AAAR 32nd Annual Conference
September 30 - October 4, 2013
Oregon Convention Center
Portland, Oregon, USA
Abstract View
Probabilistic Modeling and Bayesian Updating of Concentrations of Carbon Monoxide and Fine Particulate Black Carbon in Fort Collins, Colorado for Exposure Estimation
DANIEL MENDOZA, Amy L. Stuart, Getachew Dagne, University of South Florida
Abstract Number: 382 Working Group: Health Related Aerosols
Abstract Estimates of commuter exposure to air pollution are necessary to inform urban design and public health policy. The lack of highly resolved spatial and temporal exposure estimation methods along with incomplete uncertainty estimates have been major hindrances toward obtaining representative exposure information.
We present the development and first stage of application of a modeling system that will ultimately estimate urban commuter exposures at high resolution with specific representation of uncertainty. We first generated emissions of select traffic-related pollutants at a high spatiotemporal resolution for the study area of Fort Collins, Colorado. Uncertainty and variability in dispersion model input data, including emissions, were also identified and tabulated. This information was used to define probabilistic distributions of inputs. A stochastic code was applied to sample the input distributions and perform ensemble dispersion simulations. The sampling strategy was designed to produce probabilistic estimates of spatial fields of concentration for a sample defined by hour of the day, day type, and season. Fields were produced at 500m x 500m spatial resolution for the 10 km x 10 km study region. Concentration results were compared to monitoring data for evaluation of the concentration estimates. A Bayesian updating procedure was subsequently developed to assimilate monitoring data into posterior estimate of concentrations. The next step of modeling system development will use the posterior concentrations to estimate route- and mode-specific exposures.
Prior estimates of spatial fields of ensemble mean concentration for two pollutants (carbon monoxide and fine particulate black carbon) are presented and discussed. Fields representing cumulative distribution statistics for these estimates highlight spatial locations and hours of high variability/uncertainty. Results of the graphical and statistical comparison of prior model estimates with data from stationary and personal monitoring provide evaluation information. Preliminary results of posterior concentration distributions produced by the Bayesian updating procedure are also provided.