Neural Network Helps Locate Sand and Peat Deposits
Researchers at the Trofimuk Institute of Petroleum Geology and Geophysics of the Siberian Branch of the Russian Academy of Sciences (IPGG SB RAS), working with specialists from oil and gas companies, have applied a neural network algorithm to accelerate the search for sand and peat deposits in the southeastern part of the Yamalo-Nenets Autonomous Okrug.
The system reanalyzes data from previously completed 3D seismic surveys to identify areas where unconsolidated sediments occur in volumes large enough for commercial extraction. The algorithm is resilient to seismic noise and does not require separate retraining for each new area.
Scientists from the Trofimuk Institute of Petroleum Geology and Geophysics SB RAS, together with industry partners including RN-Geologiya Issledovaniya Razrabotka, RN-Proyektirovaniye Dobycha and Kharampurneftegaz, applied the neural network algorithm to locate sand and peat deposits. No new field surveys were required. Instead, the team relied on existing 3D seismic data originally collected during hydrocarbon exploration at greater depths. The upper 30-100 meters of the subsurface, known as the low-velocity zone, is typically treated as seismic interference. Yet this is precisely where unconsolidated materials such as sand, clay and peat suitable for construction are found. The neural network analyzed the dataset, separated useful signals from seismic noise and generated a map highlighting deposits thick enough for commercial development.
Sand and peat are essential for construction projects and environmental protection under Arctic conditions. Sand is widely used to build construction pads and access roads. In the Yamalo-Nenets Autonomous Okrug, the ground is often weak, waterlogged or permanently frozen. Before drilling rigs can be installed, roads constructed or well pads developed, engineers create sand embankments to provide stable foundations. The required volumes are enormous, often reaching hundreds of thousands or even millions of cubic meters. Protective sand berms are also built around industrial facilities to contain liquids in the event of accidental spills. For Arctic energy development and infrastructure construction, nearby sand resources are critical because transporting material over long winter roads or through complex logistics networks is both expensive and operationally risky.
Peat is equally important for land reclamation and environmental restoration because northern regions have only a very thin layer of fertile soil. It is also an effective natural sorbent capable of absorbing crude oil and petroleum products. Northern peat is more porous than peat from southern regions and can absorb 1.5 to 2 times more oil.
Arctic IT
The newly developed neural network identifies large sand lenses with a high degree of confidence, making it possible to plan quarries close to future industrial sites, reduce transportation costs and lessen environmental impacts. Major infrastructure projects in the Yamalo-Nenets Autonomous Okrug may require more than one million cubic meters of sand alone. When materials must be transported, logistics becomes a major challenge because haul distances can reach 300 kilometers. That is not merely a cost issue – transporting such volumes also places significant pressure on the Arctic's fragile ecosystems.
Before the data is fed into the algorithm, it undergoes additional processing using seismic refraction and multichannel analysis of surface waves to extract rock characteristics. The neural network then filters out artifacts and noise that inevitably arise during seismic data acquisition.
The algorithm also eliminates the need for retraining for every new survey area. As a result, operators no longer have to spend months adapting the model to local geological conditions. Just as importantly, the technology reduces the time required to identify and evaluate common construction minerals for quarry development from months to just a few weeks.
Neural Networks Gain Ground
To support the widespread adoption of AI across Russia's geological sector, the country standardized the format of government geological maps in 2023 and began a comprehensive digitalization program. The resulting datasets are now well suited for neural network applications that are increasingly being deployed throughout mineral exploration.
The work covered geological maps of Russia and its continental shelf at a scale of 1:1,000,000. Exploration companies can also draw on several geological databases. The largest is Yediny fond geologicheskoy informatsii o nedrakh (Unified State Geological Information Fund). This marks an important milestone because extracting valuable insights from terabytes of geological information depends on access to large, high-quality digital datasets.
The IPGG SB RAS development stands out for its cost efficiency. Solving an additional task – identifying shallow construction mineral deposits – requires no additional exploration spending, while a single geophysical dataset is put to productive use for a second time. In effect, combining legacy oil exploration data with an intelligent algorithm has created a faster and more cost-effective approach to supporting construction projects across the Arctic.
