It seems counterintuitive that as the volume of tweets increases, the variance increases. What could be the explanation for this?
Critically, the baseline is fairly static, so the extremes in this graph can only be reached by fairly major events that are capable of drowning out the typical everyday tweets.