Well, pick your proprietary platform that doesn't publish how they do it. Steam's is okay. Oculus's is okay. But the reason I'm writing my own is that normally the ones by/for sighted people sacrifice precision for cheap computational power and tricks and things (save me from ambisonics, for the love of god).
The thing about volumetric sources is that volume is conveyed by a lot of other stuff. You don't just hear with your ears. You actually hear with your whole body. Temperature differences, wind, vibration through the ground, pressure in the air, all of these things matter. Put on headphones and you've given all of that up. Then add in the fact that HRTF in general isn't individualized to you and there's that disconnect. And at the end of that you end up with--it's not quite binary. But it's, at best just a slider that says "vaguely big" vs. "definitely not big". In theory I could get you "big and somewhere in front and to the right" but you're never going to get shape out of it really, not by itself. you'd have to like animate the sources and make them move around and things like that, which you can, Synthizer is fast enough to let you for one thing and OpenALSoft probably is too, but as soon as you say that anything big is 10-20 moving sources you're eating a ton of processing power so you can only get a couple things like that per scene. Plus, you'll have to feed each of those animated sources different audio, or they just converge into one. And even after all that it's going to be kind of obviously artificial.
But mind you, "here's a source for each train car" probably works out super well even without volumetric anything. Maybe I should do that as a demo, actually.
But you won't get it to the point where it's really useful for navigation or anything like that, and you probably have to buy it. No one in OSS land has it to my knowledge unless OpenALSoft did it since I last looked. But even if you did pay for it the best you can do is "vaguely big, and we always put the source on the closest point to the player so that they know if they're close to the edge". it's just not solvable unless you give me some sort of amazing full-body vr chamber if you want it to sound natural.
Which puts us back at "give the player a radar". If you want something navigable then the player will need unnatural-sounding tools to do it with. There's no real way around this. So, the answer is yes if you just want it to sound cool, but definitely not if you want players to be able to follow the river without some auxiliary thing to tell them where the path is.
My BlogTwitter: @ajhicks1992