Voter rolls inflate before every election, then names disappear after the election

DataJeff, Analytics Lead for US Election Integrity Project, says:

We saw this pattern in the Colorado Data. Every county I have looked at so far has the same build-up towards an election and then a culling immediately after an election. We found this pattern last year. Blue line is the voter roll growth (From the SoS website) since 2016 and red line is the voting population growth (from the CO demography office) since 2016

f61d4ebc-a714-11ec-b007-d6d3907c63d6

The y axis is the percentage growth since 2016 for total voters (active, inactive, and pre-registered) from the SoS voter rolls and the percentage growth in voting population (those who were 17 at the time of the election for each year).

Voter rolls accessed from the Colorado Secretary of State’s website: example: https://www.sos.state.co.us/pubs/elections/VoterRegNumbers/",years[i],"/January/statistics.xls where years[i] was the year that the voter roll corresponded to (e.g., 2017).

County level population retrieved from https://demography.dola.colorado.gov/population/data/county-data-lookup/

I guess it make sense that more people would register to vote just prior to an election.

The excess # of voters compared to the population growth is certainly troubling.

But why are there so many registered voters de-registering after each election? I cannot think of a legitimate explanation for this.

Can anyone explain?