There's tea in your cup, with leaves on the bottom or floating. You stir the tea with the spoon. Let's define z the vertical axis of the cup, i.e. the axis normal to the base of the cup. The plane x-y is thus parallel to the base of the cup. By stirring, you are forcing the motion of the fluid by imparting an angular velocity. The rotation you're forcing is around the axis z. If the fluid were a solid body or non-viscous fluid, the motion will be perfectly circular in every cross section (parallel to x-y) of the cup. No vertical mixing would occur.
But your tea is a real fluid, thus it is viscous. If it's viscous, it dissipates energy through frictional processes and its velocity at the wall must be zero.
Now, let's assume the cup is a cylinder without the top surface. We thus have 2 walls: the vertical circular one, and the bottom plane one. On each wall, the fluid velocity must be zero.
Let's now consider the qualitative profile on the fluid on vertical (i.e. containing the vertical axis) and horizontal (i.e. parallel to the base) cross sections. On horizontal ones, the fluid is rotating with a velocity profile (intuitively quadratic) that is zero at the center and zero at the circular wall boundary. The maximum will be somewhere in between. On vertical cross sections, the fluid must have zero velocity on the bottom and side boundaries. The maximum will be close to the top center, i.e. the furthest from the wall. To accommodate these boundary conditions, the fluid motion should be again a circular one.
Now combine the motions. You have a helicoidal motion that brings the fluid from the top to the bottom and vice-versa. As the tea leaves are heavier than water, they follow the downward motion but they cannot come back up. Thus, they will concentrate in the least disturbed point of the 3D domain: the center of the bottom surface.
Finally, apply the same model to the river motion by making the following substitutions:
- stirring spoon ---> Coriolis force + possible bend
- tea leaves ---> erosion products (sand, particles of rock,...)
I assume, the Coriolis force is responsible for this effect. As an example, you can observe a similar effect at railway tracks. One side is more stressed, than the other and in northern areas opposite to southern one's.