Heat Maps Help Geologists Find Uranium, Gold and Tin
Researchers at the Russian Academy of Sciences’ Institute of Ore Deposits, Petrography, Mineralogy and Geochemistry (IGEM RAS) and Transbaikal State University (ZabGU) have proposed a method that could significantly improve the accuracy of mineral exploration. They digitized data on 269 deposits and 1,198 mineral occurrences and used them to create “heat maps” showing the density of known mineral deposits.

The method is based on kernel density estimation (KDE). While a conventional geological map marks individual deposits, a heat map highlights areas with a high concentration of known mineral occurrences. Brighter areas indicate where more uranium, gold, tin, molybdenum, tungsten and other minerals have been found; paler areas indicate fewer occurrences. The study was conducted as part of the Far Eastern track of Russia’s Priority 2030 program.
Two Maps, One Goal
The researchers created two types of maps. The first, an extensity map, includes all known occurrences, even small ones with no commercial significance. The second, an intensity map, includes only large deposits. Comparing the two helps distinguish areas with numerous small occurrences from those containing substantial deposits.
The researchers overlaid maps for different metals to identify areas where their distributions coincide. Uranium and molybdenum, for example, frequently occur together, with a spatial overlap coefficient of 0.86 on a scale of 0 to 1. The researchers believe this pattern reflects the geology of the Streltsovskoye uranium ore field in Transbaikalia, where the two metals are associated with shared geological processes. Gold and arsenic, by contrast, show a strong association only among smaller occurrences, not major deposits. This gives geologists a useful lead: if small-scale indications are present but no large deposit has been identified nearby, a larger deposit may lie deeper underground or slightly farther away. Comparing the datasets can therefore reveal patterns in the spatial distribution of mineral resources and help identify areas for further exploration.

Technology Reserves
Russian researchers and companies are developing digital tools designed to make mineral exploration faster and less expensive.
For example, researchers at Novosibirsk State University have developed a service that combines neural networks with conventional seismic data analysis in a single interface. The software helps geologists map faults and geological structures more accurately. These structures can control the movement of mineralized fluids, offering clues about where to look for lithium in subsurface brines. The developers say the algorithms have already been validated using real-world data, with the results documented in more than 10 scientific publications.
Polyus has also developed AI-based geological algorithms, including tools to predict ore lithology from drilling X-ray fluorescence (XRF) data and to forecast ore fields, helping identify potential deposits. An XRF analyzer at the drill site scans the core and produces a table within seconds showing the concentrations of different elements in the rock. A neural network uses those readings to identify the rock type. The software then combines data from all drilled boreholes, identifies patterns and predicts what lies between them.
Before developing the heat maps, IGEM RAS researchers published a study on using the AlexNet convolutional neural network to predict polymetallic deposits in southeastern Transbaikalia. A later paper combined Earth observation data, GIS analysis and neural network modeling to identify prospective areas in the Priargun zone.

A Three-Year Project for the Future
Researchers at IGEM RAS and ZabGU have developed a digital tool that uses previously collected data to identify promising areas for mineral exploration, potentially reducing the number of low-yield exploration campaigns.
The project’s main long-term goal is to move beyond individual digital maps and build an integrated geospatial platform capable of processing large volumes of diverse geological data. A joint laboratory run by IGEM RAS and ZabGU is already developing the system’s architecture. This year, the team began a three-year project to build the Digital Geological Navigator of Transbaikalia, a platform that will bring together archival maps, satellite imagery and AI algorithms. By 2028, the project aims to develop a unified predictive exploration model for the entire Russian Far East and reduce mining companies’ costs by at least 25%. Over time, neural network algorithms could also be trained to account for more complex relationships among geological indicators, including fault locations, geochemical anomalies, satellite data and the age of rock formations.









































