What is the JRC global surface water dataset?

What the dataset actually is

The JRC Global Surface Water dataset comes out of the European Commission's Joint Research Centre. A researcher named Jean-François Pekel led the original analysis, which is why a lot of people in the field still call it "the Pekel dataset" or "Pekel layers" when they're pulling it into QGIS. It's built from the full Landsat archive, every usable scene from 1984 through the present, classified pixel by pixel into water or not-water, then rolled up into change detection, seasonality, occurrence, and recurrence layers at 30 m resolution.

For a planner working at the basin scale, the value sits in the history behind any single scene. You can ask the dataset where water has been present every year for the past four decades, where it only shows up in wet years, where a reservoir's footprint has crept or shrunk, where a wetland that used to flood every spring stopped doing that around 2004. That's a different question than how much water is there today, and the JRC layers answer it well because the archive goes back further than almost anything else public.

Where it helps and where it runs out of road

Basin offices use the JRC data for baseline work: establishing historical extent for a lake before a permitting fight, checking whether a channel realignment actually happened or just looks that way in a flyover photo, filling in the long-term context section of a water allocation plan. It's free, it's global, and the methodology has been published and picked apart for years, so you can defend it in a technical review.

The limits show up fast once you're running the basin day to day. Landsat's native revisit is on the order of two weeks per path, less in practice once you account for cloud cover, so the occurrence and recurrence layers read as a multi-decade climatology rather than a weekly feed. The global processing also means local calibration can miss things a basin office would catch: a narrow braided channel that reads as noise, a small irrigation reservoir near the classifier's resolution limit, surface water sitting under tree canopy. None of that is a knock on the dataset. It answers what this basin's surface water has looked like since 1984, which is a different job than tracking what it looks like this week.

That gap between long-run history and this-week operations is the one most basin authorities end up patching with whatever's on hand: a field visit, a phone call to the irrigation district, a gauge reading that only tells you about the channel it sits in. Gauges are accurate where they are. A basin also has lakes, reservoirs, and reaches between stations that a gauge network was never built to see.

Using the archive alongside a weekly layer

The move that closes that gap is pairing a multi-decade record like the JRC archive with something that updates often enough to track a season while it's still happening. Surface Water Monitoring was built around that pairing: a weekly wide-swath pass gets turned into a surface water mask for every lake and reach in your basin, then stacked against the long-run archive so you can see, water body by water body, where this year's extent is running above or below its own history. You get the decades of baseline the JRC data is good for, plus a current-conditions layer on top of it, without re-deriving the history yourself every time someone asks how this spring compares to prior years.

If your basin's planning files still lean on gauge readings and an occasional Pekel pull for context, take a look at how that combination of archive and current read works on our home page.

Get in touch if you want to see your own basin's water bodies read against that kind of long-run record.