About







Jordan Richards, Ph.D.

I am a lecturer in statistics at the School of Mathematics, University of Edinburgh. I am mainly interested in the intersection of extreme value theory, spatial statistics, and deep-learning, with a particular focus on applications to natural hazards and extreme climate risk.

In 2017, I joined the STOR-i centre for doctoral training, and there I completed both a Masters of Research and a Ph.D. in Statistics and Operational Research, with the latter under the supervision of Jonathan Tawn and Jennifer Wadsworth, both of Lancaster, and Simon Brown of the Hadley Centre for Climate Science and Services at the UK Met Office. My PhD research broadly concerned methodological advancements and application of models for spatial extremes with a specific focus on modelling the extremal behaviour of aggregates of spatial processes; much of my doctoral research focused on modelling extreme spatial aggregates of UK precipitation, with a view to mitigating fluvial flood risk.

Between 2021 and 2024, I was a postdoc at the King Abdullah University of Science and Technology (KAUST). I worked in the extreme statistics (extSTAT) group led by Raphaël Huser. My postdoctoral research focus was on the development of sparse models for spatio-temporal extremes.

Research

Publications

  1. Cisneros, D., Richards, J., Dahal, A., Lombardo, L., and Huser, R. (2024). Deep graphical regression for jointly moderate and extreme Australian wildfires. Spatial Statistics, 59:100811.
  2. Richards, J., Huser, R., Bevacqua, E., and Zscheischler, J. (2023). Insights into the drivers and spatio-temporal trends of extreme Mediterranean wildfires with statistical deep-learning. Artificial Intelligence for the Earth Systems, 2(4):e220095.
  3. Richards, J., Tawn, J. A., and Brown, S. (2023). Joint estimation of extreme spatially aggregated precipitation at different scales through mixture modelling. Spatial Statistics, 53:100725.
  4. Richards, J. and Tawn, J. A. (2022). On the tail behaviour of aggregated random variables. Journal of Multivariate Analysis, 192:105065.
  5. Richards, J., Tawn, J. A., and Brown, S. (2022). Modelling extremes of spatial aggregates using conditional methods. The Annals of Applied Statistics, 16(4):2693-2713.
  6. Richards, J. and Wadsworth, J. L. (2021). Spatial deformation for nonstationary extremal dependence. Environmetrics, 32(5):e2671.

Submitted

  1. Richards, J., Alotaibi, N., Cisneros, D., Gong, Y., Guerrero M. B., Redondo, P., and Shao., X. (2023+). Modern extreme value statistics for Utopian extremes. arXiv.
  2. Sainsbury-Dale, M., Richards, J., Zammit-Mangion, A., and Huser, R. (2023+). Neural Bayes estimators for irregular spatial data using graph neural networks. arXiv.
  3. Richards, J., Sainsbury-Dale, M., Zammit-Mangion, A., and Huser, R. (2023+). Neural Bayes estimators for censored inference with peaks-over-threshold models. arXiv.
  4. Shao, X., Hazra, A., Richards, J., and Huser, R. (2022+). Flexible modeling of non-stationary extremal dependence using spatially-fused LASSO and ridge penalties. arXiv.
  5. Richards, J. and Huser, R. (2022+). Regression modelling of spatiotemporal extreme U.S. wildfires via partially-interpretable neural networks. arXiv.

Other

  1. Richards, J. and Huser, R. (2024+). Extreme Quantile Regression with Deep Learning. Book chapter, for Chapman and Hall/CRC Handbook on Statistics of Extremes. Preview.
  2. Richards, J. (2021). Extremes of Aggregated Random Variables and Spatial Processes. Lancaster University, PhD Thesis.

Talks/Posters

Southampton; Exeter; Edinburgh seminars, 2024 - Neural Bayes estimators for likelihood-free and amortised inference for spatial extremes. Slides.

KAUST Statistics workshop, 2023 -

  1. Advancements in neural Bayes estimation for spatial processes. Poster.
  2. Dual extremal cross-frequency interactions in brain connectivity. Poster.
  3. A new dependence measure for extremal brain connectivity. Poster.
  4. Causal analysis for both tails in time series: with application to China's derivatives market. Poster.

Spatial Statistics, 2023 - Deep compositional models for non-stationary extremal dependence. Poster.

EVA; ICSA symposium, 2023 - Neural Bayes estimators for fast and efficient inference with spatial peaks-over-threshold models. Slides.

EGU (different title); CMStats, 2022 - Partially-interpretable neural networks for extreme quantile regression: With application to Mediterranean Europe wildfires. Slides.

ENVR workshop, 2022 - Partially-interpretable neural networks for extreme quantile regression. Poster.

EVAN (different title); JSM, 2022 - Partially-interpretable neural networks for high-dimensional extreme quantile regression: With application to U.S. wildfires. Slides.

Bath; Lancaster seminars, 2021 - Joint estimation of extreme precipitation aggregates at different spatial scales through mixture modelling and conditional methods. Slides.

EVA, 2021 - Modelling the extremes of spatial aggregates of precipitation using conditional methods. Slides.

CMStats, 2020; vEGU, 2021 - Modelling the tail behaviour of precipitation aggregates using conditional spatial extremes. Slides.

WET workshop; EVA, 2019 - Aggregation of multivariate extremes. Poster.

Software

CensoredNeuralEstimators.jl. Julia scripts for training neural Bayes estimators for censored data. Supports Richards, Sainsbury-Dale, Zammit-Mangion, and Huser (2023+).
pinnEV. R package for fitting extreme value (and other) regression models using the partially-interpretable neural network (PINN) framework proposed by Richards and Huser (2022+). Currently in development.
scePrecip. R code for fitting spatial conditional extremes models to censored data, e.g., precipitation. Accompanies Richards, Tawn, and Brown (2022, 2023).
sdfExtreme. R package for performing spatial deformations to handle nonstationarity in spatial extremal dependence. Accompanies Richards and Wadsworth (2021).
CondExtremesPy. Functions for fitting Heffernan and Tawn (2004) conditional extremes models in Python (ongoing).
INARMA_RJMCMC_R/INARMA_RJMCMC_py. R/Python code for reversible jump MCMC to determine order and estimate parameters of INARMA models.

Contact

Email

jordan (dot) richards (at) ed (dot) ac (dot) uk

LinkedIn

www.linkedin.com/in/jbrich95

GitHub

https://github.com/Jbrich95

ResearchGate

https://www.researchgate.net/profile/Jordan-Richards-3

Design template from we.graphics. Updated 16/04/2024.