B1164
Title: Modelling coastal profiles with a functional space-time model
Authors: Francesco Finazzi - University of Bergamo (Italy) [presenting]
Alessandro Fasso - University of Bergamo (Italy)
Philipp Otto - University of Glasgow (United Kingdom)
Abstract: Many modern environmental applications involve the collection of profile data across space and over time. In order to make spatio-temporal predictions, a statistical model able to take into account spatial and temporal correlation is needed. We discuss a functional space-time model for profile data based on latent variables with Markovian dynamics in time and spatially-correlated innovations. The model is estimated by means of the maximum likelihood approach using the expectation-maximization algorithm. Model estimation is implemented within the D-STEM software which is capable of handling complex space-time data sets with missing data. The model is applied to measured coastal profiles of the Sylt island in the German Bight between 1980 and 2017. Location and timing of the measurements are strongly related to the placement of nourishments. These are placed on a yearly basis at different locations along the island, depending on erosion losses in the previous year. Although profiles are measured on average 1-2 times per year, the spatial, temporal and functional support is heterogeneous. External forcing conditions that cause changes in the coastal profiles are measured in front of Sylt and are included in the statistical model as covariates. Furthermore, change points could be assigned to profiles that are abruptly changed by a nourishment itself.
