Thermal resistance models are just like electrical resistance and are added in series and reduced in parallel etc
The key determinant is φjc which is made up of φjc + φca ( case to ambient ) , now some very coarse rules apply, the case to pcb thermal resistance can be ignored and the rough rule I use is 500 W/C for 1cm2 of double sided pcb connected via an array of thermal via . The component φja is effectively ignored as it’s a thermal resistance in parallel with a much lower die to base thermal resistance
Ie φca = 500 for 1cm2 , or 50 for 10cm2 etc.
With that you can now do a rough calculation of the temp rise over ambient using a pcb as the heat sink , for single sided , use 1000C/W for 1cm2.
Note these are very rough figures
So a double sided board board with a largely unbroken copper planes of 80 x100mm on both side would have a case to ambient of around 6.25 W/C or 13 W/C for a single sided plane , assuming the case was connected directly to that plane and was in the centre of the plane
Often in double sided boards the SMD is mounted on a small pcb pad ( ie DPAK ) connected via thermal visa to a ground plane , such vias have around 250-300 C/W thermal resistance and each one is in parallel , so say you had 16 under the Dpak , that would add circa 20 C/W which would make the total for the ground plane underneath around 30-35 C/W for 80 x100mm
Obviously there are complications where the thermal plane under the DPAK can’t be electrically connected to the underlying ground plane etc
That size plane could comfortably handle 1W with room for a rise in ambient. Obviously this assumes one heat source ( multiple sources can be approximation by addition ) a 1.5 to 2x safety margin is advisable and of course these figures are open air , static cooling , 1 ounce copper
This is very back of a napkin stuff and there’s can be serious discrepancies in real life
A very good AP note is
http://www.ti.com/lit/an/snva419c/snva419c.pdfDave