From May to October each year, the United States enters what climate scientists and emergency managers now call the “danger season.” It’s been a six-month ordeal as hurricane season peaks, heat domes settle over cities for days, wildfires spread across millions of acres, and flash floods tear through communities with little warning.
AI-powered predictions are constantly improving and can help keep some people safe by helping us understand when it’s time to stock up on food, get off the roads, and evacuate before a major storm. But whether in an uninsured mobile home park in America or a rural village in Madagascar or Nigeria, millions of people may receive the same warning, but with no safe haven to flee to or the financial means to protect their livelihoods.
More than 15 years of experience developing models that use satellite data to predict droughts and assess the impact of floods has shown us that the barrier to saving lives is rarely the accuracy of the data, but rather the lack of the policy infrastructure needed to act on it. Billions of dollars are being poured into data centers, but physical infrastructure for resilience, such as sea walls and grain bins, is lagging behind. This AI-first bias also overlooks “ground truth” data and fact-based local validation from tools like rain gauges and soil sensors. Without this ground-level input, AI lacks the credibility local policymakers need to take life-saving actions.
The funding gap reveals the most obvious mismatch. The main international relief mechanism, the Loss and Damage Response Fund, receives less than 0.1% of actual needs, leaving insufficient funding for interventions such as early warning and pre-planned disaster response.
And in the United States, a systemic overhaul of FEMA could seriously undermine disaster response, including cuts to critical disaster mitigation programs, changes in leadership, and shifting financial burdens and responsibilities to states.
Global funding disparities are enormous, but the growing movement around anticipatory action (the practice of deploying resources before, rather than after, disasters) and planned transfers (moving vulnerable communities out of harm’s way before disaster strikes) is a reminder of what is possible when better predictions, faster coordination, wiser resource deployment, and true local leadership show the world what is possible when the world chooses and acts before disasters strike. Although these approaches are still limited in scope and inconsistently implemented, they show that acting early, such as pre-positioning supplies, activating automatic financing, and moving populations before storms make landfall, is much more cost-effective than reactive aid.
Federal Insurance Mitigation Administration data consistently shows that every $1 invested in disaster mitigation saves $6 in future recovery costs. In Europe, research has found that investments in coastal flood adaptation provide a return of €6 for every €1 invested. Even more impressive, every dollar invested in climate change adaptation around the world generates a return of more than $10.50 over 10 years, an average return of 27% per project.
These are not abstract numbers. They represent real lives saved, livelihoods protected, and communities that remain intact rather than being displaced. But data alone is not enough.
Without continued investment in local leadership and an integrated framework for delivering large-scale disaster interventions, their promises will remain undelivered. Bridging this gap is essential, especially given the clear and compelling return on investment that disaster prevention efforts continue to deliver.
Addressing escalating climate risks requires combining physical capabilities with AI and data infrastructure. Crews will clear roads, the power grid will keep power flowing, and supplies will be pre-positioned before the storm hits. The goal is something like FEMA-level response capacity for each country, but that capacity varies widely. If it doesn’t exist, even the most sophisticated predictions will lengthen the wait for the inevitable. High-tech alarms save lives only if local actors have the financial wherewithal, legal authority, and trained personnel to act.
The 2026 Danger Season should be the moment when funding priorities finally shift away from AI-first and AI-only strategies. While big tech companies are leading the way in AI and prediction, international financial institutions can’t take their eyes off the ball. Digital tools are only as effective as the physical infrastructure and field expertise they support. By prioritizing the brick-and-mortar and human capabilities of local organizations, investors can finally bridge the gap between intelligence and action.
I believe that for every penny allocated to AI and digital infrastructure, approximately $10 should be devoted to the local technical capacity and physical infrastructure needed for response and action. Without this rebalancing, world-class forecasting will remain a digital blueprint with no people or means to build it.
Katherine Nakarembe is a professor of translational geoAI at the University of Maryland, where she bridges geographic science and artificial intelligence to address real-world problems. He is also a Public Voices Fellow on Technology in the Public Interest for the OpEd project.
thisarticleteeth,Yale Climate Connections First published inCreative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
