Evaluating the Role of Aerosol Mixing State in Cloud Droplet Nucleation using a New Activation Parameterization
Here, a simple scheme is proposed to evaluate the performance of activation parameterizations against a spectrum of mixing states, and two schemes commonly used in global models are studied using this framework. It is shown that each scheme exhibits systematic biases when a complex mixing state is present. To help resolve these issues, a new scheme is derived using Polynomial Chaos Expansion to build meta-models representing a full complexity parcel model. The meta-models are shown to accurately handle activation in both single-mode and mixture cases. In addition, a global sensitivity analysis is applied to benchmark the performance of the meta-models and the activation parameterizations against a detailed parcel model, and it is shown that the meta-models tend to more accurately attribute variability in activation dynamics to each input parameter and their interactions with others when compared to the physically-based parameterizations. A variety of experiments with different quadrature rules and orders are tested to optimize the performance of the chaos expansion meta-models in terms of accuracy and computational complexity.
In addition, a chaos expansion meta-model is implemented into a new global aerosol-climate model built by coupling the MIT Aerosol Model and the NCAR Community Earth System Model (CESM). The CESM already implements a physically-based activation parameterization, and estimates of anthropogenic aerosol indirect effects using both parameterizations are compared.