences in δ¹⁵N and δ¹³C between new and old growth of Hormosira, and between δ¹⁵N from different tissues of Galeolaria and Mytilus were small compared with spatial differences, and all showed very similar spatial patterns for in situ specimens (Fig. S2) and similar temporal patterns for transplanted specimens (Fig. S3). Notably, muscle tissue from transplanted Galeolaria and Mytilus adopted the δ¹⁵N of their new site more slowly than did their crowns and gills (Fig. S3). In all subsequent analyses, at each site measurements from the three main sampling times (March 2012, June 2012 and September 2012), new and old growth of Hormosira, and both tissue types for Galeolaria and Mytilus, were averaged.

Botryocladia had more epiphytes when attached to moorings by surgical tubing rather than in (more shaded) plastic cylinders (Parry 2013). The value of δ¹⁵N of Botryocladia was slightly higher in tubes than in plastic cylinders (Fig. S4). Because both replicates of Botryocladia in tubes were lost from some moorings, mean δ¹⁵N of Botryocladia at each site was estimated using all replicates (tubes and cylinders), after making a small correction for differences in δ¹⁵N caused by different deployment techniques, by using the equation shown in Fig. S4. There was no significant difference in δ¹³C between Botryocladia deployed in tubes and those in cylinders.

Spatial distribution of δ¹⁵N and δ¹³C in biota

Offshore biota

At the offshore sites, the general pattern in Fig. 4a–c is a broad 10–15 km wide zone off the WTP of high DIN, high δ¹⁵N and δ¹³C for Botryocladia and high δ¹⁵N for Mytilus. This zone extends north-west to the Yarra River, whereas its width and magnitude both reduce northward. The zone extends into the Geelong Arm immediately south of the WTP but does not usually extend west of Point Wilson. Some notable differences among patterns shown by DIN, δ¹⁵N and δ¹³C include the following:

1. DIN concentrations had a smaller footprint than did δ¹⁵N for Botryocladia and Mytilus, on the basis of a comparison of Figs 2 and 5, which share the same sampling sites.

2. δ¹⁵N values for Botryocladia declined progressively offshore, but high values nearest the outfalls may not reflect the very high DIN concentrations measured there. For example, the mean δ¹⁵N measured in the 145W discharge was 24.3‰ (Table 2), but δ¹⁵N of the closest Botryocladia, deployed 500 m away, was 18.5‰.

3. δ¹⁵N values for Mytilus showed even less variation across the width of the plume than did those for Botryocladia.

4. Modelled distributions of DIN showed spatial patterns similar to those of δ¹⁵N of Botryocladia and modelled flagellate distributions mirrored δ¹⁵N of phytoplankton consuming Mytilus, except possibly in Hobsons Bay.

5. Two adjacent sites offshore from the north-east of the Bellarine Peninsula had high and low values for δ¹³C for Botryocladia that were unrelated to their proximity to the WTP.

The five sites with the lowest δ¹⁵N of Botryocladia and Mytilus were in Swan Bay, three sites in Corio Bay and along the northern coastline between Corio Bay and WTP (Figs 1b, 5a, c, d).

Fig. 5.

Spatial variation during September 2012 in species deployed on offshore moorings between June and September 2012. (a) δ¹⁵N in Botryocladia, (b) δ¹³C in Botryocladia (c) δ¹⁵N in Mytilus and (d) collected in situ on pylons δ¹⁵N in Mytilus in September 2012.

The δ¹⁵N of Mytilus collected in situ on pylons showed a spatial pattern similar to that of Mytilus deployed on moorings 50 m away (Fig. 4c, d). The values of δ¹⁵N of Mytilus on 2-month deployments on moorings converged towards values of δ¹⁵N of Mytilus that had lived on adjacent pylons for their lifetime (Fig. 6).

Fig. 6.

Plot of (δ¹⁵N of Mytilus growing in situ on pylons − δ¹⁵N of Mytilus deployed on moorings 50 m away) against δ¹⁵N of Mytilus growing in situ on pylons. Measurements were obtained in September 2012. The grey area is δ¹⁵N of Mytilus on deployment, based on measurements in June and September 2012 at Pinnace Channel, where all deployed mussels were collected.

Although the WTP was the major influence on spatial patterns of δ¹⁵N in Port Phillip Bay, there were several minor influences. The low δ¹⁵N of the Werribee River (δ¹⁵N of DIN = 13‰), which discharges within the WTP zone (δ¹⁵N of DIN at 15 East outlet = 17‰), appeared to cause a region of lower δ¹⁵N adjacent to the coast for both Botryocladia and Mytilus north and south of the Werribee River (Fig. 5a, c). The footprint of the small Altona STP (<100 kg N day) probably declined sharply with distance from the outlet, but its apparent footprint shown by contours in Fig. 5a was exaggerated by a lack of nearby sampling stations.

Transplant experiments

The values of δ¹⁵N of Capreolia, Hormosira, Galeolaria and Mytilus transplanted to Point Cook and Williamstown increased, but biota transplanted to Point Wilson quarry did not because, despite its proximity to the WTP, δ¹⁵N values of in situ biota there were similar to those at Mount Martha and Pinnace Channel (Figs 1a, 6).

At the transplant sites, δ¹⁵N of control/in situ Galeolaria and Mytilus showed little change during the 3–5 month transplant experiment, whereas δ¹⁵N of control/in situ algae Capreolia and Hormosira varied (Fig. 7).

Fig. 7.

Comparison of changes in δ¹⁵N of in situ and transplanted species between June and December 2012 (months are indicated by their first letters). (a) Capreolia transplanted from Mount Martha to Point Cook and Point Wilson, (b) Galeolaria transplanted from Mount Martha to Point Cook and Point Wilson, (c) Hormosira transplanted from Mount Martha to Point Cook and Point Wilson, (d) Hormosira transplanted from Mount Martha to Point Cook and Williamstown, (e) Mytilus transplanted from Pinnace Channel to Point Cook and Point Wilson.

Following transplantation to Point Cook, the rate that δ¹⁵N increased varied greatly among species. The rate of change for Capreolia and Hormosira was much greater in the week following their transplantation than subsequently (Fig. 7). The value of δ¹⁵N of Capreolia increased for ~6–12 weeks when an asymptote appeared to be reached (Fig. 7), but there was a temporal trend in δ¹⁵N of in situ/control Capreolia, making it difficult to determine precisely when/if the transplanted and in situ measurements converged (Fig. 7). The value of δ¹⁵N of Hormosira transplanted to Point Cook and Williamstown approached asymptotes after ~10 weeks, when δ¹⁵N of Hormosira transplanted to Williamstown also converged with in situ values, although there was substantial temporal variation in the latter (Fig. 7).

Values of δ¹⁵N of Galeolaria transplanted to Point Cook increased slowly for ~2.5 months, when an asymptote may have been reached, but at a lower value than for in situ samples (Fig. 7). An exponential curve describing changes in δ¹⁵N of Galeolaria for the first 2.5 months is as follows:

Values of δ¹⁵N of Mytilus transplanted to Point Cook increased slowly for the 4 month duration of the transplant experiment (Fig. 7). At this time, δ¹⁵N values of transplanted Mytilus were still much lower than δ¹⁵N values of in situ Mytilus (Fig. 7). An exponential curve describing changes in δ¹⁵N of Mytilus for the 4 months of transplantation is as follows:

The nutrient footprint of the WTP

DIN measurements, modelled DIN plumes and stable isotope ratios in Botryocladia and Mytilus all broadly indicate that the footprint of WTP-sourced N extends 10–15 km offshore from the WTP along the north-western shoreline, extending into the Geelong Arm immediately south of the WTP, but does not usually affect the Geelong Arm west of Point Wilson or the eastern two-thirds of Port Phillip Bay. The footprint of δ¹⁵N of Mytilus extended ~10–15 km offshore, with the extension being slightly larger than that of Botryocladia (~10 km), because Mytilus consumes phytoplankton that disperse after taking up WTP-sourced N.

Measured and modelled DIN concentrations and δ¹⁵N of Botryocladia showed similar spatial patterns (compare Figs 2 and 4a). The pattern of δ¹⁵N of Botryocladia should reflect that of the N-enriched plume of the WTP, because it obtains its N from the water column directly. The model shows that the WTP plume is very variable and highly dependent on wind conditions (Fig. 4). Although short-lived events do not dominate the 2-month mean, fluctuations in the plume still expose monitored species to WTP effluent and allow ¹⁵N to be taken up over a broad region that will vary with seasonal winds. The modelled DIN plume periodically impinges on the coast of the northern Bellarine Peninsula (Fig. 4), explaining the high δ¹⁵N values of Galeolaria and Mytilus at Site 5 (Fig. 3a).

There were important deviations from the broadly similar patterns shown by DIN, δ¹⁵N and δ¹³C. Several of these deviations are likely to be due to fractionation on uptake of N. Fractionation results from biological processes preferentially consuming the lighter N isotope (¹⁴N), resulting in the residual reactant pool becoming enriched in the heavier isotope (¹⁵N) (Kendall et al. 2007; Montoya 2008).

Typically, fractionation causes δ¹⁵N to increase by 3–4‰ with each trophic level (Peterson and Fry 1987; Post 2002; Jennings and Warr 2003), being similar to the mean difference in δ¹⁵N between co-occurring intertidal plants and filter-feeders (3.2–3.8‰).

Fractionation owing to mineralisation of organic matter is consistently near zero (Brandes and Devol 1997; Montoya 2008), whereas fractionation by nitrifying bacteria in sediments typically causes isotope effects of 15–25‰, and by denitrifying bacteria of 20–30‰ (Montoya 2008). However, the isotope effect across the sediment water interface (ε_sed) is often near zero, because nitrification and denitrification are limited by the available N (Brandes and Devol 1997; Lehmann et al. 2007; Devol 2015; Montoya 2008; Glibert et al. 2019). In addition, nitrification and denitrification have compensatory effects on ε_sed because the light NO₃ from nitrification becomes the input to denitrification that creates heavier NO₃ and light N₂ (Devol 2008). The value of ε_sed is closest to zero at the most oxygenated sites with the minimal reactive organic matter (Lehmann et al. 2004; Alkhatib et al. 2012), conditions similar to those found in most sediments in Port Phillip Bay, whereas ε_sed is most likely to be greater than 0‰ where denitrification efficiency is low and surface sediments are poorly oxygenated (Lehmann et al. 2004; Granger et al. 2011; Alkhatib et al. 2012; Kessler et al. 2014).

Fractionation owing to uptake of N by algae is often dependent on DIN concentration (Swart et al. 2014). ¹⁵N is detected by Botryocladia with great sensitivity near the edge of the plume, but nearshore where DIN is high, δ¹⁵N values may be damped by preferential uptake of ¹⁴N. The very high concentrations of DIN and very high δ¹⁵N values in WTP discharges were not mirrored by very high δ¹⁵N values in Botryocladia nearest the WTP. This difference is likely to be due to fractionation on N uptake. Botryocladia, like other plants (Dudley et al. 2010; Swart et al. 2014; Orlandi et al. 2017; Pruell et al. 2020), probably takes up ¹⁴N preferentially at high DIN concentrations, whereas at low DIN concentrations near the edge of the plume, all N is taken up and there is no fractionation. Swart et al. (2014) found that fractionation of nitrate by macroalgae ceased when its concentration was ~2 μM.

Minimal variation shown in δ¹⁵N of Mytilus across the width of the WTP plume is also probably the result of fractionation. In contrast to fractionation shown by algae, filter-feeders (Daphnia) fed algae with high N content showed near-zero fractionation, whereas those fed algae with low N showed fractionation of ~6‰, possibly owing to greater excretion of ¹⁴N when N intake is lower (Adams and Sterner 2000). Similarly, an experimental study found that phytoplankton continuously cultured in high DIN concentrations showed a depletion of δ¹⁵N of up to 9‰ compared with algae cultured in low DIN, but δ¹⁵N of mussels feeding on phytoplankton from high DIN treatments had δ¹⁵N values only 2‰ lower than those on feeding on low DIN phytoplankton (Pruell et al. 2020). A conceptual model of fractionation near the WTP, on the basis of Adams and Sterner (2000), across two trophic levels in areas of high and low DIN, is shown in Fig. 8.

Fig. 8.

Conceptual diagram showing fractionation processes at two trophic levels near (high DIN) and distant from the WTP (low DIN), based on Adams and Sterner (2000).

The Bubbles model incorporates N utilisation, recycling and eventual losses to reproduce the DIN plume. For example, the salinity averaged over the same 3 months shows no significant correspondence with the distribution of NH₄+ or flagellates (compare Figs 9 and S5). Salinity is a conservative tracer, not subject to the complexities of the N cycle. Similarly, PO₄³⁻ shows a much more even distribution than does NH₄⁺ (Fig. 2) because P is abundant in the bay, and studies have shown that N, rather than P, is the limiting nutrient (Harris et al. 1996) for flagellates.

Fig. 9.

Outputs from the Bubbles model (Jenkins and Black 2021) for the period 14 July to 3 September 2012, when Botryocladia and Mytilus were deployed in the bay, showing (a) NOx + NH₄⁺ concentration (μM) in Port Phillip Bay, and (b) average flagellate concentration (individuals.L⁻¹).

The Bubbles model of DIN concentrations (Figs 2d, 9a) closely predicts the patterns shown by δ¹⁵N in Botryocladia (Fig. 4a), and the predicted flagellate distribution (Fig. 9b) closely mirrors the patterns shown by δ¹⁵N in the phytoplankton consuming Mytilus (Fig. 4c). The latter has the larger footprint because phytoplankton disperse after consuming WTP-sourced N.

The values of δ¹⁵N in Hobsons Bay (δ¹⁵N of Botryocladia ~12.0) were substantially higher than the δ¹⁵N of the high volume of NOx discharged to this region by the Yarra River (δ¹⁵N = 8.0, Table 2), and higher than δ¹⁵N values of Capreolia at Intertidal sites 21 (δ¹⁵N = 9.0) and 22 (δ¹⁵N = 7.5), which are more influenced by the Yarra River plume, which follows the eastern shoreline. The elevated δ¹⁵N values in Hobsons Bay may be due to infrequent excursions of WTP-sourced N, but could also be influenced by fractionation in the sediments. Hobsons Bay has characteristics typical of other areas with ε_sed greater than zero, including low denitrification efficiency (28%, Berelson et al. 1998), high organic load and poorly oxygenated surface sediments (Lehmann et al. 2004; Alkhatib et al. 2012).

Values of δ¹³C of Botryocladia deployed in the bay showed broadly similar, but more complex, spatial patterns than those of δ¹⁵N. Plants discriminate against ¹³CO₂ during photosynthesis, but the degree of discrimination varies among species using different photosynthetic pathways (e.g. C₃, C₄ and CAM plants). Macroalgae with δ¹³C smaller than −30‰ (such as Botryocladia) typically occur in subtidal or shaded intertidal habitats where they are often light rather than carbon limited, and rely on CO₂ diffusion for carbon uptake, a process that actively discriminates against ¹³C (Marconi et al. 2011). Proximity to the WTP was expected to reduce the likelihood of nutrient limitation and increase the likelihood of carbon limitation, which in turn will decrease the discrimination against ¹³C, and so increase δ¹³C. The values of δ¹³C of Botryocladia offshore from the north-eastern end of the Bellarine Peninsula were inconsistent with the broad pattern shown by δ¹⁵N and with proximity to the WTP. These δ¹³C outliers probably resulted from interactions among carbon, nutrient and light limitation. Light available to Botryocladia was probably affected by shading from nutrient-dependent epiphytes, and those along the shipping channel by turbid plumes from passing ships. The complexity of the δ¹³C pattern is consistent with previous studies that have shown δ¹⁵N to provide a more reliable estimate of nutrient/sewage influence than does δ¹³C (Roelke and Cifuentes 1997; Gaston et al. 2004; Dudley et al. 2010).

Aquatic macrophytes with δ¹³C between −10 and −30‰ (such as Capreolia and Hormosira) typically inhabit intertidal zones and use HCO₃⁻ as an inorganic carbon source, and usually have a carbon-concentrating mechanism (Marconi et al. 2011). Whereas δ¹⁵N values for intertidal algae and filter-feeders were all spatially correlated, some of the spatial variation in δ¹³C in Capreolia (e.g. high values at Sites 4, 22 and 23, Fig. 3) appeared unrelated to the likely exposure to the WTP plume. However, the Capreolia nearest to the WTP, at Point Cook (Site 17, Fig. 3), had the highest δ¹³C value and this was similar to the high δ¹³C values measured adjacent to another treated-sewage discharge (Parry 2021).

Regions of low ¹⁵N, N-fixation by seagrass

Nitrogen fixation by sulfate-reducing bacteria associated with seagrass roots is a major source of light N (δ¹⁵N of ~0) and often provides for a significant proportion of the nitrogen needs of seagrasses (Welsh 2000), including Zostera muelleri (Udy and Dennison 1997). In this study, the lowest δ¹⁵N signatures of Botryocladia and Mytilus were recorded in Corio Bay and Swan Bay, where near-zero δ¹⁵N signatures were also recorded for seagrasses (Cook et al. 2015; Jenkins et al. 2015; Hirst et al. 2016; Russell et al. 2018). There have been several interpretations of the low values of δ¹⁵N of seagrass at particular sites in Port Phillip Bay. Cook et al. (2015) suggested that the small difference between δ¹⁵N of Ulva and seagrass (Zostera muelleri) at the same sites in Port Phillip Bay was due to isotopic fractionation in the sediment rather than N-fixation, and their direct measurements suggested that N-fixation was only a minor (~15%) source of N in seagrass. Hirst et al. (2016) suggested that low δ¹⁵N of seagrass in Swan Bay may be due to either N-fixation or recycling within the sediment. Subsequently, Russell et al. (2018) found that δ¹⁵N of pore water was linearly related to that of seagrass, and where δ¹⁵N of seagrass was near zero (Corio Bay and Swan Bay), so was δ¹⁵N of NH₄⁺ in pore water.

The low δ¹⁵N in both pore water and the water column at Corio Bay and Swan Bay suggests that the ultimate source of this nitrogen is nitrogen-fixing bacteria. The enclosed geography of these bays looks to be important in the retention of this nitrogen. Both bays have small entrances that reduce water exchange and ensure minimal export of local seagrass production. Seagrass leaves shed by storms or when rhizomes are consumed by black swans become detritus and are buried and broken down by microbes, releasing their N to pore water, which is then recycled by plants. Both bays also have limited fetch, which helps retain fine sediments, which in turn retain more NH₄⁺ in pore water (Hirst and Jenkins 2017). The low export of seagrass production contrasts with sites on the east of Port Phillip Bay (including Blairgowrie, Rye and South Sands, Hirst and Jenkins 2017) where most seagrass production is exported, sediment is coarser with less retained organic matter, NH₄⁺ concentrations in pore water are an order of magnitude lower and, in contrast to Swan Bay, the addition of nutrients to the sediment increases the rate of growth of seagrass (Hirst and Jenkins 2017).

The lower δ¹⁵N values found for all species in the Geelong Arm (Sites 5–12, Fig. 3a) than those on the east of Port Phillip Bay (Sites 20–27, Fig. 3a) are likely to be in part the result of seagrass contributing more light N in this region. Biota at some sites along the northern shoreline of the Geelong Arm also had unexpectedly low δ¹⁵N values, suggesting that this area also has a source of light nitrogen as well as being influenced by the plume of the WTP. The δ¹⁵N value was low at Sites 10 and 12, but higher at Site 11. The lower δ¹⁵N values that occur at some sites in the Geelong Arm may result from their proximity to geographic features that trap large amounts of seagrass wrack that release light ¹⁴N. Similar fine-scale variation in δ¹⁵N of in situ macroalgae was found to be due to decomposition of organic matter from seasonal seaweed standings (Lemesle et al. 2016).

The exceptionally low δ¹⁵N values found in all species suspended in the water column in seagrass habitats indicates that some of the N fixed by seagrass is exported to the water column. This will be greater where extensive meadows of seagrass occur (Corio Bay, Swan Bay, and Point Wilson quarry) and probably where seagrass drift accumulates because of the groyne-like geographic features. The amount of N exported to the water column as a result of nitrogen fixation from seagrass is difficult to estimate from available data, but an approximate estimate was obtained for Swan Bay, using best estimates of δ¹⁵N for water outside and within Swan Bay. The value of δ¹⁵N of water outside Swan Bay was estimated from its value at the entrance to Port Phillip Bay, estimated from the mean δ¹⁵N of Capreolia and Hormosira at Point Lonsdale [4.8 ± 0.3 (N = 8)]. The value of δ¹⁵N of water in Swan Bay was estimated from δ¹⁵N of in situ Mytilus in Swan Bay [7.5 ± 0.4 (N = 2) – 3.5 ± 0.3 (trophic fractionation) = 4.0 ± 0.7]. These values suggest that, if δ¹⁵N of N fixed by seagrass is 0, then the proportion of nitrogen in the water column fixed by seagrass is (4.8–4.0)/4.0 = 0.20, but the error range includes zero. More measurements are clearly needed, particularly of δ¹⁵N of in situ macroalgae that source their N from the water column.

Changes following transplantation

The rate of change of δ¹⁵N, when an organism is introduced into an environment with a distinctively different ¹⁵N signal, depends on tissue growth and N turnover associated with tissue maintenance. The contribution of these two processes will vary, depending on the rate of growth (Fry and Arnold 1982; Gaye-Siessegger et al. 2004; McIntyre and Flecker 2006; Ankjærø et al. 2012). The rate of change of δ¹⁵N will be slowest when there is no growth and all change is due to tissue turnover for maintenance. Where there is new growth, the newly incorporated N will increase with growth rate, because new tissue would be expected to carry primarily the ¹⁵N signature of the new environment.

The rapid increase in δ¹⁵N of Capreolia and Hormosira in the first week, and of nitrogen content in the first 3 weeks, were probably due to ‘luxury uptake’, where nutrients are placed in intracellular storage rather than used immediately for protein growth (Kilham and Hecky 1988). Subsequent increases in δ¹⁵N occurred more slowly because they were due to algal growth. The rate of change of δ¹⁵N in transplanted Galeolaria anomalously reached an asymptote well below the δ¹⁵N of in situ Galeolaria, probably because they were placed in a non-optimal micro-environment where growth slowed.

There are few measurements of tissue-stable isotope turnover rates (Dattagupta et al. 2004; McIntyre and Flecker 2006), although primary producers have higher variance in δ¹⁵N than do secondary consumers (Cabana and Rasmussen 1996; Post 2002; McIntyre and Flecker 2006). Aguiar et al. (2003) suggested that the nitrogen pool in the macroalgae Ulva turns over completely every 12–15 days. This estimate of turnover of nitrogen is much faster than those of macroalgae from this study (6–12 weeks and 10 weeks), probably for two reasons. First, Ulva is widely recognised as a species associated with areas of high nutrients and has a high rate of growth. Second, the estimate for Ulva by Aguiar et al. (2003) was an extrapolation of the rate over the first 8 days, when most nitrogen probably went to storage (‘luxury uptake’) rather than growth. Estimated turnover rates of nitrogen in Galeolaria and Mytilus (T_1/2 = 90 days and 234 days respectively) appear similar to the limited data available for other species. McIntyre and Flecker (2006) reviewed available estimates of turnover rates and found half-lives (T_1/2) between 156 and 784 days for krill, 69 days for Elimia (snail), 99–231 days for Littorina (snail), and very variable rates for fish (3 to >173 days), depending largely on their size and growth rate. Subsequent estimates of T_1/2 for nitrogen in fast-growing Atlantic cod, in heart and white muscle, were 31–78 days (Ankjærø et al. 2012). However, Dubois et al. (2007) found much more rapid turnover rates, T_1/2 = 14 days for Pacific oysters and T_1/2 = 15 days for Mytilus. These latter estimates are higher than estimates for algae, and appear unrealistically high, especially as mussels only grew 38% over 90 days.

Preferred species for ongoing monitoring

Values of δ¹⁵N in all species showed similar spatial patterns around Port Phillip Bay, suggesting that all are satisfactory for monitoring sources of N with distinctive δ¹⁵N values. But the absence of intertidal reef habitat adjacent to the WTP limited the usefulness of species dependent on this habitat, especially Capreolia and Hormosira. Both Galeolaria and Mytilus can also establish on small coastal man-made structures. The absence of Hormosira on the intertidal reef at Point Cook is consistent with its low tolerance of high DIN (Parry 2021), which suggests that it is the least suitable species for monitoring δ¹⁵N of WTP-sourced N.

The suitability of different species depends on the time scale of the questions being posed, and the costs of sampling and analysis. Species and tissues with high turnover rates (e.g. algal tissues) are more responsive to short-term changes in the discharge of ¹⁵N, whereas slower-growing taxa (e.g. Galeolaria and Mytilus) and tissues with longer turnover rates (e.g. muscle, Lorrain et al. 2002) are better suited to providing a time-integrated measure of the area of influence of discharged ¹⁵N (Gartner et al. 2002; Fertig et al. 2009; Raimonet et al. 2013). The costs of collection and analysis are also important. Collection costs are lower for intertidal species than for those deployed on moorings. But the mean time taken to dissect and grind samples was generally greater for intertidal species (Capreolia 20–25 min, Hormosira 5 min, Galeolaria 35 min) than for those deployed on moorings (Botryocladia 5 min, Mytilus 14 min). Species attached to offshore moorings provide the only means of measuring δ¹⁵N in non-coastal regions and are also better suited to monitoring long-term trends from sources with a large footprint such as the WTP, because, unlike intertidal species, they are not affected by small local inputs.

Of the five species considered in this study, Mytilus appears the best species for monitoring long-term change in ¹⁵N, and for ground-truthing nutrient models. The current study deployed large mussels, but these are not ideal for detecting temporal trends or for ground-truthing models because δ¹⁵N measured at the end of their deployment is affected by their initial δ¹⁵N. Correcting for the influence of the latter requires a complex model of Mytilus growth. However, if small fast-growing Mytilus were deployed, essentially all the ¹⁵N in the Mytilus then results from new growth during the deployment period, so no correction for the starting ¹⁵N signal is required. Post-larval shrimp, which had a four-fold increase in weight, showed essentially constant isotopic values, reflecting their new diets (Fry and Arnold 1982). Similarly, Trueman et al. (2005) found that following a diet switch in Atlantic salmon, the δ¹⁵N of muscle and liver reached equilibrium with their new diet after a three-fold increase in body mass, equivalent to 8 months of growth under laboratory conditions. Commercial hatcheries in Port Phillip Bay currently produce 1 mm length mussels suitable for deployment, and when deployed on mussel farms they grow to ~5–10 mm over 2 months, when their δ¹⁵N should just reflect their exposure to ¹⁵N over the period of their deployment.