I'm curious how this is possible. The data in question, based on the ADF, we strongly fail to reject the null of a unit root. Yet in natural logs, the series is stationary apparently (ie, reject the null as p-value is 0.00).
But even in natural logs, the ACF shows high persistence, and it may as well be an MA(inf) based on the ACF. Also the PACF has a spike at lag=1, but nothing else anywhere. Hence the ACF/PACF seem to indicate a unit root even in natural logs, but the ADF doesn't.