Fourier Transform Infrared (FT-IR) Organic and Elemental Carbon (OC & EC) Measurements in the Chemical Speciation Network (CSN)

ANAHITA AMIRI-FARAHANI, Andrew Weakley, Bruno Debus, Satoshi Takahama, Ann Dillner, University of California Davis

     Abstract Number: 551
     Working Group: Instrumentation and Methods

Abstract
The CSN is a component of the National PM2.5 Monitoring Network, whose data is used for characterizing annual and seasonal spatial variation of aerosol and comparison to chemical speciation data collect from other networks. Representing OC and EC using FT-IR method at CSN sites improves the cost, efficiency throughout the network. It is simpler in sample handling and thermal optical reflectance laboratory analysis and less sampler maintenance is needed.

Previous work suggests that CSN sites may require partitioning into two distinct groups of atypical and typical EC for accurate calibration. Evidence in aerosol collected from some sites like Elizabeth, NJ in 2013 indicated elevated concentrations of diesel particulate matter relative to samples collected from the other sites. When samples were pooled together for calibration biased predictions were observed at these sites. A two-calibration solution was devised, one for the atypical sites and the other for the typical sites.

In this work , our first objective is to identify and separate atypical and typical samples using a year of CSN samples to capture seasonal variability. Using a global calibration we designate a site as atypical if, on average, FT-IR predictions errors are much larger than those at the other sites. Our second objective is to select typical and atypical sites to use for calibration. Atypical sites are partitioned using hierarchical cluster analysis (HCA) as the basis for their atypical behavior is unknown. HCA partitions atypical sites using median relative percent bias as well as the ratio of the median pyrolysis (char) artifact measured during thermal analysis (i.e., OP) to EC ratio and median OP/OC. Our third objective is to measure FT-IR OC and EC for all samples. Calibrations are then developed for each cluster determined by HCA. As typical sites are reasonably similar in terms of carbon composition no cluster-modeling is required. Following cluster-modeling, select CSN sites are selected to represent the 140 sites considered in this study. Sites designated as National CORE Multi-pollutant Monitoring Station (NCORE) and Speciation Trends Network (STN) are prioritized. Typical sites are selected for calibration following the application of HCA to samples on six speciation measurements OC/ NH4+, EC/ NH4+, OP, NH4, NO3-, and SO42-. Overall, only 15 (8 atypical + 7 typical) sites appear to represent, to a reasonable degree, the composition of aerosol carbon in the CSN. The impact of COVID on identifying typical versus atypical sites and on prediction capabilities will be investigated.