Habitat mapping for native oyster restoration using hydrodynamic modelling and experimental mortality data

doi:10.21203/rs.3.rs-4725384/v1

Download PDF

Research Article

Habitat mapping for native oyster restoration using hydrodynamic modelling and experimental mortality data

https://doi.org/10.21203/rs.3.rs-4725384/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Ostrea edulis has been harvested in Galway Bay, Ireland, for centuries, but the oyster aquaculture in the bay is facing multiple threats One major concern for oyster farmers in the bay is the recurrence of heavy-rainfall, high-runoff, low-salinity events. In this work, a high-resolution hydrodynamic coastal model of Galway Bay is used together with experimental data on oyster mortality to identify suitable areas for native oyster restoration.

Here, a new coastal circulation model of Galway Bay is presented, and its performance is evaluated against in-situ temperature and salinity observations. Oysters were exposed at varying conditions of temperature and salinity in the laboratory and experimental cumulative mortality curves were determined. These experiments were used to develop an oyster mortality model that, together with the temperature and salinity predictions from the hydrodynamic model, determines the total mortality over a given period. A new Habitat Suitability Index is proposed, based on the distribution of total cumulative mortality.

oyster

model

mortality

salinity

restoration

Oyster reefs constitute important habitats that provide several ecosystem functions and services (Thomas et al. 2022), including food provision (Peterson et al. 2003; Grabowski et al. 2012) and erosion prevention (Scyphers et al. 2011; Cheong et al. 2013). Historically, the flat oyster (Ostrea edulis) was widely distributed along European coasts, and has been harvested for centuries, with oyster fishing records in Ireland documented back to the 16th century (Went 1962; Tully and Clarke 2012). An increase of fishing activity led to the overexploitation of this valuable resource (Pogoda et al. 2023) and caused the collapse of oyster populations and reef habitats throughout most of Europe during the 20th century (Lotze et al. 2005; Thurstan et al. 2013). In coastal bays and estuaries, in addition to overfishing (Smyth et al. 2009), irregular recruitment or high post-settlement mortality caused by unfavorable environmental conditions also limits stock productivity and stability. The introduction of the parasite Bonamia ostreae from the USA into Europe added a significant additional pressure and caused widespread mortality (Elston et al. 1986). Physiological stress in specimens of O. edulis over ranges of temperature and salinity have been quantified in the laboratory, and show that salinities below 19 cause a significant reduction in the scope for growth. (Hutchinson and Hawkins 1992). These conditions may occur in estuarine environments and contribute to mortality or reduced productivity.

In the context of the UN decade on Ecosystem Restoration (UNEP/FAO 2020) and the new EU Biodiversity Strategy for 2030 (EU COM 2020), there currently exist numerous initiatives to restore oyster populations. Oyster ground restoration programmes typically involve the assessment of habitat suitability for oyster growth and development in a particular area and during the different stages of the life cycle, from the spawning and larval pelagic phases to the settlement and adult benthic stage. In order to identify optimal sites for restoration, a common approach is to define a Habitat Suitability Index (HSI), taking these relevant parameters into consideration. For example, Puckett et al. (2018) developed a GIS-based HSI integrating physical, biological and logistical factors, of which salinity, larval dispersal and oxygen concentration were found to be the most critical to oyster site selection. Also, Pogoda et al. (2022) used multicriteria decision analysis (MCDA) for the restoration of O. edulis in marine protected areas in the North Sea to produce habitat suitability maps, and Stechele et al (2022) used Dynamic Energy Budget (DEB) theory to identify habitats that could support oyster growth.

This manuscript examines the specific case of Galway Bay, Ireland, where a flat oyster fishery has been exploited since ancient times but, as is the case elsewhere in Europe, has experienced a severe decline over the last number of decades (Tully and Clarke 2012). Besides overfishing, massive and widespread mortalities have occurred in this stock and have persisted and increased in the past decade (Tully et al. 2023). B. ostreae infection and exposure to unfavorable combinations of temperature and salinity are thought to be significant drivers of this mortality. Oyster mortality rates are examined in the laboratory under different temperature and salinity conditions, and this experimental data is used to build a mortality response surface M_d(T, S, Δt), where daily mortality depends on temperature, salinity and exposure time. This mortality model is then applied to the environmental conditions in the bay during a period of ten years (2012–2021), as simulated with a hydrodynamic coastal circulation model. A Habitat Suitability Index (HSI), in relation to salinity and temperature, is defined based on the total cumulative mortality after that period. The proposed methodology is especially appropriate for estuarine areas with large spatial and temporal gradients in seawater salinity and can be reproduced in other areas with similar characteristics.

Area of study

Galway Bay, on the mid-west coast of Ireland (Fig. 1), is characterized by tidal and wind-driven circulation. The focus of this study is on the inner part of the bay east of Black Head, which is shallow with maximum depths of about 30 meters. The area is influenced by a humid oceanic climate, with high annual precipitation rates and large freshwater runoff from both surface water inputs and submarine groundwater inputs from karstic limestone areas. Riverine input plays an important role in the distribution of salinity across the bay. Large spatial and temporal gradients of salinity occur, oscillating with the tidal cycle due to the interaction between the incoming salty shelf water during the flood tide, and the freshwater plumes that propagate across the bay. As such, salinity is highly dynamic, making it difficult to predict accurately at a given time and location. The effect of river plumes is more predominant in shallow waters in the southeast of the bay, where the freshwater can extend along the whole water column under the effect of tide and wind-driven mixing, affecting the benthic biota. It is precisely in these shallow waters to the southeast where the main beds of native oyster and oyster restoration activities are located, and where aquaculture of Pacific oyster also takes place. Field observations are gathered periodically in a joint effort between Cuan Beo and the Marine Institute to evaluate the status of the stocks of native oyster.

The main sources of freshwater discharge in the bay are: (a) the Corrib River (Fig. 1.c), in Galway City, which largely constitutes the most important freshwater source, although its associated freshwater plume normally leaves the bay along the northern shore, and rarely affects the south bay; (b) the Clarin and Dunkellin rivers (Fig. 1.c) which are small freshwater sources to the eastern extent of the bay, close to the farming sites; and (c) the Submarine Groundwater Discharge (SGD) flows into the bay through intertidal springs around the shoreline of Kinvara town (Fig. 1.c). A remarkable characteristic of the hydrogeology of this area is the existence of a complex karst system along the southern section of the bay adjacent to the limestone region known as the Burren. Even though it is not possible to obtain direct measurements of the intensity of this SGD flow into Kinvara Bay, a NARX model taking inputs of the past five days’ flood volume, rainfall and tidal amplitude has been developed to estimate the freshwater discharge (Basu et al. 2022). This NARX was developed on the outputs of a semi-distributed 1D/2D model of the 500 km² karst catchment with the Infoworks ICM (Innovyze) urban drainage software to simulate groundwater-surface water interaction for turloughs (ephemeral lakes) in the area, and calibrated against field data that had been collected for over 25 years (see Morrisey et al. 2020).

Hydrodynamic model

The hydrodynamic model consists of a three-level nested application of the Regional Ocean Modelling System, a split-explicit, free-surface, terrain-following, primitive equations model (Shchepetkin and McWilliams 2005). A description of each of these three nested applications is provided below, and further details of the model configuration are shown in Table 1 for clarity and reproducibility. A multi-year hindcast, starting from 2012 until present, exists for each of these models, running operationally to deliver a 3-day forecast of physical parameters of the ocean, including seawater temperature and salinity, sea surface elevation, and the barotropic and baroclinic components of circulation. The latest hindcasts and 3-day forecast are open-access through a dedicated THREDDS server at https://milas.marine.ie.

In the three nested domains described here, atmospheric forcing was derived from the ECMWF 0.1º-resolution grid and incorporated into the model using bulk fluxes. Light absorption and shortwave heating were modelled as a double exponential function (Paulson and Simpson 1977), using the parameterizations for a Jerlov IB water type (Jerlov 1976). MPDATA is used for the advection of active tracers (Smolarkiewicz and Margolin, 1998). A logarithmic bottom friction was used with a bottom roughness length of Z_ob = 1 cm.

Table 1

Overview of the three ROMS nested domains. θ_s, θ_b: surface and bottom stretching parameters, h_c: critical depth, L_ρ, M_ρ: grid size in the longitude and latitude dimensions.
	Northeast Atlantic	Connemara	Galway Bay
SW coordinates	55°56'50.6"N 30°43'04.4"W	52°57'03.2"N 10°47'54.6"W	53°06'48.2"N 9°12'43.2"W
NW coordinates	63°52'39.0"N 15°38'29.4"W	53°43'44.8"N 10°47'54.6"W	53°16'50.2"N 9°12'43.2"W
NE coordinates	51°12'42.5"N 9°28'37.9"E	53°43'44.8"N 8°53'47.4"W	53°16'50.2"N 8°52'49.4"W
SE coordinates	40°08'02.0"N 5°35'57.5"W	52°57'03.2"N 8°53'47.4"W	53°06'48.2"N 8°52'49.4"W
Grid rotation angle	-43.5º	0º	0º
Horizontal resolution	~ 2 km (Lρ = 1200, Mρ = 750) ~ 1.2 km around Ireland	~ 200 m (Lρ = 640, Mρ = 440)	~ 70 m (Lρ = 336, Mρ = 283)
No. vertical levels	40	20	8
θ_s	4	4	4
θ_b	0.85	0.85	0.85
h_c	10	4	0
Wet & Dry	No	No	Yes
Nested in	Copernicus GLOBAL_ANALYSISFORECAST_PHY_001_024	Northeast Atlantic	Connemara
Atmospheric forcing	0.1º ECMWF	0.1º ECMWF	0.1º ECMWF

The Northeast Atlantic model (Nagy et al. 2020a) is a ~ 2 km resolution domain covering a large part of the northeast Atlantic Ocean, including the Bay of Biscay, Celtic Sea, Irish Shelf, Malin Shelf, Irish Sea, Porcupine Seabight, Porcupine Bank, Rockall Trough, Rockall Bank, and further offshore areas including all of Ireland’s EEZ waters (Fig. 1.a). The model is one-way nested within the Copernicus GLOBAL_ANALYSISFORECAST_PHY_001_024, which provides open boundary forcing. Tidal elevations and barotropic velocities for ten major tidal constituents are prescribed at the open boundaries, and these are derived from the TPXO8 tides 1/30°global inverse barotropic tide model (Egbert and Erofeeva 2002). The bathymetry was obtained from a combination of multiple sources, including the GEBCO_08 grid (Weatherall et al. 2015), the Irish National Seabed Survey (INSS), and the Integrated Mapping for the Sustainable Development of Ireland’s Marine Resource (INFOMAR).

The Connemara model (Nagy et al. 2020b) is a ~ 200 m resolution domain covering the mid-west coastal waters of Ireland, extending from the Co. Clare coast to the south, up to Killary Harbour to the north, enclosing the Aran Islands and Galway Bay (Fig. 1.b). Open boundary conditions are received from the Northeast Atlantic model. The Connemara model provides open boundary forcing to the Galway Bay model, a ~ 70 m resolution grid covering inner Galway Bay east of Black Head from 53º06’48.2”N, 9º12’43.2”W to 53º16’50.1”N, 8º52’49.6”W (Fig. 1.c). At the open boundaries, clamped boundary conditions have been imposed for 3-D momentum and tracers, whilst a combination of Chapman and Flather conditions have been applied for the free-surface and the barotropic velocity. The maximum water column depth in the Galway Bay model is less than 30 m and the bathymetry was entirely derived from an INFOMAR 5 m resolution LiDAR product. Wetting and drying (Warner et al. 2013) is enabled in the Galway Bay model for an appropriate representation of the intertidal areas, and a critical depth of 0.5 m was used (the water column thickness, below which a cell is considered to be dry). Daily river flows for the Corrib, Clarin and Dunkellin are operationally prescribed into the model using actual flow measurements or water level recordings from the Office of Public Works (OPW), publicly accessible at https://waterlevel.ie. Water levels are converted to flows using a rating curve: a functional relationship between the flow and the water height at a gauging station. For the Kinvara SGD, freshwater flows provided by the NARX model developed by Basu et al. (2022) were used. River temperature observations from the Corrib River were obtained from the Office of Public Works. Due to the lack of temperature measurements from the other freshwater sources, the same freshwater temperature was assumed for all of them. Freshwater flows and temperatures are shown (Fig. 2).

Hydrodynamic model validation

In order to assess the Galway Bay model’s ability to reproduce the temperature and salinity conditions in the bay, CTD surveys were carried out on 18 May 2021, 18 August 2021, 4 November 2021, 23 March 2022, 21 July 2022, 8 January 2023 and 10 July 2023. CTD casts were sampled with a YSI CastAway CTD, except for the last survey, where an RBR Brevio CTD was used. A total of 180 CTD profiles were sampled in the bay. Temperature and salinity observations were compared with Galway Bay model predictions. In addition, a near-surface temperature and salinity time series from an NKE sensor deployed at Trout Buoy (53°15'10.2"N 9°01'05.2"W) in spring 2022 was compared against the Galway Bay model temperature and salinity fields. The Trout Buoy series spans from 24 March to 19 April 2022. Finally, temperature and salinity observations from the Killeenaran farming site (53º11´51.6"N 08º56´30.5"W) were also compared against the Galway Bay model. These observations consist of: (1) a continuous seawater temperature series from a permanent mooring, and (2) weekly temperature and salinity observations from a CO 310 hand-held sensor. The locations of the sampling sites used for model validation are shown (Fig. 1.c).

Oyster mortality model

European flat oysters Ostrea edulis were exposed to 25 (5 x 5) different conditions of temperature and salinity, with temperatures of 5, 10, 15, 20, 25 ºC, and salinities of 5, 11, 17, 23, 29 g L^− 1 in the laboratory. Oysters were sourced from Galway Bay in November 2020 and transferred to a flow-through seawater holding tank at temperature and salinity reflective of ambient conditions (temperature ranging from 12 to15 ºC, salinity above 30 g L^− 1). They were held there for seven days prior to distribution to the experimental conditions. Oysters (30 individuals adding up to approximately 1.5 kg in weight to each container) were transferred from the holding tank to each of the experimental containers while the containers were still at ambient temperature and salinity. Temperatures were then gradually reduced or increased and salinities were gradually reduced over a period of five days until the experimental temperature and salinities were reached in order to avoid temperature shock. Oysters were fed with phytoplankton cultures (Isochrysis and Chaetocerus). All containers were aerated using pressurized air diffused through airstones. Pseudofaeces and detritus were removed from the tanks each day by suction and filtering onto a 63-µm mesh. Temperatures and salinities were monitored daily. Oyster mortality was monitored daily for 30 days. Dead (gaping) oysters were removed from the experimental containers daily. The mortality data were used to determine the experimental O. edulis cumulative mortality curves as a function of exposure time to combinations of salinity and temperature.

The experimental cumulative mortality curves were used to design an oyster mortality model. The idea was to define a function M_d(T, S¸ Δt): the daily oyster mortality, as a function of temperature, salinity and exposure time. This function can then be applied to any temperature and salinity time series (either observed or modelled) to determine the total mortality at the end of the time period under consideration. If N_t is the number of living individuals on day t, and M_t is the daily mortality on day t, then N_t+1 = N_t - N_t * M_t is the number of survivors on day t + 1. First, the 25 experimental cumulative mortality curves are modelled as the best-fitting logistic functions:

$\:\varvec{C}\varvec{u}\varvec{m}.\:\varvec{M}\varvec{o}\varvec{r}\varvec{t}.\left(\varDelta\:t\right)=\frac{1}{1+{\varvec{e}}^{-\varvec{k}(\varDelta\:\varvec{t}-{\varDelta\:\varvec{t}}_{0.5})}}$ (Eq. 1)

where Δt_0.5 is the exposure time corresponding to a cumulative mortality rate of 50% of the individuals, and k is the logistic growth rate or steepness of the curve. The daily rate of change of these logistic curves is the daily mortality. These daily mortality values are then interpolated to the entire temperature and salinity ranges naturally occurring in the bay, from 5 ºC to 25 ºC and from 5 to 29 g L^− 1. For conditions outside these ranges, nearest-neighbour extrapolation was used to determine an appropriate value of daily mortality. Finally, it is necessary to define a rule for the exposure time Δt, since the laboratory conditions were kept constant, but this is not the case in the marine environment. Temperature and salinity conditions that resulted in a cumulative mortality < 10% in the laboratory after 30 days were considered non-lethal, whereas when the cumulative mortality exceeded 10% the conditions were considered lethal. Following this approach, it was found that non-lethal conditions occur as long as T < 22.5 ºC and S > 20 g L^− 1. In a daily-averaged T-S series, when the conditions enter the lethal domain, the exposure time is increased by 1 day. As soon as the conditions enter the non-lethal domain, the exposure time is restarted to 0 days.

Hydrodynamic model validation

Model temperature and salinity predictions are compared against observations in Fig. 3 for the NKE sensor at Trout Buoy (Fig. 3.a), the CPT logger and CO 310 sensors at the Killeenaran farming site (Fig. 3.b) and the CTD observations from the different surveys in Galway Bay (Fig. 3.c), whereas Estimated Accuracy Numbers (EANs) are provided in Table 2. These EANs are the Mean Bias Error (MBE), Root Mean Squared Difference (RMSD) and Correlation Coefficient (CORR). Most notable is how the model successfully captures the short-scale variability in the NKE temperature series (Fig. 3.a, top), resulting in a high correlation (R = 0.924) between both series. The model not only provides an accurate representation of the natural variability of the series, but also modelled temperature values are close to those recorded by the instrument, with a Root Mean Squared Difference (RMSD) of just 0.246 ºC. The model is slightly negatively biased, with a Mean Bias Error (MBE) of -0.072 ºC.

Table 2

Hydrodynamic model validation. MBE: Mean Bias Error. RMSD: Root Mean Squared Difference. CORR: Correlation Coefficient. N: Number of data points.
	Temperature			Salinity
	MBE (ºC)	RMSD (ºC)	CORR	MBE	RMSD	CORR	N
Trout Buoy	-0.072	0.246	0.924	0.568	1.270	0.757	638
Killeenaran CPT logger	0.092	0.720	0.984	-	-	-	33471
Killeenaran CO 310	-0.254	0.800	0.981	1.310	2.505	0.904	220
CTDs (All)	0.093	0.488	0.992	0.509	1.013	0.846	6087

Due to the interaction between incoming oceanic water and freshwater discharge, the observed salinity series (Fig. 3.a, bottom) presents much larger fluctuations and, as such, is generally harder to predict everywhere in the bay. The site at Trout Buoy is particularly challenging due to the proximity of the Corrib River. Yet the model still provides a precise representation of the natural variability and the short-term trends in salinity that occur in the series. For example, the steady decline in seawater salinity from ~ 33 on 31 March to ~ 28 on 6 April, followed by a rapid recovery back to similar salinity values on 8 April. The correlation is R = 0.757 and the RMSD is 1.270. The model tends to overpredict salinity with an MBE of 0.568, but this value is small when considering the large range of fluctuations of salinity in the bay.

Figure 3.b shows the comparison between the vertically-averaged Galway Bay model seawater temperature and salinity predictions and actual observations at Killeenaran (53.198ºN, 8.942ºW, see Fig. 1.c). Killeenaran is an important native oyster farming site, located close to St. George’s bed, where native oyster restoration efforts are currently in place. Therefore, this validation exercise is of great importance to assess the suitability of the hydrodynamic model for habitat mapping for native oyster restoration efforts. It is also a challenging area to assess model performance due to the proximity to the freshwater sources of the Clarin and Dunkellin rivers, where large fluctuations of salinity occur on a regular basis.

There is an excellent match between the Galway Bay model seawater temperature and the CPT logger recordings (Fig. 3.b, top). The model reproduces not only the seasonal cycle but also the short-scale variability with great accuracy, resulting in a correlation coefficient of R = 0.984. When compared to Trout Buoy, the error is larger but still small (RMSD = 0.720 ºC), and here the model presents a slight positive bias (MBE = 0.092 ºC). The match between the model and the weekly CO 310 measurements is also excellent, with a correlation coefficient of R = 0.981 and RMSD = 0.800 ºC. Here, the MBE is larger, with a negative bias of -0.254 ºC, which can be partially explained due to the extreme high values (~ 20 ºC) recorded by the sensor in summer 2020, which were captured neither by the model nor the CPT logger. It is important to mention that the Killeenaran area dries out during the low tide, and the water may occasionally be subject to extreme temperatures when a thin layer of water is exposed to solar radiation, but these temperatures are rare.

The seawater salinity at Killeenaran, as recorded by the CO 310 weekly measurements, are compared against the Galway Bay model (Fig. 3.b, bottom). It is interesting to note how the model successfully reproduces most of the fluctuations and low salinity events detected by the instrument and captures the different salinity regimes along the annual cycle, including higher (> 30), more steady salinities during summer and rapidly fluctuating winter salinities exhibiting frequent low salinity episodes. This results in a high correlation coefficient (R = 0.904). An accurate representation of low salinity events at an oyster farming site is of great importance for the reliability of the oyster mortality estimations derived from this model, since it is during these episodes when larger mortality rates can occur. On the other hand, this very shallow area is exposed to large, tidal fluctuations of salinity that sometimes are not properly reproduced by the model, and this explains the larger error magnitude at Killeenaran (RMSD = 2.505, MBE = 1.310) when compared to a deeper area like Trout Buoy.

Finally, the CTD vs. Galway Bay model comparison (Fig. 3.c) shows an excellent performance of the model in reproducing the naturally occurring seawater temperature throughout the bay, with an RMSD of 0.488 ºC, a positive MBE of 0.093 ºC and a correlation coefficient of 0.992. Again, salinity poses a bigger challenge due to the high short-scale variability which results from the strong tidal circulation, with freshwater plumes from rivers being pushed onshore during the flood tide, and then extending offshore during the ebb tide. The fit between the Galway Bay model and CTD observations is better when the salinity remains high and relatively constant during drier periods with low freshwater discharge, whereas low salinity episodes are more difficult to reproduce accurately. Nevertheless, the Galway Bay model performance to represent these events is high, resulting in an overall correlation coefficient of 0.846 and an RMSD of 1.013. The salinity is slightly overpredicted, with an MBE of 0.509, very small compared to the large variance of the salinity within the bay.

Oyster mortality model

Oyster mortality rates are elevated at low salinities and high temperatures. Seawater temperatures in Galway Bay rarely exceed 20 ºC, and survival is high if temperature is less than 20 ºC and salinity is above 20 g L^− 1. The mortality starts to increase at salinity less than 20 g L^− 1 and raises dramatically at salinities between 11 g L^− 1 and 17 g L^− 1. When the salinity is below 11 g L^− 1 and temperature is 5 ºC, oysters show first signs of mortality after 13–14 days and reach 100% mortality after 20–22 days. For salinities of 17 g L^− 1, mortality does not occur until 20 days from the beginning of the experiment and 90% mortality was observed after 38 days (Fig. 4.a).

At temperatures of 10 ºC (Fig. 4.b), which are common in the bay during wintertime, and salinities below 11 g L^− 1, mortality appears earlier and spreads faster. After 12 days, 34% of the individuals exposed to salinities of 5 g L^− 1 are affected, and 10% of the individuals under 11 g L^− 1 of salinity are dead. Full mortality occurs after 19 days when salinity is 5 g L^− 1 and after 22 days when salinity is 11 g L^− 1.

When water temperature is 15 ºC (Fig. 4.c), the first signs of mortality appear after 3–7 days for salinities below 17 g L^− 1, showing a much lower tolerance to freshwater as temperature increases. A mortality of 90% occurs after 10 days at salinity of 5 g L^− 1. Almost full mortality occurs quickly at 20 ºC (Fig. 4.d) at salinities below 11 g L^− 1. Daily mortality increases dramatically when the salinity drops from 17 g L^− 1 to 11 g L^− 1. Finally, at 25 ºC (Fig. 4.e), even salinities of 23 g L^− 1 are no longer tolerated, causing a drastic mortality rise after only eight days.

The experimental cumulative mortality functions (Fig. 4.a-e) were modelled as the best-fitting logistic curves following Eq. 1. The response surface was interpolated to cover the range of temperature and salinity occurring in the bay and an example is shown (Fig. 4.f). The rate of change of the cumulative mortality curve is the daily mortality rate M_d(T, S¸ Δt). Notice the increase in mortality rate caused by a subtle drop in salinity from 15.6 g L^− 1 to 14.8 g L^− 1.

The resulting response surface M_d(T, S¸ Δt), shown for Δt = 3, 6, 9 and 12 days (Fig. 4.g-j), agrees with the experimental oyster mortality data. High-temperature, low-salinity conditions cause elevated mortalities (M_d > 0.2) even for short exposure times (Δt = 3 days, Fig. 4.g). After longer exposure times (Fig. 4.h-j), the mortality gradually spreads to the rest of the T-S domain. After Δt = 12 days (Fig. 4.j), the mortality is (or has already been) high for salinities below 15 regardless of the water temperature. Notice that mortality tends to zero for very long exposure times, when most of the mortality has already occurred (see Fig. 4.f).

Application of the oyster mortality model

3.1.1 Killeenaran farming site

An example of the application of the oyster mortality model is shown for Killeenaran (Fig. 5). Killeenaran is an interesting site to test the model due to both: (a) It is an important oyster farming site located close to St. George’s bed, where native oyster restoration initiatives are ongoing, and (b) It is a shallow area close to freshwater sources: mainly the Clarin and Dunkellin Rivers. As a result, during episodes of high river runoff, the seabed salinity can drop to very low values (S < 20 g L^-1), especially during the low tide. These ranges of salinity are particularly sensitive for the native oyster (Fig. 4).

The behavior of the oyster mortality model is shown (Fig. 5). The near-seabed, daily-averaged temperature (Fig. 5.a) and salinity (Fig. 5.b) series from the Galway Bay model are provided as input to the model. Based on the fitted M_d(T, S¸ Δt) response surface, as estimation of the daily mortality is obtained (Fig. 5.c). It is easy to identify the events characterized by low salinities and higher mortalities. It is this type of events that stakeholders in the oyster farming sector are mostly interested in identifying and forecasting. In particular, the high mortality event in early 2020 is quite remarkable. According to the Galway Bay model, this period was characterized by unusually low daily-averaged salinities (S < 25 g L^-1) that persisted over several days. The prolonged exposure time is the most important feature when compared to other events in the series with similar salinities but shorter in duration. As a result, the daily mortality estimation peaks to 5.4% in winter 202. It is also possible to estimate the total cumulative mortality over time (Fig. 5.d), with annual mortalities of 37.9% in 2020, 5.7% in 2015 and < 5% in other years. This again highlights the impact of the 2020 event.

Actual estimations of annual mortality rates based on field observations are much higher than the estimations presented here. Field estimations approximate to an ~ 80% mortality rate per year. Therefore, except for unusual events like the one in 2020, the contribution of poor temperature and salinity conditions to total oyster mortality would be very small (< 5%) most of the time. It is believed that Bonamia ostreae infection is the main driver of mortality for O. edulis population in the bay. On the other hand, even though no exceptional native oyster mortality was observed at Killeenaran in 2020, there were field observations from the nearby area of Island Eddy that suggest widespread mortality on the local wild bivalve populations. These field observations were taken on 8th April 2020, just one month after the low salinity event in 2020, and revealed lots of recently opened bivalve shell material, mainly from dead clams and razor shell. It is possible that this high mortality could have been driven by the high freshwater runoff and low salinity conditions few weeks earlier.

3.1.2 Habitat suitability mapping for Galway Bay

The spatial distribution of the frequency and recurrence of high mortality events can be used to produce risk maps to help in oyster restoration initiatives. The oyster mortality model was applied to the 10-year (2012–2021) seabed temperature and salinity series from the Galway Bay model to obtain a map of the estimated 10-year total mortality throughout Galway Bay (Fig. 6). The distribution of cumulative mortality is the result of the interaction of freshwater sources with shallow bathymetry. In other words, the areas with higher cumulative mortality are those that are both close to freshwater sources and shallow enough so that the freshwater affects the seabed. Low salinity at the seabed results in more frequent mortality events and overall higher cumulative mortality. This total cumulative mortality can be considered as a Habitat Suitability Index (HSI), where the local time series of seabed temperature and salinity determine the magnitude of the HSI. In this case, there is an inverse relationship between the HSI and the suitability for oyster restoration, with lower values of the HSI (i.e. lower cumulative mortality) meaning higher suitability.

A hydrodynamic model for Galway Bay was developed and combined with experimental data on oyster mortality over ranges of salinity and temperature to assess habitat suitability for O. edulis in the bay. This included a wetting and drying scheme for the representation of the dynamics in intertidal areas, and near-real time operationalization of river flows. The simulated temperature and salinity fields were extensively validated against CTD and mooring observations, and proved to be reliable even in challenging areas near freshwater sources, where large fluctuations of salinity occur and where a large part of the oyster farming activity takes place.

A new methodology is presented to predict O. edulis mortality due to temperature and salinity changes by combining hydrodynamic modelling and experimental oyster mortality data. Other factors affecting oyster health that could be incorporated into a more sophisticated model are sub-diurnal salinity fluctuations, food availability, sediment resuspension (Wilber and Clarke 2010), and prevalence of Bonamia ostreae infection. Increased concentrations of suspended sediments reduce filtration and feeding rates in oysters (Loosanoff and Tommers 1948). Larval settlement occurs on hard substrates and adult oyster shell; larger sedimentation rates and the presence of even a thin layer of sediment can inhibit larval settlement and recruitment (Galtsoff 1964). B. ostreae is a parasitic protozoan that causes oyster mortality in tidal and subtidal areas during the second and third years of growth (Baud et al. 1997).

Another limitation of the present study is that only the adult phase has been considered, whereas the success during the larval stages, from spawning to settlement and recruitment must be taken into account for a comprehensive research on habitat mapping. Aspects such as larvae mortality in brackish waters (Eierman and Hare 2013), the influence of both food availability and seawater temperature on larvae growth rates (Rico-Villa et al. 2009), and the connectivity between spawning grounds and suitable settlement areas can be inspected using particle-tracking techniques that include vertical swimming behaviours, which have an influence on the dispersal of oyster larvae (North et al. 2008).

Other works on habitat mapping and oyster ground restoration have developed more comprehensive Habitat Suitability Indexes (HSI) considering a wider range of factors, not just physical, such as salinity, but also biological factors, such as larval dispersal, logistical factors and permitting factors, such as presence of seagrasses or nursery areas for finfish and shellfish populations, where it is not allowed to develop oyster restoration projects (Puckett et al. 2018). By including multiple factors in the definition of a HSI it is possible to investigate the linkages between landscape ecology, larval connectivity, and spatial planning, providing a robust and unified approach to oyster ground restoration (Puckett et al. 2018). The multi-criteria decision analysis (MCDA) approach by Pogoda et al. (2023) also included multiple exclusion and suitability factors, including physical (temperature, salinity, shear stress, etc.), biogeochemical (chlorophyll concentration, turbidity, dissolved oxygen, etc.), and exclusion areas such as military and munition dumping grounds, wind farms, wrecks and offshore cables and pipelines.

The total cumulative mortality HSI proposed in this study displays a distinct distribution in the bay (Fig. 6). This HSI is entirely based on the hydrographic characteristics of the seawater: temperature and salinity. This HSI can be useful in areas with large gradients of seawater salinity to evaluate the potential impact of freshwater on mortality, such as in shallow water estuaries with large freshwater discharges. This is the case of Galway Bay, where interaction with stakeholders from the oyster farming sector has revealed a great concern over the impact that low salinity episodes have on oyster health and mortality. The presented oyster mortality model and derived HSI are entirely based on observed, experimental oyster mortality data across varying conditions. Another interesting feature of this HSI is the possibility of identifying events where extreme and persistent low salinity conditions may have had an impact on the bivalve populations, such as the low salinity event in winter 2020 (Fig. 5).

The objective of this research is to present a new methodology for the assessment of suitable areas for native oyster restoration efforts, based on the combination of hydrodynamic modelling and experimental oyster mortality data. This methodology was applied to Galway Bay. First, a ROMS hydrodynamic model for Galway Bay was developed and operationalized. This model is the result of a 3-step downscaling from the Copernicus GLOBAL_ANALYSISFORECAST_PHY_001_024: the Northeast Atlantic, Connemara and Galway Bay models (Table 1). Some of the most important features of this Galway Bay model are: (a) the implementation of a wetting and drying scheme, which is necessary for an appropriate representation of the dynamics in intertidal areas, and (b) the operationalization of near-real time river flows, as derived from water level gauges, for an accurate modelling of the salinity distribution in the bay. Both features are very important for an adequate representation of the processes affecting oyster health and development. The Galway Bay model temperature and salinity distributions were thoroughly validated against CTD profiles and fixed mooring series, demonstrating a high level of skill (Fig. 3).

The long-term seabed temperature and salinity distributions from this hydrodynamic numerical model have been provided as input to an oyster mortality model, which consists of a response surface M_d(T, S¸ Δt): the daily mortality rate for some given seawater temperature, salinity and exposure time. The oyster mortality model is purely based on experimental oyster mortality data, where individuals were exposed to varying temperature and salinity conditions in the laboratory, and cumulative mortalities were examined. The application of this mortality model on a long-term (2012–2021) hydrodynamic model hindcast has resulted in a map showing the expected distribution of cumulative mortality throughout the bay (Fig. 6). This map can help native restoration efforts to decide which areas are and are not suitable for oyster growth and development.

Funding Declaration

This work has been funded by the European Union’s Horizon 2020 FORCOAST project under grant agreement No. 870465.

Competing Interests:

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.

Author Contribution

All authors contributed to the study conception and design. Hydrodynamic model development, validation and operationalization were performed by D.P., T.D. and K.L. Research on oyster mortality under varying environmental conditions was developed by O.T. and G.O’H. Sampling of temperature and salinity data for hydrodynamic model validation was performed by D.P., E.D., D.K., C.O’D. and G.N. Hydrological model to simulate the underground freshwater discharge developed by P.M., B.B. and L.G. The first draft of the manuscript was written by D.P. and all authors commented on previous versions of the manuscript. All authors read, reviewed and approved the final manuscript

Acknowledgement

This work has been funded by the European Union’s Horizon 2020 FORCOAST project under grant agreement No. 870465.

Data Availability

Hydrodynamic coastal model data from the Galway Bay model is open-access and freely available at milas.marine.ie for the last month. Older, archived data can be accessed under request to the Ocean Modelling Department of the Marine Institute, Ireland.

Basu B, Morrissey P, Gill LW (2022) Application of Nonlinear Time Series and Machine Learning Algorithms for Forecasting Groundwater Flooding in a Lowland Karst Area. Water Resour Res 58. https://doi.org/10.1029/2021WR029576. e2021WR029576
Baud JP, Gérard A, Naciri-Graven Y (1997) Comparative growth and mortality of Bonamia ostreae -resistant and wild flat oysters, Ostrea edulis, in an intensive system. I. First year of experiment. Mar Biol 130:71–79. https://doi.org/10.1007/s002270050226
Cheong SM, Silliman B, Wong PP, Van Wesenbeeck B, Kim CK, Guannel G (2013) Coastal adaptation with ecological engineering. Nat Clim Change 3:787–791. https://doi.org/10.1038/nclimate1854
Egbert GD, Erofeeva SY (2002) Efficient Inverse Modeling of Barotropic Ocean Tides. J Atmos Ocean Tech 19:183–204. https://doi.org/10.1175/1520-0426(2002)019<0183:EIMOBO>2.0.CO;2
Eierman LE, Hare MP (2013) Survival of oyster larvae in different salinities depends on source population within an estuary. J Exp Mar Biol Ecol 449:61–68. https://doi.org/10.1016/j.jembe.2013.08.015
Elston RA, Farley CA, Kent ML (1986) Occurrence and significance of Bonamiasis in European flat oysters Ostrea edulis in North America. Dis Aquat Organ 2:49–54. https://dx.doi.org/10.3354/dao002049
EU COM (2020) BiodivERsA the network promoting European Research on Biodiversity, Ecosystem Services and Nature-based Solutions: Eur Commiss. https://www.biodiversa.org/
Galtsoff PS (1964) The American Oyster Crassostrea virginica Gmelin. Fishery Bulletin, v. 64. Limnol Oceanogr 11:312–312 United States Government Printing Office, Washington, D. C. iii + 480 p, $2.75
Grabowski JH, Brumbaugh RD, Conrad RF, Keeler AG, Opaluch JJ, Peterson CH, Piehler MF, Powers SP, Smyth AR (2012) Economic Valuation of Ecosystem Services Provided by Oyster Reefs. Bioscience 62:900–909. https://doi.org/10.1525/bio.2012.62.10.10
Hutchinson S, Hawkins LE (1992) Quantification of the physiological responses of the European flat oyster Ostrea edulis L. to temperature and salinity. J Mollus Stud 58:215–226. https://doi.org/10.1093/mollus/58.2.215
Jerlov NG (1976) Marine Optics. Elsevier, New York, p 231
Loosanoff VL, Tommers FD (1948) Effect of Suspended Silt and Other Substances on Rate of Feeding of Oysters. Science 107:69–70. https://doi.org/10.1126/science.107.2768.69
Lotze HK, Reise K, Worm B, Van Beusekom J, Busch M, Ehlers A, Heinrich D, Hoffmann RC, Holm P, Jensen C, Knottnerus OS, Langhanki N, Prummel W, Vollmer M, Wolff WJ (2005) Human transformations of the Wadden Sea ecosystem through time: a synthesis. Helgol Mar Res 59:84–95. https://doi.org/10.1007/s10152-004-0209-z
Morrisey P, McCormack T, Naughton O, Johnston PM, Gill LW (2020) Modelling groundwater flooding in a lowland karst catchment. J Hydrol 580:124361. https://dx.doi.org/10.1016/j.jhydrol.2019.124361
Nagy H, Lyons K, Nolan G, Cure M, Dabrowski T (2020a) A Regional Operational Model for the North East Atlantic: Model Configuration and Validation. J Mar Sci Eng 8:673. https://doi.org/10.3390/jmse8090673
Nagy H, Lyons K, Dabrowski T (2020b) A Regional Operational and Storm Surge Model for the Galway Bay: Model Configuration and Validation. Ocean Sciences Meeting, San Diego, CA USA
North E, Schlag Z, Hood R, Li M, Zhong L, Gross T, Kennedy V (2008) Vertical swimming behavior influences the dispersal of simulated oyster larvae in a coupled particle-tracking and hydrodynamic model of Chesapeake Bay. Mar Ecol Prog Ser 359:99–115. https://doi.org/10.3354/meps07317
Paulson CA, Simpson JJ (1977) Irradiance Measurements in the Upper Ocean. J Phys Oceanogr 7:952–956. https://doi.org/10.1175/1520-0485(1977)007<0952:IMITUO>2.0.CO;2
Peterson CH, Grabowski JH, Powers SP (2003) Estimated enhancement of fish production resulting from restoring oyster reef habitat: quantitative valuation. Mar Ecol Prog Ser, 264:249–264. jstor.org/stable/24867515
Pogoda B, Hausen T, Rothe M, Bakker F, Hauser S, Colsoul B, Dureuil M, Krause J, Heinicke K, Pusch C, Eisenbarth S, Kreutle A, Peter C, Pesch R (2023) Come, tell me how you live: Habitat suitability analysis for Ostrea edulis restoration. Aquat Conserv 33:678–695. https://doi.org/10.1002/aqc.3928
Puckett BJ, Theuerkauf SJ, Eggleston DB, Guajardo R, Hardy C, Gao J, Luettich RA (2018) Integrating Larval Dispersal, Permitting, and Logistical Factors Within a Validated Habitat Suitability Index for Oyster Restoration. Front Mar Sci 5:76. https://doi.org/10.3389/fmars.2018.00076
Rico-Villa B, Pouvreau S, Robert R (2009) Influence of food density and temperature on ingestion, growth and settlement of Pacific oyster larvae, Crassostrea gigas. Aquaculture 287:395–401. https://doi.org/10.1016/j.aquaculture.2008.10.054
Scyphers SB, Powers SP, Heck KL, Byron D (2011) Oyster Reefs as Natural Breakwaters Mitigate Shoreline Loss and Facilitate Fisheries. PLoS ONE 6:e22396. https://doi.org/10.1371/journal.pone.0022396
Shchepetkin AF, McWilliams JC (2005) The regional oceanic modeling system (ROMS): a split-explicit, free-surface, topography-following-coordinate oceanic model. Ocean Model 9:347–404. https://doi.org/10.1016/j.ocemod.2004.08.002
Smolarkiewicz PK, Margolin LG (1998) MPDATA: A Finite-Difference Solver for Geophysical Flows. J Comput Phys 140:459–480
Stechele B, Maar M, Wijsman J, Van der Zande D, Degraer S, Bossier P, Nevejan N (2022) Comparing life history traits and tolerance to changing environments of two oyster species (Ostrea edulis and Crassostrea gigas) through Dynamic Energy Budget theory. Conserv Physiol 10(1):coac034. https://dx.doi.org/10.1093/conphys/coac034
Thomas S, Collins K, Hauton C, Jensen A (2022) A Review of the Ecosystem Services Provided by the Native Oyster (Ostrea edulis): Implications for Restoration. IOP Conf Ser: Mater Sci Eng 1245:012010. https://doi.org/10.1088/1757-899X/1245/1/012010
Thurstan RH, Hawkins JP, Raby L, Roberts CM (2013) Oyster (Ostrea edulis) extirpation and ecosystem transformation in the Firth of Forth, Scotland. J Nat Conserv 21:253–261. https://doi.org/10.1016/j.jnc.2013.01.004
Tully O, Clarke S (2012) The status and management of oyster (Ostrea edulis) in Ireland. Ir Fisheries Investigations No. 24, Marine Institute. hdl.handle.net/10793/828
Tully O, White E, Kelly D, O’Halloran G, Chopin N, Daly P (2023) Status and restoration of native oysters in Galway Bay 2018–2023: Final report January 2023 to the Marine Institute. EMFF 2014–2020 Marine Institute Report Series, Marine Institute
UNEP/FAO (2020) The UN Decade on Ecosystem Restoration 2021–2030. https://wedocs.unep.org/bitstream/handle/20.500. 11822/30919/UNDecade.pdf?sequence = 11
Warner JC, Defne Z, Haas K, Arango HG (2013) A wetting and drying scheme for ROMS. Comput Geosci 58:54–61. https://doi.org/10.1016/j.cageo.2013.05.004
Weatherall P, Marks KM, Jakobsson M, Schmitt T, Tani S, Arndt JE, Rovere M, Chayes D, Ferrini V, Wigley R (2015) A new digital bathymetric model of the world’s oceans. Earth Space Sci 2:331–345. https://doi.org/10.1002/2015EA000107
Went AJ (1962) Historical notes on the oyster fisheries of Ireland. P Roy Irish Acad C, 62:195–223. jstor.org/stable/25505107
Wilber D, Clarke D (2010) Dredging activities and the potential impacts of sediment resuspension and sedimentation on oyster reefs. P W Dredg Assoc Tech Conf, June 6–9, 2010, San Juan, Puerto Rico, USA (2010), pp. 61–69

No competing interests reported.

Download PDF

Editor assigned by journal
12 Jul, 2024
Submission checks completed at journal
12 Jul, 2024
First submitted to journal
11 Jul, 2024

You are reading this latest preprint version

Habitat mapping for native oyster restoration using hydrodynamic modelling and experimental mortality data

Status:

Version 1

Abstract

Figures

1 Introduction

2 Methods

3 Results

3.1.1 Killeenaran farming site

3.1.2 Habitat suitability mapping for Galway Bay

4 Discussion

5 Conclusions

Declarations

Funding Declaration

Author Contribution

Acknowledgement

Data Availability

References

Additional Declarations

Status:

Version 1