A dedicated calibration site, as investigated in this Thesis, monitors successive passes of the altimeter at an individual location using an extensive array of space geodetic and sea level observing systems. The in situ ‘comparison point’ is located along one of the satellite altimeter ground tracks, enabling the comparison of altimeter SSH with in situ SSH for each overflight. In the case of the T/P and Jason-1 missions, this translates to one comparison every 9.9 days along a particular ascending or descending ground track. The overflight frequency is higher if the comparison point is situated at an altimeter crossover location, in which case both the ascending and descending altimeter passes will be observed within the 9.9 day period.
In its most basic form, the altimeter absolute SSH bias (BiasAlt) is defined as:
Alt Alt InSitu
Bias
=SSH
−SSH
Eqn 1-3where:
SSHAlt is the sea surface height derived from the altimeter; and
SSHIn Situ is the sea surface height derived from the in situ instrumentation.
A negative (-)ve bias is indicative of SSH being measured too low by the altimeter (i.e., the altimeter range is too long or the orbit position is too low).
This seemingly straightforward problem has three main difficulties:
1) The comparison point must be located sufficiently out in the open sea to avoid land contamination to any of the altimeter instruments (constrained primarily by the large footprint size of the water vapour radiometer). This introduces considerable logistical and operational difficulties.
2) The SSH estimates from the in situ instrumentation must be computed in a reference frame consistent to the altimeter-derived estimates to allow direct comparison.
3) The accuracy required at the in situ comparison point (a dynamic open sea environment) represents a demanding geodetic problem at the very limit of currently known techniques.
For the calibration of T/P, the NASA and CNES agencies operated dedicated absolute calibration sites at Harvest platform offshore from California (Christensen et al., 1994) and offshore from Lampedusa Island in the Mediterranean Sea (Menard et al., 1994). The Harvest platform site (Figure 1-3) illustrates the concept of the in situ calibration site. The platform is located on an ascending altimeter ground track approaching the coast of California. Sea level at the platform is observed using a series of sea level measurement systems mounted to the platform structure. Filtering techniques are used to remove the effect of wind waves and swell on the in situ estimate of SSH. The position of the platform (and hence the tide gauge) is fixed in an absolute reference frame using Global Positioning System (GPS) data from a receiver mounted to the top surface of the platform (readers not familiar with geodetic applications of GPS are referred to Herring (1999) for an in depth review of contemporary techniques). The in situ SSH estimate at the time of overflight is used to directly estimate the altimeter absolute bias (see Christensen et al., 1994 for site details). The measurement of additional variables, such as meteorological, atmospheric and sea state parameters, allows further investigation into the various error sources of the altimetric measurement system.
Figure 1-3 The Harvest platform dedicated calibration site (NASA) for T/P and Jason-1. (image courtesy B. Haines).
A number of techniques are available to measure in situ SSH at a chosen comparison point. GPS equipped buoys have been used by both the NASA and CNES sites for the direct measurement of SSH around a specific comparison point (see Born et al., 1994 and Bonnefond et al., 2003b for example). The concept, which consists of a GPS antenna and receiver mounted on a floating platform, ideally suits altimeter calibration experiments. Data from the GPS buoy allow the estimation of sea level on an epoch-by-epoch basis in an absolute reference frame, directly comparable to the altimeter. GPS buoy technology forms the focus of the improved calibration methodology presented in this Thesis, and is the subject of the following Chapter.
The buoy technique also allows the estimation of a local precise geoid at the altimeter comparison point (Key et al., 1998 and Bonnefond et al., 2003a for example). Precise knowledge of the geoid allows the extrapolation of coastal tide gauge SSH to an offshore comparison point. The extrapolated SSH estimates remain in the same reference frame as the altimeter enabling direct comparison against the altimeter SSH, and computation of the absolute SSH bias. A similar technique may be utilised in areas covered by precise regional geoid models. Studies such as Dong et al. (2002b) and Woodworth et al. (2004) have successfully integrated geoid models, such as the European Gravimetric Geoid (EGG97), with GPS equipped tide gauge data, and tidal difference modelling to estimate absolute bias in areas surrounding the United Kingdom.
In relation to individual calibration sites, the direct estimation of SSH at the comparison point using an in situ platform or a GPS buoy is accepted as the most accurate technique, as it is not limited by extrapolation methods and the use of
geoid models. A further review of results from different in situ calibration techniques in relation to the Jason-1 mission is presented in Chapter 4.
Whilst a dedicated calibration site offers the ability to solve for the absolute SSH bias, the limited number of observations (one bias estimate every 9.9 days in the case of T/P and Jason-1) precludes any statistically significant estimation of bias drift over a short period. Over the 6-month calibration phase of the Jason-1 mission for example, an individual calibration site will derive a maximum of 18 absolute SSH bias estimates. The precision (σ, at 1 standard deviation) of a single absolute bias estimate is determined by assimilating error terms from both the in situ ground system and the altimeter measurement system (see Table 1-2). Assuming independent observations, the precision of the mean absolute bias estimate will be
σ
18
. This small sample size prevents any statisticallysignificant estimate of bias drift from an individual calibration site over such a short period (especially at the 1 mm/yr level). To achieve statistically significant estimates of bias drift, either the measurement precision must be significantly reduced, or the number of observations needs to be significantly increased. Given
σ is dominated by the error terms from the altimeter, the only possible option over such a short period is to significantly increase the number of observations, in effect creating many ‘calibration sites’.
This introduces the concept of the second well established calibration technique which effectively increases the number of observations to allow the determination of statistically significant estimates of bias drift. The technique known as the “tide gauge network calibration” achieves this by sacrificing the ability to determine absolute bias, as per the previously discussed methodology. The two very different techniques complement each other in the overall calibration process.