INTRODUCTION

Basin-scale changes in North Atlantic ecosystems have been observed but the mechanisms and pathways between physical forcing and specific ecosystem responses have not been elucidated. Ecosystem shifts have been identified by proxies such as temperature, which can have direct and indirect effects on various trophic levels. For Arcto-Norwegian cod, there is a strong relationship between early larval survival to recruitment and temperature through effects on feeding, metabolism, and growth, or as a proxy for other climate parameters, such as the advection of zooplankton-rich Atlantic waters from the Norwegian Sea onto the adjacent shelves. For Georges Bank cod and haddock, recruitment can be more complex, but recent studies have indicated that growth and survival of larvae was related to low surface salinity, a proxy for the influx of fresh Scotian Shelf water, ultimately leading to an increase in copepod populations as prey for larvae.

The objective of the proposed study is to develop an understanding of the processes controlling recruitment of cod and haddock on Georges Bank and cod in the Norwegian Sea sufficient to parameterize useful recruitment models and to forecast likely changes in abundance under a range of climate change scenarios. We will use the detailed information and understanding gained from modeling, broad-scale and process studies completed for a limited number of years in both systems to develop recruitment models and a suite of proxies that can be used to recreate the patterns and trends of longer recruitment time series for cod and haddock.

Implicit in our approach is the hypothesis that recruitment of cod and haddock is determined by variability in survival during the egg and larval stages that is constrained by density- and habitat-dependent factors operating during the early juvenile period. Variability in egg and larval mortality is independent of density and results from a complex interaction of biological and physical factors operating on a range of temporal and spatial scales. While their importance may vary among systems, critical factors include the biomass, condition and age structure of the spawning stock, transport and retention of egg and larval, feeding conditions, growth and predation rates. The physical environment sets the stage for these processes and plays a large part in determining the outcome.

As part of the synthesis phase of GLOBEC we propose to establish comparative biophysical coupled model studies for transport and growth of larval and early juvenile fish in the two marine ecosystems Georges Bank and the Norwegian shelf/Barents Sea (the northern and southern extremes of the distribution of Atlantic cod; see Figs. 1, 2a,b). The study will be carried out in collaboration with colleagues at the Institute of Marine Research (Norway; letter attached). These comparative studies will contribute to basic understanding of the interactions between fish populations and zooplankton and how these interactions are influenced by climate variability and change. The joint efforts will include:

• Modeled basin-scale circulation fields with increased resolution within the regional domains of the two ecosystems
• Lagrangian (particle tracking) models for application within the regional domains
• Individual-based trophodynamic models for larval and early juvenile fish growth to be embedded in the regional circulation models
• Hybrid (full life-cycle) recruitment models that build on results and understanding gained from the detailed process studies and biophysical models.

Rationale for a comparative study. The synthesis phase of GLOBEC is presently underway in order to fully benefit from regional and national research efforts over the last ten years. Particularly, a comparative study between US and Norway GLOBEC programs (see Sundby, 2005 for a description of the latter) has great synergistic potential. There are several reasons for this:

The gadoid stocks of the Georges Bank and the Barents Sea are among the most explored in the North Atlantic with respect to processes acting on eggs, larvae, juveniles and adults. In both regions, coupled biophysical models for larval and juvenile growth have been developed.

With respect to marine climatology and physical setting, Georges Bank and the Norwegian Sea/Barents Sea regions have their characteristic and very distinct features, and in many ways these two areas display extremes as gadoid habitats of the North Atlantic (Figs. 2a and b). The early recruiting stages of Arcto-Norwegian cod are exposed to a low ambient temperature; the early stages of Georges Bank cod and haddock have considerable higher ambient temperatures. Tidally induced mixing is a major source for turbulence on Georges Bank; at the Norwegian shelf turbulence is principally generated by wind mixing, and only in limited areas is tidally induced mixing a major source of turbulence. The Norwegian Coastal Current is more vertically stratified than Georges Bank due to fresh water runoff from the European continent and the Norwegian coast. The light conditions are very different in the two regions creating a very different setting with respect to predator-prey interactions. On the Norwegian shelf, larvae have sufficient light to feed nearly continuously after the end of May. On Georges Bank there is a distinct diel light cycle with photoperiod ranging from 11 hours in February to a maximum of less than 17 hours per day in June. The forcing from the basin scale differs for the two areas. On Georges Bank, larval loss into the offshore Slope Waters and Gulf Stream is considered to be a source of recruitment variability. On the Norwegian shelf, the dynamics of the Norwegian Coastal Current and the adjacent Atlantic Current makes it less likely that larval loss contribute substantially to recruitment variability.

The biological setting in the two regions is different. The Arcto-Norwegian cod mature at an age of about 6 years, while Georges Bank (GB) cod mature at age 2-3 years. Size at maturity differs as well as the fecundity. The vertical distribution of fish eggs and larvae differs considerable because of differences in egg buoyancy and vertical mixing regimes. In Norwegian waters, C. finmarchicus nauplii and copepodites dominate the diet of cod larvae and early juveniles from first feeding to 2-3 month old pelagic juveniles. On GB the prey species are more diverse and also comprise Pseudocalanus spp., Oithona spp., and Centropages spp., with Pseudocalanus spp. comprising the bulk of the diet during the larval stage. Advection of C. finmarchicus from the Norwegian Sea proper onto the Norwegian shelf and into the Barents Sea is considered to be an important process for food supply of the recruiting larvae and juveniles. Advection is also an important process on GB where C. finmarchicus, Pseudocalanus spp. and Oithona spp. have off-bank sources in the Gulf of Maine. The diet of adult fish is also different with capelin lacking as a prey species on GB in contrast to the Barents Sea where the strongly oscillating abundance of the capelin stock gives highly variable feeding conditions for the cod.

In spite of the different physical and biological settings for the two regions, the parameterizations of the basic biological processes for application in individual-based models (IBMs) are similar, and we propose to continue to develop IBMs to analyze the specific responses in the two regions. Currently, the Norwegian models have a more realistic and detailed foraging component. The US models deal more explicitly with metabolism and temperature effects. Moreover, we are developing similar physical models on the basin scale as well as on the regional scale, making it possible to explore the influence of climate change in a coherent way for the two regions. We will use the Regional Ocean Model System (ROMS) as implemented by IMR for the regional and the basin-scale dynamics with enhancement and refinement for the NW Atlantic.

Hypotheses for the NW Atlantic and Norwegian Sea Systems

NW Atlantic/Georges Bank (Atlantic cod/haddock). The Gulf of Maine/Georges Bank region has a pronounced seasonal cycle of weather, hydrography, and biology.  Peak spawning of cod and haddock on Georges Bank occurs in March and April primarily on eastern part. Eggs and larvae are carried southwesterly along the southern flank with the developing zooplankton populations (Lough and Bolz 1989, Smith and Morse 1985). The highest zooplankton population biomass is generally found on the southern flank (Davis 1987), downstream of the highest phytoplankton production on the Northeast Peak (Cura 1987). The dominant cold-water copepods in the spring are C. finmarchicus, Pseudocalanus spp. and Oithona spp., and they are the primary prey for cod and haddock larvae (Kane 1984, Lough and Mountain 1986). Seed populations for these species are advected onto Georges Bank from various sources in the Gulf of Maine (GOM) and Scotian Shelf (Durbin et al. 1997, 2000). The abundance of Calanus populations is largely determined by the abundance of adult females returning to the surface during early winter from the deep basins of the GOM, rather than advective events that can determine abundance of Pseudocalanus and Oithona (McGillicuddy and Bucklin 2002).

The biomass concentration of copeopd prey on Georges Bank increases through June (Buckley and Durbin, 2004).  Consequently, peak larval cod and haddock abundance is always ahead of, or ‘mismatched’ with their prey.   Moreover, at least for haddock, individuals surviving to the fall come disproportionately from the early portion of the spawning curve (Lapolla and Buckley, 2005).  This mismatch is likely due to a number of factors that favor larger individuals through the first year of life including higher predation rates latter in the year (Davis 1984; Garrison et al 2000) and competition among newly settled individuals.  While this strategy results in lower growth rates (due to lower temperature, shorter photoperiod and lower prey levels compared to later in the season), it minimizes predation mortality by maximizing size before predators reach peak abundance.  In effect, cod and haddock are maximizing size in the seasonal cycle rather than size-at-age.  It also places a premium on the value of larger, older repeat-spawners that spawn early in the season and produce large eggs.

Environmental factors favoring high prey concentrations on GB in March and April favor rapid larval growth, survival and eventual recruitment.  On Browns Bank just to the north, recruitment of haddock was related to the timing of the spring phytoplankton bloom with an early bloom favoring high recruitment (Platt et al. 2003).  During the GLOBEC Program, years of high growth and survival of cod and haddock larvae were associated with low salinity in surface waters and high zooplankton prey biomass (Buckley et al. submitted; Mountain et al., in prep). Moreover, interannual differences in prey abundance and larval growth and survival were seen early in the season and persisted through the period of larval drift, suggestive of processes operating on large temporal and spatial scales.
 
Durbin et al. (2003) outlined a process whereby large differences in the winter phytoplankton abundance in the GOM (Gulf of Maine) can influence zooplankton abundance and productivity ultimately reaching GB. An unusual phytoplankton bloom occurred in winter 1999 in the GOM as well as on the Scotian Shelf (Platt et al. 2003). The initial spread of low salinity Scotian Shelf water across eastern GOM in early 1999 apparently resulted in a shallow, surface mixed layer that allowed a bloom to develop. The winter bloom also may have allowed an additional generation of Calanus prior to the normal March-April spring bloom. Zooplankton abundance in the GOM was an order of magnitude higher in 1999 than 2000 or in the previous GLOBEC years. The high abundance of zooplankton in the GOM during 1999 also was observed on GB where larval growth rate and prey biomass were highly correlated (Buckley and Durbin, 2005). Larval haddock survival and recruitment also were high in 1999 on the Scotian Shelf and GB. 
 
· Hypothesis 1: strong and early influx of Scotian Shelf water to GOM leads to an early phytoplankton boom with increased zooplankton abundance downstream to Georges Bank resulting in increased larval cod/haddock growth.  
 

Norwegian Shelf/Barents Sea (Arcto-Norwegian - AN - cod). Spawning of AN cod occurs in March and April along the mid to northern Norwegian shelf and the eggs and larvae are carried north along the coast into the Barents Sea by June/July as early juveniles (Helle & Pennington, 1999; Sundby, 2000). Compared to other Atlantic cod stocks there is a long route of pelagic free drift for the offspring from spawning in March until the juveniles settle to the bottom in the Barents Sea in Sept to Oct. During this time, the eggs, larvae and pelagic juveniles drift 600 to 1200 km (Vikebø et al. 2005). There is a positive correlation between early juvenile abundance and zooplankton abundance, and between zooplankton and the influx of warm water from the Norwegian into the Barents Sea. C. finmarchicus over-winters in the Norwegian Sea below 600m, the females rise to the surface in the spring to spawn, with the timing of the inflow influencing the number of females, eggs and nauplii transported into the Barents Sea. Thus, interannual variations in temperature and variable volume transport onto the northern Norwegian shelf and into the Barents Sea are important factors for the population abundance of Calanus. Synchrony in growth and recruitment for Barents Sea cod, haddock, and herring is in this way related to temperature fluctuations with warmer years favoring better larval feeding and growth, and higher recruitment (Ellertson et al 1984 and 1990; Ottersen and Loeng 2000; Sundby 2000).

Individual growth of AN cod larvae and juveniles show functional relationships to biotic and abiotic parameters (Vikebø et al. 2005). One of the most important effects is through the influence of temperature on feeding intensity, metabolic rates and thereby growth (Otterlei et al., 1999). Similarly, light conditions and wind-induced turbulence are important abiotic parameters in relation to behavioral responses affecting growth of larvae (Sundby and Fossum, 1990; Suthers and Sundby, 1996; Fiksen et al. 1998). Indirect effects of variables such as temperature through lower trophic levels are similarly important, particularly, the production at lower trophic levels. The copepod species, C. finmarchicus, is the dominant meso-zooplankton in the Subarctic Gyre of the northern N. Atlantic and the main prey item for the AN cod larvae and juveniles. It is adapted to the spring bloom of the region with feeding and reproduction during spring and summer and hibernation at depth during winter. Advected by the Norwegian Atlantic Current, the C. finmarchicus production spills over from the core regions onto the adjacent shelves, e.g., the shelf off Norway and the Barents Sea. Sundby (2000) suggested therefore, that the recruitment-temperature relation of the Atlantic cod is a proxy for the food abundance during the early stages, explained by the advection of warm C. finmarchicus-rich waters from the core production regions to the habitat of cod.

 
· Hypothesis 2: Advection of warm, zooplankton-rich Atlantic water from the Norwegian Sea onto the shelves (Barents Sea) results in increased larval cod growth and survival.

The proposed hypotheses in the two ecosystems will help guide our research activities. They integrate local biotic and abiotic effects, while introducing important effects associated with upstream, basin-scale/climate effects resulting from interactions of the biology and physics of the open ocean and shelf regions that we have not been able to explicitly consider to date.

Prior Support and Background
 
The PIs in the present proposal participated in modeling, field and synthesis efforts in the US GLOBEC Georges Bank Program from 1994-2002. [F. Werner (with J. Runge, co-PI) was funded by NSF grant OCE-9806565: Productivity of Calanus finmarchicus and fluctuations in growth and survival of cod and haddock larvae on Georges Bank: A synthesis of observations and modeling. Oct 1998-Sept 2001; $264,917. Publications are asterisked in the Reference list.] These studies dealt with circulation modeling, larval fish and zooplankton population modeling, in a spatially-explicit, coupled physical-biological setting. The PIs’ modeling activities focused on historical data and its scientific interpretation. Among other aspects, modeling studies established climatological mean physical fields from the Nova Scotian shelf to southern New England and perturbations to them; examined the impacts of these on the life history of specific planktonic species and their interactions; and related these to observations of abundance and distribution of the target species. Herein, we will build on these tools and their extensions. During Phases I-III of the GLOBEC program, we gained considerable knowledge of the physical and biological processes controlling the retention and loss and feeding and growth of cod and haddock larvae on the bank, which we briefly summarize as follows:
 

Interannual zooplankton variability. There was considerable interannual variability in mean zooplankton abundance on Georges Bank. Surprisingly, all taxa showed similar trends over the five year sampling period (Fig. 3 and Fig. 4) with 1995 and 1996 being low years, 1997 and 1998 intermediate, and 1999 a very good year. These changes were negatively correlated with salinity (Durbin and Casas, submitted). The high abundances in 1999 were associated with a phytoplankton bloom that took place during winter in the central GOM and led to high reproductive rates and abundances of the off bank taxa as well as C. typicus (Durbin et al. 2003). This in turn led to high abundances being advected onto Georges Bank, It was suggested that low surface salinities present during winter increased stability of the water column and led to the bloom. There was also a negative correlation between salinity and chlorophyll on Georges Bank itself (Durbin and Casas, submitted) reasons for which are not apparent at this time, but may

include increased grazing pressure.

Retention/Loss. We showed that for late winter/early spring seasons, Georges Bank is most retentive at depth (20m or greater) and shoalward of the 70m isobath (Werner et al. 1993, 1996 and 2001, Lough et al. 1994 and Page et al. 1999). We found that larvae lost from the Bank were primarily in the surface layers, and that the loss depended on the magnitude, timing and direction of wind stress. In more recent studies, we explored the observed 3-D structure of the tidal frontal zone in 1999 (Lough and Manning, 2001), and using data assimilative methods we computed its time-evolution and its possible role in retaining larval fish (Aretxabaleta et al. 2004).

Egg and larval mortality. Mountain et al. (2003; submitted) found that egg and larval mortality varied among years and was related to salinity (Figs. 5a and b); note that 1995 had the highest salinity and the highest mortality. Lapolla and Buckley (2005) used counts of otolith daily rings to determine the hatch date distributions of young-of-year juvenile haddock collected in the fall on GBank. Recruitment appeared continuous rather than episodic. A comparison with hatch date distributions from late stage eggs (Mountain et al. 2003) suggested that individuals hatched early in the season had a strong survival advantage over those spawned later in the season.

Observed growth rates. Recent growth rates estimated from RNA/DNA and water temperature showed pronounced trends with larval size (Fig. 6a) and Julian Day (JD). The JD effect was attributed to an interaction of photoperiod, water temperature and prey availability (Fig. 6b), all of which increase during the larval period on GB (Buckley et al. submitted). Empirical models incorporating larval size and photoperiod explained 48% and 61% of the observed variability in individual growth rates of cod and haddock respectively. We used residual growth rates (the difference between observed growth rate and that predicted using these models) to examine the affects of environmental variability on larval growth. Strong seasonal and interannual differences were seen in larval growth (Figs. 7a and b). Growth rates and residual growth rates were low in 1995 and higher in 1997-1999. Most potential prey taxa were more abundant during 1997-1999 than in 1995. Growth rate and residual growth rate were related to prey biomass concentration (Buckley and Durbin, 2004). Both the prey biomass concentration and residual growth rate were inversely related to salinity (Figs. 8a and b). These studies established that on Georges Bank growth rates of cod and haddock larvae were at times food limited (Buckley et al. 2004) and that maximum growth and optimum temperature for maximum growth varied with prey availability (Lough et al. 2005; Buckley et al. submitted).

Larval Fish Trophodynamic Model. To evaluate the variability on the larval cod and haddock feeding environment, we developed an individual based model (IBM) described in Werner et al. (1995, 1996, 2001), and Lough et al. (2005). The core of this model is the standard bioenergetic supply-demand function, in which growth is represented as the difference between the amount of food absorbed by a larva and the metabolic costs of its daily activities (Laurence 1985). The formulation includes: (i) variable composition of prey fields including 13 stages of 4 prey types (C. finmarchicus, Pseudocalanus, Centropages and Oithona) as well as protozoans during yolk-sac stages (Quinlan et al. 1997); (ii) effect of turbulence, swimming behavior and satiation on encounters and ingestion of larval fish and their prey (e.g., Dower et al. 1997, Lough et al. 1997a, MacKenzie and Kiørboe, 1995 and 2000); (iii) light limitation on ingestion rates at low and at high light intensities (Huse 1994, and Gallager et al. 1995, 1996) and (iv) effects of temperature on metabolic costs, ingestion rates and growth (e.g., Laurence, 1978; Buckley et al. 2000).

Feeding and growth of larval cod. Our studies on the larval feeding environment have shown that during late winter/early spring, effects of turbulence on contact rates and post-encounter capture can be significant (Werner et al. 1996, 2001). For late spring conditions Lough et al. (1997a) found that prey aggregations near the pycnocline are sufficient for rapid growth without turbulence-enhanced contact rates. Werner et al. (2000), using broadscale data for key zooplankton species: Calanus, Pseudocalanus, and Oithona (the 3 principal prey species found in the stomachs of cod larvae for this time period; Lough, unpubl. data), found that for March-April 1995 Pseudocalanus provide the highest growth rates for 6-10 mm larval cod but that the inclusion of Oithona and Calanus was necessary for model larval fish growth rates to be comparable to those in the field. Lough et al. (2005) found that the differences in observed growth of cod larvae in 1993 versus 1994 were best explained using a bioenergetics model when the prey was restricted to Pseudocalanus spp. and the larvae followed their weighted mean depth. Growth rate and residual growth rate of cod and haddock larvae were directly related to the biomass concentration of Pseudocalanus spp., but not Calanus (Buckley and Durbin, 2004).

Proposed Research

In this proposal, we address the overall hypothesis that recruitment variability in cod and haddock populations depends on the growth and survival through the larval stage, constrained by density-dependant mortality during the juvenile stage. Our approach is to combine detailed mechanistic models of transport, growth and survival of early life stages of cod in the NW Atlantic and the Norwegian Sea systems with recruitment models that capture adult population dynamics and juvenile mortality. In collaboration with our Norwegian colleagues we will examine selected years at each site with the objective of determining the factors that cause recruitment variability and formulate recruitment models. We will consider how variations in physical processes, copepod prey abundance, cod and haddock spawning timing, location and abundance, and predation mortality, contribute to recruitment variability of the target fish species. The information and understanding gained from the detailed mechanistic modeling studies will be used to develop proxies and recruitment models (e.g., Bailey et al. 2003) that can be used to recreate the patterns and trends of longer recruitment time series for cod and haddock. For each system, the following approaches will be implemented.

Physical fields. Develop realistic physical conditions hindcast for selected years at basin scales run from a common model and forced by a common set of variables, with increased resolution within regional domains to assess the role of local versus far field forcing in the variability of the advection mechanisms affecting the loss or retention on Georges Bank and Norwegian Shelf, the variability of the temperature and feeding environment, and ultimately the variability in the observed larval fish growth.

Basin scale physical and lower trophic models are available through our collaboration with our colleagues at IMR. For the physical fields (e.g., currents, temperature, salinity) we will use the the Regional Ocean Modeling System (ROMS version 2.1) implemented by P. Budgell (IMR, Norway). In its present form, it has been used to conduct a simulation of ocean and ice variability in the North Atlantic and Arctic regions over the period July 1980 to Dec 2004. The development of ROMS is described in a series of articles (Song and Haidvogel, 1994; Haidvogel and Beckmann, 1999; Haidvogel et al., 2000; Shchepetkin and McWilliams, 2003; Budgell, 2005).

The domain of the present (IMR) model’s implementation extends from 20°S in the South Atlantic northwards through the North Atlantic and including the whole Arctic Ocean to the Bering Sea, and eastwards from the Gulf of Mexico to the Baltic Sea (Fig. 9). The model boundaries are closed, with the exception that there is a constant inflow of 1 Sverdrup (Sv) through the Bering Sea and a corresponding outflow of 1 Sv through the southern boundary. The grid is created from the stretched spherical Mercator projection of Bentsen et al. (1999), with poles located in the Pacific Ocean and in central Asia. The grid resolution varies from 80 km on the southern boundary to 15 km in the Arctic, with typical resolution of 20-40 km in the western North Atlantic. The vertical resolution is a generalized sigma-coordinate system (s-coordinates) with 30 levels placed so as to provide enhanced vertical resolution in the surface mixed layer. In depths of 6000 m the first model level is at a depth of 15 m, at 3000m depths the first level is at 8 m, and at 100 m depths the first level is at 1 m. At 6000m depths, 5 model levels are placed in the upper 200 m. The turbulence closure is the KPP scheme of Large et al. (1994).

The forcing fields are the daily mean fluxes from the NCEP/NCAR Reanalysis (Kalnay et al., 1996) with adjustments to the applied fluxes dependent upon model surface conditions (Bentsen and Drange 2000). Precipitation is obtained from the NCEP/NCAR Reanalysis and evaporation was derived from the latent heat flux values. Freshwater runoff was obtained by combining the NCEP/NCAR Reanalysis accumulated surface runoff over land areas with the Total Runoff Integrated Pathways (TRIP) discretization and network routing of Oki and Sud (1998).

Tides were not included in this simulation, but they will in the next version, which will also be global. The global model will still be focused on the Atlantic and Arctic Oceans, with even higher resolution in the western North Atlantic than in the current simulation, but will avoid issues associated with forcing at open boundaries. One year of model simulation requires 36 hours on a cluster of 16 64-bit Opteron processors on 8 nodes. Storage is roughly 40 GB per year of simulation. The model time step is 900 seconds.

A lower trophic model is run off-line (i.e., using the physical flow fields after these have been computed; M. Skogen, pers. comm.) to estimate nutrients, diatom and flagellate production (i.e., large and small phytoplankton) over the model domain. Zooplankton is not computed, but is specified externally as a grazing term on the primary producers. Model results for selected years, based on the Georges Bank field years (1995-1999), were examined (see Fig. 10 for monthly February values for 1995 and 1999). The results capture the lower salinity values in 1999, compared to those in 1995, as well as the earlier diatom bloom in 1999. These results are consistent with field observations.

Zooplankton (larval prey fields). Data collected during selected years will be used to examine the space-time variability of the larval fish feeding environment. The distribution and evolution of the zooplankton fields will be specified based on the observed structures. If available, we will include evolving prey (zooplankton) fields computed by Davis et al. in their proposal (letter attached) for the NW Atlantic, with similar prey fields provided for the Norwegian System computed by G. Huse, Ø. Fiksen at U of Bergen and their colleagues at the IMR.

Larval fish individual based trophodynamic models – estimates of growth and survival. We will enhance the larval fish trophodynamic models to include revised formulations for foraging and prey selection, behavior, recent findings of larval sensitivity to light intensity and wavelength, and a re-examination of the haddock bioenergetics. The larval bioenergetics and foraging formulations used in most earlier cod and haddock IBMs were largely based on work by Laurence and Beyer (Laurence 1985). We recently completed a major revision of the cod model (Lough et al. 2005) that incorporated light limitation, recent data on cod metabolism (i.e., Finn et al. 2002; Peck et al. submitted), and implemented a more realistic mechanism to model temperature dependence. We now track stomach contents and larvae feed only when space is available in the gut. Digestion is a linear function of temperature. Modeled growth rates better match those observed in the field and the effect of temperature is dynamic, yielding temperature maxima for growth that are dependent on prey availability (lower at low prey abundance and higher at high prey abundance).

Fiksen and MacKenzie (2002) published a process-based model for foraging in larval cod that incorporated process important to prey selection and environmental regulation of feeding. We will combine our bioenergetics model with the foraging component from Fiksen and MacKenzie (2002) to examine the apparent differences in diet composition and preference of cod on GB and Norwegian Sea. This work builds on an ongoing collaborative effort between the U of Bergen (Ø. Fiksen and T. Kristiansen), NMFS (Lough) and UNC (Werner). The haddock model is still problematic in particular there are very limited metabolic data for larval haddock. A project is underway in one of our laboratories (Buckley) that should provide sufficient metabolic data to construct an improved bioenergetics model for haddock larvae.

Predation will be implemented in the model as a function of larval size and water temperature. This base mortality rate will be scaled depending upon the distribution and abundance of potential predators in different zones as the eggs/larvae advect through them or as they develop in the waters along side them. Because we are interested in understanding how and where population regulation may take place, we will require spatially explicit predation mortality fields for larval fish. The broadscale surveys (Madin et al. 1996; Sullivan and Meise 1996) will provide 3D maps of predator abundance, along with maps of predation rate (% removed day^-1 m^-2) on particular prey types; data on vertebrate predators is available from the Coastal Ocean Program predation studies, e.g., Garrison et al. (2000 and 2002).

Recruitment models – full life-cycle representations. We will extend the results of the preceding objectives by linking them to the dynamics of adult stocks to complete the life-cycle representation of the target fish species. First, we will develop recruitment models for cod and haddock on GB with multiplicative and interactive environmental terms using the state-space approach to analysis of multiple time series. Second, we will involve the development of forward projection models for cod and haddock in which the specific mechanisms explored in this proposal affecting growth and survival will be incorporated into stochastic simulations. Simple hybrid recruitment models similar to those developed by Bailey et al. (2003) for walleye pollock and Pacific hake will also be developed and evaluated.

Selection of Years of Emphasis

NW Atlantic/Georges Bank. During the GLOBEC field program, 1995-1999, the largest year class of haddock was observed (1998) on Georges Bank in the past 20 years, although still larger year classes were produced later. The 1999 year class of haddock was also unusually high on the southern Scotian Shelf (Platt et al. 2003). Cod failed to produce a strong year class during the 90s despite management measures taken to increase spawning stock biomass of both cod and haddock. SSB on GB fell to minimum levels in 1993 and 1995 in haddock and cod respectively, before gradually increasing. Based on the monthly GLOBEC broadscale surveys of fish eggs and larvae, the high recruitment of haddock was attributed to higher survival of larvae, approximately the first month from hatching (Mountain et al., in prep). The greater survival of larvae may be related to the better quality and larger eggs produced by the older spawning year classes entering the population. First-time spawners are known to have high egg mortality, with >50% loss due to inferior eggs.

Environmental conditions during the pelagic period also may have contributed to the greater growth and survival of larvae in 1998 and 1999. What is the role of environmental factors versus spawning stock size/composition that led to the greater survival and recruitment of haddock? Why did haddock larvae survive better than cod larvae when they generally spawn in the same area and time and feed on similar prey? The spawning curve of haddock over the GLOBEC field years extends from January to June with peak spawning occurring between February and May. Peak spawning can be very sharp, limited to essentially one month and can occur any month during February-April. In 1998, peak spawning was broader than in other years, extending over the February-April period. Egg production was not greater for haddock in 1998, but survival through the egg and larval stage was significantly greater. In 1999, haddock egg production was higher than 1998 but egg survival was lower. For cod, spawning can occur between December and June with peak spawning occurring in February or March. The 1997 season had the highest abundance of eggs produced but the lowest egg survival.

Our strategy is to look specifically at three candidate years from the GLOBEC time series, 1995, 1998, and 1999, where there is sufficient data to compare and contrast haddock and cod growth and survival during years of high and low survival/recruitment, and where environmental conditions also provide the greatest contrast of physical conditions/processes in order for us to better understand the possible recruitment mechanisms. Characteristics of these years include:

· The 1995 season had low haddock and cod survival during a warm year where Scotian Shelf intrusion was observed in March and the shelf/slope front moved on-bank to the 60-m isobath during May. What effect did these large physical events have on the residing populations?

· The 1998 field season recorded minimum salinity due to the intrusion of Labrador Slope Water which was observed in the Northeast Channel and eventually came on to Georges Bank from the Gulf of Maine. High Calanus abundance also was noted that year from the broadscale surveys, as well as high haddock survival.

· The 1999 season was warmer, more stratified, and an earlier Calanus bloom was noted which may have led to a 3^rd generation (Durbin et al. 2003). Contrasting 1999 with 1998, where haddock survival also was high, will serve to check the hypotheses relating recruitment to secondary production.

Synoptic variation of physical processes for these periods will be provided from the basin scale model. We will also examine variation in life stages of Pseudocalanus, Calanus, and to the extent possible, Oithona that serve as prey for the development stages of fish larvae found on the bank.

Norwegian Shelf. Our Norwegian colleagues have examined several years in their studies. One particular case is their investigation of the direct effects of transport and temperature on distribution and growth of AN cod in 1985 and 1986. These years were chosen because of their differences both in distribution and the individual weight of the O-group cod. Also, the total abundance of O-group cod was higher in 1985 than in 1986, and the center of biomass was displaced further west. The distribution in 1985 covered a larger area than in 1986 and the average length of O-group cod, and therefore the weight (Ellertsen et al., 1989), were significantly lower in 1986 than in 1985 (Ottersen and Loeng, 2000). They used a circulation model combined with individual based larval and juvenile models. The models simulated transport of larvae and pelagic juveniles for the two years (Fig. 11), while keeping record of the individual temperature histories enabling calculation of temperature dependent growth, from spawning until settlement at the nursery grounds.

The present simulations reproduced the variations in horizontal concentration of these two years, though differences in average weights are not: the year 1985 is known to have significant higher average weight (and length) of O-group cod than 1986 (Ottersen and Loeng, 2000). The main reason this is not captured in the simulations is thought to be because the lateral boundary conditions are taken from climatology and do not contain inter-annual variations; their inclusion would also enable estimation of year-to-year variations in fish recruitment (Sundby et al. 1989) and juvenile condition (Ottersen and Loeng, 2000). This topic is planned for continued study. A more sophisticated individual based model where growth depends on food availability, turbulence and light, in addition to transport and temperature, is also envisioned. This will enable us to assess the trade-off between feeding conditions and temperature.

TASKS

The work to be carried out by the US PIs and our Norwegian colleagues are fundamentally the same, with variations in emphasis of the specific domain, and particular field years. Focusing on the activities of the US counterpart, specific steps in our work plan are:

Task 1 - Implement the model solutions for higher resolution NW Atlantic simulations (Werner and Mountain). Initially we will focus on the 1990’s GLOBEC field years and compute the physical fields for 1995-2000. The basin-scale ROMS will be run with focus on the domain approximately corresponding to Fig. (10). We will also compute the lower trophic level response offline and store the data for inclusion in the analysis of the interannual variability. Werner will spend three months (in the Spring of ‘06) at IMR during his Research and Study Leave from UNC to complete this work in collaboration with P. Budgell. Comparison of observed and modeled NW Atlantic hydrographic and flow fields will be an essential part of this effort.

Task 2 – Compile prey fields from observations (Durbin, Lough and Buckley). Existing data in the URI Oracle database will be augmented with additional data from the process studies. Mapping of the staged-data onto the model domain will proceed as in previous efforts (e.g., Runge et al. 2000). The zooplankton data will be used in estimating (using IBMs) the growth of larval cod for the GLOBEC years.

Task 3 – Develop and Implement “Holistic” models (Buckley, Lough , Mountain and Werner). In modeling larval haddock growth and survival Heath and Gallego (1997, 1998) described a “holistic approach” in which temperature serves as a proxy for a wide range of factors which presumably affect the growth of larval fish in the sea. They coupled a hydrodynamic particle-tracking scheme with a logistic growth model incorporating a parabolic temperature term derived by Campana and Hurley (1989) for haddock larvae on Browns Bank in the NW Atlantic. They used this model to investigate the effects of the physical environment on recruitment process of North Sea haddock. Subsequent work has shown that the relationship between temperature and larval growth is highly variable in nature depending upon a variety of factors including prey concentration, photoperiod and ecosystem (Gallego et al. 1999; Buckley 2005; Buckley et al. submitted). Similar approaches have been used in estimating the growth of cod in the Norwegian Shelf System (Vikebø et al. 2005). We have developed empirical growth models for cod and haddock that relate growth rate to larval size and photoperiod. We suggest that photoperiod is a better predictor of the seasonal trend in growth rate than is temperature. We will determine the extent to which these empirical models are ecosystem specific and will use these models to estimate larval fish growth in areas where prey fields cannot be explicitly specified.

Task 4 – Augment and Implement larval fish bioenergetics models (Buckley, Lough and Werner). There are several published bioenergetics models for larval cod growth. Most are based on the early work by Beyer and Laurence (Laurence 1985). Results from our latest version (Lough et al. 2005) are in good agreement with observed growth on GB, capturing the dynamic interaction between temperature and prey availability. We will determine if our GB model reproduces observed growth in the Norwegian Sea system; augment and couple our model to the Fiksen and MacKenzie (2002) process-based model for foraging; and re-evaluate our treatment of search volume and turbulence based on recent findings (Galbraith et al. 2004). Starvation mortality will be implemented using a death barrier. Predation mortality will be implemented in the model using a specified probability of death based on larval size and water temperature (Winemiller and Rose, 1993). A scalar will be included to account for spatial and temporal differences in predator abundance.

Task 5 – Compile a life table for cod and haddock (Lough, Buckley and Mountain). Georges Bank spawning stock biomass and year-class size at age (numbers and biomass) are available from virtual population analysis (VPA) back to 1978 (NEFSC 2002). Egg production, recruitment and total mortality from eggs to recruit (age 1 or 2) will be estimated for these years. Data on egg and larval abundance are available for the years 1977-1984 (MARMAP) and 1995-1999 (GLOBEC). For these years we will be able to estimate mortality during the egg and larval stages (to about 8-10mm standard length). Since sufficient data are available for only two years (1986 and 87) to estimate mortality during the pelagic juvenile stage (Serchuck et al. 1994), we may not partition mortality during the pelagic and demersal juvenile phases. Similar data for the Norwegian Sea will be compiled and life table compiled.

Task 6 – Development of proxies for retention, growth and survival (all PIs). Extension of our recruitment modeling to data-poor years requires the development of indices or proxies for key processes. Rather than relying on time-series data from a few fixed sites, we will use indices based on synoptic variation in the basin-scale and regional model runs. For example, egg and larval retention may be indexed as the deviation from the average retention time of water on the bank between February and May. Surface salinity in the GOM/GB region appears to be a useful proxy for larval feeding conditions, growth, and survival. Stratification and the timing and intensity of the winter/spring diatom bloom (from the NPZ model) will also be evaluated as proxy for feeding conditions and growth. We will evaluate these proxies against the GLOBEC and MARMAP time series. Similar proxies will be developed for the NSS, where temperature is likely to be a useful proxy for feeding conditions, growth, and survival. Mortality during the juvenile stage is most likely dependent on the abundance of vertebrate predators including older cod and haddock. We will use assessment data (NEFSC, 2002) and the NEFSC food habits database to scale juvenile mortality within the range generated from the life table.

Task 7 – Hybrid recruitment models (Fogarty, Buckley and Lough). The models developed in the preceding sections focus on key elements of the recruitment process growth and survival during the early life stages.  We will extend these results, and the associated proxies, by linking them to the dynamics of the adult stock to complete the life-cycle representation.  The classical stock-recruitment relationships entail consideration of population egg production (knowing the spawning stock biomass) and factors affecting pre-recruit survival, but typically do not explicitly include the role of exogenous forcing factors.  Extended recruitment models that do include environmental effects have been developed (e.g., Bailey et al. 2003).  These models typically represent environmental factors as multiplicative effects.  A recent application of this general approach to cod populations throughout the N. Atlantic is provided by Brander and Mohn (2004) who examined the effect of the NAO on cod recruitment.  Stenseth et al. (2002) have proposed extended population models with interactive effects between environmental forcing and population density (see also Ciannelli et al. 2004).  Given a sequence of estimates of recruitment, adult abundance and environmental factors, model estimation is generally undertaken in a time-series analysis context and can be represented in a general state-space framework.
Extensive time series of cod and haddock recruitment and adult population sizes are available for Georges Bank (NEFSC 2002) and a major update of these estimates is scheduled for August of 2005.  Comparable time series are available for cod off Norway.  We will explore two distinct approaches to the development of extended recruitment models.  In the first, we will develop recruitment models for cod and haddock on Georges Bank with both multiplicative and interactive environmental terms using the state-space approach to analysis of multiple time series.  This approach will also be applied to Norwegian/Barents Sea cod.  The environmental terms to be included in the models will be based on the results of the preceding sections of this proposal. The second approach will involve the development of forward projection model for cod and haddock in which the specific mechanisms explored in this proposal affecting growth and survival will be incorporated in stochastic simulations (see Fogarty, 1993).  These projection models will also allow consideration of the interactive effects of harvesting and environmental change on cod and haddock populations.

Using the life table described above, we will parameterize simple hybrid recruitment models (Bailey et al. 2003) for cod and haddock. Stochastic simulations will be completed in which the range of mortality for successive early-life stages is adjusted according to environmental conditions, and juvenile mortality is a function of density and predator abundance.  This approach does not entail estimation or fitting of time series data to determine the role of environmental forcing but rather uses independent information to set probable ranges.

Significance and Project Integration

The model-based synthesis proposed directly addresses GLOBEC4b program goals to synthesize the research findings of the Georges Bank program into the context of basin-scale phenomena in the N. Atlantic and the use of that knowledge to investigate the GBank ecosystem response to future climate variability with particular reference to the target species cod and haddock.

Continental shelf and marginal sea ecosystems – such as the GBank and Norwegian Sea systems – are affected by basin-scale forcing, and on interannual to decadal scales cannot be studied in isolation. Modulations in the circulation and feeding environments of marginal seas can result in increased connectivity of distinct, previously isolated populations, or they can affect growth and possibly population recruitment through changes in transport and the feeding environment. Large-scale circulation features often define the geographical distributions of species and changes in physical oceanic conditions at basin scales will affect the organisms’ growth and survival directly, e.g., transport of larvae or prey, changes in temperature affecting vital rates, or indirectly by changes in nutrient/food supply (deYoung et al. 2004; Runge et al. 2004; Werner et al. 2004).

The proposed study provides an inter-comparison of results from the NWA/Georges Bank with other Atlantic Basin scale programs such as the Norwegian Sea ECOBE, the EU Baltic Sea UNCOVER, and others, where the physical environment affects cod differently in all there systems. Sufficient data, model development, comparative methodology, and synthesis will be available for these systems to begin a more in-depth comparison with respect to local and large scale physical forcing and the broader implications of climate change.

Our study can be integrated with other synthesis efforts that propose to generate the evolving zooplankton fields on GBank during the GLOBEC field years (Davis et al.), consider the basin-scale variation of Calanus (McGillicuddy and Haidvogel et al.), and study NAO-associated changes in ocean circulation patterns and impact on marine ecosystems (Greene et al.).