We use cookies to ensure that we give you the best experience on our website. You can change your cookie settings at any time. Otherwise, we'll assume you're OK to continue.

Institute of Advanced Study

Quantifying Output Uncertainty in Models used for Climatic Change Research

Professor Brian Huntley (Biological and Biomedical Sciences)

Professor Michael Goldstein (Mathematical Sciences)

Background and Key Research Questions

Models are widely used by Durham researchers, especially those whose research addresses various aspects of climatic change. Two broad classes of model are used by these researchers:

Firstly, models that simulate the dynamics of some component of the Earth system (e.g. ice-sheet dynamics, population and range dynamics of species, seasonal snow accumulation and melting). These models use simplified representations of the underlying processes and require estimates of (large numbers of) parameter values based upon (often limited) data, such estimates often being obtained using a statistical model.

Secondly, statistical models of the relationship between some component of the Earth system and one or more environmental variables, such models being used both in forward mode, to simulate how the Earth system component may respond to altered environmental conditions, and inverted to infer past environmental conditions from measurements of the Earth system component (e.g. to infer past July temperatures from assemblages of fossil Chironomids).

Key general questions to be addressed by the planned research include:

  • How close are model outputs to reality?
  • How uncertain are estimated parameter values, and how does this uncertainty influence uncertainty in the model outputs?
  • How robust are models when used to infer, or to simulate processes under, environmental conditions unlike those prevailing during the period from which the data used to develop the model were obtained?

More specific research questions relevant to research currently undertaken at Durham include:

  • How to develop inverse modelling approaches that will provide realistic uncertainty estimates for the environmental variable(s) being inferred?

Inverse models currently are used by researchers in SBBS and Geography to infer past environmental conditions (e.g. relative sea levels, seasonal temperatures) from assemblages of fossils of various groups of organisms (e.g. foraminiferans, Chironomids, pollen, diatoms).

  • How realistically to assess uncertainties in parameter estimations used in process models and the impact of these uncertainties on uncertainties in the model outputs, especially when simulations are made for past or future environmental conditions unlike those of the present?

Researchers in SBBS, Geography, Earth Science and several other departments use a variety of process models, often with many parameters, frequently also applying these models to simulate the behaviour of Earth system components under environmental conditions unlike those for which the data used to estimate the model parameters were collected.

Aims of the Programme

The overall aims of the proposed programme of events are:

  • To develop an increased awareness amongst modellers both of the need to quantify uncertainties in model outputs that arise from uncertainties in the input data, parameter values and/or process specifications, and of the methods available for achieving such quantification.
  • To identify statistical tools and methods that can be applied to quantify uncertainties in model outputs, where these already are available.
  • To stimulate collaborative research between statisticians and those developing and applying models that will generate new tools able to provide more realistic uncertainty estimates for both (forward) process-based or dynamic models and for statistical models, including inverse models.
  • To stimulate the development and writing of grant proposals to various funding agencies seeking support for the research ideas that emerge from the discussions amongst researchers.

Programme of Events

The first two aims will be achieved through a programme of guest lectures/seminars delivered by modellers and statisticians. Guest speakers will be invited each to spend 2-3 days in Durham following their lecture to enable interactions with Durham researchers.

The third aim will be achieved by convening a workshop attended by relevant Durham researchers, selected researchers from other UK institutions and relevant international partners. The primary workshop outcome will be the stimulation of new collaborative research between modellers and statisticians. Financial support for this research will be sought through one or more standard or consortium research proposals to a UK or EU funding agency, initial drafting of which will take place during the workshop.


Date Speaker Title
3 November 2011 Jonty Rougier The statistical challenges of limitations in dynamical models of climate
11 November 2011 Professor John Haslett Reconstructing Palaeoclimates
17 November 2011 Professor Stephen Sitch Vegetation Dynamic Models - links to GCMs
1 December 2011 Mark O'Malley Energy Systems: Susceptibility to Climate Change
19 January 2012 Richard Chandler
2 February 2012 Peter Challenor
Professor Wolfgang Cramer Cancelled
15 March 2012 Professor Philippe Huybrechts Modelling Ice Sheets
22 March 2012 Professor Tony Payne Ice-Sheet Dynamics
29 March 2012 Professor Paul Valdes Modelling Palaeoclimates
3 May 2012 Professor Wilfried Thuiller Species' Distribution Models

These seminars will take place at the Institute of Advanced Study, 1.00-2.00pm.