It's handwavy but I can't help but think that because of the way symbolic dynamics works that you can't find families of approximating functions inside turbulence even though I can't see it being practical, like sensitive dependence on initial conditions will make noise overwhelm your ability to take measurements. See