Publications

Stefanelli, M., Jansen, E., Aydoğdu, A., Federico, I. and Coppini, G. (2025).
Data assimilation for advanced cross-scale unstructured-grid ocean modelling.
Frontiers in Marine Science, 12:1656879.
DOI: 10.3389/fmars.2025.1656879

Coastal zones face growing threats from climate change, including sea-level rise and intensified storm activity. Accurate numerical modelling is essential to predict the impacts of anthropogenic and climate stressors on the coastal zone. However, it is also a very challenging environment due to complex coastlines, rapid topographic changes, and high spatial-temporal variability.

Unstructured-grid models offer a promising solution, yet their integration with advanced data-assimilation (DA) methods remains limited. This study presents the implementation of a three-dimensional variational data-assimilation (3DVar) scheme, OceanVar, within the unstructured-grid ocean model SHYFEM. A key innovation involves generalising the first-order recursive filter for horizontal background-error covariances to work with triangular unstructured meshes.

An experiment was conducted over the period 2017–2018, assimilating Argo in-situ profiles and sea-level anomaly (SLA) data from satellite altimetry missions. Results show substantial skill improvement relative to a control run without assimilation, particularly in the 100–500 m depth range, where the mean absolute error was reduced by 25–30% through data assimilation. SLA assimilation had a more modest effect, improving MAE by about 3% overall and up to 20% locally, without degrading temperature or salinity estimates.

The study demonstrates the feasibility and benefits of applying a 3DVar scheme to unstructured-grid ocean models, paving the way for more accurate and efficient coastal forecasting systems.

Read the paper

Stefanelli, M., Jansen, E., Aydoğdu, A., Federico, I. and Coppini, G. (2025).
A neural-network model-measurement-based observation operator for weather radar reflectivity assimilation.
Preprint, arXiv.
https://arxiv.org/pdf/2512.18289

In three-dimensional variational data assimilation (3DVar) for numerical weather prediction (NWP), the observation operator \(\mathcal{H}\) plays a central role by mapping model state variables to an observation-equivalent field. For weather radar, however, specifying \(\mathcal{H}\) is particularly challenging: reflectivity is a nonlinear, microphysics-dependent diagnostic quantity that relates only indirectly to the model’s prognostic variables, making traditional parameterised radar operators complex, regime-dependent, and difficult to tune.

In this study, we propose a neural-network (NN)-based observation operator for radar reflectivity and apply it within a 3DVar data-assimilation (DA) framework. Using five years (2019–2023) of radar reflectivity data from the Lisca radar and 4.4 km-resolution short-range forecasts from the ALADIN model over Slovenia, we train a convolutional encoder–decoder neural network to map model temperature, humidity, horizontal wind components, and mean sea-level pressure fields to radar reflectivity.

Statistical evaluation showed that the trained observation operator can represent the broader distribution and spatial occurrence of weak-to-moderate precipitation, but remains limited in reproducing the frequency, intensity, and spatial organisation of the strongest radar reflectivity features.

The practical performance of the framework was evaluated using an extreme precipitation event that caused widespread flooding in Slovenia on 4 August 2023. In this case, assimilating the full radar disc reduced the domain-averaged reflectivity root-mean-square error (RMSE) from 5.99 dB(Z) to 3.47 dB(Z) and improved the alignment between the analysed and observed convective bands.

Embedded within 3DVar, the Jacobian of the NN observation operator allows radar reflectivity observations to inform the model state variables, producing corresponding analysis increments. The proposed NN radar observation operator offers a flexible alternative to traditional parameterised radar operators for improving convective-storm forecasts.

Read the paper