skip to primary navigationskip to content

C405: Machine learning for Megacities: Understanding the impact of future climate extremes on energy resilience (Priority project with CASE partner) (Lead Supervisor: Scott Hosking, British Antarctic Survey)

Supervisors: Scott Hosking (British Antarctic Survey), Alex Archibald (Chemistry), Dr Emily Shuckburgh (British Antarctic Survey) and Dr Richard Turner (Engineering)

Importance of the area of research:

Regional and local-scale extreme events (such as heatwaves) will become more frequent over the next few decades, with rising mean temperature and increased climate variability. While climate models capture broad-scale spatial changes in climate phenomena, they struggle to represent extreme events on local scales. Such events are crucial to providing actionable and robust climate information to forecast, among other things, energy demand.

Around 50-55% of the world's population currently live in cities, accounting for 60-80% of energy consumption worldwide. Current projections estimate the proportion of population living in cities will rise to around 70% by the year 2050, concentrating power infrastructure further.  In addition, unique features of cities, such as the urban heat island effect can exacerbate extremes and fuel energy consumption (e.g. for air conditioning). Simultaneously, the occurrence of various types of extreme weather events is expected to increase, presenting a major source of uncertainty for forecasting power generation (e.g., wind turbine efficiencies) and power distribution (e.g., the complete destruction of pylons), and thereby threatening energy security.

Project summary:

The student will apply machine learning tools in new and innovative ways to help transform the field of environmental data science. Deep-rooted in this project is the use of intelligent and self-learning computer algorithms which automatically identify patterns in Big Data to make sense of highly complex systems. There is an abundance of data relevant to forecasting of power demand (e.g., details of the built environment, socio-economic forecasts). However, such data are not routinely incorporated into future climate risk projections.

Aims: 1. Develop a machine learning framework to predict increases in energy usage linked to heatwaves (e.g., due to increase use of air conditioning) within cities where we have a wealth of relevant climate and non-climate data: e.g. temperature, humidity, air quality, population, energy.

2. Apply this framework to emerging megacities

What the student will do:

Initially the student will focus on two exemplar cities: a large city like Madrid, which has an abundance of atmospheric measurements and energy data; and an emerging city such as Cairo, which has an urgent need for building resilience within its power networks against future climate extremes.

State-of-the-art machine learning techniques from the Cambridge Machine Learning group (built on TensorFlow) will be used to develop a statistical downscaling framework to capture these localised urban environments.

This framework will then be applied to climate simulations for a range of future scenarios (including 1.5 °C and 2 °C warming worlds) using a range of 'weather-resolving' climate simulations, e.g., CMIP6 and PRIMERVERA. Future socio-economic projections, such as population, age distributions and technological advancements, will also be incorporated within this framework.

Please contact the lead supervisor directly for further information relating to what the successful applicant will be expected to do, training to be provided, and any specific educational background requirements.


Iwata, T. & Ghahramani, Z. 2017, Improving Output Uncertainty Estimation and Generalization in Deep Learning via Neural Network Gaussian Processes, arXiv:1707.05922 [stat.ML]

United Nations, Department of Economic and Social Affairs, Population Division, 2015. World Urbanization Prospects: The 2014 Revision, (ST/ESA/SER.A/366),

Follow this link to find out about applying for this project.

Other projects available from the Lead Supervisor can be viewed here.

Filed under: