Developing Model Ensembles and Emulators for Next-Generation City Simulation

This PhD project will be part of an exciting new initiative entitled “Data Assimilation for Agent-Based Models (DUST)” (https://dust.leeds.ac.uk/). The project will develop important new methods that can be used to integrate data that emerge from smart cities (e.g. traffic counters, social media activity, environment sensors, etc.) into large-scale urban simulations in real time. This is an important methodological challenge that has yet to be overcome. Without these, it is very difficult to understand how to manage cities in the short-term, or how to respond in situations that require dramatic interventions such as an evacuation. The successful applicant will join a team of PhD students and researchers who are working together to improve the ways that scientists can model cities.

This PhD will contribute to the project by developing companion methods that allow simulations, such as agent-based models, to be rigorously applied to the real-world. Ensembles are groups of models that are executed in parallel to provide a means of exploring the range of possible outcomes from probabilistic models. These are an ideal way to better understand uncertainty in model results, but are not regularly used in the field of agent-based modelling. A related concept, emulators, refers to simple models that approximate a more complex and computationally demanding model. Developing good emulators for agent-based models could be extremely valuable, especially in the context of running ensembles of hundreds or thousands of models, but again have not been extensively used in the field. This PhD will extend the state-of-the art in methods such as emulators and ensembles – and other techniques such as Bayesian inference that might be appropriate – to allow the urban simulations that are developed as part of DUST to be applied to real-world urban systems.

How to apply: http://www.geog.leeds.ac.uk/study/phd/how-to-apply/