Nudged and forecast simulations in a multiscale modeling framework - maximizing the use of high value intermittent observations

Conventional climate model evaluation techniques, which often compare heavily averaged or subsampled measurements and simulation results, cannot fully evaluate the representation of cloud-aerosol processes that are intermittently observed and occur on naturally short time scales. In this study, complimentary methods of "nudging" and "forecast initialization" are used for the first time to observationally constrain simulations in a super-parameterized version of the Community Atmosphere Model (SP-CAM). SP-CAM applies the multiscale modeling framework (MMF) approach to climate modeling by embedding two-dimensional cloud resolving models (CRMs) within each CAM grid cell to explicitly represent sub-grid convection and replace conventional cloud and boundary layer parameterizations. The MMF approach has shown improvements in the pattern and evolution of convection from diurnal to inter-annual time scales, leading to more realistic precipitation intensity and variability compared to standard CAM, which are important controls on the processing and removal of aerosol particles. By constraining SP-CAM to an observed weather state, CRM driven convection, aerosol processes, and their interactions can be evaluated on time scales relevant to their residence times as they respond to a realistic large-scale forcing. Nudging, which relaxes the temperature, humidity, and wind fields in the model to analyzed observations, can be run continuously in a constrained mode for long periods of time. Forecast initialization starts the model with the best available representation of the observed atmosphere and then lets it run freely, allowing for short integrations before the simulation naturally diverges from observations. The model is then evaluated against satellite observations, ground measurements, and the conventional CAM convective parameterizations, not only statistically, but as observed in time following realistic weather events, maximizing the use of high value intermittent observations and field campaigns.