The key distinction
Physical dispersion is real mixing caused by aquifer processes. Numerical dispersion is artificial spreading introduced by discretization and solution methods. In transport modeling, numerical dispersion can look like dilution because concentration is averaged over cells and time steps.
Why robustness comes first
Early transport simulations should usually be robust. A stable first simulation gives users the big picture: plume direction, migration pathway, capture behavior, discharge location, and broad risk pattern. Overly aggressive accuracy settings can fail or obscure the first-order story before users understand the system.
Numerical dilution
Numerical dispersion represents an averaging effect. If grid cells are too large, the model can average high concentration over too much volume, producing artificial dilution. This can happen horizontally and vertically, especially where plume thickness is small relative to cell size.
How to reduce it
- Reduce grid size where concentration gradients matter.
- Use more computational layers for vertical plume structure.
- Use unstructured grids to refine locally without over-refining the whole model.
- Use submodels to focus resolution where the plume, well, or receptor matters.
- Switch transport solver settings after the dominant behavior is understood.
Workflow
Start robust. Understand the picture. Then refine accuracy where dispersion, dilution, and decision sensitivity matter.