Effects of a Storm-Induced Barrier Breach on Community Assemblages and Ecosystem Structure Within a Temperate Lagoonal Estuary a Post Hurricane Sandy Analysis

National Park Service U.S. Department of the Interior

Natural Resource Stewardship and Science Effects of a Storm-Induced Barrier Breach on Community Assemblages and Ecosystem Structure within a Temperate Lagoonal Estuary A Post Hurricane Sandy Analysis

Natural Resource Report NPS/NCBN/NRR—2018/1702

ON THE COVER Aerial view of the New Inlet from the south, looking north across Great South Bay, New York on an incoming tide. Photograph courtesy of John Vahey and Charles Flagg, April 18, 2013

Effects of a Storm-Induced Barrier Breach on Community Assemblages and Ecosystem Structure within a Temperate Lagoonal Estuary A Post Hurricane Sandy Analysis

Natural Resource Report NPS/NCBN/NRR—2018/1702

Janet A. Nye1, Michael G. Frisk1, Robert M. Cerrato1, Matthew Sclafani2, Charles N. Flagg1, Skyler R. Sagarese3, Jill A. Olin1

1Stony Brook University School of Marine and Atmospheric Sciences Stony Brook, New York 11794

2Cornell University Cooperative Extension 423 Griffing Avenue Riverhead, New York 11777

3NOAA – National Marine Fisheries Service Southeast Fisheries Science Center Miami, Florida 33149

August 2018

U.S. Department of the Interior National Park Service Natural Resource Stewardship and Science Fort Collins, Colorado

The National Park Service, Natural Resource Stewardship and Science office in Fort Collins, Colorado, publishes a range of reports that address natural resource topics. These reports are of interest and applicability to a broad audience in the National Park Service and others in natural resource management, including scientists, conservation and environmental constituencies, and the public.

The Natural Resource Report Series is used to disseminate comprehensive information and analysis about natural resources and related topics concerning lands managed by the National Park Service. The series supports the advancement of science, informed decision-making, and the achievement of the National Park Service mission. The series also provides a forum for presenting more lengthy results that may not be accepted by publications with page limitations.

All manuscripts in the series receive the appropriate level of peer review to ensure that the information is scientifically credible, technically accurate, appropriately written for the intended audience, and designed and published in a professional manner.

This report received formal peer review by subject-matter experts who were not directly involved in the collection, analysis, or reporting of the data, and whose background and expertise put them on par technically and scientifically with the authors of the information.

Views, statements, findings, conclusions, recommendations, and data in this report do not necessarily reflect views and policies of the National Park Service, U.S. Department of the Interior. Mention of trade names or commercial products does not constitute endorsement or recommendation for use by the U.S. Government.

This report is available in digital format from the Natural Resource Publications Management website. If you have difficulty accessing information in this publication, particularly if using assistive technology, please email [email protected].

Please cite this publication as:

Nye, J. A., M. G. Frisk, R. M. Cerrato, M. Sclafani, C. N. Flagg, S. R. Sagarese, and J. A. Olin. 2018. Effects of a storm-induced barrier breach on community assemblages and ecosystem structure within a temperate lagoonal estuary: A post Hurricane Sandy analysis. Natural Resource Report NPS/NCBN/NRR—2018/1702. National Park Service, Fort Collins, Colorado.

NPS 962/147743, August 2018 ii

Contents Page

Figures...... v

Tables ...... vii

Abstract ...... ix

Acknowledgments ...... xi

Introduction ...... 1

Methods ...... 5

Field Surveys ...... 5

Statistical Analyses ...... 7

Modeling Physical Regimes ...... 7

Environmental Variables ...... 7

Analysis of Community Metrics and Assemblage Structure ...... 8

Analysis of Species Environmental Preferences ...... 9

Ecosystem Analysis ...... 10

Data sources of regional abundance time series used in Ecosim...... 13

Results ...... 15

Environmental Variables ...... 15

Modeling of Circulation and Water Properties...... 19

Community Metrics ...... 21

Assemblage Structure ...... 26

Species' Environmental Preference ...... 29

Ecosystem-Level Differences ...... 32

Fitted Ecosim Model ...... 32

Network Analysis ...... 36

Model Limitations and Caveats ...... 38

Discussion ...... 39

Conclusions ...... 41 iii

Contents (continued) Page

Literature Cited ...... 43

Figures

Page

Figure 1. A) Study area and B) sampling region depicting the major bays (South Oyster Bay, Great South Bay, Bellport Bay, Moriches Bay), tributaries (Carmans River, Connetquot River) and connections to the Atlantic Ocean (Fire Island Inlet and breach location) along the south shore of Long Island, NY...... 2

Figure 2. Distribution of surface salinity (psu) measurements throughout Great South Bay, New York, pre- and post-Sandy...... 16

Figure 3. Contour maps depicting the seasonal bottom salinity (psu) measurements in Great South Bay, New York, in 2007, 2013, 2014 and 2015 with darkening reds representing a gradient of lower to higher salinities...... 17

Figure 4. Contour maps depicting the seasonal bottom temperature (°C) measurements in Great South Bay, New York, in 2007, 2013, 2014 and 2015 with blues and reds representing a gradient of lower to higher temperatures, respectively...... 18

Figure 5. Tidal-mean streamlines from the Finite Volume Coastal Ocean Model (FVCOM) showing pre-breach (upper panel) and post-breach (lower panel) under winter conditions in Great South Bay, New York...... 20

Figure 6. Residence time (in days) before (upper panel) and with (lower panel) the breach in Great South Bay, New York with darker colors representing short turnover and lighter colors representing long turnover...... 21

Figure 7. Catch composition (% catch per unit effort) from the Great South Bay, New York otter trawl survey by year...... 25

Figure 8. Trends in community composition, A) catch per unit effort (CPUE), B) species richness, C) Shannon-Wiener diversity, D) total length and E) derived-biomass, calculated from the Great South Bay, New York otter trawl survey...... 26

Figure 11. Catch per unit effort (CPUE) of Blue Crabs sampled during the otter trawl survey conducted in 2007 and from 2013-2015 in Great South Bay, New York...... 30

Figures (continued) Page

Figure 12. Catch per unit effort (CPUE) of Lady Crabs sampled during the otter trawl survey conducted in 2007 and from 2013-2015 in Great South Bay, New York...... 31

Figure 13. Ecosim modeled trends between observed (points) and predicted (line) biomass (t/km2) estimates of important commercial and recreational fishes from 2007 to 2015 in Great South Bay, New York...... 33

Figure 14. Ecosim modeled trends between observed (points) and predicted (line) biomass (t/km2) estimates from the GSB trawl conducted in 2013-2015 in Great South Bay, New York...... 34

Figure 15. Catches forced in Ecosim model. Note different scales for each species...... 35

Tables

Page

Table 1. Model inputs for the balanced Ecopath model including trophic level (position in the food web), biomass (B; t/km2), P/B (ratio of production to biomass; equivalent to total mortality rate), Q/B (ratio of consumption to biomass), EE (ecotrophic efficiency; fraction of the total mortality that is used within the modeled system) and harvest (t/km2; derived from commercial and recreational landings data)...... 11

Table 2. Summary of surface salinity (psu; median ± SD) measured from different locations in Great South Bay, New York before and after the breach...... 15

Table 3. Summary of the environmental conditions measured during cruises in Great South Bay, New York...... 19

Table 4a. Total number of individuals for each taxonomic group caught within Great South Bay, New York during the trawl surveys conducted in 2007 and from 2013 - 2015...... 21

Table 4b. Catch data for species caught within Great South Bay, New York during the trawl surveys conducted in 2007 and from 2013 - 2015...... 22

Table 5. Results of the permutation tests for the canonical correspondence analysis (CCAs) of catch per unit effort (CPUE)...... 26

Table 6. Summary of the optimal model selected using Akaike's Information Criterion (AIC), AIC weight (AIC wi) is ratio of ∆AIC values for each model relative to the whole set of candidate models and the deviance [Dev (%)] explained for the occurrence (PA) models (using binomial distribution) and abundance (ABUN) models (using the lognormal distribution) for each species...... 32

Table 7. A summary of metrics for structural properties of Great South Bay, New York as modeled in Ecopath for 2007-2015...... 37

vii

Abstract

Since the 1800’s, Great South Bay (GSB), NY, has had reduced connectivity to the ocean system, a reduction in filter feeders including oysters, hard clams, and menhaden, a reduction in marine top predators such as sharks and has become dominated by lower trophic level pelagic species. In October 2012, Hurricane Sandy created a breach in Fire Island at Bellport Bay, increasing the exchange between eastern GSB and the ocean resulting in higher salinity levels throughout almost the entire bay. The breach created an opportunity to evaluate whether increased connectivity to the ocean would alter the species assemblage of this lagoon system. We hypothesized that higher salinity in GSB would result in greater occurrence of marine species and increased abundance of migratory finfish and invertebrates. Otter trawl surveys were conducted seasonally (i.e., spring, summer and fall) from 2013 through 2015 throughout GSB from Fire Island Inlet to Bellport Bay. Results indicate an increase in species richness and diversity following the breach in GSB compared with assemblages sampled during 2007 pre-breach. Eight species showed an increase in catch per unit effort (CPUE) following the breach; of these species, five are considered to generally prefer higher salinity habitats, and include Summer Flounder, Windowpane Flounder, Bluefish, Sea Robin spp. and Mantis Shrimp. The higher CPUE of Summer Flounder and Bluefish in GSB is important given their economic importance. Also important to note is the increase in CPUE of Horseshoe Crab following the breach. The most dramatic change pre- and post-breach has been the shift in CPUE of Blue Crab and Lady Crab, with a ~84% decline and a >150% increase in CPUE, respectively, following the breach. Community-wide evaluation indicated that the fish and mobile invertebrate assemblage in GSB in 2007 differed from 2013-2015, with greatest differentiation between 2007 and 2013. The increase in diversity, species richness, and CPUE of higher trophic level predators, such as Summer Flounder, following the breach suggest that the connectivity of GSB with the ocean has increased the trophic complexity, an indication of initial recovery of ecosystem maturity. Many, but not all metrics from our Ecopath model indicate that the breach increased maturity of GSB. However, the species assemblage appears to still be transitioning from the effects of the breach.

Acknowledgments

The authors thank Tyler Abruzzo, David Bowman, Tara Dolan, Chris Harter, Lis Henderson, Irvin Huang, Evan Ingram, Jake Labriola, Emily Markowitz, Chris Martinez, Kellie McCartin, Maren Mitch, Adelle Molina, Cecilia O’Leary, Jared Reed, Nick Rogers, Jason Schweitzer, Mark Wiggins, Haikun Xu, Adham Younes, Joshua Zacharias and Catherine Ziegler for assistance with field collections. Claudia Hinrichs completed the FVCOM modelling work. Funding was provided by New York Sea Grant (NYSG), the New York State Department of Environmental Conservation (NYSDEC) and the National Park Service (NPS) to MGF, JAN, RMC, MS and CF.

Introduction

Great South Bay (GSB), located along the southern shore of Long Island, New York (Figure 1), is a shallow, well-mixed, highly productive temperate lagoonal ecosystem (Carpenter & Brinkhuis 1991). Its border is defined by South Oyster Bay to the west, Fire Island to the south, and Moriches Bay to the east. The system extends 40 km east to west, with varying widths, measuring 300 m to 11 km (Schubel 1991), and providing approximately 290 km2 of lagoonal habitat (Hanlon 1983). The mean low water depth of GSB ranges from 0.3 to 3.6 m and averages 2 m (Hanlon 1983; Schubel 1991). Historically, GSB supported a diverse array of finfish, including Winter Flounder Pseudopleuronectes americanus, Bluefish Pomatomus saltatrix, Striped Bass Morone saxatilis (Gabriel 1921; Schreiber 1973), and shellfish such as oyster and hard clams (Gabriel 1921; Greene 1982), and served as an important nursery area for many species of waterfowl and migratory birds (Hanlon 1983). The GSB has provided commercial, recreational and ecological benefits to New York and the eastern United States seaboard for over a century (Gabriel 1921; Dickinson 1938; McHugh 1991; Gobler et al. 2005). Prior to October 2012, Fire Island Inlet, where depths reach greater than 12 m (Hanlon 1983), provided the only direct influx of ocean water into GSB, although Moriches and Jones Inlets provide sources of indirect flow (Schubel 1991). The Carmans and Connetquot Rivers provide the main surface freshwater contributions to GSB. Eelgrass beds (Zostera spp.), salt marshes (primarily cordgrass: Spartina alterniflora and S. patens), mud, sand and shell areas comprise the majority of the benthic habitats.

In 2012, Tropical Storm Sandy created breaches at three locations in the barrier island that separates the bays from the Atlantic Ocean, one of which remains open (Old Inlet). This breach is located at the eastern end of Great South Bay (GSB), providing an additional connection to the marine environment, affording a unique opportunity to obtain biological insights into the influence of a breach on the physical and chemical environment of GSB and associated community (Figure 1).

Storm systems influence lagoonal ecosystems, most notably through storm surge and Barrier Island breaches altering flushing rates and physical properties (e.g., salinity) of coastal lagoons (Morton & Sallenger 2003, Anthony et al. 2009, Smith et al. 2010). An estimated 28 breaches and inlets (Leatherman 1985) have influenced GSB's circulation patterns (Greene 1982, Hinga 2005, Ashton et al. 2008) and have likely impacted community assemblages for centuries (Nuttall et al. 2011, Cerrato et al. 2013). While the effects of breaches and other natural alterations on physical and geological processes are well-documented (e.g., Morton & Sallenger 2003, Paerl et al. 2006), there is a paucity of research exploring these impacts on ecosystem structure and health, community composition, and spatial distributions of coastal species. Shifts in species composition are likely to occur given elevated salinity and improved circulation (Bird 1992, Morton & Sallenger 2003) under the conditions of a breach. The demand for ecosystem-based fisheries management (EBFM), recently required by the Reauthorized Magnuson-Stevens Fishery Conservation and Management Act and the National Ocean Policy, highlights moving beyond traditional abundance estimates of living marine resources (Pikitch et al. 2004). For the EBFM holistic approach to be successful, increased understanding of habitat and ecological interactions in a changing environment will be necessary to accurately reflect ecosystem dynamics (Pikitch et al. 2004, Christensen & Maclean 2011).

Given the economic, social, and ecological importance of organisms within coastal ecosystems, it is imperative to assess how environmental drivers influence species dynamics (Buchheister et al. 2013) and how changes in hydrography can impact coastal ecosystems. Physical processes are capable of altering species' distributions, abundances, and trophic interactions (Alpine & Cloern 1992, Scavia et al. 2002, Mackenzie et al. 2007). In Western Australia, breaching of a normally-closed estuary increased salinity of the lagoon and dramatically increased ichthyofaunal diversity and structure (Young & Potter 2002). Marine species often inhabit preferred habitats as a means of regulating metabolic processes (i.e., higher growth at some optimal temperature; Fry 1971) or increasing food availability (Stoner et al. 2001, Campana & Joyce 2004). Species commonly shift their distributions to remain in optimal environmental conditions, as evident by “temperature keepers” which change their depth to remain in preferred temperatures and include Haddock (Melanogrammus aeglefinus) and Silver Hake (Merluccius bilinearis) (Perry & Smith 1994, Nye et al. 2011).

Mature ecosystems tend to have higher biomass, high complexity and importantly are more resilience and resistant to perturbations (Odum 1971). Maturity of GSB has declined since the 1800s, and this suggests decreasing system complexity, stability and resilience (Nuttall et al. 2011). Nuttall et al. (2011) suggested that system consumers were becoming less efficient at utilizing available system production. The filtration and clearance of menhaden rival and may surpass those of shellfish, indicating the drastic reduction of this species may have had long term consequences for the utilization of primary production (Nuttall et al. 2011). Consistent declines in various indices of trophic complexity further indicated a trend towards simplifying the trophic structure of the ecosystem (Nuttall et al. 2011).

By changing salinity, breaches are expected to impact lower trophic-level species and impart trophic pressures in addition to affecting migratory keystone species. Salinity was identified as the major environmental gradient structuring demersal fish community composition and biodiversity in the Chesapeake Bay estuary (Buchheister et al. 2013). We evaluated the effects of a barrier island breach on GSB using three approaches, 1) assessing changes to species assemblages relative to temperature and salinity alterations; 2) assessing species' distributions and predicted changes under environmental scenarios (e.g., salinity regime after Sandy); and, 3) comparing ecosystem structure before and after Sandy. All fishes and mobile invertebrates were analyzed, although emphasis was given to the most commercially important species, including Blue Crab (Callinectes sapidus), Summer Flounder (Paralichthys dentatus) and Winter Flounder (Pseudopleuronectes americanus). Specific hypotheses that were tested include:

1. As all species associate with specific environmental conditions (e.g., salinity and temperature) increased exchange with the ocean following Sandy will cause the following:

a. An increase in habitat for marine species as indicated by an increase in salinity throughout much of the bay.

b. A greater occurrence of marine species and associated communities.

c. Greater exchange of energy between GSB and ocean habitats through increased abundance of migratory finfish and invertebrates.

2. The fish and mobile invertebrate assemblage in the vicinity of the new breach at Old Inlet will be more closely related to the assemblage in the vicinity of Fire Island Inlet in 2013 than it was pre-Sandy (i.e., 2007).

3. Observations of community structure and projected trends will indicate recovery of system maturity including:

a. Ratios of production/total respiration (PP/R), primary production/total biomass (PP/B) and relative ascendency (A/C) will show decreasing trends with increased maturity.

b. An increase in system omnivory index (SOI), mean organism size (P/B-1), Finn’s (FCI) and predatory cycling index (PCI), path length (PL) and system overhead (O).

c. An increase in filter feeding, total upper trophic and migratory biomass. 3

Methods

Field Surveys Data were collected by an otter trawl survey conducted in 2007 (pre-breach) and in 2013, 2014 and 2015 (post-breach). The survey operated three cruises per year; spring, summer and fall, following the methods of Frisk & Munch (2008). Sampling locations for the survey were selected using a stratified random sampling design based on location within the Bay and a cruise consisted of 40–45 randomly selected stations per season within GSB (n = 485; Figure 1).

The survey was conducted aboard the 8.5 m aluminum hull R/V Pritchard equipped with 2 Rotzer TH1 winches and lifting boom. A 9.9 m otter trawl with a head rope of 7.5 m, footrope of 9.9 m, 2.0 m in height, with 2 cm mesh and a 0.6 cm cod-end, and tickler chain (in 2014 and 2015) was towed for ~12 minutes at 2.5 knots at each station. These tows occurred in depths of 1.6 to 10 m and were conducted during daylight hours. Bottom temperature (°C) and salinity (ppt) were measured with a hand held YSI Professional Plus multi-parameter meter (Yellow Springs Instruments). Depth (m) was determined using an onboard depth finder.

During the surveys, species were recorded in the field unless species-level identifications were not easily made. Unidentified specimens were transported to the lab where they could be identified with a taxonomic key. All finfish and crustaceans were enumerated and 30 individuals for each taxa in each tow were measured for total length (mm) in the case of fishes and carapace/disc width (mm) in the case of crabs and rays. Data for all individuals were entered in an Excel database. To develop the species-specific gear correction factors for trawls conducted in 2007 and 2013, an experimental survey with paired tows (with/without tickler chain) at each of 45 stations was conducted in summer of 2015. Correction factors based on field comparisons were then applied to standardize catch data for changes in trawling gear, when necessary.

For each tow, catch data were converted into catch per unit effort (CPUE) data based on swept area estimated from the speed, time towed and net width over which each tow occurred. Community composition was described with species richness and Shannon-Wiener diversity measures. Biomass or fish weight was estimated for measured individuals using species-specific literature-derived length-weight relationships.

Statistical Analyses

Modeling Physical Regimes We used an existing Finite Volume Coastal Ocean Model (FVCOM; Chen et al. 2003) to examine circulation changes in Great South Bay pre- and post-Sandy. FVCOM is a fully 3-D primitive equation hydrodynamic model that uses an unstructured grid (triangular elements) to provide a realistic representation of the flow field within a complicated estuary or coastal area. First, we updated the existing pre-breach FVCOM model from version 2.7 to version 3.2. The forcing included pre-breach temperature and salinity distribution, bottom freshwater, freshwater from 61 rivers and streams, tidal forcing with six constituents on the open boundary, and local winds. We then modified the unstructured grid in the “pre-breach model” to correspond to the changes in bathymetry caused by the breach, creating before-breach and after-breach models both of which were tuned to oceanographic data. Bathymetry data were obtained from several sources. The basic bathymetry comes from a combination of NOAA NOS and modern multi-beam surveys of the GSB (Clapp and Flood 2004). The bathymetry around the breach comes from a November 2013 Lidar survey conducted by USGS (Nelson et al. 2017) combined with a multibeam survey conducted in September 2013 and an aerial overflight also in September 2013. We compared the pre-breach and post-breach models against the corresponding observed tidal and salinity data from the GSB Observatory. Temperature and salinity were collected from eight stations around the bay with SBE SeaCats that have been deployed since 2005. Meteorological data are collected from the south tower of the Smith Point Bridge. Those data include temperature and salinity from the Smith Point SeaCat, short and long wave radiation from Eppley pyronometers, and wind speed, wind direction, air temperature, humidity, barometric pressure and rainfall rate from a Vaisala WXT520. In 2010, a buoy in the center of the bay was deployed that reports wind speed and direction, air temperature and humidity, water temperature and salinity, chlorophyll-a fluorescence, turbidity and nitrate concentration. USGS maintains two tide gauges in the bay. More information and the data can be found at https://you.stonybrook.edu/greatsouthbay/monitoring/observatory/. These data from the GSB Observatory were also used to compare the distributions of salinity before and after the breach.

To estimate residence time before and after the breach, we used a passive dye tracer technique. After model run-up, the dye was established in a local area and followed for sufficient time to determine how long it took for the total dye concentration in the area of interest to fall below 1/e of the initial amount of dye. This scheme is one of several that have been used to estimate residence times but has the advantage that it includes the effect of tidal sloshing in and out of the region and so gives a better overall estimate. For more details on the physical modeling of GSB please see Hinrichs (2018).

Environmental Variables To characterize environmental conditions in GSB, contour maps of bottom temperature and salinity were created using empirical Bayesian kriging in ArcMap 10.4.1 (ESRI, Redlands, CA). Contoured data include environmental data measured during the otter trawl cruises. We then used a two-way analysis of variance (ANOVA) followed by Tukey’s HSD post-hoc test to evaluate the seasonal (spring, summer, fall) and annual (2007, 2013, 2014, 2015) differences in salinity and bottom temperature value when a significant result was observed in R 3.1.2 (R Development Core Team

2016). Environmental variables were evaluated for normality using quantile–quantile probability plots and for homogeneity using box plots. Where the assumptions of normality and equal variance were not met, the data were log-transformed.

Analysis of Community Metrics and Assemblage Structure Patterns of community metrics were examined using the following measures for each station: index of abundance (CPUE), species richness, Shannon-Wiener diversity, length and derived biomass (length converted to weight for each individual measured). The temporal differences in these metrics among years was examined using either a one-way analysis of variance (ANOVA) or a Kruskal- Wallis Rank Sum test, followed by Tukey’s honestly significant difference (HSD) test (α = 0.05). We also classified species as freshwater, estuarine or marine according to the guild scheme for estuarine fishes proposed by Elliot et al. (2007). Freshwater fishes include freshwater migrants and freshwater stragglers. Estuarine fishes include estuarine residents, migrants and putative amphidromous species. All marine species were marine migrants, and no species was identified as a marine straggler.

We examined associations between species assemblage structure based on otter trawl CPUE, environmental variables, and spatial and temporal attributes using canonical correspondence analysis (CCA). The approach therefore offers insights into species associations that are not readily obtained by univariate methods (Legendre & Legendre 1998). Essentially, CCA is a multivariate extension of multiple regression where species and sites are ordinated in a manner that maximizes the variation explained by a set of explanatory variables (ter Braak 1986, McGarigal et al. 2000). The amount of variation, termed inertia, is based on chi-square distances, which are distances weighted by species totals (ter Braak & Smilauer 2012). As with multiple regression, the significance of an explanatory variable can be tested (Legendre & Legendre 1998). CCA is appropriate for large environmental gradients and assumes unimodal distributions of species (i.e., approximately Gaussian response curves) along those gradients (ter Braak 1986). An advantage of CCA over standard univariate analysis of species abundance (e.g., multiple regression) is that the method simultaneously depicts the strength and direction of species responses to explanatory variables by the position and spread of species in ordination space. Prior to analysis, individual species that occurred in <5% of tows were grouped together as minor taxa, as CCA can be sensitive to rare species (Borcard et al. 2011).

The explanatory variables in our analysis included physical characteristics of bottom water (temperature, salinity), spatial attributes (depth) and temporal periods (year). Measured variable data were displayed in the ordination space described by the first two CCA axes of the GSB CPUE data by projecting bi-plot vectors for continuous environmental variables (temperature, salinity, and depth) and centroids for the categorical variable year. Significance of these variables in explaining community variation was evaluated using a permutation test (Legendre & Legendre 1998; Manly 1991). In ordinations, there is no reason to assume a priori that relationships between the environmental variables and the ordination space will be linear. To address this issue, we used generalized additive models (GAMs) to fit response surfaces for each explanatory variable to the two-component ordination space. A smooth 2-d function of the sample scores on CCA axes one and two was generated using thin-plate splines and used to model the individual measured environmental variables in turn (Wood 2003). The degree of smoothness in the thin-plate spline was estimated by

generalized cross-validation. Ordinations and response surfaces were implemented through package 'vegan' (Oksanen et al. 2016) in R.

Analysis of Species Environmental Preferences Spatial distribution of Blue and Lady Crab, two species which responded to the altered breach environment, were estimated using generalized additive models (GAMs) that predict species abundance and occupancy resulting from changes to the GSB ecosystem. Species' distributions are frequently modeled using GAMs which quantitatively relate occurrence or abundance to environmental and ecological drivers (Guisan et al. 2002, Leathwick et al. 2006, Heinänen et al. 2008, Damalas et al. 2010). Considered a semi-parametric extension of the generalized linear model (GLM) (McCullagh & Nelder 1989), GAMs utilize a smoothing function (Wintle et al. 2005) that can easily handle non-linear relationships and uncover hidden structure between variables missed by traditional linear methods (Hastie & Tibshirani 1990, Guisan et al. 2002). A GAM is defined by the following relationship:

(1) = + ( ) +

𝑖𝑖 𝑖𝑖 where the response𝑌𝑌 variable,𝛼𝛼 ∑ 𝑓𝑓 Y, 𝑋𝑋is a function𝜀𝜀 of the sum of the explanatory variables (Xi) altered by unspecified nonparametric smoothing functions (fi), an intercept term (α), and an error measure (ε).

Two models were constructed for each species. The first (PA) predicted the probability of occurrence using a logit link function and a binomial error distribution. The second (ABUN) predicted the abundance (based on CPUE data) using a log link function and a lognormal error distribution. All GAMs were built in R, with the package 'mgcv' (Wood 2006, 2011) using cubic regression splines and 5 knots, with the optimal model chosen based on the lowest Akaike’s information criterion (AIC; Akaike 1973) with small-sample bias adjustment (AICc; Hurvich & Tsai 1989). Akaike weights (wi) were calculated to interpret the weight of evidence for the best-fitting model with evidence ratios used to compare among models (Johnson & Omland 2004).

Unbiased estimates of each optimal model’s predictive performance were obtained through a v-fold cross validation evaluation. Occurrence (PA) models were tested for discrimination and accuracy using the packages 'PROC' (Robin et al. 2011) and 'Presence-Absence' (Freeman 2007), respectively. The ability of the model to discriminate between presence and absence sites was described using area under the curve (AUC; Brotons et al. 2004, Leathwick et al. 2006), with values between 0.7 and 0.9 considered reasonable and values >0.9 good, as the true positive rate was high relative to the false positive rate (Swets 1988, Pearce & Ferrier 2000). Validation of abundance (ABUN) models was assessed using model performance estimators, including calibration, correlations and mean error (Potts & Elith 2006, Heinänen et al. 2008). Calibration was measured with a simple linear regression between observed and predicted values, with the intercept term indicative of bias and the slope reflective of the consistency in the predictions (Potts & Elith 2006). The strength of the relationship between observed and predicted values was assessed using Pearson’s correlation coefficient (r), although a perfect correlation (r = 1.0) may still display bias in a consistent direction (Potts & Elith 2006, Heinänen et al. 2008). The similarity between ranks of observed and predicted values was assessed using Spearman’s rank correlation (rsp), with a high value indicating a correct order of

predictions (Potts & Elith 2006). Lastly, both root mean square error of prediction (RMSE) and average error (AVE) were calculated as in Potts & Elith (2006).

Ecosystem Analysis We originally proposed to evaluate the effects of Tropical Storm Sandy on the GSB ecosystem by creating a post-breach Ecopath model and compare it to the 2000s model published by Nuttall et al. (2011). The Ecopath approach utilizes a mass-balanced framework composed of trophically-linked biomass pools representing all major ecosystem functional groups (Polovina 1984, Christensen & Pauly 1992, Pauly et al. 2000). Production (Pi) is expressed as a function of all loss and gain processes to a stock’s (i) biomass and is estimated with the following equation:

(2) = + ( 2 ) + + + (1 )

𝑖𝑖 𝑖𝑖 𝑖𝑖 𝑖𝑖 𝑖𝑖 𝑖𝑖 𝑖𝑖 𝑖𝑖 where Yi = yield𝑃𝑃 from𝑌𝑌 fishery,𝐵𝐵 𝑥𝑥(B𝑀𝑀i x M2i) 𝐸𝐸= total𝐵𝐵𝐵𝐵 predation𝑃𝑃 rate− 𝐸𝐸𝐸𝐸(M2i) acting on the total biomass of the stock (Bi) (estimated with equation below), Ei = net migration rate (emigration – immigration), BAi is the biomass accumulation rate, and Pi x (1 – EEi) = ‘other mortality’ rate that represents mortality from sources outside the system; compared to production used by the system which is represented by EEi (ecotrophic efficiency).

Ecopath then calculates the total consumption, or removal of system production, with the following equation:

(3) ( 2 ) = + + … 𝑄𝑄 𝑄𝑄 𝑄𝑄 𝑖𝑖 𝑖𝑖 1 1 1𝑖𝑖 2 2 2𝑖𝑖 𝑛𝑛 𝑛𝑛 𝑛𝑛𝑖𝑖 𝐵𝐵 𝑥𝑥 𝑀𝑀 �𝐵𝐵 𝑥𝑥 𝐵𝐵 𝑥𝑥 𝐷𝐷𝐷𝐷 �𝑖𝑖 �𝐵𝐵 𝑥𝑥 𝐵𝐵 𝑥𝑥 𝐷𝐷𝐷𝐷 � �𝐵𝐵 𝑥𝑥 𝐵𝐵 𝑥𝑥 𝐷𝐷𝐷𝐷 � where Bn is the biomass of predator n, Q/Bn is the ratio of the consumption of prey to the biomass of predator n, and DCni is the percentage that prey i composes predator n’s total diet. This method assumes mass balance of these equations to present a model of trophic flows within the system. The program does not require estimations of all parameters to solve the algorithms although estimates of Yi, Ei, BAi, and DCij for stock i are necessary. Three of the four parameters (P/B, Q/B, EE, and P/Q) are needed to be estimated by the modeler while the fourth can be calculated by the software.

Initial analyses indicated that comparing two static Ecopath models was not appropriate. Significant changes were made to the collection and processing of harvest data in New York State, since the 2000s Ecopath model was constructed making a direct comparison to the 2000s model impossible. Instead, we created an Ecopath model for 2007 and fitted times series of abundance and catch data (see data sources below) in the dynamic modeling framework Ecosim to represent annual system changes for the period 2007-2015. The 2007 Ecopath model was parameterized based on Nuttall et al. (2011), with updated catch and biomass inputs and included 40 species and/or functional groups. The Ecopath model was then balanced through an iterative process and provided initial values on biomass per km2, P/B, Q/B, EE and harvest per km2 (Table 1). The Ecopath model served as the initial parameter values for Ecosim and further balancing was conducted to increase the statistical fit of the times series used in the dynamic simulations. Importantly, the Ecopath model provides the biomasses of all species groups while Ecosim is fitted to relative time series of biomass/abundance.

The Ecosim model was fitted to the observed relative biomass/abundance time series derived from regional stock assessments and scientific surveys for Weakfish, Summer Flounder, Bluefish, Striped Bass, Tautog, Winter Flounder and Scup (See data sources below). These species are present seasonally in the GSB and regional trends in abundance provide a good metric of the relative abundance in the GSB system. The three-year otter trawl survey of GSB provided abundance data for 2013, 2014 and 2015. The Ecosim model was fitted to time series from the GSB survey for American Eel, Black Seabass, Bluefish, Blue Crab, Butterfish, Horseshoe Crabs, Menhaden, Northern Kingfish, Spider Crab and Oyster Toadfish. The model calculated catches for Black Sea Bass, Bluefish, Tautog, Sea Robin, Hard Clams, Blue Crab, Forage Fish, Scup, Striped Bass, Summer Flounder, Weakfish and Winter Flounder. The times series of catches were forced; thus, they were not fit statistically and are not considered in the calculation of the model sum of squares.

The rate of flow between vulnerable and susceptible prey pools is defined by the value of vulnerability (v) estimated for each predator-prey interaction (Walters & Juanes 1993). Larger values of v indicate top-down control while smaller values represent bottom-up control. The model was fitted by iterating the initial parameters in Ecopath and statistically fitting v in the vulnerability search module in Ecosim. We allowed Ecosim to estimate 30 of the 211 v values in the model and the remaining values were set to 2, reflecting a balance between bottom-up and top-down control.

To estimate basic properties of the ecosystem, we quantified lower trophic level activity (net primary production, PP), upper trophic level activity (system respiration, RESP), and total biomass flows within the system (total system throughput, TST). Ecosystem maturity was examined by calculating various indices including PP/RESP, B/TST, and omnivory index (OI), among others (Odum 1969, Christensen et al. 2009, Nuttall et al. 2011). The relative impact of particular stocks within each model as well as their importance was assessed using 'keystoneness' (Libralato et al. 2006, Nuttall et al. 2011).

Species/Group Trophic level Biomass P/B Q/B EE Harvest Benthic fauna 2.0 5.047 4.580 24.360 0.850 − Macrobenthos 3.1 0.252 1.000 1.667 0.804 − Sand shrimp 2.1 1.618 3.750 24.900 0.999 − Shrimp 2.3 2.698 3.300 16.100 0.860 − Mantis shrimp 3.0 0.100 1.340 7.430 0.950 − Squid 2.3 0.008 5.817 7.000 0.990 0.004 Forage fish 3.1 1.947 1.290 15.000 0.999 0.070 Suspension feeders 2.1 34.664 0.300 2.182 0.910 0.089 Hard clams 2.1 4.000 1.112 5.100 0.950 1.030

Table 1 (continued). Model inputs for the balanced Ecopath model including trophic level (position in the food web), biomass (B; t/km2), P/B (ratio of production to biomass; equivalent to total mortality rate), Q/B (ratio of consumption to biomass), EE (ecotrophic efficiency; fraction of the total mortality that is used within the modeled system) and harvest (t/km2; derived from commercial and recreational landings data).

Species/Group Trophic level Biomass P/B Q/B EE Harvest Tropical fish 3.1 0.172 1.268 7.967 0.999 0.019 Crabs 2.9 3.584 1.470 8.500 0.954 0.003 Blue crab 3.2 0.442 1.200 4.000 0.950 0.120 Horseshoe crab 3.0 0.251 0.600 3.000 0.500 0.058 Spider crab 2.5 0.000 182.721 1218.140 0.950 − Skate 3.5 0.027 0.150 4.100 0.990 0.004 American eel 2.0 0.259 0.220 3.700 0.500 0.008 Menhaden 2.1 1.573 1.100 28.000 0.990 − Black seabass 3.2 0.261 0.400 6.900 0.990 0.103 Blackfish 3.8 0.070 0.144 3.100 0.990 0.010 Bluefish 3.6 1.403 0.800 15.403 0.700 0.760 Butterfish 2.6 0.002 0.800 5.500 0.990 − Cunner 3.1 0.006 0.480 7.500 0.990 0.002 Dogfish 3.5 0.011 0.200 4.770 0.950 0.002 Drums & Croakers 3.0 0.066 0.800 3.900 0.990 − Flatfish-other 3.3 0.002 1.728 7.033 0.950 − Gadidae 3.2 0.012 1.000 3.660 0.999 0.004 Northern kingfish 3.8 0.100 0.600 5.900 0.335 0.018 Northern pipefish 3.3 0.024 0.700 6.900 0.999 − Oyster toadfish 3.5 0.010 0.360 6.200 0.556 0.002 Scup 3.1 0.060 1.270 5.500 0.999 0.010 Striped bass 3.7 2.256 0.390 2.450 0.500 0.440 Sea robins 3.3 0.218 0.430 7.150 0.990 0.023 Summer flounder 3.8 0.600 1.481 4.000 0.444 0.233 Weakfish 3.4 0.090 0.572 4.100 0.999 0.012 Windowpane flounder 3.2 0.000 0.470 6.200 0.990 − Winter flounder 2.9 0.126 0.740 6.300 0.999 0.020 Zooplankton 2.0 1.722 25.000 90.909 0.900 − Phytoplankton 1.0 4050.000 60.000 0.000 0.001 − Ctenophores 2.5 0.224 8.800 35.200 0.900 − Detritus 1.0 1.000 − − 0.001 −

Data sources of regional abundance time series used in Ecosim. • ASMFC (2014). Atlantic States Marine Fisheries Commission: Interstate Fishery Management Plan for Winter Flounder.

• ASMFC (2015). Stock Assessment Update of Summer Flounder for 2015.

• ASMFC (2016). Atlantic States Marine Fisheries Commission: Atlantic Striped Bass Stock Assessment Update.

• ASMFC (2016). Atlantic States Marine Fisheries Commission: Tautog Stock Assessment Update.

• ASMFC (2016). Atlantic States Marine Fisheries Commission: Weakfish Benchmark Stock Assessment and Peer Review Report.

• NEFSC (2015). Bluefish Benchmark Stock Assessment for 2015 60th SAW Assessment Report.

• NEFSC (2016). Bluefish Stock Assessment Update: Data Update through 2015.

• NEFSC (2016). Scup Benchmark Stock Assessment for 2015.

Results Environmental Variables The physical environment changed in GSB following the breach as evident from observations that were made continuously at different stations (Figure 2) in the bay over time. Consistent with our first hypothesis, there was a persistent shift towards higher mean salinity at all stations that were continuously monitored except the US Coast Guard Station at Fire Island Inlet. The change is most pronounced at stations closest to the breach on the east side of GSB especially the Bellport location at +5 practical salinity units (psu). The increase of salinity is also accompanied by a decrease in variability (Table 2). These results are confirmed by an analysis of salinity measurements over several years at six locations pre-Sandy and post-Sandy (Table 2).

Table 2. Summary of surface salinity (psu; median ± SD) measured from different locations in Great South Bay, New York before and after the breach. For locations see map inset in Figure 2.

Site Before Breach With Breach Change

Bellport 24.0 ± 1.97 29.0 ± 1.36 +5.0 ± -0.61

Blue Point 24.8 ± 1.53 27.1 ± 1.08 +2.3 ± -0.45

Islip 25.3 ± 1.54 26.8 ± 1.33 +1.5 ± -0.21

Barrett Beach 26.2 ± 1.46 28.2 ± 1.16 +2.0 ± -0.30

Tanner Park 28.7 ± 1.49 29.7 ± 1.28 +1.0 ± -0.21

US Coast Guard Station, Fire Island Inlet 30.3 ± 1.23 30.4 ± 1.02 +0.1 ± -0.21

The timing of the bottom trawl surveys differed slightly in each season and year so care should be taken in the interpretation of the results especially for temperature. However, the salinity post-breach was higher than prior to the beach in summer and fall. Comparing salinity across years indicated that there was a smaller west-east gradient in all years post-breach compared with 2007 (Figure 3). An east-west salinity gradient was observed in July 2007 while high salinity is observable near both ocean exchange points in August 2013, July 2014 and July 2015 (Figure 3). In 2013, and in the years following, there appears to be gradients from both ocean exchange location (Figure 3). For example, summer profiles of salinities in the eastern portion of GSB indicated a salinity range of 22-24 psu in 2007, whereas salinities of 28-34 psu were recorded in the same area post-breach in 2013 (Figure 3). This increase in salinity post-breach in areas adjacent to the breach, would support the hypothesis of increased habitat for marine species throughout much of the bay.

Bottom temperatures generally increased in the summer from spring, and generally decreased in the fall (Table 3). The much lower values recorded in the fall of 2013 reflect the later sampling period of October-November in that year. Minimal changes in the temperature contours of GSB were observed (Figure 4). For example, temperatures were relatively consistent across GSB, ranging between 26 and 28°C in summer 2007 before the breach, and between 22 and 24°C in summer 2013 following the breach (Figure 4).

Figure 2. Distribution of surface salinity (psu) measurements throughout Great South Bay, New York, pre- and post-Sandy. Median (± SD values) for each location are shown in Table 2.

Table 3. Summary of the environmental conditions measured during cruises in Great South Bay, New York. Data are mean (± SD).

Year Season Salinity (psu) Temperature (°C) Depth (m) 2007 Spring 25.6 ± 1.5 17.0 ± 1.1 3.0 ± 1.5 Summer 26.3 ± 2.3 26.7 ± 1.4 3.0 ± 1.6 Fall 25.0 ± 1.8 19.6 ± 2.5 2.9 ± 0.8 2013 Spring 25.7 ± 1.0 19.9 ± 1.3 3.0 ± 1.1 Summer 28.6 ± 0.9 23.4 ± 0.9 3.2 ± 1.3 Fall 30.5 ± 0.6 10.5 ± 2.4 3.3 ± 0.9 2014 Spring 25.4 ± 0.9 16.7 ± 1.1 3.0 ± 0.9 Summer 26.4 ± 0.9 24.0 ± 2.0 2.7 ± 0.9 Fall 27.9 ± 2.1 22.9 ± 2.0 2.6 ± 1.0 2015 Spring 29.0 ± 0.9 17.8 ± 1.5 2.4 ± 1.2 Summer 29.7 ± 1.1 25.3 ± 1.0 2.5 ± 0.9 Fall 30.9 ± 0.8 25.2 ± 0.8 2.7 ± 1.2

Modeling of Circulation and Water Properties The purpose of this component of the project is to provide a model-based description of the changes in circulation and water properties that have resulted from the breach in Fire Island. We generated a new bathymetry for the model representing breach conditions as they were in September 2013, roughly one year after the breach was formed. We made several runs with this new bathymetry under winter conditions to examine the how the circulation has changed with the breach under tidal forcing. In the pre-breach conditions, the eastern part of Great South Bay was characterized by numerous small eddy features (Figure 5). Under the post-breach conditions, the tidal-mean currents clearly show that eddies in Bellport Bay and eastern Patchogue Bay have been replaced with a through-flow condition (Figure 5). This accompanied with the change from recirculating eddies to an east-to-west current in the eastern part of the bay is a change in residence time across the bay (Figure 6). Associated with these changes are much higher salinities in the eastern Bay than before the breach while the residence time in Bellport Bay has been reduced from about a month (pre-breach) to less than 10 days (post-breach; Figure 6).

Figure 5. Tidal-mean streamlines from the Finite Volume Coastal Ocean Model (FVCOM) showing pre- breach (upper panel) and post-breach (lower panel) under winter conditions in Great South Bay, New York.

Figure 6. Residence time (in days) before (upper panel) and with (lower panel) the breach in Great South

Bay , New York with darker colors representing short turnover and lighter colors representing long turnover.

Community Metrics The GSB otter trawl sampled 485 stations during the 12 cruises conducted in 2007 and from 2013– 2015. A total of 44,062 individuals were collected within this timeframe (Table 4a). There was an overall increase in the total number of species observed in the GSB following the breach relative to pre-breach conditions. Specifically, the total number of species was 36 in 2007, 49 in 2013, 62 in 2014 and 63 in 2015. Although, the total number of species recorded increased annually, the surveys were dominated by Anchovy spp., Blue Crab, Lady Crab, Spider Crab and Weakfish, accounting for 87.6% of all trawled catch (Table 4b). There was a shift in dominant species recorded pre- and post- breach in GSB (Figure 7). The most dramatic change pre- and post-breach has been the shift in CPUE of Blue Crab and Lady Crab (Figure 7), with a ~84% decline and >150% increase in Blue and Lady Crab, respectively following the breach. When comparing species that were collected pre- and post-breach, eight species showed an increase in CPUE following the breach; of these species, five are considered to generally prefer higher salinity habitats consistent with our hypothesis of greater occurrence of marine species, and include Summer Flounder, Windowpane Flounder, Bluefish, Sea Robin spp. and Mantis Shrimp (Figure 7). Also important to note is the increase in CPUE of Horseshoe Crab in years following the breach (Figure 7).

Table 4a. Total number of individuals for each taxonomic group caught within Great South Bay, New York during the trawl surveys conducted in 2007 and from 2013 - 2015.

Taxonomic Category Total number

Crustaceans 11414 Molluscs 119 Fishes 32397

Table 4b. Catch data for species caught within Great South Bay, New York during the trawl surveys conducted in 2007 and from 2013 - 2015. Species are grouped by major taxonomic categories and listed in descending order of abundance. Life-history strategy (LHS), total number, relative abundance (% catch) and body size (mean ± SD; range for measured individuals) are given for each species.

Taxonomic Relative Category Species Common name LHS1 Total number abundance Body size (mm) Crustaceans Callinectes sapidus Blue Crab E 6862 15.6 83.4 ± 36.0; 10-244 Ovalipes ocellatus Lady Crab M 2626 6.0 63.6 ± 17.0; 10-155 Libinia emarginata Spider Crab E 1560 3.5 50.4 ± 22.4; 7-190 Squilla empusa Mantis Shrimp M 206 0.5 118.6 ± 24.2; 35-180 Limulus polyphemus Horseshoe Crab E 134 0.3 223.7 ± 71.2; 20-330 Cancer irroratus Rock Crab M 35 0.1 50.2 ± 20.9; 18-100 Carcinus maenus Green Crab M 7 ≤ 0.05 36.1 ± 24.8; 11-68 Portunus spinnimanus Swimming Crab M 4 ≤ 0.05 41.3 ± 7.5; 35-50 Molluscs Squid spp. Squid spp. M 119 0.3 60.5 ± 29.3; 10-205 Fishes Anchoa spp. Anchovy Spp. E 25973 58.9 59.3 ± 17.7; 21-130 Cynoscion regalis Weakfish M 1605 3.6 106.6 ± 51.6; 21-412 Syngnathus fuscus Northern Pipefish E 455 1.1 145.7 ± 41.0; 45-240 Prionotus spp. Sea Robin Spp. E 428 1.0 152.6 ± 90.3; 23-390 Paralichthys dentatus Summer Flounder M 427 1.0 251.4 ± 110.1; 60-600 Pomatomus saltatrix Bluefish M 413 0.9 127.0 ± 42.9; 38-330 Sphoeroides maculatus Northern Puffer M 305 0.7 79.5 ± 43.0; 10-263 Leiostomus xanthurus Spot M 305 0.7 162.0 ± 30.7; 119-240 Menticirrhus spp. Kingfish Spp. M 290 0.7 187.8 ± 64.5; 44-358 Clupeidae spp. Ray-finned Fish Spp. M 289 0.7 66.9 ± 46.7; 28-354

1Note: We classified species as freshwater (F), estuarine (E) or marine (M) according to the guild scheme for estuarine fishes proposed by Elliot et al. (2007). Freshwater fishes include freshwater migrants and freshwater stragglers. Estuarine fishes include estuarine residents, migrants and putative amphidromous species. All marine species were marine migrants.

Table 4b (continued). Catch data for species caught within Great South Bay, New York during the trawl surveys conducted in 2007 and from 2013 - 2015. Species are grouped by major taxonomic categories and listed in descending order of abundance. Life-history strategy (LHS), total number, relative abundance (% catch) and body size (mean ± SD; range for measured individuals) are given for each species. Taxonomic Total Relative Category Species Common name LHS1 number abundance Body size (mm) Fishes (cont.) Menidia spp. Silverside Spp. E 285 0.7 83.5 ± 24.7; 20-102 Stenotomus chrysops Scup M 248 0.6 122.2 ± 51.4; 39-288 Pseudopleuronectes americanus Winter Flounder E 229 0.5 127.7 ± 55.7; 24-383 Tautog onitis Tautog M 198 0.5 129.3 ± 69.2; 28-438 Peprilus triacanthus Butterfish M 123 0.3 102.6 ± 55.8; 10-220 Ammodytes americanus American Sand Lance M 122 0.3 149.0 ± 38.7; 38-196 Scopthalmus aquosus Windowpane Flounder M 96 0.2 133.2 ± 81.2; 49-430 Chilomycterus schoepfi Striped Burrfish M 91 0.2 153.7 ± 72.2; 20-230 Centropristis striata Black Sea Bass M 79 0.2 122.6 ± 72.5; 35-263 Opsanus tau Oyster Toadfish E 62 0.1 98.4 ± 45.1; 42-186 Synodus foetens Inshore Lizardfish M 53 0.1 164.3 ± 45.7; 56-316 Tautogolabrus adspersus Cunner M 49 0.1 51.9 ± 14.0; 33-102 Etropus microstomus Smallmouth Flounder M 36 0.1 104.4 ± 41.0; 32-190 Trinectes maculatus Hogchoker E 36 0.1 159.1 ± 25.0; 100-204 Pogonias cromis Black Drum M 32 0.1 206.3 ± 60.4; 23-244 Anguilla rostrata American Eel F 31 0.1 127.0 ± 73.6; 45-235 Selene vomer Lookdown M 25 ≤ 0.05 77.6 ± 22.9; 46-113 Raja eglanteria Clearnose Skate M 24 ≤ 0.05 650.0 ± 57.2; 575-740 Gobiosoma bosci Naked Goby E 15 ≤ 0.05 34.6 ± 9.8; 28-52 Caranx hippos Crevalle jake M 11 ≤ 0.05 62.5 ± 18.8; 37-88 Selene septapinnis Atlantic Moonfish M 8 ≤ 0.05 101.5 ± 45.2; 40-144 Mustelus canis Smooth Dogfish M 7 ≤ 0.05 568.6 ± 183.1; 170-670 Dasyatis centroura Roughtail Stingray M 7 ≤ 0.05 975.0 ± 134.4; 880-1070 Hippocampus erectus Lined Seahorse E 6 ≤ 0.05 77.0 ± 45.4; 46-140 1Note: We classified species as freshwater (F), estuarine (E) or marine (M) according to the guild scheme for estuarine fishes proposed by Elliot et al. (2007). Freshwater fishes include freshwater migrants and freshwater stragglers. Estuarine fishes include estuarine residents, migrants and putative amphidromous species. All marine species were marine migrants.

Table 4b (continued). Catch data for species caught within Great South Bay, New York during the trawl surveys conducted in 2007 and from 2013 - 2015. Species are grouped by major taxonomic categories and listed in descending order of abundance. Life-history strategy (LHS), total number, relative abundance (% catch) and body size (mean ± SD; range for measured individuals) are given for each species. Taxonomic Total Relative Category Species Common name LHS1 number abundance Body size (mm) Fishes (cont.) Decapterus punctatus Round Scad M 6 ≤ 0.05 45.8 ± 6.2; 38-54 Gadidae spp. Cod Spp. M 5 ≤ 0.05 125.1 ± 33.7; 79-200 Melanogrammus aeglefinus Haddock M 4 ≤ 0.05 60.3 ± 15.0; 45-75 Microgadus tomcod Tomcod M 3 ≤ 0.05 43.3 ± 1.5; 42-45 Belonidae spp. Needlefish Spp. M 2 ≤ 0.05 163.5 ± 95.5; 96-231 Carangoides bartholomaei Yellow Jack M 1 ≤ 0.05 196 Gasterosteus aculeatus Three-Spine Stickleback E 1 ≤ 0.05 47 Dasyatis americana Southern Stingray M 1 ≤ 0.05 570 Cottidae spp. Sculpin Spp. M 1 ≤ 0.05 40 Epinephelus niveatus Snowy Grouper M 1 ≤ 0.05 39 Mugil cephalus Striped Mullet M 1 ≤ 0.05 194 Bairdiella chrysoura Silver Perch M 1 ≤ 0.05 81 Cyprinodon variegatus Sheepshead Minnow E 1 ≤ 0.05 41 Fistulariidae spp. Cornetfish M 1 ≤ 0.05 185 Sphyraena borealis Northern Sennet M 1 ≤ 0.05 208 Leucoraia erinacea Little Skate M 1 ≤ 0.05 450 Myoxocephalus octodecimspinosus Longhorn Sculpin M 1 ≤ 0.05 60 Mycteroperca microlepis Gag Grouper M 1 ≤ 0.05 113 Seriola zonata Banded Rudderfish M 1 ≤ 0.05 49 1Note: We classified species as freshwater (F), estuarine (E) or marine (M) according to the guild scheme for estuarine fishes proposed by Elliot et al. (2007). Freshwater fishes include freshwater migrants and freshwater stragglers. Estuarine fishes include estuarine residents, migrants and putative amphidromous species. All marine species were marine migrants.

Figure 7. Catch composition (% catch per unit effort) from the Great South Bay, New York otter trawl survey by year. Anchovy spp. was excluded from calculations because it dominated catch and makes discerning the other species.

Additional indicators of species composition change characterized the baywide-response of the fish community to the breach. Variation in CPUE was observed across years (ANOVA: F3,450 = 9.883, P = 0.000; Figure 8), with post-hoc analyses indicating that CPUE was significantly higher in 2007 and 2013 relative to 2014 and 2015, but that both of these years did not differ from one another (Figure 8). Both species richness (K-W: χ2 = 27.396, df = 3, P = 0.0001) and diversity (K-W: χ2 = 25.159, df = 3, P = 0.0000) were significantly higher after the breach, particularly in 2015, than before the breach (Figure 8). As noted above, this increase in richness and diversity is largely attributable to greater occurrence of marine and increased exchange of energy between the bay and ocean, consistent with our hypotheses. Total length (ANOVA: F3,450 = 3.071, P = 0.017) and estimated- biomass (ANOVA: F3,450 = 6.950, P = 0.000) also differed significantly by year, with post-hoc analyses indicating significantly higher values in 2015 relative to 2007.

Assemblage Structure The assemblages varied annually in GSB, reflecting the dynamic temporal characteristics of this system (Figure 9). Overall the CCA explained 14% of the total inertia in the dataset, with the first CCA axis explaining 9% of the fitted relationship and the second CCA axis explaining 5%. Permutation tests indicated that a relationship occurred between the assemblages and three environmental variables; salinity, temperature and year (Table 5). The 2007 assemblage is associated with lower salinity and the 2015 assemblage is associated with higher salinity. The proximity of the 2013 and 2014 centroids corresponds to their respective similarity in average species composition (Figure 9). As well, these assemblages ordinated intermediate between 2007 and 2015. The effects of salinity and temperature were largely independent, as indicated by the perpendicular vectors (Figure 9). These findings are consistent with our hypotheses that increased connectivity between the ocean and bay would provide more suitable habitat, i.e., higher salinity, and access for marine species, shifting the overall assemblage. Specifically, higher catch rates are observed for Mantis Shrimp, Lady Crab, Squid Spp., in 2015 compared with all other years, shifting the overall assemblage to one more indicative of ocean influence (Figure 9, 10).

Table 5. Results of the permutation tests for the canonical correspondence analysis (CCAs) of catch per unit effort (CPUE).

Environmental variable df χ2 F P

Temperature 1 0.254 28.137 0.005* Salinity 1 0.084 9.323 0.005* Depth 1 0.063 6.973 0.015 Year 3 0.218 8.082 0.005* Residual 447 4.041 − −

* indicate significance

Figure 9. Canonical correspondence analysis (CCA) ordination for catch per unit effort (CPUE) of species collected in trawl surveys conducted in 2007 and 2013 - 2015 in Great South Bay, New York. Arrows indicate the increasing gradient of the explanatory variables with the origin representing the mean. Squares represent the centroids for each year of sampling. Species scores are represented by points (AN: Anchovy spp.; BC: Blue Crab; BF: Butterfish; BL: Bluefish; BS: Black Sea Bass; CL: Clupeid spp.; CU: Cunner; HC: Horseshoe Crab; KG: Kingfish spp.; LC: Lady Crab; MA: Mantis Shrimp; MINOR: species < 5% cumulative frequency; NP: Northern Puffer; PF: Pipefish spp.; PG: Scup; SB: Striped Burrfish; SC: Spider Crab; SF: Summer Flounder; SM: Smallmouth Flounder; SO: Spot; SQ: Squid spp.; SR: Sea Robin spp.; SV: Silverside spp.; TG: Tautog; WD: Windowpane Flounder; WF: Winter Flounder; WK: Weakfish).

Figure 10. Ordination from catch per unit effort (CPUE) of species collected in trawl surveys conducted in 2007 and 2013 - 2015 in Great South Bay, New York with response surfaces for the main environmental variables, A) salinity and B) temperature. Species are represented as follows: AN: Anchovy spp.; BC: Blue Crab; BF: Butterfish; BL: Bluefish; BS: Black Sea Bass; CL: Clupeid spp.; CU: Cunner; HC: Horseshoe Crab; KG: Kingfish spp.; LC: Lady Crab; MA: Mantis Shrimp; MINOR: species < 5% cumulative frequency; NP: Northern Puffer; PF: Pipefish spp.; PG: Scup; SB: Striped Burrfish; SC: Spider Crab; SF: Summer Flounder; SM: Smallmouth Flounder; SO: Spot; SQ: Squid spp.; SR: Sea Robin spp.; SV: Silverside spp.; TG: Tautog; WD: Windowpane Flounder; WF: Winter Flounder; WK: Weakfish.

The percentage of deviance explained by the response surfaces of the CCA axis scores is 47.9% and 20.1% for temperature and salinity, respectively. The fitted response surface for both variables was statistically significant (temperature: F = 44.81, P < 0.000, df = 6.75; salinity: F = 11.84, P < 0.000, df = 5.74). For salinity, the majority of species were associated with salinities > 28 psu, highlighting their higher catch rates in higher salinity waters (Figure 10A). For temperature, the majority of species were associated with temperatures < 23°C, showing higher catch rates at lower temperatures (Figure 10B).

Species' Environmental Preference The distributions of Blue and Lady Crabs changed following the breach. Blue Crabs were distributed throughout GSB in 2007 (Figure 11). After the breach, Blue Crab density was highest in the north portion of GSB (Figure 11), with fewer individuals being collected in the otter trawl in the western portion of GSB. Overall, blue crab CPUE decreased 2013 through 2015. In 2014, Blue Crabs were largely collected in the eastern portion of the bay in lower densities (Figure 11). In contrast, Lady Crabs exhibited the opposite trend in GSB (Figure 12). For example, prior to the beach, in 2007, lady crab CPUE was very low and the highest CPUEs were observed in the western portion of GSB adjacent to Fire Island Inlet (Figure 12). Lady Crab expanded its range with a more uniform distribution throughout GSB by 2015 (Figure 12) and GSB CPUE increased 2013 through 2015. As all species associate with specific environmental conditions, the higher catch rates of Lady Crab post- breach are consistent with our hypothesis that increased salinity in GSB would result in changes in abundance and distribution of species with preferences for higher salinities.

For both species, the probabilities of occurrence and abundance were influenced by abiotic and biotic factors (Table 6). The occurrence model (PA) for Blue Crab identified bottom temperature, salinity and depth (Table 6) as factors influencing the probability of presence of this species in GSB. The probability increased at temperatures of ~23°C, salinities of ~28 psu and depth < 3m. The temporal factor year also was identified to affect the presence of Blue Crab in the GSB. The abundance model (ABUN) for Blue Crab similarly identified the importance of bottom temperature and year (Table 6) as factors affecting the CPUE of the species. The presence model for Lady Crab identified temperature and depth (Table 6) as factors influencing the probability of occurrence of this species in GSB. The probability of Lady Crab presence was greatest at temperatures ~20°C and depth > 3m. Salinity was not included in the PA model, probably because few Lady Crabs were collected during 2007, the period with the lowest salinities. Year was also identified to influence the occurrence of Lady Crabs (Table 6). The abundance of Lady Crabs was affected by all three continuous environmental variables, salinity, temperature and depth (Table 6). Specifically, CPUE of Lady Crab was highest at temperatures of 23-26°C, salinities > 30 psu and depths > 3 m.

Figure 11. Catch per unit effort (CPUE) of Blue Crabs sampled during the otter trawl survey conducted in 2007 and from 2013-2015 in Great South Bay, New York. Data are CPUE.

Figure 12. Catch per unit effort (CPUE) of Lady Crabs sampled during the otter trawl survey conducted in 2007 and from 2013-2015 in Great South Bay, New York. Data are CPUE.

Species Model n AICc AIC wi Dev (%)

Callinectes PA ~ factor(Year) + s(Temperature) + s(Salinity) + s(Depth) 325 442.47 0.943 29.9 sapidus, ABUN ~ factor(Year) + s(Temperature) 325 1013.70 0.271 30.5 (Blue Crab) Ovalipes PA ~ factor(Year) + s(Temperature) + s(Depth) 111 408.91 0.483 24.0 ocellatus, ABUN ~ s(Temperature) + s(Salinity) + s(Depth) 111 390.01 0.519 18.1 (Lady Crab)

Ecosystem-Level Differences Fitted Ecosim Model The fitted Ecosim model resulted in a log sum of squares of 32.91. The predicted abundances matched observed trends fairly well for Weakfish, Bluefish, Tautog and Winter Flounder (Figure 13). The predicted abundance generally followed trends of Summer Flounder abundance with the exception of recent predicted values increasing above observed (Figure 13). The 2015 Summer Flounder discrepancy may result from a large biomass of adult Summer Flounder and several years of poor recruitment (ASMFC 2015); thus, the model is responding to declines in harvest and assuming average recruitment. Striped Bass abundance trends were only generally predicted by the model. The Ecosim model was fitted to the three years GSB survey data to both calibrate the Ecosim run and provide information of the trends in species abundance for American Eel, Black Sea Bass, Bluefish, Blue Crab, Butterfish, Horseshoe Crab, Menhaden, Northern Kingfish, Spider Crab and Oyster Toadfish (Figure 14). The model estimated fishing mortality based on assumed (forced) observed catches (Figure 15).

Figure 14. Ecosim modeled trends between observed (points) and predicted (line) biomass (t/km2) estimates from the GSB trawl conducted in 2013-2015 in Great South Bay, New York. Note different scales for each species

Figure 15. Catches forced in Ecosim model. Note different scales for each species.

Network Analysis Network analyses were performed using the Ecosim network analysis module. Several network analysis and structural metrics were estimated including system primary production (PP), respiration (R), biomass (B), primary production/biomass ratio (PP/B), primary production/respiration (PP/R), total system throughput (TST), system omnivory index (SOI) and Kempton’s Q (Q). Primary productivity measures the productivity of lower trophic groups. System R provides a measure of upper trophic system productivity. Total system throughput provides a measure of flow in the system and represents the size of the system (Ulanowicz 1986). The system omnivory index is an indicator of the trophic level (TL) of foraging for the system (Christensen et al. 2000). Kempton’s Q index provides a metric of the biomass of species above a TL of 3.0. Maturing systems are expected to show PP/R approaching 1.0, increases in SOI and PP/R and a decrease in PP/B. Increases in Q suggest an increase in system upper trophic diversity.

Consistent with our third overarching hypothesis, the impact of ocean water exchange through the breach was evident in changes across all system structure with an increase in nearly all maturity metrics in the years following the storm (Table 7). However, contrary to our hypothesis PP/R and PP/B suggest a slightly less mature system; while, SOI indicates a maturing system. Importantly, system biomass and Q increased following the breach suggesting that the system has increased in biomass and upper-trophic species abundance (Figure 16A). Moreover, consistent with our hypotheses an overall increase in filter feeding, total upper trophic and migratory biomass was observed (Figure 16B). The system structure and maturity appear to have been impacted by the breach; however, the system’s response appears to still be developing. These results should be viewed with caution for the reasons expressed in the model caveats section.

Figure 16. A Trends in total biomass and Kempton’s Q (measure of the biomass of species with trophic level>3) from 2007 to 2015. B Trends in biomass (t/km2) of Ecopath groupings from 2007 to 2015. Filter feeders include: hard clams, oysters, suspension feeders, Menhaden and Shrimp spp. Benthic biomass include: benthic fauna (e.g., whelks), Mantis Shrimp and crab spp. Upper trophic levels include species/groupings with TL >3. Migratory species include: squid, forage fish, tropical fish, skates, sharks, Black Sea Bass, Bluefish, Butterfish, Gadoids, Drums, Striped Bass, Scup, Sea Robin spp., Flatfish spp. and Weakfish.

Table 7. A summary of metrics for structural properties of Great South Bay, New York as modeled in Ecopath for 2007-2015.

Category Indicator 2007 2008 2009 2010 2011 2012 2013 2014 2015 System Primary production (PP) 243000.0 242998.3 243031.1 243090.5 243021.8 242839.2 242767.0 243118.6 243220.1 Indices Production (P) 242654.8 242645.9 242732.4 242781.6 242620.8 242293.1 242340.6 242892.2 242994.8 Respiration (R) 345.2 352.3 298.8 308.9 401.0 546.1 426.4 226.3 225.3 Production/Biomass Ratio 59.1 59.1 59.0 59.1 58.9 58.8 58.9 59.1 59.2 (PP/B) Production/Respiration 703.9 689.8 813.5 786.9 605.9 444.7 569.3 1074.1 1079.7 (PP/R) Network Total System Throughput 486534.0 486548.6 486523.0 486592.3 486496.8 486218.3 486078.1 486645.9 486858.5 Analysis (TST) Indices System Omnivory Index 0.33 0.32 0.29 0.28 0.27 0.29 0.31 0.33 0.30 (SOI)

Model Limitations and Caveats Catch Time Series The Nuttall et al. (2011) model was created on catch data that was available before the NOAA’s current Marine Recreational Information Program (MRIP) methodology was implemented. We were unable to use the Ecopath structure of the 2000s model published by Nuttall et al. (2011) as it was not comparable to current catch estimation methods. In many ways the Ecosim model presented represents a better method to compare the impacts of superstorm Sandy on the GSB system. However, the Ecosim process is far more involved and time-intensive than originally allocated for the Ecopath model. Throughout the modeling process the catch data remained a significant challenge for the Ecosim model. The harvest values for all finfish showed great inter-annual variability; which was not likely accurate or precise. Ecosim had trouble statistically fitting the catch trends as the inter- annual variation placed a large burden on the dynamic model to reproduce the catches and produce a reasonable biomass time series for individual species. While the Ecosim model results are reasonable, future work is needed to rectify or perhaps smooth the catch data, to more accurately reflect the fishery. Abundance Time Series The Ecosim model requires a longer timer series than provided by the three year GSB otter trawl survey. Given the observed variability in the system, additional years of data would improve model fits. We were able to use time series for six important species derived from regional surveys and stock assessments; however, the potential remains to extend both the abundance and catch series to early periods to capture longer-term dynamics of the GSB ecosystem. The GSB survey provided data for 10 species that were helpful in the calibration of the Ecosim runs. Ecosim Modelling The results presented in this report should be viewed as preliminary as further model vetting is needed before the model can be considered final. Model evaluation can be achieved statistically through the fitting of time series and also by evaluating whether the model produces ‘credible behavior’ (Heymans et al. 2016). The fitting of time series allows initial evaluation of the model predicted behavior but additional vetting is needed to insure model outputs are realistic. For example, the model predicted exploitation rates have only been rudimentary vetted. Further, evaluation of model structure is needed to explore whether multi-stanzas should be added for key species to better represent species specific dynamics. Finally, time series fit could be evaluated using Akaike’s Information Criteria (AIC) to compare various model structures and key runs (Heymans et al. 2016).

Discussion

Observations and models indicate a clear change in environmental conditions before and after the breach followed by an increase in diversity and overall CPUE of finfish and mobile invertebrates including higher trophic level species like summer flounder. These initial findings suggest that the connectivity of GSB with the ocean is increasing the trophic complexity and maturity of the system. The increase in biomass and diversity did not happen immediately in 2013, emphasizing the importance of sampling in 2014 and 2015 and systematically into the future. Rather, over three years of sampling we observed the biomass, species abundances and diversity changing slowly. A slow transition following a perturbation makes sense given that these changes are the result of population- and community-level processes including recruitment, growth and survival if annual cohorts.

Network analysis was performed using the Ecosim network analysis module. Several network analysis and structural metrics were estimated including system primary production (PP), respiration (R), biomass (B), primary production/biomass ratio (PP/B), primary production/respiration (PP:R), total system throughput (TST), system omnivory index (SOI) and Kempton’s Q (Q). Primary productivity measures the productivity of lower trophic groups. System R provides a measure of upper trophic system productivity. Total system throughput provides a measure of flow in the system and represents the size of the system (Ulanowicz 1986). The system omnivory index is an indicator of the trophic level of foraging for the system (Christensen and Walters 2000). Kempton’s Q index provides a metric of the biomass of species above a trophic level of 3.0. Maturing systems are expected to show PP/R approaching 1.0, increases in SOI and PP/R and a decrease in PP/B. Increases in Q suggest an increase in system upper trophic diversity.

Conclusions

In conclusion, our study demonstrates that there are clear differences in the temperature, salinity and, the fish and mobile invertebrate species assemblage of GSB pre- and post-breach. There was a persistent bay-wide shift towards higher mean salinity following the breach, with 2015 having the highest reported salinity overall. This increase in salinity post-breach supports the hypothesis of increased availability of habitat for marine species throughout much of the bay and not limited to the western bay near Fire Island Inlet. There was a clear increase in the number of species, species diversity and derived biomass observed in GSB, consistent with our hypothesis that the system would be more mature after the breach. As noted above, this increase in richness and diversity is largely attributable to greater occurrence of marine species, specifically, Summer Flounder, Squid spp., Windowpane Flounder, Bluefish, Sea Robin spp., Lady Crab and Mantis Shrimp using the higher salinity waters of GSB. Although marine species were more abundant after the breach, this finding is best represented by a switch in the dominant crab species from the estuarine Blue Crab to the more marine Lady Crab. Moreover, as all species associate with specific environmental conditions, the higher catch rates of Lady Crab post-breach are consistent with our hypothesis that increased salinity in GSB would result in changes in abundance and distribution of species with varying preferences for temperature and salinity. Whether these abiotic and biotic changes to GSB are advantageous for the ecosystem remains to be understood. Yet, the increase in diversity, species richness and CPUE of higher trophic level predators, such as Summer Flounder following the breach is indicative of initial recovery of ecosystem maturity. The patterns in the finfish and invertebrate community suggest an increase in connectivity with the ocean, which may result in an increase in ecosystem maturity following the breach.

Some Ecosim model output metrics suggest that the breach, during and since, Tropical Storm Sandy impacted system structure and maturity. However, the metrics did not all indicate an increase in system maturity. An increase in system maturity was reflected in Kempton’s Q (Q); which measures the biomass of species with a trophic level >3. The trend in Q indicates that a greater exchange with the ocean, and/or internal system dynamics, have resulted in an increase of upper trophic species. Upper trophic species play an important role in system maturity and are commonly targeted by recreational and commercial species. Similarly, total system biomass (B) increased indicating a positive impact on the size of the biological community following the breach. Caution should be used when interpreting the model results until further development of the model is conducted. Finally, the system appears in a transition period following the breach, during and since, Tropical Storm Sandy and it is likely continuing. Collectively these results suggest that restored connectivity between GSB and the ocean will continue to raise the level of resilience of GSB nekton assemblages.

Literature Cited

Akaike, H. 1973. Information theory and an extension of the maximum likelihood principle. Pages 267-281 in B. N. Petran and F. Csaki, editors. International Symposium on Information Theory, 2nd Edition, Budapest, Akademiaiiadi.

Alpine, A. E., and J. E. Cloern. 1992. Trophic interactions and direct physical effects control phytoplankton biomass and production in an estuary. Limnology & Oceanography 37:946-955.

Anthony, A., J. Atwood, P. V. August, C. Byron, et al. 2009. Coastal lagoons and climate change: ecological and social ramifications in the US Atlantic and Gulf coast ecosystems. Ecology & Society 14(1):8.

Ashton, A., J. Donnelly, and R. Evans. 2008. A discussion of the potential impacts of climate change on the shorelines of the Northeastern USA. Mitigation & Adaptation Strategies for Global Change 13(7):719-743.

Atlantic States Marine Fisheries Center (ASMFC). 2015. Stock Assessment Update of Summer Flounder for 2015. 17 pp.

Atlantic States Marine Fisheries Center (ASMFC). 2016. Weakfish Benchmark Stock Assessment and Peer Review Report. 263 pp.

Atlantic States Marine Fisheries Center (ASMFC). 2014. Interstate Fishery Management Plan for Winter Flounder. 16 pp.

Atlantic States Marine Fisheries Center (ASMFC). 2016. Tautog Stock Assessment Update. 143 pp.

Bird, E. C. F. 1992. Submerging coasts: The effects of a rising sea level on coastal environments. John Wily & Sons, Chichester, England, 192 pp.

Borcard, D., F. Gillet, and P. Legendre. 2011. Numerical Ecology with R. Springer-Verlag, New York, 306 pp.

Brotons, L., W. Thuiller, M. B. Araújo, and A. H. Hirzel. 2004. Presence-absence versus presence- only modelling methods for predicting bird habitat suitability. Ecography 27(4):437-448.

Buchheister, A., C. Bonzek, J. Gartland, and R. Latour. 2013. Patterns and drivers of the demersal fish community of Chesapeake Bay. Marine Ecology Progress Series 481:161-180.

Campana, S. E., and W. N. Joyce. 2004. Temperature and depth associations of porbeagle shark (Lamna nasus) in the northwest Atlantic. Fisheries Oceanography 13(1):52-64.

Carpenter, E. J., and B. M. Brinkhuis. 1991. Primary Production and Nitrogenous Nutrients in Great South Bay. Pages 5-8 in J. R. Schubel, T. M. Bell, and H. H. Carter, editors. The Great South Bay. State University of New York, Albany Press.

Cerrato, R. M., P. V. LoCicero, and S. L. Goodbred. 2013. Response of mollusc assemblages to climate variability and anthropogenic activities: A 4000-year record from a shallow bar-built lagoon system. Global Change Biology 19:3024-3036.

Chen, C., H. Liu, and R. C. Beardsley. 2003. An unstructured grid, finite-volume, three-dimensional, primitive equations ocean model: application to coastal ocean and estuaries. Journal of Atmospheric & Oceanic Technology 20:159-186.

Christensen, V., A. Beattie, C. Buchanan, H. Ma, S. J. D. Martell, et al. 2009. Fisheries ecosystem model of the Chesapeake Bay: methodology, parameterization, and model explanation. U.S. Department of Commerce, NOAA Tech. Memo. NMFS-F/SPO-106.

Christensen, V., and C. Walters. 2000. Ecopath with Ecosim: methods, capabilities and limitations. Pages 79–105 in D. Pauly, T. J. Pitcher, editors. Methods for evaluating the impacts of fisheries on North Atlantic Ecosystems. Fisheries Centre, University of British Columbia, Vancouver, BC.

Christensen, V., and J. Maclean. 2011. Ecosystem approaches to fisheries: A global perspective. Cambridge University Press, New York, 342 pp.

Christensen, V., and D. Pauly. 1992. Ecopath II—a software for balancing steady-state ecosystem models and calculating network characteristics. Ecological Modeling 61(3):169-185.

Damalas, D., C. D. Maravelias, S. Katsanevakis, A. P. Karageorgis, and C. Papaconstantinou. 2010. Seasonal abundance of non-commercial demersal fish in the eastern Mediterranean Sea in relation to hydrographic and sediment characteristics. Estuarine, Coastal and Shelf Science 89(1):107-118.

Dickinson, C. L. 1938. Commercial Fisheries. Pages 15-46 in A Biological Survey of the Salt Waters of Long Island, Part I. New York State Conservation Department.

Freeman, E. 2007. Presence-Absence: An R package for presence-absence model evaluation. USDA Forest Service, Rocky Mountain Research Station, Ogden, Utah.

Elliott, M., A. K. Whitfield, I. C. Potter, S. J. M. Blaber, D. P. Cyrus, et al. 2007. The guild approach to categorizing estuarine fish assemblages: a global review. Fish and Fisheries 8:241-268.

Frisk M. G., and S. B. Munch. 2008. Great South Bay Fishery Independent Survey. Stony Brook, NY, 21 pp.

Fry, F. E. J. 1971. The effect of environmental factors on the physiology of fish. Pages 1-98 in W. S. Hoar, and D. J. Randall, editors. Fish Physiology, Vol 6. Academic Press, New York.

Gabriel, R. H. 1921. The Evolution of Long Island. Yale University Press, New Haven, Connecticut, 194 pp.

Gobler, C. J., D. J. Lonsdale, and G. L. Boyer. 2005. A Review of the Causes, Effects, and Potential Management of Harmful Brown Tide Blooms caused by Aureococcus anophagefferens. Estuaries 28(5):726-749.

Greene, G. 1982. Hard Clams, Competitors, Predators, and Physical Parameters in Great South Bay, NY. Draft Report submitted to US Environmental Protection Agency, Region II, New York, NY, by Wapora Inc. pp 157.

Guisan, A., T. C. Edwards, and T. Hastie. 2002. Generalized linear and generalized additive models in studies of species distributions: setting the scene. Ecological Modeling 157(2-3):89-100.

Hanlon, J. R. 1983. Fish and Wildlife Resource Studies for the Fire Island Inlet to Montauk Point, New York, Beach Erosion Control and Hurricane Protection Project Reformulation Study, Estuarine Resource Component. Long Island Sub Office, US Department of the Interior, Fish and Wildlife Service, Region 5.

Hastie, T. J., and R. J. Tibshirani. 1990. Generalized Additive Models. Chapman & Hall, New York, NY. 352 pp.

Heinänen, S., M. Rönkä, and M. Von Numers. 2008. Modelling the occurrence and abundance of a colonial species, the arctic tern Sterna paradisaea in the archipelago of SW Finland. Ecography 31(5):601-611.

Heymans, J. J., M. Coll, J. S. Link, S. Mackinson, J. Steenbeck, C. J. Walters, and V. Christensen. 2016. Best practice in Ecopath with Ecosim food web models for ecosystem-based management. Ecological Modelling 331:173-184.

Hinga, K. R. 2005. Water quality and ecology of Great South Bay (Fire Island National Seashore Science Synthesis Paper). Technical Report NPS/NER/NRTR—2005/019. National Park Service. Boston, MA.

Hinrichs, C. 2018. Assessment of hydrodynamic changes in the Great South Bay after breaching during Hurricane Sandy. Doctoral Dissertation. Stony Brook University, New York.

Hurvich, C.M., and C. Tsai. 1989. Regression and time series model selection in small samples. Biometrika 76: 297–307.

Johnson, J. B., and K. S. Omland. 2004. Model selection in ecology and evolution. Trends in Ecology & Evolution 19:101-108.

Leatherman, S. P. 1985. Geomorphic and stratigraphic analysis of Fire Island, New York. Marine Geology 63(1-4):173-195.

Leathwick, J. R., J. Elith, and T. Hastie. 2006. Comparative performance of generalized additive models and multivariate adaptive regression splines for statistical modelling of species distributions. Ecological Modeling 199(2):188-196.

Legendre, P., and L. Legendre. 1998. Numerical Ecology, Volume 24. 2nd Edition, Elsevier Science 852 pp.

Libralato, S., V. Christensen, D. Pauly, D. 2006. A method for identifying keystone species in food web models. Ecological Modelling 195:153–171.

Mackenzie, B. R., H. Gislason, C. Mollmann, and F. W. Koster. 2007. Impact of 21st century climate change on the Baltic Sea fish community and fisheries. Global Change Biology 13(7):1348-1367.

Manly, B. F. J. 1991. Randomization and Monte Carlo Methods in Biology. Chapman & Hall, London, UK. 480 pp.

McCullagh, P., and J. A. Nelder. 1989. Generalized Linear Models Second Edition. Chapman & Hall, London, UK. 532 pp.

McGarigal, K., S. Cushman, and S. Stafford. 2000. Multivariate Statistics for Wildlife and Ecology Research. Springer-Verlag, New York, New York. 283 pp.

McHugh, J. L. 1991. The hard clam fishery past and present. in J. R. Schubel,, T. M. Bell, H. H. Carter (Eds.), The Great South Bay. SUNY press.

Morton, R. A., and A. H. J. Sallenger. 2003. Morphological impacts of extreme storms on sandy beaches and barriers. Journal of Coastal Research 19(3):560-573.

Nelson, T. R., J. L. Miselis, C. J. Hapke, O. T. Brenner, R. E. Henderson, B. J. Reynolds, and K. E. Wilson. 2017. Bathymetry data collected in October 2014 from Fire Island, New York—The wilderness breach, shoreface, and bay: USGS Data Series 1034, https://doi.org/10.3133/ds1034.

Northeast Fisheries Science Center (NEFSC). 2015. Bluefish Benchmark Stock Assessment for 2015 60th SAW Assessment Report. 536 pp.

Northeast Fisheries Science Center (NEFSC). 2016. Bluefish Stock Assessment Update: Data Update through 2015. 28 pp.

Northeast Fisheries Science Center (NEFSC). 2016. Scup Benchmark Stock Assessment for 2015. 323 pp.

Nuttall, M. A., A. Jordaan, R. M. Cerrato, and M. G. Frisk. 2011. Identifying 120 years of decline in ecosystem structure and maturity of Great South Bay, New York using the Ecopath modelling approach. Ecological Modeling 222(18):3335-3345.

Nye, J. A., T. M. Joyce, Y. Kwon, and J. S. Link. 2011. Silver hake trachs changes in Northwest Atlatnic circulation. Nature Communications 2:412.

Odum, E. P. 1971. Fundamentals of Ecology. WB Saunders Co, Philadelphia. 574 pp.

Odum, E. P. 1969. The Strategy of Ecosystem Development. Science 164:262-270.

Oksanen, J., F. G. Blanchet, M. Friendly, R. Kindt, P. Legendre, et al. 2016. Community Ecology Package in R. 2017-01-17 19:41:27

Paerl, H., L. Valdes, A. Joyner, B. Peierls, et al. 2006. Ecological response to hurricane events in the Pamlico Sound system, North Carolina, and implications for assessment and management in a regime of increased frequency. Estuaries & Coasts 29(6):1033-1045.

Pauly, D., V. Christensen, and C. J. Walters. 2000. Ecopath, Ecosim, and Ecospace as tools for evaluating ecosystem impact of fisheries. ICES Journal of Marine Science 57(3):697-706.

Pearce, J., and S. Ferrier. 2000. Evaluating the predictive performance of habitat models developed using logistic regression. Ecological Modeling 133(3): 225-245.

Perry, R. I., and S. J. Smith. 1994. Identifying habitat associations of marine fishes using survey data: an application to the Northwest Atlantic. Canadian Journal of Fisheries & Aquatic Sciences 51(3): 589-602.

Pikitch, E. K., C. Santora, E. A. Babcock, A. Bakun, et al. 2004. Ecosystem-based fishery management. Science 305(5682):346-347.

Polovina, J. 1984. An overview of the ECOPATH model. Fishbyte 2(2):5-7.

Potts, J. M., and J. Elith. 2006. Comparing species abundance models. Ecological Modeling 199(2): 153–160.

R Development Core. 2016. R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. http://www.R-project.org

Robin, X., N. Turck, A. Hainard, N. Tiberti, F. Lisacek, J. Sanchez, and M. Muller. 2011. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 12:77.

Scavia, D., J. Field, D. Boesch, R. Buddemeier, et al. 2002. Climate change impacts on U.S. Coastal and Marine Ecosystems. Estuaries 25(2):149-164.

Schubel, J. R. 1991. Introduction in J. R. Schubel, T. M. Bell, and H. H. Carter, editors. The Great South Bay. State University of New York Press, Albany, NY. 107pp

Schreiber, R. A. 1973. The Fishes of Great South Bay. MS Thesis. Stony Brook University, Stony Brook, NY. pp 199.

Smith, L. M., S. Whitehouse, C. A. Oviatt. 2010. Impacts of climate change on Narragansett Bay. Northeastern Naturalist 17(1):77-90.

Stevens, P., D. Blewett, J. P. Casey. 2006. Short-term effects of a low dissolved oxygen event on estuarine fish assemblages following the passage of hurricane Charley. Estuaries & Coasts 29(6):997-1003.

Stoner, A. W., J. P. Manderson, and J. P. Pessutti. 2001. Spatially explicit analysis of estuarine habitat for juvenile winter flounder: combining generalized additive models and geographic information systems. Marine Ecology Progress Series 213:253-271.

Swets, J. A. 1988. Measuring the accuracy of diagnostic systems. Science 240(4857):1285-1293.

Ter Braak, C. J. F., and P. Smilauer. 2012. Canoco Reference Manual and User's Guide. Microcomputer Power, Ithaca, NY, 496 pp.

Ter Braak, C. J. F. 1986. Canonical Correspondence Analysis: A New Eigenvector Technique for Multivariate Direct Gradient Analysis. Ecology 67:1167-1179.

Ulanowicz, R. E. 1986. Growth and Development: Ecosystem Phenomenology. Springer Verlag, New York. 203 pp.

Walters, C. J., and F. Juanes. 1993. Recruitment limitation as a consequence of natural selection for use of restricted feeding habitats and predation risk taking by juvenile fishes. Canadian Journal of Fisheries & Aquatic Sciences 50:2058-2070.

Wintle, B. A., J. Elith, and J. M. Potts. 2005. Fauna habitat modelling and mapping: a review and case study in the Lower Hunter Central Coast region of NSW. Austral Ecology 30(7):719-738.

Wood, S. N. 2006. Generalized Additive Models: an Introduction with R. Chapman & Hall/CRC, Boca Raton, FL, 410 pp.

Wood, S. N. 2011. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society (B) 73(1):3-36.

Wood, S. N. 2003. Thin plate regression splines. Journal of the Royal Statistical Society (B) 65(1):95-114.

Young, G. C., and I. C. Potter. 2002. Influence of exceptionally high salinities, marked variations in freshwater discharge and opening of estuary mouth on the characteristics of the ichthyofauna of a normally-closed estuary. Estuarine, Coastal & Shelf Science 55(2):223-246.

The Department of the Interior protects and manages the nation’s natural resources and cultural heritage; provides scientific and other information about those resources; and honors its special responsibilities to American Indians, Alaska Natives, and affiliated Island Communities.

NPS 962/147743, August 2018 49

National Park Service U.S. Department of the Interior

Natural Resource Stewardship and Science 1201 Oakridge Drive, Suite 150 Fort Collins, CO 80525

EXPERIENCE YOUR AMERICA TM