The paper considers the development of a model for precipitation field nowcasting using the data obtained from the Himawari-8 satellite and a GFS numerical forecast model. The nowcasting method employs a convolutional and recurrent neural network architecture. A peculiarity of the developed model is a possibility to make a forecast using no ground-based meteorological radars data. The authors present preliminary research results as exemplified by the precipitation field nowcasting for a 30-minute period and the 60-minute forecast of the cloud cover optical depth distribution. Finally, the paper outlines the areas for further research with the account to the identified drawbacks of the existing forecasting algorithm software implementation.