Report of the Broad-Scale Data Coordination Workshop
The Broad-Scale Data Coordination Workshop was held on July 19-20, 2000 in the Carriage House at WHOI. The goals of the workshop were:
To insure the preliminary analyses of the Broad-Scale data sets collected on Georges Bank are done and stored in a consistent manner for subsequent synthesis and modeling activities.
To determine which Process data sets have a sufficient spatial extent to be integrated with the Broad-Scale data, and how to accomplish the integration of different data sets.
To identify any processing or analysis steps that would make the data sets more readily available or useful in synthesis and modeling activities.
To identify important or unusual events or characteristics evident in the data sets which could become topics of analysis in subsequent synthesis and modeling activities.
The agenda for the workshop is listed in Appendix A. Before the workshop, attendees were asked to provide a list of the research questions they hope to address with the Broad-Scale data, as well as a listing of the Broad-Scale data sets they had. The input received are given in Appendix B and Appendix C.
Research Questions
To begin the meeting most of the investigators provided a brief description of the research questions they hope to address using Broad-Scale data. Most presentations also provided a status of the data processing and analysis. This required more time than anticipated, but also proved informative. Major points presented were:
David Mountain - for the ichthyoplankton and hydrography Broad-Scale investigators, the primary question to be addressed is what controls the growth and survival of the cod and haddock larval populations. Addressing this will require collaboration with the BrSc zooplankton and predation groups, as well as the circulation modelers. A request was made for a timeline of when the BrSc ichthyoplankton data will be available.
Peter Wiebe - using the acoustic data from the Greene Bomber and the BIOMAPER-II, a major issue will be separating the signal due to the biology from that due to the physics. Under the physics acoustic backscatter could be associated with internal waves, fronts, and secondary circulation cells. The displacement volumes from net tows will be used to calibrate the acoustic backscatter and help solve the separation problem.
Ted Durbin - the BrSc zooplankton data indicates high Calanus nauplier abundance in April 1996 and 1998. It is possible that small shifts in the timing of the cruises may have resulted in missing the peak abundance in other years, particularly in 1997. The peak in abundance of older copepodites during April of each year was due to a combination of the G1 and G2 generations. Ted pointed out the survival during diapause may be greatly influenced by the bottom water temperatures in the Gulf of Maine during the previous fall and winter. If high, the metabolic requirements might be greater than the energy stored and result in greater mortality and a smaller initial population the following spring. For analysis, a primary interest will be in the spatial and temporal variability in the zooplankton species dynamics.
Erich Horgan - the predation group has completed its data processing and is in the process of getting the data on predator abundance, biomass, and size frequency into the GLOBEC database.
Karen Fisher - the temperature, salinity, and fluorescence along track data from various cruises is being analyzed using wavelet analysis. The primary interest is to look at the spatial character of variability in these parameters and its year to year changes. The vertical structure also will be investigated using the CTD data.
Jim Bisagni - two sets of satellite data are available via the web. The first is the images in JIF format, which are available through the GLOBEC data system. The other set contains the actual data values and is available through the DODS system. For research the interannual variability in cross-over events and in the paths of warm-core rings are of primary interest. The intent will be to compare the nutrient and biological (ichthyoplankton) character of the cross-over events with the rest of the Bank.
Dianne Gifford - sampling the prey field for the zooplankton and larval populations, with the focus on the diet of copepodites in Phase I and on the nauplii in Phase II. The broadest spatial coverage was in 1995.
Larry Buckley - measurements of the biochemistry of cod and haddock larvae for determining growth rates. The samples are from both some BrSc surveys and from Greg Lough's processes cruises. Results indicate the about 60% of the variability in growth rate can be explained by temperature. This is more than one would expect, which might mean that there is a relation between temperature and prey availability.
Charlie Flagg - The ADCP data have been processed and are available for 75 to 80 cruises. A minor error in processing was discovered that resulted in some increase in hort term jitter, but this is being corrected. The ADCP data from a number of the BrSc cruises on ALB IV were not recorded and, thus, not available. Work is progressing on up-dating the tidal model for the Bank.
Jeff Runge - Estimates of egg production rates of target species and others. A revised reproductive index for Calanus in food limited condtions is being developed with B. Neihoff from the BrSc survey data. A similar analysis will be done for Pseudocalanus. The intent is to get a total integrated production estimate copepods as a measure of prey for the larval fish.
Greg Lough - focusing on the development and fate of the larval fish population in 1995, following a cohort on three subsequent cruises.
David Townsend - The chlorophyll data indicate that production already going in January. The lowest values are characteristically in May. The water column warming in May and June promotes the recycling of silicate. From January to March the diatoms are dominant. In May and June the flagellates become dominant. On a separate note, there was substantial interannual variability in the nutrient sources to the Gulf region, as indicated by values at station 25. In particular, there were low values in 1998 when the Labrador Slope Water was observed entering the Gulf .
Karen Wishner - The primary contribution of data for BrSc interests will be from the sources and sinks work in 1997. The work focused on features exporting water from the Bank and cross-over features that were sources to the Bank.
Data Base Tools
Bob Groman presented an overview of various data access tools that are available or are being developed. He started with a brief description of DODS (the Distributed Oceanographic Data System), which has recently been awarded funding by NOPP to become a national system. Peter Cornillion at URI is the PI for the DODS. DODS is basically an enhancement of the JGOFS system (on which the GLOBEC database system is based). The DODS will handle data other than ascii - e.g., binary data, matrixes, and vectors.
Bob then gave a description and demonstration of a number of new tools available through the GLOBEC website for accessing and analyzing GLOBEC data. The tools relate to data retrieving, data display, and data searching. The online capabilities are continually being enhanced and Investigators can check the website to learn about the latest advances.
Combining Data Sets
John Quinlan probably has spent as much or more time than any other PI in bring pieces of various BrSc data sets together for an integrated analysis. He was asked to give any insights he might have on the types of problems he has encountered. John is working with Greg Lough and others to look at what happened to the cod and haddock larval populations from March through May of 1995. This has required using data from both the BrSc survey cruises and from a number of process cruises. A first problem was that since much of the data were not year available online, he had to work directly with various PI's to get the relevant data. Another issue was that the level of spatial detail varied considerably between the different types of cruises (survey vs process). In mapping the parameter fields to a model grid, the inclusion of uncertainty fields, as done by OAX and by kriging, was important in keeping track of the difference in detail in the original data sets. Another situation that caused problems with some biological data sets was knowing when no information at a station meant that there were there were no individuals of the species of interest found at the station, or that the samples from the station had not been sorted. He said it was very important to have someone who collected the data involved in working them up.
Dan Lynch discussed an new interpolation method being developed by the modeling group, Oafe (objective analysis finite element). Oafe takes the flow field pattern into account in determining the correlation scales to be used in the objective analysis. Generally parameter patterns are coherent over longer distances along the direction of flow than in the cross flow direction. The method is not available for general use yet, but will be in the near future.
David Mountain presented some comparisons of estimates cod larvae abundance on the Georges Bank derived using different interpolation methods (figure 1). The original data was from two BrSc survey cruises. The two spatial methods used were kriging and linear interpolation to map the BrSc data to the Bank 150 grid. In addition, a non-spatial method, referred to as the delta-distribution, was also used. The delta method assumes the observations are random draws from a population that has both zero values and non-zero values that are log normally distributed. For such populations, it provides an unbiased estimate of the mean and variance, and is used in analyzing data from the adult fish surveys. For the two spatial methods the larval data was log transformed before the methods were applied. Then the values at the grid points were estimated and, knowing the area each grid node represented on the Bank, the total population of larvae on the Bank was estimated. The kriging and linear interpolation methods were applied using various numbers of original data points to make each estimated value. As the number of points was increased, the estimated population level greatly decreased - by up to an order of magnitude. This decrease was due to the larval distribution being quite patchy, generally with one, large high-density patch on the Bank. The smoothing, by either method, of the transformed values greatly reduces the influence of the few, very high larval catches. The linear interpolation suffered this problem less than did the kriging, and was much closer to the estimate from the delta-distribution method.
Modeling
Dennis McGillicuddy provided an example of using adjoint modeling in the study of the population dynamics of Pseydocalanus. With sufficient information on the distribution of the species at two times, the modeling can estimate the distribution of the source or sink of organisms during the intervening period. It was felt that this could be a major tool during the synthesis phase of the program.
One limitation of the method is that the flow field is assumed accurate and any error in the estimated flow field contributes to the estimated biological source/sink term. Dennis pointed out that having a number of species can provide a constraint on the physics and indicate if the flow field has a problem.
Analysis Topic Areas
An extended discussion took place to identify primary topic areas for analyses that would be based on the BrSc data sets:
- Larval trophodynamics - Greg Lough and others will be focusing initially on 1995, as was outlined by John Quinlan.
- Zooplankton trophodynamics - adjoint modeling by Bucklin/McGillicuddy and population dynamics analyses and modeling by Durbin and by Miller.
- Variability in the water properties - on a regional scale considering the role of advective vs local changes, using the BrSc hydrography, mooring data sets, and the circulation models.
- Statistical analyses of the community structure - whether there are any fundamental ecological characteristics of the system that can be investigated such as size frequency distributions or energy flow between trophic levels.
- Nutrient and primary production modeling - looking at the limitation on production by different nutrients at different times in relation to productivity, as outlined by David Townsend.
- Relationships between water column conditions and surface conditions observable by satellite.
- Comparisons with other systems around the North Atlantic Ocean to consider climate connections and local manifestations of climate variability. This work would be done in close cooperation with European investigators.
Phase IV Preparations
In preparation for Phase IV proposals, it was decided that another workshop should be held in November when issues relating to specific analyses (such as those listed above) can be addressed. In preparation for the November workshop two activities were agreed upon.
First a Mapping Working Group was formed to further investigate the various methods of mapping or interpolating the BrSc data sets to the model grids. In addition the group would work with the modelers to develop a model grid suitable for the BrSc analyses. The group members are Wiebe, Mountain, Durbin, Quinlan, McGillicuddy, Chu, Lynch. See the report of a short session from the group which meet right after the close of the meeting (Appendix D).
Second, a summary of the variability and major events observed in the various data sets will be prepared by the appropriate PI's, along with time series of distributional plots for the key variables. The key data sets and the suggested lead PI(s) are:
Hydrography - Mountain & Taylor
Nutrients/chl - Townsend
Plankton displacement volume - 150 & 335 um - Green, Durbin, & Madin
Zooplankton by taxa - Durbin & Madin
Ichthyoplankton - eggs & larvae for cod & haddock - Mountain/Green/Berrien
High frequency acoustics and ADCP backscatter - Wiebe & Flagg
SST (AVHRR) - Bisagni
Along track t/s/f - Fisher
Met data - TBD
ADCP currents - Flagg
Drifters - Limeburner
Moored measurements - Irish
It is hoped that the availability of brief summaries and key time series plots will help identify relationships, events and/or questions that could be topics of Phase IV analyses.
Finally, to help compare the biological observations, it will be helpful to use a common set of units for reporting the various parameters. The following list of suggested units was developed:
by area by volume
Zooplankton #/m2 #/m3
Macrozooplankton #/m2 #/1000m3
Ichtyoplankton #/10m2 #/100m3
Nanoplankton #/m2 #/ml
#/l
ugC/l
Microplankton #/m2 #/l
ugC/l
Chlorophyll mgChl/m2
Fluorescence volts
The workshop adjourned at 2:00 on Thursday, July 20, 2000.
Appendix A
Draft Agenda
GLOBEC Broad-Scale Data Coordination Workshop
July 19-20, 2000
Carriage House, WHOI
Wednesday, July 19
9:00 Introduction, logistics, review agenda
9:15 Research questions for Broad-Scale data
Input from all participants
10:15 Database tools - R. Groman
GLOBEC/JGOFS & DODS
11:00 Combining data sets -
a) Experience of working with BrSc data - John Quinlan
b) Which data sets will be involved?
Noon Lunch
1:00 Combining data sets
c) methods for/experience with interpolation/integration
OAX, OAX-fm, Kriging, other
2:00 Modeling considerations - Dennis McGillicuddy
2:30 Discussion (one or many groups)
How to address the original research questions?
5: 00 Adjourn
Thursday, July 20
9:00 Phase IV preparations
synthesis topics and groups
Noon Lunch
1:00 continue discussions (for the hardy, dedicated PI's) - 'til you drop.
Appendix B
Research Questions Submitted by Participants before the Workshop
Bucklin:
How do our analyses of species' distribution and abundance compare to other Bank-wide biological and physical data sets? What are the interesting and important points for integrated analysis and synthesis?
I am interested in summaries of the meaning (i.e., how have the data been interpreted?) of all data sets. In some cases, I do not know the questions being asked by the owners of the data. I am looking more for shared questions than for answers.
Bisagni:
1) What are the number, transports and timing of Scotian Shelf Water "cross-overs" to Georges Bank from the Scotian Shelf? What are the chemical and biological implications of these "cross-overs", i.e., nutrients, zooplankton and fish eggs & larvae? Is there strong interannual variability (IAV) in the chemical and biological makeup of these "cross-overs"? (Use satellite SST, broad-scale hydrographic, plankton and nutrient data.)
2) What is the IAV of potential "new" primary production and secondary production over Georges Bank? (Use satellite SST and broad-scale hydrographic and plankton data.)
Durbin
Questions: Below are a number of very general questions.
1. What are the spatial and vertical distributions of zooplankton? Are there day-night differences in depth-distribution? Are there differences between species in their distribution patterns? What is the relationship of these distribution patterns to the physics? How do these patterns vary interannually? Included in this could be an analysis of expatriate species of zooplankton on the Bank as indicators of the presence of warm core ring water.
2. How are zooplankton being transported onto Georges Bank from the GOM and from Scotian Shelf waters? What is the relative significance of these different source regions and how does it vary temporally? How does the source affect the subsequent age structure and population dynamics of the populations on the Bank?
3. What is controlling the distribution patterns of the dominant copepods on the Bank. Advection, in situ production and mortality (predation)?
In addition there are many more specific questions which would combine survey and process cruise data and models. For example, Charlie Miller's individual-based model could be used to investigate a number of questions which have come from our initial analysis of Calanus survey data. We might also wish to try to hindcast Calanus distribution seen on survey cruises with improved models.
The survey and process cruise data could be used to come up with improved secondary production rate estimates of the major zooplankters on Georges Bank. This will require Broad-Scale survey data, and estimates of in situ growth and biomass.
There is a need to develop models for Oithona and Pseudocalanus that incorporate food-limited growth. With Barbara Niehoff, we are working to come up with a reproductive index for Pseudocalanus. Our plan is to use this with selected samples to look at whether there are differences in biological responses of the two species and to compare their responses to those observed for Calanus. This would be a most interesting exercise for the April-May time period when we have evidence of food-limitation of growth and reproduction of Calanus.
Miller and Crain
Ideas for data coordination we may be interested in:
1) A study of patterns of individual and spatial/environmental variables in C. finmarchicus C5's from 1998 Broad-Scale cruises.
Jen has been putting this data set together.
The idea behind this is to see what different "life choices" the individuals may be making in response to environmental conditions. The data are only from 1998, but might dovetail nicely with a multi-year study of lipid volume versus environmental conditions (see idea number 2).
Individual variables Data source
Prosome length Miller and Crain
Lipid volume Miller and Crain
Gonad type Miller and Crain
Gonad development stage Miller and Crain
Jaw phase Miller and Crain
Environmental variables Data source
Latitude event logs
Longitude event logs
Station Group event logs
Depth event logs
Julian Day event logs
Season event logs
Time event logs
Day/night event logs and navy website
Temperature D. Mountain
Salinity D. Mountain
Chlorophyll D. Townsend
Population dynamics T. Durbin
2) A study of patterns of individual lipid volumes in C. finmarchicus C5's from 1996-1998 Broad-Scale cruises compared with spatial and environmental variables.
Charlie has been working on this data set.
The idea behind this is to look at seasonal and multiyear patterns of oil sac volumes with an eye toward spatial and temporal patterns and environmental and community variables. This project might dovetail nicely with idea number 1 above.
Lipid Volumes Data source
1996 Miller and Crain
1997 Miller and Crain
1998 Miller and Crain
Spatial variables Data source
Latitude event logs
Longitude event logs
Station Group event logs
Depth event logs
Temporal variables Data source
Year event logs
Julian Day event logs
Season event logs
Time event logs
Day/night event logs and navy website
Environmental variables Data source
Temperature D. Mountain
Salinity D. Mountain
Chlorophyll D. Townsend
Biological community variables Data source
Population dynamics T. Durbin
3) With Ted Durbin as the lead PI and Melissa Wagner designated as the postdoc:
A comparison between oil sac volume and RNA:DNA ratio and protein content in C. finmarchicus C5's from 1997 Broad-Scale cruises. This project is in a planning phase.
Wishner:
I would like to be able to fit the process results into the larger context (spatial and temporal) of zooplankton (especially Calanus) distributions, hydrography, and currents. How representative are our detailed studies of conditions at those locations? What upstream and downstream distributions are likely during our sampling times? Scientific questions I am dealing with include: downstream effects especially from the northeast peak-cold plume to the south flank (1997), effects of detrainment into the Slope Water along the south flank and Slope Water-cold plume input in the northeast (1997), and cross-frontal exchange on the northeast peak and south flank (1999). Also, our vertically-stratified day and night sampling from the process cruises might help interpret some of the broad-scale results. How should we proceed to mesh these data? Also, it might be good to have a discussion of the protocol of co-authorship with regard to the broad-scale data.
NMFS Broad-Scale (Mountain, Taylor, Green, Berrien, Johnson, Sibunka):
What are the physical and biological factors contributing to the spatial and temporal variability in gadid egg survival, larval growth, and larval mortality?
What are the atmospheric and circulation factors determining the spatial and temporal variability in Bank-wide water properties?
What are the physical and biological factors contributing to the spatial and temporal variability in fluorescence and chlorophyll/florescence ratio?
Appendix C
Listings of the Data Sets Provided by the Participants
Bucklin:
Summary: At UNH there are archived alcohol-perserved samples fromall Broad-Scale cruises, including the MOC-01 Net 5 (downhaul) and splits of MOC-01 Nets 1 - 4.
We have data on relative abundances of Pseudocalanus newmani and P. moultoni (converted to absolute abundances using URI counts) for all priority 1 and 2 Broad-Scale stations (N=~20) for 1997 (depths 40 - 15 and 15 - 0) and 1999 (depth 40 - 15), and are now working on 1998 (depth 40 - 15). We will also analyzed some samples from the June 1999 process cruise by Karen Wishner et al, to examine cross-shelf transport.
Bisagni:
1) Pathfinder-processed AVHRR SST images and SST frontal images, 512 X 512 pixel, 1.4 km-resolution, GLOBEC "standard" domain (see Web-based documentation), 1985-1996 all passes (~2 images per day). (***NOTE: Not Available on the GLOBEC Website ***)
2) MCSST-processed AVHRR SST images, 512 X 512 pixel, 1.4 km-resolution, GLOBEC "standard" domain (see Web-based documentation), 1993-present all passes (~2 images per day).
3) MCSST-processed AVHRR SST images, 512 X 512 pixel, 1.0 km-resolution, GLOBEC "blowup" domain (see Web-based documentation), 1995, 1997 & 1999 field periods all passes (~2 images per day).
4) MCSST-based, cloud-free, optimally-interpolated daily AVHRR SST 512 X 512 images and 5-day averaged SST ASCII grids, ~7.0 km-resolution, reduced GLOBEC "standard" domain, 1993-1998.
5) SST frontal ASCII files (raster & geolocated), 1985-1996 (Pathfinder-derived) and 1999 field year (MCSST-derived), all passes.
Miller and Crain
1) Lipid volumes - Oil sac volumes from Calanus finmarchicus C5's from shallow and deep nets at an average of 7 stations around the cruise track on all Broad-Scale cruises for 1996, 1997 and 1998.
2) Prosome Lengths - from the above data set and a number of other stations and nets that were subsampled for other projects
3) Images - Calanus finmarchicus C5's from the above data set. These have been useful to other researchers for determining such things as projected area for calibrating Optical Plankton Counter data.
4) A catalogue of all Broad-Scale subsamples stored at Oregon State University
NMFS Broad-Scale:
All Broad-Scale survey cruises, at all or most stations:
Temperature, salinity, fluorescence profiles
Extracted chlorophyll (3 size fractions, 3 depths, Full stations in '95 & '96)
Plankton displacement volume (bongo and MOCNESS)
Gadid egg abundance (by stage)
Gadid larval abundance (by length and by age)
Wishner:
See attached spread sheet file.
Appendix D.
The mapping working group report (D. Chu, T. Durbin, D. Lynch, D. McGillicuddy, D. Mountain, J. Quinlan, and P. Wiebe).
The issues were grids to map to and the mapping method.
A good portion of the mapping group got together this afternoon to discuss the issues that came up with respect to 1) a standard grid to use when creating contoured plots of broad-scale data and 2) which mapping routine to use in making the contour plots. Associated with the latter is the issue of how to create unbiased estimates of property values for the Bank as a whole or sub-regions. John Quinlan strongly recommended that we use the "MARMAP-2" grid which has been used by Dennis and Ann for the Pseudocalanus mapping and is now being used by John in his work with the fish larvae. But there was some desire on the part of others to have a more restricted region, but one larger than the "Bank 150" grid so that all of the broad-scale stations were within its boundaries. We came up with a non-rectangular box that we thought would be useful for most of the broad-scale data sets, but would not have too many far-field data points. The region we came up with is defined by the following turning points.
1) 41 45'N; -69 30'W
2) 42 30'N; -67 45'W
3) 42 30'N; -65 30'W
4) 41 30'N; -65 30'W
5) 40 00'N; -67 15'W
6) 40 00'N; -69 30'W
The idea is that we would use the finite element grid points that fall within the border defined by the turning points to fit kriged, OAXed, or Oafed data to. Dan Lynch was asked in a subsequent email to look at the region so defined and see if it is compatible with any flow field considerations that he might have? He was asked to make up the grid for the program based on a selection of points from the larger Quoddy finite element grid. Chu said it would not be a big job to incorporate this grid and the MARMAP2 grid into Easy-Krig 2.0 so that one could choose to make a regular rectangular grid as is now done, or choose to use one or the other of the finite element grids. John Quinlan said he would provide the MARMAP-2 file to D. Chu and others. It is intended that the file with the "broad-scale base grid" go to D. Chu and R. Groman for posting/linking on the GLOBEC web site.
The other portion of the discussion concerned what the best procedure might be to obtain unbiased estimates of property values for the Bank. To address this, D. Mountain will be redo the experiments he described to the workshop yesterday and expand them in several ways. The results are intended to provide insight into how best to get the desired estimates and which mapping routine to use.