Mining related Groundwater evolution

[rapheul]

Hi, there,

I am studying on the groundwater evolution process in a mining area, I have 4 aquifers with one coal mining layer in the very middle (above are two aquifers, one is the first aquifer-unconfined, another is the coal seam direct roof aquifer; and below are the other two limestone aquifers). each aquifer has time series water samples as well as the mining stope water. There are no isotopes data.
We aim at getting knowledge of the origins of mine water and surface pit lake water, and also some connections check, for instance, shallow aquifers with deep aquifers . then we have several assumptions before modeling:
1) Mine water may mainly come from singlely or mixture of direct roof and floor aquifers near to it.
2) The first unconfined aquifer is probably connected with the bottom limestone aquifer as the bottom aquifer has a outcrop area and faults in the mining area.
3) The lower two limestone aquifer may have as well connections and the bottom aquifer water may flow into upper aquifers through fractures.
4) For subsidence pit lakes water, we would also like to know its origin. however, we do not have rainwater samples. we assume the water is from the mixture of mine dewatering system (mine water), the phreatic aquifer and rainfall.
Based on those conditions. am I correct that we should use inverse modeling to deduce the origin of each concerned water? But what should be done with the connections check? Or what else concept models can anyone suggest?

Thanks in advance!     

[dlparkhurst]

I would try inverse modeling. You can hypothesize, for example, that water in the stope is a mixture of waters above and below plus additional (oxidation) reactions in the stope. Inverse modeling may provide a set of mixing fractions and additional reactions that indicate the hypothesis is plausible.

However, inverse modeling can be more difficult than it appears. First, because you need to develop a set of plausible reactants, and second, because inverse modeling can fail to find a result that seems perfectly clear. Still, the exercise is worthwhile and much more direct than forward modeling.

[rapheul]

Hi David,

Thanks for your suggestions. I will try inverse modeling for mine water. I have some information of potential reactants.

But for the other waters, for example, surface pit lake water without rainfall data, What do you think how it should be done? I think I need to find some rainwater samples if still using inverse modeling? 

And the aquifers hydraulic connections check, like bottom aquifer is recharging upper aquifers. can inverse modeling also help or what else methods can I apply in PHREEQC? Upper aquifers all have their own rock water interactions, can these reactions be included while doing inverse modeling at the same time?

Best regards

[dlparkhurst]

Pure water is probably close enough to rain water. Even concentrated 10 fold, concentrations are pretty small relative to groundwater.

I think inverse modeling is the most direct method. If you use any other PHREEQC capabilities, you have to specify all of the reactions, which is what you are trying to deduce. Thus, it is a trial and error process.

PHREEQC considers one or more initial waters and a set of reactions. It could include reactions from any aquifer or conceptual model. However; if you want to consider reactions in the upper aquifer, I think I would first use inverse modeling to try to determine the evolution of the upper aquifer alone. It will be simpler if you can isolate the reactions in as simple a conceptual model as possible.

[rapheul]

Thanks David, your comments are really helpful. Regards

[rapheul]

Hi David,

I have some inverse modeling questions on correct understanding. I tried inverse models to find out the mixing fractions of each water, for example, 2 waters to form mine water(with potential reactants), defined with 3 solutions. I am a little confused that under what conditions should I use phase H2O(g)? Only when we assume that the initial water may be evaporated or? Because the results may differ if I do not add this phase. By adding it, the uncertainty can be set to a smaller value which seems more preferable. I then obtained results showing H2O(g) with positive value, does it mean dilution or more vapour is condensed into it? 
I think when run for such underground mining related simply mixture processes, there should be no such dilution process and even if there is, should be rather small. But for pit lake water or shallow water involved inverse modeling then the phase H2O(g)(also CO2 and O2) is necessary?

Another question is about the fractions that I got after run, if the results shows "0" zero of solution 1 or 2 (still the mine water example with 3 solutions). does it indicate the mine water is singly from minerals dissolution or mixture with another water not zero valued?  The "0" initial water then should not be considered as a source water in this case?

If the fractions of 2.7e+1 means the water must be 27 fold concentrated, then does a value of 2.7e-1 indicate dilution?

About the uncertainty, sometimes I need to give a quite large value like 50% to find a model, in this case, can I say that the two waters have very limited connections? I wonder normally what is the reasonable limit of the uncertainty. Or if larger than a value, the model does not make any sense.

Maybe these questions are a little basic, but I think I need to confirm my understandings for further modeling design. Thanks a lot for your kind help.

[dlparkhurst]

Yes, adding the phase H2O(g) allows dilution or evaporation. Evaporation makes more sense for a system in contact with the atmosphere like a pond or, possibly, a mine working. Possibly addition of dilute rainwater makes sense in some cases.

A think a pit lake is usually subject to significant evaporation, as well as exchange of CO2 and O2.

A mixing fraction of zero means that the water is not part of the mole-balance model. If all of the initial solutions are zero, or sum to less than 1, then water has been derived by mineral reactions. The simplest example of such a reaction would be to dissolve 27+ mol of gypsum to arrive at one kilogram of water. Usually, these models without an initial solution are not plausible, and are usually ignored. It is possible to use -mineral_water false to ignore water that is derived from mineral reactions. However, to consider evaporation, it is necessary to allow the "mineral" H2O to react and affect the mass of water in the final solution.

You would expect a value for a mixing fraction that is greater than 1 if evaporation occurs. A value of 27 would be accompanied by a loss of 26*55.5 mol of H2O. A mixing fraction of 0.27 (without a mixing fraction of 0.73 for a second solution) indicates you are adding water, either from H2O(g), or other mineral reactions (say gypsum).

I like to have uncertainties less than 20%, and preferably 5 to 10 percent, with exceptions for trace elements, or sometimes other elements that are not too important to the model. A value of 1.0 will allow any water to evolve to any other water. A value of 50 may indicate the waters are not related, or it may mean there are other processes that you have not thought of.

[rapheul]

Thanks so much. Now all the questions are clear. Regards