Typically for this kind of output the letters put the treatments into groups such that anything within that group isn't statistically significantly different from one another. So there isn't enough evidence that anything in group A is different from anything else in group A. Treatments that don't share any group letters are statistically different from one another though.
For example Temp=5, Location=Ringsted has Index A so it's not different from Temp=5, Location=Solrod but it is different from Temp=10, Location=Gravlev. That's just a single example and I'm playing a bit loose with the language but hopefully it illustrates what that particular encoding is supposed to represent.
Thank you, Dason. In this example I've combined 4 different treatments (Temp = temperature). If I make a two-way ANOVA for Temp = 5 only it says Ringsted-Gravlev and Ringsted-Solrod is different from each other. How do I see that in the index letters?