Artificial Neural Network for the Short-Term Prediction of Arctic Sea Ice Concentration

Abstract

In this paper, we applied an artificial neural network (ANN) to the short-term prediction of the Arctic sea ice concentration (SIC). The prediction was performed using encoding and decoding processes, in which a gated recurrent unit encodes sequential sea ice data, and a feed-forward neural network model decodes the encoded input data. Because of the large volume of Arctic sea ice data, the ANN predicts the future SIC of each cell individually. The limitation of these singular predictions is that they do not use information from other cells. This results in low accuracy, particularly when there are drastic changes during melting and freezing seasons. To address this issue, we present a new data scheme including global and local SIC information, where the global information is represented by sea ice statistics. We trained ANNs using different data schemes and network architectures, and then compared their performances quantitatively and visually. The results show that, compared with a data scheme that uses only local sea ice information, the newly proposed scheme leads to a significant improvement in prediction accuracy.

The Influence of Internet Usage Frequency on Women’s Fertility Intentions—The Mediating Effects of Gender Role Attitudes

“…Thirdly, at present, many scholars use artificial intelligence tools (ANN, LSTM, etc.) to conduct influence factor analyses and predictive analyses [ 32 , 33 , 34 ]. It is also of great significance to analyze the influencing factors of women’s fertility intentions and predict the changing trends surrounding the female fertility rate, which is also the research field we will pay attention to in the future.…”

Two-Stream Convolutional Long- and Short-Term Memory Model Using Perceptual Loss for Sequence-to-Sequence Arctic Sea Ice Prediction

“…Second, we introduce a new feature-based loss function to produce better qualitative prediction maps by comparing their perceptual characteristics. As addressed by us and others in previous studies [13][14][15][16], ice, atmospheric, and oceanic variables may be helpful in improving predictability. Finally, we test the significances of several input variables through the direct training of the model without using relative sensitivity tests.…”

“…Conventional numerical, statistical, or ensemble models for sea ice prediction often use various ice-, ocean-, and atmosphere-related properties as input variables [9,11,12]. Based on previous DL-based sea ice prediction studies [13][14][15][16], we also used several sea ice and atmosphere-related variables. The details of the variables are provided in Table 1.…”

“…The proposed model determines the nonlinear relationship between sea ice and climate variables. Choi et al [15] used a gated recurrent unit, a type of recursive neural network, to provide 15-day predictions. They used the spatial information of sea ice extent and SIC as input variables and compared the proposed model to that of the LSTM.…”

“…They used the spatial information of sea ice extent and SIC as input variables and compared the proposed model to that of the LSTM. These studies [13][14][15] were based on one-dimensional vector data; however, Choi et al incorporated the image coordinates into the input vector data [12]. Kim et al [16] proposed the convolutional neural network (CNN)-based models by incorporating various sea ice, atmospheric, and oceanic variables to provide one-month predictions.…”