Conservation?
 

I used first-order upwind scheme to simulate the Zaleska problem. The interface is continuous, not sharp. The equation is conservative, so do the sheme. But the scalar quantity is not conservated.

Anyone can tell the possible reason?

Thanks