Comparison between geodetic and oceanographic approaches to estimate mean dynamic topography for vertical datum unification: evaluation at Australian tide gauges

Abstract

The direct method of vertical datum unification requires estimates of the ocean’s mean dynamic topography (MDT) at tide gauges, which can be sourced from either geodetic or oceanographic approaches. To assess the suitability of different types of MDT for this purpose, we evaluate 13 physics-based numerical ocean models and six MDTs computed from observed geodetic and/or ocean data at 32 tide gauges around the Australian coast. We focus on the viability of numerical ocean models for vertical datum unification, classifying the 13 ocean models used as either independent (do not contain assimilated geodetic data) or non-independent (do contain assimilated geodetic data). We find that the independent and non-independent ocean models deliver similar results. Maximum differences among ocean models and geodetic MDTs reach >150 mm at several Australian tide gauges and are considered anomalous at the 99% confidence level. These differences appear to be of geodetic origin, but without additional independent information, or formal error estimates for each model, some of these errors remain inseparable. Our results imply that some ocean models have standard deviations of differences with other MDTs (using geodetic and/or ocean observations) at Australian tide gauges, and with levelling between some Australian tide gauges, of ${\sim }\pm 50\,\hbox {mm}$. This indicates that they should be considered as an alternative to geodetic MDTs for the direct unification of vertical datums. They can also be used as diagnostics for errors in geodetic MDT in coastal zones, but the inseparability problem remains, where the error cannot be discriminated between the geoid model or altimeter-derived mean sea surface.

Appendices

Appendix 1: Sensitivity analysis of MSL epochs

Using the 2003–2007 MSL epoch, three out of 32 tide gauge sea level records used have a 1-year gap over the 5-year period, but comparison of MSL values from adjacent tide gauges indicated that the missing year would cause no more than ${\sim }10\,\hbox {mm}$ difference to the 5-year MSL at these sites. Another consideration is that this relatively short epoch may not be representative of long-term MDT in this area, especially in the northern Australian seas where large seasonal and interannual differences in MSL occur (Ridgway and Godfrey 2015; Condie 2011). In addition, the interannual signal from ENSO (El Niño–Southern Oscillation) contributes to sea level variations (a decrease during El Niño years) at the 50–100 mm level at Australian coasts (Pariwono et al. 1986).

To test the sensitivity of tide gauge geodetic MDT to shorter time periods (cf. Coleman et al. 1979), MSL was computed for different 5-year epochs; 1993–1997, 1998–2002, 2003–2007, and 2007–2011 (the latter with an overlapping year due to the data ending in 2011), and for the 19-year-long period 1993–2011. The 1993–2011 epoch was subtracted, tide gauge by tide gauge, from all 5-year MSL values (Fig. 11; Table 7). The 1993–1997 epoch shows MSL for this period below the 19-year average with the mean difference −45 mm, and largest magnitude of −93 mm for tide gauges #20 (WYND) and #22(KARU). There is no data for #15 EXMO as this tide gauge record starts in 1998. During the 1998–2002 epoch, MSL is above the 19-year average, reaching $+$ 61 mm, but mostly around 10 mm to 20 mm. The 2007–2011 epoch is shown to be as much as $+$71 mm higher than the 1993–2011 MSL, with a mean difference of 40 mm. The largest differences occur across the northwest of Australia and Gulf of Carpentaria for all 5-year epochs (cf. Amin 1993; Forbes and Church 1983; Ridgway and Godfrey 2015; Tregoning et al. 2008). On the other hand, the south eastern corner of Australia indicated differences of only ${\sim }20\,\hbox {mm}$ for all 5-year epochs compared to the 1993–2011 MSL.

In contrast, the 2003–2007 tide gauge MSL epoch does not differ by more than ${\sim }20\,\hbox {mm}$ for most tide gauges (mean difference − 13 mm), suggesting it is more representative of MSL over the 19 years covering 1993–2011. It is ${\sim }-30~\hbox {mm}$ at tide gauges #14 (CARN), #16 (ONSL), and #22 (KARU), reaching ${\sim }-\,50~\hbox {mm}$ at tide gauge #15 (EXMO), indicative of the seasonal and interannual variations of MSL in this region. It shows agreement with 1993–2011 in the Gulf of Carpentaria and Cape York. The larger difference at EXMO should be treated cautiously because the 1993–2011 average may be biased at this tide gauge because it is missing data from 1993 to 1997.

Appendix 2: Tide gauge and GNSS information

See Table 8.