| Popis: |
The operational Atmospheric Transport Modelling (ATM) system deployed and used at CTBTO produces source receptor sensitivity (SRS) fields, which specify the location of the air masses prior to their arrival at any radionuclide station of the International Monitoring System (IMS) network. The ATM computations support the radionuclide technology by providing a link between radionuclide detections and the regions of their possible source. If an IMS station detects an elevated level of radionuclide, the ATM in a backward mode is used to identify the origin of air masses. In the case of a single detection, the FOR (Field of Regard) is computed, which denotes the possible source region for a material detected within one single sample. On some occasions, multiple detections occur at one or more IMS stations. Depending on the nature of these detections and on prevailing meteorological conditions, it is possible that all these detections may come from a unique source. For this case, the PSR (Possible Source Region) is computed for each grid point in space and time by calculating the correlation coefficients between the measured and simulated activity concentration values (SRS fields). Obviously, the result will depend on the algorithms used for that purpose. Currently, in the WEB-connected GRAPhics Engine (WEB-GRAPE) software, designed and developed by the International Data Centre (IDC) to visualize and post-process of the ATM results, three different PSR algorithms are implemented: two based on the Pearson’s correlation coefficient and one based on the Spearman’s rank correlation coefficient. For the quality assessment of these PSR algorithms, subsets of datasets developed in the framework of the 2nd and 3rd ATM Challenge will be used, which satisfy the condition that the agreement between Xe-133 measured and simulated values is very good. In this sense, the selected samples will represent “ground truth” data, where the contribution from all dominated sources (e.g. Isotope Production Facilities or Nuclear Power Plants) is included. For these selected samples, the results produced by the different PSRs algorithms will be assessed, taking into account both spatial and temporal variations. |
