SNARF Ver 2.0 Construction Prof. Thomas Herring, Department of Earth Atmospheric and Planetary Sciences, MIT Cambridge, MA
Overview Analysis of GPS solutions submitted for SNARF Version 2.0 Vertical motions in the SNARF realization Application to Plate Boundary Observatory Reference Frame 1/18/2019 GPS Lec 01
Submissions to SNARF Ver 2.0 Code AC File Stat UStat Rew CBN CBN/NRCAN CBN06P30v4oc.snx 206 178 10 NAR NAREF/NRCAN NAR06P44v6uc.snx 727 504 70 PUR EC/Purdue calais_nov06.snx 681 490 10 UAF JF/UAF UAF_SNARF.stacov 135 123 1200 UNR GB/UNR frame_ITRF0--.stacov 32 32 700 Stat -- number of stations in solution. Multiple entries for stations with breaks. Ustat -- number of unique stations Rew -- Multiplier of covariance matrix to make velocity sigmas realistic. Reweighting factors were determined from the square root of chi^2-per-degree of freedom (normalized root mean square scatter, NRMS) of the differences between the velocity estimates and SNARF version 1.0 1/18/2019 GPS Lec 01
RMS difference between new solutions and SNARF Ver 1.0 Code # WRMS (mm/yr) V01 NRMS V01 N E U N E U CBN 14 0.42 0.44 1.79 1.17 1.00 0.96 NAR 69 0.66 0.55 2.41 1.14 1.11 1.22 PUR 69 0.45 0.56 1.54 0.55 0.80 0.92 UAF 10 0.25 0.24 1.67 0.32 0.47 1.09 UNR 15 0.61 0.63 1.68 1.30 1.41 1.04 The reweighting coefficients were set to make the average NRMS from the V01 comparison be approximatey unity. 1/18/2019 GPS Lec 01
RMS difference between new solutions and SNARF GIA model Code # WRMS (mm/yr) GIA NRMS GIA N E U N E U CBN 14 0.56 0.24 1.58 1.57 0.55 0.85 NAR 69 0.49 0.65 1.52 0.85 1.32 0.77 PUR 69 0.67 0.79 1.62 0.81 1.13 0.97 UAF 10 0.58 0.66 1.44 0.75 1.26 0.94 UNR 15 0.69 0.42 1.12 1.48 0.95 0.69 Notice new solutions match the GIA height rates better than SNARF Ver 1.0 1/18/2019 GPS Lec 01
Center to Center comparison Comparison between centers for velocity estimates NAR PUR UAF UNR # WRMS NRMS # WRMS NRMS # WRMS NRMS # WRMS NRMS (mm/yr) (mm/yr) (mm/yr) (mm/yr) CBN 34 0.47 0.56 25 0.59 0.53 62 0.62 0.20 2 -- -- NAR 367 0.96 0.49 39 0.79 0.54 27 0.49 0.42 PUR 33 0.71 0.50 27 0.78 0.61 UAF 3 0.64 0.61 # -- Number of common stations. The NRMS < 1 suggests re-scaling may be too large, although similar values for all comparisons suggests one over all scale factor. 1/18/2019 GPS Lec 01
Center to Center Position comparison Center # NE RMS # NEH RMS CBN.NAR 35 2.87 mm 35 8.66 mm CBN.PUR 26 4.77 mm 26 9.98 mm CBN.UAF 63 3.26 mm 63 8.81 mm NAR.PUR 367 3.99 mm 367 371.18 mm NAR.UAF 39 2.60 mm 39 14.06 mm NAR.UNR 27 1.98 mm 27 25.34 mm PUR.UAF 33 2.08 mm 33 14.29 mm PUR.UNR 27 4.65 mm 27 532.49 mm Differences show height errors up to 6 meters in one case. 47 sites have height differences greater than 10 cm. 1/18/2019 GPS Lec 01
SNARF Ver 1.0 GIA Model 1/18/2019 GPS Lec 01
SNARF Ver 2.0 Preliminary 1/18/2019 GPS Lec 01
PUR contribution 1/18/2019 GPS Lec 01
CBN+NAR Contribution 1/18/2019 GPS Lec 01
Summary Height errors in current files make rigorous combination impossible at the the moment. 47 sites have significant height errors and there could be problems at other sires. Noise models could be improved. There are large variance re-scaling factors which compensate from correlated noise at stations. Development of site specific noise models would improve the overall reliability of the final solution New model needs to be incorporated into the GIA ensemble model for the final frame alignment. Consistency of horizontal and vertical motions seems to be a problem. Need for additional non-North America sites (Pacific plate) to control the rotation of the frame should be considered. RMS differences between solutions are smaller than RMS to SNARF Ver. 1.0 and its GIA model indicating that SNARF Ver. 2.0 will be more accurate and contain many more stations than SNARF Ver. 1.0 1/18/2019 GPS Lec 01