A new mixture model for spatiotemporal exceedances with flexible tail dependence

TL;DR

This paper introduces a flexible mixture model for spatiotemporal streamflow exceedances that captures various dependence regimes in the tail, using simulation-based inference with random forests.

Contribution

It presents a novel mixture model framework and an estimation method for extreme streamflow events that accounts for different tail dependence structures.

Findings

Model effectively captures asymptotic dependence and independence.

Simulation results demonstrate accurate estimation of extreme event probabilities.

Application to US streamflow data shows practical utility.

Abstract

We propose a new model and estimation framework for spatiotemporal streamflow exceedances above a threshold that flexibly captures asymptotic dependence and independence in the tail of the distribution. We model streamflow using a mixture of processes with spatial, temporal and spatiotemporal asymptotic dependence regimes. A censoring mechanism allows us to use only observations above a threshold to estimate marginal and joint probabilities of extreme events. As the likelihood is intractable, we use simulation-based inference powered by random forests to estimate model parameters from summary statistics of the data. Simulations and modeling of streamflow data from the U.S. Geological Survey illustrate the feasibility and practicality of our approach.