B0808
Title: Time varying autoregressive gamma shot noise model for wildfires
Authors: Roberto Casarin - University Ca' Foscari of Venice (Italy)
Federico Bassetti - Politecnico Milano (Italy) [presenting]
Matteo Iacopini - Queen Mary University of London (United Kingdom)
Abstract: Motivated by the analysis of wildfires, a novel time-varying shot noise Cox process is proposed for modelling time series of spatial data. The model assumes that a latent sequence of autoregressive gamma random measures drives the random intensity of the Cox process. To perform inference, a Bayesian approach is employed combined with a Markov Chain Monte Carlo algorithm. Several properties of the latent sequence of time-dependent gamma random measures are derived, essential for computing moment, predictive, and pair correlation measures of the proposed shot noise process. The flexible and tractable model makes it suitable for capturing spatial patterns and temporal dynamics in forest fires. Furthermore, the approach offers a potential solution to the challenging problem of estimating global trends and seasonality using high spatial-resolution fire data from a wide region. By adopting the Bayesian approach, uncertainty is quantified in the estimates and forecasts, a critical aspect of climate-risk analysis. In the application, NASA moderate resolution imaging spectroradiometer (MODIS) data is utilized.
