~ Is likely some remote sensing calculations/methods need some adjustments.
Taking advantage of persistent "Big Blue" of 2016 spring, a truly remarkable insolation bombardment , the right term "relentless" onslaught of sunshine, we can check and find if refraction gained insights (written here) are true on a planetary scale. NOAA daily climate composites are very good, so we look at its sea surface temperature setting or "surface skin" temperatures VS surface temperatures.
NOAA May 4 2016. Daily mean
Click on GIF image to expand and use your mouse pointer to make comparisons.
If there is a calculation error with NOAA surface temperatures, it would be "averaged out" eventually because Ta>=Ti , a fact gained by multiple horizon observations, this feature will show up over a longer term:
NOAA May 1-5 2016. Composite mean makes it much harder to find
Hypothetically, the longer we average out the likely Algorithm error, the more impossible it would be to find Ta < Ti.
March 1 to May 5 2016, literally impossible to find top of sea ice temperature
NOAA surface temperature looks better but still has small daily flaws.
May 9, 2016 NOAA daily composites offer surface air temperature feature which performs slightly better than 1000 mb, if you click on extreme cold surface air temperature near land it will be likely erroneous, making surface air colder than ice. Which has never been observed optically.
Likewise looking back longer term:
April 9-May 9 average If you find a spot where Ta< Ti , let me know. There are none I can find.
NOAA remote sensing temperatures are quite good, but I would look at every case when
sea ice is warmer than surface air, double check the calculations and the physics. I don't know if this is the error which causes sea ice models to err in making good melt projections. A 4 C warmer sea ice than surface temperature (Chukchi anomaly very top GIF above) would make the horizon extremely low and that has never been observed, on top of the underlying thermal physics which would be hard to explain. WD May 7 and May 11, 2016