Statistical methods to quantify and visualise the complex behaviour of clouds in the climate system

Lead Research Organisation: University of Leeds
Department Name: School of Earth and Environment


This project will develop statistical methods to understand the highly complex behaviour of clouds and improve quantification of how clouds affect climate change. The PhD will apply and further develop uncertainty quantification and visualisation methods to high-resolution cloud model simulations to understand how clouds respond to many interacting environmental factors. By exploring visualisation of high-dimensional sparse model data the aim is to help scientists understand and interpret how changes in environmental conditions cause cloud fields to change dramatically under some conditions. The candidate will use a Large Eddy Simulation (LES) model developed by the US National Oceanic and Atmospheric Administration (NOAA).

Background and rationale
Shallow clouds have a strong influence on climate by reflecting solar radiation and causing a cooling effect. The clouds respond in complex ways to changes in air pollution (aerosols) as well as many environmental factors that are expected to change in a future climate. Cloud models are very complex and slow to run, so it has not been possible to understand how all these factors work together to influence climate. One major challenge is that shallow clouds are particularly sensitive to atmospheric aerosol concentrations and meteorological conditions - i.e., small changes in the temperature, humidity or aerosols can cause sharp discontinuities in behaviour. Climate models are currently unable to quantify the resulting uncertainty in shallow cloud simulations since the clouds form and develop on smaller scales than a climate simulation. Cloud-resolving models have been developed to better understand cloud responses to atmospheric aerosol and meteorological conditions but a quantification of the uncertainty in shallow cloud simulations is still lacking.
Uncertain quantification of complex computer codes typically relies on continuous, smooth "response surfaces". However, preliminary analysis of shallow cloud simulations has shown that cloud behaviour tends to shift between different "regimes" quite sharply as conditions are changed. The latest methodological development suggests using Voronoi tessellations to partition the regimes of cloud behaviour and to use Gaussian Process emulation to produce independent response surfaces for each regime. The studentship will further develop this new methodology following two strands of research:

Key research questions:
1. How can uncertainty be visualised and interpreted across different cloud regimes?
First the candidate will carry out sensitivity analysis on the independent Gaussian processes and work closely with scientists to develop visualisations of the resulting high-dimensional data that can be used to interpret the behaviour of the cloud field and better understand the effect of the uncertainty in shallow cloud simulations.

2. How do changes in the environment caused by climate change affect the behaviour of shallow cloud fields?
The candidate will work with scientists to design new model experiments to explore the variability in the simulated regime changes with the objective of understanding how climate change might affect changes in cloud behaviour. Sensitivity analysis of the discontinuity itself will require some developments in statistical methodology and will provide valuable information on why and where such regime changes occur.

The candidate will have the opportunity to work closely with scientists in NOAA supported by the world-leading cloud modeller Graham Feingold and with the Centre of Excellence for Modelling Atmosphere and Climate within our School to run simulations themselves. All training in computer programming for cloud simulation and statistical analysis will be given where appropriate.


10 25 50