It's much more that a frame of reference issue. For example, to calculate the trajectory of a shell you can assume an initial velocity and direction and you calculate the path by taking the following into account: mass and inertia of the projectile, rotation and position wrt to its path, drag coefficient and other effects of wind force and so on: you know the forces on the shell and where it will end up. Meanwhile, you have 'summarized' the effect of wind into a few simple coefficients like so
If you want to determine this drag coefficient you can do various tests, or, you try to simulate the flow as accurately as possible. That means that would you've just done for the shell you now need to repeat for each molecule in the flow. Now, this is a bit too much, so you use the Navier-Stokes equations that apply to a small volume of the flow. Basically, the Navier-Stokes equation are nothing more than Newtons second law: Force equals mass times acceleration, applied to a flow. The tricky part is that, compared to a single shell, there isn't really a single mass to deal with. Now, acceleration is the change of speed, so Newton actually said that force equals mass times the change in velocity. For Navier-Stokes you first need to go to force equals the change in mass times the change in velocity. or
F= m*a --> F=m*change(v) --> F=change (m*v)
Speed times mass is called momentum, so basically newtons second law is a more primitive version of force = change in momentum.
Modern calculation methods rely on subdividing the flow into many small volumes, taking into account that fluid that flows from one volume to the next is equal, that forces equal out and flow through each volume. So, you know how the fluid flows and forces interact. The problem is that solving the flow motion in each small flow volume is a massive undertaking. Ships are hundreds of meters long while the flow near the hull consists of small vortices much smaller than millimeters: tracking them all is currently impossible because you hardly have the memory to subdivide the flow around the hull in all these volumes that all capture the smallest vortices. This is where the turbulence modeling comes in: you assume the flow doesn't really have these vortices but that the flow is more viscous than it really is in real life because of these vortices. This works pretty well but you need to be really, really careful how you go about it. Near the hull our calculation volumes are very very thin, more so than a sheet of paper; further away from the hull the volumes become more cube-like and larger. Using these approaches makes the problem solvable (we use several 10s of millions of volumes for normal calculations).
And while you do this, you STILL need to take the mass and velocity of the ship or shell into account to predict its motions. And if you're really precise, you need to determine how much the ship will deform under the fluid loading and predict the influence of ship deformation back onto the fluid. At the moment this is all way too much for our computers to handle all at the same time but progress is very fast
When we calculate the self-propulsion of a ship at one speed, we use 256 processors and calculate for 1 week for one answer...