When you plot my lake, it's a solidly linear trend of fewer days of ice coverage ever since they began measuring it.
The neat atmospheric CO2 correlation is presumably the same thing that has caused that same correlation to exist in pre-historic times: CO2 has been a correlated trailing indicator of global temperature through the entirety of Earth's history.
If there has been a long-term global trend for Earth temperatures, like my little (not so little actually) lake seems to indicate, some of the assumptions that establish CO2 radiative forcing as a cause rather than an effect go out the window entirely.
But don't take my word for it. Read this Nature paper, and add in the postulate that global temperature since the 1860s is some linear trend similar to my lake's temperature, and watch what comes out:
