The way sea-surface temperatures are archived seems to vary widely from dataset to dataset. In REGRID parlance, we make a distinction between sea-surface temperature and skin temperature. A sea-surface temperature field represents the temperature of the water (whether that's the temperature at the water surface or temperature of some layer at the top of the water). Over land, this field doesn't really have any meaning, and in many datasets the values are smoothly interpolated from water points. Example:

.

A skin temperature field represents the temperature of the surface of the earth, whether that surface is land or water. Thus, in summer for example, you may wind up with large temperature gradients between warm land points and cold water points. Example:

The skin temperature can be tricky to use within the MM5 modeling system, because of the horizontal interpolation to the MM5 grid. Horizontal interpolation to a point near the coast can wind up being an interpolation from a land point and a water point. If, for example, the summertime ground temperature is considerably warmer than the water temperature, you may wind up with unreasonably cold ground temperatures and unreasonably warm water temperatures at the coast. This will be particularly noticeable when interpolating to an MM5 grid that has a much higher horizontal resolution that the source dataset.

At the REGRID stage, fields SST and SKINTEMP are treated identically. They are simply interpolated to the MM5 horizontal grid, with no concern for what may happen at the coast.

Later in the MM5 modeling system, in program INTERPF, fields called SST and SKINTEMP are interpreted differently. If an SST field is found, INTERPF will interpret that field as one that can be used as an instantaneous field appropriate for water temperatures, which may vary in time. INTERPF will create a lower-boundary-conditions file (LOWBDY) with an SST field for each analysis time. This field will be called TSEASFC. If there is no SST field, but a SKINTEMP field is found, INTERPF will average SKINTEMP for all times, and write that average as a single, time-invariant TSEASFC field to the LOWBDY file. This averaging is done to remove the diurnal variation of the skin temperature over land in the averaged, time-invariant TSEASFC field.

In short, fields called SST will be interpreted as instantaneous water-temperature fields and will allow time-varying sea-surface temperatures