highly significantp < 0.001
The agreement between the classification based on both campaigns and each single campaign dataset was highly significant (99.3%, k = 0.991, p < 0.001 for the test, when the Baffin Bay was the training set and the Fram Strait dataset was predicted and 92.9% of agreement, k = 0.905, p < 0.001 when reversed, i.e., Fram Strait as training dataset and the Baffin Bay as predicted).