We first considered each calendar year separately and subsequently grouped together consecutive years that were not statistically different.
http://plosone.org/article/info:doi/10.1371/journal.pone.0000139#article1.body1.sec2.sec11.p1
Imagine a sequence such as 0,0,2,2,4,4,6,6,8,8 where any difference greater than 1.2 is significant. Add a noise sequence such as 1,1,-1,0,1,-1,1,0,-1,0. The resulting sequence would be 1,1,1,2,5,3,7,6,7,8. After grouping the noise free sequence is grouped as [0-1],[2-3],[4-5],[6-7],[8-9] while after noise one obtains [0-3],[4],[5],[6-9]. I am not claiming that this limits the validity of your results or even that it is probabilistically likely that this impacts the analysis, but it could deserve a justification analysis.