Deep-Learning Models for meteorology prediction

Rionowcast - Deep-Learning Models for meteorology prediction

Project Site

Related Areas

Applied Ai (AAi)
Meteorology
Deep Learning Architecture and Peformance
Big Data Processing techniques, systems and challenges
Data Centric AI

In this project, we aim at improving the forecast of meteorology variables using a combination of spatial and temporal signals. We have developed two models: a CONV-LSTM and a STCONVS2S. These models are trained using the output of numerical simulation. In the former case, data from different models are ensembled through different channels of the convolution. In the latter case, the temporal signal is capturted by a 1D convolution and a new module is placed after a spatial convolution to extend the tenporal forecast.