Adaptive Kernel Density Estimation for Improved Sky Map Computation in Gamma-Ray Astronomy

Abstract

We introduce an alternative method for the calculation of sky maps from data taken with gamma-ray telescopes. In contrast to the established method of smoothing the 2D histogram of reconstructed event directions with a static kernel, we apply a Kernel Density Estimation (KDE) where the kernel size of each gamma-ray candidate is related to its estimated direction uncertainty. Exploiting this additional information implies a gain in resulting image quality, which is validated using both simulations and data. For the tested simulation and analysis configuration, the achieved improvement can only be matched with the classical approach by removing events with lower reconstruction quality, reducing the data set by a considerable amount.