Left: MAD vs. median magnitude in WISE1 bands. For convenience, the diagram only shows the RSG sample and AGB populations. The median value of single-epoch MADs (each epoch is approximately five to ten days) of each target is shown as the error. The optical and MIR spectroscopic RSGs are shown as open diamonds and open squares, respectively. The MIR variability of RSG sample increases along with luminosity. In total, ∼21% of targets have MAD_WISE1 >  0.01 mag (dotted line) and ∼2% have MAD >  0.1 mag (dashed line). The inset shows only the RSG sample with the limit of MAD_WISE1 = 0.01 mag in the range of 13.0 <  Median_WISE1 <  7.0 mag and 0 <  MAD_WISE1 <  0.025 mag. The RSG sample is separated into two subsamples. One subsample represents high-variability and red targets with Median_MAD >  0.1 mag and J − K_S >  0.89 mag, which may be more at risk of being contaminated by AGBs (the increasing symbol sizes indicating the increasing variabilities). The other subsample (gray) represents low variability (MAD ≤ 0.1 mag) or blue (MAD >  0.1 mag and J − K_S ≤ 0.89 mag) targets. One M5e AGBs is identified (arrows) and excluded from our RSG sample, based on its significant variability with moderate magnitude. The normalized histogram (NC; upper left) indicates the range of 0 <  MAD_WISE1 <  0.05 mag, where the distributions of all (gray), brighter (Median_WISE1 <  9.0 mag; red), and fainter targets (Median_WISE1 ≥ 9.0 mag; blue) are shown (for clarity, slightly different bin sizes are used for different groups). The dash-dotted lines indicate the median values of corresponding groups. See text for details.