2025 Internship Project Overview

As part of the European Space Agency’s “Medicanes” Project, we developed machine learning (ML) pipelines that use satellite measurements to improve the understanding and prediction of medicanes. Medicanes are tropical-like cyclones in the Mediterranean Sea region characterized by a deep warm core (warm anomaly extending from the middle to upper troposphere at the storm center) and a nearly closed rotation of the wind field.

We curated a dataset (correcting sensor inconsistencies), surveyed state-of-the-art ML approaches and delivered models for this novel satellite-based detection task.

To automatically detect warm cores, we implemented a semi-supervised anomaly detection method that utilizes an autoencoder’s reconstruction error (the difference between observed and reconstructed data) and evaluated it on data from 800+ cyclones.

To detect the nearly closed rotating wind field, we preprocessed wind data from Advanced Scatterometer (ASCAT) on board the European MetOp satellites. We trained a U-Net semantic segmentation model (which assigns a label to each point) with partial convolution layers to handle missing data over land.

ORGANIZATION/RESEARCH GROUP

National Research Council of Italy, Institute of Atmospheric Sciences and Climate (CNR-ISAC)

LOCATION

Rome, Italy

 

Interns

Angel Li ’27
Major: Computer Science
Minor(s): Mathematics
"What I found most interesting was the intersection between computer science and weather. I was surprised by how much machine learning was being used to study climate and remote sensing. This is something I would not be exposed to if I had not done this internship. This experience taught me how much I enjoy research and has made me lean towards going to graduate school."
Chinmaya Saran ’28
Major: Computer Science
Minor(s): Statistics and Machine Learning, Sustainable Energy
"I gained a wider, more strategic view of the research and industry landscape, including how organizations like the European Space Agency and partner institutions are shaping the future of Earth observation and environmental monitoring. Seeing work beyond the U.S. helped me understand global trends at the intersection of environment and technology. I also developed fluency with remote sensing data, skills that will be increasingly important as climate change intensifies and as machine learning is used to turn large volumes of environmental data into actionable insight."

Click to view the end-of-project presentation

 

Mario Marcello Miglietta
Research Director

Mentors

  • Leo Pio D’Adderio, Postdoctoral Research Associate, CNR-ISAC
  • Giulia Panegrossi, Senior Research Scientist, CNR-ISAC
  • Stefano Sebastianelli, IT Remote Sensing, CNR-ISAC