Climate & disasters

Storm Daniel and flood risk in Thessaly: what the satellites mapped

Some of the floodwater from September 2023 was still on the fields in mid-December. That alone says a lot about how flood risk works in a flat, farmed basin.

Published: 6 min readKassandra Labs
A seafront promenade lined with palm trees under floodwater and mud, beside a beach under an overcast sky.
Photo: José Antonio Otegui Auzmendi · Pexels License

What happened

Storm Daniel hit central Greece in early September 2023. According to NASA's Earth Observatory, the four-day storm began on 4 September and dropped record rain on Thessaly on 5 September. It was a cut-off low, a low-pressure system that had stalled after separating from the jet stream, and it was part of an unusual "omega" pattern over Europe.

The numbers are extreme. NASA reports that the highest rainfall, 754 mm, fell on Zagora on Mount Pelion, near Volos. Over the four days, stations of the Greek weather service Meteo recorded about 910 mm at Zagora and 659 mm at Karditsa, even though some stations stopped sending data during power cuts. For scale, Athens averages about 400 mm of rain in a whole year.

In Volos, the Krafsidonas river broke its banks and destroyed a bridge. The water then spread across the Thessalian plain. On 10 September a Sentinel-2 image showed the Larissa area still badly affected by the overflowing Pinios river, and by then firefighters had rescued more than 2,850 people, according to the EU Space programme.

What Copernicus EMS mapped

The EU's Copernicus Emergency Management Service (EMS) turns satellite images into maps for civil protection. Its rapid mapping was activated on 5 September 2023 at the request of Greece's General Secretariat for Civil Protection, under the code EMSR692. The first notice, written while rain was still falling, already reported one person dead and one missing around Volos and the Pelion coast. The request was for a first rough estimate, the flood extent and monitoring over time.

The activation grew as the water moved inland. Its public record lists 7 areas of interest (Magnesia, Palamas, Larissa, Stefanovikio, Karditsa, Keramidi and Kalamaki) and 29 products. They are flood delineation maps, repeated as "monitoring" updates through to 22 September. They combine radar images from Sentinel-1, which see through cloud, with very high resolution optical images from commercial satellites.

By 21 September, the EU Space programme reported that the floods had caused damage to over 700 km² of Greece's central agricultural belt, with some fields still under water.

Three months later: Lake Karla

The water did not simply drain away. Copernicus EMS opened a second, slower activation, EMSN184, under its risk and recovery mapping, dated 23 November 2023. It was for Greece's Ministry of Rural Development and Food, to support recovery planning in the area north and north-west of Lake Karla. Using a Sentinel-1 image from 18 December 2023, refined with a Sentinel-2 image from the next day, the analysis found about 10,000 hectares still flooded. An impact assessment combined that extent with reference data and WorldPop population estimates.

National data mattered too. A Copernicus in-situ case study notes that a Greek 5-metre digital terrain model, provided through the General Secretariat for Civil Protection, was used to estimate water depth.

Lessons for flood planning

We draw these from the public record above. They are our reading, not official conclusions.

  1. Plan for the long tail, not just the peak. In a flat basin, the damage continues for months. Recovery plans for farms, roads and services need flood duration, not only maximum extent.
  2. Combine satellites with terrain. A satellite shows where the water is. A good terrain model shows how deep it is and where it will go next. Each needs the other.
  3. Keep hazard maps alive. The EU Floods Directive requires flood hazard maps, flood risk maps and flood risk management plans, reviewed in six-year cycles (the current one runs 2022–2027). Events like Daniel are the evidence those reviews should use.
  4. Replay the event. The EMSR692 maps form a day-by-day record of how the water spread. Running a flood model against that record is one of the best ways to test it before the next storm.

Where a digital twin fits

A flood twin would join rainfall forecasts, terrain, rivers, land use and the location of people and assets, so planners can ask "what if this much rain falls here?" before it does. It cannot remove uncertainty, and it does not replace official warnings or the Floods Directive maps. Kassandra Labs is at concept stage; our flood risk page sets out how such a twin could support planning.

Related

Early access

Want to read it first?

Join the waitlist for new articles and the first pilots.

We only use your details to contact you about Kassandra Labs. You can ask us to delete them at any time.