An Assessment of Predation by the Lobate Ctenophore Mnemiopsis leidyi (Agassiz, 1865) on Ichthyoplankton in Narragansett Bay, Rhode Island

I investigated the importance of predation on fish eggs and larvae by the lo bate ctenophore, Mnemiopsis leidyi, in Narragansett Bay, RI, USA, by measuring the abundance and distribution of M. leidyi and its ichthyoplankton prey at five stations in the East and West Passages of the bay from May to August, 2002. During early-June, the M. leidyi population reached an abundance of 682 m-3 in the mid-bay, while fish egg densities were 3.2 m-3 and fish larvae were absent. In late June, a maximum larval fish density of 34 1 oom-3 was observed, and ctenophore abundance was <33 m-3 at all stations. These data confirm that predator and prey co-occur in the bay, but there is not a substantial amount of temporal overlap. Coincident with the ctenophore abundance measurements, I conducted in situ gut content analyses of 1,031 M leidyi during the period of highest ichthyoplankton abundance. This revealed that 6.9% of the ctenophore guts examined contained at least one fish egg and one fish larva was observed from May-August. During June, 14.6% contained at least one fish egg. These data provide evidence that fish eggs are consumed ~y ctenophores in the bay. There was no evidence for frequent predation on fish larvae. Individual feeding rates ranged from 0.04 to 0.6 fish eggs ingested per ctenophore h1• Predation rates on fish eggs were calculated from the numbers of ichthyoplankton prey found in M leidyi, temperature-specific digestion times determined in the laboratory, and the field densities of predator and prey. Accordingly, estimates of predation on fish eggs in Narragansett Bay ranged from <1 % to 330% of the standing stock of the fish eggs ingested h1 during periods of low and high ctenophore abundance, respectively. Predation on fish eggs was not detected in samples taken after June 26. An electivity analysis was performed to compare the proportions of fish eggs versus other prey in the diet of M leidyi and that found in the environment. Electivity of M leidyi was examined at the Fox Island station and was found to be positive 23% of the time with respect to fish eggs and negative 8% of the time. All positive and negative values were significantly different from 0 (p<0.0005). The date of peak abundance of ctenophores in the bay coincided with positive electivity for fish eggs. Also, the predators were found to have no selection 69% of the time with respect to fish eggs. Electivity was also examined at Dutch Island and was positive 42% of the time with respect to fish eggs and negative 16% of the time. No selection occurred 42% of the time. All positive and negative values were significantly different from 0 (p<0.01). These results support that M leidyi preys upon fish eggs in Narragansett Bay. M. leidyi, zooplankton, fish eggs, and fish larvae abundances were sampled through the summer of2002 in Narragansett Bay. These data were combined with literature values of microzooplankton abundance and physiological processes and a bioenergetics model was developed to simulate seasonal ctenophore biomass from June to July at Dutch and Fox Island. The goal of the model was to examine which prey groups were most important in supporting the observed M leidyi biomass during the initiation or rapid population growth. The magnitude, rate of biomass increase, and timing output by the model was compared with field estimates of M leidyi biomass. The model simulations clearly show that the carbon present in copepod biomass alone could account for the magnitude and high rate of increase of the ctenophores in early to mid-June. Later in the season after the copepods have been depleted, it appears that other sources of carbon become increasingly important to M leidyi.

: Results of the analysis of the period of greatest fish egg decline. The exponential model, N1=N0 e-k\ was applied to determine the percent fish egg decrease d-1 , where N0 =initial number of prey, N1 =number of prey at time, t, e = base of natural logarithm, t = time in days, and k = the fraction of the standing stock of prey removed d-1 • The loss rate of 20-40% due to flushing is based on a tidal mixing model for the lower West Passage of Narragansett Bay    Keller et al., 1999. Unpublished 2001 leidyi and ichthyoplankton data are courtesy of B. Sullivan and G. Klein-MacPhee, respectively.
Ichthyoplankton sampling frequency and site location was similar in all studies, but the sampling frequency of M. leidyi and sites sampled varied among studies.
Mesh sizes used to determine M leidyi abundances were as follows: 6 mm, 19716 mm, -72, 1974153 µm, 1973153 µm, and 1974153 µm, -76 (Hulsizer 1976 and ; 1.8mm (summer and fall) and 153 µm (winter), 1975 (Deason 1982);1mm,19831mm, , 1985; 505 µm, 1990 (Keller et al. 1999); 500 and lOOOµm . Mesh sizes used to determine ichthyoplankton abundances were as follows: 505µm, 1972, 1990(Bourne and Govoni 1988, (Keller et al. 1999)., and (Klein-MacPhee, (closed diamonds) and at Fox Island in the middle of the West Passage measured once each week (1992, (open circles) (Hawk 1998)  Predation on the early life history stages of marine fish is an important contributor to their overall mortality (Bailey and Houde 1989). Marine scientists noticed the importance of ichthyoplankton predation by invertebrates early in the 20th century (Mayer 1917;Joubin 1924;Bigelow 1926: all cited in Alvarifio, 1985. Recent research has continued to focus on mortality due to predation because starvation appears to account for only a minor fraction of ichthyoplankton mortality (Bailey and Houde 1989). One study estimated that predation losses can range up to 95% over the duration of the egg stage (Hunter 1976).
Unfortunately, in spite of this long history, progress in this field has been slow due to the experimental difficulties associated with in situ predation studies (Bailey and Houde 1989;Purcell 1985). Previous studies have examined predation by gelatinous zooplankton on fish eggs and larvae using a variety of methods such as laboratory observations, quantification of digestion rates, gut content analyses, and modeling of predator-prey dynamics.
Among the known gelatinous predators of ichthyoplankton, cnidarians and ctenophores have substantial predatory potential because of high population densities and ingestion rates. Numerous studies have shown that pelagic cnidarians and some ctenophores prey on fish eggs and larvae (Burrell and Van Engel 1976;Purcell 1985;Purcell 1989;Arai and Hay 1982;Fancett and Jenkins 1988;Bailey and Houde 1989;Purcell 1994). Most studies that have reported predation by ctenophores on ichthyoplankton targeted the total diet of the predator and the majority found that fish eggs comprised only a small portion of the diet. For example, Burrell and Van Engel (1976) observed that, out of 3,300 M leidyi guts examined, only 1 % had ingested a fish egg and 0.4% had ingested a fish larva. Similarly, investigations that examined the role of larval fish in the diet of ctenophores found that they made up a relatively small portion of the total. Vanderveer (1985) suggested that Pleurobrachia pileus was a significant predator of flounder larvae, despite finding only 9 larvae in 15,000 ctenophores guts. Purcell (1989) found 3 fish larvae out of3,566 Pleurobrachia bachei guts examined in Kulleet Bay, Vancouver Island, British Columbia. These studies, however, examined tentaculate ctenophores, which may not be as effective at capturing fish larvae as their lobate couterparts.
The lobate ctenophore, Mnemiopsis leidyi, may be an important predator in coastal systems given its high population densities and ingestion rates, and multiple feeding strategies such as lobe or auricular capture. Mnemiopsis leidyi consumes a wide range of prey, including: crustacean larvae and copepods , Annelid larvae (Burrell and Van Engel 1976), and fish eggs and larvae (Cowan and Houde 1993). The mechanisms used by M leidyi to capture prey include the creation of a low-velocity current with the auricles and use of the oral lobes . The ctenophore's ability to use both lobes and auricles in concert for prey capture makes it an effective predator . A laboratory study by Cowan and Houde (1992) indicated that M leidyi preyed on Chesapeake Bay goby larvae, Gobiosoma bosci, from 2.7-9.4 mm SL, and that the ctenophore's slow swimming speed resulted in lessened escape responses by larger fish larvae. However, there is little supporting evidence from field studies to confirm this laboratory finding.
A study of M. leidyi in the Chesapeake Bay (Purcell et al. 1994) provides in situ clearance rates offish eggs of 128 ± 58 l d-1 predatof 1 and predation estimates of 0-38% of the prey consumed d-1 • Large ctenophores were collected for gut content analysis and data for 75 individuals were reported. Of those, 51 contained no fish eggs, 24 contained one or more fish eggs, and none contained fish larvae. However, this study focused on one fish species, Anchoa mitchilli, and was limited by a small sample size and use of only large ctenophores for gut content analyses.
Because of container effects, studies which determine the predatory potential of ctenophores from laboratory feeding experiments should be considered with caution.
Three laboratory feeding experiments and one mesocosm study have been conducted that used both 40-1 and 20 to 25-1 vessels , 5-1 vessels (Tsikhon-Lukanina et al. 1994), 15 liter containers (Monteleone and Duguay 1988) and 3.0 m 3 enclosures (Cowan and Houde 1993), respectively. An experiment to determine the effect of container volume on the feeding rates of M leidy i demonstrated that small containers ( <50 l) significantly reduce ctenophore feeding behavior (Monteleone and Duguay 1988). Unfortunately, the one large enclosure study only provided one experiment in which alternate prey were available and predation on only one species, Anchoa mitchilli, was investigated (Cowan and Houde 1993). Thus, more field estimates of predation rates are needed to compare with existing estimates of predation from laboratory studies.
Mnemiopsis sp. was first reported in large "rafts" in northern coastal waters in 1881(Fewkes1881). However, surprisingly little was known about the ecology of ctenophores because they are difficult to sample using conventional methods such as plankton nets, and they do not preserve well. Large-scale quantitative studies involving M feidyi in New England waters did not begin until the 1970s . Since the 1970s, M leidy i has been reported in large concentrations in Narragansett Bay, Rhode Island , where the ctenophore population is typically larger than in warmer southern waters (Kremer 1994). The timing and maintenance of such immense growth events ("blooms") has substantial trophic effects on zooplankton and, potentially, ichthyoplankton populations, as well as indirect impacts on phytoplankton. For example,  observed diminished copepod abundance in Narragansett Bay following an increase in ctenophore abundance. The relaxation of grazing pressure by copepods then allowed a summer phytoplankton bloom in the bay.
In Narragansett Bay, a temporal shift in peak abundance of M. leidyi has been documented concurrent with increasing water temperatures .
Potential ecological consequences of this shift include a spatial-temporal overlap of M leidyi and fish eggs and larvae during warm years, decreased survival of larval fish due to competition with M. leidyi for their zooplankton food source, and a decrease in overall ichthyoplankton abundance due to the top-down control exerted by the ctenophores. I speculate that the latter is potentially supported by a documented 2-4 fold decrease in ichthyoplankton abundance since the 1970's (Keller et al. 1999) which 18 coincides with a significant increase in M. leidy i abundance from the same time period   (Fig. 1-1 1970 1975 1980 1985 1990 1995 2000 2005 Year .,,  Keller et al., 1999. Unpublished 2001 leidyi and ichthyoplankton data are courtesy of B. Sullivan and G. Klein-MacPhee, respectively. Ichthyoplankton sampling frequency and site location was similar in all studies, but the sampling frequency of M leidyi and sites sampled varied among studies. Mesh sizes used to determine M. leidyi abundances were as follows: 6 mm, 19716 mm, -72, 19746 mm, (Kremer 1975153 µm, 1973153 µm, and 1974153 µm, -76 (Hulsizer 1976 and ; 1.8mm (summer and fall) and 153 µm (winter), 1975 (Deason 1982);1mm,19831mm, , 1985; 505 µm, 1990 (Keller et al. 1999); 500 and 1 OOOµm . Mesh sizes used to determine ichthyoplankton abundances were as follows: 505µm, 1972, 1990(Bourne and Govoni 1988, (Keller et al. 1999), and (Klein-MacPhee, unpublished), 333µm (this study).

Field Sampling
Narragansett Bay is a temperate, relatively well-mixed estuary on the northeast coast of the United States (Hicks, 1959

Fish egg and larvae densities
Ichthyoplankton density was determined from three replicate oblique tows, each at 3.7 km h-1 (Herman 1958) and lasting 2 minutes (Keller et al. 1999). The samples were collected using a 333-µm mesh net with a 0.5-m diameter opening equipped with a flowmeter. The average amount of water filtered per tow was 50 m 3 . This mesh size is a mid-sized mesh and was chosen based on a range of sizes, 280 to 505µm, used in previous ichthyoplankton surveys (Keller et al 1999) as well as its suitableness for capturing both fish eggs and larvae in Narragansett Bay (Klein-MacPhee, pers. comm.).
One sample was counted live and the ichthyoplankton removed and used in laboratory digestion rate experiments, while the remaining two were preserved in 37% buffered formalin.

Zooplankton densities
Other constituents of the zooplankton (such as copepods, crab zoea, and veliger larvae) were sampled by vertical tows taken with a 64-µn mesh net with a 0.25-m diameter opening also equipped with a flowmeter. The chosen mesh size is appropriate for both larval and adult stages of members of the zooplankton in Narragansett Bay (Durbin and Durbin 1978). These samples were preserved in 37% buffered formalin and the contents enumerated. Data from these tows was used for calculation of selectivity of prey and was provided by Sullivan and Van Keuren (unpublished).

Gut Content Analyses
I also collected ctenophores at each station using a long-handled bucket and a plankton net. I immediately examined their gut contents under a dissecting microscope.
Shipboard microscopy is the most direct method to determine the actual diet of the predators. This method avoids the artifacts of over-handling, especially the ejection of gut contents, which have plagued laboratory studies. Fish eggs and larvae were identified to species level and enumerated in the ctenophore gut contents. As many ctenophores as possible were examined at each station with the target number being a minimum of30 organisms from each of two size classes (<1 cm, >1 cm). This target number was selected based on desired confidence intervals for statistical significance.
This was not always possible due to time constraints or lack of organisms.

Digestion Times
It is important to accurately determine the length of time a fish egg or larva can be identified in the predator's gut (D), so that those values can be used in combination with frequency of prey per predator (G) to calculate predation rates from the equation I = GID, when I = ingestion predator per hour. Laboratory studies were performed to determine the digestion times of M leidyi of multiple size classes. Freshly collected ctenophores were placed in 8-1 containers with 20-µip mesh filtered seawater in an environmental chamber and held overnight to clear their guts. The temperature was set within 1-2°C of the temperature of ambient seawater in the bay at the time ctenophores were collected. Prior to measurements, the ctenophores were maintained in the environmental chamber at the same temperature and light/dark cycle as they would encounter naturally. Two temperature treatments, a low range of 7.5-13°C and a high range of21.5-24°C, were chosen based on the observed temperature range in Narragansett Bay during 2002. Approximately the same temperature ranges were used m previous M. leidyi predation experiments on zooplankton of Narragansett Bay (Kremer 1979). Eight-liter containers were used in the digestion experiments, because container effects were not important as these studies were concerned with ctenophore digestion time and not feeding behavior. The prey consisted of net-collected zooplankton and net-collected and cultured ichthyoplankton. Fish egg sizes used in the determination of ctenophore digestion rates ranged from 0. 74-1.15 mm in diameter and fish larvae used were 2.78-3.0 mm in total length ( innovation which served three purposes: first, it allowed the eggs to be easily observed in the transparent ctenophore gut; second, it decreased the amount of handling of the ctenophore during the experiment; and third, it allowed for more frequent observations of gut contents. Live prey items were added to the experimental container and the ctenophore was allowed to feed until 1-2 prey were detected in the gut. The ctenophore was then removed from the prey container, placed into another 8-liter container with filtered seawater and no prey items. I examined the ctenophore at 2-3 minute intervals to determine a functional digestion time (the time when an ingested prey item could no longer be positively identified) and an actual digestion time (the time when the prey was 25 completely digested). An interval of 2-3 minutes allowed for observation of several ctenophores at the same time.
26 fish larvae being digested in as few as 15 minutes, so the 2-3 minute observation intervals resulted in more tightly constrained digestion rates than in previous studies (Monteleone and Duguay 1988).

Ingestion Rates and Predatory Impact
For each sampling time, ingestion rates of the ctenophores were calculated from the average number of eggs or larvae per ctenophore from 30 or more organisms in each size class in field collections and digestion time measured in the laboratory. The ingestion model used was: I = GID , where I = number of eggs ingested per ctenophore per hour, G =number of fish eggs or larvae per ctenophore (from field collections), D = egg or larvae actual digestion time (h) (from laboratory study). The actual digestion time was used instead of the functional time because actual digestion times were available for all experiments. This technique of using the model I= GID, described by Purcell ( 1997), reduced laboratory artifacts and revealed the actual diet of the ctenophore by relying on gut content data from field. collected ctenophores. However, this equation assumes that there is steady-state feeding by the predators and that food identified from gut content analysis (G) is the same food that will be used in the measurements of digestion times (D). In addition, this method assumes that the animals collected for gut contents are representative of the population at that location. A % clearance rate was calculated from the ingestion rate and densities of prey in the water column (I/egg or larvae density). These ingestion rates were multiplied by the number of predators to estimate the percent of the prey population that could potentially be consumed. The calculation of% clearance rate relies upon "I" or the calculated ingestion rate and is therefore dependent on the gut content approach.

Statistical Analysis
Descriptive statistical analyses such as mean and standard deviation of M. leidyi and ichthyoplankton abundances were performed using the Microsoft® Excel 2000 software package. I used the Sigmastat® statistical software package to perform one way ANOV As to determine the variability of both ctenophore and fish egg and larval abundance by station and date. Because there was a significant difference among stations for ctenophore abundance, I used Sigmastat® to perform an All Pairwise Multiple Comparison Test or Tukey Test. In order to examine if digestion times of fish eggs and larvae were affected by independent variables such as ctenophore size, date, prey size, and temperature, a multiple linear regression analysis was performed using Sigmastat® software. Pearre's (1982) electivity index, C, was used to assess M. leidyi's prey-selectivity in situ at Fox Island and Dutch Island. Electivity analysis could not be performed at the other stations, because total gut content data was not collected, only ichthyoplankton data.

Seasonal Abundance of Predator and Prey
Mnemiopsis leidyi was the most abundant gelatinous predator collected in Narragansett Bay during the study period. The M leidyi population in Narragansett Bay was characterized by a rapid increase in abundance which spanned several orders 29 of magnitude and then a steep decline in late summer through early fall ( Fig. 1-3 Total predator densities were <100 m-3 until mid-June, when a dramatic increase in total ctenophore abundance was observed at both stations ( Fig. 1-3 (French 1991). But the maximum egg population appears to have occurred earlier at Fox Island and Dutch Island than the time of increase ofMnemiopsis ( Fig.1-4a). Of the 14 identified species of fish eggs and larvae in the ichthyoplankton samples, the species composition was dominated by butterfish, cunner, tautog, and searobin eggs ( Fig. 1-5A). The same four species were the dominant eggs found in the gut contents of M. leidy i during the sampling period (Fig 1-5B).
The highest total mean fish egg density (of three tows), 73.2 ± 82.2 m-3 , was recorded on May 29 at Fox Island when mean ctenophore abundance (of duplicate tows) was low at 2.9 m-3 ( Fig. 1-4A). Fish larvae reached a peak average abundance (of three tows) of 34.1±59.8 lOOm-3 larvae on June 18 at Prudence Island and were inversely related to ctenophore abundance at 3 out of 5 stations. Prudence Island and Warren River did not exhibit this pattern and ctenophore abundance and fish larvae increased simultaneously (Appendix B). Abundance of fish eggs and larvae was not significantly different among stations during the survey (one-way ANOV A, p=0.318).

Sea robin 9%
Butterfish 39% Further analysis of the abundance patterns of fish eggs and M. leidyi at each station was performed during the period characterized by the greatest fish egg decline (

Gut Content Analyses
Jn situ gut content analyses of M. leidyi during the period of highest ichthyoplankton abundance (May 22-June 7) revealed that 19 .6% of the ctenophores examined had consumed at least one fish egg with a range from 0-71 % over the sampling season. Of 1,031 ctenophore guts examined from May to August, 6.9% contained at least one fish egg, with one having 5 in the gut. This study shows that a broad size range of ctenophores are capable of ingesting fish eggs. The largest ctenophore to have ingested a fish egg was 10.l cm total length, the smallest was 0.5 cm, and the mean ctenophore size that ingested a fish egg was 5.2 cm. Only 2 ctenophores < 1 cm consumed a fish egg and the median size of M. leidyi that ingested an egg was 5.4 cm and the mean size was 5.2 cm total length. The size distribution of the % of ctenophores that consumed fish eggs revealed that the percentage of M leidyi that ingested fish eggs was roughly 5 times higher when the predators were larger than 0-.9 1-1.9 2-2.9 3-3.9 4-4.9 5-5.9 6-6.9 7-7.9 8-8.9 9-9.9 10-10.9 M. leidyi total length, cm

Predation rates on fish eggs
Predation rates by M leidyi on fish eggs were calculated from the numbers of ichthyoplankton prey found in the predator's gut contents, the temperature-specific digestion times, and the field densities of predator and prey. Accordingly, estimates of predation on fish eggs in Narragansett Bay ranged fr~m <l to 111 eggs consumed m-3 d 1 • These values are comparable to previous estimates that range from 0 to 14. 7 and 21 to 174 (Monteleone and Duguay, 1988) and 10 to 79 eggs m-3 d-1 (Purcell, 1994). The percent clearance ranged from less than 1 % to over 300% h-1 during periods oflow and high ctenophore abundance, respectively (

Temperature and Salinity
The year 2002 was unusually warm and had the second highest mean annual water temperature (12.5°C) since 1956 (Hawk 1998) (Fig. 1-7). The seasonal temperature pattern in Narragansett Bay is characterized by a spring increase, a peak in early fall, and an autumnal decline ( Fig. 1-8).
The Warren River station and in the West Passage at Fox Island. The highest salinity, 31.3 psu, occurred at Dutch Island on July 1, 2002. Ctenophores were found throughout the entire salinity range.

DISCUSSION
Predation on the early life history stages of marine fish is potentially the single most important source of mortality. A likely result of predation is the regulation of fish egg and larvae abundances, which may, in tum, affect recruitment (Bailey and Houde, 1989). To affect fish recruitment, ctenophore predation must remove a significant number of fish eggs, which prevents maturation into larvae and the escape of ctenophore predation. M. leidyi has been shown to prey successfully on fish eggs and larvae in both laboratory and field studies. In the field, timing of fish spawning and explosive ctenophore population growth dictate the extent to which M leidyi can decrease the ichthyoplankton in Narragansett Bay. If most of the fish spawn and the eggs mature into larvae before ctenophores become abundant, then the fish are not subjected to intense predation because of their timing relative to the ctenophore abundance. Therefore, perhaps the most important factor to consider when examining predation of fish eggs and larvae by ctenophores is the degree of spatial-temporal overlap of predator and prey (Frank and Leggett 1982).
The degree of temporal overlap was considered by examining the predator-prey abundance patterns that show that the peak of M leidy i occurred at the end of the period of highest fish egg density. In other words, ctenophore abundance was relatively low during the period of the greatest decrease of fish eggs at each station. Although, gut content analyses revealed that M. leidy i consumed fish eggs at all stations, eggs were only present in 6.9% of the ctenophore guts examined. This data supports the conclusion that M. leidy i preys upon fish eggs. However, the low percentage of ctenophores that ingested fish eggs combined with the lack of temporal overlap between predator and prey means that M leidyi did not substantially diminish fish eggs at the stations sampled. Other fish egg predators such as fish and crustacean larvae, fish egg maturation, and flushing from the bay might explain the decline observed in the bay during early-mid June.
Another important result of the gut content analyses was that very small M .
/eidyi (0.5 cm) can consume fish eggs, but do so very infrequently. The former differs from the findings of a previous laboratory study that stated "larval tentaculate ctenophores ( <0.9 cm) did not consume bay anchovy eggs" (Monteleone and Duguay, 1988). Ctenophores larger than 4 cm had higher feeding rates, particularly in the first two weeks of June, but all sizes did sometimes consume fish eggs contrary to the findings of previous studies (Appendix C). Specifically, gut content data from the present study show that 2 ctenophores < 1 cm consumed a fish egg. The mean size of M leidyi that ingested an egg was 5.2 cm. Of the 1,031 ctenophores examined for gut contents, 36% were <l cm total length, so there was not an equal representation of smaller ctenophores.
Gut content data also suggest that the decline in fish larvae does not appear to be due to ctenophore predation. Only one fish larva was found out of 1,031 ctenophore guts examined. The lack of predation on fish larvae agrees with findings of Purcell et al (1994) as they failed to find any fish larvae in the gut contents of M. leidyi (n=75) in Chesapeake Bay. Plausible explanations for this include: rapid digestion times of M .
/eidyi for fish larvae found in this study 0.4 ± 0.05 hrs. (24°C), low in situ densities of fish larvae also observed in the present study (<35 lOOm-3), and/or low swimming speeds of the ctenophores which may have resulted in fewer encounters with fish larvae (Cowan and Houde 1992). Also, if the escape response of a fish larva can be assumed to approximate that of an adult Acartia tonsa, then the larva may be strong enough to escape an intial contact with the ctenophore . Given the numerous factors that may prevent researchers from finding larvae in field-collected ctenophores, it seems that alternate approaches warrant consideration. I propose that a laboratory experiment to compare M leidyi's ability to capture fish larvae with its capture of other prey types (e.g. copepods) would be useful in addressing this issue.
In recent literature, the application of immunological techniques has been suggested as a method with which to identify highly-digested fish larvae in ctenophore guts (Purcell 1985). A preliminary assessment of the Ouchterlony Immunoassay technique was conducted in conjunction with the present study (n=50) and it was unable to detect fish eggs or larvae in ctenophores that had ingested each prey type (Feller et al. 1979). This was likely due to extreme dilution caus~d by the high water content of M.
leidyi (96%) . Therefore, this approach does not appear to be sufficient to further our understanding of M leidyi predation on ichthyoplankton.
Ctenophore predation was quantitatively estimated by calculating individual ingestion rates using the ingestion model, I = GID. I calculated a range of ingestion rates from 0.04 to 0.6 eggs h-1 • Based on the small degree of temporal overlap of M leidyi and fish eggs, I conclude that results obtained using the ingestion model should not be extrapolated over a 24-hour period as the steady state feeding assumption is most certainly violated in this case.
I also calculated individual clearance rates and percent clearance rates, which were highly variable. This is because they respond to changes in both predator and prey densities. For example, the range of individual clearance rates for M leidyi > lcm total length was 10.6-578.1 liters d-1 g wet weighf 1. A laboratory study by  examined M . leidyi clearance rates on copepods and reported a range of 0.61-2.03 liters g wet weighf 1 for ctenophores larger than 8g wet weight. Individual clearance rates from this study differ greatly from those Kremer observed and this can likely be attributed to the patchiness of fish eggs in Narragansett Bay. At Dutch Island and Fox Island, the % clearance rates ranged from 70-330.7% of the eggs cleared h-1 , whereas values in the upper-bay were 0.5-13 .3% of the eggs cleared h-1 • A recent study in Narragansett Bay estimated percent clearance of M. leidy i on fish eggs between 4.6-62.5% h-1 (Sullivan unpublished). The peak abundance of ctenophores in the study by  was 350 m-3 . In the present study, M leidyi obtained a maximum abundance of 846 m-3 at Fox Island, which is 2.5 times higher than the peak abundance recorded in Sullivan's survey. Tow-derived estimates of predator abundance were characterized by small amount of variability (Appendix D), which is incorporated into the % clearance estimates. As such, the extremely high % clearance values in the lower-bay are attributed to very high densities of predators and relatively low densities of prey. Abundance data from the lower bay stations show that the period when fish eggs are abundant does not coincide with the period of peak ctenophore abundance ( Figure 1-4). As a result, my predation estimates seem too high to be reasonable in Narragansett Bay given the small amount of temporal overlap between predator and prey.
Furthermore, the percent of ctenophores with fish eggs present in their guts was greater in the upper bay despite higher % clearance rates from lower bay stations (Table   1-3). The discrepancy between observed ctenophore ingestion and the predation impact estimated by the ingestion model using the predator and prey densities is evidence that the latter is not appropriate in this type of application. As a result, a mechanistic numerical model of M. leidyi biomass was developed as an alternate approach for examining ctenophore predation on ichthyoplankton in the bay (manuscript #2).
Numerous factors determine the extent to which M leidyi can impact ichthyoplankton stocks in coastal ecosystems. Some factors that were not addressed in the present study include diel periodicity in ichthyoplankton as Govoni and Olney (1991) observed peak densities of fish eggs in the Chesapeake Bay from dusk to dawn and vulnerability of fish larvae to predation (Paradis et al. 1996). So, these estimates may be conservative or underestimates of both fish egg density and ingestion by ctenophores. Also, an on-going study in Narragansett Bay is addressing diel differences in M. leidyi predation. The findings ofthis study may provide important information concerning M. leidyi predation on ichthyoplankton. It is likely that interannual variation of predation is considerable.
This study has shown that M leidy i is a predator of fish eggs in Narragansett Bay, but did not considerably reduce their standing stocks during this year. M leidyi predation on fish larvae is rare. At the outset of marine ecosystem modeling,  described "two opposite approaches" used to test hypotheses in the field of oceanography. The first method was "descriptive" and relied on measurements analyzed with statistics, while the other was an assumption-based, synthetic approach that sought the "mathematical derivation of relationships" . In manuscript 1, I used a traditional descriptive approach to examine the possible predatory impact of the secondary consumer, Mnemiopsis leidyi, on ichthyoplankton by calculating their ingestion and % clearance rates based on gut content analyses coupled with laboratory-determined digestion times and predator and prey densities. The conclusion was that M leidy i is a predator of fish eggs in Narragansett Bay, but does not considerably reduce their standing stocks and that predation on fish larvae is rare.
The present study is representative of the alternative approach, whereby a carbon budget model was developed to examine the ~mount of prey carbon necessary to support the observed changes in ctenophore biomass in Narragansett Bay. Recently, it has been suggested that climatic warming is increasing water temperatures in Narragansett Bay and that this allows M leidyi to experience explosive population growth during the spawning season of local fish species . This suggests that M. leidyi may have a better opportunity to reduce ichthyoplankton stocks in the bay ifthere is sufficient spatial-temporal overlap between predator and prey and M leidyi eat a significant number of fish eggs. In other words, the carbon in fish eggs and larvae may be contributing to the rapid population increase of M. leidyi in June.
One condition that allows ctenophores to increase their populations rapidly is suitable food (in terms of quantity and quality). Understanding the nutritional value of the prey present in the environment and how efficiently the predator uses the procured nutrients can give researchers insight into the causes of massive population growth events.
Carbon comprises 1. 7% of the ctenophore' s dry weight. This is equivalent to 4% of the organism's wet weight, which is a large amount given the high water content of the ctenophore . Based on the body composition of the organisms, carbon is the most important constituent and plays a key role in overall protein metabolism, the major energy source for ctenophores . Therefore, a carbon budget model can provide important information about the increase in ctenophore biomass observed in the summer in Narragansett Bay.
Numerical models of gelatinous zooplankton have been developed for a variety of estuarine systems (i.e. the Black Sea, the Chesapeake Bay, and Narragansett Bay) to examine population dynamics and the factors that control seasonality of the organisms.
For example, a population model of the carnivorous ctenophore, M. leidyi in Narragansett Bay showed that food availability was the main factor controlling the predator's abundance rather than predation, as had been previously suggested . Also, the model indicated that predation brought about the decrease of the biomass of ctenophores in the fall, but it did not limit the maximum seasonal abundance.
Over the last decade the Black Sea ecosystem received a great deal of attention from biologists as the invasive ctenophore, M leidyi, reportedly destroyed stocks of commercially important species such as anchovy and the Mediterranean horse mackerel (Mutlu 1994;. More recently, models of the Black Sea ecosystem suggest that gelatinous zooplankton may play a much smaller role in the documented fish stock decline than previously reported and that over-fishing may have led to the fisheries crash . In this case modeling was used to successfully address tropho-dynamics within a complex ecosystem

Field Sampling
Narragansett Bay is a temperate, relatively well-mixed estuary on the northeast coast of the United States (Hicks, 1959 comm.).

Carbon Budget Model of M leidyi Population Biomass
A dynamic numerical model was developed to simulate the seasonal biomass of M /eidyi in Narragansett Bay (Figure 2-1) and I compared the model results with those of a previous study, which suggest that fish eggs are a small component of the ctenophore's diet and fish larvae are rarely consumed (manuscript 1). Field-generated model inputs include: temperature and biomass data for the following forced compartments: M. leidyi, mesozooplankton (Acartia sp. was used as a proxy for this category), fish eggs, and fish larvae. Microzooplankton biomass was based on abundance data from the literature (Verity 1984). Carbon composition of the mesozooplankton and microzooplankton was estimated using values obtained from the literature (Durbin et al. 1992;Verity 1984) while values for fish eggs and larvae were calculated based on the size of the prey item and the assumption that 50% of the dry weight is carbon . Only seasonal biomass of the M leidyi population was computed mechanistically using the software package Stella®, Version 6. The in situ amount of ctenophore carbon at Dutch and Fox Island was examined using both the mean size of ctenophore present and the size frequency data of the population (Figure 2-2).

AE
Biomass was simulated at Dutch and Fox Island, because the time series were extensive at these stations and mesozooplankton data were also available.

Growth Terms
Ingestion rates were calculated according to the following equation: I (mg C m-3 d-1 ) = (prey biomass)*(l-exp(-Grazing*M. leidyi biomass*dt))   (2002) B. Sullivan, unpublished (2002) P. Verity (1984) Author's data (2002) Author's data (2002) D. Gifford, unpublished (2002) P.  C. Oviatt and P. Kremer (1977) J.  where prey biomass is equal to one or more of the following prey types: mesozooplankton, microzooplankton, fish eggs, and fish larvae (mg C m-3). An assimilation efficiency (AE) of 75% was input into the model and is based on findings for the congener of M leidyi, M. mccradyi . In the schematic diagram of the model, A= Ingestion* AE and growth= A *Ctenobiomass. Ctenophore grazing (G) was computed as in : Kremer' s formulation is based on prey removal experiments conducted in 20-25 liter tanks over forty-eight hours. Kremer's feeding rate model is used despite the results of a study by Monteleone and Duguay (1988), which show that small containers (<501) significantly reduce ctenophore feeding behavior. The container size used in Kremer's study ranged from 20-40 liters and was deemed sufficient given that studies in which larger containers are used may encounter problems with patchiness of prey during the experiments.
Alpha ( where 13 .5°C is the mean and the amplitude is 10°C. ( As a first order approximation, an assumed organism size of 5 grams wet weight was input into the model. This weight represents a mid-range value as determined in  M leidyi population model.

Loss Terms
The loss terms in the model are predation by butterfish, flushing from the bay, respiration and excretion, and egg production. A constant loss due to predation by the butterfish, Peprilus triacanthus, of 10% of the ctenophore biomass per day was used in the initial runs of the model (Oviatt and Kremer 1977). A range of20-40% was determined appropriate for the amount of hydrodynamic exchange between the lower West Passage of Narragansett Bay and Rhode Island Sound in a model developed by . As such, a flushing rate of20% of the ctenophore biomass per day was used in the initial model simulation. Respiration was determined to be a function of temperature only and was input into the model according to Kremer

Sensitivity Analysis
The model's sensitivity to variations in various parameters was examined. M.
/eidyi weight was varied to determine how important the value was in the model runs.
In the sensitivity analyses, a logical array of values was tested based on approximate ranges of each from Narragansett Bay. For example, ctenophore weights of 1 and 10 grams wet weight were tested and temperature values of 11.5, 15.5, and 20.5 °C. Other parameters were tested with a sensitivity analysis similar to that of  where the model was run with Yz and 2 times the initial values from the standard run.

RESULTS
Simulations for Dutch Island were initiated on June 1 with an initial M leidy i biomass of 25 mg C m-3 whereas for Fox Island, the in~tial biomass was 7 mg C m-3 . M. leidyi biomass was also simulated when all prey categories were available, but the change was negligible in the beginning of the season and increased throughout the remainder of the season (Figure 2-4).
Sensitivity analyses were run to test the sensitivity of the model to initial conditions and changes in model parameters. In the sensitivity analysis of temperature and M. /eidyi weight, a logical array of values was tested based on approximate ranges of each from Narragansett Bay. Varying the temperature by ±2 °C resulted in a relatively small change in the estimated ctenophore biomass. Increasing the temperature from an initial 13.5°C to 20.5°C resulted in a difference of 46.5% at Dutch Island (Table 2-2). It appears that the model is fairly robust to temperature changes within the range that is typical of the summer months in Narragansett Bay (Figure 2-5B). Also, when the weight of M. leidyi was decreased to 1 g at Fox Island and Dutch Island, the difference was 28.6% at each station. These results indicate that the model is fairly insensitive to changes in the weight of the ctenophores that make up the population.
The model appears to be very sensitive to mesozooplankton biomass, predation by butterfish, and choice of flushing rate (Figure 2-5A). The mesozooplankton results are not surprising as Kremer ( 197 5) observed that "the exact choice of food concentration proved to be the most critical parameter of all." This was attributed to forcing the compartment with no daily feedback from predation by the ctenophores to the mesozooplankton biomass .
The sensitivity of the model to predation is attributed to the fact that it was

7S
The flushing rate was set at a mid-range value, 20%, which may not be appropriate for the summer months in Narragansett Bay. However, in this case, the mid-range value was deemed sufficient as a first order approximation was the goal of the model.

DISCUSSION
A numerical model of M leidyi seasonal biomass was developed and has been applied to two mid-bay stations, Dutch and Fox Island, in Narragansett Bay. In preliminary runs of the model, an average ctenophore biomass was input into the model and the model output agreed with the field data. However, when I compared the average ctenophore biomass with ctenophore biomass based on size frequency of the in situ population, I found that the biomass based on the size frequency data was different from the average (Figure 2-2). As a result, all runs of the model presented herein use the size frequency ctenophore biomass and do not agree well with the field data. Future efforts will focus on re-parameterizing the model to the size frequency based ctenophore biomass.
I also compared the model predictions with field data to examine the extent to which mesozooplankton, fish eggs, fish larvae, and microzooplankton could support the observed ctenophore biomass. Plankton studies show that initiation of the summer increase of M leidyi coincides with a rapid decline in the copepod population in Narragansett Bay . Based on these results, the model was run with mesozooplankton as the lone prey source and these runs show that copepods would be an adequate carbon supply for the field biomass estimates in magnitude and the rate of biomass increase. However, the timing of the peak ctenophore biomass was 76 not as well captured at Fox Island or Dutch Island. The model was re-run at both stations with mesozooplankton, microzooplankton, fish eggs, and fish larvae as carbon sources and the results did not change in the beginning of the season. I conclude that the initiation of rapid ctenophore population growth in the bay is primarily supported by copepods and that fish eggs, fish larvae, and microzooplankton are not an important source of carbon during this time. Later in the season when the copepods are diminished, however, there seems to be an increase in ctenophore biomass due to the addition of these alternate prey sources.
Sensitivity analysis of the model showed that mesozooplankton biomass,