Bias in reported seismic arrival times deduced from the ISC Bulletin

The accurate timing of seismological data is crucial for most quantitative examinations in seismology. We present evidence that traveltime data from many stations contain systematic variations in timing which can be identified by checking the median of station delay times as a function of time. This function is expected to be constant but many deviations are found. Several hundred stations that report arrival times to the ISC have been examined. The median station delay times of almost 8 per cent of these stations show changes of more than 1 s and thus exceed the structural signal in the data. Temporal variations of,0.5-1 s are common. Changes in the distribution of observed earthquakes and other possible explanations of such variations have been tested and fail to explain most of the observations. Therefore, the bulk of the observed changes must be caused by flaws in the timing of the data or by biased picking of arrival times. For instance, at one station with a strong annual variation of noise level, the arrival times are on average picked several tenths of a second later during months with a high noise level. Because of their systematic nature, these errors will not necessarily cancel out by using the large number of traveltimes in the ISC Bulletin and may therefore introduce a bias in many investigations. If the observed timing variations are due to the recording equipment at the stations, the errors will be present in the digital waveform data as well. Tomographic studies could potentially be affected, but in particular studies of temporal variations of Earth structure based on traveltime data, e.g. inner core rotation, need to be looked at with caution as results might be influenced by station effects. The exact nature of the bias is study-dependent and needs thorough investigation in each individual case.