That's just it: if the changes are only visible if you chop off the bottom of the graph, the probability they aren't significant increases.
(My point is that temperature is probably an odd man out in graph terms, our usual reference points are not actually at zero.)