Preliminary test result of PRESTo application to the southern Korean peninsula

KMA(Korea Meteorological Agency) and KIGAM(Korea Institute of Geoscience and Mineral Resources) have started a project to construct EEWS(Earthquake Early Warning System) from 2007 in South Korea. KIGAM has been operating ElarmS(Earthquake Alarms Systems) developed by UC. Berkeley Seismological Lab. in the real time mode from the middle of 2010. PRESTo (PRobabilistic and Evolutionary early warning SysTem) is a software platform coded by RISSC-lab in Italy for regional earthquake early warning. In comparison to ElarmS, PRESTo has the different approach in estimating parameters of an earthquake by adapting probability function. We conducted online and offline tests to evaluate feasibility of PRESTo about seismic network of the southern part of the Korean Peninsula. The 3D velocity model grids of P- and S-wave were calculated from 1D velocity model used by KIGAM for routine work. Two kinds of magnitude equations, High Mag to events with magnitude over 4.0 and Low Mag to others, were used without any change of the parameters applied to Naples in Italy in order to investigate patterns of the difference between KIGAM's local magnitudes and PRESTo's. Off and online test of PRESTo was done at two PCs with Windows7, Intel Core i7-2 3.4 GHz CPU and 8 GB memory. Offline simulation was applied to the total 162 earthquakes occurred at inland or offshore area of the southern Korean peninsula with magnitude over 2.0 from January 2007 to May 2012. As a result of offline test, the number of events by PRESTo was 132 that were about 80.5 % of the total earthquakes. The 91 % of the detected events had reasonable resolution with origin time error within 15 seconds and location error within 10 km. The 106 events, 80 % of the processed ones, had good resolution with error less than 5 km. The events with magnitude less than 4.0 showed the smaller pattern as amount of 0.3 ~ 0.8 than the magnitude from KIGAM's bulletin. The large magnitude over 4.0, however, showed similar pattern to that of bulletin. Due to the limitation of computer resources which PRESTo requires in online processing, we separated the study area into 4 zones for online test. The KIGAM's bulletin with magnitude over 2.0 had 13 earthquakes from May to June 2012. PRESTo detected 12 events among them in the real time processing mode. The location accuracy of these events showed slightly higher resolution than the results of ElramS. PRESTo used small velocity grid because we had separated study area as 4 parts. ElarmS, however, observed events over all areas. PRESTos of four areas made different results about the same events and they triggered many false events when noisy stations were included to the localized network. As a result of test operation of PRESTo, the magnitudes showed smaller pattern than that of KIGAM's bulletin in the size of about 0.3 ~ 0.8. The parameter tuning is required to determine adequate size of magnitude in regarding the seismic characteristics in the Korean peninsula. In the current system of PRESTo, it is possible to get detailed analysis in the local area, however, it gives larger burden on computing system as increasing the size of area and stations. It is necessary to split one package into several modules in order to reduce system resource and control large number of data channels the future.