Atmospheric Inverse Modeling with Passivity-Based Input Observers: Estimating Uncertain Aerosol Microphysics

Atmospheric models simulate the concentration of gaseous pollutants and aerosols in the atmosphere. These

models include significant mismatch from the true atmosphere resulting from uncertainty in the inputs including

meteorology and emission fields, extrapolation and simplification of experimentally-derived relationships

for the process dynamics, and possibly un-accounted for processes. Uncertainty in aerosol microphysical

processes is a significant source of uncertainty in estimates of climate change. In this thesis we estimate

time-varying parameters that adjust uncertain aerosol dynamics to account for some of the model mismatch.

We show how to design a passivity-based input observer (PBIO), which estimates model input parameters,

using proportional feedback so that the model-measurement error exponentially decays. The PBIO considers

a set of inventory variables instead of the full model state-space. Furthermore, the PBIO not only requires

observations of the inventories but also their derivatives.

As the first implementation of the PBIO in an atmospheric model, we estimate uncertain parameters by utilizing daily satellite measurements of aerosol optical depth. We integrate distributed PBIOs with a three-dimensional chemical transport model (CTM) to estimate time-varying sea spray emission scaling factors across the Southern Ocean band. We find the measurement error decreases from 305% to 55% by utilizing distributed PBIOs.

To estimate aerosol nucleation, emission, and growth due to organic vapors, we aim to utilize the measured aerosol size distribution. First, we focus on developing and applying the estimation methodology in a zerodimensional “box” model as a proof-of-concept. The PBIO is first tested on a dataset of synthetic and perfect measurements that span diverse environments in which the true particle emissions, growth, and nucleation rates are known. The inverse technique accurately estimates the aerosol microphysical process rates with an average and maximum error of 2% and 13%, respectively. Then, the PBIO is applied to a global 3-D CTM

utilizing size distribution measurements at seven field sites in rural Western Europe. The overall model performance, in terms of the inventory variables, is improved with the PBIO at the seven sites as well as at six other sites not included in the online PBIO framework.