Rob Shooter | Searcharxiv

arXiv2022

Multivariate spatial conditional extremes for extreme ocean environments

The joint extremal spatial dependence of wind speed and significant wave height in the North East Atlantic is quantified using Metop satellite scatterometer and hindcast observations for the period 2007-2018, and a multivariate spatial conditional extremes (MSCE) model, ultimately motivated by the work of Heffernan and Tawn (2004). The analysis involves (a) registering individual satellite swaths and corresponding hindcast data onto a template transect (running approximately north-east to south-west, between the British Isles and Iceland), (b) non-stationary directional-seasonal marginal extreme value analysis at a set of registration locations on the transect, (c) transformation from physical to standard Laplace scale using the fitted marginal model, (d) estimation of the MSCE model on the set of registration locations, and assessment of quality of model fit. A joint model is estimated for three spatial quantities: Metop wind speed, hindcast wind speed and hindcast significant wave height. Results suggest that, when conditioning on extreme Metop wind speed, extremal spatial dependence for all three quantities decays over approximately 600-800 km.

stat.ME↗

arXiv2021

A marginal modelling approach for predicting wildfire extremes across the contiguous United States

This paper details a methodology proposed for the EVA 2021 conference data challenge. The aim of this challenge was to predict the number and size of wildfires over the contiguous US between 1993 and 2015, with more importance placed on extreme events. In the data set provided, over 14\% of both wildfire count and burnt area observations are missing; the objective of the data challenge was to estimate a range of marginal probabilities from the distribution functions of these missing observations. To enable this prediction, we make the assumption that the marginal distribution of a missing observation can be informed using non-missing data from neighbouring locations. In our method, we select spatial neighbourhoods for each missing observation and fit marginal models to non-missing observations in these regions. For the wildfire counts, we assume the compiled data sets follow a zero-inflated negative binomial distribution, while for burnt area values, we model the bulk and tail of each compiled data set using non-parametric and parametric techniques, respectively. Cross validation is used to select tuning parameters, and the resulting predictions are shown to significantly outperform the benchmark method proposed in the challenge outline. We conclude with a discussion of our modelling framework, and evaluate ways in which it could be extended.

stat.ME↗