Bridging the Water Data Divide – A Q&A with Donghai Wang

By Quinn Fenger and Donghai Wang

Water managers today are swimming in data. River gauges, reservoir sensors, weather forecasts, and hydrologic models churn out terabytes of information, yet integrating these diverse streams into coherent insight remains a challenge. This fragmentation means that even seasoned professionals spend valuable time piecing together spreadsheets, portals, and static maps instead of focusing on analysis and action.

Donghai Wang has been tackling this problem for decades. As a registered civil engineer and principal software engineer at GEI Consultants, he leads the firm’s Information Management Division. Over his career he has delivered more than a hundred water resource information management applications for local, state, and federal agencies. For the 2025 Floodplain Management Association (FMA) conference, he’s presenting a session titled “Tools for Water Data Management: Enhancing Retrieval and Reporting with GIS and Large Language Models.” In this interview, Donghai explains why data integration matters, how geographic information systems (GIS) and artificial intelligence (AI) can bridge the gap, and why he’s excited about the next wave of technological innovation.

Q: Why should FMA attendees care about your session on water data management?

A: Water data is everywhere, but getting at it is hard. Today, even a simple request like “Show me the current water level near my house” requires hopping across multiple websites or databases. You might find a flood‑risk map in one portal, dam details in another, and real‑time streamflow data somewhere else. Interpreting all that and packaging it in a way that managers or the public can understand is challenging. My presentation will show how we’re using GIS and AI to make data retrieval and reporting faster and easier.

Q: Your resume lists you as both a civil engineer and a software engineer. How did you end up linking those two worlds into your current position?

A: It was a long process. I started in hydraulic modeling, which requires enormous amounts of data—everything from terrain and land use to flow measurements. To run models, you have to preprocess the inputs, post‑process, the outputs and visualize the results. That naturally led me into data management. Over time I took on more software development to support modeling, and recently, with advances in machine learning and AI, I see opportunities to automate what used to be painfully manual.

Q: What are the biggest pain points you see when agencies try to retrieve and share water data?

A: For large organizations, data is often siloed within different branches or units. One team manages flood maps; another handles real‑time hydrology; a third oversees dam operations. These groups may not even know what each other are collecting. Smaller local agencies face a different challenge: they often lack funding to install sensors or collect high‑resolution data, and they don’t have IT staff to manage or analyze what data they do have. So there’s an equity issue on top of a technical one.

Q: GIS is a backbone of water management. How are you using it in your tools?

A: GIS has long been used for mapping—people pull up a flood‑risk layer or an inundation map—but the underlying data often stays disconnected from other information. We’re trying to bridge that gap by integrating GIS with databases, documents and real‑time feeds. A user should be able to enter an address and ask: “What’s the current water level at the nearest river? Which dams are upstream? If a dam failed, would my property be inundated?” Our tools use spatial queries to locate the relevant features, then pull data from the dam database, flood maps, and real‑time sensors to assemble a complete picture.

Q: You’re also incorporating large language models (LLMs). What do they add to the mix?

A: People don’t just want numbers; they want stories. If a system spits out a table of water levels and dam heights, most users won’t know what to do with it. LLMs can take the query results and generate a human‑readable report—“The American River is currently at X feet; the nearest dam is Y; under a 100‑year flood your area would be inundated.” They can also assist with search, forecasting, and summarizing. We’re still in the early stages, but I see huge potential for LLMs to translate data into actionable insight.

Q: What about integrating all those disparate data sources—stream gauges, weather forecasts, reservoir levels, ground sensors? How do you handle that?

A: That’s the hard part. LLMs don’t have access to proprietary data, so we have to bring everything into our own system first. Our initial step is to present all the data we have, but the goal is to go further: combine time series from streamflow and weather with dam operations and other model outputs to produce actionable recommendations. That involves standardizing formats, ensuring data quality, and building analytics on top.

Q: Has anything surprised you as you’ve developed these tools?

A: One thing that struck me is that even in large organizations with dozens of data portals and applications, managers still struggle to get a holistic view. I’ve heard from clients about how even though they’ve been at the agency for their whole career, they may not immediately know where to find the right data themselves. If they want data from a groundwater level, he has to call the groundwater data team; for a water quality data, he calls another group; for reservoir operations, yet another. It shouldn’t be that hard to get a high‑level summary, but the vast amounts of data we deal with creates that challenge. We’re aiming for a system that meets 80–90 % of users’ needs automatically, so they only have to reach out to specialists for deep dives or further analysis rather than questions about where to find the data.

Q: Looking ahead, what emerging technologies are you excited to incorporate?

A: AI is advancing almost daily or weekly. In other data industries, image recognition and pattern analysis are already routine—for example, radiologists use AI to spot patterns in medical scans. We want to leverage similar techniques for water resources: use video or imagery to identify objects on a levee or along a riverbank and help inspectors categorize and prioritize issues. I’m also interested in digital twin and real‑time feeds that can integrate sensor data directly into models. There’s a lot we can borrow from other fields.

Q: Any final thoughts for FMA attendees or readers?

A: Be open to AI and machine learning. What we’re doing is still exploratory, but it shows that we can leverage technologies from other industries to advance the water sector. If you come to the presentation you’ll see some concrete examples, and hopefully you’ll leave with ideas for how to integrate GIS, AI, and modern data management into your own work.