Агроклиматическое районирование фитосанитарного риска вспышек нестадных саранчовых в Казахстане на основе климатических данных и нейросетевого моделирования

Abstract

Background and objective. Given the ongoing shifts in global climate, enhancing predictive models for the dispersal of non-invasive locusts—a major threat to Kazakhstani agriculture—has become a critical priority. By leveraging advanced spatial data analytics and artificial intelligence, researchers can significantly refine phytosanitary forecasting. This research aims to conduct an agro-climatic classification of Kazakhstan and construct a predictive model for identifying high-risk locust breeding grounds, utilizing historical monitoring data, climatic variables, and deep learning architectures.

Materials and methods. The study relies on extensive, long-term phytosanitary monitoring records from across Kazakhstan. Key agro-climatic predictors—including surface and air temperatures, precipitation levels, soil moisture, and NDVI—were derived from ERA5-Land, Open-Meteo, and satellite-based Earth observation data. The analytical framework integrates geoinformation systems with a hybrid CNN–LSTM neural network, enabling the simultaneous evaluation of both spatial and temporal dynamics of the environmental factors involved.

Results. The study successfully mapped Kazakhstan into distinct zones based on the level of phytosanitary risk (low, medium, and high) for locust outbreaks. Findings indicate that the likelihood of infestation is driven by a synergy of agro-climatic conditions, with risk levels rising as one moves from humid, mountainous regions toward semi-arid and arid landscapes. The implementation of the CNN–LSTM model significantly improved the accuracy of spatial risk assessments, allowing for a more precise identification of vulnerable agricultural areas.

Conclusion. This methodology enhances the precision of phytosanitary surveillance and provides a robust scientific foundation for predicting locust migration. These findings offer a valuable framework for building digital decision-support systems, optimizing survey logistics, and streamlining the implementation of agricultural pest control strategies throughout Kazakhstan.