Goodbye Mr. Weatherman, Hello AI Forecasts

Welcome to this week’s Deep-Fried Dive with Fry Guy! In these long-form articles, Fry Guy conducts in-depth analyses of cutting-edge artificial intelligence (AI) developments and developers. Today, Fry Guy dives into AI’s ability to predict the weather. We hope you enjoy!

*Notice: We do not receive any monetary compensation from the people and projects we feature in the Sunday Deep-Fried Dives with Fry Guy. We explore these projects and developers solely to showcase interesting and cutting-edge AI developments and uses.*

You packed your beach bag, slathered on sunscreen, and picked the perfect spot on the sand—only for the sky to turn gray, the wind to pick up, and the weatherman’s “clear skies” forecast to betray you. Sound familiar?

For years, weather predictions have been hit or miss, sometimes leaving us soaked in unexpected downpours. But now, AI is stepping in to take the guesswork out of forecasting. From predicting extreme storms with pinpoint accuracy to tracking wildfires before they spread, AI-driven models are revolutionizing how we prepare for and respond to the forces of nature.

Let’s dive into how AI is making weather forecasts smarter, faster, and (hopefully) less likely to ruin your beach day.

AI’S FORECASTING WIZARDRY

There are many companies, both in the United States and abroad, that are looking for ways to use AI to more reliably predict the weather. Many of these methods leverage machine learning to pick up on past weather patterns and render predictions based on live data.

One company that is actively researching ways to use AI for weather predictions is Chinese tech company, Huawei. Huawei has been developing an AI model called “Pangu-Weather” which can predict weekly weather patterns around the world much more quickly and accurately than traditional forecasting methods.

The Huawei researchers constructed Pangu-Weather by developing a deep neural network that learned from 39 years of reanalysis data. This network integrates historical weather observations with contemporary models. In contrast to traditional approaches that examine weather variables individually—a time-consuming process which can last hours—Pangu-Weather can analyze all variables simultaneously within a matter of seconds to issue accurate forecasts. If adopted, these models could be used alongside conventional weather prediction methods to more accurately and efficiently predict bad weather, which will give people more time to prepare for rainy days.

Although Pangu-Weather excels at normal, everyday weather predictions, Lingxi Xie, a senior researcher at Huawei, said one area to be further developed is AI’s ability to track the severity of weather. Xie said, “AI will tend to underestimate extreme weather.” Nevertheless, additional AI models could provide support in that regard. One such model is NowcastNet, which utilizes a physics-based generative approach to forecast extreme rainfall much quicker than conventional methods. NowcastNet can successfully predict intense rainfall up to three hours ahead. Combining this sort of system with Pangu-Weather could lead the way to the future of weather prediction.

Beyond Pangu-Weather and NowcastNet, Google researchers have released a weather prediction model named “NeuralGCM,” which, like Pangu-Weather, merges machine learning with traditional forecasting methods. However, NeuralGCM does this a bit differently.

NeuralGCM bridges the gap between fast, data-driven machine learning techniques and the highly accurate but computationally intensive general circulation models (GCMs). So instead of learning from mere historical weather patterns, NeuralGCM uses AI to simulate the earth’s atmosphere and enhance small-scale predictions, such as cloud formations and microclimates, to predict weather patterns. NeuralGCM has demonstrated accuracy comparable to the European Centre for Medium-Range Weather Forecasts (ECMWF) for up to 15 days in advance.

Experts see potential for NeuralGCM beyond local forecasts, envisioning applications in predicting large-scale climate events and extreme weather risks. The possibilities could range from predicting tropical cyclones with more notice to modeling more complex climate changes that are years away. In this way, NeuralGCM’s efficiency could revolutionize climate modeling, making long-term predictions more accessible and cost-effective.

Probably the most popular AI weather prediction model is Google’s GenCast. GenCast uses a diffusion model—similar to the tech behind image and video generation—tailored specifically for Earth’s spherical shape. It is trained on historical weather data to generate over 50 forecasts, each representing a possible future scenario. Using real-time weather updates from traditional systems, GenCast evaluates all these scenarios to offer both day-to-day and extreme weather predictions, including confidence ratings about such predictions.

It’s no secret that accurate weather forecasts can save lives and inform the allocation of resources, especially during extreme events like hurricanes and heatwaves. GenCast generates predictions faster than traditional systems, producing forecasts in minutes instead of hours. GenCast also improves prediction accuracy, particularly for high-impact events, up to 15 days in advance. By making its model and data freely available, Google wants to empower researchers, governments, and organizations to enhance their own forecasting systems, improving global disaster preparedness and resource management.

Now, will AI completely replace weather reporters and meteorologists? This is unlikely in the immediate future. AI will likely not be the primary cause of weather reporters losing viewership. As of 2023, before AI started being explored as a primary source for weather prediction, only 27% of people turned on their local news to get information about the weather. Most people (53%) rely on weather apps as their primary source for information, regardless of whether AI is involved in the predictions. So those few who enjoy watching a human report on these findings will probably be able to do so for the foreseeable future. Moreover, technology has continually transformed the work of meteorologists over the past few decades, and they will likely be happy to have more tools in their arsenal.

NATURAL DISASTER RELIEF

AI isn’t limited to predicting weather patterns; it is also being used to be more proactive when it comes to natural disasters like fires and floods.

Google Research has partnered with experts to develop FireSat, a satellite constellation dedicated to wildfire detection. Wildfires are increasingly common due to hotter, drier climates, and traditional satellite imagery often fails to detect fires until they have grown significantly. In 2023, 55,571 fires burned 2,633,636 acres of land. This is where FireSat steps in to help. The project uses AI and advanced infrared sensors to monitor potential fire activity and help prevent disasters. FireSat can spot fires as small as 5x5 meters, roughly the size of a classroom, and can provide global updates every 20 minutes. If a fire is identified, FireSat can immediately notify first responders.

FireSat is a game changer for firefighting efforts, allowing authorities to respond before fires get out of hand and become incredibly destructive. With $13 million in funding from Google.org and support from the Earth Fire Alliance and the Moore Foundation, FireSat aims to massively improve the ability to track wildfires going forward, throwing out traditional methods and leaning into the power of AI. This initiative is part of Google’s broader commitment to combating climate change with cutting-edge technology and research.

AI is not only limited to fighting fires; it is also being used to mitigate the impact of floods. MIT scientists have developed a tool to help communities prepare for hurricanes and other large storms by visualizing potential flooding. The method, called the “Earth Intelligence Engine,” combines AI with a physics-based flood model. Through this approach, the research team has been able to create realistic satellite images showing how a storm might impact a region. As a test, researchers applied the method to Houston, generating images of flooding similar to what occurred during Hurricane Harvey in 2017. These AI-generated visuals closely mirrored real post-storm satellite images.

By combining science and technology, the Earth Intelligence Engine is poised to become a vital tool for improving public readiness and saving lives in the face of climate disasters. If this approach continues to prove reliable, it could inform decisions about where to allocate resources and could guide residents in deciding whether or not they should evacuate.

A MORE RELIABLE DAY AT THE BEACH?

So, will AI finally put an end to those ill-fated beach trips ruined by surprise storms? Maybe not entirely—weather will always have a few tricks up its sleeve—but it’s certainly making forecasts more reliable than ever.

With AI models like Pangu-Weather, NeuralGCM, GenCast, and more, we’re moving toward a future where storm warnings come earlier, flood predictions are more precise, and wildfires are spotted before they spread. These breakthroughs don’t just make life more convenient—they save lives and help communities prepare for extreme weather events with confidence.

So next time you check the forecast before heading to the beach, you might just owe your perfect sunny day to AI. And if it still rains? Well, at least you can blame the machines this time.