EPA SON Precipitation Chemistry Data

The following discussion was taken from personal correspondence from Luther Smith, Ph.D., to Richard Artz, NOAA Air Resources Laboratory, Silver Spring, MD, 23 December 1997.

Monitoring Sites

Precipitation samples were collected at the following sites:

Data Analysis and Processing

Approach

The EPA State-Operated Network (SON) data were treated in a manner as close to NADP data as possible. Though the data were at least nominally similar to NADP, the network had a somewhat different protocol and provided a somewhat different data set than NADP. The process of data "clean-up" included eliminating duplicate observations, checking observations for suspicious or erroneous information, converting codes to "NADP-like" form, and checking agreement between the documentation and the data sets themselves. Where possible, incorrect entries in an observation were changed based on a consideration of all information in the observation and, if appropriate, observations near it in time from the same site. Changes involving the raw data were limited to correcting dates when buckets were either removed or placed on a collector.

Very little changing of chemistry values was done. One change was to invalidate data for sample periods longer than eight days. The only other changes to chemistry data were to convert ionic concentrations indicated to be below the detection limit to half the detection limit value. Detection limits were gleaned from information in one of the annual SON reports.

Quality assurance type codes or comments for each individual sample were obtained from the SON documentation. No data were changed based on such codes for the following reasons. The data contained instances of samples coded as contaminated (in one way or another) that had chemistry values missing and other such samples that had chemistry values present; such a scenario is consistent with NADP treatment of samples. In addition, the SON data sets contained codes that were clearly in error, based on the provided documentation, and included some cases with chemistry values where the documentation said the sample should have been voided. In such cases, it was not clear if codes had been mistyped or if the documentation did not completely describe how such coding was treated.

Another point relating to quality assurance is the following. The SON network data came with dates indicating when samples arrived at the laboratory. If such dates (and the corresponding bucket on and off dates) were correct, it appeared that some sites were holding some samples for a relatively lengthy period before shipping them. However, there were also instances of some clearly impossible "lab arrival dates"; for example, a lab arrival date equal to the date the bucket was placed on the collector. Many of the "odd" lab arrival dates were suspected of being data entry errors, and again, no chemistry data were changed based on lab arrival dates. However, these dates are included in the weekly database for the SON data.

Results

The following data files are available for the SON network: a site file including latitudes and longitudes; weekly data file; file of annual precipitation-weighted means and deposition totals; file of quarterly precipitation-weighted means and deposition totals. Latitudes and longitudes were supplied in network documentation for SON, and these values were checked against a road atlas. The first quarter was defined as January-March, the second as April-June, and so on.

The standard NADP data completeness criteria were applied to the summary periods, and if an ion failed to meet any of them, it was left missing in the summary files. The summary files contain data completeness measures. (Note that since the SON did not add water to samples for chemical analysis, but simply measured until no sample was left, data completeness criteria were separately calculated for each ion.)

A note of caution is in order about some of the ions reported. The SON measured fluoride, bromide, and nitrite as well as the NADP suite of ions. In the SON data base, fluoride, bromide, nitrite, and phosphate were frequently reported as below the detection limit. In addition, a caveat appeared in some documentation from SON that indicated that the fluoride data were all suspect due to the chemical analysis methodology employed. Save for that mentioned above, no adjustment to the data for these ions was done.

The years and quarters meeting the data completeness criteria were somewhat sparser than one might have hoped. The results were particularly disappointing for West Virginia; no year and only one quarter at any site met the criteria there. The final results are summarized below.

The following tables summarize the years and quarters meeting the completeness criteria.

Comments

While it appears that the SON network is not of the same quality as NADP, it is believed that it may be useful to CADM. Although the EPA-SON network suffered from a lack of coordination, the database is documented well enough that one can locate the data gaps. The reported summary numbers are probably fairly good. Certainly the site locations are nicely placed for CADM purposes. The SON is no longer operating, so basically "What we have is all there is."

The SON data from West Virginia is not worth pursuing further. This would perhaps not be a terrible loss, since three of the sites were both urban and far removed from the Chesapeake Bay watershed.

Additional investigation was conducted to determine why sites failed to meet the data completeness criteria. There were several instances where the only criterion that was not met was the requirement that precipitation be monitored 90% of the time. If a rain gauge could be found that was near enough to each site, one could substitute the values from these gauges for some of the missing time periods. This approach holds out the possibility of "filling in" the following missing years in the above table:

• Lum's Pond, DE: 1990
• Elms, MD: 1985 and 1990
• Rocky Gap, MD: 1985, 1987, 1988, 1989, 1990, 1992, and 1994

If successful, one would then have reasonable deposition estimates (in addition to those from the other networks) for at least two and as many as nine years.

Go to the EPA SON Precipitation Chemistry Data Page