Users has been questioning why the land mask does not appear to work in MUR Level SST dataset. For example, over the Hawaiian Islands, only the Big Island is masked out (fill values are set). The other islands appear "underwater"... (Kauai, Oahu, Molokai, Lanai, Maui).
Here is the MUR data provider’s response (by Mike Chin):
“The "fillvalues" in the MUR SST field is NOT the land mask. Users need to use a separate and/or provided landmask (mask variable in file, mask = 2 -> land). The landmask was generated independently from the SST field using the GMT software and OSI SAF ice concentration data. The "fillvalue" pixels in MUR SST field should only be used to identify some globally prominent geographical landmarks, such as locations of the continents and large enough islands.”
“The "fillvalue" pixels in MUR are fairly arbitrarily placed, since MUR can interpolate SST everywhere in the world, including under Mt. Everest. Ideally, the front lines of the fillvalues would follow world's coastlines, but coastlines can move (more 1km) daily, especially due to tides.”
“Such is not necessary a big problem around Hawaii, but it would be for some other users looking at shallow estuaries etc. So, "we" decided to apply fillvalues less aggressively, for those users who might need some SST references during high tide, for example. The quality flags in L2P inputs should be the primary determinant for where the fillvalues in MUR should go, but the L2 producers also need to deal with things like tides. Peter Minnett, for example, warned me that the SST returns over shallow tide water are not so accurate to begin with. MUR conveys these returns to the users as long as L2P labels them "high quality" and even though the pixels may be dry-land half of the day each day.”
The short answer is that users should apply the landmask (mask variable = 2) by themselves to view the MUR SST with the islands masked out properly.