Data Processing at the International Data Centre

Every day, around five gigabytes of raw data pour into the International Data Centre from monitoring stations around the world. Most of the data is collected and transmitted in a continuous stream, and it is the IDC's function to process, analyse and transmit this data to the Member States. States can request all the raw data, or they can chose to receive it in bulletin format. States can also request the IDC to provide data according to their own specific selection. Seismic, hydroacoustic and infrasound data are timeseries, and they go through a similar processing and analysis 'pipeline' when they arrive at the IDC. Small differences in the processing and analysis between one technology and another are generally related to the processing parameters applied, the execution of particular algorithms in some cases, and the time at which processing steps are executed within the pipeline. Within minutes of arrival at the IDC, timeseries data undergoes automated station processing. The output of this process is a list of signals detected within the timeseries from a single station, and a suite of attributes that describes each detected signal (e.g. arrival time, amplitude, frequency, direction, estimate of the travel path). The list of signal detections and attributes from all stations in the network are next input into the automated network processing system. The data undergoes three rounds of automated network processing, resulting each time in the production of a Standard Event List (SEL) at two, six and twelve hours respectively. The third SEL (SEL3) is reviewed by human analysts, who compile a Reviewed Event Bulletin (REB), currently within four to six days after the occurrence of an event. Sophisticated graphical decision support and analysis software is used to review the raw timeseries data in conjunction with SEL3, and to validate, or correct if necessary, the automated solutions. Senior analysts and scientists review the work of the first-line analysts as a means of quality assurance. Standard event screening criteria are then applied with the goal of screening out events consistent with natural or non-nuclear, manmade events.

Radionuclide processing differs in a number of details. One of these is the time schedule, as the propagation of radionuclides from the causative event to the sensors could take up to two weeks. Subsequent sample collection and measurement in the field takes up to two days for noble gas analysis, and up to three for particulate analysis. The radionuclide spectra then undergo automated station processing, which is followed by interactive analysis by human experts. Automated characterization and screening then categorizes the spectra into levels according to the number, type and relative amounts of nuclides present. The identification of anomalous levels of particular radionuclides triggers an atmospheric modeling operation to track the origin of the sample. This modeling is based on retrieved meteorological data and makes fusion of data from all four technologies possible. Standard Screened Radionuclide Event Bulletins (SSREBs) are produced when an anomalous radionuclide level is detected. On average, Member States receive close to 21,000 segmented data and product deliveries per month from the IDC. The system continues to be developed and refined so that the IDC can meet demands for an increasing volume of data and range of IDC services.