HIDRA-D: deep-learning model for dense sea level forecasting  using sparse altimetry and tide gauge data

Rus, Marko; Ličer, Matjaž; Kristan, Matej

doi:10.5194/gmd-19-2177-2026

Articles | Volume 19, issue 5

https://doi.org/10.5194/gmd-19-2177-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-19-2177-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 19, issue 5

Development and technical paper

|

17 Mar 2026

Development and technical paper |

| 17 Mar 2026

HIDRA-D: deep-learning model for dense sea level forecasting using sparse altimetry and tide gauge data

Marko Rus, Matjaž Ličer, and Matej Kristan

Data sets

Training and Test Datasets, Pretrained Weights and Predictions for HIDRA-D Marko Rus et al. https://doi.org/10.5281/zenodo.15790578

Model code and software

Code for HIDRA-D: Deep-Learning Model for Dense Sea Level Forecasting using Sparse Altimetry and Tide Gauge Data Marko Rus et al. https://doi.org/10.5281/zenodo.15799686

HIDRA-D Marko Rus https://github.com/rusmarko/HIDRA-D

Video supplement

Video supplement for HIDRA-D Marko Rus et al. https://doi.org/10.5446/70892

Short summary

This paper introduces HIDRA-D, a novel deep-learning model for dense, gridded sea level forecasting from sparse satellite altimetry and tide gauge data. By forecasting low-frequency spatial components, HIDRA-D offers a faster alternative to traditional numerical models. Evaluated on the satellite altimetry data in the Adriatic Sea, it outperforms the NEMO general circulation model, reducing the mean absolute error by 28.0 %. The model is robust but shows limitations in complex coastal areas.