ET03-2006: Detection and quantification of asymmetric peaks

One of the goals of a recent U.S. Environmental Protection Agency air pollution study was to relate the participants’ activities to their exposures to particulate matter (PM) during their routine activities. PM concentration data were collected continuously with a real-time monitor, downloaded daily in CSV format and concatenated using AWK. Human activity diary information was collected via pen and paper and recovered using scannable decoding software (TeleForm). All data were converted into SAS. The asymmetric response peaks of the particulate matter concentrations made systematic detection and analysis of the area under their respective curves problematic. To overcome this problem, a program was written to locate each peak and generate an integrated value for its response area. Data were smoothed using a running average, inflection points were identified relative to the start and stoppage of each peak, and a moving baseline calculated. Incorporation of human activity descriptors allowed for the quantification of human activity pattern influence upon potential exposures to particulate matter under real-world conditions. INTRODUCTION Many daily activities affect the level of particulate matter (PM) in the air. These activities include cleaning (dusting, sweeping and vacuuming), cooking and personal hygiene (using spray products like hair spray, deodorant and perfume). Increased levels of PM have been associated with various health effects including cardio-pulmonary and respiratory diseases. We needed to determine the exposure level of this pollutant and associate that level with daily activities of a subpopulation of cardiovascular-impaired residents of central North Carolina. PM mass concentrations are measured on a semi-continuous scale. One means of estimating PM mass concentrations is determining the area under the curve using a continuous monitor. Graphing the data over time, it is easy to locate major PM contributions to an individual's personal exposure by noting the "peaks" generated by the monitor. Graphically, these can be easily seen. Locating them programmatically is a different matter. Most programs written to locate peaks assume a symmetric peak. The beginning and ending of each peak are indicated by a change in the sign of the slope of the tangent line to the curve. These points of change are commonly known as inflection points. Once the slope of the tangent line to the curve is zero, the top of the peak is assumed. The area under the front and rear portions of the halves under the curve are assumed to be equal. In PM mass concentrations, an exponential decline of the rear half of the peak is often observed, the results from some integration programs greatly underestimated the true area under the curve. We will be discussing two elements in this paper. The first discussion point is the conversion of the data from multiple sources into SAS. The data were collected in various methods including paper questionnaires, computer assisted technician interview, and electronic transfer from the monitoring instrument. The second element is the program that was written to locate the peaks in the data. DATA TRANSFERS The first and often the most arduous step of any analysis is getting the data into the proper structure and format. Since various methods of data capture were used, each type of data will be examined separately. PM DATA Exposure monitoring data were electronically downloaded daily from the monitoring instrument in a comma separated value format (CSV). Thus one file per day per participant was generated. A sample of the data follows: pDR-1000, , , User ID: 09841S20010308 Tag Number: 01, , Number of logged points: 1438 Start time and date: 08:11:00 08-Mar Elapsed time: 23:58:00, Logging period (sec): 60, Calibration Factor (%): 100.0, Max Display Concentration: 0.371 mg/m3 Time at maximum: 11:23:03 Mar 08