Count the fruits: the new model will warn about a possible crop failure at no additional cost

Russian scientists have developed a model that allows predicting the risk of crop failure in advance without expensive satellite imagery. The system creates a digital twin of a plant and simulates its development, taking into account weather conditions and soil characteristics. The technology can also be used for forecasts for decades, including to assess how different cultures will feel in a particular region against the background of climate change. This will allow farmers to choose the most suitable plants in advance and, if necessary, change the structure of their crops. According to experts, the technology can be useful both for individual farms and for solving the country's food security problems.

The global agricultural model for Russia

Scientists from the Space Research Institute of the Russian Academy of Sciences, Lomonosov Moscow State University and Chengdu Chinese University of Technology have developed a method capable of predicting crop failures in advance based only on weather data and soil information in large areas. Weather data and soil characteristics are sufficient for this. The researchers took the WOFOST system, which is widely used in the world, as a basis and adapted it to Russian conditions. The model creates a digital twin of a plant and calculates its development depending on the surrounding soil and climatic conditions. The accuracy of forecasts reaches 80%.

— Obtaining digital plant twins, taking into account the characteristics of each region, allows, firstly, to make an early assessment of the condition based on regional soil and climatic norms. This is, in fact, the basis of an early warning system for possible crop failures. Secondly, using climate scenarios for 2050 and 2100, it allows us to assess the potential yield of Russian key crops in the face of climate change - that is, to establish which regions and which crops may become unprofitable and appear to be at risk if left "as it is," said the project manager, head of the Methods sector. remote assessment of the condition and monitoring of used lands of the Space Research Institute of the Russian Academy of Sciences Dmitry Plotnikov.

Photo: IZVESTIA/Eduard Kornienko

Based on soil and climatic data, the system simulates the growth and development of selected crops in a specific area. One of the main problems in adapting the model to Russian conditions was the lack of accuracy of the available soil information. Global soil registries for Russia are not always reliable due to the lack of necessary verification. Therefore, the researchers used the soil map of the RSFSR of the Soviet period, which covers almost the entire territory of the modern country. These data are characterized by high thematic accuracy, but were not originally intended to be uploaded to digital models. The scientists had to manually compare the information with the information used by the system.

— In modern conditions, achieving food independence is one of the strategic goals of our country. Dynamic simulation ideally allows real-time assessment of potential yields at the regional, district, and individual field levels. At the same time, for Russia, which covers various natural zones, a variety of climatic and soil conditions is becoming one of the factors determining the ability of crop varieties to realize their yield potential," explained Yulia Meshalkina, Associate Professor of the Department of General Agriculture and Agroecology at the Faculty of Soil Science at Lomonosov Moscow State University.

Let's go to the tubers: the new device will increase potato yield in agricultural holdings

How will the introduction of technology affect the prices of the popular vegetable

Minimize errors in crop forecasting

To set up the model, the scientists selected the three most common crops in the studied regions — corn, barley and sunflower. For their digital counterparts, the phenological parameters, photosynthesis characteristics, and nutrient distribution features were adjusted.

— Forecasting yields without the mandatory use of remote sensing data is of considerable interest, primarily when assessing large areas. This approach reduces the dependence on the availability of satellite observations of the required quality and time resolution, as well as reduces the amount of spatial data processed. This is especially important in cases where obtaining representative satellite images is difficult due to clouds or insufficient frequency of shooting," said Valeria Gresis, senior lecturer at the Agrobiotechnology Department of the RUDN Agricultural Institute of Technology.

Photo: IZVESTIA/Eduard Kornienko

At the first stage of the study, scientists sought to synchronize the estimated dates of flowering and ripening of crops as accurately as possible with the real annual averages for each area. The model was then adjusted in such a way as to minimize the error in predicting yields.

"Our accuracy is quite high and comparable to the results of high—level foreign studies performed under similar conditions — based on global soil and climate datasets, without using satellite or point meteorological and field information," said Dmitry Plotnikov.

Sowing instinct: irradiation of seeds increased wheat yield by a third

The new technology stimulates plant growth and makes them more resistant to infections.

For sunflower, the average error on the verification data was about 24%, for corn — 26%, for barley — 24%. At the same time, it was possible to achieve almost complete leveling of systematic errors in estimating yields. The developers note that for the first time in Russia, it was possible to configure the model in such detail that it not only predicts crop yields, but also identifies in advance deviations in plant development from the climatic norm — one of the main indicators of crop failure.

Photo: IZVESTIA/Eduard Kornienko

— Such forecasts are in demand both at the level of individual farms, as well as on a regional and federal scale. They allow you to pre—estimate the gross harvest, plan the harvesting campaign, logistics of storage, transportation and processing of the crop, as well as analyze the risks associated with adverse weather conditions," said Valeria Gresis.

In the future, the authors of the project plan to develop machine learning models that will use both the created technology and satellite surveillance data.