Inverse model for conduit flow karst spring with unknown recharge

[Grivia1]

I'm attempting to determine source water of a karst spring that has unique chemistry relative to other springs located less than a mile away. The recharge/source for this spring is unknown. It should be discharging from the same unit as the others, but extensive faulting in the area may be changing its chemistry. I have cation/anion data for this spring, 5 nearby springs, and stream data that the other springs interact with. I'm having trouble with uncertainty errors which may be because I'm approaching the model the wrong way.

Thanks

[dlparkhurst]

For starters, I would try to make the spring waters from pure water. If you can find plausible reactants that account for the evolution of the waters, then you can consider the differences between the one spring and the others.