different salt and temp result between turn on and turn off T_PASSIVE

[zhiguoMEI]

Hi everyone! I just find a problem if turn on the T_PASSIVE CPP option.
I have a Western Pacific model based on ROMS (version 3.9) with a resolution of ~5.5km. I chose 4-th order centered advection for tracers and moment.
The model had run for 5 years successfully. I want to add passive tracers at the beginning of the 6th year.
I set the boundary and initial value of dye_01 as 0 and turn on T_PASSIVE. I chose HSIMT advection for the passive tracer.
After one month simulated I got undulatory surface salt and temp(below are the surface salinity), which disappeared with the T_PASSIVE option turned off.

[zhiguoMEI]

After one month simulated the results are obviously different between Passive tracer and without passive tracer.
Surface salinity appears to ripple near the boundary and in the middle of the model region when the T_PASSIVE switch is turned on and disappeared after the switch is turned off.
Surface salinity along latitudinal lines also appears wavy especially near boundary areas.
Temp is similar to salinity so I only upload salt pictures.

[zhiguoMEI]

Here is the log file with passive tracer turned on(log-1999_dye.log) and turned off(log-1999.log)
The results left me wondering if the passive tracer switch would affect the temperature and salinity calculations, and aren't they calculated separately?