AI Models For Predicting Extreme Weather Events

Google DeepMind and Google Research have developed a new experimental AI model to predict tropical cyclones, and the results on recent hurricanes like Hurricane Erin are really exciting. Watch Olivia Graham from our team explain it below. 🌀Tropical cyclones cause immense destruction and seriously impact communities. Improving the accuracy and timeliness of our forecasts is critical for protecting property and saving lives. 💧Traditional models💧 Physics-based models struggle to accurately predict both a cyclone's path and its intensity. This is because a cyclone's path is influenced by vast atmospheric currents, while its intensity depends on complex turbulent processes within and around its core. ✨Our new model✨ Our new experimental model is a single system that overcomes the traditional tradeoff between track and intensity. It's trained on two distinct types of data 1. A vast reanalysis dataset that reconstructs global weather patterns from millions of observations. 2. A specialized database containing key information about the track, intensity, size, and wind radii of nearly 5,000 observed cyclones from the past 45 years. This allows the model to learn from historical events in a way that traditional models cannot. We’re working with the National Hurricane Center to test this experimental model out this season.  It was really gratifying to see this writeup from the former chief of the hurricane specialist unit there on Hurricane Erin, the strongest Atlantic storm this year. According to his analysis, our model (GDMI in the graphs) had the most accurate forecasts for both track and intensity for the first 72 hours, outperforming a number of the best physics-based models and even the consensus models used by forecasters. This model is now live on Weather Lab, where it's generating 50 possible scenarios for potential future outcomes. If you'd like to explore the model yourself, check it out on Weather Lab: https://lnkd.in/gY9z5wCK Analysis on Hurricane Erin from the former chief of the hurricane specialist unit at the National Hurricane Center: https://lnkd.in/gTHHXdup

Thrilled to unveil our latest work: multi-modal machine learning to forecast localized weather! We construct a graph neural network to learn dynamics at point locations, where typical gridded forecasts miss significant variation. Paper: https://lnkd.in/eBmfsJin Weather dataset: https://lnkd.in/ejCG8bKs Code: https://lnkd.in/eQg-JzQJ AI weather models have made huge strides, but most still emulate products like ERA5, which struggle to capture near-surface wind dynamics. The correlation between ERA5 and ground weather station data is low due to topography, buildings, vegetation, and other local factors. In this work, we forecast near-surface wind at localized off-grid locations using a message-passing graph neural network ("MPNN"). The graph is heterogeneous, integrating both global forecasts (ERA5) and historical local weather station data as different nodes. What do we find? First off, ERA5 interpolation performs poorly, failing to capture local wind variations, especially in coastal and inland regions with complex conditions. An MLP trained on historical data at a location performs better than ERA5 interpolation, as it learns from the station's past observations. However, it struggles with longer lead times and lacks the spatial context necessary to capture weather patterns. Meanwhile, our MPNN dramatically improves performance, reducing the error by over 50% compared to the MLP. This is because the MPNN incorporates spatial information through message passing, allowing it to learn local weather dynamics from both station data and global forecasts. Interestingly, adding ERA5 data to the MLP does not improve its performance significantly. The MLP struggles to integrate spatial information from global forecasts, while the MPNN excels, highlighting the importance of combining global and local data. Large improvements in forecast accuracy occur at both coastal and inland locations. Our model shows a 92% reduction in MSE relative to ERA5 interpolation overall. This work showcases the strength of machine learning in combining multi-modal data. By using a graph to integrate global and local weather data, we were able to generate much more accurate localized weather forecasts! Congrats to Qidong Yang and Jonathan Giezendanner for the great work, and thanks to Campbell Watson, Daniel Salles Chevitarese, Johannes Jakubik, Eric Schmitt, Anirban C., Jeremy Vila, Detlef Hohl, and Chris Hill for a wonderful collaboration. Thanks also to our partners at Amazon Web Services (AWS) for providing cloud computing and technical support!

For those of you following developments with AI & Science, particularly around weather forecasting… At Google Research and Google DeepMind we have introduced an experimental model for tropical cyclone prediction, which can predict a cyclone’s formation, track, intensity, size and shape – generating 50 possible scenarios, up to 15 days in advance. And as we head into this year’s cyclone season, we’re partnering with the US National Hurricane center to support their forecasts and warnings. We’re publicly sharing this experimental model in Weather Lab, a new platform to access experimental weather forecast visualizations, and we hope to gather feedback and enable researchers and forecasters to leverage our models and predictions to inform their own work. You can learn more in our blog post (https://lnkd.in/geG62c2v) or this New York Times story (https://lnkd.in/gAFPbUrD).

AI-Powered Weather Forecasting: The GraphCast Innovation In the quest to enhance weather forecasting and provide early warnings for extreme events like hurricanes, artificial intelligence has emerged as a potent tool. Traditional forecasting systems have undoubtedly improved over the years, but AI's ability to swiftly analyze historical data and make predictions is transforming the field. Google #DeepMind's innovative AI tool, #GraphCast, has demonstrated its potential by outperforming conventional models and significantly expediting forecast delivery. Weather forecasts serve a crucial purpose beyond helping us decide our daily attire; they offer a lifeline in anticipating and preparing for severe weather events such as storms, floods, and heatwaves. However, traditional weather forecasting demands immense computational power. It involves processing hundreds of variables across various atmospheric layers worldwide. GraphCast takes a fundamentally different approach. Instead of attempting to model intricate atmospheric processes, it leverages machine learning to analyze extensive historical weather data, including output from the European Centre for Medium-Range Weather Forecasts (ECMRWF) model, to understand the evolution of weather patterns. This AI-driven approach enables it to predict how current conditions are likely to change in the future, with remarkable precision. GraphCast has demonstrated exceptional accuracy, outperforming traditional models on more than 90% of the factors crucial for weather forecasting. Moreover, it produces forecasts in under a minute, utilizing only a fraction of the computing power required by traditional numerical weather prediction (NWP) models. An illustrative example of its success is its prediction of Hurricane Lee's landfall in Canada in September. The AI tool accurately forecasted the storm's path nine days in advance, surpassing the ECMRWF's six-day prediction window. This extended lead time for forecasting can be pivotal in preparing for extreme weather events, potentially saving lives and mitigating property damage. Crucially, AI models like GraphCast do not supplant traditional weather forecasts but complement them. These AI models rely on data generated by traditional approaches, emphasizing the symbiotic relationship between AI and traditional meteorological methods. Despite advances, climate change brings unpredictable weather extremes, challenging AI models with data quality issues. Rising ocean temperatures introduce a previously unseen variable that can accelerate storm intensification — like Hurricane Otis's swift escalation from a tropical storm to a Category 5 hurricane within 24 hours. GraphCast by Google DeepMind represents a significant advancement in weather forecasting. As climate change continues to reshape weather patterns, AI's role in forecasting becomes increasingly crucial in safeguarding communities worldwide. #JordiPlusJavis Note: This is an #AI generated image

AI is rapidly improving weather forecasting. Today, Google DeepMind published a new paper in Nature presenting GenCast, a new high resolution ensemble model that includes 50 or more predictions for weather forecasts up to 15 days in advance. GenCast is a diffusion model, similar to generative AI models for image, video, and music generation. However, instead of generating frames or pixels, it produces a probabilistic forecast of atmospheric and weather variables encompassing a range of possible future outcomes. This is more useful than a single prediction for things like extreme weather prediction. And it works well! GenCast is more accurate than the leading operational ensemble model, ECMWF ENS, on 99.8% of forecast variables at lead times greater than 36 hours. https://lnkd.in/gT9-BcNU #AI #weather #google #deepmind