A new statistical method for characterizing the atmospheres of extrasolar planets
Physical and Biological Sciences
LAMAT (http://stemdiv.ucsc.edu/lamat/)
By detecting light from planets outside our solar system, we can measure their compositions and bulk physical properties. The technologies used to make these measurements are still in their infancy, and a lack of self-consistency suggests that previous observations have underestimated their systemic errors.
Ordinarily, models for planetary atmospheres would be fit to the brightness profiles measured, but in the case of inconsistent photometry, finding a best fit model is more difficult because some measurements disagree with each other. We demonstrate a statistical method, newly applied to exoplanet characterization, which uses a Bayesian formalism to account for underestimated errorbars (that is, overestimated precision in the measurements). This allows us to formally address problematic data in our model fitting.
We use this method to compare photometry of an exoplanet, GJ 758b, with custom atmospheric models. Our method produces a probability distribution of atmospheric model parameters including temperature, gravity, cloud model (fsed), and chemical abundance for GJ 758b. This distribution is less sensitive to highly variant data, and appropriately reflects a greater uncertainty on parameter fits.