Analyses of the latest Gaia data release (DR2) by various groups have revealed that the dark matter (DM) halo around the solar neighborhood is in disequilibrium, and an O(1) fraction of the DM may have been accreted during recent dwarf galaxy mergers. These results indicate a departure from the Standard Halo Model (SHM) of DM velocity distribution, which could significantly affect DM signals at direct detection experiments. In this talk, we parametrize the local DM velocity distribution through a model with three distinct components: an isotropic halo, an anisotropic substructure component (the so-called Gaia Sausage), and a coherent stream of DM associated with the S1 stellar stream. We illustrate the effect of astrophysical uncertainties on discriminating signals from various DM models using the novel framework of information geometry. In particular, we review the Euclideanized signal method introduced by Edwards & Weniger that allows us to rapidly scan over a large-dimensional parameter space without resorting to computationally expensive Monte Carlo simulations that rely on a finite number of arbitrary benchmark points.