Photo:The machine used to digitise the old film reels. The machine is located at Stanford University in the USA but was brought to Lyngby to digitise the radar archive. 8Photo: GEUS.)

Researchers have digitised the film reels and assembled the images into coherent radar pictures. The printed technical reports with coordinates and dates were scanned and processed through an OCR (optical character recognition) programme. A significant amount of manual work was required to ensure the information's accuracy.

"Sometimes the OCR program would read a 0 as an 8 or miss a minus sign in front of a coordinate, which made the flight lines look completely wrong. We had to go back to the reports, identify the error, and correct it manually," explains Nanna B. Karlsson.

Figure: The blue lines on the map represent the flight lines from 1971, 1972, 1974, 1978, and 1979. (Excerpt from Figure 2.)

Much more than just ice thickness

When digitising old data, the quality is often uncertain. The primary goal of collecting radar data in the 1970s was to measure ice thickness. Therefore, the data were logged with coordinates, and the developed images, like the one below, were used to record the depth from the ice surface to the bedrock.

Figure: Example of a radar image showing a vertical slice through the Greenland Ice Sheet from the surface to the bedrock. The green box indicates the recording date, which in this case is May 1979. The black box shows the flight line number (flight line 9), and the yellow box highlights the depth to the bedrock. The bedrock signal is seen as a line beneath the ice. The blue box shows the unique ID number assigned to each radar image. This radar image spans approximately 5.6 km in length, and the ice thickness ranges from 1.7 km to 1.9 km. (Figure 3.)

The quality has proven to be so good that in many places, the internal layers of the ice sheet are visible. This gives Nanna B. Karlsson and her colleagues hope that the data can provide new insights into the distribution of fast-flowing ice in the ice sheet.

"There are areas where fast-flowing ice exists now. We want to know if there was also fast-flowing ice in the same areas in the 1970s, or if the areas have expanded since then," says Nanna B. Karlsson.

Fast-flowing ice is one of the most significant contributors to the mass balance of the ice sheet. The ice sheet has lost mass for 27 consecutive years. When the ice sheet loses mass, global sea levels rise. To create realistic estimates of future global sea level rise, researchers need to understand the dynamic ice sheet, and every piece of data helps.

Knowledge about the ice's movement speed before the 1990s is limited due to scarce data. Therefore, having 50 years of data instead of the 30 years previously available makes a significant difference.

Over ten years in the making

Nanna B. Karlsson has led the extensive digitisation project, which has required considerable coordination and patience. The digitisation process has been more than a decade in the making.

"I first heard about these film reels with radar data from the 1970s at a conference in 2013, where Preben Gudmandsen, the original radar pioneer, mentioned the data stored in a basement at DTU. Since then, the project has been a complex puzzle of skilled people's hours and funding. But we succeeded," says Nanna B. Karlsson.

The late Preben Gudmandsen was a professor at DTU and initiated the radar data collection via overflights in the 1970s. He witnessed parts of the digitalisation process before his death.

There was no doubt among the researchers that the reels should be brought out of the basement and made accessible to everyone. It has also been refreshing for the researchers to collaborate on something analogue.

"At a time when there’s so much focus on AI data analysis approaches that promise fast payoffs from ingesting large amounts of data, it's refreshing and inspiring to work on a collaborative project dedicated to recovering a half century of ice-sheet subsurface history from slow, careful analysis of remastered archival data," says Dustin M. Schroeder, an associate professor at Stanford University who purchased and operated the scanner that enabled efficient, high-quality digitisation of the reels.

Photo: The photograph is from Kangerlussuaq, Greenland, circa 1978. Radar data were collected using American Hercules LC-130 aircraft. The radar antennas can be seen under the wing. (Photo: Niels Skou).