Be careful with those calculators. Use them as first quick calculation, never as an accurate, definitive tool.
When current passes through a PCB trace there is an equilibrium between the generated heat and the heat going away, basically, by NATURAL CONVECTION of the surrounding air.
Heat transfer by natural convection is highly empirical and depends on many factors, like geometry (horizonta, vertical, top face, bottom face) as well as non-idealities (holes in a plate, air flow perturbations, conducting/ non conducting objects, etc.)
When designing, most of the times there is a maximum environment temperature (like 60 C or 80 C) plus a maximum temperature in your trace (depends on the PCB quality as well as the mean life you want to give to your product, it is around 150 - 200 C), so your temperature rise in a standard application would be the difference.
Keep in mind if there are other components or traces that could add local additional heat to your trace 'under test'.
Also, final thickness of copper traces are different from 'nominal' values (35 microns for 1 ounce)
I suggest one of these three ways:
Go and use a FE/CFD tool (Finite Elements/ Computational Fluid Dynamics) if you like theories and simulations and have money for computers and not for prototypes.
Go and build a prototype and measure the actual temperature. Correct your model if necessary.
Use the calculator and be very conservative (choose a thicker trace)