Latest news on ongoing CMLs projects in Africa/Madagascar and new research on rainfall measurement from sound sensors
Marielle Gosset1, Modeste Kacou2, Rodrigo Xavier3, Armel Kodji2
1 Institut de Recherche pour le Développement, Toulouse, France
2 Université Félix Houphouët-Boigny, Abidjan, Côte d’Ivoire
3 Universidade Federal do Ceará, Fortaleza, Brazil
In this presentation, we will present our latest research on opportunistic rain sensors in Africa and Brazil, carried out in the framework of the RAINSMORE project funded by IRD. First, we will share our latest experiences collecting and testing CMLs data with a new test bed in Madagascar, discussing the promises and limitations of the current experiment. Second, we will present our new research line on the use of sound-recorders (initially developed and installed for biodiversity monitoring) for detecting and estimating rainfall using machine learning, with exciting results in the Amazon region.
Quantitative Precipitation Estimation with GEOsat and a conditional Generative Adversarial Network
Selina Janner1, Christian Chwala1
1 Institut für Meteorologie und Klimaforschung Atmosphärische Umweltforschung (IMK-IFU), Karlsruhe Institute of Technology, Garmisch-Partenkirchen, Germany
Unlike traditional precipitation monitoring systems, geostationary satellites are not locally limited; but they do not directly measure precipitation. This presentation will showcase a generative deep learning model for precipitation estimation based on Level 1.5 data from SEVIRI. By using a conditional Generative Adversarial Network (cGAN), we aim to develop a method which has the ability to create probabilistic, spatio-temporal consistent, near-real-time rainfall fields with a realistic distribution of rain rates. This talk will provide an overview of the model, assess its current state, and outline future plans (e.g. applying this approach in data-sparse regions).
