Frequently Answered Question

Question

Evaluation performance of national stations using PQLX

We are working on National Earthquake Data Center project where all the seismological data and earthquake information have been stored, archived and published in real time to the researchers in Turkey. Now, we would like to evaluate the performance of seismic stations operated by us from 2005 to now. To perform this task, we are able to use PQLX software and draw PDFs for each station. However, we have a problem on the graph. We hope that you might help us to reach a final solution.

We have selected a pilot station from IRIS and obtained PDFs. Then, perform the evaluation seismic station performance using the same time range, station and channel. Finally, compare the PDFs obtained from IRIS and evaluated ones by our PQLX server. We realized that there are small differences especially for out of the HLNM boundaries even the graphs are mostly the same. As far as I know, this is because of out-of-nominal noise conditions. How can we reduce this and make the graphs wealthy as much as ones available in IRIS using our PQLX server? Is there any available open source tool to clean these kinds of noise from waveform data file that you can suggest us?

Answer

The waveform data that we have here at IRIS for your example KO.ELL..BHE are the same as the data that you have. We don’t clean noise from the data in any way. As a test, I ran the data through PQLX and produced a plot very similar to the one in your attachment. It looks like the difference between the QUACK and PQLX PDF plots is related to the gaps in the data. While QUACK and PQLX use largely the same methodology in generating PDFs, there is apparently a difference in their algorithms when calculating PSDs on data that has gaps. I believe that in some instances PQLX will pad the data gaps with zeros while QUACK does not. However, I have not verified this with Richard Boaz who wrote the PQLX code.



Updated: 05/17/2017