The development and application of sensitivity tools for investigating microphysical processes in atmospheric models
Sheyko, Benjamin Andrew
MetadataShow full item record
We present the development of the adjoint of a physically based cirrus formation parameterization that computes the sensitivity of formed crystal number concentration to numerous model variables (e.g., updraft velocity, soluble aerosol geometric mean diameter and number concentration, insoluble aerosol geometric mean diameter and number concentration, and ice deposition coefficient). The adjoint is demonstrated in the CESM Community Atmosphere Model Version 5.1, where sensitivity information is computed and used to quantify which variables are most responsible for modeled variability in formed crystal number concentration. The sensitivity of formed crystal number concentration to updraft velocity is positive and largest over the tropics where regions of deep convection are collocated with large sulfate number concentrations. Sensitivity to sulfate number concentration is largest over the tropics where updraft cooling is sufficient and sulfate number concentration is low, pointing to a sulfate limited regime. Outside of the tropics, crystal production is dominated by heterogeneous freezing; unexpectedly, sensitivities to insoluble aerosol number concentration for accumulation and coarse mode dust, black carbon, and organic carbon are negative in sign here. This is a result of infrequent, anomalously high updraft velocity events causing shifts in the dominant modes of freezing which act to bias sensitivity information when annually averaged. Updraft velocity is responsible for ~95% of the variability in formed crystal number concentration in the high latitudes of the Northern Hemisphere. In the tropics, sulfate number concentration controls variability in formed crystal number concentration since crystal production here is sulfate limited. Insoluble aerosol species play a secondary role in influencing the variability of crystal concentrations; coarse mode dust is the largest contributor to crystal number variability at nearly 60%, although the spatial extent of this influence is small and concentrated over highly localized dust events. When globally averaged, nearly 90% of the variability in crystal number concentration can be described by only updraft velocity, sulfate number, temperature, and coarse mode dust number concentration. Although these results depend on parameter assumptions, the robustness of the underlying physics of the cirrus formation parameterization used throughout this work suggests that this approach can be a powerful method for efficiently identifying the origin of microphysical dependencies within large scale atmospheric simulations.