One difficulty for ocean modeling is the lack of velocity data for specifying the initial condition. Diagnostic initialization is widely used; it integrates the model from known temperature (Tc) and salinity (Sc) and zero velocity fields while holding (Tc, Sc) unchanged. After a period (around 30 days) of the diagnostic run, the velocity field (Vc) is established, and (Tc, Sc, Vc) fields are then treated as the initial conditions for the prognostic numerical modeling. During the diagnostic initialization period, the heat and salt 'source/sink' terms are generated at each time step. Maximum time rates of absolute change of the monthly mean T, S (0.1¿/day, 0.1 ppt/day) are taken as the standard measures to identify the strength of the thermohaline 'sources/sinks'. Twenty four times of the standard measures (0.1¿/hr, 0.1 ppt/hr) represent strong 'sources/sinks'. Ten times of the strong 'sources/sinks' (1¿/hr, 1 ppt/hr) represent extremely strong 'sources/sinks'. The Princeton Ocean Model implemented for the Japan/East Sea is used to demonstrate the existence of extremely strong thermohaline sources and sinks generated by the diagnostic initialization with the annual mean Tc, Sc from the Navy's Global Digital Environmental Model. The effects of extremely strong and spatially nonuniform initial heating/cooling (salting/freshening) rates on thermohaline and velocity fields need to be further investigated.