抄録
We assessed the performance of a permanent displacement detection algorithm, which was developed by the authors at previous study, by using long-term Real Time Kinematic-GPS（RTK-GPS）data processing. False detection of permanent displacement should be avoided for a reliable warning system. We examined the detection rate using RTK-GPS time series with various baselines and different reference sites for a long time period. The false detection rate was ∼0.25% with a 4 sigma confidence level for a single baseline. This false detection rate is inadequate for practical use in a large GPS network consisting of many stations, such as GEONET. We improved the permanent displacement detection algorithm to mitigate the false detection rate. In the improved algorithm, an earthquake occurrence is defined as when the target GPS site and all its neighboring GPS sites detect a displacement. We applied the improved algorithm to an actual data set. With our improved algorithm, the false detection rate decreases to less than 0.002%. When we can use several reference sites for the RTK-GPS data processing and compare results for each reference site, the false detection rate will become almost zero. The improved algorithm is also useful for the detection of small displacements, because it may be possible to lower a threshold value from 4 to 2 sigma while keeping the false detection level to acceptable limit.
