On the causes of trends in the seasonal amplitude of atmospheric CO₂