Nicolo' Caron
Correcting Hydrodynamic Model Biases in Western Baltic Sea Storm Surges via Neural Networks.
Rel. Vincenzo Nava. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Per L'Ambiente E Il Territorio, 2026
|
Preview |
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (22MB) | Preview |
Abstract
Storm surges are the primary coastal hazard in the Danish and western Baltic seas, where forecasting relies on hydrodynamic models such as the Danish Meteorological Institute's DKSS. Despite improvements in physics and forcing, modelled sea levels still deviate from observations, and the discrepancy intensifies with rising water levels. At the evaluated stations, the full-record root-mean-square error (RMSE) of 8.4 cm grows to 12.8 cm above the 99th percentile. Because tide gauges are sparse, this error is only partially observable. The challenge is thus not merely to correct the model where it is measured, but to learn a correction that generalizes to ungauged coasts.
This thesis separates the model–gauge discrepancy into a station-dependent vertical offset (removed as a per-station constant that also defines the comparison baseline) and a dynamic, time-varying modelling error, which becomes the target of a supervised learning problem on ordered time series
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
Ente in cotutela
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
