If you think about weather and climate, they are inherently uncertain, highly non-linear, and turbulent systems. Even with all the advances we have made, we still cannot fully resolve many of the governing equations computationally. We rely on approximations, what we call parameterizations, to represent key processes. That’s where this convergence with the AI/ML world becomes so powerful. 

By coupling traditional physics-based CFD models with AI/ML, we can start to get around some of those limitations. These systems can learn from massive amounts of data and begin to recognize patterns and relationships in ways that complement, and sometimes extend, what our physical models can do. In a sense, they are learning from experience, much like humans do. 

With more frequent and intense extreme weather events that are directly impacting infrastructure, economies, and communities, the question becomes not just how we advance the science, but how fast we can translate that science into usable solutions. 

For me, Climate + AI is really about that bridge: connecting climate and weather predictions to real-world impact models, so decision-makers have better, faster, and more actionable information to prepare for what is coming. 

At the DPI Climate Hub, a big part of my team’s focus right now is on developing AI-enhanced weather forecasting systems for the State of Illinois. What we are really trying to do is build on the strong foundation of physics-based models, which are incredibly important, but also very computationally intensive.   

By bringing AI/ML into the workflow, we can speed things up and significantly improve the resolution of our predictions. What excites me is that we are now getting to a point where we can look at weather at a much more granular level, down to hyper-local, even at street-scale, and quickly. That is a big shift from how forecasting has traditionally worked. This gives us an opportunity to use advanced CPUs and GPUs to connect to real-world impacts, e.g., anticipating flooding at critical locations, understanding heat stress in a neighborhood, or preparing for severe storms: information that people can actually use. 

We at the DPI Climate Hub are using AI/ML in our Climate Hub projects, which complement our computationally intensive physics-based models. These projects are helping reach unprecedented high resolution at hyper-local (street) scale. We are also preparing to use AI/ML to provide faster and more accurate information for emergency management, infrastructure planning, and public safety across the state by delivering more accurate and timely weather intelligence.   

At the DPI Climate Hub, our mission is to advance actionable science for climate-resilient investments and decisions on the ground. We are not just studying the climate/weather risks, but also working closely with government agencies across jurisdictions, industry, and academic partners to build something I can truly call “a climate and weather intelligence hub”. One of the things I find most important is that we operate across multiple time scales. 

When severe storms or other extreme weather events are unfolding, we focus on high-resolution forecasting across Illinois to support emergency preparedness and rapid response. 

We also focus on longer-term planning to account for climate. For example, our team works on where we should place critical infrastructure so it is less vulnerable to flooding. How do we help utilities prepare for increasing stress on the grid during extreme heat? And how can we bring in nature-based solutions, especially in places like Chicago and the Quad Cities, to reduce future flood risks?  

For me, it is really about connecting the science to decisions that shape cities and infrastructure in a more resilient way. 

Now we can analyze extremely large datasets and uncover signals and patterns that would be difficult or impossible to detect using traditional approaches alone. For climate science, this means we can better capture complex relationships across systems: how atmospheric processes interact with land, infrastructure, and human activity, and see it as an integrated system with real-world consequences. 

Another breakthrough is using ML as a kind of fast “emulator” for complex physical models. Traditionally, these models take a huge amount of computational time, which limits how many scenarios we can explore. With AI/ML, we can run simulations much faster, test different scenarios, and improve the resolution of our forecasts. This allows us to move beyond just predicting weather or climate variables. We can start connecting those predictions directly to impacts, i.e., things like stress on infrastructure, energy demand during extreme heat, flood risks, or disruptions to transportation systems. 

AI/ML can help us shift from just predicting weather and climate conditions to delivering accurate, actionable climate intelligence that people can actually use to make decisions. 

Ashish Sharma leads the Climate Hub at the Discovery Partners Institute. He also serves as Director of the Clean Energy and Equitable Transportation Solutions (CLEETS) Global Center, based out of DPI in Chicago. In recognition of his leadership and contributions to climate sustainability, he was named to Crain’s Chicago Business 40 Under 40 Class of 2023.