The manuscript present analysis Arctic surface skin temperature and surface type, obtained from airborne thermal measurements during HALO-(AC)3. In general, the manuscript provide updated assessment of sea ice surface in the marginal ice zone (MIZ) using high-resolution airborne thermal images, which will promote research on sea ice dynamic and thermodynamic process in the MIZ. 

Although, i have a few minor question for the author

1. It seems you used Level 3 MODIS daily IST for radiative transfer simulations and evaluation of airborne IST algorithm, and you mentioned the temperal mismatch between airborne and satellite data. Assuming there were 1 or 2 HALO flight in one day during the compaign, have you tried Level2 MODIS swath data, i.e. MOD29/MYD29 instead of Level 3 gridded product? the swath data suffer more from cloud contamination of course, but it  will help reduce the time different to less than an hour, and possibly, you would find better agreement between airborne and satellite data.

2. if i get it right, the class Sea Ice-Water Mixture (IWM) is not included in the training set, but only classified Open-water pixels with surface temperature below -2.5℃. it seems the IWM only account for a minor portion of area (Fig.7), and its surface temperature is very close to thin ice(TI). the question is, how is the threshold determined? Also,  since you used Sentinel-2 MSI for labeling of the images in training set, It is recommended to check the distribution of corresponding reflectance (from MSI) for different surface class in the training set, IWM samples might appear in the OW samples as anomaly.

3.about the size distribution of surface features in the MIZ, it seems the thin ice area is connected into a very large segment in Fig.7. i'm not sure if is it a common phenomenon in your data set, but it seems the TI is the dominant class in Fig.11.  It is recommended to break down the large TI segment into pieces by any means, otherwise the FSD/SSD result for TI could be misleading.