The
Grid Group met on Friday, December 13, 2002, in Room 204 Redfield at WHOI. Attending were Jeff Runge, Dennis
McGillicuddy, Peter Wiebe, Dezhang Chu, and David Mountain. The primary charge to the group is to
recommend a consistent procedure for gridding the various data sets in the
Broad Scale component of the program.
This requires deciding upon 1) the area (or box) to be gridded, 2) the
distribution of the grid points within the box, 3) the interpolation scheme to
estimate values at the grid points and 4) the data sets to be used in the
gridding exercise.
The
strawman box suggested at the October workshop was felt, at that time, to
extend too far beyond the actual Broad Scale station locations. David Mountain was charged with recommending
a smaller box and did so for the Grid Group meeting. The discussion reflected two desires. The first to keep the box limits near the data points so that any
plots or calculations would represent areas of meaningful data. The second was to include within the box
areas that might be of interest to the modelers, such as likely nearby source
or fate regions for variables measured on the Bank (since the data on the Bank provide information concerning of the
conditions in those upstream or downstream areas at a different time). In the second case areas where little or no
information actually exists could be eliminated (or not displayed) on the basis
of the error estimate from the interpolation scheme. After much discussion, the first approach won consensus. The smaller box submitted by David Mountain
was felt to be acceptable for the Broad Scale needs. While some modest changes were considered, in the end, the
feeling was to use the box as submitted.
In addition, it was decided that a wider area grid that included all of
Georges Bank and the Gulf of Maine, the MARMAP2 grid, would be appropriate for
the modeling interests and that copies of all of the data sets use to make
gridded products should be retained for later use in any modeling
investigations. The gridding of
selected data sets to the MARMAP2 area would be a simple task, that could be
done when needed.
At
the October workshop the value of using the finite element grid points or a
regularly space grid pattern was discussed.
That discussion continued in the Grid Group meeting. A consensus was reached that the regular
spaced grid was preferable. The finite
element grid spreads points further apart where the bottom is flat, on the
basis of the physics. However using a
regularly spaced grid was felt to offer a more consistent representation of the
biological fields, particularly when the gridding/interpolation is done on log
transformed data. The grid spacing for
the initial testing by the Group will be 0.05 degrees in latitude and
longitude. The format for the grid will be: grid point #, latitude, longitude,
area represented by point. This grid
will be referred to as the Georges Bank Broad-Scale Standard Grid.
Two
primary methods of interpolation are to be tested B krigging and
objective analysis. The krigging will
be done using easykrig developed by D. Chu.
The objective analysis will be done using OAX developed by C. Hannah. Some estimate of the spatial statistics for
the variable being interpolated is needed B in the form of a variogram for
easykrig and a correlation length scale ellipse for OAX. The standard OAX package includes the
correlation values appropriate for the physics within the grid node file. To develop the needed variogram(s) for each
parameter, the associated investigator will meet with D. Chu to use easykrig
for determining the appropriate variogram.
Variables:
The
list of variables to use in the grid testing includes
A physical variable - surface salinity
(D. Mountain)
A phytoplankton variable - chlorophyll (D.
Townsend)
A plankton variable- biovolume (P. Wiebe
through J. Green)
A zooplankton variable - Calanus egg
production rate (J. Runge)
A fish variable - larval cod (D.
Mountain)
A nutrient variable - to be determined
(from D. Townsend)
An along track variable - to be determined
(P. Wiebe)
A
number of test sets (i.e., cruises) for each of these variables will be created
in a format: latitude, longitude,
value. For each data set and for each
variable, the output of the interpolation will be a file containing: gird point #, latitude, longitude, value,
value*area for the grid point, error.
Once
the test data sets are created and the appropriate variograms are developed,
the Grid Group will meet again to decide how (who, when, Y) the gridding will
be done and how the results will be reviewed and evaluated. The data sets both original and mapped will
be stored on the GLOBEC database.
The Grid data
The Grid picture