We have investigated the regional Pn backazimuth estimation capabilities of the NORESS seismic array as a function of element spacing, frequency band, and time window to determine which parameters are optimal for reducing azimuth errors. We used a broadband frequency-wavenumber estimator to calculate backazimuths from the Pn arrival for each of 274 regional events recorded at NORESS for 126 parameter combinations. The large data base provides a wide range of signal-to-noise ratio (SNR) (0 to 70 dB), distance (up to 10.5°), and azimuth characteristics, and includes identified earthquakes and explosions as well as “unknown” sources. Most of the errors in backazimuth are less than 20° when appropriate parameters are used, and mean backazimuth errors are close to zero. The best results are obtained using a 13-element array configuration that has a 1.4 km aperture and a maximum station spacing of about 600 m. With the 13-element configuration and the data filtered to include frequencies between 3 and 10 Hz, the mean errors for the 274 event data set are less than 1.4°, and standard deviations are as small as ± 11.1°. The entire array also produces good results for 3 to 10 Hz, and a 9-element configuration (two inner rings) performs well at high frequencies. The five-element B-ring (600 m aperture) appears to be important in obtaining good backazimuth estimates for regional Pn waves. Of the frequency bands considered in this study, the 3 to 6 Hz, 4 to 8 Hz, and 5 to 10 Hz bands yield the most reliable backazimuth estimates, even better than an “optimal” band that calculates the azimuth in the fixed frequency band that has the largest average SNR. The time-window length has little effect on the backazimuth estimates. Other factors investigated include SNR, source region, and phase type. We found that event backazimuth accuracy degrades if the SNR of a beamed Pn signal is less than 5 dB in the frequency band of interest. Conversely, the backazimuth estimation statistics improve if only events with 5 dB of SNR are included in the event set. These data sets yield near-zero mean errors and smaller standard deviations than the entire data set. For specific source regions, standard deviations are as low as 2° for some parameter combinations, but there can also be large biases in the backazimuth estimates. Events farther than 500 km from NORESS tend to have larger azimuth errors than the closer events, but combinations of small aperture configurations and middle (3 to 10 Hz) frequency bands work well for events over the entire distance range of 40 to 1200 km. The Pn arrival and RONAPP azimuths calculated from the Lg phase have similar accuracy statistics for a 220 event common data base, implying that the two phases work equally well for regional event backazimuth estimation. In fact, averaging of the Lg and Pn estimates provides the most accurate backazimuths relative to PDE reference information.