Modelling pollution dispersal around Solomon Islands and Vanuatu

To assess potential dispersion of pollutants around Honiara, Solomon Islands, and Port Vila, Vanuatu, 3D ocean circulation models were developed using Telemac-3D. A series of scenarios then explore the vulnerability of the system and test potential control measures. Results show that high coastal concentrations are most likely during the wet season, with increased volumes of discharge as well as favourable wind speed and direction. Buoyant plumes ﬂow along the coastline, and high concentrations build up in enclosed bays. Control measures tested focus on consolidating existing outﬂows at depth oﬀ-shore. This results in an overall reduction of surface concentrations along the coastline. However, the reduction is dependent on the depth, oﬀ-shore positioning, and volume of outﬂow. With increased concentrations then found at depth, the subsequent impact on oﬀ-shore and benthic ecosystems would also need to be considered.


Introduction
The South Pacific contains many small island developing states (SIDS). SIDS face unique challenges in their development associated with their island locations, with populations dispersed across small islands, with limited land resources. They are often remote from neighbouring countries, particularly in the vast Pacific Ocean. These island nations are then dependent on the surrounding ocean for many reasons, and are also sensitive to changes, e.g. in sea level rise, fish stocks, or water quality (CMEP, 2018). Water quality issues have become increasingly more important as population, tourism and coastal infrastructure grow and expand (Morrison, 1999).

The Commonwealth Marine Economies Programme (CMEP) is working with SIDS across the South
Pacific, to assess the current environmental status of their oceans, and advise on how they can manage sustainable development through the future (McManus et al., 2019). In this study, we particularly consider pollution dispersal, with its potential impacts on water quality around the Solomon Islands and Vanuatu.
These two nations are formed of archipelagos in the southwest Pacific. However, the focus here is on their main centres of population. For Solomon Islands, the capital of Honiara is found on Guadalcanal; for Vanuatu, the capital Port Vila is found on Efate. As populations grow rapidly within limited coastal areas, this can place significant stress on local infrastructure, and hence the surrounding environment.
With limited observations to date in these coastal regions, a major aim of the CMEP has been to gather data. However, with observations still sparse in both space and time, hydrodynamical modelling can be a vital tool for understanding conditions throughout the region. As part of the CMEP water quality assessment, the aim of this hydrodynamic modelling project has been to assess the likely extent of pollutant dispersion, from Honiara and Port Vila, and test potential control mechanisms. Ocean models can be used to assess dispersion from different locations, and under different oceanographic conditions.
The models developed here have sufficient resolution to provide detail along the local coastline (with grid spacing ≤ 100 m) and therefore represent circulation around these two locations, but also cover a wide enough extent to include impact of larger-scale ocean currents. The results here demonstrate the capability of model simulations to show how concentrated pollutants released directly through coastal outfall pipes, or into rivers that outflow to the ocean, can disperse or accumulate depending upon physical conditions.

Geography and physical setting
The Solomon Islands consist of an archipelago of 6 major, and over 900 smaller islands, formed of both volcanic islands and coral atolls. The archipelago is found along the western edge of the Pacific Ocean, extending between ∼ 6 − 11 o S. The administrative capital, Honiara, is situated on the northern coast of one of the larger islands, Guadalcanal (Figure 1a).
The topography around the Solomon Islands is complex, separating the relatively deep ocean floors of the Equatorial Pacific and Solomon Sea, to the northeast and southwest, respectively. To the southwest of the islands, deep ocean trenches are found along the boundary of the Pacific and Australian Plates (depths > 5000 m; Figure 1). The islands themselves are also separated by deep channels and canyons, with lateral shelf gradients between channels and the coastal zone of up to 13%.
Vanuatu, also in the South Pacific, is formed from an archipelago of over 80 islands. The chain of islands is of volcanic origin, lying over the subduction zone between the Australian and Pacific plates. The capital of Vanuatu is Port Vila, situated on the island of Efate (Figure 1b).
The Solomon Islands lie within the western Pacific warm pool (e.g. Ganachaud et al., 2014), and therefore have an ocean-equatorial climate, with average temperatures around 27 • C and high humidity. While there is little seasonality in temperature, there are pronounced changes in wind stress and rainfall throughout the year. June to August is a cooler, drier period, with prevailing south-easterly winds (e.g. Figure 2). Weaker north-westerly winds from November to April bring frequent rainfall and occasional cyclones. Climate variability, in terms of both wind stress and rainfall, will be influenced by the South Pacific Convergence Zone (SPCZ), and the El Nino Southern Oscillation (ENSO) (e.g. Ganachaud et al., 2014). Vanuatu has some similar seasonality to the Solomon Islands, but being further south, has a cooler and less humid climate overall. There is a seasonal change in wind stress, leading to alternating wet and dry seasons during the year. Weaker winds typically occur during December to March, compared with stronger south-easterlies during May-September ( Figure 2). These winds are associated with the hot and wet, and cool and dry seasons, respectively. Variability in climate will be influenced by changes in ENSO and the SPCZ. However, Vanuatu also has an increased risk of tropical cyclones throughout the wet season, with the most significant of recent impacts being cyclone Pam, in March 2015. The influence of such storms can be seen as peaks in wind speed, and alternating direction, in research campaigns, including the Southwest Pacific Ocean Circulation and Climate Experiment (Ganachaud et al., 2014), have led to a significant increase in observations. Regional modelling efforts have also increased, to represent large-scale currents as well as smaller scale features in the Solomon and Coral Seas (with grid spacing 3 − 4 km; e.g. Djath et al. (2014); Hristova et al. (2014)). This combination of models and observations has helped to map the pathways of major currents and their variability, connecting the South Pacific and regional Solomon and Coral Seas.
The broad South Equatorial Current flows westward in the South Pacific. As this encounters island chains in the western Pacific, the current splits around the islands, forming jets. For example, the North Vanuatu Jet then flows westward between the Vanuatu and Solomon Islands. Transport also occurs on smaller scales through narrow passages within the islands (e.g. Figure 3). However, the location and strength of these is dependent on the bathymetry. For example, Iron Bottom Sound (off-shore from Honiara) is bounded by both islands and shallower bathymetry, restricting flow through the region. Stronger currents are found to the north and south of Guadalcanal, particularly through the deeper channel of Indispensable Strait (e.g. Figure 3a; Djath et al. (2014)).
Variability in wind stress also results in variability in ocean currents. Around Guadalcanal, this leads to stronger currents from the Pacific during June-September, and weaker currents during November-January ( Figure 3a-b). Currents within the Vanuatu archipelago are typically weaker than those found between the Solomon Islands, with a less visible seasonal cycle (Figure 3c-d). Around Efate, a mean anticlockwise flow is seen around the island through most of the year.
While the recent research effort has significantly improved our understanding of circulation within and between the deeper ocean basins, there is still limited information for the shallow coastal seas. Model resolutions of 3 − 4 km struggle to represent the coastlines around individual islands (e.g. Figure 1), and the  )). Many global and regional models also neglect the influence of tides, which can have a significant impact on coastal currents.

Core model description
To investigate the spread of contaminants in the coastal ocean, regional models have been developed for both the Solomon Islands and Vanuatu. The modelling system used here is Telemac-3D (v7p2r2; Electricité de France (EDF) (2017); Hervouet (2007)). This model provides a variable grid resolution (with an unstructured triangular mesh), and varying complexity, dependent on the region and processes of interest.
To assess potential impact on surrounding populations, the focus here will be on surface concentration of contaminants rather than depth-averaged concentration. While Telemac can be used in a barotropic (2D) mode, a 3D model is then necessary, to inform on both the vertical and horizontal extent of dispersion. This also allows the model to resolve flows in both the shallow coastal zones of interest, and through deep trenches between the islands. Further details on the model grid, forcing and treatment of baroclinic processes can be found in the following sections (3.2 -3.5).
The Telemac-3D model configuration here uses the hydrostatic approximation. Whilst this may not be realistic on the scale of individual estuaries, given the wider bays of interest here (resolutions of < 100 m up to 5 km), this is a reasonable approximation for the configuration. Use of the non-hydrostatic approximation also increases computation time significantly (runtimes increase by approximately 50%), so would limit the periods and scenarios that would be feasible to explore here.
Treatment of tidal flats is included within both configurations, such that the free surface gradient is corrected wherever dry areas are detected. Advection of velocity and tracers is then handled using the MURD (Multidimensional Upwind Residual Distribution) scheme adapted for use with tidal flats. Horizontal turbulence is modelled using the Smagorinsky scheme, which provides a turbulent viscosity dependent on the size of resolved motions and therefore on the size of the mesh spacing. The k − ε model was considered as an alternative, however this proved to be less stable than Smagorinsky in these configurations, and was therefore discounted for this study. A mixing length model is used for vertical turbulence, with the Quetin option chosen to allow representation of wind drift in the surface boundary layer (Gauthier and Quetin, 1977). Bed friction in the model is applied by using the Law of Nikuradse with a roughness length equal to 0.01 m, suitable for a rippled non-cohesive granular bed. This is applied as a constant over the model domain.

Domain and bathymetry
The study here focuses on pollutants released from the main centres of population for Solomon Islands and Vanuatu, Honiara and Port Vila, respectively. Regional models are therefore centred on these cities, with the resulting model domains shown in Figure 1. Resolution is increased along the coastlines of interest, with grid spacing ∼ 100 m along the coast of Honiara, and ∼ 20 m at the coast in Port Vila Bay. Fully resolving each individual outflow or river is not feasible for the domains considered here. However, final resolution does ensure that key features of coastline are resolved, which could influence circulation in the proximity of these outflows (e.g. Point Cruz peninsula in Honiara, and complex coastline and varying bathymetry within Vila Bay). Given this small grid spacing, a time step of 1 s is used for both models.
Both model domains extend into the deep ocean, off-shore, to ensure that circulation near the coast can evolve without being overly influenced by the boundary conditions. To allow for more efficient simulations (especially given the short time step), progressively coarser resolution is used away from the coast, increasing 6 to at least 5 km spacing at the boundary ( Figure 1). Both models have 10 vertical sigma-levels, with resolution focused at the surface and base of the water column. These layers allow for more accurate representation of currents and dispersion near the sea surface, with the depth of each level, as a percentage of full depth, defined as: 1, 2, 10, 20, 50, 80, 90, 95, 98, 100 (where 1 represents the seabed, and 100 represents the sea surface). The resulting 3D

Boundary and Surface Forcing
For both models, tidal forcing is provided at the open ocean boundaries. Eleven tidal constituents are provided from the Topex Poseidon crossover solution (TPXO), Pacific Ocean 1/12 • regional model (Egbert and Erofeeva, 2002). These TPXO harmonics are used to drive the model, with prescribed velocity and elevations along the open boundary. Tracers evolve freely through the boundary.
For the studies considered here, the major sources of pollution are at the sea surface (either in river outflows or coastal pipe discharge). Surface currents and mixing are likely to be influenced by the local wind stress. Therefore, wind forcing is applied for each model. Wind data was obtained from European Centre for

Medium-Range Weather Forecasts (ECMWF) atmospheric reanalysis product, ERA5 (Copernicus Climate
Change Service (C3S), 2017). The same wind forcing is applied over the domain, but varies in time, with wind speed updated at daily frequency. These daily speeds were taken from locations in Mele Bay and Iron Bottom Sound (the regions of interest for the model domain), and are shown in Figure 2.

Tracers and sources
For the study regions here, the sources of contaminants occur predominantly through freshwater outflows along the shore. Temperature and salinity are included as active tracers within the model, and determine 7 the density of the ocean. We then assume that the transport of these buoyant plumes will be dominated by salinity gradients between the outflows and surrounding seawater. While temperature gradients are found with depth (especially in the deep, open ocean), there is little difference in temperature expected in the surface mixed layer around the coast. Variability in open ocean surface salinity is also likely to be negligible compared to the difference between the open ocean and fresh sources.
For simplicity, these idealised experiments then set a constant temperature throughout the domain, but salinity differs between the ocean (initially 34.5 throughout) and each river or pipe outflow (∼ 0 PSU).
The constant temperature and salinity in the open ocean then neglects the impact of baroclinic currents.
However, the salinity gradient between the ocean and freshwater sources will result in buoyancy-driven circulation, associated with these sources (for example determining the behaviour of river plumes). Future work will investigate the impact of both temperature and salinity variability, and resulting baroclinic currents, throughout the region.
At the time of this study, there is no data available from river gauges in either the Solomon Islands or The majority of the simulations here will consider the largest value (PWWF), as a "worst-case" scenario.
However, it is worth noting that these peak flows are based on the population in 2017. By 2022, with an increase in population (from 105,453 to 125,245) and increased proportion connected to sewage system (from 55% to 62%; Solomon Islands Water Authority and Hunter H20 (2017a)), the PDWF for Honiara is forecast to increase to 168.8 l/s (Solomon Islands Water Authority and Hunter H20, 2017b). There is little information on flowrate for Port Vila. Therefore, given similar population size of Honiara and Port Vila, the same discharge values are assumed for both model domains. For both locations, the total discharge is then distributed between the known pipe locations. Whilst flow rates are likely to vary with time (and throughout the day), both models assume a constant flow rate for simplicity.
To investigate water quality, pollution in the model has been represented by a passive tracer. This passive tracer has been released from river sources, as well as known locations of outfall pipes along the shore. In addition to runoff or discharge rates, a concentration of tracer is then set for each source.
Water quality can be defined by a number of different parameters, many of which have been observed during the CMEP field campaigns (Devlin et al., 2019(Devlin et al., , 2018. Here, the concentration of passive tracer is scaled by the level of Dissolved Inorganic Nitrogen (DIN). This variable was chosen due to availability of observations in the region, to enable scaling of sources along the coastline. The caveat for any comparison with observations here is that DIN is an active tracer, varying with time due to chemical and biological processes in the ocean. The model does not represent these processes, and is therefore not a true water quality model, but is rather a tool for understanding the circulation in the region, and therefore potential spread of contaminants. To clarify that this is a passive tracer, hereafter we refer to the tracer as simply "pollutant". Due to the lack of previous studies around these islands, this is still a significant step forward in understanding their coastal environment. This suggested concentration is likely an overestimate for individual pipes, given that other sources may exist along the river, and the concentration at the mouth may also be influenced by nearby pipes along the coast. However, the order of magnitude is justifiable given the total flux of DIN observed, and is also consistent with nutrient concentrations expected for untreated wastewater (e.g. Vazquez-Montiel et al., 1996). Based on these observations, the concentration for each pipe is then estimated as 10, 000 µg/l. Given the uncertainty in river runoff volume, and the fact that chemical/biological processes are not included, ADWF total, consolidated outflow at 20 m this is a reasonable assumption to apply for the simulations. This magnitude is also further supported by observations from one pipe in Vanuatu, where a concentration of 4, 451 µg/l was observed in 2016.
As with the flow rate, the source concentration of each tracer is held constant with time in the model.
Tracer diffusion coefficients were set to 0.1 m 2 /s in the horizontal, and 0.1 m 2 /s in the vertical.  Table 1, with scenarios outlined as follows:

Experimental design
• Scenario 1: Control simulations under December wind conditions (Sol N W and Van wet ).
• Scenario 2: Impact of seasonality in wind forcing for Solomon (Sol SE ).
• Scenario 3: Impact of runoff rates in Vanuatu (Van dry ).
These experiments provide guidance for when concentrations are likely to be higher or lower in the region.
For the Solomon Islands, Scenario 2 considers varying wind forcing between prevailing north-westerlies or south-easterlies, typical during the wet or dry seasons, respectively. Figure 2 shows that there is seasonal variability in both the speed and direction of wind forcing at Honiara. Scenario 2 then provides insight into how sensitive the dispersal of pollutants may be to changing wind stress for this location.
For Vanuatu, while there is some seasonality in wind speed (and variability associated with storms), the mean direction remains predominantly southeasterly throughout the year. The scenarios tested here then use the same period of forcing for each experiment (December, during the wet season). However, Scenario 3 investigates the impact of varying the outflow rates within pipes, comparing ADWF with PWWF conditions.

Scenario 4 is used to test potential mitigation methods for both Honiara and Port Vila. Whilst there
is no real substitute for sewage treatment before disposal, this investigates potential future development of the pipeline system (e.g. as suggested in Solomon Islands Water Authority and Hunter H20 (2017a)).
It is worth emphasising again that these are idealised scenarios. The lack of chemical and biological processes in the model means that these results should not be considered as indications of actual DIN concentrations in the surrounding ocean. Rather than focusing on potential threshold values for "safe" conditions, the focus here is instead on variability between different scenarios, and hence how local concentrations are likely to vary.

Validation of tidal circulation
While current observations exist for the wider vicinity of these islands (e.g. Section 2; Hristova and Kessler (2012) (south-western Vila Bay). As the circulation in these regional models is driven initially by tidal forces, these can then be used to assess the tidal circulation. expected. While these models have relatively high resolutions along the coastline, they are still unlikely to represent small-scale coastal features, such as variability in depth and port infrastructure, where these tide gauges are situated. However, this demonstrates that while there is some difference between model and observations, both models should perform well in reproducing the tidal circulation and therefore mixing at these locations.

Scenario 1: Control
Control simulations for both Honiara and Port Vila demonstrate the impact of PWWF outflows during wind conditions that would be typical for the wet season (December; Sol N W and Van wet , Table 1). Results show that the highest concentrations are found near the point of source, as expected (Figures 5a and 6a).
However, Port Vila and Honiara differ in the geometry of their coastlines and regional bathymetry. Honiara has a narrow coastal shelf, with depth then dropping steeply away from the coast. Vila Bay contains portions of both shallow and deeper water within the sheltered bay, but remaining < 50 m deep. This means that while the surface concentration drops off relatively quickly away from the shore around Honiara (in depths the outfall pipes (10, 000 µg/l), or in the Mataniko River (210 µg/l), the volume of discharge (∼ 42.0 m 3 /s and 3.8 m 3 /s, for the Lungga and Mataniko rivers, respectively) means that there is a wider impact on the coastline due to the plume extent. In the southern hemisphere, coriolis forces will encourage buoyant freshwater plumes to flow with the coast to their left. However, the influence of residual tidal circulation, and the prevailing wind direction, does cause variations along the coastline (and will be discussed further in Subsection 5.2).
In central Honiara, Point Cruz peninsula is found to the west of the Mataniko river mouth. This restricts flow from the Mataniko flowing further west along the coast. Along with a lower number of outfall pipes along the western coastline, this contributes to overall lower concentrations in the west, compared with the east. However, an outflow pipe on the western side of Point Cruz does still cause the high concentrations visible here (Figure 5a).

Scenario 2: Impact of wind stress
Both locations show variability in wind speed during the year (Figure 2). For Honiara, there is a marked change in both wind speed and direction between the seasons. Therefore, Scenario 2 investigates the impact of this seasonality (Sol N W and Sol SE , Table 1). While the change in wind stress is typically also accompanied by rainfall, both these experiments consider PWWF, but with wind forcing from either December (Sol N W ) or July (Sol SE ). Figure 7 shows that this change in wind stress has an impact on the surface ocean currents. While the large-scale residual currents appear similar (Figure 7a-b), the anomalies show that weaker, northwesterly winds during the wet season can lead to reduced off-shore flow from the coast (Figure 7c). Surface concentrations of pollutant also show similar large-scale patterns, with the largest concentrations at the point of source along the coast, compared with dispersion off-shore (Figure 5a-b). However, the anomalies between these simulations show that there is a higher surface concentration at the coast during Sol N W (Figure 9a).
This shows that aside from any change in runoff volume during the wet season, the wind conditions will be favourable for increased surface concentrations. This result can initially be attributed to the change in surface currents, reducing off-shore dispersion (Figure 7c). However, along with the change in wind direction, there is also an increased wind speed during July (Sol SE , Figure 2). Increased wind stress will increase mixing in the surface layer, distributing pollutant through greater depth, and therefore reducing surface concentrations. It is likely that the results for this Scenario are down to a combination of these forcing mechanisms.
Port Vila sees some similar seasonality, with a mean reduction in wind speed during the wet season.
Although this has not been tested here, it is therefore likely that such conditions would also be favourable for higher surface concentrations. However, the peak wind speeds associated with tropical storms are likely to increase mixing in the region. Such processes would have to be considered along with the increased rainfall Sol 50 -Sol NW associated with the storm, and therefore increased runoff in the region, to decide on the likely dominant impacts.

Scenario 3: Impact of runoff rates
Both Honiara and Port Vila experience large variability in rainfall during the year. Increased rainfall will lead to larger volumes of runoff through drainage systems as well as rivers. With the absence of river gauge data, the focus here is on varying outfall pipes, considering a typical average dry weather flow (ADWF) compared to peak wet weather flow (PWWF) discharges. With limited information on how concentration may vary during the year, the concentration of each outfall is kept the same in this scenario. Therefore, the volume of pollutant released will vary directly with changes in outflow volume. While Scenario 2 considered the impact of wind stress in Honiara, Scenario 3 focuses on Port Vila, comparing Van dry and Van wet , respectively.
As expected, reducing the discharge volume leads to lower concentrations throughout Vila Bay (Figures   6a-b and 10a). As the river runoff is kept constant, there is little impact along the shore of Mele Bay.
However, there is a visible reduction in concentration associated with flow exiting Vila Bay.
While there is a significant reduction in surface concentration overall, there are still regions of elevated concentrations. In addition to high concentrations in the immediate vicinity of outflows, Fatumaru Bay still has widespread concentrations exceeding 25 µg/L. The shallow sill at the entrance to this bay still leads to an accumulation of pollutant. Similarly, the southeastern corner of Vila Bay is partially enclosed by surrounding bathymetry. Here, although there are fewer pipes discharging into the area, concentrations also increase due to restricted flow through the region.
These results demonstrate that while concentrations are likely to be lower during dry weather, there are regions within the Bay that are likely to be at risk from higher concentrations throughout the year.
Restricted flow into and out of enclosed bays may also mean that elevated concentrations are retained for longer periods in these regions following sporadic wet weather (e.g. during tropical storms), or other discharge events. However, to consider concentration evolution or longer residence timescales, decay rates of contaminants as well as local chemical or biological processes, would also need to be considered.

Scenario 4: Impact of moving pipes to depth off-shore
A major cause of high surface concentrations along the shoreline is release of pollutant from pipes at the sea surface along the coast. Given that these are typical freshwater sources, they will be more buoyant than the surrounding seawater. The highest concentrations then remain at the surface, with stratification limiting the strength of vertical mixing. With this in mind, Scenario 4 considers a proposed method of reducing impact of sewage along the coast, by consolidated release through pipes at greater depth, off-shore.
This should lead to greater mixing (dilution) with the surrounding seawater, as the buoyant plume tends to rise, thereby reducing concentrations that reach the surface.
The experiments carried out for Honiara and Port Vila all consider consolidated outflows, and they also consider either varying depth and location (Sol 20 and Sol 50 ), or varying outfall rates (Van dry20 and Van wet20 ).
For all these experiments, there is a reduction in surface concentration along the coastline . However, the impact is more apparent for Honiara than Port Vila, with a consistent reduction across the region seen in Figure 9b, as opposed to the more spatially-variable response seen in Figure 10c show that, as expected, there is a large reduction in surface concentration. This is shown further in Figure   9b, which shows a large reduction along the coast, in the vicinity of current surface outfalls. Although not shown, the difference between surface concentrations for Sol 50 and Sol 20 is relatively small. However, looking at the difference sub-surface, it is clear that release at 20 m could have a larger impact on coastal waters (Figures 11 and 12). Ocean currents have a tendency to follow contours of equal depth (e.g. Figure 7). This means that the release at 50 m remains further from the shore as it disperses in the ocean (Figure 11a). However, the narrow shelf allows the plume from 20 m depth to bring pollutant close to shore ( Figure 11b). As bathymetry drops steeply from the shore of Honiara, the 20 m depth release is only ∼ 300 m from the shore, compared to ∼ 1 km for release at 50 m. For a shorter, shallower pipeline, release is then closer to the coast, and closer to the surface.
In contrast to Honiara, results for Port Vila show spatial variability in anomalies within the bay (Figure   10b-c). Bathymetry is shallower than 50 m throughout Vila Bay. Therefore, only one location of release is considered here, at the 20 m isobath, close to existing pipeline systems at the northern end of the Bay ( Figure   13). However, as with the surface release, experiments consider the impact of either PWWF (Van wet20 ) or ADWF (Van dry20 ). Both Van wet20 and Van dry20 show reduced concentrations in the vicinity of existing pipe locations. Both also show a large reduction within Fatumaru Bay. However, there is a differing response in central and southern Vila Bay ( Figure 10).
As the location of release has not changed, this shows how the volume of outflow plume may affect its Van wet20 Van dry20 Van dry20 . Locations of the cross section in Vila Bay is shown in Figure 13. Grey dashed lines indicate the depth levels of terrain-following coordinates used within the model. trajectory of dispersion through the region. With a larger outflow, the increased buoyancy allows the plume to penetrate closer to the surface, and then also further horizontally throughout the bay (Figure 14a and   13a). With a reduced outflow, the plume remains predominantly at depth in the water column, and has a limited horizontal extent (Figure 14b and 13b). As the Van wet20 plume reaches the surface off-shore, it contributes to increased surface concentrations in the central bay (Figure 10), whereas Van dry20 results in increased concentrations in southern Vila Bay, as the plume spreads towards the south.
This study shows how, for release at depth, the volume of outflow can affect the trajectory of dispersal as well as the concentration. As the rising plume mixes with surrounding seawater, vertical dispersal depends on the resulting level of neutral buoyancy in the water column. As currents often vary with depth, the horizontal spread of the plume will also be influenced by its depth of neutral buoyancy. This should be kept in mind, along with any regional bathymetry that may constrain spread at this depth. A deeper plume is likely to be more constrained than one that reaches the surface.
In the studies considered here, the water column is considered to be initially fully-mixed, with constant density. Including further stratification within the water column could then impact both the vertical and horizontal extent of the plume, and is worth further investigation. While the use of 10 vertical levels provides sufficient resolution in the coastal ocean, the terrain-following grid becomes relatively coarse in the deeper ocean (especially at mid-depth within the water column; Figures 12 and 14). This may lead to an overestimate of dispersion through the water column. Assessing the impact of vertical resolution on these scenarios would then also be worth further consideration.
For each of these experiments, it is also worth keeping in mind that while surface concentrations are reduced, the increased concentration at depth will still have an impact on ocean ecosystems. This could then have associated impacts on human health through marine food resources. In the absence of waste water treatment before disposal, such impacts would need to be considered further before future development.

Discussion and Conclusions
This study assesses the potential dispersion of pollutants around Honiara, Solomon Islands, and Port Vila, Vanuatu. Using 3D ocean circulation models, scenarios were tested to investigate the vulnerability of the system and potential control measures. While the pollutant release here has been scaled by available observations of DIN, the results here are applicable to any contaminant or bacteria released through the rivers or drainage systems along the coastline. For example, this includes E. coli and faecal coliforms, which pose risk to public health. Results show that high coastal concentrations are most likely during the wet seasons, due to increased volumes of discharge, as well as favourable wind stress. Buoyant plumes flow along the coastline, and high concentrations build up in enclosed bays. Testing consolidated outflows at depth off-shore shows overall reduction of surface concentrations along the coastline. However, the impact is not necessarily consistent within each region, with dispersal dependent on the depth, off-shore positioning, and volume of outflow.
The results here will help to inform future observations or monitoring programmes. For both regions it is likely that the current discharge system will allow any associated pollutants to remain close to shore, and in the surface ocean. To mitigate high surface concentrations, runoff could be discharged at sufficient depth and distance from the shoreline. However, it must be noted that discharge at depth will still have an impact on the ocean ecosystem, and potential food sources, with increased contaminant then at depth off-shore. There is no real substitute for fully-functioning waste-water treatment system.
The study here focuses on idealised scenarios. Future work should initially increase the complexity of the model, to demonstrate the impact of varying temperature as well as salinity, and large scale baroclinic ocean currents. With increased stratification in the ocean, the impact of non-hydrostatic processes and vertical turbulence models should also be explored, to assess the impact on vertical dispersal from buoyant outfalls.
In addition to this, introducing water quality processes will inform on likely timescales for any pollutants to 25 decay in the region following release. Such analysis would be required to more accurately assess the off-shore extent, or recovery time following one-off events.
For Vanuatu, tropical cyclones pose a significant threat to the region. While this study has demonstrated potential influence of changes in either runoff or wind stress, the impact of extreme weather events (causing significant changes in both factors) has not been considered here. This is worth further study, to assess both the short term and potential long-term impact of extreme weather in the region.
Further observations and monitoring are also vital to establish the current baseline conditions and variability in the region. This will further inform model simulations, as well as validation of results. There are a number of caveats in this study, and prospects for future work. However, given the lack of previous studies available, this article presents an important first step in our understanding of coastal circulation and therefore transport of pollutants around these Pacific island states.