Monday, 9 July 2018
Regency A/B/C (Hyatt Regency Vancouver)
Gaussian process emulation allows one to derive multi-dimensional, non-linear relationships from sparse data. In the atmospheric sciences, this technique has so far been used to investigate the relationship between model response and uncertain parameters associated with physical parameterizations. We illustrate that emulation can also be applied to explore relationships and sensitivities of cloud field properties such as albedo or drizzle to atmospheric variables such as temperature, humidity, boundary layer depth, and aerosol. We apply emulation to large-eddy simulations of marine stratocumulus for a range of meteorological and aerosol conditions. By capturing a range of stratocumulus-topped boundary layers, the emulator technique provides a more comprehensive picture than individual case studies. The emulated range of boundary layer states can be interpreted as the “state space” of a single dynamical system. Boundary layer dynamics translate to an evolution in this state space. It is governed by "attractive" states, which are reached asymptotically; "basins of attraction” describe how the asymptotic state depends on the initial meteorological and aerosol conditions.
In contrast to case studies, our approach allows us to
- distinguish typical and rare situations, including aerosol-meteorology co-variability;
- determine regime-dependent susceptibilities with fewer simulations than would be needed for a traditional linear-fitting approach;
- comprehensively describe the contribution of boundary-layer dynamics to buffering or amplifying perturbations;
- conceptually interpret LES output and observations in the context of idealized models of the stratocumulus-topped boundary layer.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner