A team of scientists from the North Caucasus Federal University has won a grant from the Russian Science Foundation (RSF) to create a yield forecasting system. An intelligent system based on neural networks will analyze data from unmanned aerial vehicles (UAVs), satellites and weather stations, assess the state of vegetation and soil in order to make agricultural forecasts based on them.

The project of young scientists led by Valentina Arustamyan, a junior researcher at the North Caucasus Center for Mathematical Research at NCFU, became the winner of the Russian Science Foundation competition, which was held for small scientific groups in 2025. The amount of financing amounted to 1.5 million rubles. The development integrates the approaches of agronomy, computer vision, climatology and machine learning and is focused on solving the applied problem of digitalization of agricultural production management based on domestic technologies.

– The key principle of science at our university is the transformation of ideas and research projects into practical ones. We are prioritizing the development of technologies that should bring real benefits to the economy and be focused on technological leadership. In this regard, our scientists are developing a number of projects in the field of AI and big data analysis in the interests of medicine, the chemical industry, and agriculture," commented Acting Rector of NCFU Professor Tatiana Shebzukhova.

According to the authors of the study, modern approaches to monitoring agricultural land in the context of climate change and unstable weather factors require a transition from periodic field observations to integrated digital vegetation analysis systems. The use of remote sensing data obtained from drones and satellites makes it possible to form a spatiotemporal picture of the state of crops.

– We plan to train a neural network to process images provided from heterogeneous information sources: to remove noise and omissions, i.e. to perform spectral normalization of data and coordinate spatial solutions. Algorithms for analyzing these images and a module for processing temperature and sedimentary series will also be developed. The key result will be the creation of a predictive model capable of monitoring agricultural fields and making yield forecasts," said Valentina Arustamyan, a graduate student and a junior researcher at the North Caucasus Center for Mathematical Research at NCFU.

The created neural network architecture will allow farmers to quickly respond to changes in plant conditions, plan fertilizing, watering and other measures, reducing crop losses and increasing its sustainability. The scientific novelty of the project is to combine diverse sources of information and implement transformational architectures to solve forecasting problems in the agricultural sector.

The authors of the project are confident that the created neural network model will be in demand both in scientific research on agroecology and agriculture, as well as in application platforms for agricultural holdings, crop management centers and regional phytosanitary monitoring services. In addition, the developed architectures and methods can be adapted to other tasks, such as crop yield assessment, crop degradation detection, irrigation control, and climate change impact modeling.

For reference: in December 2025, NCFU received 12 grants from the Russian Science Foundation (RNF), which allow for the implementation of research by small individual scientific groups. The total amount of financing for these projects is 36 million rubles.