My research aims to improve the scientific basis of policy decisions regarding air pollutants and greenhouse gases. Air pollution and greenhouse gases are trace gases that alter our environment, with direct and indirect consequences for society. Our society must, therefore make difficult and expensive decisions about how to mitigate trace gas emissions or adapt to the consequences. When making these decisions, we rely on the science community to estimate the costs and benefits related to trace gas concentrations. The characterization of unknown concentrations (unmeasured past or future) depends on model simulations, which should represent a codified state of the science. The actual models used for policy decisions make practical compromises between computational tractability and pure science.

The modeling community must be vigilant in tracking how models are used, in order to identify applications that are inconsistent with the embedded simplifications. We also watch for model usage that goes beyond what the current scientific state of the knowledge supports. My research seeks to identify and improve model deficiencies by integrating regional and global chemical transport models and observations. This type of research is critical to the goals of several federal institutions, and is being undertaken in partnership with those entities and leading research and academic institutions.

My research first identifies model weaknesses by comparing model outputs with observations. Observations of trace gases (air pollutants and greenhouse gases) come from a variety of sources. Aircraft and satellite data provide spatial coverage that allows for examination of models in environments where relatively little is known. Evaluating the model in new environments can reveal areas of poor model skill. When the model deficiency is a result of model simplification, it is important to identify these conditions where a higher level of detail is required. This type of information prevents the inappropriate application of model data to problems the model is not fit for. When the poor model performance cannot be explained by model simplifications, the disagreement highlights an uncertainty or gap in the science that underlies the model.

The next step is to identify the key model processes and to constrain the model based on the scientific uncertainties involved. In higher-order models, like chemical transport models, there are many processes, each of which has its own associated uncertainty. To test individual processes, I develop targeted modeling systems that remove other processes through boundary conditions or stochastic process representations. Using these targeted modeling systems, I mathematically constrain the key process using Bayesian inference techniques. Constraining models with observations provides a rigorous platform for questioning modeling and scientific assumptions. The results from a well-designed, field-based, inference experiment provide direction for laboratory studies. When the model, field observations, and laboratory studies are brought together in this way, the policy-relevant science is advanced more than by any individual technique.

Correlation of LAI and model differences.

Evaluating and constraining models is often funded by the U.S. EPA, NOAA, NASA, NSF, and private institutions. EPA, NOAA, and NASA provide funding for independent scientific institutions to evaluate observational data that they produce. The EPA and private institutions fund evaluation and improvement of models to ensure that the policy decisions are informed by the best science available. I have participated in the writing of several project and grant proposals. I have done research with funding from the Houston Advanced Research Center and the NASA RoMANS project. I have also received a post-masters research fellowship (3 yrs) and, subsequently, a post-doctoral fellowship at the U.S. EPA.

During my dissertation, I developed new techniques that enable me to constrain problems facing organizations that apply models. The funding of my dissertation and post-doctoral fellowship demonstrates my ability to work with funding organizations, and the relevance of my research to these groups. As a faculty member looking for funding, I would start by expanding collaborations already in progress and applying to the organizations with which I have existing relationships. I would then leverage my successful history of working with stakeholders to expand my funding base.

Likelihood estimation using population-based statistics.