Calculation of northern hemisphere sea ice area using recurrent neural networks

• Published in: IOP Conference Series: Earth and Environmental Science
• Main Authors: Viatkin, D, Zhuro, D, Zakharov, M, Malysheva, S
• Format: Article in Journal/Newspaper
• Language: unknown
• Published: IOP Publishing 2021
• Subjects: Arctic; Sea ice
• Online Access: http://dx.doi.org/10.1088/1755-1315/937/4/042094; https://iopscience.iop.org/article/10.1088/1755-1315/937/4/042094; https://iopscience.iop.org/article/10.1088/1755-1315/937/4/042094/pdf

Abstract Ice covering water surfaces causes difficulties for ship traffic in the northern regions. Developing a sustainable logistics system that describes and manages ship traffic requires consideration of many factors, one of which is the area of sea ice covering the waterways. Most of the volume of sea ice in the northern hemisphere is concentrated in the Arctic zone. The paper describes the process and results of data preparation and development of a recurrent neural network to determine the value of ice area change in the next 50 days relative to the last day of sea ice area measurement. The prediction is made based on the previous 30 measured values of sea ice area and a user-specified value of the day for which the prediction will be made. The work uses NSIDC open dataset on sea ice area for the northern hemisphere. This model allows us to calculate the change of sea ice area for 1 day ahead with an accuracy of 0.581%. For the 50-day prediction of ice area, the accuracy is 4.017%.