Station corrections for the Kaapvaal seismic network: Statistical properties and relation to lithospheric structure

Station corrections for body wave travel times are required to compensate for lateral variations in the crust and uppermost mantle in the analysis of seismic travel times that are used to determine deep Earth structure by various methods, including tomography. Station corrections to be applied to P wave arrival times from teleseismic earthquakes recorded by the Kaapvaal seismic network were estimated by five different methods: (1) averaging, (2) computing the median, and (3) weighted averaging of residuals; (4) least-squares regression, and (5) weighted least-squares regression. The corrections display variations that are related to the tectonic features of southern Africa inferred from surface geology, clearly delineating the southern and central areas of both the Kaapvaal and Zimbabwe cratons as regions of early arrivals, and the area around the Bushveld complex by later arrivals. Use of a simple ray method for generating synthetic station corrections suggests that lateral variations in the top 230 km of the Earth can explain the observed pattern of variations in station corrections. A satisfactory way of compensating for the biasing effects of outliers in the individual estimates of station corrections is through adaptation of a method originally developed by Jeffreys, which involves ascribing weights to the observations that reduce the standard deviation on a single estimate of a station correction from 0.123 to 0.096 s. Methods (2), (3) and (5) avoid serious bias by outliers, although methods (3) and (5) are preferred, because they also provide information on the causes of outliers. The presence of some outliers cannot be explained by errors in the measurement process, but must be caused by timing errors at the stations during recording, and/or errors introduced during the process of constructing the archived data files from the field data.