Absorption photometry of patterned deposits on IMPROVE PTFE filters

The IMPROVE program (Interagency Monitoring of PROtected Visual Environments) tracks long-term trends in the composition and optics of regional haze aerosols in the United States. The absorptance of red (633-nm) light is monitored by filter photometry of 24 h-integrated samples of fine particulate matter (PM_2.5). These are collected every third day at about 150 rural and often remote locations. Systematic reanalyses of archived samples have established the reproducibility of the optical absorption measurements across decades and rebuilt instrument systems, with a consistent calibration that is traceable back to 2003. IMPROVE samples for nondestructive sequential analyses by photometry, gravimetry, and X-ray fluorescence are all collected on ring-mounted membranes of expanded PTFE (polytetrafluoroethylene). Although chemically inert, low in blank mass, and optically thin, these collection media yield visibly nonuniform deposits that do not admit direct interpretation according to a naïve "Beer-Lambert" formulation of optical absorption. Most IMPROVE PTFE deposits exhibit fine-scale "pixelation," a pattern shaped by the perforated metal screen that supports the membrane during sample collection. This paper extends the traditional Beer-Lambert interpretive model to accommodate the patterned deposits generated on PTFE filters by IMPROVE and similar sampling protocols.<i>Implications</i>: The ambient absorption coefficients historically reported from IMPROVE filter samples are based on the conventional Beer-Lambert interpretation of measured transmittance by filter deposits assumed to be uniform. The neglect of observed patterning in actual deposits yields predictable underestimates for the summed absorption cross-sections of the collected particles. This pixelation bias can be substantial for individual samples that are taken from elevated concentrations of absorbing aerosols, near wildfires or urban centers. Over the regional-haze-tracking scale of the predominantly rural IMPROVE network, the aggregate bias relevant to radiative forcing is more moderate, estimated at roughly 10%.