The estuaries of the northern Gulf of Mexico, United States, are nowhere as extensive as in the Louisiana coastal zone, where an area of 56,000 km² supports productive commercially and recreationally important species (Keithly and Roberts, 2017). This region is also suffering the highest rates of wetland loss in the United States due to coastal erosion and relative sea level rise (SLR); the amount of land already lost between 1932 and 2016 is close to 5,000 km² (calculated using data from Couvillion et al., 2017). Projected accelerated rates of SLR are expected to increase this loss in the future (IPCC, 2014). To mitigate this decline, coastal protection and restoration projects are being planned and implemented. The 2017 Louisiana Coastal Master Plan (CMP) is the latest plan in an adaptive management approach that provides a list of projects that build or maintain land and protect coastal communities (CPRA, 2017). The restoration measures that are implemented or proposed to be employed in the CMP have been selected using an extensive simulation process, using multiple models that determine the effect of all projects on, e.g., land building, water quality, vegetation growth, and coastal flooding, under three different future scenarios that vary SLR, hurricane frequency and intensity, precipitation patterns, and subsidence (rate of land compaction and sinking). Most models that support the CMP are actively linked within one framework, the Integrated Compartment Model (ICM; CPRA, 2017). The ICM simulates the effect of several proposed restoration and risk reduction projects, with a focus on building and maintaining land. Additional models during the plan development process were added to examine the effects of flood risk reduction projects. A planning tool that made use of simulation output—including economic considerations—informed the decisions of which restoration and protection projects to include in the CMP (CPRA, 2017).

Any changes in the coastal area have the potential to change the habitat suitability of the fish and shellfish that live in that area. We developed a new ecosystem model and coupled that to the ICM using an automated one-way coupling approach, which allowed us to simulate effects of the restoration and protection projects on the biomass and distribution of living marine resources within the Louisiana coastal area. ICM outputs used as environmental drivers include salinity, total Kjeldahl nitrogen (TKN), temperature, total suspended solids (TSS), and percent cover of marsh. With these linked models, we addressed the following research questions: (1) How does the biomass and spatial distribution of economically and ecologically important fisheries species change with the implementation of the Louisiana CMP as compared to a future without action (FWOA)? and (2) How does SLR affect these species in a future with and without restoration over 50 years?

The ecosystem model developed to this purpose is an Ecospace model built within the Ecopath with Ecosim (EwE) software, representing the local food web with 55 groups. EwE is an open source ecosystem modeling software, originally developed by Polovina (1984) to model trophic interactions and to estimate mean annual biomass on a Hawaiian coral reef ecosystem. The Ecosim and Ecospace components were developed later (Walters et al., 1997, 2000). The utility of the model has expanded from investigating changes in mean annual biomass of fish species to operational ecosystem management approaches such as investigations into combined effects of fishing impacts and environmental drivers (Coll et al., 2016; De Mutsert et al., 2016; Hernvann et al., 2020) and marine protected area placement and environmental drivers (Dahood et al., 2020), environmental impact studies (Raoux et al., 2018), and spatial–temporal modeling for marsh restoration effects on fish species (De Mutsert et al., 2017).

To identify how the restoration and protection projects selected for the 2017 CMP differentially affect living marine resources from changes occurring over 50 years under different rates of SLR, an FWOA scenario was simulated as well. Both the CMP and the FWOA were simulated under three future SLR projections (Meselhe et al., 2017); here we show output from the “high” and “low” SLR projections (details of these two projections are provided in section “Materials and Methods”). By directly linking the models through an executable version of Ecospace (the EwE Console), we evaluate effects of proposed projects on biomass and distribution of living marine resources within the estuaries under two different SLR scenarios. This information provides managers and other stakeholders with a fuller picture of trade-offs that are part of the decision-making process, a better understanding of what to expect when selected projects are implemented, and estimates how different SLR projections can affect the outcome.

The changes resulting from plan implementation are species-specific and basin-specific (Figure 5). In general, the eight species of interest we focus on had more biomass at the end of the 50-year simulation when the Coastal Master Plan was implemented as compared to an FWOA. We focused our comparisons on the juvenile life stages with two exceptions, because the species highlighted make most use of the estuaries of the Mississippi River Delta during that stage in their life. The exceptions were eastern oyster, for which we chose to display the sack oysters (adults) that occur within the model area, and spotted seatrout, another species that occurs in the estuaries during their adult life stage.

Since we are comparing model output with the two SLR scenarios as “samples” (see Supplementary Table S6 for description of SLR scenarios), sample size for the Wilcoxon rank test was small and did not provide enough power to return significant differences. Rather, the SLR scenarios should be viewed as a visualization of the level of uncertainty in the response of living resources to CMP implementation in the face of uncertain future climate. The difference between biomass under the low and high SLR scenarios visualized with spatial distribution maps reveals the large effect the SLR scenario chosen has on the outcome (Figures 6–14). We will point out the difference in model output with and without plan implementation, but since we lack sufficient model output samples, we are unable to determine whether these differences are significant.

Future rates of SLR appear to impact plan implementation effects on biomass and distribution of fisheries species (Figures 6–13). In other words, SLR adds uncertainty in the predictions of CMP effects. The SLR scenario simulated also has an influence on fish biomass and distribution in an FWOA. To evaluate what effect plan implementation has under different rates of SLR, the biomass difference in year 50 between CMP and FWOA over all basins under the two SLR scenarios was compared (Figure 14). With positive values indicating that more biomass is present in year 50 under plan implementation that in an FWOA, only blue crab under the high SLR scenario shows an average lower biomass with plan implementation than in an FWOA. All other species responded positively to plan implementation. For blue crab, black drum, Gulf menhaden, and largemouth bass, the effect of plan implementation is larger under the low SLR scenario than under the high SLR scenario. For brown shrimp, eastern oyster, spotted seatrout, and white shrimp, the effect of plan implementation is larger under the high SLR scenario than under the low SLR scenario.

The Louisiana CMP is designed to mitigate coastal wetland losses and habitat changes that are expected due to both anthropogenic and natural causes (CPRA, 2017; White et al., 2017), which in turn affects the biomass of living marine resources. This reduction of habitat loss due to plan implementation is more successful under low SLR scenarios, but still noticeable under high rates of SLR (White et al., 2019). How much the climate will change is unknown; therefore simulations of two scenarios have been compared here, one considered on the low end of change in SLR and subsidence and one considered on the high end. The complications of SLR rate scenarios, and how they will be impacted by new knowledge (e.g., of ice sheet melt dynamics), and by climate change mitigation and emission reduction efforts at the global scale, are a continuously evolving and advancing area of scientific study¹. Research conducted since the development of these scenarios is exploring the relative likelihoods of these (and others) SLR scenarios (Kopp et al., 2014; Sweet et al., 2017); however, it still remains difficult to project, with any level of confidence, exactly which SLR scenario will be the closest to future sea level trends in coastal Louisiana; a factor that is further compounded by subsidence rates. What is clear is that future sea levels in coastal Louisiana will result in changed conditions that will only be partially mitigated by an extensive restoration plan. Restoration and mitigation alone will not prevent the loss of coastal habitat for living marine resources. Climate change is a global process and should therefore be considered when developing ecosystem restoration plans. In fact, Busch et al. (2016) outlines a Climate Science Strategy for the US National Marine Fisheries Service (NMFS) and directly calls for “robust projections of future ocean conditions and the likely responses of ecosystems, LMRs, and human communities on appropriate temporal and spatial scales.” The work herein is an exciting step to aligning with the plan put forth by the NMFS.

Planned restoration projects that will have the most effect on the fish and shellfish in the Mississippi River Delta (our focus area) are large sediment diversions and marsh creation projects. These projects are expected to increase habitat as compared to an FWOA by mitigating wetland loss, increasing sediment and nutrient loads by diverting Mississippi River water into the estuaries, and reducing salinity for the same reason. While the land-building capacity of river diversions has been shown in various projects (Day et al., 2018), stakeholders are concerned about potential negative impacts to fish and fisheries (Barra, 2020), mostly due to salinity reductions (Das et al., 2012). Explicitly simulating effects of planned restoration projects on living resources in the estuaries as was done in our study provides an estimate of the effects and impacts of these projects on living resources. One species of concern is eastern oyster, which will be affected by lowered salinities (Soniat et al., 2013). We have put much care into the simulation of oyster response to environmental changes by including OECLs (De Mutsert et al., 2017); a different habitat suitability subroutine than the habitat capacity model available in Ecospace that reflect effects of environmental change on daily rather than monthly time steps. We found similarities in our oyster distributions to Soniat et al. (2013); suitable locations for eastern oyster are at the marine end of the lower estuaries and depend on the presence of hard substrate. Perhaps surprisingly, we found that eastern oyster experienced higher biomass under plan implementation than in an FWOA in year 50 of the simulation. This does not mean oyster biomass increased over the 50-year simulation with plan implementation; oyster biomass in both the FWOA and CMP scenarios decreased over the course of the simulation under the low SLR scenario. The distribution maps show that higher rates of SLR might actually befits oysters; plan implementation under high SLR resulted in highest biomass, which is due to a combination of increased nutrients and increased salinity.

Our results indicate that the effects of the combined projects in the 2017 Coastal Master Plan in the Mississippi River Delta are species-specific and basin-specific. However, the general trend is that the differences in species’ biomasses are small, with mostly slightly higher biomass with plan implementation versus an FWOA. The fact that an FWOA does not result in more biomass of living resources in a 50-year simulation indicates that the need for estuarine habitat at some stage of each of these species life histories outweighs the costs of lowered salinities. In addition, these species are all estuarine species with a salinity optimum below full seawater (35); the saltwater intrusion as result of relative SLR may increase the salinity too much for these species in this area. Results from White et al. (2018), for example, indicate that the existing freshwater diversion in Barataria Basin may be inadequate in maintaining present-day salinity dynamics under the same high SLR scenario as we used in this study. One example of this in our results is the decline in brown and white shrimp biomass over time, with biomass reduced in the Chandeleur Sound (at the marine end of Breton Sound and lower Pontchartrain Basin). This may be a result of saltwater intrusion due to SLR, and outside of the effect area of restoration plans.

The positive effects on the biomass of living resources with plan implementation are encouraging since the plan is focused on mitigating land loss and is not explicitly designed to benefit fisheries species. In addition to mitigating the loss of fish habitat, this effect can be explained by the increased nutrients as a result of diverting Mississippi River water into the estuaries as part of the restoration efforts, represented in the model as TKN driving primary productivity (Supplementary Figure S2). Fish and fisheries generally show strong responses to changes in nutrient loads (Breitburg et al., 2009). Our results show that especially for Gulf menhaden, which has biomass concentrated around nutrient-rich areas because their phytoplankton-centric diet links them closely to nutrients. This concentrates Gulf menhaden juveniles around the Birdsfoot Delta in an FWOA, since that is the main source of nutrients in that scenario, and suitable habitat seems to disappear elsewhere. Under plan implementation, suitable areas for Gulf menhaden juveniles expand into the wetlands of Barataria Bay and Breton Sound because of increased nutrient levels in response to large river diversions. Since nutrient changes elicit such a strong response, a useful addition to a suite of models such as the ICMs would be a biological model that explicitly simulates phytoplankton growth in response to all physical and biogeochemical changes occurring in the model scenarios, such as presented in Baustian et al. (2018), and potential formation of hypoxia with high nutrient loads, such as presented in Fennel et al. (2011). While the current drivers of change in the model are salinity, nutrients, and habitat, future improvements to the Ecospace model should include accounting for effects of increased sediments loads on phytoplankton, fish, and shellfish, effects of flow increase on these organisms during diversion openings, and effects of hypoxia.

A previous study in the Mississippi River Delta with a similar scope and objective found that the negative effects of the salinity reductions as a result of large sediment diversions outweighed the positive effects of reduced marsh loss on estuarine fish and shellfish species (De Mutsert et al., 2017). The results showed a redistribution of species to more favorable areas rather than overall biomass loss. The current study found that for most species, the biomass was higher with implementation of restoration projects, of which some reduce salinity, than in an FWOA. De Mutsert et al. (2017) and the current study were linked to different models delivering the environmental driver data, and used different Ecospace models to simulate the effects on living resources. The current study uses ICM projections with future scenarios that include SLR at two different rates, while the model used to drive the De Mutsert et al.’s (2017) study did not include environmental changes through subsidence and/or SLR (Baustian et al., 2018). Without the inclusion of SLR, the amount of marsh loss and saltwater intrusion under an FWOA is likely underestimated. Because of that fact, the difference in available juvenile habitat between an FWOA and coastal restoration is smaller, and the salinity reduction through river diversions is stronger. A second reason for the difference is that phytoplankton was explicitly modeled in Baustian et al. (2018), while this was not the case in the ICM (White et al., 2019). De Mutsert et al. (2017) reported a reduction in secondary biomass close to diversion outfalls because high sediment load reduced phytoplankton growth, even at high nutrient levels. Without phytoplankton explicitly modeled in the ICM, nutrients are used in the current study to drive primary production in Ecospace, which does not account for changes in light availability to phytoplankton. Therefore, the primary production and secondary production near the outfall of large sediment diversions or the Mississippi river mouth are likely overestimated in the current study. Despite the differences in the use of models and results, neither study found strong negative impacts on living resources of coastal restoration projects designed to mitigate wetland loss. Both studies found small effects on biomass that were species-specific and basin-specific. These basin-level differences could help inform localized management responses to the changes estimated by the models. By providing spatial outputs (maps) with these studies rather than line graphs, estimates of spatial redistribution could provide resource managers with basin specific knowledge to implement the most appropriate restoration strategy in that area.

By comparing June and October, we compared a month after a period of high diversion flow (June) and a month after a period of diversion closure (October) under the simulated diversions operation plan. By considering these results, recovery of high salinity species from a period of low salinity can be evaluated. We found that the biomasses observed in both June and October were similar between a number of the target species (e.g., brown shrimp, blue crab, spotted seatrout, and black drum). This outcome may mean that the magnitude of the diversions maintains a new status quo of an estuarine nekton community structure and distribution that is tolerant of lowered salinities. However, our simulations do not explicitly address the effect of salinity changes on spawning success or larval recruitment to the estuaries of fish and shellfish, which could be affected by spring and summer diversion opening (Peyronnin et al., 2017). The operation schedule and thereby flow regime are not finalized yet, and discussions have pointed to winter openings being potentially more optimal (Peyronnin et al., 2017). The June output can be more broadly interpreted as conditions when the large sediment diversions are open, and October as conditions when the large sediment diversions are closed.

The simulations show species-specific and location-specific effects which are a result of local biological factors. For example, the spatial distribution maps of brown and white shrimp show that most of the shrimp biomass is at the marine end of the lower estuaries, which may be why plan implementation or SLR effects such as wetland loss have less impact than on other species. Das et al. (2012) in their study in the same area concluded that river diversions strongly affect salinities only in the middle section of the estuary, so spatial distribution of organisms and their overlap with environmental change on small scales affect the species-specific outcomes. While diversions do not affect salinities much in the upper estuary (since it is already fresh; Das et al., 2012), they can affect the biomass of upper estuary low-salinity species. Salinity is likely the main driver for higher largemouth bass biomass with plan implementation as compared to an FWOA, due to expansion into the mid-estuary with the opening of the large sediment diversions under CMP. Busch et al. (2016) explicitly states that climate projections need regional and local down-scaling to adjust for the relevant and local biological factors that drive smaller-scale systems. EwE software provides a user-friendly interface to do just that, which facilitates the transfer and applicability of the approach described here to other areas and issues.

This work fills an important gap in the literature by evaluating landscape-scale impacts on estuarine food webs and changes in species biomass and distributions in response to environmental changes. Small-scale, site-specific studies that evaluate impacts of restoration on nekton species are widely available in the literature. However, landscape-scale impacts of marsh restoration on fish and fisheries have been less studied. This type of study would not be possible without the availability of long-term monitoring data from stations covering the coastal area provided by the Louisiana Department of Wildlife and Fisheries. Using available monitoring data over larger regions in combination with known restoration projects could elucidate broad-scale patterns of nekton response over time and space. Recent large-scale studies evaluate impacts of the Deepwater Horizon Oil Spill on fisheries using data from the same monitoring program (Van der Ham and De Mutsert, 2014; Ward et al., 2018). Monitoring data provide an avenue for managers to implement adaptive management strategies in a system laden with much uncertainty, and allow for the development and validation of large ecosystem models as presented in this study that are increasingly needed to addressed ecosystem-based management questions.

Coastal areas need to prepare for a changing environment, which includes wetland loss. Coastal restoration and risk reduction projects are part of the mitigation strategy. While these activities will undoubtedly affect the living resources in these areas, simulations such as the ones conducted in this study provide insights into what responses can be expected. Current simulations do not show large negative impacts on living resources of implementing restoration projects designed to mitigate wetland loss. A main reason for this outcome is that an FWOA is a changing environment as well, undergoing habitat loss and saltwater intrusion, represented here by including two SLR scenarios in the model projections. Simulations such as these can provide information to help managers prioritize restoration strategies and anticipate localized response needs to changes in living resource biomass. The simulations show species-specific and location-specific effects that emphasize that species-specific relationships to environmental and location-specific drivers will need to be included in a model when applying this methodology to other locations globally. Resource managers can use the resulting information to implement precautionary measures of ecosystem management and adaptation.

The data used to for the development of the model can be received upon request to the Louisiana Department of Wildlife and Fisheries. Input and output data from the Louisiana Coastal Master Plan simulations used in this study are free and available upon request; email data requests to masterplan@LA.gov or the corresponding author. ICM validation statistics, data, and figures are available online at https://coastal.la.gov/our-plan/2017-coastal-master-plan/. Model source code for EwE is free and available upon request at http://sources.ecopath.org/trac/Ecopath. Model source code for the other ICM components are open source and available online at https://github.com/CPRA-MP.

KM and KL developed the model. JB helped troubleshoot software issues. EW and JB coupled the Ecospace model to the ICM. KM, KL, JB, and EW wrote the manuscript. All authors contributed to the article and approved the submitted version.

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