If you can create a relationship of (for example) Sor as a function of OWC, then its easy. When you have a moderate degree of linkage only it becomes more difficult, and you will just have to think through what makes most sense for your situation.

I do not have any sheet to calculate appropriate pairings to most efficiently calculate a response surface for a given number of simulation runs - I've seen many different tables though that for a desired number of runs suggest optimal parings to gain most information - those papers should be a good start, otherwise I think the likes of JMP can help. I normally use the package EnABLE to do it all in the background automatically for me.