Collaborating Authors

Google says new AI model allows for near 'instantaneous' weather forecasts

Daily Mail - Science & tech

Google is throwing the power of its AI and machine-learning algorithms behind developing faster and more accurate weather forecasts. In a blog post, Google describes a new model developed by the company called'nowcasting' which it says has shown initial success in being able to accurately predict weather patterns with'nearly instantaneous' results. According to a new paper, the method is able to produce forecasts for up to six hours in advance in only five to 10 minutes - figures that it says outperform traditional models even in early stages. While some traditional forecasts generate massive amounts of data, they can also take hours to complete. '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,' Google writes.

AI Contributing to Better Accuracy and Precision in Weather Forecasting - AI Trends


Traditional models of weather forecasting are based on statistical measures based on data collected from deep space satellites, such as NOAA's Deep Space Climate Observatory, weather balloons, radar systems, and sometimes from IoT-based sensors. Today, AI is finding a role in weather forecasting with machine learning being employed to process more complex data in less time, with the hope of improving accuracy. For example, the Numerical Weather Prediction (NWP) site from NOAA offers a range of data sets for use by researchers, from temperature and precipitation data to wave heights, according to a recent account in Analytics Insight. The site offers vast data sets relayed from weather satellites, relay stations, and radiosondes to help deliver short-term weather forecasts or long-term climate predictions. Besides machine learning, other AI techniques for weather predictions include Artificial Neural Networks, Ensemble Neural Networks, Backpropagation Networks, Radial Basis Function Networks, General Regression Neural Networks, Genetic Algorithms, Multilayer Perceptrons and fuzzy clustering.

Predicting Weather Uncertainty with Deep Convnets Machine Learning

Modern weather forecast models perform uncertainty quantification using ensemble prediction systems, which collect nonparametric statistics based on multiple perturbed simulations. To provide accurate estimation, dozens of such computationally intensive simulations must be run. We show that deep neural networks can be used on a small set of numerical weather simulations to estimate the spread of a weather forecast, significantly reducing computational cost. To train the system, we both modify the 3D U-Net architecture and explore models that incorporate temporal data. Our models serve as a starting point to improve uncertainty quantification in current real-time weather forecasting systems, which is vital for predicting extreme events.

How Companies Like IBM Are Helping Predict Weather Better With AI


Each year, natural catastrophes result in severe economic costs and impact the lives of millions globally. While the forces of nature cannot be controlled, accurate and reliable weather forecasts will enable governments and administrations to take appropriate measures and contain the magnitude of the resultant damage. Likewise, precise weather predictions would benefit businesses by allowing them to take more appropriate decisions by factoring in weather related impacts. Over the years, weather predictions have greatly improved, however, there is still room for growth. Here's a look at how advanced technologies such as Artificial Intelligence (AI) are being leveraged to transform weather forecasting.

Machine Learning for Precipitation Nowcasting from Radar Images Machine Learning

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.