Grid with huge dimensions
 

Hi every one
I want to study a multiphase flow (water+ metallic particles (Ca)) which enters a lake with huge dimensions (4km*2km*2m). The target is specifying the dispersion and sedimentation of metallic particles inside water. It's worth to mention that the lake doesn't have any output (channel, soil diffusion, and etc.) except water vaporization from its interface with the surrounding air.
I uploaded a simple drawing of the problem.
http://s8.picofile.com/file/8347065284/Drawing1.jpg
Could anyone guide me through grid generation/independency + model validation policy? In other words, I thought about two ways till now:
1- Generation the mesh at its original dimensions plus, using some solution like using size function or link meshes in order to decrease and control mesh volume.
2- Performing grid independency and validation of the model at lower scales based on a solved numerical or experimental model then using scale factors (in fluent) to reach the real dimensions.
which one is better or applicable? is there a better solution to reach a suitable grid?
I'm using Gambit 6.2.4 and ANSYS Fluent 19.2 through this problem.
Any help will be appreciated.
Thanks.