Editor’s Word: This CNN sequence is, or was, sponsored by the nation it highlights. CNN retains full editorial management over material, reporting and frequency of the articles and movies throughout the sponsorship, in compliance with our policy.
CNN
—
On the northern fringe of the Rub al-Khali, there are secrets and techniques buried within the sand.
The huge 250,000 sq. miles (650,000 sq. kilometer) desert on the Arabian Peninsula is named “The Empty Quarter.” And to most, apart from waves of ocher dunes, it does look empty.
However to not synthetic intelligence.
Researchers at Khalifa College in Abu Dhabi have developed a high-tech resolution to looking large, arid areas for potential archaeological websites.
Historically, archaeologists use floor surveys to detect potential websites of curiosity, however that may be time-consuming and troublesome in harsh terrains just like the desert. Lately, distant sensing utilizing optical satellite tv for pc photographs from locations like Google Earth has gained recognition in looking massive areas for uncommon options — however within the desert, sand and mud storms usually obscure the bottom in these photographs, whereas dune patterns could make it troublesome to detect potential websites.
“We wanted one thing to information us and focus our analysis,” says Diana Francis, an atmospheric scientist and one of many lead researchers on the mission.
The staff created a machine studying algorithm to research photographs collected by artificial aperture radar (SAR), a satellite tv for pc imagery approach that makes use of radio waves to detect objects hidden beneath surfaces together with vegetation, sand, soil and ice.
Neither know-how is new: SAR imagery has been in use for the reason that Nineteen Eighties, and machine studying has been gaining traction in archaeology. However using the 2 collectively is a novel utility, says Francis, and to her information, is a primary in archaeology.
She educated the algorithm utilizing information from a web site already recognized to archaeologists: Saruq Al-Hadid, a settlement with proof of 5,000 years of exercise that’s nonetheless being uncovered within the desert exterior of Dubai.
“As soon as it was educated, it gave us a sign of different potential areas (close by) which might be nonetheless not excavated,” says Francis.
She provides that the know-how is exact to inside 50 centimeters and might create 3D fashions of the anticipated construction that can give archaeologists a greater thought of what’s buried under.
In collaboration with Dubai Tradition, the federal government group that manages the positioning, Francis and her staff carried out a floor survey utilizing a ground-penetrating radar, which “replicated what the satellite tv for pc measured from area,” she says.
Now, Dubai Tradition plans to excavate the newly recognized areas — and Francis hopes the approach can uncover extra buried archaeological treasures sooner or later.
Utilizing SAR imagery will not be widespread in archaeology, attributable to its cost and complexity.
However using it to determine buried websites is “actually thrilling,” says Amy Hatton, a PhD scholar on the Max Planck Institute for Geoanthropology, who’s researching deep studying fashions to detect archaeological buildings in northwest Saudi Arabia.
Hatton notes that, by utilizing SAR imagery, which bypasses the issue of sunshine scatter from mud particles, Francis and her staff solved technical particulars that make distant sensing troublesome in desert areas.
Khalifa College isn’t alone in utilizing synthetic intelligence to detect potential websites.
Amina Jambajanstsan, one other PhD scholar on the Max Planck Institute, is using machine learning to hurry up the “tedious job” of looking by way of high-resolution drone and satellite tv for pc photographs for potential websites of curiosity. Her mission, which focuses on medieval burial websites in Mongolia — a rustic of greater than 1.56 million sq. kilometers, practically the scale of Alaska — has uncovered 1000’s of potential websites that Jambajanstsan says she and her staff may by no means have discovered on the bottom.
Jambajanstsan says that whereas the fee and computational demandsof SAR imagery could possibly be a barrier to utilization for lots of researchers, the tactic is effective for desert areas the place different applied sciences wrestle — and is one she would think about for the Gobi Desert in Southern Mongolia, the place her “regular optical imagery” will not be yielding outcomes.
Machine studying is discovering an increasing number of purposes in archaeology, though not all researchers are enthusiastic about it.
“There are two distinct perception programs,” says Hugh Thomas, an archaeology lecturer on the College of Sydney and co-director of the prehistoric AlUla and Khaybar excavation mission in Saudi Arabia. On one facet, persons are pursuing technological options like AI to determine websites; on the opposite, those that consider you want a “educated archeological eye” to determine buildings, he explains.
Whereas know-how may assist determine and monitor archaeological websites — significantly ones beneath risk from land use modifications, local weather change, and looting — Thomas is cautious of over-reliance on it.
“The way in which that I wish to use this sort of know-how is on areas that maybe have both no or a really low likelihood of archaeological websites, due to this fact permitting researchers to focus extra on different areas the place we count on there to be extra discovered,” says Thomas.
The actual take a look at — and hopefully, validation — of the know-how will occur subsequent month when excavations start on the Saruq Al Hadid advanced, of which solely an estimated 10% has been uncovered throughout a 2.3-square-mile (6.2-square-kilometer) space, in accordance with Dubai Tradition.
If archaeologists discover the buildings the algorithm has predicted, Dubai Tradition plans to make use of the know-how to unearth extra websites.
Francis and her staff published a paper on their findings final 12 months, and they’re persevering with to coach the machine studying algorithm to enhance its precision, earlier than increasing its use.
“The thought is to export (the know-how) to different areas, particularly Saudi Arabia, Egypt, perhaps the deserts in Africa as properly,” she says.
