AI at the Train Station: Why Floodplain Managers Should Pay Attention to Artificial Intelligence

By Rachael Orellana & Skye King

You’ve probably heard the buzz about AI, but what does it mean for floodplain management? It turns out, quite a lot. From satellite imagery to community reports, AI is already helping us make sense of complex data and act faster during emergencies.

Think of AI like a bullet train: it’s pulling in and out of stations whether you’re ready or not. You don’t have to get on right away. It’s fine to watch one pass and decide to board the next. But pretending the tracks don’t exist? That’s not an option.

At its core, AI is just another tool in the floodplain management toolbox that can speed up analysis, improve communication, and help us prepare for a future with more extreme weather. In this blog, we’ll review a few ways AI is already being applied by flood risk managers and discuss how these examples may shape the practice moving forward.

Why AI Matters for Floodplain Managers

Floods are striking harder and more often across the U.S. In 2024, they caused an estimated 3.9 billion in property and crop damage. Traditional systems like paper reports, spreadsheets, and slow simulations are often stretched to the limit. AI offers a way to turn overwhelming complexity into actionable clarity.

Instead of processing mountains of data yourself, AI can sift through satellite imagery, water-level sensors, social media posts, and inspection records in minutes. It’s not a replacement for expertise. The real work still falls to managers, engineers, and communities to make decisions, but the insights come faster and with more precision.

Importantly, not every solution requires a huge investment. Some platforms, like ISeeChange, are free for residents and affordable for agencies, making AI accessible even for smaller communities. And as Adobe’s Bill Kirst has put it, the real question isn’t “can you trust AI?” but “can you trust yourself with AI?” Used wisely, it strengthens judgment rather than replacing it.

Example 1: FloodHub – Forecasting at Scale

Google’s FloodHub is an AI-powered platform that provides early flood forecasts, up to 7 days in advance, by combining a Hydrologic Model (river flow) and an Inundation Model (flooded areas and water levels) to show where and how severely floods will impact communities. It currently covers river basins in more than 100 countries, reaching more than 700 million people, and has issued millions of alerts in regions where traditional warning systems are limited.

In the U.S., FloodHub is being piloted in select river basins. This short video contains a brief overview about the publicly available data the model is trained on. But FloodHub also highlights an important point: forecasts are only as good as the systems and communication around them. In a recent European flood, models accurately predicted the disaster days in advance, but the warnings didn’t translate into effective response (Reuters).

Insights for Practice: Forecasts are only useful if they connect to your communication and response systems. AI can sharpen the signal, but floodplain managers still need to act on them.

Example 2: DeltaERT – Smarter Emergency Response

In California’s Sacramento-San Joaquin Delta, levee failures threaten both communities and statewide water supply. The Delta Emergency Response Tool (DeltaERT), developed for the California Department of Water Resources by GEI helps by rapidly testing breach scenarios and suggesting response strategies. Think GPS rerouting during a traffic jam.

Its machine learning feature proposes tailored actions while keeping humans in the loop for oversight. DeltaERT demonstrates how floodplain managers can move from fixed plans to flexible ones that adapt as flood conditions unfold.

Insights for Practice: Adaptive AI tools can help floodplain managers move beyond static plans. Use them to stress test scenarios in advance and to support quicker adjustments during a crisis.

Example 3: ISeeChange – Community Data Meets AI

ISeeChange collects community and customer observations—photos, posts, 311, and inspection reports—and translates them into usable, structured insights for agencies.

 

In one case, a resident reported two feet of street flooding. ISeeChange AI analyzed the flood height, extent, and volume and confirmed that more severe flooding was happening nearby, out of sight and threatening property. Crews checked, and the AI prediction proved accurate.

This kind of “ground truthing” fills gaps in official datasets and gives communities a direct role in shaping flood-risk decisions. Beyond technical modeling, AI can also help agencies harness lived experience and strengthen interagency data sharing.

Insight for Practice: Community observations, when paired with AI, can fill critical data gaps. Consider ways to integrate resident reports into your monitoring or planning systems, especially in areas where sensors are sparse.

Risks and Realities

AI isn’t perfect and doesn’t deliver value on its own. It depends on good data, careful training, and effective human oversight. Even the best AI tools can fall short without policy guardrails, adequate funding, and local expertise to guide how they’re used. These systems also rely on timely, accurate information. If the inputs are wrong or delayed, the outputs can mislead, especially in the middle of a disaster.

Another consideration is over-reliance. AI may recommend strategies or generate forecasts, but it can’t replace the context and judgment that experienced floodplain managers bring to the table. That expertise is the antidote to false confidence and overlooked local nuance.

Responsible use of AI in flood risk management means being transparent about when and how it’s applied, monitoring for environmental costs and information bias, and always pairing machine insights with human analysis. If we don’t start engaging with these tools, we risk standing on the platform while others decide where the train is headed.

Conclusion: Standing in the Station

If you’re reading this, you’re already in the AI train station. That’s progress. You don’t have to leap onto the very next train, but you do need to notice the schedule. The trains are coming, one after another, and eventually you’ll want to be on board.

Floodplain managers and other leaders who start experimenting now will shape the standards and practices the rest of the field will follow. With climate-driven floods accelerating, AI is quickly becoming part of the toolkit. It’s not a replacement for your expertise, but a way to make your expertise go further.

You don’t need to be a coder or an early adopter to start. Begin small:

• Review your agency or organization’s policy on digital tools.
• Try AI for simple everyday tasks like summarizing long reports, creating agendas, or even brainstorming questions for upcoming meetings.
• Explore platforms already available, like FloodHub and IseeChange.
• Notice where AI tools could save time at work by helping you to sort options, categorize documents, unlock data in PDFs, or forecast what could come next.

The learning curve is real, but you’re not facing it alone. Every conversation, experience, and pilot builds collective resilience. The question isn’t whether AI will reshape floodplain management. The real question is whether we will guide that change or simply watch it pass us by. Want to explore further? Contact Rachael (rorellana@geiconsultants.com) and Skye (skyeking@geodelifeworks.com).