Restore or retreat? saltwater intrusion and water management in coastal wetlands

Abstract Coastal wetlands perform a unique set of physical, chemical, and biological functions, which provide billions of dollars of ecosystem services annually. These wetlands also face myriad environmental and anthropogenic pressures, which threaten their ecological condition and undermine their capacity to provide these services. Coastal wetlands have adapted to a dynamic range of natural disturbances over recent millennia, but face growing pressures from human population growth and coastal development. These anthropogenic pressures are driving saltwater intrusion () in many coastal systems. The position of coastal wetlands at the terrestrial–marine interface also makes them vulnerable to increasing rates of sea‐level rise and changing climate. Critically, anthropogenic and natural stressors to coastal wetlands can act synergistically to create negative, and sometimes catastrophic, consequences for both human and natural systems. This review focused on the drivers and impacts of in coastal wetlands and has two goals: (1) to synthesize understanding of coastal wetland change driven by and (2) to review approaches for improved water management to mitigate in impacted systems. While we frame this review as a choice between restoration and retreat, we acknowledge that choices about coastal wetland management are context‐specific and may be confounded by competing management goals. In this setting, the choice between restoration and retreat can be prioritized by identifying where the greatest return in ecosystem services can be achieved relative to restoration dollars invested. We conclude that restoration and proactive water management is feasible in many impacted systems.


Introduction
Environmental stressors impacting coastal wetlands have both natural and anthropogenic sources. Coastal zones are dynamic and subject to changing environmen tal conditions caused by natural variations in climatic, oceanographic, and ecological processes such as flood ing, drought, long term climate cycles, storm surges, hurricanes, winds, herbivory, and changes in sea level. At the same time, increasing human population and development in coastal areas have increased point source and non point source pollution (Sampathkumar 2015), habitat destruction (Hefner andBrown 1984, Li et al. 2014), and hydrological alteration (Niemi et al. 2004). Moreover, due to their position in the landscape, coastal wetlands are vulnerable to both sea level rise (SLR) and changing temperature/rainfall regimes driven by climate change. In human dominated landscapes, these natural and anthropogenic stressors frequently overlap with synergistic, and sometimes catastrophic, consequences for both natural and human systems. This phenomenon is perhaps best evidenced by the envir onmental, economic, and social damage caused by Hurricane Katrina as it passed through the severely degraded wetlands of the Mississippi River Delta (MRD; Laska and Morrow 2006, Petterson et al. 2006, Deryugina et al. 2014. Considering these multiple, interactive stressors, a central question for land managers is how to proact ively manage and/or restore coastal wetlands in a future of changing climate, land use, and shoreline modification. Here, we review our current understanding of the state and projected trajectory of coastal wetland change, with a focus on the potential for improved water manage ment to mitigate the worst impacts of saltwater intrusion (SWI) and best prepare coastal wetlands to adapt. First, we review how natural and anthropogenic SWI drivers alter hydrology and water quality in coastal wetlands and describe the subsequent ecological response. Next, we provide a framework for assessing the potential for restoration interventions in systems degraded by SWI and summarize approaches for their restoration. Among the numerous causes of coastal change, SWI is a primary stressor in many coastal wetland systems. SWI refers to landward and/or upward displacement of the freshwater-saltwater interface in coastal aquifers (Knighton et al. 1991) and increased saltwater penetra tion in estuaries (Barlow 2003). SWI drivers can be natu ral, anthropogenic, or synergistic. Natural drivers of SWI include storm surges, hurricanes, climatic fluctuations, SLR, and subsidence (Fig. 1). Anthropogenic SWI drivers include land drainage, pumping of coastal freshwater aquifers, reduction in freshwater discharge from dam construction, water withdrawals, or other water diver sions, and hydrological/hydraulic structures and land use changes within watersheds (Fig. 2). While SLR is a natural driver of SWI, anthropogenic climate change exacerbates SLR effects and may also alter the timing and magnitude of climatic cycles and extreme events; these effects are addressed in section SWI and anthropogenic climate change.

Natural drivers of SWI
Natural drivers of SWI occur over a gradient of time scales and frequencies (Fig. 1). Chronic and acute drivers of SWI both act to shape the coastal environmental and dictate the direction and rates of change in coastal wet land ecosystems. Chronic drivers of SWI include SLR and geologic uplift and subsidence. These drivers act over millennial timescales, although changes in their relative rates can drive large scale changes in coastal wetland structure and function (DeLaune and White 2012). Acute drivers of SWI include tsunamis, hurricanes, droughts, and climate oscillations. Importantly, acute drivers may elicit both short-term and long-term effects on coastal swamps and marshes that can be positive (Baustian and Mendelssohn 2015) or negative (Brisson et al. 2014, Leonardi et al. 2016. At the shortest timescales, tsunamis can act as phys ically destructive forces on the coast and carry saline water far inland. For example, the 2011 Tohoku oki tsu nami carried a large load of sediment, nutrient, and salin ity 5 km inland (Chagué-Goff et al. 2012). Affected rice paddies and other freshwater wetlands became saline or brackish and remained at highly elevated salinity lev els for up to 5 months. Even after surface water salinity decreased, salt concentrations in the soils remained ele vated in some areas for the remainder of 2011 (Chagué Goff et al. 2012). The Indian Ocean tsunami of 2004 illustrated the physically destructive potential of these events on coastal wetlands at a large spatial scale. An assessment by Indonesia's State Ministry of National Development Planning estimated that 25%-35% of coastal wetlands had been destroyed in tsunami-affected areas (Srinivas and Nakagawa 2008).
Like tsunamis, hurricanes are stochastic events that can transport saline water, sediments, and bound nutri ents far inland (Rejmánek et al. 1988). Hurricanes have both short term physical impacts (winds, wave, and surge in the case of hurricanes) and longer term impacts from subsequent flooding with saline waters. In many regions, coastal salt marshes or mangroves ring the continents and can serve to weaken hurricane wind and wave damage and the associated storm surge (Moller et al. 2014), greatly reducing the potential for damage to Fig. 1. Schematic representation of the spatial and tem poral scales of natural saltwater intrusion (SWI) drivers. Temporal scale varies from fast acting and low frequency (i.e., acute) to slow acting and consistent (i.e., chronic). Spatial scale indicates the size of the area impacted by SWI. Fig. 2. Schematic representation of the spatial and tem poral scales of anthropogenic saltwater intrusion (SWI) drivers. Many anthropogenic drivers act over a wide variety of spatiotemporal scales as a function of development intensity (e.g., local agricultural water abstraction in a small river vs. damming a large river). Note reduction in axis scaling from Fig. 1, indicating that anthropogenic drivers of SWI are generally limited to centennial timescales and regional spatial scales. One exception is anthropogenically exacerbated climate change, which can drive larger spatiotemporal processes (indicated by dashed upper border). the coastline and inland freshwater wetlands. However, coastal marshes and mangroves are being lost, decreas ing the storm surge attenuation potential and increasing the vulnerability of adjacent freshwater wetlands (Moller et al. 2014).
The magnitude and duration of hurricane surge induced SWI is strongly driven by hurricane intensity, wind direction, tide, and local hydrological conditions. For example, salinity increased from 4 to 15 parts per thousand (PPT) in a Louisiana brackish coastal marsh during Hurricane Rita in 2005 and remained elevat ed for more than 3 months. Salinity in the same marsh was unchanged during Hurricane Katrina due to differ ences in wind direction (Steyer et al. 2007). Vegetation in freshwater and brackish wetlands is not adapted to the high salinity water associated hurricane surge and often dies back completely, leaving bare soil that is vul nerable to erosion. After inundation, wetland recovery depends on several factors including post intrusion salinity and flooding level (Flynn et al. 1995). If saline water is trapped inland by levees or roads, recovery can be extremely slow. Also vital are the location, connec tivity, and type of propagules to re populate the wet land if significant plant mortality occurred (McKee and Mendelssohn 1989). Recent research into community re-assembly after hurricane storm surge points to addi tional factors, including species-specific reproductive strategies (Middleton 2009), sediment and wrack deposi tion (Tate and Battaglia 2013), and stochastic community assembly (Guo et al. 2014).
Droughts and yearly to decadal climate oscillations are longer term drivers of SWI ( Fig. 1), which can play a significant role in the long-term survival of coastal wet lands. During periods of drought, brackish and fresh water wetlands face stress from reductions in rainfall and freshwater flows with subsequent influences on SWI in surface water and groundwater (Kaplan et al. 2010a, b, Kaplan andMuñoz Carpena 2014). Without relief from saltwater stress, vegetation dies back, eventually con verting to salt marsh or open water (Drexler and Ewell 2001). Droughts can have particularly strong effects on coastal wetlands when coupled with SLR induced SWI, leading to rapid declines in both species richness and regeneration (Desantis et al. 2007). In addition to vegeta tion effects, drought-induced changes in biogeochemical cycling can cause coastal wetlands to release nitrogen (Scott et al. 2003), with potential effects on marsh pro ductivity, accretion, and coastal eutrophication.
Climate oscillations affect coastal wetlands through impacts to a variety of abiotic factors. For example, the North Atlantic Oscillation (NAO) and Atlantic Multidecadal Oscillation (AMO) affect hurricane fre quency and tracks. In the eastern United States, the NAO influences whether hurricanes enter the Gulf of Mexico or travel up the Atlantic Coast (Ting et al. 2009), and the AMO alters sea surface temperatures (SSTs) in the Atlantic Ocean, which also affect hurricane routes (Knight et al. 2006, Wyatt et al. 2012. AMO, NAO, and El Niño/Southern Oscillation (ENSO) all influence the mag nitude and distribution of precipitation and temperature anomalies, impacting coastal and marine ecosystems by changing the coastal freshwater delivery (Karamperidou et al. 2013, Nye et al. 2014. Dry phases of these climate cycles represent long term droughts, which can cause persistent increases in coastal wetland salinity, driving vegetation mortality, and reorganization (Drexler and Ewell 2001, Tolan 2007, Angelini et al. 2016. At the longest timescales, chronic stressors include SLR and subsidence, which act over hundreds or thousands of years (Fig. 1). SLR decreases the relative elevation of coastal wetlands and subsidence decreases their absolute elevation. Both drivers act at a slow, but consistent pace, so that the elevation of coastal wetlands is relatively well poised to "keep up" via accretion of organic mat ter and trapping of inorganic sediments (Morris et al. 2002). Sea level has risen throughout the current inter glacial period (USGS 2000) and particularly during the mid to late Holocene era, though not always at the same rate (Wanless 1989). For example, global sea level rose quickly (2-5 mm/yr) between 3,200 and 6,500 yr before the present, but slowed to approximately 0.4 mm/yr over the next 3,200 yr, allowing for the development of broad coastal wetlands around the continents. More recently, the rate of SLR has increased rapidly to between 2.8 and 3.6 mm/yr (Church et al. 2013). While there are uncer tainties about the future rate of ice sheet collapse and glacial melting (Nicholls and Cazenave 2010), anthro pogenic climate change is projected to rapidly accelerate SLR over rates observed in recent millennia. The ability of coastal wetland ecosystems to keep up with this accel eration is spatially variable and largely uncertain (see section SWI and anthropogenic climate change).
Along with SLR, compaction, subsidence, and uplift are natural processes acting over long timescales in coastal wetlands (Fig. 1). Coastal wetlands are built from surface layers of largely uncompacted peat. Over time, peat at lower depths is compacted and dewatered by the weight of the overlying soil and water (Morton et al. 2002), leading to reductions in elevation. Without accretion of organic and mineral sediments, coastal wet lands become inundated, experiencing SWI in much the same way that is expected with SLR (DeLaune and White 2012). It is estimated that compaction and tectonic driven subsidence in the MRD account for 80% of relative SLR in that region (Dokka 2006), though this problem is likely exacerbated by anthropogenic activities (Penland and Ramsey 1990, see section SWI and environmental modification). Uplift and subsidence are larger-scale phenomena that typically occur in relation to geologic and tectonic activity (Lambeck et al. 2014), and in coast al regions are driven by isostatic adjustment following glaciation, deglaciation, and/or mountain formation. In North America, isostatic rebound following the retreat of the Wisconsin ice sheet is estimated to be from −3 to 3 mm/ yr along the U.S. eastern seaboard. In Asia, uplift of the Himalayas is causing subsidence of the Bengal basin, accelerating SWI in that region (Alam 1996).

Anthropogenic drivers of SWI
In contrast to natural drivers, many of the current SWI driven threats to coastal wetland sustainability have anthropogenic (and/or synergistic) origins (Fig. 2). We identify two primary subcategories of anthropogenic SWI drivers, namely anthropogenic climate change and environmental modification.

SWI and anthropogenic climate change
Coastal wetlands are uniquely vulnerable to landscape scale ecological change driven by global climate change (Burkett andKusler 2000, Day et al. 2008). Climate change affects coastal wetlands through changes in the frequen cy, magnitude, and duration of acute events (e.g., hurri canes, droughts, climate oscillations) and via changes in the rates of chronic stressors such as SLR, all of which play a role in SWI (section Natural drivers of SWI). Climate change thus exacerbates natural SWI drivers, yielding accelerated or more extensive effects (Fig. 2).
Climate change is expected to increase hurricane intensity largely due to increased SST (Mendelsohn et al. 2012). This expectation is supported by paleocli mate data from the past 200 yr (Donnelly et al. 2015) and observations over the past 30 yr (Knutson et al. 2010), which show increased hurricane activity and intensity during periods of warm SST. Given current limitations in our understanding of the processes that drive hurri cane frequency and intensity (Zwiers et al. 2013), con siderable debate remains on how these extreme events will change in the future. Despite this limitation, any increases in hurricane intensity can be expected to pro duce larger storm surges with concomitant increases in coastal wetland damage from scour, erosion, soil com paction, vegetation burial, and SWI (Lin et al. 2012, Thomas et al. 2015. What is unknown is whether coastal systems, which have adapted to survive periodic distur bance, can survive a future with either more frequent or more severe storms (Michener et al. 1997, Day et al. 2008, Knutson et al. 2010, Leonardi et al. 2016.
Like hurricanes, droughts are expected to increase in intensity as regional temperature and precipitation regimes shift with changing global climate (Trenberth et al. 2014). Generally, these changes will be seen as increasing regional temperatures and increased probabil ity of extreme temperature events. Changes in precipita tion regime are less certain and more regionally variable. In general, wet areas are expected to get wetter and dry areas drier (International Panel on Climate Change [IPCC] 2013). As an example, under Representative Concentration Pathway (RCP) 8.5, which represents a future with the highest greenhouse gas emissions, the U.S. Atlantic and Gulf Coasts will see an increase in maximum and minimum temperatures, mean and maxi mum 1 day precipitation, and the number of consecutive dry days (Wuebbles et al. 2014). Globally, these types of climate projections promise increasing frequency and severity of droughts. For coastal wetlands, less availa ble freshwater may be available to mitigate SWI. While the impacts of anthropogenic climate change on yearly to-decadal climate oscillations (NAO, AMO, ENSO) are uncertain , Knutson et al. 2010, cli mate change is expected to increase the magnitude of these climate cycles (Cai et al. 2014, Mann et al. 2014, meaning that coastal wetlands will likely face larger and more frequent extremes of wet and dry periods. Most pressingly, current and accelerating rates of SLR exceed those observed in recent millennia. Sea level rise may eliminate as much as 22% of the world's coastal wetlands by 2100 (Nicholls et al. 1999), though regional impacts would vary (Michener et al. 1997). Under the "best case" emissions scenario (RCP 2.6), sea level is expected to rise by 0.4-0.6 m by 2100 AD. In contrast, the unmitigated warming scenario (RCP 8.5) has temperature rising by 4.5°C, yielding 0.7-1.2 m of SLR by 2100 (Horton et al. 2014). Either outcome is likely to drive large scale disturbance and reorganization in many coastal wet lands. It is important to note that SLR rates and projec tions by IPCC (2013) and Horton et al. (2014) are based on global averages. Other factors such as subsidence, uplift, and accretion rate modify the local rate of SLR in a spe cific region (Morton et al. 2002). There needs to be better understanding of these local variables to improve early detection of accelerated SLR and enhance wetland sur vival and resilience at the local level (Haigh et al. 2014).
A primary determinant of how SLR will restructure different coastal wetland communities is accretion rate (Morris et al. 2002), which varies across ecosystem types, latitude, tidal range, suspended sediment concentra tions, and local hydrodynamic setting (Morris et al. 2002, Cahoon 2007. While a complete review of these factors is beyond the scope of this work, we briefly summarize coastal wetland accretion rates and how they are expect ed to respond to accelerated rates of SLR. Tidal fresh water forests accrete 1.3-2.2 mm of soil per year (Craft 2012), less than current SLR rates of 2.8-3.6 mm/yr (IPCC 2013). These forests face the threat of conversion to open water, freshwater marsh, or brackish marsh. Coastal freshwater and brackish marsh both have accretion rates that are higher than the current rate of SLR (Craft 2012). How accretion rates in these systems will respond to increased flooding duration and salinity remains uncer tain, but they are expected to be more resilient than tidal forests due to increased carbon sequestration and miner al sediment accumulation (Craft 2012).
The review by Kirwan and Megonigal (2013) seeks to identify threshold levels of SLR beyond which tidal wetlands transition to open water. These authors point to important biophysical feedbacks (e.g., sediment availability, tidal range, marsh productivity, SLR, and accretion rates) and anthropogenic factors that dictate this threshold rate. Importantly, the ability of salt marsh to maintain its elevation in the face of accelerated SLR is largely dependent on upstream flows of freshwater, nutrients, and sediments. With changing climate and other anthropogenic changes to watersheds, these flows cannot be assumed to be stationary. Some coastal wet lands will be able to keep pace with SLR for 50-70 yr, but will eventually transition to other wetland types or open water. Critically, this type of analysis does not account for added stress from anthropogenic SWI, which likely shortens the survival window in many systems (Thorne et al. 2015).
Finally, anthropogenically exacerbated climate driven drivers of SWI are often interconnected, which can ampli fy their effects. For example, more intense hurricanes bring more powerful storm surges, which when coupled to accelerated SLR and local subsidence can drive inun dation far more than either driver alone (Yang et al. 2014). Similarly, more frequent droughts coupled with acceler ated SLR can push coastal systems beyond their capaci ty to rebound (Desantis et al. 2007, Angelini et al. 2016, illustrating the perilous situation for coastal wetlands in regions where climate models predict that drought will become more frequent and severe (Karl 2009).

SWI and environmental modification
Beyond impacts to the global climate, humans have also directly modified the coastal environment, with wide spread impacts on SWI and coastal wetlands. These anthropogenic modifications act synergistically with the natural and climate change drivers discussed above and include river modification, groundwater abstraction, land drainage for agriculture and urban development, subsidence due to extraction of underground resources, and land use change (Fig. 2). In many cases, several of these drivers work simultaneously.
River modification causes SWI via construction of canals, changes in bathymetry, and dredging. Canals are used as shipping routes to major rivers and for oil and gas exploration, but also facilitate SWI. For example, sim ulation modeling of canals in coastal Louisiana showed that, under similar environmental conditions, the 5 and 10 PPT isohalines moved approximately twice as far inland when canals were widened and deepened (Wang 1988). To maintain river widths and depths required to accommodate ever increasing vessel sizes, natural rivers are also dredged and widened, causing similar SWI and associated ecological impacts. For example, mangrove encroachment was observed after dredging the Tanshui River in Taiwan, and substantial changes to tidal range and SWI occurred after dredging in the Pearl River in China (Liu et al. 2001, Yuan andZhu 2015). In addition, many rivers have been straightened to improve trans portation and flood control (Bechtol and Laurian 2005). These fast, deep, and straight channels act in a similar way to canals in permitting upriver SWI.
In addition to changes in surface water systems, groundwater pumping can cause SWI when aquifer recharge rates are lower than the rate of abstraction. Many coastal wetlands are dependent on the delivery of fresh groundwater from coastal aquifers either direct ly (i.e., discharge wetlands) or via complex interactions between surface water, groundwater, and soil water in the root zone (Kaplan et al. 2010a, Sánchez Martos andSánchez 2013). Aquifer overexploitation can lead to reduced freshwater delivery and increasing groundwa ter salinity in these systems. Examples of groundwater abstraction driven aquifer salinization are numerous globally (Sadeg and Karahanoðlu 2001), and in most sys tems, groundwater use plays a larger role than SLR in driving SWI (Ferguson and Gleeson 2012).
Land-use change, watershed modifications, and hydraulic infrastructure construction in coastal watersheds can also play a large role in driving SWI. Land use changes can increase impervious surface area, reduce aquifer recharge, and lower freshwater heads and facilitating SWI (Ranjan et al. 2006). Drainage of wetland areas for agricul tural and urban development in some coastal areas has also lowered groundwater elevations, leading to SWI in aquifers and surface waters (Holman and Hiscock 1998). Dam construction and surface water withdrawals can also cause reduction or elimination of freshwater flows to the coast, increasing SWI (Bunce et al. 2010). For example, the Aswan High Dam on the Nile River in Egypt has reduced freshwater outflows to the Mediterranean by more than 75%, leading to increased salinity in the wetlands of the Nile River Delta (Johnson 1997).
Finally, accelerated subsidence driven by human activi ties is a major driver of SWI in many coastal regions. For example, historical rates of subsidence in the MRD region were approximately 3 mm/yr, driven primarily by com paction and isostatic rebound (Wolstencroft et al. 2014). Current rates of subsidence are as high as 23 mm/yr driven by extraction of subsurface hydrocarbons and brine, which destabilizes overlaying sediments (Morton et al. 2002, Ko andDay 2004). Subsidence rates in the delta were high est after extraction peaked in the 1970s and have slowed with decreasing rates of extraction (Morton et al. 2002). Similar trends have been observed in the coastal wetlands of Texas (Morton et al. 2006). Schmidt (2015) illustrated the global scope of this problem with examples from Vietnam, Indonesia, and Thailand, where high rates of natural resource and groundwater extractions are also leading to accelerated subsidence and SWI.

Impacts of SWI on Coastal Wetlands
The impacts of SWI on coastal wetlands are variable in scale as a function of the magnitude, duration, and fre quency of the SWI stressors summarized above. These impacts include changes in primary production, com munity composition, and the provisioning of ecosystem services (Fig. 3).

Primary productivity
Coastal wetlands rely on primary production as a source of autochthonous organic matter for soil accretion. While salt marshes can increase productivity and accretion to match contemporary rates of SLR (Cahoon 2006), specific thresholds for marsh survival are locally variable and strongly impacted by human activities (Kirwan and Megonigal 2013). In contrast, SWI is strongly correlated with decreased primary production in coastal freshwater wetlands, where it can instigate increased subsidence and flooding (Fig. 3). Under increasing salt stress, coastal freshwater marshes have reduced above and below ground biomass Mendelssohn 1989, Neubauer 2011). This trend is also present in tidal forested wetlands, where net primary productivity can be twice as high in forests with low salinity relative to those with high salinity (Pierfelice et al. 2015). At longer timescales, tree mortality in high salinity forests may open the canopy, increasing net primary production (Fig. 3). Critically, not only do salt stressed forests have lower primary produc tivity, but even sub lethal salt levels can lead to weaker seedlings (Pezeshki et al. 1990), pointing to the possibility of a change in community composition due to insufficient recruitment (see section Community composition).
There is some evidence that salt stressed coastal for ests can return to background primary productivity levels if freshwater is returned to the system (Fig. 3). A managed freshwater pulse intended to keep oil from washing ashore after the 2010 Deepwater Horizon inci dent increased aboveground productivity in salt stressed Taxodium distichum swamps in the area (Middleton et al. 2015). These results indicate that properly timed fresh water pulses can mitigate the effects of SWI in impacted coastal forests and marshes, representing an important management strategy in systems with the appropriate hydraulic infrastructure (see section SWI Restoration Goals and Tools).

Community composition
In addition to changes in the productivity of particular species, SWI can cause wholesale shifts in community composition (Kaplan et al. 2010a). In coastal wetlands, plant community is largely structured by abiotic factors, including water level and tidal range, salinity regime, and soil biogeochemistry (Isacch et al. 2006, Mitsch andGosselink 2015). The response of plants to SWI is a func tion of the level, duration, and abruptness of exposure to saline water and varies widely across species (McKee and Mendelssohn 1989), but in general, salinity is the most important abiotic factor in determining coastal wet land habitat types (Lin et al. 2012). Under increased salinity and flooding regimes, freshwater wetland soils become more anaerobic and have higher interstitial sulfide concentrations (Flynn et al. 1995). Increased salin ity, decreased oxygen, and increased sulfide in these soils disfavor freshwater and brackish plants not adapted to these conditions (McKee and Mendelssohn 1989). Phy siological effects of flooding and salt stress include decreased stomatal response and reduced photosyn thetic rates; however, the magnitude of the effect is species-specific (Pezeshki et al. 1990).
The magnitude and duration of SWI events is a strong determinant of whether coastal wetland communities will recover or whether a different community assem blage will take its place. If SWI is temporary, impacted wetlands may recover, with recovery rates depending on post intrusion salinity and inundation (Flynn et al. 1995) and the presence of a propagule source to re populate the wetland, if necessary. If SWI is sufficiently frequent or extreme, the plant community will transition to a more salt-and/or flood-tolerant wetland type (Webb and Mendelssohn 1996), presuming that a source of more salt tolerant propagules is available (McKee and Mendelssohn 1989). For example, Spartina alterniflora will Fig. 3. Schematic representation of the impacts of saltwater intrusion (SWI) on coastal wetland structure and function. In many cases, wetland ecosystem services (ES) are degraded by SWI; however, the overall effect is a function of the spatial and temporal scale of the impact. This situation is illustrated by comparing (1) and (2), where (1) indicates loss of ES with SWI, followed by conversion to another wetland type with equivalent (though perhaps different) ES values, and (2) represents a continued trajectory of ES decline. Similarly, net primary pro ductivity (NPP) is reduced under SWI stress, but may follow several trajectories according to the magnitude and rate of SWI. These pathways are demonstrated by comparing (3), (4), and (5), where (3) represents conversion from one productive wetland type to an equally productive type over time, (4) rep resents conversion to a less productive wetland system, and (5) represents conversion to mudflat or open water. Finally, wetland species richness (SR) generally decreases monotonically with SWI disturbance, as the vegetated end member is likely to be salt marsh (dominated by monospecific stands of Spartina alterniflora) or mangrove forest (dominated by a small number of mangrove species).
colonize bare ground left behind after SWI in brackish and freshwater marshes (Sutter et al. 2015). Once estab lished, S. alterniflora is resilient to future SWI, allowing the system to transition to salt marsh. Similar directional transitions have been observed in coastal freshwater for ests where SLR and modified water management have caused plant community shifts to salt marsh and man grove ecosystems (Desantis et al. 2007).

Ecosystem services
Associated with these changes in productivity and com munity composition, SWI alters the types and amounts of ecosystem services (ES) provided by coastal wetlands (Fig. 3). Ecosystem services derived from coastal wet lands include fisheries production, carbon sequestration, coastal erosion/shoreline stabilization, tourism and recre ation, water quality, and biodiversity support (Blair et al. 2015). Understanding how the ecosystem services of coastal wetlands may change with increased SWI is criti cal for envisioning resilient coastal human communities and prioritizing funding for the management and resto ration of the coastal ecosystems (Rugai and Kassenga 2014); however, relatively few studies address this ques tion quantitatively. In a modeling study of coastal wet lands in Georgia, Craft et al. (2009) found that tidal freshwater swamps, tidal freshwater marshes, and salt marshes will decline in areal extent by as much as 34%, 39%, and 45%, respectively, under the maximum IPCC SLR scenario. Taken together, these changes will yield a reduction in nitrogen soil sequestration by 23% and reduce potential denitrification by 25%, a finding sup ported by Hines et al. (2015). Other ES negatively affected by SWI are carbon sequestration (Neubauer et al. 2013) and storm wave attenuation (Wamsley et al. 2010). It is important to note, however, that conversion from one wetland type to another may change ES provisioning without substantially degrading its overall value (Fig. 3).
One benefit of the ES approach is the ability to con vert ES value into economic terms through ES valuation (Costanza et al. 1989(Costanza et al. , 1997. Valuations put a price tag on consumptive uses, such as fisheries production, and nonconsumptive uses, such as carbon sequestration (Coen and Luckenbach 2000). In an ES valuation for the MRD, Batker et al. (2014) found that the system currently pro vides $12-47 billion of ES annually and investigated how that value would change under three scenarios, includ ing no restoration, prevent future land loss, and large scale restoration. They found that no restoration would result in a loss of $41 billion in ES. Efforts to prevent future land loss would prevent this loss, but not provide additional value. Large-scale efforts to increase the size of the MRD would prevent the $41 billion loss and pro vide an additional $21 billion of ES. While ES valuation methods are inherently uncertain (Johnson et al. 2012), these dollar amounts highlight the economic importance of coastal wetlands and lead us to a critical question for coastal resource managers: Should we proactively man age and restore coastal wetlands? If so, how do we set restoration goals, and what approaches are available and most likely to succeed? In the following section, we address these questions in the context of SWI and coastal water management.

SWI Restoration Goals and Tools
Deciding whether and how to pursue restoration inter ventions in coastal wetlands impacted by SWI requires the assessment of four interrelated questions: 1. What are the drivers of degradation? Are they natural, anthropogenic, or synergistic? 2. Which elements of ecosystem structure and function are degraded? 3. To what ecological condition do we seek to restore the ecosystem? 4. Can we modify the drivers of degradation? If so, what tools are available to do so?
The first two questions are addressed in sections Drivers of Saltwater Intrusion in Coastal Wetlands and Impacts of SWI on Coastal Wetlands of this review, which summarize the suite of SWI drivers leading to coastal wetland degra dation and subsequent impacts on wetland structure and function (Figs. 1-3). In this section, we aim to address the second two questions by briefly reviewing approaches for setting restoration goals and outlining the set of tools available for SWI management in coastal wetlands.

Setting restoration goals for coastal wetlands impacted by SWI
Restoration goals describe the abiotic states and biotic conditions that an ecological restoration effort attempts to achieve. Setting restoration goals is critical for deter mining project successes and learning from failures (Society for Ecological Restoration [SER] 2004), but can be problematic for coastal ecosystems impacted by natu ral and anthropogenic SWI. Particularly challenging aspects of restoration are (1) identifying an appropriate reference system to guide project success and (2) assess ing the vulnerability and/or resilience of restored sites to present and future environments. The selection of n reference system helps restoration planning by identifying one or more intact ecosystems that the restored site will emulate. Many restoration pro jects in North America explicitly or implicitly use "pre European conditions" as a temporal bound for their reference models (White and Walker 1997). This goal is idealistic, but impractical, particularly for coastal eco systems. Clearly, climate patterns, hydrology, fauna, and other critical drivers in coastal wetlands have changed in the past 500 yr so that using a "pre European" reference model is inappropriate. Instead of restoring historical conditions, Choi (2004) argues that we should restore for future conditions based on best available knowledge about projected climate, hydrology, and flora/fauna. This approach recognizes the need to quantify the dynamic range of variation in reference and restored systems. This "dynamic reference model" approach (Hiers et al. 2012, Kirkman et al. 2013) is particularly critical for restoring coastal wetlands impacted by SWI as it enables practi tioners to incorporate current and projected SLR and expected increases in storm frequency and intensity into restoration goal-setting efforts.
Restoration planning also requires an assessment of the resilience of the restored system to future environmental conditions (Harris et al. 2006). Ecological resilience is defined as the ability to recover structure or function following a disturbance (Lake 2013). Some coastal wet lands are thus resilient to SLR, but only up to a cer tain threshold (see section Drivers of Saltwater Intrusion in Coastal Wetlands), after which they will transition to another ecosystem type (see section Impacts of SWI on Coastal Wetlands). One critical, and often over-looked, element of restored ecosystem resilience is the ability of the restored area to support the regeneration of the species necessary for continued stability or development along a successional trajectory (SER 2004). For example, habitat suitability models based on adult tree survival will underpredict SWI impacts on coastal forests because they overlook life cycle requirements of seeds and seed lings (Kaplan 2010). It is also important to draw a con trast between resilience and resistance (i.e., the ability of a system to avoid damages) when considering restora tion of SWI impacted coastal wetlands. Many large scale coastal ecosystem restoration projects rely on heavi ly engineered solutions to control, limit, or even apply ecosystem disturbances. This "Command and Control" approach (Holling and Meffe 1996) aims to make sys tems more resistant to disturbance, but often ends up decreasing system resilience (Pittock and Finlayson 2013) by reducing systems' range of natural variation and their capacity to respond to future disturbances (Gunderson 2000, Hilderbrand et al. 2005. Reduced resilience can be particularly problematic for projects with large infra structural requirements (including some of the coastal water management tools we describe below) and can be a major limitation of complex restoration projects.

Freshwater management to mitigate the impacts of SWI on coastal wetlands
With explicit restoration goals set and an understanding of the likely resilience of restored systems, practitioners may then explore restoration approaches to ameliorate SWI. The potential for improved freshwater manage ment to mitigate the worst effects of SWI is based on the strong inverse relationship between river discharge (i.e., freshwater flow) and salinity in coastal rivers and estuar ies (Kratzer and Grober 1991, Kaplan et al. 2010b). Several authors have leveraged this flow-salinity relationship to explain observed changes in coastal ecosystem function and in support of coastal ecosystem restoration. For example, there was a significant decrease in salinity dur ing periods of high freshwater flow from the Mississippi River in Louisiana and observed improved ecosystem function during this time (Middleton et al. 2015). This relationship has also been used to develop restoration flow thresholds for coastal wetlands impacted by SWI (Kaplan 2010) and to improve freshwater flow manage ment to mitigate severe oyster reef decline tied to SWI (Seavey et al. 2011, Kaplan et al. 2016. In short, upstream flow management may be of fundamental importance in the fight against SWI. Efforts to ameliorate SWI fall into three overarching categories: (1) modified groundwater and surface water withdrawals, (2) modified groundwater and surface water deliveries, and (3) the use of engineered structures. All approaches rely to some extent on the use of models to develop alternative management scenarios. While our focus is on SWI in coastal wetlands, we also consider SWI in coastal aquifers. In both cases, the strong relationship between freshwater quantity and salinity makes infer ences drawn from either type of study useful for under standing how freshwater management can be improved to ameliorate SWI impacts.

Modified water withdrawal
Modified surface water and/or groundwater withdraw als are changes to water use that seek to prevent or min imize SWI and include (1) optimization of the amount, location, and timing of withdrawals and (2) conservation efforts to reduce the magnitude of water diverted from natural systems (Table 1). In the first category, a large body of work has advanced groundwater pumping opti mization schemes that reduce the risk of aquifer SWI (e.g., Willis andFinney 1988, Zekri et al. 2015). Also, see reviews in Bear et al. (1999), Cheng et al. (2004), Dhar and Datta (2007), and Werner et al. (2013). These efforts cou ple spatially distributed groundwater hydrology models with optimization methods to meet a specific manage ment goal (e.g., maximize pumping rate and minimize saline intrusion; Dhar and Datta 2009). While these stud ies usually focus on the protection of aquifers, the same approach can be used to protect and restore coastal wet lands by including objective functions that address eco system requirements (e.g., maintaining freshwater for phreatophytic vegetation or sustaining fresh groundwa ter delivery via springs or diffuse seepage). A numerical model was developed to understand how past anthropo genic activities and SLR have combined to cause SWI into coastal freshwater forests in the Po River Plain in Italy (Giambastiani et al. 2007). Model results suggested that reduced groundwater withdrawals during periods of low recharge would be required to reduce SWI into this ecologically and historically important ecosystem. Similarly, groundwater modeling on the Kona Coast of Hawai'i was used to balance water use and submarine groundwater discharge, which supports important near shore ecosystems (Duarte et al. 2010). Further applica tion of the SWI modeling approaches in coastal aquifers is an important area of study for the management and restoration of coastal wetlands.
Similar optimization approaches balance the amount and timing of water withdrawals from surface water with the needs of coastal ecosystems. For example, Nobi and Gupta (1997) developed a SWI model for a coupled coastal stream-aquifer system in southwest Bangladesh to demonstrate how planned increases in groundwater pumping would cause undesirable increases in estuarine salinity. Model results showed that prevention of estu arine SWI was possible, but only if plans for expand ed groundwater pumping were abandoned or if fresh surface water was diverted from another nearby river. Optimization modeling was also applied to allocate sur face and groundwater resources to balance social, eco nomic, and environmental demands in the Pearl River Delta in China, which is experiencing SWI (Liu et al. 2001). The model developed by these authors could be used to propose modified surface water withdrawals during the dry season to limit SWI. In this case, howev er, the authors proposed water system engineering pro jects (reservoir storage, river diversions, and distributed pumping stations) to satisfy cumulative water demand.
In the examples, above, models were used to optimize the magnitude and timing of water withdrawals while limiting SWI. Water conservation schemes, in contrast, aim to reduce consumptive use across one or more sectors, leaving a larger proportion of freshwater avail able for natural systems (or other sectors). Mirroring the discussion above, the literature describing water conser vation is expansive (Anderson 2003, Inman and Jeffrey 2006, Pujol 2013), but only a subset of studies address efforts directed specifically to limit SWI. For example, Laraus (2004) points to expansion of irrigated agriculture in driving widespread SWI in the southern and eastern Mediterranean and proposes reductions in agricultural water usage by using drought tolerant crops and rain fed production systems. In coastal North Carolina, rain water harvesting systems provide an alternative irrigation water source and improve aquifer recharge in urban areas, potentially helping to manage SWI that threatens aquifers and ecosystems in that region (DeBusk et al. 2012). Simple conservation efforts to reduce the impacts or likelihood of SWI have been developed in tourist areas where surface and groundwater use outpace supply and/ or recharge (Gössling 2001, Kent et al. 2002. While direct study of the impacts of water conservation practices on SWI is limited, opportunities for improved conservation are abundant across the agricultural, municipal, and industrial sectors (Anderson 2003, Low et al. 2015 and have the potential to beneficially augment freshwater flow in coastal systems impacted by SWI.

Modified water delivery
While modified water withdrawals seek to prevent or minimize SWI, modified water deliveries usually aim to treat existing SWI symptoms. Several studies and projects have sought to develop coastal wetland management or restoration plans by linking water management, water quality, and habitat suitability or threshold models to identify beneficial environmental flow regimes in sys tems impacted by SWI. Coupled models have been used to provide guidance on the timing and magnitude of freshwater flow required to restore coastal floodplain forests impacted by SWI (Kaplan 2010) and to set flow augmentation goals for restoration of the Florida Everglades (Koch et al. 2015). Regarding resilience, Koch et al. (2015) showed that enhanced freshwater flows Table 1. Water management strategies, approaches, and tools used to ameliorate the impacts of saltwater intrusion in coastal systems.

Strategy
Approach Tools References

WHITE AND KAPLAN Saltwater intrusion in coastal wetlands
Volume 3(1) v Article e01258 Ecosystem Health and Sustainability would enhance mangrove peat accumulation, helping these coastal wetlands keep up with SLR and serving to stabilize coastlines and limit SWI. In a novel study of SWI mitigation in the Pearl River estuary in southern China, the design of a coastal wetland network was mod eled to optimally store and convey freshwater runoff to the system to reduce river and estuarine salinities during periods of SWI .
In addition to augmenting surface water flows, increased freshwater delivery to coastal systems impact ed by SWI can also be achieved by increasing recharge. Enhanced recharge may be achieved via changes at the land surface that increase the proportion of precipitation that infiltrates and is then available to surface water or groundwater resources (McLaughlin et al. 2013), or via "artificial recharge" to pump water into the subsurface (Todd 1959). Land surface changes include modifica tions of surface topography, land use/land cover (LU/ LC), or the soil profile to facilitate infiltration. These changes include construction of infiltration ponds and basins (Dillon 2005), LU/LC conversion to more perme able cover types (Scanlon et al. 2005), and soil manage ment to maximize infiltration rates (O'Leary 1996). For example, groundwater modeling was used to assess the coupled use of infiltration ponds and pumping reduc tions to freshen salinized aquifers and restore native coastal dune vegetation in the Netherlands (Walraevens et al. 2002) and to minimize SWI in aquifers in Australia (Narayan et al. 2003). Where "enhanced recharge" strat egies use modification of the surface environment, "arti ficial recharge" is the direct transfer of water below the surface to increase aquifer storage for later recovery or to serve as a hydraulic barrier to intruding seawater from pumping further inland (Werner et al. 2013, Maliva 2014, Asano 2016. Taken together, enhanced and artificial recharge strategies are collectively referred to as "man aged aquifer recharge" (Dillon 2005). This approach can be used to protect both aquifers and coastal ecosystems from SWI, particularly where coastal ecosystems rely on groundwater discharge (Duarte et al. 2010).

Engineered structures
Engineered structures are used widely around the world to modify the delivery of surface water and groundwater in systems impacted by SWI. Engineered solutions can be very effective at preventing or limiting disturbances, though this added protection can decrease system resil ience (Pittock and Finlayson 2013) by making the system less adaptable to future disturbances (Gunderson 2000). Structures can also be expensive to design, build, and maintain. Structures reviewed here include freshwater diversions, saltwater barriers, dams, and other hydrolog ical connections and cutoffs (Table 1).
Freshwater diversions are structures that connect, or re-connect, coastal floodplain wetlands to their riv er channel to mimic historical flooding and sediment delivery. Hydrodynamic and water quality assessment models can assess the potential for freshwater diversions to be implemented in support of coastal wetland resto ration goals in the MRD (RTMRD 2016). The results of studies to assess the potential for these structures to slow wetland loss have been mixed. Increased sediment deliv ery and salinity reduction in the wetlands of the northern Breton Sound Basin were observed after the opening of the Caernarvon Diversion (DeLaune et al. 2003), sug gesting that the diversion has the potential to slow down or reverse wetland loss trends. Snedden et al. (2007) found that this structure provided the largest source of sediments to the sinking wetlands, but that sediment delivery was far lower than historical values and insuf ficient to support marsh accretion relative to SLR. These authors note that the diversion's primary design goal is to maintain optimal salinity for shellfish production, lim iting the amount of freshwater discharge and associated sediments allowed through the diversion. These studies point to the potential utility of diversions in combatting SWI, but also highlight the challenge of using engineered structures to support multiple ecological functions and stakeholders simultaneously.
The MRD is also home to the Davis Pond Diversion (DPD), which is the largest diversion in the world (Das et al. 2012). The DPD connects the Mississippi River to the Barataria Estuary, a complex of lakes, bays, and over 200,000 ha of fresh, brackish, and saline wetlands. Like the Caernarvon Diversion, the DPD was designed to counteract SWI and may reduce site-specific salini ty in associated coastal freshwater forests (Middleton et al. 2015). This potential was investigated by Das et al. (2012), who found that despite the large volumes of water passing through the DPD, its effects on salinity were limited. In the upper and lower estuary, salini ties were primarily influenced by upstream freshwater and downstream marine waters, respectively, limiting the effect of the DPD. In the central region, diverted flow volumes strongly affected predicted salinity, with modeled differences as high as 10 PPT, suggesting that management of flows through the DPD can be used to manage for specific vegetation communities in specific places. Additional reviews of the current efficacy and future potential for freshwater and sediment diversions in the MRD can be found in Allison and Meselhe (2010), Paola et al. (2011), andKemp et al. (2014). In general, these studies argue for the utility of large diversions to benefit marshes and suggest continued monitoring to better understand the potential and limitations of this engineered approach.
While the MRD is the world's testing ground for large-scale flow diversions, the technique has also been employed in other regions. Fresh, brackish, and saline floodplain wetlands on the Yellow River (China) severe ly degraded by SWI are being restored using freshwa ter flow diversions to re-connect wetlands to the river channel (Cui et al. 2009). The diversion yielded increased wetland hydroperiods, improved surface water and soil salinity conditions, increased soil nutrient and organic matter, expanded and more diverse vegetation, and an increase in avifauna use relative to the pre restoration landscape (Cui et al. 2009). A similar project was imple mented on the Nueces River in Texas, where freshwater flow reductions had caused severe SWI in the river and its associated wetlands and delta (Ward et al. 2002). River diversions constructed as part of this project increased freshwater flow to the upper Nueces Delta by 700%, restoring a more natural salinity gradient, with positive effects on the abundance and diversity of wetland vege tation and benthic communities.
While river diversions are designed to encourage flow of freshwater and sediments into SWI impacted areas, saltwater barriers are designed to prevent upriver trans port of saline water. Saltwater barriers include gates, dams, dikes, levees, and other structures that physically block the upstream flow of saline water and have been used in river systems around the world. For example, nat ural resource managers have built concrete and rock "bar rages" and earthen blocks to prevent upstream SWI in the Lower Mary River (Northern Territory, Australia) during the dry season and reduce floodplain drainage rates dur ing the wet season (Applegate 1990). These structures have been generally successful in halting the further upstream advance of SWI, which has impacted nearly 20,000 ha of freshwater wetlands over the past 70 yr (Mulrennan and Woodroffe 1998). In an update to this study, Miloshis and Fairfield (2015) developed a scoring system to rank the effectiveness of "traditional" (i.e., engineered) vs. "ecoengineering" and "do nothing" approaches to manage coastal wetlands in Australia's Northern Territory coastal wetlands. They found traditional engineering approach es (including barrages or levees), were "neither required, nor beneficial" for ecosystem management and that "ecoengineering" (conservation and gentle rehabilitation) was the best management response (discussed further in section The Path Forward: Restore or Retreat?). The "Chiaro Pontazzo" water management approach described by Giambastiani et al. (2007) is another example of an engi neered artificial embankment. In this case, the structure creates a shallow freshwater "lake" that prevents SWI into a coastal pine forest.
Dams, gates, and other types of mechanical barri ers block saline waters in major river systems affected by SWI, though in many cases these are primarily used to reduce flood damage by storm surge in urban areas. While construction of saltwater barriers in coastal rivers is primarily for the protection of drinking and agricul tural water supplies, we were unable to find published studies describing their number, distribution, or impacts on water quality and ecosystem integrity. For some engi neered structures, it is their removal that best restores natural salinity distributions in coastal wetlands. For example, Yang et al. (2010) investigated the impact of breaching or removing dikes to restore estuarine and coastal ecosystem function. Modeling results suggested that dike breaching would restore hydrology and salinity in support of four habitat types and substantially increase the area flushed with freshwater. Moreover, by inference, the large number of studies demonstrating how existing dams drive or exacerbate SWI (Ge et al. 2015, Hutchinson 2015, Webber et al. 2015 illustrate the potential for salt water barrier removal to alleviate this effect. Finally, we note that where optimizing delivery of existing flows, enhancing recharge, and deploying engi neered structures is insufficient to manage SWI, it may be possible to develop new freshwater resources. Several of the examples above implicitly or explicitly include suggestions for new resource development, such as res ervoir construction (Kaplan 2010), river diversions and other engineered solutions (Nobi andGupta 1997, Liu et al. 2001), and non point source recharge (McLaughlin et al. 2013). Clearly, water is neither created nor destroyed in the global hydrological cycle, so in these cases, it is important to understand not only how the diverted resource reduces SWI stress in the affected system, but also the potential for detrimental effects in the systems from which these "new" resources were diverted.
Other strategies and "best practices" to support coastal wetland restoration While this review focuses specifically on SWI drivers, effects, and management, there are several other impor tant tools and methodological approaches that support coastal wetland restoration and management. Here, we provide an initial set of references for further investiga tion in each of these areas. Efforts to facilitate sediment delivery and accretion are central to coastal marshes "keeping up" with SLR. Reviews of sediment dredging, diversions, and other modes of delivery are summarized in Ford et al. (1999), Day et al. (2005), and Tong et al. (2013). In some cases, facilitated or assisted migration of plant and animal species may be necessary to support coastal adaptation in areas with substantial built infra structure (Smith and Lenhart 1996, Doyle et al. 2010, Minteer and Collins 2010, Dawson et al. 2011. In other cases, the "do nothing" option may be preferable (Burton 1996, Dolan and Walker 2006, Miloshis and Fairfield 2015, allowing for managed decline and transition via ecosystem self-organization (Odum 1989). In all cases, restoration ecologists, ecosystem managers, and other stakeholders may seek to balance responsive behaviors (i.e., restoration) with proactive behaviors (i.e., improved management and direct climate mitigation) to meet this global challenge (Harris et al. 2006).
We also identified many studies that aimed to summa rize "best management practices" for wetland manage ment in the face of climate change (Table 2). Early reviews highlighted the importance of increasing protections for healthy ecosystems and removing existing non climate stresses from degraded systems to enhance the potential for these ecosystems to adapt to non stationary conditions (Burkett andKusler 2000, Erwin 2009). These reviews also suggested land management and policy actions, including developing setbacks to allow for SLR and hab itat transition, building water and sediment diversions, restoring connectivity between fragmented wetlands and waterways, securing water allocations for wetlands, using water control structures where needed to secure particu lar functions, and restoring degraded wetlands. Many of these recommendations directly address the issues of SWI, and applications of these practices are discussed in the sections above. To make coastal human communities more adaptable to future conditions, these practices must be combined with improved monitoring, training, and education about coastal systems (Erwin 2009).
More recent reviews advocate for several broad goals, including prioritizing integrative and resilience focused management as opposed to schemes that seek to resist or avoid disturbance (Koch et al. 2015). This more holistic paradigm considers the coupled terrestrial-freshwatermarine ecosystem and seeks to develop comprehensive regional/local governance and planning frameworks (Koch et al. 2015). Given the constraints of widespread development, environmental degradation, and a non stationary climate, Wiens and Hobbs (2015) advise eco system managers and restoration practitioners to set realistic goals and advocate the use of the dynamic refer ence concept (Hiers et al. 2012) to help do so, particular ly in heavily altered systems. This approach stresses the importance of embracing uncertainty, using the adaptive management framework to adjust goals and objectives if necessary, and enlisting and maintaining public support during this dynamic process (Wiens and Hobbs 2015).

The Path Forward: Restore or Retreat?
In this review, we have shown that where SWI is driven by anthropogenic and/or synergistic drivers, some amount of coastal wetland restoration is likely achievable. The ques tion remains: Is restoration advisable given the number of SWI drivers described in section Drivers of Saltwater Intrusion in Coastal Wetlands? The choice between restora tion and retreat is not a simple dichotomy. The preferable action will depend on local physiographic setting, the spe cifics of SWI drivers and impacts, desired ecosystem ser vices, and economic considerations. No single "cookbook" or if-then scenario analysis tool is sufficient to support decision-making for a specific situation. However, in many cases, the best option may be a compromise between restoration and retreat that maximizes the value of ES pro vided by coastal wetlands while minimizing invested time, money, effort, and risks to human health and the built environment. Performing such an analysis requires an ES valuation of the restored and unrestored systems (section Ecosystem services), coupled with an assessment of the time, energy, and money required to pursue restora tion. The magnitude of this investment will be driven by the size and proportion of a site that is degraded, the eco logical "distance" from the desired state (SER 2004), and the actions required to achieve this state.
From an economic perspective, it makes sense to pur sue restoration if the marginal value of the ES provided by the restored ecosystem relative to the unrestored case is greater than the cost of restoration, monitoring, and future adaptive management (i.e., if the benefit-cost ratio [BCR] is >1; Gregory et al. 2006). Retreat is preferable if the opposite is true. In the real world, decision making is unlikely to be this clear cut. Some issues that may con found the process include differing objectives of stake holders, valuations of a given ecosystem service, levels of knowledge regarding science and terminology. Despite these complications, framing the decision in monetary terms can help open the discussion to a broader range of stakeholders (Brauman et al. 2014). Whether a specific system should be restored can be evaluated by considering both the BCR of restoration actions and the site's current and future susceptibility to SWI. From an ES perspective, BCR is defined as eco system services value gained relative to restoration and maintenance costs (Barendregt et al. 1992, Turner et al. 2000, Dubgaard 2004. While selection of a thresh old BCR to justify restoration is arbitrary, values ≫1 indi cate that restoration is advisable and values ≪1 suggest that retreat makes more sense. At BCRs closer to 1, the decision to restore or retreat is less clear and may be strongly driven by the issues listed above (largely social), along with other site-specific environmental considera tions. For example, restoration of sites with BCR values <1 may be appropriate if restoration activities are tied to the protection of threatened and endangered species, historical/cultural sites, or other non use values, which

Authors
Best Practices Burkett and Kusler (2000) Increased protection/remove stresses Develop setbacks Sediment diversions Link fragmented wetlands and waterways Use water control structures to enhance particular functions Secure water resources for wetland conservation Wetland restoration Erwin (2009) Significantly reduce non climate stressors Protect coastal wetlands and accommodate SLR (acquisition, setbacks, restoration) Monitoring, training, and education Incorporate climate oscillations Medium and long range planning: strategize conservation priorities Koch et al.
Integrative & resilience focused management Paradigm that considers coupling of connected terrestrial, freshwater, & marine ecosystems Develop comprehensive regional/local governance and planning frameworks Wiens and Hobbs (2015) Frame realistic and complementary goals Embrace uncertainty Enlist public support are not necessarily valued as highly in ES assessments (Laurila Pant et al. 2015).
Threshold BCR values may also be influenced by current and future susceptibility to SWI of a particular site. Current and future SWI impacts include predict ed changes in SLR projections, regional climate, future groundwater abstractions, and land use, all of which could lower groundwater recharge or modify freshwater flows. As SWI susceptibility increases, a higher BCR may be required to justify restoration activities given a higher risk and greater potential for future failure. The assess ment of a system's susceptibility to SWI requires explic it definition of a time window (e.g., 30 yr in the future) over which to assess BCR. While not a precise methodol ogy, this approach can be used by decision makers and restoration practitioners to prioritize restoration projects that maximize return on investment and minimize risk.
We close by noting that while changes in global climate that drive SLR and rainfall availability are beyond the scope of local control, choices about the delivery of fresh water to the coast in support of resilient coastal ecosys tems are not. This review points to the variety of technical approaches that can be used in these efforts; however in all cases, water management decisions come down to choices and compromises about conservation, allocation, and infrastructural investment. Where this decision making framework is codified in environmental and water law, communities are likely in a better position to balance water allocation with natural resource protection. For example, Florida state law mandates development of minimum flows to prevent harm to natural resourc es (s. 373.042, Florida Statutes), and this framework has been used to quantify flow requirements in coastal rivers to prevent SWI into coastal wetlands. While the science of "environmental flows" has advanced greatly in recent decades (Tharme 2003), implementation of environmen tally protective flow recommendations via explicit legal frameworks and enforcement of these laws lags behind (Kiwango et al. 2015). As stressors on coastal wetlands continue to increase, wider application and enforcement of environmental laws, which allocate a portion of the water budget to the natural environment, will be critical for maintaining or restoring a resilient coastal environment.