LONG-TERM DROUGHT FORECAST IN UZBEKISTAN BASED ON A DYNAMIC-STOCHASTIC ROOT MODEL

The article presents a methodology for long-term forecasting of atmospheric droughts in Uzbekistan with a one-month lead time based on the SPI index. The algorithm is centered on a predictive regression equation constructed using the characteristic roots method. The procedure involves forming an augmented matrix containing columns for both the predictand and predictors, followed by the calculation of a correlation matrix. Predictors include the mean monthly precipitation for each of the three months preceding the forecast, the Southern Oscillation Index (SOI), and Wolf numbers. Decomposition based on the system of eigenvectors of the correlation matrix ensures high temporal stability of spatial links. This approach allows for the extraction of the most energetically significant oscillations of the climate system while effectively filtering out small-scale noise. The physical advantage of this model over standard regression methods lies in the interpretation of the natural modes (characteristic roots) as reflections of the longterm inertia of moisture transport processes and global teleconnections. Model verification was conducted using independent data over an 8-year period. The results confirmed the high efficiency of the methodology: the average sign consistency ρ= 0.914, the correlation coefficient R= 0.789, the root mean square error =0.381, and the share of absolute errors exceeding unity max >1 averaged 12.8%. Despite a natural decline in accuracy during transitional seasons (March–April, October–November) due to the restructuring of circulation processes and increased frontal activity, the obtained scores remain high for long-term forecasting tasks. Based on these quantitative assessments, the model is recommended for implementation into the operational practice of Uzhydromet and other relevant agencies for drought monitoring and early warning.