B1076
Title: A functional ground motion model for partially observed response profiles
Authors: Teresa Bortolotti - Politecnico di Milano (Italy) [presenting]
Riccardo Peli - Politecnico di Milano (Italy)
Giovanni Lanzano - Istituto Nazionale di Geofisica e Vulcanologia (Italy)
Sara Sgobba - Istituto Nazionale di Geofisica e Vulcanologia (Italy)
Alessandra Menafoglio - Politecnico di Milano (Italy)
Abstract: Driven by the role played by ground motion models in seismic hazard analysis, the problem of formulating a ground motion model is addressed as a concurrent functional model in the presence of partially observed response profiles. A novel functional regression routine is proposed, which employs data reconstruction techniques on the partially observed trajectories and addresses the inherent uncertainty arising from the reconstruction. The methodology defines observation-specific functional weights, which enter the estimation process to reduce the reconstructed trajectories' impact on the final estimates. The classical methods of smoothing and concurrent functional regression are adapted to incorporate weights. The advantages of the proposed weighted methodology are evaluated using synthetic data. The weighted functional analysis applied to seismological data is shown to provide a natural smoothing and stabilization of the coefficients estimates of the considered ground motion model.
