Not sure if this is the answer or not, and I don't feel like looking at the climate data to see if I'm right or not, but they could be doing it based on how statistically significant each average is... For instance if the -9.1 in Lansing is 2 standard deviations below normal for that location and the +5.5 is 3 standard deviations above normal for its location then it might make more sense... but that's literally a random guess