In time series (Fig.8.) it can be seen that HSMN, TSFT, and RDCT, three stations with longest baseline, shows more random noise than other stations in the horizontal direction and this can be explained that the accuracy of double difference method is mainly affected by the baseline, the longer base line the less accuracy, while this also can explain the time series of NAD83 shows more random noise than the result from Topcon Tool since the baseline is generally about kilometers in OPUS process.
  However the GPS observation in vertical (UD) direction (Fig. 7 and 8.) generally demonstrates more random noise signal than the horizontal one, which indicates the resolution of the vertical observation is more sensitive to the length of the baseline and thus needs the shorter baseline to attain the same resolution as the horizontal one.

     The GPS result show the horizontal displacement rate in this area is about 2 mm/yr and the subsidence is about 5 mm/yr. While compare the result from different reference frame (Fig.12, 13, 14 and 15.) the effect of the reference frame can be significant in this case and thus a more stable local reference is critical for precisely dileanting local ground deformation over time and space. 

    The groundwater level in this area has been rising but still below the preconsolidation head and thus subsudence is still occuring. Shallower real-time groundwater data will be used to filter the GPS sugnal in this area in order to examine how the faulting and subsidence interact.