Characterising Uncertainty in Complex Environmental Simulations for Public Engagement with Climate Change Conscious Sustainable Planning and Design

The Grantham Centre for Sustainable Futures focuses on advancing the science of sustainability and connecting it with the policy debate around how humans can live in a more sustainable way.
grantham.sheffield.ac.uk

We are recruiting Grantham Scholars who will combine outstanding intellect with a strong commitment to public engagement, leadership and action. If these principles match your ambitions, you are invited to apply for one of our interdisciplinary PhD research projects to help solve the challenges of sustainability. You will be supported by the Grantham Centre through a unique training programme, designed to equip to with the skills to become a policy advocate and leader in sustainability matters.

Your applications for this studentship should be accompanied by a CV and a 200 word supporting statement. Your statement should outline your aspirations and motivation for studying in the Grantham Centre, outlining any relevant experience.

Please submit this in the online application. Please note: in online application process please select ’standard PhD’ not DTC option.

Project Description
Urban neighbourhoods and buildings designed or retrofitted with future climate in mind are more likely to perform sustainably. To do so, the planning and design process will require uses of detailed site-specific climate projections, and complex urban microclimate and building climate models. Although computationally intensive, the multi-scale environmental modelling and simulation can be used to systematically explore a large number of planning and design parameters and options, to examine the likely effect on sustainability over time. However, the computer simulation involved is often computationally expensive and contains uncertain elements, and the implication of compounded uncertainties in the complex multi-scale environmental simulations is not well understood. This research aims to identify, quantify and visualise such uncertainties.

With improved characterisation of uncertainty through statistical quantification and data visualisation planning and design professionals will better incorporate uncertainty in computational models in their planning and design practice. Also, researchers and developers of computational models will be better informed of the end-user requirements of seeing and comprehending uncertainty in the models. Furthermore, public understanding and engagement with sustainable planning and design policies in response to climate change could be transformed because of the greater transparency made possible by well-researched visual communication of uncertainty.

The previous research on multi-scale environmental simulations for site-specific climate change conscious urban neighbourhood design has identified three interlinking areas: (a) generate present-day microclimate profiles (e.g. air temperature, atmospheric humidity, wind speed, wind direction, etc.) for an existing or proposed urban neighbourhood site using data from the weather station closest to the site; (b) generate indoor environmental profiles for individual buildings in the neighbourhood based on the present-day microclimate profiles; and (c) repeat (a) & (b) for future years according to the UKCP09 probabilistic climate projections.

In computational fluid dynamics (CFD) based simulators, an urban microclimate is modelled as complex spatial-temporal interactions of air, surface (including soil) and vegetation. The operation of these prognostic models often takes days to complete the simulation for a limited time span and relies upon many unknowns due to lack of reliable measurements available for a particular urban location. Also, in predicting building thermal performance and energy consumption using CFD based simulators, the operation requires substituting built-in ‘defaults’ for missing input data often without users’ knowledge, and without accounting for the uncertainty in doing so.

This project will employ advanced statistical methods such as Gaussian process emulators for overcoming computational costs, Monte Carlo methods for the forward propagation of uncertainty, and Bayesian updating methods for combining data and model predictions to reduce uncertainties. The project will have another supervisor Professor Helen Kennedy (Professor of Digital Society) to advise how methods of data analytics and visualisation can be developed for communicating uncertainty in complex environmental simulations to non-experts.

Key Words
uncertainty in computer models, climate change projections, sustainable neighbourhood and building design processes, environmental modelling and simulation, data visualisation