My research expertise is in extreme value theory, methods and its applications. My research is driven by new problems arising from applications such as oceanography, hydrology, climatology, reliability, economics, reinsurance, finance, medicine and sport.
2017: Invitation to Royal Garden Party and Buckingham Palace in recognition of service to UK Flood Resilience Review.
2019: Lancaster University Staff Prize: Outstanding Contribution Award, (joint with Professors Eckley and Glazebrook) for leading three successful CDT bids.
Research in Extreme Values
A wide interest range of in extreme value theory, methods and applications. This is the study of rare events enabling the estimation of the probability of the occurrence of events that are more extreme than those already observed.
Theory and methods: multivariate extreme value and joint probability problems, time series and covariate modelling for extreme values, spatial modeling, and structuring asymptotically justified models to capture process structure. Specifically the methodology developed with collaborators provides:
the first statistical inference for multivariate extremes for componentwise maxima and multivariate peaks over thresholds methods;
identifies a class of dependence structures, asymptotic independence, not covered by all previous statistical and probability theory for multivariate extremes;
develops dependence measures for multivariate extremes that have become the standard in a range of application areas;
a new approach to multivariate extremes based on conditioning that opens multivariate extremes to general dimensions and broad dependence structures.
efficient inference methods for max-stable processes and new models for asymptotically independent spatial processes.
Applications: environmental modelling in hydrology and oceanography and meteorology, financial modelling, medicine and sports records.
Other Research Interests
Extreme value theory and its applications to environmental processes and finance. Environmental statistics, inference methods, asymptotics, spatial statistics, reliability and copulas, non-regular estimation, and modelling complex stochastic systems.
Research Impact
My research has had substantial impact with applications which have effected UK and global policies in terms of safeguarding lives, homes and infrastructure. It led to 2 Impact Case Studies in REF2013 (2 of 3 for Mathematics UoA), both 4* scored, and will give 2 Impact Case Studies in REF2020 (2 of 5 for Mathematics UoA). Examples include:
Developing the statistical theoretically method and deriving the statistical estimates for optimising the height of all coastal flood protection schemes in the UK, influencing a total spend of $0.9B on 900 schemes over the period 2008-13. (JBA, NOC, Royal Haskoning, EA). Methods are still in use and in 2018 I was part of a team that updated these;
Providing evidence of fundamental importance on the likely cause of the sinking in 1980 of the MV Derbyshire to the $11M High Court Reopened Formal Investigation. Reasons for its sinking, the largest UK ship to have sunk, were a mystery until the High Court Investigation. I was asked to be an expert witness when it was clear that earlier statistical analysis were contradictory with the extensive evidence collected. The statistical problem was to combined information on the waves encountered, their likely impact on the ship and the ship condition and how it was controlled. The statistical analysis was key to the investigation and was treated as the fundamental evidence in the Judge’s conclusions;
Setting new worldwide mandatory design standards for bulk carriers, ore carriers and combination carriers. Specifically, the strength of hatch covers has been increased by 35% from the previous design standards. In 2008-13 this impacted on the design of 1720 new bulk carriers, and strengthening as well as new inspection and maintenance procedures for the 5830 previously built bulk carriers (Lloyds’ Registry of Shipping). The design standards are still in place;
Optimising oshore oil-rig designs to protect them from jointly occurring directionally varying extreme waves heights and other extreme met-ocean variables (Shell). This work has had major impacts across oshore risk assessment and is the core for an Impact Case Study in REF2020;
Developing the first rigorous tools to undertake spatial flood risk assessment for the insurance and reinsurance industries. The tool uses the Heffernan and Tawn (2004) method to model widespread river flooding covering large numbers of rivers simultaneously and at various time lags (JBA);
With JBA, I have provided vital work to inform the UK Risk Register and the 2016 Government National Flood Resilience Review. Analyses included studies of storm events likely to hit the Humber and the Thames rivers and for estimating the probability of somewhere in the UK receiving a 100 or 1000 year event (UK Government). This work will form the basis of an Impact Case Study in REF2020;
Identifying the effect of climate change on: extreme sea levels and waves over the North Sea which is of critical importance to future oshore and coastal infrastructure (Shell), and work on European temperature records to inform the UK Government of the present and future risk of heatwaves, prolonged periods of extreme heat (Met Office).
Research Training Grant
2017-22: Director of STOR-i CDT the joint EPSRC, industry and LU funded doctoral training centre for statistics and operational research.
2010-17: PI on the £4.5M grant from EPSRC (with CoI: Glazebrook and Eckley) for STOR-i.
Research with Government and Industry
Substantive research projects has been undertaken with the following Government agencies, multi-nationals and SMEs:
AstraZeneca
ATASS
CEH Wallingford
Department for Environment, Food and Rural Affairs
Department for Transport
EDF
Environment Agency
Forensic Science Service Ltd
HR Wallingford
JBA Consulting
JBA Risk
Lloyds Registry of Shipping
MAN Investments
Met Office
Mouchel and Partners Ltd
National Oceanography Centre
Royal HaskoningDHV
Shell
Sparx
Supervision of postgraduate and postdoctoral students:
I put a great importance on the development of PhD students and Post-Doctoral researchers. All my PhD students have successfully completed within 4 years, most a year earlier. About half my students are now in academic posts and the rest are in industry. Coles, Ledford and Wadsworth were the 1st, 3rd and 12th recipients respectively of the RSS prize for PhD research (awarded every two years). Wadsworth won a PhD Plus prize and Dean’s award for the best 3rd year PhD student in the Faculty at Lancaster University. Papastathopoulos was awarded the Eleneio Doctoral Thesis Award in Statistics by Greek Statistical Institute for the best dissertation written by a Greek statistician for the two year period 2011-2012.
Previous Ph.D. Students:
Stuart Coles (1991). Statistical Methodology for the Multivariate Analysis of Environmental Extremes.
Stuart’s PhD developed new models and inference for the point process approach to multivariate extreme values and also developed the first threshold-based modelling strategy for max-stable processes. Stuart was the first recipient respectively of the RSS prize for PhD research (awarded every two years).Since his PhD he has held lecturing posts at the universities of Nottingham, Lancaster, Bristol and Padova. He authored the Springer book: An Introduction to Statistical Modeling of Extreme Values, which is of one of best and most accessible extreme values books with over 1200 citations. He currently works for Smartodds a statistical research and sports modelling company.
Saralees Nadarajah (1994). Multivariate Extreme Value Methods with Applications to Reservoir Flood Safety. [Joint supervision with C. W. Anderson].
Saralees’ PhD developed multivariate extreme value methods for reservoir safety by accounting for dependence between rainfall events, previous rainfall and winds. His work also looked at the impacts of ordering constraints of extreme values which is relevant when rainfalls of different temporal windows are being considered. Saraless’ PhD was funded by CEH Wallingford.
Since his PhD he has held a series of lecturing posts in the UK and USA and is currently senior lecturer at Manchester University. He has published three text books on distribution theory, with one, Extreme Value Distributions: Theory and Applications with Kotz a valuable review of extreme value theory. He has a highly prolific publication track record in a wide range of distribution theory.
Marks’ PhD and post doc work was focused on developing methods for a systematic spatial analysis of UK sea level extremes and applying them to a rich network of observational and synthesized data. His work had major impact in optimising the height of all coastal flood protection schemes in the UK and in explaining and overcoming the poor performance of maximum likelihood inference for extreme value problems. Mark’s PhD was funded by the MAFF via the Proudman Oceanographic Laboratory.
Since his PhD he held lecturing posts at Newcastle and City universities. While doing his PhD he developed an interest in sports modelling leading to an appearance on the BBC’s popular science programme Tomorrow’s World. This research has expanded substantially and he since founded and is the director of ATASS a company with a portfolio of enterprises in research, sport and education technology. However for most statisticians the company is associated with its highly successful sport statistics and betting.
Anthony Ledford (1995). Dependence within Extreme Values: Theory and Applications.
Anthony’s PhD developed the first probability theory and statistical methods for the asymptotic independence class of distributions that are important in multivariate extremes. Anthony was the third recipient of the RSS prize for PhD research (awarded every two years). Anthony’s PhD was funded by the CEH at Merlewood.
Since his PhD he became a reader in statistics at the University of Surrey before joining MAN Investments , the world’s leading hedging company. He is currently chief scientist of the AHL component of the company (the world-leader in systematic trading strategies) and has overall responsibility for AHL’s strategic research undertaken at the Oxford-Man Institute of Quantitative Finance.
Paola Bortot (1997). Analisi Della Dipendenza Tra Valori Estremi. [External supervision for University of Padova].
Paola’s PhD focused on exploring the use of asymptotically independence models for time series extremes and oceanographic extremes. Her work led to the new methods for assessing the impact of wave extremes in conjunction with still water sea level extremes for design of coastal flood protection through its inclusion in the JOINSEA software produced by HR Wallingford.
Paola’s PhD was co-funded by the MAFF via HE Wallingford.
Since her PhD, Paola has been at the University of Bologna working on a mixture of extreme value methods and biostatistics. Her homepage is: www2.stat.unibo.it/bortot/
Louise Harper (1997). Model-based Geostatistics in Environmental Sciences [Joint supervision with P. J. Diggle].
Louise’s PhD was on spatial generalized linear models and spatial latent process models for extremes essentially paving the way for the new area termed mode-based geostatistics.
Since her PhD she has worked for Westlakes and now works as a statistician for AstraZeneca.
Her married name is Louise Garside.
Mike Robinson (1997). Statistics for Offshore Extremes.
Mike’s PhD developed a range of methods for the analysis of sea-current extremes. This involved research of directional extremes and on the impact of changes in sampling rates on extremes. He also produced some high profile research on the validity of the world record achieved by the Chinese athlete Wang Junxia. Mike’s PhD was co-funded by the Proudman Oceanographic Laboratory.
Since his PhD he became a Post-Doc in statistics at the University of Surrey before joining MAN Investments , the world’s leading hedging company. He is currently Head of Directional Strategies of the AHL component of the company (the world-leader in systematic trading strategies).
Miguel Ancona-Navarrete (2000). Dependence Modelling and Spatial Prediction for Extreme Values.
Miguel’s PhD examined the impact of different assumptions on the estimation of the extremal index and was the first to explore asymptotic independence as a tool to examine pairwise extremal dependence in space.
Since his PhD he worked as a quantitative analyst at MAN Investments and is currently doing a similar role at Aviva Investor Global Services Ltd.
Carl Scarrott (2003). Reactor Modelling and Risk Assessment, [Joint supervison with G. Tunnicliffe-Wilson].
Carl’s PhD looked at the spatial variations in temperatures in the core of a nuclear reactor and cooling strategies to stabilize these. This involved a range of extreme value analyses from marginal to spatial, however the major research was on spectral analysis. Carl’s PhD was co-funded by British Nuclear Fuels Ltd.
Since his PhD, Carl has worked as a researcher for CEH Wallingford before taking up a lectureship post at the University of Canterbury, New Zealand, where is currently is senior lecturer. His research on extreme values has continued and his recent work has included accounting for threshold uncertainty.
Fabrizio’s PhD looked at extremes of financial time series, in particular focusing on deriving new estimators for the extremal index which are tuned for the variation typical for such series and for deriving the formula for the extremal index for a GARCH(1,1) process.
Since his PhD, Fabrizio has continued to work in financial extreme values and time series and has developed a range of semiparametric methods to handle non-stationarity. He is currently an assistant professor in Applied Economic Statistics at the University of Parma.
Chris’ PhD looked at a range of approaches for estimating dependence in extremes in time series and multivariate contexts. His key results concerned a new approach looking at inter-arrival timess of threshold exceedances to estimate the extremal index that did not require model assumptions or clusters to be identified in advance. He also examined within cluster behavior using the theory of strings.
Since his PhD, Chris has been a researcher in statistical meteorology at the universities of Reading and Exeter. He is currently a senior lecturer in Statistics at Exeter University, where he primarily researchers into assessing the accuracy of weather and climate model forecasts, often with an extreme value focus.
Alec’s PhD explored a range of different topics for extremes including: Bayesian estimation of univariate distributions to account for the three types, simulation methods for multivariate extremes, and inference for componentwise maxima when the dates of the maxima are known. During and post his PhD Alec has played a key role in disseminating extreme value methods but putting out a series of packages in the open source language R (evd, evdBayes, evir, ismev)
Since his PhD he has held posts at Macquarie University and Swinburne University, Australia and been Assistant Professor, National University of Singapore where he has continued to research in extreme values. During this period he has been highly successful in a number of the Kaggle competitions. Currently he is a research scientist, CSIRO, Melbourne, Australia.
Adam Butler (2005). Statistical Modelling of Synthetic Oceanographic Extremes. [Joint supervison with J. Heffernan].
Adam’s PhD focused on modelling decadal scale non-stationarity in sea-level extremes over the North Sea using gridded synthetic data over the entire sea. He found clear spatial patterns for trends over this time period which help when modelling non-stationarity observed at a single site. Adam’s PhD was co-funded by the Proudman Oceanographic Laboratory.
Since his PhD he has been an environmental statistician at Bioss in Edinburgh, where his current post is senior statistician.
Mark’s PhD looked at dimension reduction for spatial extremes when applied to synthetic sea level data from the North Sea and at temporal clustering under the conditional multivariate extremes approach including mixture distributions. Mark’s PhD was co-funded by the Proudman Oceanographic Laboratory.
Since his PhD he worked as a quantitative analyst at MAN Investments and is currently the sports analytics manager at Decision Technology, a research consultancy dedicated to the study of human decision-making. The company produce the Fink Tank Predictions, that features regularly in the Times newspaper, which provides up-to-date forecasts and ranking systems for English club football.
Tilman’s PhD was motivated by concerns about the safety of high-speed trains from strong winds. His work looked at modelling dependence in extreme winds speeds over different sites whilst accounting for their directions. He also developed a solution to resolve the masking of extreme wind speeds in different directions within an observation period.
He is now a postdoctoral researcher at the CIBER in Epidemiology and Public Health, attached to the Granada Cancer Registry in Spain.
Caroline Keef (2007). Spatial Modelling of Extreme River Flows.
Caroline’s PhD developed models for extremes river flows over networks of different rivers and explore the hydrological factors that led to strong spatial dependence. She used methods for conditional multivariate extreme values and extended the methods to allow for missing data and multivariate-temporal dependence. Caroline’s PhD was funded by the CEH Wallingford.
Since her PhD, Caroline has worked as a senior statistical at JBA Consulting working on the implementation of the methods of her PhD thesis to enable estimation of the loss distribution for insurance companies from widespread flooding and providing information for emergency planners. Caroline is currently working as a statistician for Yorkshire Water.
Bakri Adam (2007). Extreme Value Modelling for Sports Data.
Bakri’s PhD looked at modeling fastest swimming times over different events at different distances. The inter-connections between the events lead to natural ordering which he was able to impose on the extreme value extrapolations giving improvements in efficiencies.
Bakri is now a statistics lecturer at University of Putra in Malaysia and continues to work on extreme values and reliability.
Paul’s PhD focused on estimation methods in hydrology when the model was sufficiently unknown that classic Bayesian inference methods could not be used.
This was pre-ABC methods but he developed methods based on various objective functions of fit to allow uncertainty in hydrological modelling to be account for in predictions.
Since his PhD, Paul has been a statistical hydrologist post-doc at the Lancaster Environment Centre at Lancaster University.
Emma’s PhD developed a new approaches to estimate non-stationarity in extremes, the use of an extension of the GPD to account for temporal dependence for modelling cluster maxima, and hierarchical multivariate extreme distributions for air pollutants. During her subsequent period as a post doc she also developed random effects and latent process models to explain temporal clustering of apparently independent extreme values for river flows.
Since her PhD and post-doc, Emma has been a statistics lecturer at Lancaster University researching on extreme values.
David’s PhD developed particle filter and associated smoothing methods for univariate extreme value problems to enable estimation and prediction of extremes of non-stationary extremes.
Since his PhD, David has been a statistician at HR Wallingford working on flood risk management and modelling fish populations.
Sawsan Abbas (Hilal) (2010). Statistical Methodologies for Financial Market Risk Management.
Sawsan’s PhD focused on modelling multivariate time series of financial series with the aim to optimize a portfolio or investments to reduce the risk of large losses whilst making profit. It also looked at ways of hedging and account for the dependence in the extremes when hedging.
Since her PhD, Sawsan been a statistics lecturer at the University of Bahrain and continues to research in extreme values. She is now an associate professor.
Jenny’s PhD explored a range of fundamental issues in extremes including: what scale is best to extrapolate on and how do you incorporate the uncertainty of this choice in the inference; incorporating threshold uncertainty into quantile estimation; a new approach to modelling asymptotic independence in multivariate cases; and spatial dependence models which extend existing max-stable processes to incorporate asymptotic independence. Wadsworth won the Faculty PhD Plus prize and Dean’s award for the best 3rd year PhD student in the Faculty. Jenny’s PhD was co-funded by Shell.
Since her PhD, Jenny has been a post-doc at EPFL and started as a statistical research fellow at Cambridge in late 2013.
Ioannis’ PhD was motivated from the need to build statistical models that incorporate scientific knowledge and capture the characteristics of extreme multivariate laboratory variables that show potential in identifying signals of death induced by drug medication. Major challenges related to the small sample sizes of Phase II studies that limits the applicability of standard extreme value methods. He developed extensions of standard extreme value techniques that offer additional structure for the main body of the distribution of the data and improve the estimates of key extremal quantities. Ioannis’ PhD was funded by AstraZeneca.
Ioannis was awarded the Eleneio Doctoral Thesis Award in Statistics by Greek Statistical Institute for the best dissertation written by a Greek statistician for the two year period 2011-2012.
Since his PhD he has been a Brunel Research fellow in statistics at Bristol University working on graphical methods for extreme values.
Ye’s PhD focused on multivariate extreme value dependence modelling for copulas for use in finance for portfolio selection. He developed new factor models for the dependence structure of multivariate extremes and found a powerful diagnostic technique for model selection between stochastic volatility models and GARCH models when in extreme states. Ye’s PhD was co-funded by MAN Investments.
Since his PhD, Ye has been a senior analyst in JBA Consulting’s Risk Management team where he works on spatial dependence in flooding.
Darmesah Gabda (2014). Efficient inference for nonstationary and spatial extreme value problems.
Darmesah’s PhD looked at the problems of the analysis of environmental extreme values when only short series are available. She looked to see what could be gained by marginal and spatial penalty functions, exploring dependence on the mean when trends are present, and exploiting information form climate models.
Since her PhD, Darmesah has been a statistics lecturer at the University Malaysia Sabah and continues to research in extreme values.
Ross Towe (2015). Modelling the Extreme Wave Climate of the North Sea [Joint supervision with E. Eastoe]
Ross’ PhD was motivated by the offshore industries need to know what the future extreme wave climate will be like. He used extreme value statistical downscaling methods to exploit information from global climate change models to relate to the distribution of extreme waves. Ross’ PhD was funded by Shell.
Since his PhD, Ross has been a working as a Research Associate jointly between Lancaster University and JBA Consulting where he works on spatial flooding through joint modelling of river flows and rainfall.
Hugo Winter (2015) Extreme Value Modelling of Heatwaves.
Hugo’s PhD developed extreme value methods to estimate the frequency of heatwaves in space and time. He also modelled changes in the characteristics of the type of heatwaves due to future climate change through using extreme value models coupled with global climate model outputs. Hugo’s PhD was co-funded by the Met Office.
Since his PhD, Hugo has been a working as a research engineer within the EDF Energy UK R&D Centre in London. His work there involves working within the natural hazards team and leading on the programmes concerning extreme weather, coastal flooding and hazard combinations.
Tom Flowerdew (2016). Methods for the Identification and Optimal Exploitation of Profitable Betting [joint supervision with K. Glazebrook and C. Kirkbride]
Tom’s PhD was concerned with the optimal exploitation of betting opportunities in sports betting markets. He looked at different betting strategies, modelled the learning process from previous outcomes, and accounted for the associated uncertainty in estimating probabilities. His findings also apply to the financial markets. Tom’s PhD was co-funded by ATASS Sports.
Since his PhD, Tom has been working as a data scientist for Featurespace, a small company based in Cambridge who specialises in building adaptive behavioural analytic models, which are used to catch fraud in financial services, insurance and gaming
George Foulds (2016). A Study of Home Advantage in Football and Other Contributions to Sports Data Analysis [joint supervision with M. Wright and R. Brooks]
George’s PhD primarily investigated the factors which influence home advantage in football. He conducted a study across all Europe’s top leagues using data on results, goals scored, and cards awarded. Much of the work built on the Dixon and Coles model to capture the Poisson nature of such outcomes.
George’s PhD was co-funded by ATASS. Since his PhD, George has worked at Summit Media, a company working in online retail, first as a data scientist and now as Head of Data Science.
Christian Rohrbeck (2016). Statistical Methods for Weather-related Insurance Claims [joint with D. Costain and E. Eastoe]
Christian’s PhD developed extreme value and monotonic nonparametric regression methods for the statistical analysis of environmental and financial data. He analysed regionally aggregated flooding related insurance claim data across Norway and investigated how this could be explained by meteorological events. He developed methodology for extremes of discrete (count) variables which joined with non-extreme models. He further investigated the spatial heterogeneity and potential for pooling extreme value data across regions.
Christian’s PhD project involved a strategic partnership with Oslo University, and specifically co-supervision by Arnoldo Frigessi. Since his PhD, Christian has worked as a post-doctoral researcher at Lancaster University, was awarded a prestigious AXA Research Fund postdoctoral grant, and is now a Statistics lecturer in the Department of Mathematical Sciences at Bath University, where he continues to research in environmental multivariate/spatial extreme values.
Monika Leng (nee Kereszturi) (2016). Assessing and Modelling Extremal Dependence in Spatial Extremes.
Monika’s PhD was the first to look at spatial dependence of extreme wave heights, and to identify different forms of asymptotic independence and dependence with both inter-site distance and directionality. Monika also explored the previously unstudied lower tail dependence of copulas in the max-linear class, with previous focus being on the much simpler upper tails.
Monika’s PhD was co-funded by Shell. Since her PhD, Monika (now under her married name of Monika Leng) has worked as an environmental statistician for JBA risk Management (environmental flood risk), and now works as a data scientist for Equifax, UK.
Paul Sharkey (2018). Statistical Models for Extreme Weather Events.
Paul’s PhD developed extreme value models for extreme extratropical windstorms. Unlike standard extreme value analyses, where the observation locations are fixed in space and storms pass over them, Paul’s work took a Lagrangian approach to the analysis of windstorm events with the spatial properties of the storm, its temporal evolution and its track being modelled. With this model, the implications for the windstorm process at fixed observational sites can be inferred. This approach had benefits for capturing spatial-temporal physical properties of the meteorology and incorporate features not observed at observational networks.
Paul’s PhD was co-funded by the Met Office. Since his PhD, Paul has worked as an environmental statistician for JBA Consulting (environmental flood risk), as a data scientist for the BBC, and as a Senior Statistical Scientist at Cervest (a company that develops products to helps users to manage and adapt to climate risk).
Thomas Lugrin (2018). Flexible Bayesian Semiparametric Risk Estimation for Complex Extremes. [External supervision for EPFL with Anthony Davison]
Thomas’s PhD focused on improving understanding of conditional multivariate extreme value models. He studied the convergence rates to these models and developed alternative model formulations to improve convergence. He also developed the first Bayesian semi-parametric inference schemes for such models.
Since his PhD, Thomas has worked as post-doctoral researcher at EPFL on extreme values, a data scientist for the asset manager company Schroder Adveq and an analyst for the Department of Defence on the Swizz Federal Administration.
Emma Simpson (2019). Classifying and Exploiting Structure in Multivariate Extremes [joint with J. Wadsworth]
Emma’s PhD had two areas of focus. The first was on identifying important structure in multivariate extreme value problems from data. Specifically, the target was to classify which subsets of a vector variable could simultaneously reach their extremes from the same observation, and to estimate the probability of such subsets from overall extreme events. The second focused on theoretically deriving extremal dependence properties from the increasing popular vine copula families, to see how higher-order dependence properties were restricted through the pairwise construction process.
Since her PhD, Emma has worked as post-doctoral researcher at Lancaster University and then as a lecturer in the Department of Statistical Science at University College, London, where she continues to research in environmental multivariate/spatial extreme values.
Christina Wright (2019). Modelling and Inference for the Travel Times in Vehicle Routing Problems [joint with M. Nunes, K. Zografos]
Chrissy’s PhD explored models for traffic flow to use inside vehicle routing problems. The element of Chrissy’s project I supervised linked to extreme travel times which were affected by accidents or other delays. Chrissy’s work explored the causes for the frequency, sizes, and clustering of such large delays.
Since her PhD, Chrissy has been a transport modeller at Tetra Tech Europe, directly using her PhD statistical modelling skills for traffic flow.
Robert Shooter (2019). Conditional Models for Spatial Extremes [joint with J. Wadsworth]
Rob’s PhD focused on the modelling of extreme wave height sea-states to provide information for risk assessments for offshore industries. His work specifically focused on spatial extremes, developing highly efficient semi-parametric models for the conditional distribution of wave fields, given an extreme value is observed at a given site. A novelty in this work was that models could handle both asymptotic dependence and asymptotic independence, transitioning between the two at appropriate spatial lags.
Rob’s PhD was co-funded by Shell. Since his PhD, Rob has been a statistical researcher at the Met Office, directly using his PhD skills through his work on extreme meteorological event modelling.
Oliver Hatfield (2019). Statistical Methods for Detecting Match-Fixing in Tennis [joint with C. Kirkbride]
Oliver’s PhD investigated match-fixing in tennis by analysing betting odds for anomalies using Gaussian processes and state space modelling. The analysis followed the betting markets through pre-match and in-match behaviours and looked for departures from natural updates that reflected current match score or patterns in volumes of bets. He could draw particularly clear conclusions given the high-quality data available from his co-funders, ATASS.
Since his PhD, Oliver has worked as a statistical consultant at Lancaster University and is currently a statistician and data analyst at the Institute for Mathematical Innovation, University of Bath.
Toby Kingsman (2020). Optimisation of Scheduling and Routing for Oshore Wind Farm Maintenance [joint with B. Boyaci]
Toby’s PhD developed operational research tools and statistical modelling for MetOcean risk management. The aim was to optimise complex marine operations, such as a list of operations and maintenance objectives, for a spatially dispersed site such as an offshore wind farm. The targets are to minimise the costs for undertaking these tasks under resource constraints and to identify the most favourable time to undertake these operations whilst accounting for uncertainty in forecasts and hindcasts. Toby’s PhD was co-funded by JBA Consulting.
Since his PhD, Toby has worked as an operations research analyst at Microlise (a logistics/AI company), then at Tesco.
Anna Barlow (2021). Flood Events: Extreme Value Problems and Efficient Estimation of Loss [joint with C. Sherlock]
Anna’s PhD was concerned with overcoming practical and theoretical problems of implementing the conditional multivariate extremes approach of Heffernan and Tawn for the modelling of widespread flood events and from this model using Monte Carlo methods for deriving household insurance risk over thousands of locations. Anna’s PhD was co-funded by JBA Risk.
Since her PhD, Anna has worked as post-doctoral researcher at Lancaster University continuing to research in environmental multivariate extreme value problems.
Jordan Richards (2021). Extremes of Aggregated Random Variables and Spatial Processes [joint with J. Wadsworth]
Jordan’s PhD developed extreme value methods to study both spatial dependence and extremes of aggregated variables. The driving force for his work is that, for rainfalls, it is aggregates over space and time that matters. His work identified different properties of extremes of rain-fields for convective and frontal rainfalls and exploited this mixture structure.
Jordan’s PhD was co-funded by the Met Office. Since his PhD, Jordan has been working as a postdoctoral researcher in spatial extremes with environmental applications at KAUST (King Abdullah University of Science and Technology), in Saudi Arabia.
Zak Varty (2021). Statistical Modelling of Induced Earthquakes [joint with P. Atkinson]
Zak’s PhD focused on the modelling of induced seismicity, particularly the temporal aspects. Earthquake data exhibit variable data quality with considerable incompletely observed data due to changes in measurement precision. Zak’s focused on a combination of earthquake magnitude modelling and inference for the temporal clustering of events. His work gave substantial efficiency gains by identifying a time-varying earthquake magnitude level above which all earthquakes in the field are certain to have been recorded.
Zak’s PhD was co-funded by Shell, who have followed up Zak’s project with a project with Conor Murphy, see below, with Zak being co-supervisor. Since his PhD, Zak has been a Teaching Fellow in Statistics in the Department of Mathematics at Imperial College, London.
Yanyun’s PhD looks at covariate modelling of wave extremes and the spatial dependence of wave fields. Specifically her current research focuses on non-stationary max-stable processes. Yanyun is a part time PhD student funded by Shell.
Harry’s PhD focuses on ranking methods, with a particular focus on sport. He first developed an overall ranking scheme across swimmers in different distances, strokes, dates of swims, and gender categories; with the ranking based on an extreme value model of all top swimming times. His second project developed new models for pairwise comparisons, extending the widely used Bradley-Terry, to account for intransitivity between pairs – enabling the model to capture different strategies of play dependent on the opponent. During his PhD he has undertaken a year-long internship with Roche in Basle. Harry’s PhD is co-funded by ATASS.
Stan’s PhD aims to develop multivariate and temporal extreme value methods for modelling extreme offshore conditions. In research to date, he has worked on bivariate problems with a mixture structure, derived extreme value properties when distributions are expressed conditionally, not via copulas, and compared multivariate extreme models against industry standards for waves heights and period. Stan’s research is co-funded by Shell, with whom he did a Masters’ project prior to coming to STOR-i.
Eleanor’s PhD is looking at extreme value methods for estimating events with long return periods for the nuclear industry. For such cases capturing small non-stationarities and seasonality is vital. To date, Eleanor has worked on extreme sea-level inference, accounting for previously ignored differences in the seasonal patterns in the tides and skew surges and addressing key questions on the dependence between skew surges and tide (which previously was assumed to be independent). Eleanor’s research is co-funded by EDF (with the project instigated by STOR-i PhD graduate, Hugo Winter).
Daire Healy (2nd Year at Maynooth, joint with Andrew Parnell)
Daire’s PhD is looking at temporal non-stationarity in spatial climate extremes over Ireland. The work aims to bring together, in a unified framework, information about spatial extreme events which utilises both observational data and output from climate models. His work aims at quantifying how spatial extreme events are changing over time given the lack of a unified measure of extremity in the spatial context. Daire is based at Maynooth University, and I co-supervise him jointly with Andrew Parnell.
Conor’s PhD focuses on the modelling of induced seismicity, particularly the spatio-temporal clustering of extremal events. Earthquake data exhibit variable data quality with considerable incompletely observed data due to changes in measurement precision. This is of particular concern with induced earthquakes which occur at shallow depths with low magnitudes. Conor’s work involves developing spatio-temporal threshold selection methods to avoid biases due to incompletely observed data. The techniques will have generic use in spatial extremes threshold selection. Conor’s research is co-funded by Shell, working there with STOR-i PhD Graduate, Ross Towe, and co-supervised by STOR-i PhD graduate, Zak Varty, at Imperial College.
John worked of multivariate extreme value methods for sea levels and waves to quantify the benefits of joint probability approaches over direct univariate methods.
Since his post doc he has been a lecturer at the University of Central Lancashire, a statistician at Unilever and is currently an environmental statistician at Plymouth Marine Laboratory.
Paola’s PhD focused on exploring the use of asymptotically independence models for time series extremes and oceanographic extremes. Her work led to the new methods for assessing the impact of wave extremes in conjunction with still water sea level extremes for design of coastal flood protection through its inclusion in the JOINSEA software produced by HR Wallingford.
Paola’s PhD was co-funded by the MAFF via HE Wallingford.
Since her PhD Paola has been at the University of Bologna working on a mixture of extreme value methods and biostatistics. Her homepage is: www2.stat.unibo.it/bortot/
Marks’ PhD and post doc work was focused on developing methods for a systematic spatial analysis of UK sea level extremes and applying them to a rich network of observational and synthesized data. His work had major impact in optimising the height of all coastal flood protection schemes in the UK and in explaining and overcoming the poor performance of maximum likelihood inference for extreme value problems. Mark’s PhD was funded by the MAFF via the Proudman Oceanographic Laboratory.
Since his PhD he held lecturing posts at Newcastle and City universities. While doing his PhD he developed an interest in sports modelling leading to an appearance on the BBC’s popular science programme Tomorrow’s World. This research has expanded substantially and he since founded and is the director of ATASS a company with a portfolio of enterprises in research, sport and education technology. However for most statisticians the company is associated with its highly successful sport statistics and betting.
Martin worked on max-stable processes, developing an R package () for simulation, and theory for the extremal function. For multivariate extremes he quantified the inter-connections between dependence on different variables in multivariate extreme value distributions, found counter examples in multivariate extremes, and linked the central limit theory with extreme value theory by considering the norm of a vector as a function of the number of variables.
Since his post doc he has been a lecturer at the universities of Luxenburg, Hamburg and professor at the University of Gottingen. He is currently professor of mathematical statistics at the University of Mannheim, Germany where he is still highly active in multivariate and spatial extreme value theory.
Keming worked on an extension for model-based geostatistics to address the problem of high dimensionality of sites and the consequent computational problems that imposed.
Since his post doc he was a lecturer at Plymouth and is now reader at Brunel where he is known for his work on quantile regression.
Janet developed a new approach to multivariate extremes that enables high dimensions and a broad range of dependence structures to be accounted for. The approach was based on conditioning on one component of the vector variable being extreme. She also worked on the investigation of the sinking of the MV Derbyshire and the subsequent development of new safety standards.
Since her post doc she was a lecturer in statistics at Lancaster University before becoming an independent statistical consultant, trading as J. Heffernan Consulting.
Miguel’s PhD examined the impact of different assumptions on the estimation of the extremal index and was the first to explore asymptotic independence as a tool to examine pairwise extremal dependence in space.
Since his PhD he worked as a quantitative analyst at MAN Investments and is currently doing a similar role at Aviva Investor Global Services Ltd.
Emma’s PhD developed a new approaches to estimate non-stationarity in extremes, the use of an extension of the GPD to account for temporal dependence for modelling cluster maxima, and hierarchical multivariate extreme distributions for air pollutants. During her subsequent period as a post doc she also developed random effects and latent process models to explain temporal clustering of apparently independent extreme values for river flows.
Since her PhD and post-doc Emma has been a statistics lecturer at Lancaster University researching on extreme values.
Olivia worked on covariate modelling for extreme river flow data building in covariates on rainfall and soil moisture.
Since her post doc she runs an online tutoring business mA+hswork
Ross Towe 2014 & 2014-16 [with Sherlock]
Christian Rohrbeck 2017-19
Emma Simpson 2018-21
External Examination of PhDs
External examiner for PhDs at Dublin, Kent, Glasgow (2), Oxford, Newcastle, Nottingham, Sheffield and UMIST, Rotterdam (2), Lisbon, EPFL (4), Halifax Nova Scotia, Munich.
Organisation of Conferences and Research Programmes
1989: British co-ordinator for the 6th European Young Statisticians Meeting.
1993: Spruce II organising committee member.
1999-2001: Member of the RSS 2001 Conference Planning Committee.
2009: Joint organiser of a 6 month research programme on Risk, Rare, Events and Extremes at the Bernoulli Centre EPFL, following the award of £120K.
2011: Joint organiser of Environmental Risk and Extreme Events a conference in Monte Verita following award of £25K
Internal Research Related Responsibilities:
1996: Joint responsibility for the Statistics RAE return
2001: Responsible for the Statistics RAE return
2010: Appointment panel member for Faculty Associate Dean for Research
2013-7: Executive Board member of Impact Accelerator Account
Editorial work
1997: Associate Editor of Extremes journal.
1998-2002: Associate Editor of Applied Statistics journal.
By continuing to use the site, you agree to the use of cookies. more information
The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.
