Forecast of Precipitation in Ai-Petri Area Based on Artificial Neuron Network Model.

An improved method is proposed for forecasting monthly precipitation in the mountain Crimea based on a model of artificial neural network. The input parameters of the model where taken in the form of a set of climatic indices of the global ocean–atmosphere system over 1948–2020, calculated based on reanalysis data NCEP/NCAR and HadISST1. The model was verified for the control period of 2007–2020. The possibility of seasonal forecast of precipitation with a lead time of up to 6 months was demonstrated. The model was shown to adequately forecast the total precipitation in winter, summer, and autumn (September and October) months. The forecast of precipitation in winter in the mountain Crimea is of particular importance in what regards the filling of local reservoirs, as ~70% of its total sum falls in this period of the year. It is shown that precipitation for winter can be forecasted in October and, with a higher forecast accuracy, in December. The forecast quality was 62 and 56%, respectively. In April, summer precipitation and precipitation in September–October can be forecasted (at a forecast quality of 55%). Parallel to precipitation, the possibility to make seasonal forecasts of atmospheric precipitation was checked and confirmed. [ABSTRACT FROM AUTHOR]