1. Vertical velocities at tide gauges from a completely reprocessed global GPS network of stations: How well do they work? G. Wöppelmann 1, M-N. Bouin 2, Z. Altamimi 3, C.Letetrel 1, A. Santamaria 3,4, X. Collilieux 3 1 UMR LIENSS, Université de La Rochelle-CNRS, France 2 CNRM / Centre de Météo Marine, Brest, France 3 LAREG/IGN, Marne-la-Vallée, France 4 IGN, Madrid, Spain IGS Analysis Center Workshop 2008 - Miami Beach, Florida (USA), 2 - 6 June 2008 An important issue concerned with the understanding of sea level variations due to climate change is that long-term sea level records from tide gauges may be affected by vertical land motion. Fig. 1 gives some examples of vertical land motions in tide gauge records. Source: PSMSL, http://www.pol.ac.uk/psmsl/landmove.html Challenges  Rates in sea level change: ~1-2 mm/yr.  Standard errors several times smaller to be useful in climate sea level studies. Motivation for reprocessing GPS data  To use the best available data and most accurate models to reduce errors in the estimates of coordinates.  To use the same analysis strategy all over the data span to derive consistent sets of station coordinates, and to limit spurious signals in their time series. Context of the study  Total # of GPS stations: 225  IGS05 stations: 91  Time span: 1997.0 - 2006.9  205 time series > 3.5 years  160 are co-located with TG  90 CGPS@TG are not IGS Computation of realistic uncertainties on GPS velocities, according to Zhang et al. (1997): GPS data and analysis GPS-corrected sea level trends Noise characteristics Rate uncertainty Fig. 2. Distribution of the 225 GPS stations reprocessed at ULR consortium. Black circles are IGS05 stations (91); stars are CGPS@TG stations (160). (Glacial isostatic adjustment) (Co-seismic displacement) 1964 earthquake (Groundwater extraction) (Sedimentation and reclamation works) (No evidence of land motion) Table 1. Data analysis strategy Stacking of the weekly station coordinate solutions using Altamimi et al. (2002, 2007) approach:  Station positions and velocities at t 0 (2001/346), as well as time series of weekly GPS post-fit residuals and transformation parameters, and discontinuities (iterative detection procedure) Outputs This poster briefly describes the GPS reprocessing data analysis strategy that we have implemented at the ULR analysis centre. We show to what extent sea level trend estimates from long term tide gauge records benefit from our GPS-derived velocities by revisiting the analysis of Douglas and Peltier (2001). The scatter of the estimated GPS-corrected sea level trends is further reduced compared to Wöppelmann et al. (2007), either by extending the GPS data span or by changing from ITRF2000 to ITRF2005 datum. The search for Ray et al. (2008) anomalous harmonics in the spectra of non-linear position residuals indicates a clear reduction in the noise level of our GPS reprocessed time series. The noise analyses in the GPS position time series allow to compute realistic rate uncertainty estimates (Zhang et al. 1997). Those are larger under correlated noise hypothesis, though they prove still to be worth for sea level trend studies (the resulting standard errors are comparable size to those coming from the tide gauges). Summary Most long-term sea level trend analyses performed so far have applied corrections for only one of the many land motion processes that can affect tide gauges, namely the Glacial Isostatic Adjustment (GIA). We revisit here one of the most quoted studies, Douglas and Peltier (2001), which conclude in a global sea level rise of 1.84  0.35 mm/yr over the last century. Tide gauge data and analysis  Mean sea level data from the PSMSL (Woodworth and Player, 2003)  Application of Douglas (1991, 2001) selection and analysis criteria:  Tide gauge records longer than 60 years  Valid data within the records > 85%  Same grouping into oceanic regions But, we applied GPS velocities instead of GIA corrections. Working hypotheses: 1. GPS antenna vertical land motion  Tide gauge land motion 2. Land motions are linear over the tide gauge record length Table 2. Relative and absolute sea level trends from tide gauge records using different land motion corrections Fig. 6. Distribution of the 75 IGS GPS stations in common with Ray et al. (2008) Noise level and anomalous harmonics To investigate the quality of our GPS results we computed the spectra of the position time series. The idea is to examine the noise level. Is it lower than in the standard IGS results? Could it yield to a better detection of the anomalous harmonics recently discovered by Ray et al. (2008)? Input data  Non linear position residuals time series, either from IGS weekly solutions (generated in the ITRF2005 combination), or from our reprocessed GPS weekly solutions expressed in ITRF2005.  75 time series with more than 200 weekly points over the common 1997.0-2006.0 period. Spectra computation  Lomb-Scargle periodogram (Press et al. 2001).  Normalized spectra stacked and filtered following Ray et al. procedure (annual and semi-annual signals removed from each time series prior to computation of the periodogram. Fig. 5. Stacked periodograms for filtered GPS position time series in Vertical, East and North. VERTICAL EAST NORTH Wöppelmann et al. 2007 ITRF2000 1999.0 – 2005.7 ITRF2000 1997.0 – 2006.9 ITRF2005 1997.0 – 2006.9 ICE5Gv1.2 + VM4 (Peltier 2004) 1.56  2.05 1.28  1.34 1.69  1.49 1.37  1.25 1.68  1.15 RMS in sea level trends Fig. 4. Illustration of sea level trends in two regions affected by GIA Table 3. Sea-level trends summary Fig. 4. Examples of sites illustrating the issue raised by working hypothesis 1. 350m distance; monitored by precise leveling … Many geophysical phenomena can be described using a power-law process of the form (Agnew 1992): Spectral analysis shows significant lower values for k than previous studies which conclude to spectral indices close to -1 (flicker noise), e.g. Williams et al. (2003). -0.75  0.01 -0.64  0.01 -0.48  0.01 -0.58  0.01 The third line in Table 3 represents the scatter of the individual sea level trends about the mean computed by a classical RMS, whereas the fourth line follows Douglas approach by grouping the records into regions prior to the computation of the mean and its standard error. -0.71  0.01 -0.70  0.01