Nicolas Eckert - Climate change risks related to the shrinkage of the mountain cryosphere: state of the art and challenges for statistical modelling
Abstract : Mountain risks have deep socio-economic consequences when interacting with elements at risks such as settlements, their inhabitants and critical infrastructures. They result from spectacular phenomena such as snow avalanches, snow storms, permafrost sawing and glacier collapses which are deeply impacted by ongoing climate change. However, as stated by the recent special IPCC report on oceans and cryosphere, past and future trends remain ill-known. This talk provides a very quick overview of existing results regarding changes in snow extremes, snow avalanches, glaciers and related risks, and of open challenges in statistical modelling to better assess and mitigate them. Required developments involve combining extreme value theory, spatio-temporal models, formal risk modelling and knowledge resulting from many disciplines. In addition to innovative findings of wide geophysical relevance, this should eventually contribute to the design of efficient adaptation strategies.
Thomas Opitz - Spatiotemporal modeling of the risk of extreme wildfire
Abstract: Accurate spatiotemporal modeling of conditions leading to moderate and large wildfires provides better understanding of mechanisms driving fire-prone ecosystems and improves risk management. In this talk, I will present a joint model for the occurrence intensity and the wildfire size distribution by combining extreme-value theory and point processes within a novel Bayesian hierarchical model. The model is used to study daily summer wildfire data for the French Mediterranean basin during the 1995–2018 period. The occurrence component models wildfire ignitions as a spatiotemporal log-Gaussian Cox process. Burnt areas are numerical marks attached to points and are considered as extreme if they exceed a high threshold. The size component is a two-component mixture varying in space and time that jointly models moderate and extreme fires. We capture non-linear influence of covariates (Fire Weather Index, forest cover) through component-specific smooth functions, which may vary with
Sebastian Engelke - Machine learning methods for extremes
Abstract: In this talk we show how methods from machine learning can be used for risk assessment in complex, high-dimensional settings. Such situations often appear in the analysis of weather and climate extremes, and we present several applications from this domain. The first part is concerned with the prediction of extreme quantiles, conditionally on a high-dimensional predictor vector. We adapt gradient boosting and random forest using a suitable loss function to obtain methods that outperform existing approaches from quantile regression and extreme value theory. In the second part, we consider spatial extreme events, where classical models encounter challenges because of possible non-stationarity and complicated likelihoods. We propose to combine generative adversarial networks with extreme value theory to obtain a flexible spatial model with accurate extrapolation properties. The talk is based on the papers Boulaguiem et al. (2021, arXiv:2111.00267) and Velthoen et al. (2021, a
Philippe Naveau - Non-parametric multimodel Regional Frequency Analysis applied to climate change detection and attribution
Abstract: A recurrent question in climate risk analysis is determining how climate change will affect heavy precipitation patterns. Dividing the globe into homogeneous sub-regions should improve the modelling of heavy precipitation by inferring common regional distributional parameters. In addition, in the detection and attribution field, biases due to model errors in global climate models (GCMs) should be considered to attribute the anthropogenic forcing effect. Within this D\&A context, we propose an efficient clustering algorithm that, compared to classical regional frequency analysis (RFA) techniques, is covariate-free and accounts for dependence. It is based on a new non-parametric dissimilarity that combines both the RFA constraint and the pairwise dependence. We derive asymptotic properties of our dissimilarity estimator, and we interpret it for generalised extreme value distributed pairs. As a D\&A application, we cluster annual daily precipitation maxima of 16 GCMs fr
Anthony Davison - Functional peaks-over-threshold analysis
Abstract : Peaks-over-threshold analysis using the generalized Pareto distribution is widely applied in modelling tails of univariate random variables, but much information may be lost when complex extreme events are studied using univariate results. We extend peaks-over-threshold analysis to extremes of functional data. Threshold exceedances defined using a functional are modelled by a functional generalization of the generalized Pareto distribution that covers the three classical regimes for the decay of tail probabilities, and that is the only possible continuous limit for suitable exceedances of a properly rescaled process. We describe the new approach and apply it to extreme European windstorms and heavy spatial rainfall. The work is joint with Raphaël de Fondeville.
David Makowski - Is it possible to predict the occurrence of extreme agricultural yield losses and their impacts on commodity prices?
Abstract:Although prices of agricultural commodities are influenced by many factors (food stocks, input prices - particularly fertilizer -, food and feed demand, etc.), shocks on agricultural productions are often recognized as a major driver of price volatility. It has been shown that adverse environmental conditions (e.g., droughts) and the resulting decline in regional production greatly contributed to spikes in global food prices. The last decade has seen an increasing number of applications of machine learning, dedicated in particular to anticipate agricultural production shocks. The reasons are multiple. First, machine learning techniques can now be more easily applied than before using specialized packages. Second, large agri- cultural databases are becoming more and more easily accessible, especially thanks to the emergence of open-science good practices. Finally, mechanistic models were found to miss several major yield loss events in Europe and in the US, which led scientists
Loading... 
https://wecc.sciencesconf.org/data/program/WECC_2022_NicolasEckert_diffusable.pdf
