Core Sound Food Web

Information on this web page is based on a doctoral disseration from East Carolina University:

MEASURING THE ECOSYSTEM IMPACTS OF COMMERCIAL SHRIMP TRAWLING AND OTHER FISHING GEAR IN CORE SOUND, NORTH CAROLINA USING ECOLOGICAL NETWORK ANALYSIS
by Rebecca Anne Deehr
July, 2012

Director: Joseph J. Luczkovich, PhD
Coastal Resources Management Program

Abstract

The impacts of commercial trawling are well documented, especially alteration of benthic environments, removal of targeted and by-catch species, and alteration of food webs. I investigated and modeled the impacts of shrimp trawling on the estuarine ecosystem in Core Sound, North Carolina.  Since 1978, the North Carolina Division of Marine Fisheries has enforced no-trawling rules in nursery areas, but most of Core Sound is open to trawling. This “natural experiment” allowed me to compare the ecosystem impacts trawling using ecological network models.  I used field collections, fisheries data from the NC Trip Ticket program, and Ecopath network modeling software to create four network models of areas open and closed to shrimp trawling during spring (2007) and fall (2006 and 2007).  Each model consisted of 65 compartments (including non-living detritus, by-catch, producers, and various invertebrate and vertebrate consumers), and harvests by different types of fishery gears (crab pots, gill nets, haul seines, and pound nets in closed areas; shrimp trawls, skimmer trawls were added to the models in areas open to trawling). 
Approximately 12,000 shrimp trawling trips occurred from 2001 – 2007 in areas open to trawling, suggesting the potential for large trawling impacts.  Based on the benthic sampling, shrimp trawling had a major impact on the Core Sound ecosystem.  Contrary to expectation, biomass (g C/m2) of infaunal benthic invertebrates, especially deposit-feeding polychaetes, was significantly greater in areas open to trawling.  Meiofaunal biomass was significantly greater in the closed areas.  Field collections of fish and invertebrates revealed that there was more biomass (g C/m2) of benthic-invertebrate feeders (such as spot, pinfish and blue crabs) in areas closed to trawling.  These results suggest a trophic cascade due to trawling may have occurred in the open areas, whereby trawls removed benthic-feeding fishes and blue crabs, released their prey (benthic polychaetes) from predation pressure, and lowered the abundance of meiofauna (prey of the polychaetes).  Alternatively, the dead biomass from by-catch could fuel the growth in polychaetes and other benthos due to a direct subsidy from trawling.  Further experimental work is required to test these model-derived hypotheses.
Ecopath outputs were validated using stable isotopes and examined for system-wide impacts.  The concentrations of stable isotopes of d15N and d13C were compared to Ecopath effective trophic levels.  Trophic fractionation occurred across trophic levels, and results were comparable to published studies (for each unit effective trophic level increase there was a fractionation of +2.637‰ for d15N and +1.084‰ for d13C).  Ecopath whole-ecosystem metrics indicated that net primary productivity, trophic efficiency, ascendency, and net primary production: respiration ratios were greater in the areas open to trawling; total system throughput and Finn Cycling Index were greater in the areas closed to trawling.  Additional compartment-level comparisons were made using mixed trophic impacts (MTI) and keystoneness index (KSI).  The MTI analysis indicated that shrimp trawling in Core Sound caused large negative impacts only on jellyfish, a bycatch species. The KSI indicated that sea turtles and brown pelicans were keystone groups (large influence relative to their biomass) overall in Core Sound.  Spot, bluefish and Atlantic croaker also had high KSI in closed areas.  Pink shrimp, white shrimp and bluefish all had high KSIs in the open areas, suggesting that they played a key role in the ecosystem’s trophic structure where trawling was allowed.    
These Ecopath models can be useful tools for resource managers to better understand the direct and indirect impacts of (shrimp trawl) fishing in Core Sound.  Future work should include the creation of annualized models and simulation modeling using Ecosim to explore different management scenarios.

Network model of Core Sound, NC, USA after the shrimping season in Fall 2007, model data from the areas open to trawling. The brown shrimp (Farfantepenaeus aztecus) fishery in North Carolina, valued at $13 million in 2012, depends on a food web comprised of detritus, bacteria, phytoplankton, seagrass, meiofauna, benthic polychaetes, and zooplankton.  Shrimp are consumed by various fishes like the southern flounder (Paralichthys lethostigma). This graphic shows a multi-dimensional scaling statistical analysis of the feeding relationships observed in Core Sound, NC. The similarity of trophic roles of each species is shown by how close they plot together on the graph.  Similarity was computed in terms of the regular role equivalence, the degree to which they consume or are consumed by species with similar trophic roles. Colors show species with similar feeding roles.  The size of each node is proportional to the log 10-transformed biomass of the node in an Ecopath Network Model of the food web. Arrows show the flow of carbon from detritus to the top of the food web, represented here by the shrimp trawlers.       This graphic uses a food web concept, which is a paragdigm commonly known to all students of ecology and scientists. The food web model was created using a computer modeling software (Ecopath) of the Core Sound estuarine system based on data from our own work, the literature, and the State of North Carolina. The visualization output was created with social network analysis software (UCInet), which was used  to compute regular equivalence similarity coefficients and multi-dimensioanl scaling coordinates. The final network graph was made with Pajek. Organism and trawler images courtesy of Integration and Application Network, University of Maryland Center for Environmental Science (ian website at University of Maryland).

Project Goals

• Develop a food web network model with the fishery harvest of Core Sound as part of the model
• Sample the ecosystem and species present to estimate abundance (g C per m2) using benthic cores, plankton nets, trawl nets, gill nets, and visual observations
• Use the NC Division of Marine Fisheries Trip Ticket program catch records to include as harvests  in the model
• Compare the areas open to shrimp trawling with secondary nursery areas (closed to shrimp trawling May - Oct)
• Develop network models for Spring and Fall 2007 with Open to Trawling and Closed to Trawling network models

Location in North Carolina, USA, near  Cape Lookout National Seashore and Cedar Island, NC

Sampling Methods

Sampling of Core Sound performed with benthic cores, plankton net tows, otter trawls, barrier seines, wrap-around-net, and gill-nets. Trawling was performed using an otter trawl with a 3.2-m head-rope, 1-cm stretched mesh in the wings and 3.2-mm stretch mesh cod end. Three two-minute trawls covering approximately 75m were performed at each station.  Abundances (number of individuals/m²) and biomass (g/m²) for each species were calculated after the samples were measured and weighed.  Trawl two lengths were determined using a scientific echosounder operated simultaneously with the trawl deployment.  The BioSonics DTX echosounder was used to assess bathymetry, bottom substrate, and fish abundance in front of the trawl.   The echosounder was interfaced with a JVC GPS receiver and a Panasonic Toughbook CF-29 laptop computer so that precise trawl tracks and depths were recorded to a hard drive.  Barrier seining was performed in the methods similar to those performed by Christian and Luczkovich (1999) and Luczkovich et al. (2002).  Barrier seines were used in the fall 2006 and spring 2007. Biomass (g/m²) and number of individual/m² were calculated by assuming the area was constant (A=1/2(b*h) =1/2(7.62*7.62) =29m²).  The gill nets were 114.3-m long with five 23-m panels of varying sizes. The panels started at 8.9-cm and increased by 1.3-cm increments to 14-cm. Gill net sampling area was determined assuming an area was sampled equal to the length of each gill net squared (114.3m*114.3m =13,064.5 m²).  The strike net was 115-m long of # 10 monofilament twine and 2.4m deep, with 8.9-cm stretch mesh, with the last 23 m were trammel net. The strike net was deployed off the port side of the stern of the boat as the boat was making a circle to enclose an area.  Area sampled after each wrap-around net was determined by using the routes from a Garmin handheld 76S GPS and the Expert GPS software application.

1. Benthic cores (inside diameter of 9.5 cm) were collected by SCUBA divers and pushed manually into the substrate to a depth of 10 cm.  Corers were plugged with large stoppers at both ends and brought to the surface for processing.  A total of 12 cores were collected at each site.
   1. Three cores were combined to form one sample that will be processed for benthic macroinvertebrates.  Samples were passed through a 500 μm sieve, and all specimens preserved in 10% buffered formalin until further processing in the laboratory.  Specimens will be examined under a microscope and identified to lowest taxonomic level.  The biomass of all specimens will be measured as wet weight or dry weight and converted to C (the currency of the Ecopath model) utilizing the methods of Baird et al. (1998). 
   2. All subsequent cores were drained of overlying water and subsampled for benthic bacteria, microalgae, detritus, meiofauna and sediment grain size. 

                                                               i.      Benthic bacteria were collected using a 1-cm diameter syringe plunged to a depth of 1 cm and preserved in a scintillation vial filled with 19 ml of 2% bacteria-free formalin.  Bacteria will be enumerated and biomass estimated by epiflourescence microscopy with 4’6-diamino-2-phenylinodole (DAPI) using a method similar to Marsh (2007), considering Schallenberg et al. (1989).   All biomass measurements will be converted to grams Carbon using literature as necessary. 

                                                             ii.      Microalgae were collected using a 1-cm diameter syringe plunged to a depth of 1 cm, collected into a Whirlpak bag and stored on ice in a cooler until returned to the laboratory.  Microalgae will be estimated from chlorophyll a content measured by fluorometry, similar to the methods of Baird et al. (1998).  All biomass measurements will be converted to grams Carbon using literature as necessary. 

                                                            iii.      Sediment organic matter (for the detritus compartment) was collected using a 1-cm diameter syringe plunged to a depth of 1 cm, collected into a Whirlpak bag and stored on ice in a cooler until returned to the laboratory.  Sediment organic matter will be measured by loss on ignition using the methods of Baird et al. (1998).   All biomass measurements will be converted to grams Carbon using literature as necessary. 

                                                           iv.      Meiofauna were collected with a 2-cm diameter syringe plunged to a depth of 3 cm, and preserved in 10% buffered formalin with Rose Bengal.  Meiofauna will be separated from sediments using the method of Burgess (2001), passed through a 63 μm sieve, and all specimens will be identified to lowest taxonomic level and measured for conversion to biomass (Higgins and Thiel 1988; Giere 1993).   All biomass measurements will be converted to grams Carbon using literature as necessary. 

                                                             v.      Sediment samples were retained from the core (a “scoop” down to no more than 3 cm), collected into a Whirlpak bag and stored in a cooler until returned to the laboratory.  The samples were dried in a 60°C oven for 48 hours, ground with a mortar and pestle to remove large clumps and processed in the Geology Department Ro-Tap machine for 10 minutes.  Grain sizes were measured as the amount of a 100 g sample retained on sieves from 1.0 φ to 4.5 φ, corresponding to grain sizes of coarse sand to silt/clay, respectively. 

2. For plankton samples, 90 μm mesh bongo plankton nets (net diameter of 28 cm) were towed for 1 min at a constant speed.  Continuous GPS locations throughout the tows were recorded to avoid crossing previous tow tracks.  A General Oceanics flow meter with the low-speed rotor was attached to the bongo net to measure the volume of water towed.  Triplicate samples were collected for analysis of zooplankton, and triplicate samples were collected for ctenophores.
   1. All zooplankton samples were fixed in 10% buffered formalin for storage until processing.  Any ctenophores or other large gelatinous zooplankton were removed before fixing.  Using a Folsom splitter, samples were split three times, and the 1/8 sample was processed.  Five 10-ml subsamples taken with Hensen-Stempel pipettes will be counted in a Ward wheel, summed and total counts will be extrapolated.  This method should subsample 100-300 individuals at a time, an amount recommended by several sources and avoids potential errors associated with repetitive Folsom splitting of samples (Van Guelpen et al. 1982; Griffiths et al. 1984; Mallin 1991; Johnson and Allen 2005).  Counts will be converted to biomass using literature values, and all biomass measurements will be converted to grams Carbon using literature as necessary.   
   2. To estimate the abundance of ctenophores or other gelatinous zooplankton, separate 1 min tows were conducted.  Any ctenophores collected in the tows were counted and recorded on the boat.  Total counts will be multiplied by literature values of ctenophore biomass for use in the Ecopath model.  All biomass measurements will be converted to grams Carbon using literature as necessary. 
3. Water samples were collected to measure phytoplankton, particulate organic carbon (POC) and bacterioplankton biomass separately. 
   1. Carboys (1 L³) were filled with surface water at each station and stored on ice in a cooler until returned to the laboratory.  Water was filtered through glass microfiber filters (47 mm, GF/C) for aquatic chlorophyll a content to estimate the biomass of phytoplankton.  Water samples were also filtered through glass microfiber filters (47 mm, GF/C) to measure particulate organic carbon (POC). 
   2. Bacterioplankton were collected from 1 mL water samples that were placed in scintillation vials filled with 19 mL of 2% bacteria-free formalin.  Bacteria will be enumerated and biomass estimated by epiflourescence microscopy with 4’6-diamino-2-phenylinodole (DAPI) using a method similar to Marsh (2007), considering Schallenberg et al. (1989).  All biomass measurements will be converted to grams Carbon using literature as necessary. 
4. To sample fishes and other forms of nekton, an otter trawl similar to the one used by NC DMF was deployed.  The otter trawl has a headrope of 3.2 m, a body net stretch mesh of 1 cm, a cod-end stretch mesh of 0.5 cm, a tickler chain, and trawl doors measuring 90 cm by 46 cm.  Trawls were deployed for 2 min at a constant speed.  All specimens retained by the trawls were fixed in 10% buffered formalin for identification and measurement in the laboratory.  When necessary, some samples were weighed in the field using spring scales.  In the laboratory, all fish were identified, measured for length and wet weight, and stomachs removed for diet analyses (to be described later).  All biomass measurements will be converted to grams Carbon using literature as necessary. 
5. To capture small fishes in shallow sites, a barrier seine (in the shape of a right triangle) was utilized in Fall 2006 and Spring 2007.  For this study, opposite and adjacent sides were 7.62 m each, separated by a 90° angle; the hypotenuse was closed off by a bag seine.  The use of the barrier seine has been described in Luczkovich et al. (2002).  All fishes and non-fish collected with the barrier seine were fixed in 10% buffered formalin for identification and measurement in the laboratory.  All biomass measurements will be converted to grams Carbon using literature as necessary. 
6. An experimental gill net was used to collect larger, faster fishes not captured by the otter trawl or barrier seine gears.  Five-22.86 m panels of different stretch mesh (from 8.9 cm to 13.9 cm in 1.3 cm increments) were deployed for upwards of six hours and checked at least every two hours.  All specimens were tagged and stored on ice in a cooler until brought back to the laboratory or field processing site.  Specimens were identified, measured and stomachs will be removed for diet analyses (described below).   All biomass measurements will be converted to grams Carbon using literature as necessary. 
7. A strike net (or wrap-around net) was used to collect larger fishes.  This gear is similar to that used by North Carolina fishermen to actively target mullet (Mugil cephalus), but we deployed this gear without targeting schools of fish.  The net was deployed at each station to sample a known area, then pursed together to capture any fish within the area.  The use of this gear was restricted to Fall 2007 only, due to a late purchase date and the lack of an appropriate vessel from which to deploy the gear.  The strike net was 114.3 m long with 8.9 cm stretch mesh and including trammel netting in the last 22.86 m. 
8. A clam rake was used to collect mollusks from sites in shallow water.  Four 7.62 m transects were raked to collect any mollusks at the shallow sites.  Mollusks were stored on ice until returned to the laboratory for positive identification and measurements.  All biomass measurements (shell-free) will be converted to grams Carbon using literature as necessary. 
9. Visual surveys were conducted for groups such as birds, dolphins and turtles.  Bird surveys were conducted for 30 min at each site within an area of 500 m², similar to the methods used by Christian and Luczkovich (1999).  The biomass of any turtles spotted near the study sites were estimated by visual inspection (and literature) rather than capture.  Any dolphins observed near study sites were counted, and estimates of biomass will be made from literature values and converted to grams Carbon. 
10. Human predators, in the form of commercial fishing gears, will also be included in the models.  Unpublished data from the NC Division of Marine Fisheries for Core Sound will be utilized to incorporate humans into the models.  Harvest data aggregated by gear type, fish or invertebrate species and month will be used to estimate the proportion of “diet” for each gear.

 Feeding Relationships in the Model (Diet Matrix)

To construct a diet matrix, diet data were collected using stomach content analysis for several fishes (Hart 2008), while diet information for all other compartments as gathered from literature.  A sieve fractionation method (modified by Luczkovich and Stellwag [1993] from Carr and Adams [1973]) and used by Baird et al. (1998), Luczkovich et al. (2002) and Chagaris (2006) was used in this study. 

Ecopath Parameters

Information on ecotrophic efficiency and production:biomass and consumption:biomass ratios was collected from literature.  It was necessary to utilize existing datasets for this information (Peters 1983; Jorgensen et al. 1991; Christensen and Pauly 1993; Christian and Luczkovich 1999; Johnson 2006; Froese and Pauly 2006; Chagaris, unpublished Ecopath model of Pamlico River estuary).  

Species in the Model

Spring Model - Species observed by our sampling methods

Diatoms and algae	Seagrasses	Producers
Penaeid shrimp	Bivalves (scallops, clams, oysters)	Herbivores
Atlantic Menhaden	Bay Anchovies	Zooplanktivores
Atlantic Croaker	Pinfish	Benthic feeders
Sharks	Rays	Apex predators

Final Spring compartments which were aggregated from above and combined with NC Division of Marine Fisheries Catch data for ECOPATH modeling:

Fishery Nodes, data sources and assumptions: