School of Earth and Environment

Lindsay Lee Dr Lindsay Lee

Leverhulme Research Fellow

Telephone number: +44(0) 113 34 36473
Email address:
Room: 10.127

Affiliation: Institute for Climate and Atmospheric Science


I am a statistician and my research interests are using advanced statistical methods to better understand Earth science models and the visualisation of statistical results for scientific interpretation.

I joined the department in January 2010 to work on the NERC funded AEROS project and I now work on the NERC funded GASSP project. For AEROS I carried out a sensitivity analysis of the global aerosol model GLOMAP (developed here at Leeds) in order to assess the effects of uncertainty in the model parameters on model predictions. In order to carry out the sensitivity analysis I conducted an expert elicitation, designed a set of model runs for emulation, and built and validated Gaussian Process emulators for univariate model outputs. The main challenge for me as a statistician was using these methods to provide information about a global model and producing visualisations and results that could be interpreted by the modellers. AEROS provided the modellers with unprecedented information about GLOMAP and its uncertainties as well as helping us to understand the limitations of our current predictions of atmospheric aerosol and its effect on the climate. The work I did for AEROS has been well received by the scientific community and I now advise on similar projects using different models. The vast data set we have as a result of the AEROS project has also been used by MSc and PhD students to assess other model uncertainties and their effects.

In GASSP I use the sensitivity results from AEROS to help constrain the model parameters and produce a calibrated version of GLOMAP understanding the limitations of the calibration and the irreducible uncertainties. The sensitivity results are used to reduce the globe to fewer regions in which similar parameters are acting for further analysis. History matching with perfect ‘observations’ is used to quantify the value of observations taken from different regions in reducing model uncertainty so that new observations can be suggested and the vast database of current observations used most efficiently. The use of perfect ‘observations’ means we can assess the usefulness of potential observations depending on their own uncertainties. Using the newly defined global regions we can test the validity of the constrained parameter sets within regions and between regions perhaps introducing new parameters where constraints don’t match.


2010: PhD in Probability and Statistics, University of Sheffield

Title: Climate variability and its effect on the UK carbon distribution. Used Gaussian process emulation and dynamic linear modelling to quantify the sensitivities of carbon fluxes to variability in the climate using the Sheffield dynamic global vegetation model (SDGVM).

2006: MSc in Statistics, University of Sheffield

Modules included linear modelling, time series analysis, multivariate data analysis, computational inference, Bayesian inference. Dissertation used time series analysis to count layers in ice core data.

2005: BSc in Mathematics, University of Sheffield

Research Interests

My research interests are in quantifying and understanding the effect of uncertainty in complex global models. I have so far worked with a global vegetation model and a global aerosol model but I also advise on similar use of the techniques in other Earth system component models. I apply well-established statistical methods to models whose uncertainty is not yet well-understood. I first look at parametric uncertainty which I believe is the first step to really understand model uncertainty. I am currently researching how understanding parametric uncertainty can help to identify structural uncertainties and understand model diversities with the aim of producing models in whose predictions we can be more confident and whose limitations are better understood.

I have given many talks to both the scientific and statistical community both regarding my research and the more general topic of using statistics to assess environmental model uncertainty as well as public lectures in the use of statistics in environmental modelling.