Based on the data of long-term observations for 1950–2020, the ice regime of the Pechora River in modern climatic conditions is considered. A statistically significant increase in air temperature for the cold season, equal to 1.3°C, was revealed. October, March and May make the largest contribution to the overall climate warming. A significant effect of an increase in air temperature on the timing of the formation and destruction of ice formations was revealed. There is a decrease in the duration of freeze-up from 4 to 12 days. The recurrence of ice jams during the spring ice drift is considered. The spatial and temporal heterogeneity of the recurrence of ice jams is noted. According to the data of long-term observations, the frequency of ice jams near the Ermitsa HP reaches 53%, the Ust-Tsilma HP – 7%, while ordinary water levels are formed, the Oksino HP – 21%. The analysis of the highest water levels for the period of ice drift, ice jams and free channel is presented. Depending on the conditions for the formation of the highest level for the period of spring ice drift and geographical factors, a classification of river sections has been developed, consisting of 3 types. The principles of forecasting the highest water level for the period of spring ice drift are proposed, depending on the conditions of its formation. Neural network models have been developed and tested to predict the highest level for the period of spring ice drift in the area with. Oksino. The best quality of the produced forecasts is characterized by the MLP 2-8-1 neural network, the accuracy of the forecasts of which was 92%, the accuracy of the forecasts, data according to the multiple regression equation, was 79%. For MLP 2-8-1, the S / σ∆ ratio on an independent material was 0,46, which indicates the high quality of the forecasts issued.