mass transport: vertical dispersion

Greetings,

I am running a large model that was constructed by Dwaine Edington - he has posted on here several times so you may be somewhat familiar with the model.

I have set up the transient mass transport component for the model and I am simulating the release of a constituent via mass concentration BC's at selected nodes on the top slice. Currently I am disturbed by the concentrations diving deep into the bottom layers (6 through 9) of the model. I am trying to figure out what is driving the concentrations in the z-direction. I have calibrated the transient groundwater flow model and I am pleased with the results, so I am trying to avoid adjusting the material properties of the flow components. I am curious about the transverse dispersion term - from what I can tell, it is dispersion orthogonal to the direction of flow. I need dispersion in the lateral direction, but not vertical. Are there ways to focus on reducing dispersion only in the vertical direction? I understand you probably do not have enough information to answer my question, let me know what you need and I will provide as much as I can.

Thanks!
Humphrey

FEFLOW does not support a dispersivity model that distinguishes between vertical and horizontal flow direction. In the longitudinal/transverse model, longitudinal means 'in flow direction' and transverse 'perpendicular to flow direction'. This can lead to difficulties in case of vertical flow through aquitards, where dispersivity is often relatively low in vertical direction. In this case, I'd recommend to set a relatively small dispersivity value to units where you expect more vertical flow (e.g., aquitards).

Thanks Peter,

I assume you meant "less" vertical flow in the last sentence. Would a high contrast (1 order of magnitude or greater) in dispersivity values between layers/elements create numerical problems?

Humphrey

Sorry, that was not quite clear. What I wanted to say was 'where you expect a more vertical flow direction'. From my experience, contrasts of one order of magnitude should not cause instabilities - but this depends on the model, of course.