The GA was run 100 times, starting from differ- ent initial conditions, and using the inverse of RM- SEP as fitness criterion. As mentioned before, the search space for GA was restricted to selections with a fixed number of wavelengths, which was set to 14, in order to allow a comparison with the SPA solu- tion. Remark that, even with this restriction, the number of possible selections is still larger than 2= 1013. The best and worst GA realizations yielded RMSEP of 0.01 and 0.03 mg ly1, respectively.