BanditSparkle: Hands-free stellar asteroseismic parameter estimation with a coherent-mode likelihood function
in bad data, even incoherent modes look coherent
Feeney & Hogg
Copyright 2017 the authors. All rights reserved. Everything will be released under a sensible license, and soon.
Stellar oscillations are stochastically driven high-Q modes. If a star is observed through an observing campaign that is shorter than the coherence time of a mode, that mode will effectively appear coherent within that time interval. Here we exploit this to create an inexpensive probabilistic model for asteroseismology data. The huge advantage of this approach to traditional asteroseismology approaches is that it does not require the performance of a fourier transform or anything like it; this permits the measurement of asteroseismic parameters in data sets that are irregularly or sparsely sampled, or in data sets where the critical modes are near the edges of the available frequency space (near the inverse exposure time at the high-frequency end or near the inverse of the duration of the campaign at the low-freqency end). It also permits the construction of a tractable and justifiable likelihood function, even when the noise processes are non-trivial. We produce a parameterized description of an asteroseismic frequency spectrum and use it along with this likelihood function to produce posterior estimates of asteroseismic parameters for stars in the Kepler data. We sub-sample the Kepler light curves to (short duration) TESS-like and (sparsely samples) Gaia-like data sets and show that even in these much smaller data sets, we ought to be able to estimate asteroseismic parameters for many stars, and (because we are just turning the Bayesian crank) without human intervention.
- We have actually done nothing at this point; this is all vapor-ware.
- There is another paper to write about variable exposure time, and why TESS is making a huge mistake by using a uniform time sampling.