It is currently anticipated that the spectrum tapes, data base disks, and hard copy spectrum printouts will be
archived somewhere in Las Vegas.
The in situ samples are taken on a regular grid, at stakes identified by the four digits of their coordinates
north and east relative to the Oscar triangulation platform*. The full coordinate is not necessary because
knowing the island gives the first two digits. Americium-241 concentrations and counting errors will be
stored on tape in arrays so that relative matrix positions are identical to relative ground positions. Matrix
positions beyond the edge of an island will be set to zero, and there will always be at least one array row or
column beyond any edge. If the data array must be broken into subsets to meet the estimation routine's
limitation of 400 data points, the entire array will also be available in a single file for reference. When the
Pu/Am ratio has been established (see "Soil Sample Data Procedures"), similar arrays of Pu concentrations
will be calculated and stored.
The data matrix is used by the Gamma and Gamtst programs(files §, 1) to plot the raw semi-variogram and
test model fits. If a drift is present, GenCov (File 2) fits the generalized covariance. The model chosen
should fit the raw variogram reasonably well, and should make sense in light of the support of the data, the
sampling method, and previous experience.
Printouts of the raw data and numerical results of model fitting, along with plots of the raw variogram and
the fitted model, will be maintained in the results notebook. Written comments on the data and the model
will appear in the daily log.
The covariance structure will be used to make kriging estimates of average Am and Pu concentrations
(KrigIn, 16Krig, 9Krig on disk; tape files 2, 3, and 22). Estimates and standard deviations of kriging errors
will be stored on tape for averages over two different areas. Printed outputs of the estimates, 1/20 upper
bounds (16prt, 9prt on disk; tape files 6, 14), and contour intensity plots (CnfBnd; file 16) will be reviewed by
the Technical Advisor, EG&G and Eberline, then submitted to the ERSP Manager and JTG for action.
Copies will also be kept in the result notebook.
When excavated and adjacent areas are resurveyed, the new data will be placed in the proper matrix position
and stored on tape without altering the original data (i.e., in a new file). New estimates of averages and
errors will be made and stored, and the printouts submitted for review and action.
Upon completion of cleanup for an island, a certification run will be made to estimate residual
concentrations on the entire island with the most current data. The printed outputs will be prominently
marked "Certification Estimates"**,
I.

Soil Sample Data Procedures

The Eberline laboratory will store the soil sample results on magnetic tape in the form of two descriptive
strings and a 2048 word integer spectrum array for each data point, The data can be stored directly on disk
from tape (EICDB1 on disk; tape file 25), except for coded quality assurance samples, which require manual
input to decode. The results for a data point will be stored logically as strings on a single 21-record file.

Procedures for file names, disk labels and cataloging are similar to the in situ data base; details are in the

soil sample data base program documentation, along with exact data format. The program documentation
also includes examples of data retrieval. Update and correction procedures are the same as the in situ data
base***,
The physical soil samples consist of two six-sample composites from each selected in situ survey location.
The randomly-oriented pattern samples the field of view of the detector with a density approximately
corresponding to the weighting function of the detector geometry.****

*Stake locations and identifiers followed the grid numbering systems established by the surveyors.
Attempts to tie in to the Oscar system failed.
**Certification Estimates, as such, were not produced. However, final data maps were produced for
islands from which soil was removed.
***EIC has extensively modified the procedure described. Details may be found in Chapter 4.
****The pattern was based on misinformation about detector response; as a result it does not correspond

even roughly to the correct weighting function.

A-3-2

Select target paragraph3