Activation of Cloud Droplets in Large-scale Models

In this study we compare to different methods for determining how many potential Cloud Condensation Nuclei(CCN) are activated. Both methods are applicable to large-scale models where either cloud-scale vertical motions are resolved or are parameterized. One approach is to implement an on-line bin-resolving sub-grid parcel model for aerosol activation. The other approach is to construct look-up tables for aerosol activation using the same parcel model but in an off-line mode. Previously we adopted the approach to construct look-up tables for aerosol activation of cloud droplets. The first generation of this approach, described by Saleeby and Cotton (2004), predicted droplet activation based on off-line parcel model simulations for a range of updraft speed, temperature, aerosol size, and number concentration. Aerosol chemistry was hard-wired to that of ammonium sulfate. Subsequently, the look-up tables were expanded to include aerosol chemistry or hygroscopicity represented by the hygroscopicity parameter, kappa,  (Ward et al. 2001). More recently Lerach (2012) further extended the look-up table approach to consist of an external mixture of pollution aerosols and dust. This required over 1000 parcel model runs to build the expanded look-up table. In Lerach's (2012) study, dust was assumed to originate from the southwestern U.S., and assigned a value of 0.03 (Koehler et al., 2009) and an accumulation mode median radius of 0.2 µm. We now consider expanding this approach to an external mixture of pollution aerosols, dust, and sea-spray generated sea salt particles. Again, thousands of parcel model runs are needed to construct the expanded parcel model runs. This expanded look-up table represents competition for water vapor among the pollution aerosols, dust, and sea-spray. One can imagine that a global atmospheric model is interfaced(one-way or interactively) with a global atmospheric chemistry model such as GEOS-Chem. The atmospheric chemistry model provides the concentration, and chemistry (via ) of internally mixed groups of pollution aerosols, externally mixed dust with its unique kappa, sizes, and concentrations, and externally mixed sea-spray generated sea salt with its unique  (1.28), sizes and concentrations. Typically, we hard-wire the sizes of each aerosol species to reduce the degrees of freedom.

However, as more aerosol species are included in a model simulation, many more combinations are possible of the concentrations, sizes, and chemistry(via  ) of each aerosol species. Thus each added aerosol species adds three degrees of freedom to the lookup table that provides a full representation of the aerosol competition for vapor. This quickly becomes intractable, and consequently the table dimensions must be reduced by lumping together the activation properties of multiple aerosol classes into one or very few bulk categories, as was done in Lerach (2012). An alternate strategy is to perform parcel model calculations on line. In this approach the parcel model is run as an integral part of the microphysics code in the global model for any grid cell and time-step in the simulation where conditions are right for nucleation. The parcel model is initialized with the specific environment of that grid cell and with an updraft speed that is either resolved or parameterized, and its bins are populated with the unactivated hygroscopic aerosols that currently occupy that cell with the predicted concentrations,  , and size. This approach provides an accurate representation of competition for vapor among the prognosed aerosols and avoids the necessity of approximations required by the lookup table approach (ie. hard-wired sizes). However, the on line method is computationally more expensive because it is done not off line but as part of the model simulation (on-line). Hence, it is necessary to distinguish between situations where aerosol activation is likely or unlikely so that the parcel model is not run unnecessarily.

In this study we compare both the off line and on line approaches to activation of potential CCN in terms of accuracy and efficiency.