Volatile Organics Data Handling
Once the batch analysis is complete, the calibration runs are checked for completeness and linearity. The chemist then builds the calibration table. The sample concentrations are then calculated from the table to determine the concentration of VOCs in the samples. The calibration and sample reports are archived on a 250 mb "Zip" disk and periodically the complete file from raw data to report is burned on a CD.
The .CSV reports are dissected by a "Visual Basic" program in the data managers PC. The program transposes the file to make it compatible with an "ACCESS" database. A worksheet is setup and maintained for each site. The total year is kept in this format and quarterly data tables are produced from this. The quarterly tabulation contains the tables for each target compound for every site in the network.
The "Visual Basic" program VOCdat written by Sonoma Technology is used to visualize, summarize statistics and format the data for the EPA AQS database. It has proved invaluable for the PAMS data submission and is equally valuable to see "anomalies" in the toxic data for confirmation by the analytical chemist.
Samples where a compound is not detected by the analytical system are handled differently than required by the EPA for the previous studies. Now a value of zero will be submitted for any non-detects. Error codes are submitted when sampling or analytical errors occur and no value is submitted. Data analysis researchers may use or replace the zero values as needed for various analysis tasks. Currently only the 2001 and 2002 data has been submitted with zeroes for non-detect. The 1999 and 2000 will be resubmitted with zeroes in the future.
All data from 1990 to 1998 had non-detects replaced with a value of 1/2 the MDL* for that compound. This procedure causes some concern where the AGC* is lower than the MDL. Annual averages then appear higher than the AGC (and higher than what the actual value may be), when in fact the compound is not detected in many samples. This data is flagged by shading and a footnote on the data value in annual average tables (pages 17 to 20) to indicate that the user must apply judgment in interpreting the data. Methods for more stringent statistical analysis of databases with significant non-detects is beyond the scope of this report, but may be found in Rao et al., "Analysis of Toxic Air Contaminant Data Containing Concentrations Below the Limit of Detection."**
A spreadsheet file is used for comparing replicate samples and blanks for the Quality Assurance (QA) report. The summary of the data is appended to the QA section(7).
* See Table III
** Journal of the Air and Waste Management Association Vol. 41, No. 4, April 1991