Evolutionary Computational Methods for the Design of Spectral Instruments

We have developed a technique based on evolutionary computational methods (ECM) that allows for the automated optimization of complex computationally modeled systems. We have demonstrated that complex engineering and science models can be automatically inverted by incorporating them into evolutionary frameworks and that these inversions have advantages over conventional searches by not requiring expert starting guesses (designs) and by running on large cluster computers with less overall computational time than conventional approaches. We have applied these techniques to the automated retrieval of atmospheric and surface spectral signatures from Earthshine observational data. We have demonstrated that in addition to automated spectral retrieval, ECM can also be used to evaluate the discriminability of scientific results as a function of requirements placed on the spectral model. An important application of this technique is for the optimization of design parameters for spectral instruments.