University of Edinburgh logo British Antarctic Survey logo National Oceanographic Centre logo University of Leeds logo


Using Data Science to Improve Prediction of the Upper Atmosphere

In recent decades the scientific community has made major advances in forecasting both space weather and terrestrial weather in the troposphere and stratosphere. However, the mesosphere/lower-thermosphere/ionosphere (MLTI) region, covering the altitude range from ~50 to 150 km, has for the most part been neglected. Even though the MLTI is a critical transition region between the lower atmosphere and geospace, our current skill in predicting it is little better than climatology. We know at the summers in the mesosphere are cooler than the winters (unlike near the surface), but we cannot accurately predict the MLTI weather even with a few days. This lack of knowledge surrounding the MLTI is an impediment to future advances in prediction of variability of the whole atmosphere.

Historically there have been few observations of this region to help characterise it. However, in the past decade or so the number of observations has increased markedly, including those from multiple middle atmosphere observing satellite missions. By taking advantage of this golden age of middle atmosphere observations,  the new generation of “whole atmosphere” models and cutting edge data science tools it will be possible to investigate the teleconnections between the mesosphere and space weather/lower atmosphere. In doing so, his work explores pathways to improve middle atmosphere model predictability, with the potential to improve prediction throughout the atmosphere.

The supervisors leading this project are involved in a multi-million pound project to study this area of research. The student would be working as part of an 8 person team across 3 institutes and spend time at the Met Office, working with the extended version of the Unified Model as part of their placement.

Aims and objectives of the PhD project:

The overall aim of the PhD is to use satellite and ground based observations, modelling and data science to identify key teleconnections in the atmosphere that drive MLTI variations.  Specifically the student will:

  • Data mine level 2 data from satellites and ground based instrument data to identify the key modes of variability in the MLTI.
  • Identify potential teleconnections that connect MLTI variability to the lower atmosphere and geospace, using machine learning techniques such as CNN and clustering.
  • Evaluate the ability of the current state-of-the-art models to reproduce these teleconnections.

Satellite data identified for this project are: Aeolus, AIRS, COSMIC, HIRDLS, IASI, MLS, SABER and VIIRS. Ground based data identified are from: MF and Meteor radars, Lidars and airglow imagers.

The student will be based at the University of Leeds but will make regular visits to BAS as well as a 3-month placement at the Met Office. This will provide excellent training and first-hand experience of working in world-leading research facilities.