AI to Detect Hidden Soil Degradation
Scientists in Volgograd have developed an integrated system to keep deserts from encroaching on fertile farmland. Its mathematical model will make it possible to detect the problem at an early stage.
The main driver of land desertification is the slow, initially invisible degradation of soil. The loss of the black-soil layer disrupts the water cycle, affecting the climate, accelerating wind erosion and reducing biodiversity. Ultimately, it can sharply lower crop yields. Russian scientists are now developing integrated digital platforms that can detect dangerous changes in soil in time for measures to be taken to protect farmland.
Combining Technologies for More Accurate Results
A system developed by researchers at the Federal Research Center for Agroecology, Integrated Land Reclamation and Protective Afforestation of the Russian Academy of Sciences (RAS Federal Research Center for Agroecology) can detect soil degradation at its earliest stages, when there are virtually no visible signs of the problem. The approach combines three levels of diagnostics: satellite-based observation, field surveys and laboratory analysis.
“We aim to detect the onset of degradation processes during their incubation period, at latent stages when the system still appears stable, including in satellite imagery. We have developed a three-tier model that combines a space-based level, using the remote-sensing data we are all familiar with and the Opustynivanie i yego monitoring (Desertification and Its Monitoring) information and analytics system, with two levels of ground-based research: fieldwork and laboratory analysis,” said Anna Fedotova, the center’s deputy director for research, Doctor of Biological Sciences and professor.
A Mathematical Model Brings the Data Together
These technologies cannot work effectively in isolation. That is why the earliest signs of soil problems previously went undetected. Satellite data, for example, can show the condition of the land surface, but cannot reveal what is happening to microorganisms within the soil or how humus levels are changing. Conventional soil analysis can provide information about humus content, but it takes time: researchers must examine individual plots in the field and collect samples. The samples then undergo laboratory testing to assess the soil’s physical, chemical and biological properties. Even a decline of half a percentage point in humus content, however, can signal a serious problem: the depletion of a plot of land. At that stage, the damage may not yet be visible on the surface.
Artificial intelligence speeds up the analysis. “We have added a mathematical component to the model, with hierarchical clustering at its core. The data are brought together in a single matrix, standardized and then differentiated according to a resulting vector of properties. In other words, the algorithm automatically divides the area under study into categories: at-risk areas, areas in normal condition, and areas we need to identify to prevent degradation,” Anna Fedotova explained.
Russian Technologies to Protect Farmland
The new platform is intended for the early detection of land degradation in southern arid regions. It has already been tested in the Volgograd Region. After identifying and assessing problem areas, the researchers recommended a set of measures to help the farm owner restore the land. These included applying specific fertilizers, changing crop rotations and temporarily suspending fieldwork. The methodology will be further developed at a dedicated research site in the Volgograd Region.
Russia has been developing digital systems to prevent soil desertification since 2022. That year, Terra Tech JSC, part of the Russian Space Systems holding under the Roscosmos State Corporation, began work on the Opustynivanie i yego monitoring (Desertification and Its Monitoring) information platform, commissioned by the RAS Federal Research Center for Agroecology. The platform is designed to support monitoring, geospatial analysis and satellite imagery analysis. It went live in 2026.
Sberbank has developed and is deploying Verda (meaning “green”), an AI-powered land restoration technology, in the Republic of Kalmykia. Its neural network analyzes satellite images, radar data and field survey results for an area divided into 10-by-10-meter plots. It creates a digital profile for each plot and then recommends a soil restoration program.
The platform developed in Volgograd could help create a unified digital system for assessing the condition of Russia’s fertile land. The experience gained in southern regions could also prove particularly valuable in countries facing a high risk of land degradation, including Kazakhstan, Uzbekistan, Turkmenistan and Mongolia.
