A weld pool does not have a temperature the way a room does. It has a temperature field:
hottest where the arc or beam delivers energy directly, and progressively cooler moving
outward, because heat is flowing away from that point into the surrounding solid metal and
into the pool’s own advancing edges the whole time the pool exists.
The two ends of that field answer to different physics. Near the centre, where the heat
source acts, the metal sits above its liquidus temperature — superheated liquid, not merely
molten. At the trailing edge, behind the heat source’s direction of travel, the temperature
has fallen back toward the weld pool’s freezing range: metal there
sits between liquidus and solidus, part liquid and part already solidifying, which is exactly
where the finished bead’s microstructure starts to be set. A single instrument reading taken
without saying which of these two regions it looked at, or somewhere between them, is not
comparable to another reading taken at a different point in the same pool, however precise
either measurement is on its own.
That is also why “weld pool temperature” as a project specification is rarely written as one
number the way interpass temperature is: an interpass
value is a point reading, taken at a defined location on the weld metal and adjacent parent
material once the pool has solidified, immediately before the next run starts. A weld pool
has no such moment to be read at — there is no instant in its existence when it is not a
field spanning superheated liquid to a solidifying edge, so no single point on it stands in
for the whole the way an interpass reading is allowed to.
The practical consequence is a reporting rule rather than a formula: any weld pool
temperature reading is only meaningful stated alongside the position in the pool it was taken
from and the surface emissivity assumed to convert measured radiance
into that temperature — liquid metal under an arc is one of the harder surfaces to fix an
emissivity value for, because it is bright, moving and partly obscured by arc glare rather
than a settled, oxidised solid. A reading missing either of those two facts is a number
without the context needed to judge whether it means anything at all, let alone whether it
can be compared against a different pass, a different machine, or a different day.