operational methodology to immobilize arsenic during aquifer storage and recovery

[Kirkoff Xn]

Hi Brian and Jia

I come across an issue and would appreciate if you would help out , and maybe suggest how to go with my model.

Conceptually, I am trying to test an operational methodology that has shown promised in attenuating the geogenic arsenic contamination during aquifer storage and recovery. Basically, I am running two model scenarios: the first one (ASR_Buffer Zone) aims in creating a buffer zone by promoting an oxidizing environment around the ASR well during all cycle tests (3 cycles in this case). The second (ASR_No Buffer_Zone) corresponds to over-recovery that usually results in releases of arsenic up to levels higher than their maximum contaminant level (10 µg/L) in the groundwater. While the first scenario promotes formation of Fe oxyhydroxide with high sorption capacity of trace elements (hence arsenic attenuation around the ASR well), the second scenario tends to deplete oxygen around the ASR well; with oxygen depletion, arsenic would tend to remobilize following Fe oxyhydroxide reduction dissolution.

I am working with the thermos database developed by Lazareva et al. (2013) called (Thermo_GKD.tdat, see attached) and the two-layer surface complexation database (FeOH+.sdat).

As you can see on the ppt attached, the oxygen concentration in the first scenario is quite higher than in the second scenario, meaning that ideally the first scenario would promote more arsenic attenuation with formation of Fe oxyhydroxide (most of them, if not all, being weak HFO forms in my case). However, as seen on the 2^nd slide, both scenarios seem to form almost the same amount of Hydrous ferric oxides by the end of the simulation time (954 days buffer zone scenario versus 884 days in the case of no buffer zone).

I have tried to go with kinetic model scenarios of arsenopyrite, but it didn’t improve the results.

Any advice to improve my model?

 

 

[Jia Wang]

Hello,

It sounds like in the first case, you are trying to get Fe(OH)3ppd to precipitate and thus provide a surface for complexation of arsenic contamination. In the second case, you are trying to dissolve the Ferrihydrite.

When I ran your models and plotted the Fe(OH)3ppd volume at the ASR well, it seems like the mineral precipitated in both scenarios and in fact there are more Fe(OH)3ppd that forms in your case of no buffer zone than the model with the buffer zone. This doesn't seem to fit the scenario that you conceptually described. If the main difference between the two models is the presence of Fe(OH)3ppd for surface complexation, then I would suggest simplifying your model to investigate. Maybe eliminate the reactants and see what effect that has on the model. You may want to investigate fluid chemistry in React for the initial composition (your native groundwater) mixed with your injection fluid. Maybe also consider how the pH can factor into Ferrihydrite stability?

Another thing to check is to see whether you have any mistakes with the units for fluid concentrations. Looking at your input files and comparing the two, I see the same amount of O2(aq) assigned for all fluids.  How are you modeling an oxidizing environment for the first and the second with oxygen depletion? I also noticed that the native groundwater has the same concentration between the input files for all components except for Fe++.

I noticed that your X2t model does not have any flow set. Is that intentional?

Hope this helps,
Jia Wang
Aqueous Solutions LLC

 

[Kirkoff Xn]

Hi Jia,

Thank you for your insightful comment.

As far as the amount of O2 assigned goes, the native groundwater (initial system and Native_GW) has O2 of 0.01 mg/L and the injectant  has O2 of 4.5 mg/L. The operational methodology developed can be found in "Wells" pane and there you can see that both scenarios totally differ in their operations. In the case of Buffer Zone, the methodology is such that a buffer zone (see attached picture below) is maintained and there is no over recovery of water that has been injected, hence promoting more oxidizing environment around the ASR well. However, in the 2nd case of "No Buffer zone" there is over-recovery and therefore oxygen concentration is way lower than in the case of "Buffer Zone" scenario. I understand that the pH also controls the sorption capacity; maybe I need to set a fixed pH considering the buffering capacity of carbonates? But even that, it does not help. I was expecting high oxidizing conditions around the ASR well to promote more Fe(OH)3ppd for surface complexation in the buffer zone scenario, but it looks like I need to refine my conceptual model. 

Thanks

 

 

[Kirkoff Xn]

Maybe I need to further give a context of my conceptual model adding the below figure. As you can see, the cumulative storage volume as a negative correlation with arsenic concentrations, meaning that the more the buffer zone is maintained and enhanced, the more the arsenic is immobilized and the only way I am thinking of it is the formation of Fe(OH)3ppd for surface complexation around the ASR well.

 

And yes, I neglected the background groundwater flow considering the high flow rate induced by the ASR well during the operation.

And you're right that the No Buffer" scenario yields more Fe(OH)3ppd than the buffer scenario, which is quite unexpected!

 

 

 

[Jia Wang]

Thanks for providing more context on your conceptual model and the explanation.

To simplify your problem for troubleshooting, you can eliminate the sorption reaction and focus on getting the mineralogy and fluid chemistry of your model correct. Even though sorption is ultimately the reaction that you want to focus on, it’s heavily dependent on getting a model to correctly account for formation of Fe(OH)3. The best place to start would be creating a reaction model in React.

To transfer your model from React to a reactive transport model, I would suggest evaluating the role of your reactants. In a reaction path model, you can add in minerals like Calcite (moderating the pH of your system), and pyrite/arsenopyrite for the iron content to modify the composition of your system. This is typically less commonly used in a reactive transport modeling setting. If calcite is present at the start and in equilibrium with the fluid, for example, it should be swapped into the initial system’s basis. If a mineral exists in your domain but is not in equilibrium with your fluid, you may want to consider using a kinetic rate law to constrain its formation and dissolution. Also, is there any Fe(OH)3 present in the initial domain, before injection cycles?

A 1D radial model might also be another step before a 2D model. When you are working on a 2D model, you create parallel models also focus on the transport side of things. For example, in a simple model, set a Br- tracer to see the size of the zone influenced by injection and see if that matches what is expected.

Hope this helps,
Jia

[Kirkoff Xn]

Thank you, Jia!

I will try using React to simplify this conceptual model. There is no Fe(OH)3 based on the results from the aquifer matrix mineralogy. 

Question: Is there an example out there that could help in computing/correcting mineralogy and fluid chemistry using React? It looks like that's where I should have first started from.

Also, I can calculate the zone influenced by the ASR well stresses, but not in GWB. Can you elucidate how to go with it using GWB? Does it mean I can just plot Br- (or Chloride) versus X position?

Thanks

[Kirkoff Xn]

----

 

[Jia Wang]

Hello,

Thanks for attaching your files. I think there are a few things for you to consider. Above, I mentioned that you might want to consider the roles of your minerals in your system. Swapping a mineral for a component in the basis tells React to calculate that component's dissolved concentration in equilibrium with that mineral. Is that true for all minerals in your system? Given how oxidized your initial system is (as indicated by the high concentration of O2(aq)), it is unlikely pyrite arsenopyrite would be present in equilibrium with your fluid. Are these minerals present in your system initially? If so, you might want to consider using a kinetic rate law for constraining its reaction. Again, the purpose of the React simulation is to investigate the formation of Fe(OH)3, which will ultimately responsible for sorption reactions. Please see section 4.1 for more information regarding kinetic reactions.

Other things to note, it seems like the mineral volume takes up 99% of your system. Is that correct? I think the low amount of fluid in the system is not helping with the convergence issue. In your reactive transport model, the porosity is set at 0.2.

O2(aq) should almost always be set as a free constraint, although in this case I don’t think that’s a huge issue.

You might also want to suppress more stable iron minerals like you had previously, so ferrihydrate can form.

Hope this helps,
Jia

 

[Kirkoff Xn]

As you advised, I wanted to check the groundwater composition in equilibrium with the 3 main mineral phases of the aquifer matrix since in my 3D model, the reactants (the 3 minerals) did not have any influence on the ground-water composition throughout the reactive model run. Since these minerals are initially present in my system, It is possible that O2(aq) reported is not accurate. React is though limited in mimicking the operational methodology I am trying to investigate. That's why I am using React to just investigate the initial composition of the native groundwater before injection. As far as the Volume % in the mineral phase goes, I know that the aquifer is made of calcite (98%) and a small amount of pyrite (1.5%), and arsenopyrite (0.5%) with 0.2 porosity.

[Kirkoff Xn]

--