Rionowcast Project Holds Workshop on Extreme Rainfall Prediction in Rio de Janeiro

The Rionowcast consortium, composed of LNCC (National Laboratory for Scientific Computing), CEFET-RJ, UFF, and the Rio Operations Center (COR), held a technical workshop last Wednesday (July 30) to align the project's advances, challenges, and goals. The initiative aims to develop artificial intelligence-based solutions for predicting extreme rainfall events in the city of Rio de Janeiro.

The project focuses on three main areas: applied research, human resource training, and technological support to COR, directly contributing to the mitigation of the impacts caused by heavy rainfall, such as flooding, landslides, and traffic congestion in urban areas.

During the meeting, participants discussed topics related to the collection, automation, and storage of meteorological data. Data sources such as radars, satellites, rain gauges, probes, and ocean buoys were addressed, as well as the use of climate models like ERA5 and GFS.

In the modeling area, different AI-based architectures are being evaluated, including NowcastNet, UNet, ConvLSTM, GraphCast, and models based on Transformers and causality. The goal is to identify efficient approaches for very short-term forecasting (nowcasting) in complex urban areas like Rio de Janeiro.

One of the workshop’s highlights was the decision to use the Gypscie platform, developed by LNCC, as the foundation for organizing and experimenting with the project’s data and models. The platform is being expanded and will feature new instructional videos to guide researchers, students, and professionals on its use and development workflows. The goal is to ensure reproducibility, transparency, and collaboration within the project ecosystem.

New meteorological partners were also defined, aiming to strengthen the integration between computational approaches and applied meteorological knowledge. In addition, the group began planning the next scientific publication, focusing on the results of the infrastructure and models under development.

The Rionowcast project reaffirms its commitment to scientific and technological advancement applied to natural disaster prevention, contributing to a more resilient and safer city.

More information: https://rionowcast.dexl.lncc.br/