Hyper-local source strength retrieval and apportionment of black carbon in an urban area

Abstract

Neighborhood-scale air pollution hotspots have recently been identified through detailed field campaigns, including the 100x100 Black Carbon Experiment which took place in West Oakland, CA, in 2017. Here, high-resolution nested atmospheric simulations are used together with a Bayesian inversion framework to estimate source apportionment at the hyper-local scale for a neighborhood in West Oakland. Forward simulations are performed with the Weather Research and Forecasting (WRF) model using 6 grid nests from 11.25 km to 2 m horizontal resolution. On the finest grid, building geometries are resolved using the immersed boundary method. Seven point sources and four line sources at known locations are included in the forward simulation for two 1-h periods during the 2017 field campaign. Data from 12 black carbon sensors are used to perform source inversion using a Markov Chain Monte Carlo approach, which provides a probability distribution for each of the 11 source strengths. From this, a most-likely plume can be created using the peaks of the distributions, and source apportionment can be estimated for each sensor. In addition, a composite plume can be constructed to indicate 90% confidence that concentrations are above or below a specified value. With this probabilistic analysis, it is possible to determine that more than half of the neighborhood has black carbon concentrations of higher than 0.4 μg/m3, with some areas higher than 3 μg/m3 during the time periods studied.