From 1984 to 1995, NAPAP researchers exposed limestone and marble briquettes to
weathering for months to years at several different sites. They sampled the
briquettes and analyzed multiple layers for anion content (sulfate SO4, nitrate NO3,
chlorine Cl, and fluorine F). In the first four years, they analyzed a subset of
samples for fifteen cations (Mg, Al, Cd, Mn, Ba, Be, Ni, Pb, Co, Cr, Cu, Fe, Sr. V.
and Zn). They compiled the results into several BSC (briquette surface chemistry)
data files and documented the data—field formats thereof both in table headers and in
separate files.
In July of 1998 I received nineteen BSC files as email attachments: three format
files (and two duplicates) describing the fields in the data files, twelve anion data
files (four material types times three project cycles) , and two cation data files
(from the first cycle, and which were re-sent separately in September). In examining
these files, I have verified and extended their format descriptions, transformed all
to a common physical format, combined the fourteen date files into two master files,
slightly modified (recoded) and reordered identification data, and done preliminary
analyses of the measured data.
For this study, the outdoor treatment units are the briquettes while the laboratory
analytical. Units are the layers sampled from each briquette. The data files reflect
this hierarchical structure. Each line, representing a layer, has a briquette and
layer id; treatment fields describing the rock type and condition, exposure site and
rack slot, and exposure period; and analytical fields giving values for either the
four anions or fifteen cations and indicating which are below detection limits. The
briquette treatment fields are duplicated for each layer of a given briquette.
Proper statistical analysis must also reflect this two-level structure. This is
impeded in this study by the diversity of layers sampled for different briquettes.
(There are sixteen different patterns, not counting some of the control blocks ). On
the other hand, there are essentially no missing data. Overall, the data are in good
shape for statistical analysis after the few changes I made. The disk accompanying
this report contains the ready-to-analyze anion and cation files. There are three
versions to meet the differing input needs of different programs.
Lacking existing analyses to review, I performed some myself. The report text and
tables tabulate the briquette treatment variables and the below-detection indicators
for layers for both the anion and cation files. Histograms, plots, and analyses of
covariance show the following about the overall relationship between treatment and
anion content: rock type (limestone versus marble), exposure time, and layer selected
all affect each of the anions; condition (fresh versus weathered) affects SO4 and
Fluorine.
There are two directions to go for further analyses. One is to examine Subsets of
data to answer specific questions. The other is to augment the current data with
other information.