relative humidity: wrt ice or water?

[JoeU]

Relative humidity values that are output (and plotted) by HYSPLIT are very different, depending on the met data set chosen. E.g. for GDAS 1km the humidity appears to be with respect to ice at lower temperatures. However, for GDAS 0.5km (GFSG) this seems to change to water, although the name of the variable remains the same. Digging through info I found the statement:
https://ready.arl.noaa.gov/gdas1.php
"NOTE ON GDAS RELATIVE HUMIDITY: NCEP outputs relative humidity with respect to water for temperatures greater than or equal to zero degrees Celcius (0C), with respect to ice for temperatures less than -20C, and a blend between -20C and 0C. ARL does not modify NCEP's values."
This is very confusing, as both sets are described as GDAS, yet the actual variable is different. Can anyone confirm my suspicions? If so, then at the very least some documentation on GDAS 0.5 should be provided, rather than the generic one which is misleading.

[barbara.stunder]

The GDAS 1 degree (not km) has relative humidity, while the 0.5 degree has specific humidity. If you are outputting RH along a trajectory, RH is calculated from specific humidity if needed. You are correct that for that case, it is RH with respect to water. We will review and edit our documentation to make that clear.

[barbara.stunder]

NCEP outputs relative humidity with respect to water for temperatures greater than or equal to zero degrees Celcius (0C), with respect to ice for temperatures less than -20C, and a blend between -20C and 0C. ARL does not modify NCEP's values.