An autoregressive spatio-temporal precipitation model

A spatio-temporal model for precipitation is presented. It is assumed that precipitation follows a censored and power-transformed normal distribution. Through a regression term, precipitation is linked to covariates. Spatial and temporal dependencies are accounted for by a latent Gaussian variable that follows a Markovian temporal evolution combined with spatially correlated innovations. Such a specification allows for nonseparable covariances in space and time. Further, the Markovian structure yields computational efficiency and it exploits in a natural way the unidirectional flow of time. In addition, the model is space as well as time resolution consistent. The model is applied to three-hourly Swiss rainfall data, collected at 26 stations.