Sunday, February 19, 2017
Exhibit Hall (Hynes Convention Center)
Benjamin Brown-Steiner, Massachusetts Institute of Technology, Quincy, MA
Continuing advancements in computational capabilities has enabled the atmospheric chemistry and climate research community to incorporate more detailed and complex components and parameterizations and to increase the resolution and length of their simulations. However, the atmospheric research community has a long history of clever and efficient utilizations of limited computational resources that have resulted in innovative and original results and methodologies. In the hope of continuing this tradition and of maximizing the potential for uncertainty analysis, we examine the potential benefits and uncertainties provided by three chemical mechanisms of different complexities (full tropospheric, reduced hydrocarbon, and a superfast mechanism) within the CESM CAM-Chem framework and characterize the research questions that can most benefit from longer simulations of simpler mechanisms. Preliminary results indicate that while biases in the magnitudes and the absolute value of the variability in surface ozone over the United States can be different among mechanisms, the day-to-day and interannual variability, as well as the characterization of the distribution and extreme values, is well captured among all mechanisms. We demonstrate that uncertainties associated with the utilization of different chemical mechanisms is less important in the characterization of ozone distributions and variability than the choice of meteorological dataset, and that a process of calibration -- in which mean values and standard deviations from a short run of the most-complex mechanism are applied to the time series from a long run of the simplest mechanism -- can sufficiently emulate the behavior of the complex mechanism. We offer recommendations and suggestions for the length of time, the averaging window, and the size of the region in which this process is sufficient and propose to extend this analysis to other parametric uncertainties (meteorological dataset, model resolution) and structural uncertainties (via the GEOS-Chem model).