The following graphs show the ISS mean radius vector (or semi-major axis) minus 6371 km (just to show an approximate altitude) and the mean air density at the ISS position calculated with the NRLMSISE-00 atmosphere model updated with the solar and geomagnetic indices: ftp://ftp.agi.com/pub/DynamicEarthData/SpaceWeather-All-v1.2.txt.
EDIT: the blue plot represents the ISS mean radius vector minus 6371 km, which is equivalent to the sami-major axis (I use "mean radius vector" to avoid confusion with the osculating semi-major axis).
The mean radius vector is obtained from the TLEs downloaded from https://www.space-track.org/ and all the calculations are done with the CSpOC's SGP4 library downloaded from the same site.
The mean radius vector is calculated by numerical integration of the inverse of the radius vector as a function of the time for 1 orbit that starts from T/2 before the TLE epoch and ends to T/2 after the TLE epoch (T is the ISS orbital period). This is the same as integrating the radius vector against the eccentric anomaly and the result represents the average distance between ISS and Earth's barycenter (some additional info can be found in this page of my site: http://cristianopi.altervista.org/as/calc_sat.html).
The air density is about constant: probably only attitude variation or, generally speaking, configuration variation:

Big peak in the air density:

Big peak in the air density:
