Besides the stability of the on-board clocks and the precision of the Doppler measurements, the tracking network is a key factor in the success of the DORIS system. This network was deployed by the ‘Institut géographique national’ and is still maintained by its SIMB group (‘Service d'installation et de maintenance des balises’) [17,54]. The DORIS network was operational before the launch of the first DORIS satellite and has maintained since then a dense and geographically good distribution of tracking stations. Fig. 1 shows the current status of the DORIS tracking network as of January 1, 2005.

A significant number of sites (40) have continuously observed for more than ten years. These sites are also geographically well distributed all over the Earth. Only 15 sites have observed between five to ten years, while 16 stations have been installed on new sites during the last five years. Fig. 2 represents a histogram of the distribution of the observations data span per station, showing how stable this network has been with time. Fig. 2 shows in more detail how stable the DORIS network has been since 1993.

When selecting a new DORIS site and especially at the very beginning of the system, the availability of other space techniques to obtain precise a priori geodetic coordinates and velocities for the site was an important criterion. It is also a very important criterion for Terrestrial Reference Frame maintenance [1]. Fig. 3 provides the current geographical distribution of the geodetic collocations between DORIS and other space geodetic systems (VLBI, SLR, GPS). It can be seen that the DORIS–GPS collocations are the more numerous by far and are very important in the realization of the Terrestrial Reference Frame (TRF) [1,54]. The VLBI collocations are mostly in North America, for economic reasons. Several VLBI and SLR collocations were realized only temporarily through campaigns of mobile equipments and are no longer active collocations.

The local surveys between DORIS were performed or at least collected by the IGN using strict specifications. Table 1 displays a summary of the estimated precisions of these local tie surveys. A large number of them have been done with a sufficient care to be used for global geodetic techniques combinations or intercomparisons. It must also be noted that the DORIS stations are often located on remote islands, sometimes with only a few ships traveling there per year. This of course complicates the task of the geodetic survey teams. The few collocated sites for which local tie could not be performed or was performed with lesser accuracy correspond mostly to mobile VLBI or SLR stations that are no longer in existence.

DORIS is a globally distributed, all-weather system providing nearly continuous precise Doppler tracking of SPOT2, TOPEX/Poseidon, SPOT4, SPOT5, ENVISAT and Jason-1 satellites [34]. Precise Orbit Ephemeris (POE) orbits computed either by the Precise Orbit Determination and Production System (PODPS) at GSFC, or by the POD team at CNES, with different orbit determination software, have a radial accuracy of 2–3 cm for TOPEX/Poseidon, Jason-1 and ENVISAT, based on a combination of DORIS and Satellite Laser Ranging (SLR) tracking.

Orbits computed using only DORIS compare to the POEs to within 2 cm on the radial component [36].

The dense DORIS tracking allows a reduced-dynamic approach for POD [59]. In a dynamic approach, all forces applying on the satellite are supposed to be known or at least a limited number of additional parameters (e.g., atmospheric drag or solar pressure coefficients) are estimated during the orbit determination process, but the satellite position and velocity at the epoch of reference (start of the considered arc of data) are considered to be constant. In a reduced-dynamic approach, the satellite accelerations are estimated more frequently to give more freedom to the dynamic model itself, but with appropriate constraints [59]. The reduced-dynamic approach requires more data and also a continuous tracking is more appropriate, as more flexibility is introduced in the estimation process. However, when the data are of excellent quality, a reduced-dynamic is less affected by error models and the orbit will be more accurate. In the case of TOPEX, orbit overlap consistency and SLR data fits suggest that the preliminary ‘enhanced parameter’ POE offers an improvement over the ‘dynamic’ POE orbits. However, evaluation of the 2–3-cm orbits is also quite a challenge and requires new types of improved tests. Recently, such tests were done by comparing the results of different oceanographic missions, enabling the detection of previously unnoticed tiny geographically correlated errors.

DORIS offers a powerful satellite tracking capability, whose full potential for POD has yet to be realized. For example, at CNES, the DORIS-modeled observation is weakened by the current POD strategy of unconstrained adjustment of pass-by-pass measurement and troposphere biases. The application of crude constraints dramatically increases the level of signal in the DORIS residuals, and in the case of the troposphere, improves the orbit [48] as well as the geodetic derived results [54]. The investigation conducted at CNES, aiming at strengthening the modeled observation, continues to try to improve operational POD accuracy.

On a more general point of view, POD results could be improved by adopting better models or by using a more refined data analysis strategy. The gravity field models are now much more accurate since the use of the GRACE data, and improvements are foreseen in the surface force models (atmospheric drag, solar pressure, albedo) as well as in the measurement model (use of quaternions for the spacecraft attitude, better phase centre determination and monitoring [56]). In particular, better force models could improve the centering of the derived orbits [9], which is of critical importance for oceanographic studies. It could also improve geocenter monitoring derived from time series of station coordinates as discussed later. Other aspects still under discussion are proper weighting between tracking techniques [9] and parameter time constraints [55].

The mission objective for TOPEX/Poseidon and ENVISAT has been largely fulfilled with 2–3-cm POE orbits routinely produced. Further improvements are under study in order to take greater advantage of the dense tracking DORIS has to offer, possibly leading to more accurate POEs. For example, there is presently some discussions between POD groups using multi-technique analysis (DORIS/SLR/GPS) as they all need to estimate an additional parameter in their estimation that is equivalent to a DORIS time tagging offset, whose effect can be seen in the along-track component of the satellite if not taken into account (Zelensky, personal communication).

Table 2 summarizes the different POD results obtained for DORIS satellites. Major improvements were done after the adoption of a new generation of gravity fields [5,42] based on data from the GRAvity and Climate Experiment (GRACE) mission [43].

4.1 General purposes

4.2 The SPOT-4 mission

4.3 The Jason-1 mission (SPOT-5 and ENVISAT)

4.4 Future missions (CRYOSAT, Pléiades and Jason-2)

As usual, the future DIODE flight version has first been validated on-ground. To do so, we use real DORIS measurements gathered from the different satellites carrying the DORIS tracking system. Below are displayed the results for a TOPEX twelve-day arc (September 2004) (Fig. 4).

4.5 Conclusion

Instead of using the stations coordinates as known values in the POD estimations, it is also possible to use the DORIS data for geodetic purposes [8,11,41,48] and to estimate the positions (and eventually the velocities) of the DORIS tracking stations. In such a case, the satellites orbits and the station positions are estimated simultaneously in a combined free-network adjustment [24].

The precision of the geodetic results depends on the accuracy of the different model used (gravity field, solar pressure, Earth rotation...), but also on the amount of DORIS data available for this specific station.

Fig. 5a–c describes the sensitivity of the DORIS geodetic accuracy to the considered observation period in latitude, longitude and height. We have processed the DORIS data on daily basis during several months and we have computed for each day the position of all stations of the DORIS network using all data from the first day of this test to the considered day. We have then compared for each individual realization (from 1 day to 5 months) the derived geodetic reference frame (coordinates of all stations) to a better reference expressed at the mean epoch of observation (IGN04D02) [51].

Fig. 5a–c shows that the precision of the results is directly linked to the duration of the DORIS observations. One-day solutions provide 20–30 mm precision, while 7-day solution as currently provided to IDS provide 10 mm or better accuracy, making the DORIS technique very competitive with other techniques such as GPS. The observed degradation in Fig. 5 after several months is due to the fact that we do not take into account tectonic plate motions in this simple test. It must be noted that this recent improvement on precision was obtained when using the recent GRACE-derived gravity fields [43,50].

Another aspect related to the quantity of DORIS data per station during a fixed period of observation is the number of available satellites during the period of observation. Since 1993, most of the geodetic results were obtained using a limited constellation (2 satellites). Since June 2002, five satellites are now available for geodetic positioning (SPOT-2, TOPEX, SPOT-4, SPOT-5 and ENVISAT). Unfortunately, the Jason data cannot be used for geodesy due to an increased sensitivity to radiation of the oscillator over the South Atlantic Anomaly (SAA) [57,58].

In the following test, we have used the data for a specific week (February 1–7, 2004) but similar results can be obtained for any other week of DORIS data. As data from five different satellites were available for this specific week, we processed the data using either one satellite (5 choices = 1 per satellite), two satellites (10 choices), three satellites (10 choices), four satellites (10 choices) or five satellites (only 1 choice = all available satellites). For each individual solution, we have compared the position of all DORIS stations with our internal reference (IGN04D02) as before and derived a unique Weighted Root Mean Square (WRMS) value for this specific week.

Fig. 6 shows that the precision of the geodetic DORIS results improves with the size of the DORIS constellation: five satellites provide better results than one, even if the improvement is not linear and even if, for some specific cases, solution with two satellites (usually two SPOTs) would provide better results than solution with three satellites (usually including TOPEX). In general, SPOTs (98° inclination) provide better geodetic results than TOPEX (66° inclination) for stations at very high latitude because they generate more data for these stations during one week. Station positions at high latitude are also poorly determined by inclined satellites, due to the specific geometry of the passes, making it difficult to distinguish between all position components. It can also be seen that individual satellites provide different levels of geodetic precision: one satellite can provide for this specific week station coordinate accuracies ranging from 14 to 36 mm depending on the satellite orbit (better results for SPOTs and worse for TOPEX for reasons explained above). The best single-satellite solutions are obtained with SPOT-5 (equipped with a newer generation on-board receiver). Testing on a few other weeks indicates that, even though the results themselves slightly change from week to week, the general conclusions discussed above remain the same.

Another way to look at the impact of the size of the DORIS constellation on the accuracy of the derived station coordinates is to consider all the weekly results available for the full considered DORIS data set (since January 1993) and to classify the results according to the total number of satellites available per week. Table 5 provides a detailed summary of these weekly comparisons to ITRF2000 [1] or IGN04D02 reference after grouping the results with regards to the size of the DORIS constellation. In both cases, a clear improvement can be seen, with the increase in satellite availability, confirming the previous results (Fig. 6), but on a much longer time span (12 years vs. 1 week). Comparisons with IGN04D02 perform better as this reference is directly based on those same DORIS results. ITRF2000 may also degrade with time, due to normal imprecision in the velocities (positions in 2005 are extrapolated beyond the last data in 2000) [51].

In the past few years, the whole concept of maintenance of the Terrestrial Reference Frame (TRF) has slowly changed. Previously, the TRF was defined as a set of station positions and velocities for a limited number (less than a few hundred) of fundamental geodetic stations all over the Earth [2,33]. The so-called maintenance of the frame was done by computing these solutions typically on a yearly basis to take into account the increase in data availability as well as the improvements in accuracy regularly obtained by the numerous Analysis Centers for the different techniques. More recently, instead of realizing the reference frame through a set of positions and velocities of a large number of stations (cumulative solution such as IGN04D02, for example), the notion is now evolving into time series of station coordinates [2]. The concept of the origin of the frame is also going though intense discussions and very precise definitions must now be provided [6,7,12,15,54]. DORIS provides valuable and independent information for all these investigations.

DORIS can also be used to estimate the Earth Orientation Parameters (EOPs), and more specifically, the Earth's Polar Motion. As for the other geodetic results, we will show that the number of available DORIS satellites is critical to the accuracy of the results.

We have conducted for these parameters the same study as previously done for the station coordinates. In a first step, we have selected a specific week in which data from 5 satellites are available. We then estimate these parameters using only 1 to 5 satellites and looking at all the possible choices as before. In this case, we compared the DORIS X-POLE (resp. Y-POLE) with the values derived independently from a GPS analysis (JPL solution [49]). For each possible choice, we have estimated a Weighted Root Mean Squares (WRMS) of the observed differences (daily DORIS–GPS) over the seven days of the week without removing any bias or drift. We implicitly assume that the DORIS and GPS-derived EOP time series are expressed in the same Terrestrial Reference Frame.

Fig. 7 shows that the accuracy of the DORIS-derived polar motion strongly depends on the availability of DORIS satellites. The point on Fig. 7b (one satellite) that looks like an outlier corresponds to a limited satellite dataset (50% of the data were lost for this specific week for this satellite).

In a second step, we have considered all weekly results of this type, taking into account for each week all available satellites (since 1993). We have then grouped the results not by time but by the number of available satellites available during each week. Table 6 shows a direct comparison of the DORIS/IGN-JPL polar motion time series with the GPS/JPL time series (no bias or drift removed) using two different reference frames ITRF2000 and IGN04D02 to express the results.

Table 6 confirms the previous results obtained on a limited dataset (Fig. 7). Sub 1 milli-arcsecond (mas) is presently achievable with the current DORIS system when a sufficient number of satellites is available. These results represent quite an improvement on previous submissions (2 mas estimated over the full 1993–2004 period [21], as the author did not consider possible improvement during time due to the successive changes in the DORIS constellation). In Table 6, very little difference can be found with the choice of the reference frame, showing that larger errors can be found presently in the DORIS data processing and cannot be attributed to the frame.

It is also possible to derive from the DORIS data information on the atmosphere, both for the lower part (troposphere) and for its upper part (ionosphere). Concerning the troposphere, early studies [53] have shown that precision of 1 cm in zenith path delay can be obtained using three satellites (equivalent to 2 mm in Integrated Precipitated Water). More recent studies show that 6 to 8 mm can be obtained when using five satellites during the CONT02 campaign in comparison with VLBI and GPS results [39,40]. However, we must also say that these results can only be obtained after a one-month delay, making them totally useless for any assimilation in a weather prediction model. However, for climatology studies, DORIS could play a role as it provides an external reference for long time periods. DORIS is also less subject than GPS to antenna changes or equipment modifications such as radome changes.

It is also possible to use the data from the two DORIS frequencies (2 GHz and 400 MHz) to derive information on the ionosphere and more specifically on the Total Electron Content (TEC) in the direction of sight of the satellite. This can be used to derive ionospheric corrections in the case of a single-frequency altimeter like Poseidon on the TOPEX satellite [19]. It can also be used to derive information on the protonosphere (over 1000 km) as the DORIS satellites are lower (800 to 1300 km) than the GPS satellites (20 000 km) and they do not probe the totality of the ionosphere [47]. It is also possible to study ionospheric perturbations related to geophysical activities, such as earthquakes [46].

Since the launch of SPOT-2, the DORIS system has been used for Orbit Determination and for Geodesy or Geophysics. For Precise Orbit Determination, several important steps were taken in the last few years, specially using GRACE-derived gravity field, in order to obtain a 2-cm orbit on the radial component in post-processing mode or sub-centimeter when combined with GPS and Laser Ranging. Real-time orbits are currently derived with 20-cm precision for the radial component and are extremely reliable. This study shows that 5-cm level of performance could be obtained with Jason-2 with proper software upgrades. For geodesy and geodynamics, the precision of the station coordinates depends on the number of available DORIS satellites. Best results are currently obtained using 5 satellites and show that sub centimeter precision for weekly coordinates are now at reach, potentially corresponding to 1–2 mm yr⁻¹ precision in station velocity determination since 1993. Polar motion estimations can now be obtained daily with sub milli-arcsecond precision. With its dense and permanent tracking network, DORIS should play a significant role in the realization of the Global Geodetic Observing System.