A team of researchers working at Google's Mountain View research center has developed a new weather forecasting model that allows for near 'instantaneous' forecast, as shared by Google in a blog post on January 13, 2020. The firm says 'nowcasting', with the power of artificial intelligence and machine-learning algorithms, can provide more accurate weather predictions in only minutes.
Nowcast is a term that refers to a forecast for the present time or the very near future. Forecasting the weather for these future moments is complex, but can be incredibly useful for individuals and societies for event planning, farming, or disaster prevention, for instance.
The work is in its early stages of development and has yet to be included in any commercial systems, but initial findings already deliver promising results. In the paper, researchers detailed how they generated accurate rainfall predictions up to six hours ahead of time at a 1 km (0.6 miles) resolution.
With just minutes of calculation, it was already a significant improvement over existing techniques that usually take hours to generate forecasts, Google said.
The speed increase was achieved by replacing the heavy lifting of computationally intensive physics models with neural networks trained on historical data. Radar images are analyzed using convolutional neural networks or CNNS-- the same layered networks used to recognize objects in images and interpret speech.
"A significant advantage of machine learning is that inference is computationally cheap given an already-trained model, allowing forecasts that are nearly instantaneous and in the native high resolution of the input data," said Jason Hickey, a senior software engineer at Google Research.
Google's AI nowcasts have proven to be more accurate than three existing prediction techniques used as comparisons. For now, this type of AI processing is ideal for producing short-term predictions in weather patterns.
In addition to making people's lives easier, the technology could also help pave the path for major weather predictions about extreme and unusual events caused by climate change.
"As weather patterns are altered by climate change, and as the frequency of extreme weather events increases, it becomes more important to provide actionable predictions at high spatial and temporal resolutions," the researchers wrote.
Top: Image showing the location of clouds as measured by geosynchronous satellites. Bottom: Radar image showing the location of rain as measured by Doppler radar stations. Image credit: NOAA, NWS, NSSL
"Machine Learning for Precipitation Nowcasting from Radar Images" - Agrawal, S. et al - arXiv e-prints - https://arxiv.org/abs/1912.12132
High-resolution nowcasting is an essential tool needed for effective adaptation to climate change, particularly for extreme weather. As Deep Learning (DL) techniques have shown dramatic promise in many domains, including the geosciences, we present an application of DL to the problem of precipitation nowcasting, i.e., high-resolution (1 km x 1 km) short-term (1 hour) predictions of precipitation. We treat forecasting as an image-to-image translation problem and leverage the power of the ubiquitous UNET convolutional neural network. We find this performs favorably when compared to three commonly used models: optical flow, persistence and NOAA's numerical one-hour HRRR nowcasting prediction.
Featured image credit: PIRO4D/Pixabay