Bayesian treatment of a chemical mass balance receptor model with multiplicative error structure


  1. Keats, A.
  2. Cheng, M-T.
  3. Yee, E.
  4. Lien, F-S.
Corporate Authors
Defence R&D Canada - Suffield, Ralston ALTA (CAN);Waterloo Univ, Waterloo ONT (CAN) Dept of Mechanical Engineering;National Chung Hsing Univ, Taichung (TAIWAN) Department of Environment Engineering
The chemical mass balance (CMB) receptor model is commonly used in source apportionment studies as a means for attributing measured airborne particulate matter (PM) to its constituent emission sources. Traditionally, error terms (e.g., measurement and source profile uncertainty) associated with the model have been treated in an additive sense. In this work, however, arguments are made for the assumption of multiplicative errors, and the effects of this assumption are realized in a Bayesian probabilistic formulation which incorporates a 'modified' receptor model. One practical, beneficial effect of the multiplicative error assumption is that it automatically precludes the possibility of negative source contributions, without requiring additional constraints on the problem. The present Bayesian treatment further differs from traditional approaches in that the source profiles are inferred alongside the source contributions. Existing knowledge regarding the source profiles is incorporated as prior information to be updated through the Bayesian inferential scheme. Hundreds of parameters are therefore present in the expression for the joint probability of the source contributions and profiles (the posterior probability density function, or PDF), whose domain is explored efficiently using the Hamiltonian Markov chain Monte Carlo method. The overall methodology is evaluated and results compared to the US Environmental Protection Agency's standard CMB model using a test case based on
Report Number
DRDC-SUFFIELD-SL-2008-040 — Scientific Literature
Date of publication
07 Oct 2009
Number of Pages
Reprinted from
Atmospheric Environment, vol 43, 2009, p 510-519
Document Image stored on Optical Disk

Permanent link

Document 1 of 1

Date modified: