1Dept. of Applied Environmental Science, Stockholm University, 10691 Stockholm, Sweden
2Bert Bolin Centre for Climate Research, Stockholm University, 10691 Stockholm, Sweden
3The Henry Samueli School of Engineering, Department of Civil and Environmental Engineering, University of California, Irvine, USA
4Computational Geo-Ecology, Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam, Amsterdam, The Netherlands
5Department of Meteorology, Stockholm University, Sweden
6Dept. of Chemical and Environmental Engineering, The University of Arizona, Tucson, USA
7Department of Atmospheric Sciences, The University of Arizona, Tucson, USA
Abstract. This paper presents a novel approach to investigate cloud-aerosol interactions by coupling a Markov Chain Monte Carlo (MCMC) algorithm to a pseudo-adiabatic cloud parcel model. Despite the number of numerical cloud-aerosol sensitivity studies previously conducted few have used statistical analysis tools to investigate the sensitivity of a cloud model to input aerosol physiochemical parameters. Using synthetic data as observed values of cloud droplet number concentration (CDNC) distribution, this inverse modelling framework is shown to successfully converge to the correct calibration parameters.
The employed analysis method provides a new, integrative framework to evaluate the sensitivity of the derived CDNC distribution to the input parameters describing the lognormal properties of the accumulation mode and the particle chemistry. To a large extent, results from prior studies are confirmed, but the present study also provides some additional insightful findings. There is a clear transition from very clean marine Arctic conditions where the aerosol parameters representing the mean radius and geometric standard deviation of the accumulation mode are found to be most important for determining the CDNC distribution to very polluted continental environments (aerosol concentration in the accumulation mode >1000 cm−3) where particle chemistry is more important than both number concentration and size of the accumulation mode.
The competition and compensation between the cloud model input parameters illustrate that if the soluble mass fraction is reduced, both the number of particles and geometric standard deviation must increase and the mean radius of the accumulation mode must increase in order to achieve the same CDNC distribution.
For more polluted aerosol conditions, with a reduction in soluble mass fraction the parameter correlation becomes weaker and more non-linear over the range of possible solutions (indicative of the sensitivity). This indicates that for the cloud parcel model used herein, the relative importance of the soluble mass fraction appears to decrease if the number or geometric standard deviation of the accumulation mode is increased.
This study demonstrates that inverse modelling provides a flexible, transparent and integrative method for efficiently exploring cloud-aerosol interactions efficiently with respect to parameter sensitivity and correlation.