Space-Time Dynamics of the Excess of PM10 Concentration Over Threshold Levels

This study presents a spatio-temporal modeling framework to predict concentrations across Northern Italy and to estimate the probability of exceeding predefined thresholds. The statistical model combines a space-varying parameter approach, incorporating meteorological covariates, seasonal harmonics, and temporal lags, with spatio-temporal kriging to provide predictions at any spatio-temporal location. To assess exceedance probabilities, a Monte Carlo procedure is implemented that combines a parametric block bootstrap with conditional spatio-temporal simulations, thereby propagating parameter uncertainty while preserving the spatio-temporal dependence structure. The method enables efficient prediction at unobserved locations and over arbitrary temporal windows, providing a flexible and computationally feasible tool for air quality assessment and management.