## Jonathan (Jonty) Rougier's homepageReader in Statistics ## Summer School in risk and uncertainty in natural hazardsMore details here.## Exciting new book!## ResearchMy research concernsthe assessment of uncertainty in complex
systems, particularly environmental systems such as climate. Here is an
informal
summary. The main issues I think about are:
- How do we use scientific models? If they are to be used
quantatively to predict the behaviour of a system, e.g. the
climate, how do we represent the discrepancy between the model and
the system?
- Is there a general framework for combining model-evaluations,
system observations, and expert judgements, that is applicable
across different areas of science?
- What is the role of probability in representing our
uncertainty? What does a probability represent, and what are the
limitations of the probability calculus? What is the best way to
explain probabilistic concepts to non-statisticians?
- How do we implement our inferential calculations with very large models, that may take weeks or even months to evaluate? What corners do we cut?
## Some current and recent projects- CREDIBLE: one
of the two NERC-funded consortia on uncertainty and risk assessment for
natural hazards; 2012-2015.
- RATES: NERC-funded project to quantify mass trends in Antarctica using a
synthesis of model evaluations and several different datasets; 2011-2013.
- PalaeoQUMP:
"PalaeoQUMP aims to constrain climate sensitivity [...] using a
wider range of derived climate observations from the geological
past (reconstructions from sediments and geomorphological changes
for the Last Glacial Maximum and the mid-Holocene period), to
evaluate climate model predictions generated using the same
series of simulations as QUMP produced for the modern climate."
Funded by NERC QUEST: 2007-2010, but we're still analysing the
model runs!
## Publications- Here is my Google Scholar list of
publications: http://scholar.google.co.uk/citations?user=hcBGdiYAAAAJ (opens in a new window/tab).
More details are given below.
- Some excellent advice on writing
papers, http://www.diabetologia-journal.org/eicadvice.html,
including
"If you are lucky enough to find a statistician who can communicate with the non-numerate and is of the opposite sex, you should consider a proposal of marriage. It's that important."
## Technical reports- J.C. Rougier (2011), Monte Carlo methods for
Bayesian palaeoclimate reconstruction.
**Abstract.**In palaeoclimate reconstruction, the natural modelling direction is forwards from climate to sensors to proxy measurements. Statistical methods can be used to invert this direction, making climate inferences from proxy measurements. Among these methods, the Bayesian method would seem to deal best with the substantial epistemic uncertainties about climate, and about its impact on sensors. The main challenge is to perform this inference efficiently within a simulation approach. This paper reviews the Importance Sampling approach to Bayesian palaeoclimate reconstruction, and then goes on to demonstrate the value of recent advances in Markov chain Monte Carlo (MCMC) inference.Available as a pdf file. This paper was written as an addendum to the SUPRAnet report *Studying uncertainty in palaeoclimate reconstruction: A framework for research*(still under development). - J.C. Rougier (2011), A representation theorem for stochastic
processes with separable covariance functions, and its implications
for emulation.
**Substantially revised, again, 12 Oct 2011.**As this paper gets longer, I hope it gets clearer.**Abstract.**Many applications require stochastic processes specified on two- or higher-dimensional domains; spatial or spatial-temporal modelling, for example. In these applications it is attractive, for conceptual simplicity and computational tractability, to propose a covariance function that is*separable*; e.g. the product of a covariance function in space and one in time. This paper presents a representation theorem for such a proposal, and shows that all processes with continuous separable covariance functions are second-order identical to the product of second-order uncorrelated processes. It discusses the implications of separable or nearly separable prior covariances for the statistical emulation of complicated functions such as computer codes, and critically reexamines the conventional wisdom concerning emulator structure, and size of design.Available as a pdf file. - J.C. Rougier and M. Crucifix (2012),
Uncertainty in climate science and climate policy.
This is a book chapter for L. Lloyd and E. Winsberg, eds, Conceptual Issues in Climate Modeling, University of Chicago Press, forthcoming. Submitted draft chapter available as a pdf file. - J.C. Rougier, A. Zammit-Mangion and N. Schoen (2013), Visualisation for large-scale Gaussian updates.
In geostatistics (and also in other applications in science and engineering) we are now performing updates on Gaussian process models with thousands of components. These large-scale inferences involve computational challenges, because the updating equations cannot be solved as written, owing to the size and cost of the matrix operations. They also involve representational challenges, to account for judgements of heterogeneity concerning the underlying fields, and diverse sources of observations. Diagnostics are particularly valuable in this situation. We present a diagnostic and visualisation tool for large-scale Gaussian updates, the `medal plot'. This shows the initial and updated uncertainty for each observation, and also summarises the sharing of information across observations, as a proxy for the sharing of information across the state vector. It allows us to `sanity-check' the code implementing the update, but it can also reveal unexpected features in our modelling. We discuss computational issues for large-scale updates, and we illustrate with an application to assess mass trends in the West Antarctic Ice Sheet. Available as a pdf file.
## Statistics and ProbabilityJ.C. Rougier and M. Goldstein (2014), "Climate Simulators and
Climate Projections", A. Zammit Mangion, J.C. Rougier, J.L. Bamber and N.W. Schoen
(2013), "Resolving the Antarctic contribution to sea-level rise: a
hierarchical modelling framework", J.C. Rougier, M. Goldstein, and L. House (2013), "Second-order
exchangeability analysis for multi-model ensembles", I. Scheel, P.J. Green, and J.C. Rougier (2011), "A graphical diagnostic for
identifying influential model choices in Bayesian hierarchical
models", J.C. Rougier (2010), Discussion of "A statistical analysis of
multiple temperature proxies: Are reconstructions of surface
temperatures over the last 1000 years reliable?" by McShane and
Wyner, J.C. Rougier and M. Kern (2010), Predicting Snow Velocity in Large
Chute Flows Under Different Environmental Conditions. J.C. Rougier, S. Guillas, A. Maute, A.D. Richmond (2009), Expert
Knowledge and Multivariate Emulation: The Thermosphere-Ionosphere
Electrodynamics General Circulation Model
(TIE-GCM), M. Goldstein and J.C. Rougier (2009), Reified Bayesian Modelling
and Inference for Physical Systems, J.C. Rougier (2008), Efficient Emulators for Multivariate
Deterministic Functions, J.C. Rougier (2008), Discussion of 'Inferring Climate System
Properties Using a Computer Model', by Sanso et al, M. Goldstein and J.C. Rougier (2006), Bayes
Linear Calibrated Prediction for Complex Systems, M. Goldstein and J.C. Rougier (2004), Probabilistic
Formulations for Transferring Inferences from Mathematical Models to
Physical Systems, P.S. Craig, M. Goldstein, J.C. Rougier and A.H. Seheult (2001),
Bayesian Forecasting for Complex Systems Using Computer Simulators,
## Other scienceJ.C. Rougier (2013), 'Intractable and unsolved': some thoughts on statistical data assimilation with uncertain static parameters, D.B. Stephenson, M. Collins, J.C. Rougier, and R.E. Chandler
(2012), Statistical problems in the probabilistic prediction of
climate change, M. Collins, R.E. Chandler, P.M. Cox, J.M. Huthnance, J.C. Rougier, and
D.B. Stephenson (2012), Quantifying future climate change, R.M. Gladstone, V. Lee, J.C. Rougier, A.J. Payne, H. Hellmer, A. Le
Brocq, A. Shepherd, T.L. Edwards, J. Gregory, and S.L. Cornford
(2012), Calibrated prediction of Pine Island Glacier retreat during
the 21st and 22nd centuries with a coupled flowline model, P.W. Fitzgerald, J.L. Bamber, J. Ridley and J.C. Rougier (2011),
Exploration of parametric uncertainty in a Surface Mass Balance Model
applied to the Greenland Ice Sheet, N.R. Edwards, D. Cameron, J.C. Rougier (2011), Precalibrating an
intermediate complexity climate model, S. Guillas, J.C. Rougier, A. Maute, A.D. Richmond, and
C.D. Linkletter (2009), Bayesian calibration of the
Thermosphere-Ionosphere Electrodynamics General Circulation Model
(TIE-GCM), M. Crucifix and J.C. Rougier (2009), On the use of simple dynamical systems
for climate predictions: A Bayesian prediction of the next glacial
inception. J.C. Rougier, D.M.H. Sexton, J.M. Murphy, and D. Stainforth (2009),
Analysing the climate sensitivity of the HadSM3 climate model using
ensembles from different but related experiments. J.C. Rougier and D.M.H. Sexton (2007), Inference in Ensemble
Experiments, J.C. Rougier (2007), Probabilistic
Inference for Future Climate Using an Ensemble of Climate Model
Evaluations, J.C Rougier (2005), Probabilistic Leak Detection in Pipelines
Using the Mass Imbalance Approach. M. van Oijen, J.C. Rougier and R. Smith (2005), Bayesian
Calibration of Process-Based Forest Models: Bridging the Gap Between
Models and Data, ## Economics and FinanceS.C. Parker and J.C. Rougier (2007), The Retirement Behaviour of
the Self-Employed in Britain, P.R. Holmes and J.C. Rougier (2005), Trading Volume and Contract
Rollover in Futures Contracts, S.C. Parker and J.C. Rougier (2001), Measuring Social Mobility as
Unpredictability, B. Hillier and J.C. Rougier (1999), Real Business Cycles,
Investment Finance and Multiple Equilibria, J.C. Rougier (1997), A Simple Necessary Condition for Negativity
in the Almost Ideal Demand System with the Stone Price Index,
J.C. Rougier (1996), An Optimal Price Index for Stock Index
Futures Contracts, J.C. Rougier (1993), The Impact of Margin-Traders
on the Distribution of Daily Stock Returns: The London Stock Exchange,
## Books and book chaptersJ.C. Rougier, R.S.J. Sparks, and L.J. Hill (eds), 2013, - L.J. Hill, R.S.J. Sparks, and J.C. Rougier, Risk assessment and uncertainty in natural hazards, ch 1, pp 1-18.
- J.C. Rougier, Quantifying hazard losses, ch 2, pp 19-39.
- J.C. Rougier and K.J. Beven, Model and data limitations: the sources and implications of epistemic uncertainty, ch 3, pp 40-63.
## Non-peer-reviewedJ. Murphy, R. Clark, M. Collins, C. Jackson, M. Rodwell,
J.C. Rougier, B. Sanderson, D. Sexton and T. Yokohata (2011),
Perturbed parameter ensembles as a tool for sampling model
uncertainties and making climate projections, J.C. Rougier, T.L. Edwards, M. Collins and D.M.H. Sexton (2011),
Low-noise projections of complex simulator output: A useful tool when
checking for code errors, J.C. Rougier and L. Chen (2010), Comment on the paper by Diggle et
al., R. Chandler, J.C. Rougier, and M. Collins (2010), Climate
change, J.C. Rougier (2009), Notes on statistical modelling for complex systems, ver. 0.5, unpublished. Available as a pdf file. Please note the version number: this document is still evolving. J.C. Rougier (2008), Climate change detection and attribution, J.C. Rougier (2008), Formal Bayes Methods for Model Calibration with Uncertainty, in K. Beven and J. Hall (eds), Applied Uncertainty Analysis for Flood Risk Management, Imperial College Press / World Scientific. Draft version available as a pdf file. J.C. Rougier (2006), Comment on the paper by Haslett et al.,
J.C. Rougier (2005), Literate Programming for Creating and
Maintaining Packages. J.C. Rougier (2004), Comment on the paper by Murphy et al. J.C. Rougier (2001), Comment on the paper by Kennedy and
O'Hagan, J.C. Rougier (2001), What's the Point of `tensor'?, ## Miscellaneous## Selected presentations (since 2009)**Server advantages in tennis matches**, Iain MacPhee day, Durham, April 2014. Slides available. This is a blackboard presentation of MacPhee et al, 2004.**"Intractable and unsolved": the challenge of data assimilation with uncertain static parameters**, Intractability workshop, Bristol, April 2012. Slides available. This is really a talk about long-time-scale palaeoclimate reconstruction.**Nomograms for visualising relationships between three variables**, UseR! 2011 invited talk. Slides available. Here is some beta code for producing the parallel scale nomograms used in the talk: parallelNomo.R. This is the picture you should get if you run the example: nomogramBMI.pdf.After the discussion at UseR! 2011 I appreciated that I had an extra degree of freedom, which I could use to set the righthand axis so that the typical donkey had a horizontal line on the nomogram. Then I put all the donkeys' lines on the plot, and finished it off with a couple of density functions, original and predicted. The results is not exactly 'less is more': donkDensity_19Aug.pdf. After seeing that line at the top of the 'not less is more' nomogram heading off to surprising lengths I created a parallel coordinate plot of the donkey data, using parallelCoord.R, which resulted in parallelCoord_26Aug.pdf. As Kate Milner noted, "that is an exceptionally long donkey" (128cm). And in fact she checked her notebooks and this length had been misrecorded: it was in fact 102cm. And what's the UseR! connection? it was over a beer at the mixer that David Nicolaides made me understand the value of this plot for screening. **Multivariate emulation for North American mid-Holocene temperature reconstructions**, MPI-M, Hamburg, Mar 2011. Slides available. This is a longer version of the talk given at the Liverpool Marine Symposium, with a stronger focus on emulating large climate simulators (HadCM3, in our case). Note that the reconstruction at the end is*purely illustrative*.**Calibration and hypothesis testing for a model of avalanche behaviour**, IHRR, Durham, Jan 2011. Slides available. Please note that this is a preliminary analysis and we are currently investigating a more sophisticated statistical model for the relationship between the HB parameters and the measurements.**Inference using large climate simulators: HadCM3 and North American Mid-Holocene temperature anomalies**, Liverpool Marine Symposium, Jan 2011. Slides available. The slides for the talk at the MPI-M (Mar 2011) are more detailed.**Accounting for the limitations of quantitative models**, SECaM, Exeter, Jan 2011. Slides available.**Stochastic dynamical modelling in Environmental Science**, AMSTAT, Warwick, Oct 2010. See the slides for the Exeter Jan 2011 seminar.**What can pollen tell us about palaeo-climate?**, ICAS, Leeds, Oct 2010. Slides available (more detailed).**Is Building Bigger and 'Better' Climate Change Models a Poor Investment?**, CRASSH conference, Clare College Cambridge, 29 Sep 2010. text available, I was proposing the motion. Watch the debate.**Probabilistic frameworks**, INI CLP workshop on probabilistic climate prediction, Exeter, Sep 2010. See the slide! Watch the presentation (multiple formats).**A new statistical framework for analysing multi-model ensembles**, IMS 2010, Gothenburg, Aug 2010. Slides available.**A new statistical framework for analysing multi-model ensembles**, 11th IMSC, Edinburgh, July 2010. Slides available.**What can pollen tell us about palaeo-biomes?**, 11th IMSC, Edinburgh, July 2010. See the slides for the ICAS Oct 2010 seminar.**Assessing model limitations**, Governing Through Uncertainty meeting, Bristol, May 2010. Slides available.**Complex systems: Accounting for model limitations**, Research Students' Conference 2010, Warwick, April 2010. slides available.**Complex systems: Accounting for model limitations**, St Andrews, Mar 2010. slides available.**Model limitations: Sequential data assimilation with uncertain static parameters**, Edinburgh, Mar 2010. Slides available.**Uncertainty and risk in natural hazards**, Nottingham, Nov 2009. Slides available.**Introduction to computer experiments, and the challenges of expensive models**, RSS Statistical Computing Section, Bath, October 2009. Slides available.**Quantifying uncertainty in probability of exceedence (PE) curves**, NERC/KTN workshop on Catastrophe modelling for natural hazard impact, Lloyd's, London, Oct 2009. Slides available and also R code for the pictures.**Emulator-based simulator calibration for high-dimensional data**, JCGS session, JSM-2009, Washington DC, Aug 2009. Slides available.**The What, Why, and How of Multivariate Emulation**, Spring Research Conference On Statistics in Industry and Technology, Vancouver, May 2009. Slides available.**Simple models for glacial cycles**, Bath Institute for Complex Systems, Feb 2009. Slides available.
## Donkeys!Joint work with Kate Milner.Donkeys are extremely hard to weigh, but easy to measure with a measuring tape. Because weight is a important factor in health and in veterinary care, statistical models are used to predict weight on the basis of measured hearth girth and height. The Donkey Sanctuary sponsored Kate Milner to go to Kenya with a large weighing scales and collect data on donkeys. Together we are analysing the results, with the intention of producing an improved 'nomogram' (or similar) showing how to predict a Kenyan donkey's weight on the basis of its height and heart girth, and also age and condition. Beautiful picture of two Kenyan donkeys, taken by Kate Milner. |