10th International Aerosol Conference September 2 - September 7, 2018 America's Center Convention Complex St. Louis, Missouri, USA
Abstract View
Constraining Aerosol Processes with a Stability-Based Inverse Model
DANA MCGUFFIN, Peter Adams, Erik B. Ydstie, Carnegie Mellon University
Abstract Number: 181 Working Group: Aerosol Modeling
Abstract The general dynamic equation predicts the evolution of aerosol size distributions with terms that represent emission, nucleation, condensation, coagulation, and deposition of particles. This work aims to constrain key uncertain terms in the general dynamic equation using measured size distributions. Coagulation of particles is calculated based on their Brownian motion and the loss rate includes deposition to the ground and advection. However emission, nucleation, and condensation of volatile organic compounds (VOCs) are highly uncertain. Primary aerosol emissions are difficult to monitor and calculate due to the variability in emission fluxes and intensity levels within a sector. The nucleation of particles from molecular clusters of compounds possibly including sulfuric acid, water, amines, and organic compounds is still an active area of research. Secondary organic aerosols (SOA) are generated as VOCs condense to the particle phase or on pre-existing aerosol, but there are large uncertainties in the identity of VOC precursors and the condensation mechanism as well as in the SOA budget. The goal of this work is to use measured aerosol size distributions to gain knowledge on three uncertain aerosol processes: primary organic aerosol emissions, nucleation rate, and SOA production. Such measurements of aerosol mass and number size distributions implicitly contain a great deal of information about these microphysical processes, and they are available at several ground sites. We integrate size distribution measurements and an aerosol microphysics model using nonlinear process control theory. Measured size distributions are generally used to validate models through sensitivity analyses in forward models. In these cases, tuning factors or model parameters are systematically adjusted until the model reproduces a measured distribution of aerosols. However, a more robust way to do this is to generate and run an inverse model.
In general, inverse models are used to constrain uncertain processes based on a set of measurements. Inverse models usually minimize a cost function related to the sum of squared error in the set of observations by linearizing the forward model. However, inverse modeling of aerosol microphysics is challenging since it involves nonlinearities that span large spatial and temporal scales. This work develops a new, stability-based inverse modeling technique that is not computationally intensive and rigorously accounts for nonlinear dynamics.
Here, we present the development of the inverse modeling technique and preliminary applications to a box model. We use a zero-dimensional version of the TOMAS microphysical code, which models size-resolved aerosols with the discretized general dynamic equation. We generate synthetic measurements by running TOMAS with a set of “true” rates for primary emissions, nucleation, and SOA production. Then, we start a TOMAS simulation with a set of process rates biased from their “true” value. We will see if the developed inverse modelling technique will estimate the “true” process rates based on the synthetic measurements. The methodology developed here will facilitate future work that estimates primary aerosol emissions, nucleation, and SOA production in a global model based on a measurement network.