Evaluation of a Quasi-steady state approximation of the cloud Droplet Growth Equation (QDGE) scheme for aerosol activation in global models using multiple aircraft data over both continental and marine environments

This research introduces a numerically efficient aerosol activation scheme and evaluates it by using stratus and stratocumulus cloud data sampled during multiple aircraft campaigns in Canada, Chile, Brazil, and China. The scheme employs a Quasi-steady state approximation of the cloud Droplet Growth Equation (QDGE) to efficiently simulate aerosol activation, the vertical profile of supersaturation, and the activated cloud droplet number concentration (CDNC) near the cloud base. We evaluate the QDGE scheme by specifying observed environmental thermodynamic variables and aerosol information from 31 cloud cases as input and comparing the simulated CDNC with cloud observations. The average of mean relative error of the simulated CDNC for cloud cases in each campaign ranges from 17.30 % in Brazil to 25.90 % in China, indicating that the QDGE scheme successfully reproduces observed variations in CDNC over a wide range of different meteorological conditions and aerosol regimes. Additionally, we carried out an error analysis by calculating the Maximum Information Coefficient (MIC) between the mean relative error (MRE) and input variables for the individual campaigns and all cloud cases. MIC values are then sorted by aerosol properties, pollution level, environmental humidity, and dynamic condition according to their relative importance to MRE . Based on the error analysis we found that the magnitude of MRE is more relevant to the specification of input aerosol pollution level in marine regions and aerosol hygroscopicity in continental regions than to other variables in the simulation.

Physically regularized machine learning emulators of aerosol activation

The activation of aerosol into cloud droplets is an important step in the formation of clouds and strongly influences the radiative budget of the Earth. Explicitly simulating aerosol activation in Earth system models is challenging due to the computational complexity required to resolve the necessary chemical and physical processes and their interactions. As such, various parameterizations have been developed to approximate these details at reduced computational cost and accuracy. Here, we explore how machine learning emulators can be used to bridge this gap in computational cost and parameterization accuracy. We evaluate a set of emulators of a detailed cloud parcel model using physically regularized machine learning regression techniques. We find that the emulators can reproduce the parcel model at higher accuracy than many existing parameterizations. Furthermore, physical regularization tends to improve emulator accuracy, most significantly when emulating very low activation fractions. This work demonstrates the value of physical constraints in machine learning model development and enables the implementation of improved hybrid physical and machine learning models of aerosol activation into next-generation Earth system models.

Is a more physical representation of aerosol activation needed for simulations of fog?

Aerosols play a crucial role in the fog life cycle, as they determine the droplet number concentration and hence droplet size, which in turn controls both the fog's optical thickness and lifespan. Detailed aerosol-microphysics schemes which accurately represent droplet formation and growth are unsuitable for weather forecasting and climate models, as the computational power required to calculate droplet formation would dominate the treatment of the rest of the physics in the model. A simple method to account for droplet formation is the use of an aerosol activation scheme, which parameterises the droplet number concentration based on a change in supersaturation at a given time. Traditionally, aerosol activation parameterisation schemes were designed for convective clouds and assume that supersaturation is reached through adiabatic lifting, with many imposing a minimum vertical velocity (e.g. 0.1 m s−1) to account for the unresolved subgrid ascent. In radiation fog, the measured updraughts during initial formation are often insignificant, with radiative cooling being the dominant process leading to saturation. As a result, there is a risk that many aerosol activation schemes will overpredict the initial fog droplet number concentration, which in turn may result in the fog transitioning to an optically thick layer too rapidly. This paper presents a more physically based aerosol activation scheme that can account for a change in saturation due to non-adiabatic processes. Using an offline model, our results show that the equivalent cooling rate associated with the minimum updraught velocity threshold assumption can overpredict the droplet number by up to 70 % in comparison to a typical cooling rate found in fog formation. The new scheme has been implemented in the Met Office Natural Environment Research Council (NERC) Cloud (MONC) large eddy simulation (LES) model and tested using observations of a radiation fog case study based in Cardington, UK. The results in this work show that using a more physically based method of aerosol activation leads to the calculation of a more appropriate cloud droplet number. As a result, there is a slower transition to an optically thick (well-mixed) fog that is more in line with observations. The results shown in this paper demonstrate the importance of aerosol activation representation in fog modelling and the impact that the cloud droplet number has on processes linked to the formation and development of radiation fog. Unlike the previous parameterisation for aerosol activation, the revised scheme is suitable to simulate aerosol activation in both fog and convective cloud regimes.

Evaluation of a numerically efficient aerosol activation scheme by using worldwide cloud data from multiple field campaigns in continental and marine regions

<p>This research summarizes a numerically efficient aerosol activation scheme and evaluates it by using stratus and stratocumulus cloud data sampled during multiple aircraft campaigns in China, Canada, Brazil, and Chile. The scheme employs a Quasi-steady state approximation of the cloud Droplet Growth Equation (QDGE) to efficiently simulate aerosol activation and vertical profiles of supersaturation and cloud droplet number concentration (CDNC) near the cloud base. We evaluated the scheme by specifying observed environmental thermodynamic variables and aerosol properties from 36 cloud cases as input and comparing the simulated CDNC and other simulated variables with cloud microphysical observations. The relative error (RE) of the mean simulated CDNC ranges from 15.27 % for Chile to 23.97 % for China, with an average of 19.69 %, indicating that the scheme successfully reproduces observed variations in CDNC over a wide range of different meteorological conditions and aerosol regimes. Subsequently, we carried out an error analysis by calculating the Maximum Information Coefficient (MIC) values between RE and individual input variables and sorted them by aerosol properties, pollution degree, environmental humidity, and dynamic condition according to their importance. Based on this analysis we find that the magnitude of the RE is sensitive to the specification of aerosol chemical composition and updraft velocity in the simulation, which can partly explain differences between simulated and observed CDNC in some of the regions.</p>

Physically Regularized Machine Learning Emulators of Aerosol Activation

The activation of aerosol into cloud droplets is an important step in the formation of clouds, and strongly influences the radiative budget of the Earth. Explicitly simulating aerosol activation in Earth system models is challenging due to the computational complexity required to resolve the necessary chemical and physical processes and their interactions. As such, various parameterizations have been developed to approximate these details at reduced computational cost and accuracy. Here, we explore how machine learning emulators can be used to bridge this gap in computational cost and parameterization accuracy. We evaluate a set of emulators of a detailed cloud parcel model using physically regularized machine learning regression techniques. We find that the emulators can reproduce the parcel model at higher accuracy than many existing parameterizations. Furthermore, physical regularization tends to improve emulator accuracy, most significantly when emulating very low activation fractions. This work demonstrates the value of physical constraints in machine learning model development and enables the implementation of improved hybrid physical-machine learning models of aerosol activation into next generation Earth system models.

Assessing Desert Dust Indirect Effects on Cloud Microphysics through a Cloud Nucleation Scheme: A Case Study over the Western Mediterranean

In this study, the performance and characteristics of the advanced cloud nucleation scheme of Fountoukis and Nenes, embedded in the fully coupled Weather Research and Forecasting/Chemistry (WRF/Chem) model, are investigated. Furthermore, the impact of dust particles on the distribution of the cloud condensation nuclei (CCN) and the way they modify the pattern of the precipitation are also examined. For the simulation of dust particle concentration, the Georgia Tech Goddard Global Ozone Chemistry Aerosol Radiation and Transport of Air Force Weather Agency (GOCART-AFWA) is used as it includes components for the representation of dust emission and transport. The aerosol activation parameterization scheme of Fountoukis and Nenes has been implemented in the six-class WRF double-moment (WDM6) microphysics scheme, which treats the CCN distribution as a prognostic variable, but does not take into account the concentration of dust aerosols. Additionally, the presence of dust particles that may facilitate the activation of CCN into cloud or rain droplets has also been incorporated in the cumulus scheme of Grell and Freitas. The embedded scheme is assessed through a case study of significant dust advection over the Western Mediterranean, characterized by severe rainfall. Inclusion of CCN based on prognostic dust particles leads to the suppression of precipitation over hazy areas. On the contrary, precipitation is enhanced over areas away from the dust event. The new prognostic CCN distribution improves in general the forecasting skill of the model as bias scores, the root mean square error (RMSE), false alarm ratio (FAR) and frequencies of missed forecasts (FOM) are limited when modelled data are compared against satellite, LIDAR and aircraft observations.

Is a more physical representation of aerosol activation needed for simulations of fog?

Aerosols play a crucial role in the fog life cycle, as they determine the droplet number concentration, and hence droplet size, which in turn controls both the fog's optical thickness and life span. Detailed aerosol-microphysics schemes which accurately represent droplet formation and growth are unsuitable for weather forecasting and climate models, as the computational power required to calculate droplet formation would dominate the treatment of the rest of the physics in the model. A simple method to account for droplet formation is the use of an aerosol activation scheme, which parameterises the droplet number concentration based on a change in supersaturation at a given time. Traditionally, aerosol activation parameterisation schemes were designed for convective clouds and assume that supersaturation is reached through adiabatic lifting, with many imposing a minimum vertical velocity (e.g. 0.1 m/s) to account for unresolved sub-grid ascent. In radiation fog, the measured updrafts during initial formation are often insignificant, with radiative cooling being the dominant process leading to saturation. As a result, there is a risk that many aerosol activation schemes will overpredict the initial fog number concentration, which in turn may result in the fog transitioning to an optically thick layer too rapidly. This paper presents a more physically-based aerosol activation scheme that can account for a change in saturation due to non-adiabatic processes. Using an offline model, our results show that the minimum updraft velocity threshold assumption can overpredict the droplet number by up to 70 % in comparison to a cooling rate found in fog formation. The new scheme has been implemented in the Met Office Natural Environment Research Council (NERC) Cloud (MONC) LES model, and tested using observations of a radiation fog case study based in Cardington, UK. The results in this work show that using a more physically-based method of aerosol activation leads to the calculation of a more appropriate cloud droplet number. As a result, there is a slower transition to an optically thick (well-mixed) fog that is more in-line with observations. The results shown in this paper demonstrate the importance of aerosol activation representation in fog modelling, and the impact that the cloud droplet number has on processes linked to the formation and development of radiation fog. Unlike the previous parameterisation for aerosol activation, the revised scheme is suitable to simulate aerosol activation in both fog and convective cloud regimes.