LEARNING OBJECTIVES
The course provides practical and technical skills for managing large-scale meteorological radar datasets and applying Artificial Intelligence techniques for precipitation nowcasting. Participants will learn how to access cloud-native infrastructures, work with radar data repositories and apply the state-of-the-art IRENE deep learning model for short-term weather forecasting.
Course Content and Programme
Module 1 – Radar Datasets and Big Data Management (4 hours)
Understand radar data structures and modern cloud-native approaches for managing large meteorological datasets.
Structure of the Italian national radar composite dataset.
Radar variables and potential application domains.
Efficient storage and management using the Zarr format.
Accessing data through Arcodatahub and open-data services.
Cloud-native approaches for data retrieval and analysis.
Module 2 – Nowcasting Modelling with IRENE (4 hours)
Understand the IRENE nowcasting architecture and learn how to use it for operational and research applications.
IRENE model architecture and design principles.
Innovations compared with traditional nowcasting approaches.
Training pipeline, loss functions and optimization strategies.
Performance analysis and benchmark evaluation.
Inference workflows on new radar observations.
COURSE DURATION
8 hours
Participants are welcome to attend the entire course or, if they prefer, attend only one of the two modules. There is no requirement to attend both modules.
TARGET AUDIENCE
Technical professionals working in the meteorological sector.
Academic researchers interested in precipitation monitoring and short-term forecasting.
Data Scientists, AI Engineers and Machine Learning specialists focusing on geospatial and weather-related applications.
Professionals involved in the development of meteorological services and applications.
REQUIREMENTS
Basic knowledge of Python programming.
Fundamentals of Machine Learning and Deep Learning.
Basic understanding of meteorology or radar data analysis is recommended but not mandatory.
ASSESSMENT METHOD
Multiple-choice test
TEACHING MATERIALS
Presentation slides covering theoretical concepts.
Interactive Jupyter notebooks for the Radar Dataset Module.
Interactive Jupyter notebooks for the IRENE Modeling Module.
INSTRUCTOR PROFILE
Alessandro Camilletti is a researcher at the Bruno Kessler Foundation (FBK), where he develops artificial intelligence methods for weather forecasting. After obtaining a PhD in Physics, he focused his research on AI applications for seasonal forecasting and weather nowcasting. He has also contributed to the analysis of the Italian National Radar Composite (italian-radar-dpc-sri) and to the development of a probabilistic AI-based nowcasting model for Italy.
Email: acamilletti@fbk.eu
CERTIFICATE
Digital Badge
SUPPORT