The original target species meeting agenda, which addressed cod and haddock larvae on Wednesday, Pseudocalanus spp. on Thursday, and Calanus finmarchicus on Friday was revised to reflect participant availability. In the revised agenda, Wednesday was reserved for general discussion, Pseudocalanus and cod and haddock larvae were discussed on Thursday, and C. finmarchicus on Friday.

Wednesday, 19 January 2005

The first hour was used for computer and internet setup. Catherine Johnson opened the meeting by soliciting participants' priorities for discussion topics and data sharing needs. Jamie Pringle suggested that a species budget or box model for the Georges Bank / Gulf of Maine be developed in order to quantify sources of variability and uncertainty, and this topic was the subject of much of Wednesday's discussion, along with the status and availability of various data sets. A box model approach minimizes model and conceptual complexity and can be used to test hypotheses about the sources of population variability of the target species. For example, this approach could be used to compare the contribution of transport and population growth/mortality to zooplankton population variability. It could also be useful to the energy flow group. Ann Bucklin emphasized the need to focus modeling efforts on the target species and to complete description of target species variability patterns. She focused the discussion on the specifics of box model development. Key questions in box model development are (1) how to define the best domain for the model; (2) how subregions of the domain should be defined; (3) how vertical resolution should be defined; (5) what temporal resolution should be used (e.g. climatological; interannual?); and (5) how to formulate the developmental model. The answers to these questions will likely be different for different target species; for example, Georges Bank may be a reasonable domain for a Pseudocalanus box model, while the GoM/GB region may be better for modeling Calanus dynamics.

Data requirements for a box model will include population abundance(A), birth rates(B), mortality rates(D), and immigration(I) and emigration(E) rates, all of which are currently available for the GLOBEC Georges Bank region. Abundance data can be provided by Ted Durbin and Ann Bucklin. Birth rates can be calculated as a function of chlorophyll and temperature, based on functional relationships developed by Jeff Runge. Peter Wiebe noted that gridded chlorophyll data are now available. Mortality rates can be provided by Runge and Mark Ohman. Immigration and emigration rates can be calculated from target species concentrations and volume transport. Species concentrations can either come from the GLOBEC dataset or from literature review, depending on the model domain, and transport can be estimated using FVCOM (Pringle and Johnson). Peter Wiebe suggested that copepod prosome length data from the Gulf of Maine and slope water could be used to help constrain a box model, and later gave an update on the status of copepod length data and intercomparison of two different prosome length measurements. J. Pringle volunteered a temperature/salinity data set that he has compiled, for use by other GLOBEC investigators.

In the afternoon, participants worked independently or had discussions in small groups. At the end of the day, the group discussed the plan for the following day.

Thursday, 20 January 2005

Catherine Johnson opened the meeting by reviewing the agenda and goals of the meeting and asking participants what questions they felt were important to discuss, what data they needed to share, and whether they wanted to add anything to the agenda. Participants suggested returning to this question after the presentations. Johnson then summarized the previous day's discussion of using a box model approach as a way of facilitating synthesis, and asked for feedback about whether this approach would be useful.

The discussion began with alternative hypotheses and questions that could be tested using the box model approach. For example:   To what extent can we explain zooplankton processes on Georges Bank using bulk or biomass measurements? Does Georges Bank define a functional, self-sustaining population of both Pseudocalanus species? Peter Wiebe described new data assimilation modeling results by Li and McGillicuddy, to be published in the upcoming Deep-Sea Research volume, suggesting that the latter is true; however, he suggested that the influence of reseeding the Bank should still be considered since the relative importance of immigration/emigration and birth/death may vary seasonally.

Discussion returned to data sources and the appropriate levels of temporal and spatial resolution for the box model. Jamie recommended starting with climatology first, and then move to time-specific data. A 'graded approach' to determining minimum resolution was suggested, in which different levels of resolution would be tried to determine the minimum required to resolve population variability. For Pseudocalanus, birth rate functions are available from Jeff Runge. Dian Gifford noted that use of size-fractionated chlorophyll measurements, where they were available, might improve estimates of egg production. These measurements are available for process cruise stations in 1995. Jeff Runge and Mark Ohman have the best mortality rates available. Abundance estimates, from Durbin and Bucklin, will include both pooled species data and separate species. Sensitivity analyses can be used to determine what the pooled species data are missing. Transport can be estimated from FVCOM, to provide immigration and emigration rates. Similar data sources are available for Calanus; however, the spatial domain may need to include areas beyond the Bank.

Cod and haddock larvae data for use with a box model are available, but are not as complete as those for copepods. There is need for monthly tracking of larvae. Data on the factors influencing growth are available for egg through all larval stages, up to juveniles (which are caught in fall trawl survey for assessments), and can be used to recognize good year classes. This first look at the year class strength tends to correlate with later catch data, although we don't know much about biology of these species populations. Monthly data on mortality of early larval stages are good, but data are not available for later stages that settle on the bottom. Swimming behavior and residence on the bottom are additional factors to consider in formulating cod and haddock larvae box models. Jamie Pringle said that one approach is to put all the uncertainty in the models into mortality (but this implies you have absolute faith in all other parameters). Another way is to estimate mortality from the literature values, and then see whether the model closes. Uncertainty in parameters can be estimated, and used to estimate uncertainty in model results. The McGillicuddy model now constrains B and D, using food availability, mortality, and diapause. With this approach, the model results are consistent with data, e.g., a population "hole" develops on top of the Bank, similar to what is seen later in the season.

A brief period of concluding comments ended the morning discussion. Ann Bucklin suggested development of target species box models in parallel and to build in plans for cross-linkage. Beth Turner asked if the box model approach would allow for scenario testing. Scenario testing would be possible, and it would be accomplished by modifying the parameters. Peter: David Mountain has done work on larval abundance vs recruits for cod and haddock, and found haddock get 5X recruitment per larva. This suggests there is density-dependent post-larval mortality, due to change in environment from when stocks were much higher. Larry: But be careful since this conclusion results only from 5 years' worth of observations, which cover at least 2 excellent years for haddock.

Following the morning discussion, Ann Bucklin presented a talk titled 'Maintenance of Pseudocalanus spp. on Georges Bank: differential use of the Bank.' Bucklin described similar seasonal population abundance patterns between P. moultoni and P. newmani, but noted differences in spatial usage of the bank that may be due to the species' different vertical distributions. She also suggested that Pseudocalanus on Georges Bank and in the western Gulf of Maine are functionally distinct groups, although they are not genetically distinct.

Jeff Runge presented a talk titled 'The effect of spatial and temporal variation in zooplankton concentrations on larval cod growth on Georges Bank: a comparison of two years based on modeling and observations.' Runge discussed trophodynamic modeling efforts for cod, and emphasized the importance of Pseudocalanus spp. as prey for cod, especially for early larval stages in 1995. Key questions that remain in understanding the connection between cod larvae and zooplankton are evaluating the role of prey selectivity, clarifying cod gut contents analysis, perhaps by genetic analysis, and identifying vertical partitioning of larval feeding. Runge suggested that a similar trophodynamic model should be developed for haddock larvae; this would require metabolic rate and gut contents data.

After a break for lunch, Larry Buckley gave a presentation titled 'Factors affecting growth of cod and haddock larvae on Georges Bank,' which described work on cod growth in 1995, 1997, and 1999. Photoperiod and larval size explained much of the variability in cod growth during these years, and temperature explained very little. In addition, there was a strong year effect, with both growth and growth residuals lowest in 1995; this may reflect food availability or parental factors. Analysis of prey preference suggested that for early cod larvae, Pseudocalanus is an important prey item, and Calanus is less important. Variability in growth also appears to be related to salinity on Georges Bank. Causal mechanisms for this relationship are unclear.

Following Buckley's talk, C. Johnson asked if covariance of other zooplankton species with Pseudocalanus could be reinforcing the observed patterns. Buckley commented that, based on otolith data, the mean birth-dates of survivors are early compared to the overall distribution of birth-dates, despite better feeding conditions later in the season. Thus, higher survivorship for early-spawned larvae may be due to higher mortality later in the year, rather than to feeding conditions. It is generally accepted that as larvae grow, mortality decreases. Thus, fish larvae should grow as fast as possible, to get through vulnerable stages as quickly as possible to escape the highest mortality periods. This is driven by predation, but also by metabolic demands of rapid growth at higher temperatures, which generate higher mortality later in the season. There seems to be no significant interannual variation in this pattern. Peter Wiebe noted, however, that mortality on larvae is believed to be "chronic", not catastrophic with dramatic increases in mortality of larval fish. Mortality is a critical question for box model development, and data are available to determine correlation between estimated mortality rates and predation.

After discussion of the presentations was finished, the group briefly reviewed priorities for the afternoon discussion session. Priorities were as follows:

Ted Durbin then presented a talk titled 'Interannual variability of copepods on Georges Bank.' The Georges Bank copepod community can be sorted into two groups. Group I includes copepods with off-bank sources, such as Calanus finmarchicus, Metridia lucens, Pseudocalanus spp., and Oithona spp.; the abundance of species in this group increases when copepods arrive on the bank in early winter. Group II includes copepods with local, on-bank sources in the winter and spring, including Temora longa, Centropages hamatus, and C. longicornis; the abundance of species in this group increases later in the spring. In most species, the abundance of juveniles and adults were correlated; the exception was C. finmarchicus. Populations of dominant copepods on Georges Bank appear to respond similarly from year to year. Populations were high in 1998 and 1999, years of high chlorophyll, which they were low in 1995 and 1996, years of low chlorophyll.

Afternoon discussion focused on developing a plan for implementing the box model. J. Pringle and C. Johnson volunteered to develop the basic model, with functional relationships provided by other investigators. Johnson will discuss the developmental model with Wendy Gentleman, who has been developing methodology to reduce the effects of developmental diffusion. Because development in the box model is one-dimensional, it will not involve a lot of computing time. Jeff Runge and Mark Ohman will provide birth and mortality rate functions. Peter Wiebe brought up changing spawning patterns in haddock (i.e. a change from the Northeast Peak to the southwest corner of the bank) and asked if the model could include this information. Larry Buckley suggested that more boxes could be added to address this shift, and Runge suggested that population dynamics of key species be explored first, and spatial-explicit models could come later. As a first approximation for spatial subdivision of the Bank, Ted Durbin suggested using regions 2, 3, and 4 (i.e. crest, Northeast Peak, and southern flank) from the Grid Group.

The afternoon wrapped up with discussion about the key questions that should be addressed by the box model. Jeff Runge suggested going back to the original questions in the Implementation Plan, and it was decided that these would be discuss ed on Friday.

Friday, 21 January 2005

The meeting began with a brief discussion of priorities for the day, which were discussion of chronology; discussion of hypotheses presented in the Implementation Plan and prioritization of research questions; and further discussion of the box model.

Dian Gifford gave the first presentation of the day, titled 'The diet of Calanus finmarchicus on Georges Bank' The relationship between chlorophyll and microplankton was seasonally variable, and chlorophyll could be low when microzooplankton biomass was high. However, microzooplankton are part of the diet of Calanus everywhere on Georges Bank, and so they may have important effects on development and egg production; these effects would not be evident in relationships developed using chlorophyll alone.

Jeff Runge presented a talk titled 'Characteristics of Egg Production of Calanus finmarchicus and Pseudocalanus spp. on Georges Bank: 1994-1999.' For both Calanus and Pseudocalanus, chlorophyll was a good first-order proxy for food availability. Bob Campbell commented that relationships between growth rates and chlorophyll were similar. Use of size fractionated chlorophyll as a proxy for food availability was also discussed. Because these data are only available for a limited selection of stations in the first two GLOBEC-GB years, Ann Bucklin suggested that a few 'cameo' model realizations, with full food characterization, be compared with runs using only chlorophyll.

Ebru Unal gave an overview of her poster 'Interannual variability of population genetic diversity of Calanus finmarchicus in the Gulf of Maine, Northwest Atlantic.' Analysis of the metabolic citrate synthetase gene revealed interannual population genetic variability between 1997 and 1999. Dramatically low abundances of C. finmarchicus were also observed in 1998, the year of the cold slope water incursion into the deep Gulf of Maine. This time period formed a cluster genetically distinct from several years when warm slope waters were present in the deep GoM.

Bob Campbell summarized his poster 'Calanus Production on Georges Bank,' which presented estimates of Calanus production based on temperature and food availability. Food limitation was common on the bank. Despite the Calanus "hole" on the crest, food limitation on the crest was lower than in other areas, suggesting that mortality may drive low population numbers in this area.

After lunch, Jamie Pringle presented a talk, 'Sources of circulation variability in the Gulf of Maine,' which quantified the sources of inter-annual variability of the circulation in the central Gulf of Maine. The most important source of variability is the year to year change in the density field in the Gulf. Variations in the density field are driven by variations in the density entering the Gulf, and the strength of the cooling and wind stress which drive wintertime vertical mixing and the formation of Maine intermediate water.

Catherine Johnson concluded the presentations with a talk titled 'Transport and retention of dormant copepods in the deep Gulf of Maine.' Copepod retention estimates are relatively high under all conditions simulated. Two factors that are least understood -- behavior and variability in Scotian Shelf inflow -- can strongly influence retention. Alongshore wind variability enhances retention; its influence is weaker than behavior and Scotian Shelf inflow variability.

Chronologies:
Following the presentations, discussion turned to chronologies. The key question was how to go about developing a chronology, and what format would be most effective. Beardsley and Irish have developed a textual chronology describing physical events; this type of approach could also be useful for biological data.

Two kinds of data need to be included in a chronology: seasonally recurring events, such as timing of the spring bloom, the first generation of Calanus, and extreme production limitation; and unusual events, such as 'the year of the pteropod, 1998' and Scotian Shelf cross-overs. It was noted that unusual events may be valid or they may merely have been noticed on a particular cruise. Hydroid outbreaks were given as an example - it will be necessary to go through the data and confirm the 'casual chronology.' One idea for cataloging unusual events is to send out surveys to the investigators involved in the cruises.

One central question that the chronology can address is whether there is a relationship between physical events and biological response.

The format for presenting seasonally regular events was discussed. Jeff Runge asked if it should be a timeline, which can be clicked on. Peter Wiebe suggested starting with a table form, month-by-month, listing noteworthy events, which could then be formatted. Catherine Johnson suggested building seasonal cycles that can be compared between years. Runge suggested starting with a 5-year (1995-1999) X-axis, and showing timelines for each parameter. Johnson commented that events will show this way, but not processes and patterns.

Spatial subdivisions of the Bank and comparison of chronologies for different parts of the Bank were discussed. How do we define the regions? Should they stay the same for species? Maybe we need hydrographic regions as well as geographic (or bathymetric) regions? Whatever we do, we should all do the same thing, except we need to be vigilant for "wild outliers" that actually result from mis-identification of regions. The general consensus was that geographic regions are better, despite their disadvantages.

ACTION ITEMS

Peter, Ted, Bob, Jeff: Will come up with a template for which population dynamic parameters should be included for all species:

1.Distribution and abundance
2.Rate process measurements (production; egg production)
3.Biological events (whales feeding)
4.Physical events

What are important questions & can they be addressed with a box model?

Jamie Pringle requested that participants state a set of questions or hypotheses that we can address with a box model. For example, for the key species, to the extent possible, can we break down year-to-year variability into changes in food, predators, inflow from off Bank, and environment (temperature, etc.)? If you assume the system is linear, once we get a mean model, you can run the years with different forcings (food, predator concentration, inflow). Is this a good assumption for fish larvae?

Participants suggested questions. For example, do different forcing events force the events you see? Could you sustain a local population on the bank? And if not, what would you have to do to make it work?

Several methodological issues were addressed, including how the Bank should be divided up. Consensus was that Grid Group regions 2, 3, 4 are resident sub-regions, 1 is inflow, and 5 is outflow. Jamie Pringle asked if we have enough data to parameterize the processes? This is different from McGillicuddy, because this box model would not try to estimate a parameter assuming all others are correct. Instead we are trying to build the best community model. Ted Durbin replied that Dennis will get the mortality field, and compare it to Ohman, Durbin, et al.'s estimates of mortality. Durbin also recommended a sensitivity analysis. Carin Ashjian asked how this effort differed from Cabell Davis' work on Pseudocalanus (1984), and recommended revisiting this paper before starting box model development.

Discussion turned to the Implementation Plan (US GLOBEC Report No. 6) to compare current questions to the program's original questions. Jeff Runge presented the questions on powerpoint slides. For a list of hypotheses, refer to http://www.usglobec.org/reports/nwaip/nwaip.chapter4.6.html

Slide 1 (Jeff): Local Growth vs Retention/Exchange questions:
The first two are doable, the third can be estimated by vertically partitioning the boxes, and the fourth not appropriate.

Ted Durbin noted that Centropages needs a benthic source, similar to hydroids and T. longicornis, to sustain population; what do you have do to Calanus to kill it off on the Bank later in the season? Why can Pseudocalanus persist on the crest? Is it better at predator avoidance than Calanus? The Houghton group is working on nutrient and front hypotheses.

Slide 2 (Jeff): Stratification:
The box model approach is not appropriate to these hypotheses.

Slide 3 (Jeff): Episodic gains and Exchanges/Loss:
We know that the first question is true. The second question is interesting and doable, and should be focused on Pseudocalanus. The third question can be done better with particle tracking, similar to Lynch. Pringle stated that the fourth question has been answered -- confirmed by Craig Lewis' work that average winds can be used to force the model just as accurately as daily winds. However, Jeff Runge noted that question four referred to fish larvae rather than to copepods. Wiebe recalled Greg Lough's survey of fish, in which larvae disappeared after a storm. Pringle commented that storms matter to the extent that they matter to the mean. Peter asked if we could put fish larvae into a model and advect them and see where they go.

Slide 4 (Jeff): Mortality question (predation vs. starvation for fish larvae; predation vs advective loss for copepods):
This question could be addressed by the box model. Further discussion addressed the importance of off-bank sources for observed developmental stage structure patterns and for driving abundance variability in the target species.

Would the box model be a good way to sum up information collected during the GLOBEC years, leading to publication of a book? Wiebe and Pringle felt that it would not. The box model should be used to explore target species issues, not focused on a book.

Jeff Runge reiterated that Pseudocalanus dynamics have become more important with the realization that they are a primary food for larval fish. Key questions about Pseudocalanus include the following: Are birth rates sufficient to maintain Pseudocalanus on the Bank? Do we expect Pseudocalanus to be more variable than Calanus? What controls its dynamics, especially on the southern flank?

Ted Durbin said that the box model can address questions about the dynamics of re-seeding the Bank each year. A key question is the importance of re-seeding each year vs maintenance of local populations on the Bank. Relevant events include abundant Pseudocalanus in the GoM during Winter 1999.

Bob Campbell noted that larval fish growth rates are correlated with Pseudocalanus spp., based on the trophodynamic model and correlation between RNA/DNA ratios and abundance of Pseudocalanus in the net samples. Ted Durbin noted that Larry's samples come largely from April/May, when Pseudocalanus is more abundant. If we looked earlier in the year, we would perhaps see higher abundances - and infer higher importance - of Calanus. Or, maybe Pseudocalanus is more variable (less reliable), so Calanus are an important - predictable - prey species. Campbell stated that it was not necessarily that fish larvae are choosing Pseudocalanus, but Pseudocalanus abundances and fish larvae condition are correlated due to larger-scale forcing. Jeff Runge said that this is the opposite of the expectation from the Gulf of Saint Lawrence; Pseudocalanus is more variable than Calanus. Durbin suggested looking at seeding of species on the Bank: a lot or a few? Where are they from? What are interannual patterns? Another important question is, what are important sources of variability in growth? Feeding selectivity experiments need to be done.

ACTION ITEM:

Pringle and Johnson: Initial examination of Calanus and Pseudocalanus spp. using a box model approach. After using pooled data for Pseudocalanus, then do sensitivity analyses with species data from Bucklin for selected years.

Box model details not today.

Next meeting:

Informal, at WHOI. Announce in advance to gball list, so that all interested investigators may come.