I think SAIL3 pretty much hit the nail on the head. But, you might also consider the following: (1) I would like to know much more about what these arch structures really are, how they are loaded and magnitudes, how they’re braced, how the torsional loading is applied, do they have vert. legs/columns under the 180° arch, how is the arching action supported and reacted, etc. (2) If there is torsion, how much and how concentrated? However you cut it, with torsion, there can be a tendency for two corners to close and two to open up, and you never know which ones, so all four must be considered as important. I don’t think the exact type of weld is the important thing, they all will have roots which can be adversely affected. The important thing is that your weld detail leads to a good, sound, root pass or root condition. (3) I would like to see the moment, shear and torsion diagrams and stresses in these arches so I could study what to do with the fab. and welding details. (4) Remember that flange tension on the inside of a curved member has a stress component which pulls the flange away from the webs, and this stress must be included in designing the welds. Flange compression on the inside of a curve presses the flange into the web, generally not a problem because it is a compressive bearing stress, but that stress must still be considered in the weld design. On the outside of the curve this flange/web condition is the opposite in terms of pulling the flange away from the web.
My first choice for a cross section would be the following, assuming this was acceptable aesthetically. It is totally honest from the fab. and design standpoint and is most economical, so I’d fight for it. I would use 6" (+/-) wide flange plates and press or pull them to the webs for fit-up and welding. I would set the webs at 5" (+/-) wide, out-to-out, leaving .5" (+/-) of flange outstanding to receive a nice clean fillet weld. This has the potential of good, but simple, fit-up and welding, and as good a root condition as you are going to get short of a fillet on both sides of the web, and you have access problems on that account. I would cut the webs out of plate, in two or three pieces, 90° or 60° of arc. This gives you some economy btwn. plate size/waste and splice welding and clean-up. In cleaning up the cut web pieces, you could grind or nibble a 1/16th or 1/8th inch bevel on the webs, to pick up some weld throat, but this is expensive and not needing in the above cross section. This would be needed in your option #2, and possibly in SAIL’s backer bar scheme. As SAIL mentioned, your option #1 is an almost impossible fit-up problem out in the fab. shop. But, easy to draw.
Omer Blodgett has several real good books on weld design and the design of fabed. structures.