Recently NIVIDIA announced its latest generative AI model, "Climate in a Bottle,” that simulates Earth’s global climate with an unprecedented level of resolution. The news closely followed Microsoft’s Aurora announcement, an AI foundation model that promises to revolutionize not just weather forecasting but environmental prediction broadly. These announcements are just a few of many over the past year, announcing groundbreaking AI and weather modeling and pointing to a new era of weather forecasting.

As a meteorologist who works with customers in multiple industries, I see infinite possibilities for the beneficial application of AI across organizations, supply chain dependencies and public safety. I also hear the murmurs of how AI will replace human forecasting. With so many conversations around AI and weather, I want to weigh in what I believe AI means for our profession and the future of weather forecasting.

Tech giants like Google, Microsoft, and NVIDIA are pouring massive resources into weather AI models, each claiming breakthrough capabilities. Google's GraphCast made headlines for outperforming traditional forecasting systems. Microsoft's Aurora model promises to revolutionize not just weather prediction but environmental forecasting broadly. Meanwhile, NVIDIA launched Earth-2, a digital twin platform simulating weather and climate conditions in unprecedented detail.

This year the company that I work for launched the DTN Hurricane Threat Index. It is a model that uses AI to assess multiple hurricane hazards beyond wind speed to include storm surge, rainfall, flooding, and tornadic activity. This provides a more complete picture of storm threats up to seven days before landfall for all areas of impact.

These announcements have created both excitement and conversations throughout the meteorological community. Will these powerful new tools transform weather prediction as we know it? Will these improve the role of meteorologists? Or, will they render human forecasters obsolete? Yes, to the first two questions. No, to the third, which I’ll explain more below.

AI has been outperforming humans in raw weather prediction for decades. Weather science embraced computational power long before most industries even understood its potential. Why? Because weather is the original big data challenge.

Back in 1922, Lewis Fry Richardson developed the first numerical weather prediction system. His calculations were so complex that he envisioned a "forecast-factory" with 64,000 people using mechanical calculators to process a single eight-hour forecast. A task that would take six weeks to complete. By the late 1900s, computers had taken over these calculations entirely.

Today AI forecasting models can calculate hundreds of weather variables over ten days at granular resolution globally, in under one minute, a task that would typically take hours on a supercomputer.

Despite impressive advances, current AI weather models have limitations. They can only make deterministic forecasts, not probabilistic ones with ranges of possibilities. However, since the models run so quickly, they can be used to create ensembles, or large numbers of concurrent forecasts. Meteorologists use ensemble data to help determine probabilities and confidence in the forecast.

A recent study from the University of California demonstrated that current AI fails at forecasting “freak events” beyond the scope of existing training data. However, because this landscape is evolving so quickly, I’m sure the processing and data will get better over time.

I expect in the next decade we will see another technological leap in forecasting with quantum computing and physics-informed neural networks. PINNS is a hybrid approach that combines the efficiency and pattern-recognition capabilities of deep learning with traditional physics-based models, such as the European Centre for Medium-Range Weather Forecasts. PINNs are already used for specific applications, but full operational deployment holds transformational capabilities.

An exciting area where I see AI making a difference today is the integration of weather data and industry-specific data streams. This would not have been computationally possible 20 years ago and certainly not in the context of machine learning and predictive analytics.

Logistics companies are a prime example of embracing these capabilities. For example, Amazon’s inventory management system incorporates seasonal weather forecasts to pre-position products and customize delivery commitments based on real-time weather data. Similarly other logistics companies uses weather for delay management and routing management.

Utilities are also in a better position to respond to extreme weather by integrating weather data into their own operational data, such as service areas, infrastructure, and crew capacity. Outage predictions using AI modeling have shown the probability of outages to specific areas as early as a week ahead of impact. A DTN analysis showed this could reduce outage durations by as much as 50% and more effectively size and stage response crews.

Awhile back a friend shared the 2023 Gartner’s video on the prediction about AI’s impact on jobs. Out of 13 categories, weather prediction was the only one where AI outperformed humans today, and that was in 2023. My friend’s reaction was typical. AI would eventually eliminate meteorologists altogether. But that conclusion misses the mark completely. Rather than replacing meteorologists, AI is transforming our role.

Meteorologists are now freed from computational drudgery to focus on what humans do best: interpreting model outputs, applying local knowledge, and communicating weather impacts to decision-makers.

AI excels at pattern recognition and processing vast amounts of data, but meteorologists excel at understanding the "so what" behind the forecast. We translate complex atmospheric science into actionable insights, especially in marginal weather scenarios where a stop, continue or pause decision can make the difference in safety or profit.

Paradoxically, as we are discovering more about how AI can improve forecasting, I increasingly hear customers ask for a “human in the loop” or meteorologist to interpret and advise on the forecasts. One of the most common remarks is, “I see the data and I read the forecast, but I am not a meteorologist. The stakes are high, and I want to consult with an expert that understands my business.”

As a meteorologist, I'm not worried about AI taking my job. I'm excited about how it's making me better at my job and how it is going to improve weather decision-making, planning and responding. The science of meteorology has always been at the forefront of technological innovation, AI and weather is just another evolution of how we help the public, emergency responders and businesses be more weather resilient.