Artificial Intelligence is one of the recognised major players to fight climate change and build a more sustainable world. It is also revolutionising medicine.
We have recently started a collaboration with ICTP on forecasting extreme climate events, starting from precipitations. We are working on a framework in which the first step is to learn a model to predict from low resolution simulation data a high resolution precipitation map.
Since the end of 2022 we are involved, via the iNEST project, in the co-design of a digital twin of the north Adriatic sea, harnessing the power of AI to speed up and increase precision of current models used by our colleagues in OGS. We are currently working on the improvement of initial conditions for simulating models, integrating data from satellites, vertical measurements, and low resolution models.
In medicine, we work mostly on signal data. We have worked on assisted ventilation, patenting with other colleagues a simple way to detect asynchronies between patients and ventilation machines. More recently, we are collaborating with cardiology department of Trieste hospital to predict the future insurgence of diseases from ECG data. We started from insurgence of Atrial Fibrillation within five years.
Keywords: Forecasting of extreme climate events, digital twin of the ocean, AI for ECG data, synthetic generation of medical data.
In the following, there is a reasoned bibliography, in which each class of contributions is briefly described and the relevant bibliography is cited.
Precipitation projection
Precipitation-related extremes are rapidly growing due to climate change and data-driven approaches can significantly help in efficiently derive hazard projections. Here, we are developing a novel deep learning framework based on graph neural networks to downscale low-resolution (~25km) atmospheric data to high-resolution (3km) hourly precipitation distribution. The model is trained on reanalysis data, nonetheless we plan to use predictors from model simulations to derive future decades projections.