Parameterization Scheme For Aerosol Activation And Nucleation Adaptive To Horizontal Spatial Scale

The development of weather climate integration model requires a scale adaptive cloud microphysical parameterization schemes.The aerosol activation and nucleation parameterization scheme describe the aerosol-cloud interaction,which is an important part of the cloud microphysical parameterization schemes.This paper mainly studies the scale adaptive parameterization schemes of aerosol activation and nucleation.The grid scale dependence of these two kinds of parameterization schemes is mainly reflected in the use of vertical velocity driving parameterization schemes:sub grid disturbed vertical velocity(W_std)is usually used in large-scale climate models,while grid average vertical velocity(W_ave)is mainly used in high-resolution models.Therefore,the scale adaptive aerosol activation and nucleation parameterization scheme need the vertical velocity using method that can transition smoothly under different spatial resolutions.The vertical velocity data used in this paper comes from the high-resolution simulation test of WRF model.The statistical analysis results show that there are significant differences in the probability distribution characteristics of W_aveand W_stdunder different spatial resolutions,and the relative proportion of W_aveand W_stdchanges significantly.Therefore,in order to realize scale adaptation,the vertical velocity of driving parameterization scheme must take into account W_aveand W_std.Based on the original parameterization schemes of aerosol activation and nucleation based on physical process,a set of graded vertical velocity using method(PDF)using W_aveand W_stdat the same time is designed.According to the assumption of normal distribution,the vertical velocity in the model grid is divided into several possible bins,and the aerosol activation and nucleation numbers concentration in each possible bin are calculated respectively.Then,count the number concentration of aerosol activation and nucleation in the grid.The analysis and test show that the difference of the number of ice crystals and cloud droplets calculated by PDF method under different horizontal grid resolutions is very small,the transition is smooth,and it has the ability of scale adaptation.Due to the relatively large amount of computation required by PDF method,a scale adaptive parameterization scheme based on statistical relationship is further developed.In this paper,the empirical fitting formula is constructed by machine learning method.The model framework adopts shallow neural network,and the training sample data is calculated and generated by the original parameterization scheme in PDF method.The experimental results show that the statistical relationship has good fitting ability.The correlation coefficient of fitting ice nucleation is more than 0.7,and the correlation coefficient of fitting cloud droplet formation can reach0.99.In addition,the statistical relationship has well interpretability,which can reflect the physical relationship between aerosol activation,nucleation number concentration and vertical velocity.To sum up,this paper analyzes two kinds of solutions for the scale adaptation problem test of cloud microphysics parameterization scheme.The scheme based on PDF method can well realize the scale adaptation,but needs a large amount of computation.The statistical relation formula obtained by machine learning can overcome the defect of calculation amount in PDF method.However,the fitting ability of aerosol ice nucleation is limited.The work done in this paper is innovative in China,contributes to the development of weather and climate integration model,and provides suggestions for the improvement of cloud microphysical process parameterization.