10th International Aerosol Conference September 2 - September 7, 2018 America's Center Convention Complex St. Louis, Missouri, USA
Abstract View
PM2.5 Source Apportionment Using a Hybrid Environmental Receptor Model
LUNG-WEN ANTONY CHEN, Junji Cao, University of Nevada, Las Vegas
Abstract Number: 900 Working Group: Source Apportionment
Abstract Receptor model is an important tool for air quality management. Since none of the modeling approaches is without biases or uncertainties, a weight-of-evidence (WOE) approach that takes into account multiple model results is strongly recommended in practice. This paper introduces the hybrid environmental receptor model (HERM) that can perform EV-CMB and PMF, two most popular receptor models for PM2.5 source apportionment, using a unified algorithm and evaluates it with simulated and real-world datasets. The current HERM software is capable of 1) conducting EV-CMB analysis for multiple samples in a single iteration; 2) calculating EV-CMB and PMF source contributions as well as middle grounds (hybrid mode) between the two; 3) reporting source contribution uncertainties and sample-/species-specific fitting performance measures; 4) interfacing with MS Excel® for convenient data inputs/outputs and analysis. HERM allows a hybrid mode that takes partial source information such as incomplete source profiles available for the region of study to pursue a middle ground between EV-CMB and PMF. This is particularly useful since the inclusion of only reliable source profiles in the model avoids poor fitting in EV-CMB while decreasing the rotational degree of freedom in PMF analysis. HERM implements the constraints differently from the PMF software (e.g., EPA PMF5.0) in that it uses source profile uncertainties explicitly in the effective variance fitting. Initial testing with simulated and real-world PM2.5 datasets show that HERM reproduces exact EV-CMB results from existing software (EPA CMB8.2) but with more tolerance to collinearity and better uncertainty estimates. It also shows that partial source information helps reduce rotational ambiguity in PMF, thus producing more accurate partitioning between highly correlated sources. Moreover, source profiles generated from the hybrid mode can be more representative of the study region than those acquired from other regions or calculated by PMF with no source information. When practicing receptor modeling, users are recommended to first determine the possible number(s) of sources by examining the dependence of fitting performance (χ2) on source number. HERM in different modes (EV-CMB, hybrid, and PMF) should be carried out with their results compared and reconciled to support the WOE approach of source apportionment.