Interpolation Techniques for Reconstructing Galactic Faraday Rotation

Abstract The line-of-sight structure of the Galactic magnetic field (GMF) can be studied using Faraday rotation measure (RM) grids. We analyze how the choice of interpolation kernel can affect the accuracy and reliability of reconstructed RM maps. We test the following kernels: inverse distance weighting (IDW), natural neighbor interpolation (NNI), inverse multiquadric interpolation (IM), thin-plate spline interpolation (TPS), and a Bayesian rotation measure sky (BRMS); all techniques were tested on two simulated Galactic foreground RMs (one assuming the GMF has patchy structures and the other assuming it has filamentary structures) using magnetohydrodynamic simulations. Both foregrounds were sampled to form RM grids with densities of ∼40 sources deg −2 and area ∼144 deg 2 . The techniques were tested on data sets with different noise levels and Gaussian random extragalactic RM contributions. The data set that most closely emulates expected data from current surveys, such as the POlarization Sky Survey of the Universe’s Magnetism (POSSUM), had extragalactic contributions and a noise standard deviation of ∼1.5 rad m −2 . For this data set, the accuracy of the techniques for the patchy structures is ranked from best to worst as BRMS, NNI, TPS, IDW, and IM; in the filamentary simulated foreground, the ordering is BRMS, NNI, TPS, and IDW. IDW is the most computationally expensive technique, while TPS and IM are the least expensive. BRMS and NNI have the same, intermediate computational cost. This analysis lays the groundwork for Galactic RM studies with large radio polarization sky surveys, such as POSSUM.