AI Bits Traditional Weather Forecasting Methods for Short-Term Predictions
Google researchers have developed am AI system for short-term weather predictions that outperforms current solutions.
In the paper “Machine Learning for Precipitation Nowcasting from Radar Images”, scientists shared that their system has outperformed current methods in short-term weather forecasting:
“This precipitation nowcasting, which focuses on 0-6 hour forecasts, can generate forecasts that have a 1km resolution with a total latency of just 5-10 minutes, including data collection delays, outperforming traditional models, even at these early stages of development”
The innovation comes from new “physics-free” approach, where system is not modelling an actual atmosphere, but treats a problem as “as an image-to-image translation problem” that allows to use convolutional neural networks (CNNs).
Researchers report that the invention could be useful in supporting immediate decisions – “from traffic routing and logistics to evacuation planning”. Additionally, Google scientists shared that new discovery will be especially useful in times of climate change events:
“High-resolution nowcasting is an essential tool needed for effective adaptation to climate change, particularly for extreme weather”