Query about calculation of Saturation Indices

[sally.xiaosa.0212]

There are several questions about saturation indices. I'm wondering about the way to judge results and asking for help. The questions are as follows:
1. In output, several SI results of direct calculation after solution input is distinctly different from SI results in inverse modeling, what?s the reason and how to judge and choose?
2. In inverse modeling, what is the acceptable level of uncertainty for input? Sometimes, a way to eliminate the charge imbalance error is increasing the uncertainty, but what?s the limit?

[dlparkhurst]

To the extent you trust the SI calculations, an inverse model is invalid if it precipitates an undersaturated mineral or dissolves a supersaturated mineral.

In PHREEQC inverse modeling, you must define at least enough uncertainty to allow for charge balance. I prefer analyses that balance within 5 or 10 percent (cations - anions)/(cations + anions), and I like to see several analyses that demonstrate the same trends in concentrations. In general, I like to use 0.05 or 0.1 for uncertainties, but you will have to account for the quality and consistency of your analyses.

[sally.xiaosa.0212]

Thanks for your reply, Dr. Parkhurst. In my inverse model, the uncertainty of input solution2 must be increased to 0.4 to eliminate charge imbalance error, perhaps due to missing of NO3- data or large uncertainty in HCO3- data. After solving the problem of solution charge imbalance（uncertainty of input solution2 was set 0.4）, the output of inverse model path is too much, there were 75 or 155 paths. So a new question is how to simplify and choose the most reasonable one?

[dlparkhurst]

Start by using the -minimal option.

If you know some minerals must dissolve or precipitate, you can put dis or pre for those minerals, which may decrease the number of models.

Don't spend too much time on inverse modeling if your analyses have that much charge balance error. And, in general, inverse modeling is a process for you to consider the reactions that are occurring in your system, even if you do not find a quantitative model.

[sally.xiaosa.0212]

Thanks for your advice, Dr. Parkhurst. My research focuses on quantitative calculation of transfer amount of iron & manganese minerals.
Firstly the charge imbalance error of input solutions was solved. Then I selected ?minimal? option and set dis/prep states of several minerals in phases. The process of reducing modeling results was actually time-consuming as I needed to set mineral dis/prep state one by one (Seems to depend to input order of minerals). Now output models has decreased to just 2. Seems in line with expectations. But several new questions appeared:
1.   The dis/prep state was set according to SI results of initial solution. Take Hematite as example, the calculated SI of Hematite is 18.6, which indicates prep. If I set prep state in phases windows, the output model was 0. The state of Hematite must be set as dissolution to get 2 output models but it contradicts the SI results. So on this occasion, whether SI results was reliable? Should I trust the SI calculation or the quantitative model?
2.   The mineral phase input includes Hematite, but the output didn?t show the mole transfer result of Hematite, only the transfer amount of other Fe-bearing minerals, which does not reach the purpose of quantitative calculation of targeted Fe-bearing minerals.

[dlparkhurst]

First, consider that the Fe(3) is not usually measured, so the SI for Fe(3) minerals relies on the total Fe measurement and your choice of pe.

It sounds like you have an increase in Fe concentration, which means you need a mineral that dissolves. It is probably not hematite or any of the other oxyhydroxides like goethite or Fe(OH)3.

If the increase in Fe is significant, you should consider other possible sources--pyrite, aluminosilicates, others.

[sally.xiaosa.0212]

Thanks for your help, Dr. Parkhurst. I?m now carrying out forward reaction path simulation, and some new problems have emerged.
I originally planned to use the amount of mineral transfer calculated in the reverse simulation as the input for the forward simulation, but found that the units did not agree.
The unit of mineral transfer for inverse modeling is mol/kgw, but the unit of mineral transfer input in forward modeling is mol. As it is impossible to acquire the water quantity in a natural groundwater system, how to solve the unit problem and continue the forward simulation?

[dlparkhurst]

The units of mole transfers from inverse modeling are moles. The amount would change depending on the mass or volume of solution. In most cases you are using SOLUTION definitions with the default of 1 kg water, which will close to 1 L. So, effectively the mole transfers are numerically equal to mol/L.

Your water analyses are concentration, usually mg/L or ppm, which would apply regardless of the representative volume that you choose. If you use a representative volume of 1 L (~ 1 kgw) for your groundwater system, the inverse modeling results would be directly applicable. You would be tracking the evolution of 1 L of water as it moves through the groundwater system.