Exploring a new mechanism in global Earth system models: Impacts of co-condensing semi-volatile compounds on clouds globally

Summary

The project combines field observational data, global earth system modelling and statistical emulators, to quantify the impacts of semi-volatile compounds on clouds globally.

Project background

Clouds play a critical role in global climate. Just a 6% increase in cloud reflectivity can offset the warming from CO2 doubling. Cloud droplets originate from aerosol particles. Semi-volatile compounds are abundant in the air. They are moderately volatile and can partition between gas and particle phases. Gaseous semi-volatiles can co-condense with water vapour, enhancing aerosol hygroscopic growth and facilitating cloud droplet formation. These processes can deteriorate air quality and affect clouds and radiation. Yet, this co-condensation effect is not well constrained as the semi-volatiles loss during drying and heating in traditional aerosol measurements is poorly understood, and quantifying the co-condensing mass in cloud formation processes is challenging. 

To address these challenges, our recent developed method uses open-access data to estimate aerosol hygroscopic growth, considering the co-condensation effect (Wang and Chen, 2019). Applying this method in Delhi, India, we found that the co-condensation of hydrogen chloride causes 50% of visibility reduction and can halve the activation critical supersaturation needed to form cloud droplets, significantly influencing local air quality and climate (Gunthe et al., 2021). A similar significant effect is observed in Chinese megacities (Beijing, Guangzhou, and Shanghai), but with nitric acid (Wang et al., 2019, and in prep.). Collaborating with the ETH team, we are developing a cloud parcel model to quantify this effect in various environments, including urban, rural, and boreal forest, as case studies.

The PhD project will build on the newly developed parcel model and established prior knowledge, continuing collaboration with the ETH team to deepen our understanding from case studies to global scale, targeting all inorganic and organic species from different emission sources. In this project, we will design a minimal runs of parcel models and use the outputs to train a Gaussian process emulator to build up a surrogate model. This model enables us to perform sensitivity analysis of input parameters on co-condensation effect in a computationally efficient way. Combining sensitivity data with a state-of-the-art earth system model will allow us to quantify the impact of co-condensation effect on clouds globally. 

Research questions

Q1: What are the key parameters controlling the co-condensation effect of semi-volatile compounds?

Q2: To what extent does the co-condensation effect influence cloud properties on a global scale, considering temporal and regional variabilities?

Q3: Which group of components is more dominant in co-condensation: inorganics or organics? 

Q4: Are there any links to local or transported emission sources? 

Methodology

The project will combine a 1-D cloud parcel model and a Gaussian emulator to perform a sensitivity analysis and quantify the sensitivity of modelled outputs (cloud droplet number concentration) to changes in model input parameters. In addition, a 3-D Earth system model will be used together with sensitivity analysis to help us understand the importance of the co-condensation . 

To map all input parameters to the output with sufficient accuracy requires thousands of parcel model runs, which can be prohibitively expensive. To overcome this difficulty, we will build a surrogate model using a Gaussian process emulator. This method allows us to use a minimal number of simulations (300-400 runs for 20-30 input parameters) to train an emulator. The emulator learns the relationship between the input and output parameters and subsequently can predict the model output for various given initial conditions within a few seconds on your laptop, replacing the costly parcel model. This approach is often referred to as a surrogate model. This model will help us assess the sensitivity output parameter to the initial conditions. Hence the student will gain experience with state-of-the-art statistical modelling. 

Year 1, The student will use a maximin Latin Hypercube Space-filling method to design the initial conditions for the representative simulations and run the cloud parcel model. He/she will then learn to build a surrogate model and train the emulator with parcel model outputs.

Year 2, The student will analyse these results to identify the key parameters controlling the co-condensation effect of semi-volatile compounds. He/she will also attend training to learn the fundamentals of climate models and how to analyse model output. Additionally, He/she will gather emulated data from Year 1, available observational data and global model outputs. 

Year 3, The student will utilise the collected data to address the following questions: 1) Quantify the regional and global impacts of co-condensation effect on cloud properties 2) Examine detailed contributions from different components in various environments with different source emissions.

Training

A comprehensive training programme will be provided comprising both specialist scientific training and generic transferable and professional skills. The student will attend relevant atmospheric science courses at the University of Edinburgh and those available through the National Centre of Atmospheric Science (NCAS). Specialist training will include in depth training in data-science, emulation, and climate modelling by supervisors.

Requirements

The project would suit a student with a strong Natural sciences background (Chemistry, Physics, Maths, Engineering), familiarity with programming, and an interest in climate science and modelling.