There are many algorithms for direction-of-arrival (DoA) estimation in array signal processing. Most of the work is carried out in scenarios with a large number of available snapshots. That is, received signals are sampled for a certain time duration. This paper focuses on worst case scenarios of having only one signal snapshot available, while trying to resolve several target signal sources. Such scenarios might occur in automotive radar applications, with small number of available snapshots being available. Moreover, other difficult propagation channels with correlated signals might reduce the possibility to exploit an increased number of signal snapshots. This paper elaborates on improved signal covariance matrix estimation based on averaging in space instead of taking time-averaged samples.