If it's loud enough that you don't need to crank the volume knob more than a reasonable amount on playback you don't need to do anything more. If you want to get the average level up higher, you can usually find and modify a couple of the very highest peaks which are often significantly higher than all others. You may only need to do that for one or a handfull of those highest peaks to be able to raise the level of everything else enough to be comfrotable. The easiest way to modify them is to find them, zoom in until you can see the transient peak itself, then either redraw just the peak with the pencil tool or make a quick volume envelope change which just covers the peak event an lowers it to about the level of all the other peaks. That becomes a chore with more than a few, but usually there tends to be a just a few really bad offenders. A limiter can do it automatically, but is much easier to abuse unless you learn how to set it correctly and listen very closely to make sure it's not screwing up the sound.
You want to do all editing such as level changes, peak editing, and other stuff like eq or whatever before dithering to 16bit and exporting from your editor. However, simple file splitting and renaming is an exeption to that and can be done either before or after export. Check to make sure Audacity is set to automatically dither when exporting. The standard trianglar dither option is entirely adequate, you don't need anything fancier like noise-shaped dither.
Audacity has the abilitly to export files at different bit-depths, but that's not necessary if the end product you want is 16bit. You can either export the entire set as a single file, then split that into individually labled tracks using another tool (such as CDWave if you are on windows, which is excellent for spitting and renaming tracks) and that's how most folks seem to do it. Or you can do the track splitting and naming in Audacity and export the individual tracked files directly. It's mostly a workflow choice and I can't remember the specifics of tracking and renaming in Audacity. I use Samplitude and do it all within the editor, but I still use CDWave for some stuff.
It's safe and makes sense to normalize all your 'loud' concert recordings to -1.0dBfs as part of your standard processing. -15dB is a good target when recording most things, which leaves some saftey room for unexpected peaks. Lower than that is fine too, as long as you don't find that you need to amplify so much that you begin to hear the noise of your recording equipment. The tricky part is that it can be hard to tell sometimes if the noise at the bottom is from your equipment or from the environment in which you are recording. But none of that will be a problem for most loud concert recordings where the noise floor of the recording environment is usually well above that of your equipment.