This work assesses the performance of several array processing algorithms for directional of arrival and power estimation to the problem of radar ice bed tomography. A cluster-based wideband radar simulator is described which is used as the basis for the assessment. We look specifically at target configurations that are representative of a layered media with two interfaces: air-ice and ice-rock. We evaluate algorithm performance in terms of RMS error at both normal and off-normal incidence angles by adjusting algorithm parameters and optimization methods. The optimization method refers to the non-exhaustive global search for selecting the direction of arrival given the a priori knowledge that the targets are arranged into two smooth interfaces. These results are used to select the best algorithms and optimization methods for a variety of surface types. We then apply these selected routines to the problem of radar ice bed tomography and discuss the results.