Prediction of bovine brucellosis incidence in the Kabardino-Balkarian Republic using time series decomposition and spatial analysis methods

Abstract

Introduction. Bovine brucellosis remains a significant zoonotic threat in the North Caucasus. Climate change and spatial heterogeneity require integrated predictive models.
Aim. To develop a practice-oriented predictive model incorporating long-term management trends, short-term climatic fluctuations, and spatial risk differentiation for the Kabardino-Balkarian Republic.
Materials and methods. Time series analysis of incidence (2000–2024) was applied using decSARIMA decomposition, correlation and regression analysis, kriging interpolation, cross‑validation.
Results. The model achieved a determination coefficient R² = 0.84. The mean forecast error on an independent test sample was 0.16 per 10,000 examined animals. Cross–validation confirmed model stability with a mean absolute error (MAE) ranging from 0.18 to 0.21. Spatial heterogeneity in risk was identified: the foothill zone (600–1200 m above sea level) exhibited a 2.3-fold higher risk compared to the plain. Seasonal peaks were observed in May–June (+35%) and October–November (+28%). The forecast for 2025 is 11–12 cases per 10,000 examined animals (95% confidence interval).
Conclusion. The developed model combines robust annual forecasting with spatial risk differentiation and analysis of the seasonal mechanisms of weather influence. Economic assessment shows that the costs of preventive measures in areas with predicted risk are 4–6 times lower than the losses from disease outbreaks. This model can serve as an additional decision-support tool for the veterinary service of the Kabardino-Balkarian Republic.

To cite: Khitieva A.Zh., Tebuev A.Kh. Prediction of bovine brucellosis incidence in the Kabardino-Balkarian Republic using time series decomposition and spatial analysis methods. Agrarian Bulletin of the North Caucasus. 2026;16(2):22-32. (In Russ.) https://doi.org/10.31279/2949-4796-2026-16-2-22-32