Organised deep convection in weather and climate models
How can we improve prediction of energetic organised convective systems?
Mesoscale Convective Systems (MCS) are large complexes of many individual thunderstorm cells, several hundred km in diameter, and lasting for many hours (compare this to the scale of individual thunderstorms, which last around 30 minutes and have a scale of 10-15 km). MCS commonly occur across the Tropics, and over land during the warm season, including over the American Great Plains, West Africa, and Southern Europe. MCS influence the climate system in many ways. They release substantial quantities of latent heat which drives the circulation, they provide the main source of precipitation in many regions, and they have a large impact on the radiation budget.
Despite their importance, current climate models struggle to represent MCS. This is because the small-scale processes which drive MCS, including individual convective storms, are represented in models in a simplified manner through parametrisation schemes. The poor representation of MCS impacts the representation of extreme weather events and the water cycle in climate models, and hinders prediction of how these will change in a changing climate.
Even weather forecasting models, with substantially higher resolutions (approximately 10 km in space compared to 50-100 km in climate models) show large biases in the representation of MCS. The problems encountered in predicting these convective systems cascade outwards, impacting forecasts far away from the MCS in question. For example, forecast busts over the UK are generally associated with the presence of an energetic MCS over the United States.
Improving the simulation of MCS in global atmospheric models is therefore important for improving the accuracy of both global weather forecasts and climate predictions.
Aims of the Project
We seek to understand the sources of error in the representation of MCS in models, including triggering, propagation and coupling to the large-scale dynamics, particularly how it relates to assumptions made in the parametrisation process. This physical understanding will then be used to motivate improvements to the parametrisation schemes used to represent convective processes in weather and climate models.
Methods to be used
This project will study the evolution of MCS in observations and very high-resolution model simulations, and compare their evolution to models in which convection is parametrised. This allows us to characterise the ‘error’ involved in the parametrisation process. To what extent are these errors predictable (i.e. systematic), and to what extent do these errors reflect the chaotic and unpredictable nature of small-scale atmospheric processes? How can we improve our parametrisation schemes to correct for these systematic errors, but also to account for the random component of the error? Developing stochastic parametrisations to account for the random component of the error is likely to be key for improving the representation of MCS in models. Techniques from Machine Learning could be employed at various points in the project, if that is of interest to the studen
Specialised skills required
A strong background in maths or physics is essential.
Please contact Hannah Christensen on email@example.com if you are interested in this project