Abstract View
Quantifying Linear and Non-Linear Influences of Aerosol Precursor Emissions on Pollutant Concentrations Using CMAQ-hyd
JIACHEN LIU, Eric Chen, Ryan Russell, Shannon Capps, Drexel University
Abstract Number: 566
Working Group: Source Apportionment
Abstract
Chemical transport models (CTMs) are essential assets to understand complex physico-chemical reactions in the atmosphere. They provide estimates of concentrations of air pollutants based on emissions and meteorological parameters. Sensitivity analysis in CTMs has helped researchers determine the uncertainty in the CTMs and make policy recommendations. The simplest method to compute sensitivity coefficients is the finite difference method (FDM). The sensitivity coefficients are calculated by running the model multiple times with incremental values for the input variable of interest. However, this method suffers from truncation (i.e., ignoring higher-order terms) and cancellation errors (i.e., numerical issues caused by subtracting two very close numbers). The truncation error can be minimized by taking smaller perturbation step, thus eliminating the impact of higher-order sensitivity terms on the first-order result. However, the cancellation error will dominate if the perturbation step is too small. Other methods, including the direct decoupled method (DDM) and the adjoint method, involve formulating new forward sensitivity or adjoint equations to the CTM. When the CTM is updated, new equations must be modified manually, thus reducing their applicability in various complex CTMs. Here, we propose an alternate approach for sensitivity coefficient calculations in CTMs: the hyperdual number approach (Fike et al., 2011). Instead of a real number perturbation in the FDM, we applied a hyperdual perturbation to the emission variable of interest. The method is more accurate and does not depend on the perturbation step size. The method is also easier to implement compared to the DDM and the adjoint method. We applied this method in the Community Multiscale Air Quality (CMAQ) model v.5.3 to formulate the CMAQ-hyd model. We have calculated the exact first- and second-order sensitivities of aerosol and gas concentrations to select aerosol precursor emissions based on this new method.