On the changing nature of the regional connection between the North Atlantic Oscillation and sea surface temperature