Deep Learning Models Ratify ENSO's Substantial Impact on Antarctic Sea Ice Subseasonal Predictability: supplemental data
cor_spatial38c.hdf contains SIPNet skill information, comprising four subsets: 'mid', 'nino', 'nina', and 'all', representing model skill under neutral conditions, El Niño, La Niña, and the entire time range, respectively.
acc_per_spatial38.hdf is similar to cor_spatial38c but represents the model skill for anomaly persistence.
cor_spatial38_linear.hdf, like cor_spatial38c, represents the model skill but specifically for the linear SIPNet model.
sic_stddev.hdf contains sea ice variability information with three subsets: 'nino', 'mid', and 'nina', denoting sea ice variability under El Niño, neutral conditions, and La Niña, respectively.
SIPN_ENSO_ice_model1979_2022trian1980_2022test8_8_2236.h5: Antarctic weekly SIC prediction from SIPNet model. The data has two sub-datasets: 'predicted_test' and 'y_test'. 'predicted_test' refers to the predicted sea ice value, while 'y_test' is the observed value corresponding to 'predicted_test'.
SIPN_ENSO_ice_model1979_2022trian1980_2022test8_8linear_2236.h5: Similar to SIPN_ENSO_ice_model1979_2022trian1980_2022test8_8_2236.h5 but from the linear SIPNet model.
If there is anything unclear, please contact Wang Yunhe at wangyunhe@qdio.ac.cn.