This study is devoted to the problem of predicting sand and dust storms in the Aral Sea region. The Aral Sea desert, which emerged as a result of the drying up of the Aral Sea, is increasing the frequency and intensity of sand and dust storms in the region, which has a serious negative impact on ecological stability and public health. The SARIMA statistical model and the XGBoost machine learning algorithm were used in the study for the forecast based on observation data recorded at the Muynak, Kun'gorat and Nukus meteorological stations located in the Aral Sea region over the past 10 years. According to the results obtained, the hybrid SARIMA–XGBoost model demonstrated high prediction efficiency. The accuracy of the model was 0.88, the F1-index was 0.85, and the ROC-AUC value was 0.92. Forecast errors were reduced by an average of 25–30% compared to the SARIMA model, achieving MAE values of 0.27 and RMSE values of 0.39. According to the feature importance analysis, maximum wind speed was the most important predictor, with a share of 34.7%. Wind gust (21.3%) and minimum relative humidity (17.6%) also had a significant impact.

Aral Sea region, sand and dust storms, XGBoost, machine learning

Micklin P. Aral Sea disaster // Annual Review of Earth and Planetary Sciences. - 2007. - Vol. 35. – P. 47–72.

Indoitu R., Orlovsky L., Orlovsky N. Dust storms in Central Asia: spatial and temporal variations // Atmospheric Environment. - 2015. - Vol. 118. – P. 183–194.

Goudie, AS (2014). Desert dust and human health disorders . Environment International, 63, 101–113. https://doi.org/10.1016/j.envint.2013.10.011.

Shao, Y., Wyrwoll, K.-H., Chappell, A., et al. (2011). Dust cycle: An emerging core theme in Earth system science . Aeolian Research, 2(4), 181–204. https://doi.org/10.1016/j.aeolia.2011.02.001.

Terradellas, E., Cuevas, E., Baldasano, JM, et al. (2015). The WMO Sand and Dust Storm Warning Advisory and Assessment System (SDS-WAS) . Bulletin of the American Meteorological Society, 96(9), 1467–1487. https://doi.org/10.1175/BAMS-D-14-00181.1.

Indoitu, R., Orlovsky, L., & Orlovsky, N. (2012). Dust storms in Central Asia: Spatial and temporal variations . Journal of Arid Environments, 85, 62–70. https://doi.org/10.1016/j.jaridenv.2012.04.002

Zender, CS, Bian, H., & Newman, D. (2003). Mineral dust entrainment and deposition (DEAD) model . Journal of Geophysical Research: Atmospheres, 108(D14), 4416. https://doi.org/10.1029/2002JD002775.

Darmenova, K., Sokolik, IN, Shao, Y., et al. (2009). Development of a physically based dust emission module . Journal of Geophysical Research, 114, D14201. https://doi.org/10.1029/2008JD010999.

Reichstein, M., Camps-Walls, G., Stevens, B., et al. (2019). Deep learning and process understanding for data-driven Earth system science . Nature, 566, 195–204. https://doi.org/10.1038/s41586-019-0912-1.

Opp, C., Groll, M., Aslanov, I., Lotz, T., & Vereshagina, N. (2017). Aeolian dust deposition in the Aral Sea region. Journal of Arid Environments, 136, 90–101.