Drivers of anomalous wintertime temperature around Japan revealed by neural networks and dynamical mechanisms of the Indian Ocean Dipole influence

Abstract A neural network that predicts the sign of wintertime temperature anomalies around Japan from global sea surface temperature and 500 hPa geopotential height anomalies in the preceding season is first constructed to obtain its optimal inputs. It is revealed that prior to early winter warm events, a negative Western Pacific (WP)-like pattern, a positive Indian Ocean Dipole (IOD)-like pattern, and a weak positive El Niño/Southern Oscillation-like pattern are observed. On the other hand, a negative WP-like pattern along with a positive Arctic Oscillation-like pattern and a positive Pacific-North American-like pattern are seen prior to late winter warm events. Since the influence of the IOD-like pattern on wintertime temperature around Japan, which is statistically identified by the neural network, has not been systematically examined, its physical mechanisms are examined in detail. Dynamical analyses reveal that during the positive (negative) IOD in early winter, Rossby wave sources are generated over the Arabian Peninsula (the Middle East) and East Asia, and Rossby waves are emitted towards the North Pacific. An anomalous anticyclone (cyclone) centered over the northwestern Pacific in the upper troposphere strengthens (weakens) westerly winds over the Sea of Okhotsk, and that in the lower troposphere weakens (strengthens) seasonal northwesterly winds, resulting in anomalously warm (cold) winters around Japan.