The SAR (Synthetic Aperture Radar) Copernicus Sentinel-1 satellites require a high orbit accuracy of 5 cm in 3D in comparison to external processing facilities. The official orbit products delivered by the Copernicus POD (Precise Orbit Determination) Service fulfil this requirement. Nevertheless, analyses have shown discrepancies in the orbit results for the two satellites Sentinel-1A and Sentinel-1B. Since the satellites are identical in construction estimated orbit parameters like the scale factor for the radiation pressure are expected to be at the same magnitude, which is not the case. Estimation of GPS antenna offsets leads to differences between the two satellites, which might explain the discrepancies in the estimated orbit parameters. Such offset estimations are, however, very sensitive to orbit and observation modelling. It has to be assured that the results are not biased by insufficient models. First of all, stabilisation of the antenna offset estimation is achieved by improving the observation modelling by applying single receiver ambiguity resolution. The Copernicus Sentinel-1 satellites have a very complex shape with the long SAR antenna and the two large solar arrays. Antenna offset estimation based on different satellite models may give results which differ by up to 1.5 cm. The dispersion of the estimates is quite large depending also on eclipse and non-eclipse periods. Consideration of simple assumptions on satellite self-shadowing effects improves the satellite model and also the results of the antenna offset estimation. Finally, more consistent results for the two Sentinel-1 satellites are achieved by applying the antenna offset estimates.

The European Copernicus Programme

The Synthetic Aperture Radar (SAR) satellite Sentinel-1A

Figure

Artist's impression of the S-1 satellite; © ESA.

GPS is the only POD observation technique available for Sentinel-1. Assessment of the absolute orbit accuracy is, therefore, difficult. Radar measurements have already successfully been applied for orbit validation of TerraSAR-X and TanDEM-X

S-1A and S-1B satellites are identical in construction. Therefore, estimated orbit parameters allow to indirectly validate the orbit solutions by comparing the estimated parameters such as solar radiation pressure (SRP) coefficient or empirical cycle-per-revolution (CPR) parameters with each other. They are expected not to be exactly the same but they should have the same magnitude.

If the same orbit and observation models are applied for precise orbit determination of the two satellites different results for the estimated orbit parameters might be caused by erroneous satellite geometry, namely center-of-mass (COM) coordinates, GPS ARP coordinates or GPS antenna PCO

In the study presented here antenna offset estimation means the estimation of the vector from COM of the satellite to the ARP, in most cases the physical mounting point of the GPS antenna. Although a potential error cannot be assigned either to the COM or the ARP or PCO/PCV, the antenna PCO and PCVs are handled separately and are not included into this term in this case.

A main topic of this research is the sensitivity of such antenna offset estimations to orbit and observation modelling for the special case of the Sentinel-1 mission. The two satellites A and B are flying in the same sun-synchronous dawn-dusk orbit (inclination 98.18

Sun angle above the orbital plane (

The solar panels are fixed to an angle of 30

Considering the complex shape of the Sentinel-1 satellites, the mission and orbit design, the dependency of the offset estimates on the complexity of the satellite macromodel and different orbit parametrizations is analysed. Sophisticated satellite models based on ray tracing considering shadowing effects and multiple reflections proved to be beneficial to satellite macromodels

In addition, the improvement in observation modelling, namely the application of single-receiver ambiguity resolution for the Sentinel satellites

The sensitivity analysis of the S-1 antenna offset estimation is composed as follows. Section

Part of motivation for this study is that the estimated orbit parameters (e.g., SRP coefficient and empirical CPR parameters) of the two Sentinel-1 satellites show different magnitudes and or systematic signatures, which cannot be explained by other means than different antenna offset vectors of the GPS antennas. The satellites are identical in construction, but of course there are differences in the satellite mass due to the different age of the satellites. The satellite masses change over time, which is documented in the corresponding mass history files available from S-1 FOS (Flight Operations Segment). For instance the satellite masses have been 2145.057 and 2153.622 kg on 11 June 2018 for S-1A and S-1B, respectively. The differences of about 8.5 kg is not very large compared to the absolute mass of the satellites. The satellite mass is used in the non-gravitational force models as scaling factor to model the resulting acceleration. Therefore, the mass is already properly incorporated in the modelling and the SRP coefficient and atmospheric drag scaling factor should not show large differences due to this. Significant differences are, however, present in the center of mass (COM) coordinates of the two satellites. Table

GPS ARP coordinates for main and redundant antennas (top); COM coordinates (middle) and vectors from COM to the main GPS ARPs (bottom) for the two satellites as of 11 June 2018; all values are given in SRF.

Summary of models and parameters employed for the Sentinel-1 orbit determination.

Table

Estimated SRP coefficient for both Sentinel-1 satellites.

The estimated orbit parameters from S-1A and S-1B are analysed based on orbit determination results from one year of data (2018).
At first, Fig.

Estimated CPR parameters for both Sentinel-1 satellites.

Figure

To check if the differences in the estimated orbit parameters are due to different antenna offset vectors, the estimation of the

The processing of the data used for this study is based on the model and parameter settings given in Table

Solution list for antenna offset estimation.

Common for all solutions is that no PCVs are used, because it shall be avoided to introduce any possible implicit offsets from the PCVs (see

Sentinel-1 macromodel, instantaneous re-radiation

Modelling of accelerations due to atmospheric drag, SRP and ERP is done based on a macromodel of the Sentinel-1 satellite. The macromodel consists of eight planes. The surface area and normal vector in SRF of each panel are given along with the visual (vis) and infrared (IR) optical properties in Table

Schematic drawing of shadowing assumptions of solar array on backside of SAR antenna (

Azimuth angle of the sun in SRF (

Estimated

Estimated

List of shadowing conditions for Sentinel-1.

A simple box(-wing) model cannot give full consideration to the complex shape of Sentinel-1, in particular not to shadowing effects, e.g., by one of the solar arrays to the backside of the SAR antenna. Therefore, an update of the macromodel is developed to consider part of the shadowing effects. The simple shadowing assumptions have proven to be beneficial for the SRP modelling especially during the eclipse period of the Sentinel-1 satellite

Figure

Figure

Table

The results of the antenna offset estimations from the different solution types listed in Table

Estimated

Estimated

At first, the impact of the carrier-phase ambiguity-fixing on the offset estimation is shown based on the comparison of the offset estimates of solutions CFL and C (Figs.

Figures

The SRP coefficient is, therefore, fixed to 1.0 for all other solutions. The antenna offset estimates for solutions B and C are very similar for the individual satellites. The estimates differ in the sub-mm range. The standard deviations of the

Estimated

Estimated

The antenna offset estimates of solutions C–F are shown in Figs.

Mean and standard deviation (cm) of

Corrected displacement vectors between main GPS ARP and COM for S-1A and S-1B (based on solution F).

Estimated SRP coefficient for both Sentinel-1 satellites, original and F CDV solution.

Estimated CPR parameters for both Sentinel-1 satellites, original and F CDV solution.

Solution F shows equivalent results as solution C for the

When applying the estimates of solution F to the a priori antenna offset values from Table

To conclude the analysis the orbit determination (as shown in Sect.

The corrected displacement vectors between COM and GPS ARP lead to consistent estimated orbit parameters of S-1A and S-1B, which is expected for the two identically constructed satellites.

The Copernicus Sentinel-1 mission consists of two identically constructed satellites. Nevertheless, different orbit parameter estimates are present in the precise orbit determination results making an antenna offset estimation necessary. Due to the sun-synchronous orbit and the complex satellite design, the large solar arrays mounted on the two sides of the satellite bus and the long SAR antenna partly shaded by one of the solar panels, the sensitivity of the antenna offset estimation to orbit modelling has been studied. Several solutions with different macromodels for the satellite are performed. Additionally, the observation modelling was updated by applying single-receiver ambiguity resolution.

All the different solutions for antenna offset estimation show the sensitivity of the estimates on the orbit modelling and on the observation modelling. Introducing carrier phase ambiguity fixing significantly improves the antenna offset estimates due to the stiffer solution resulting in less noisy estimates. In particular the

The first main conclusion of this study is the necessity of performing carrier phase ambiguity resolution to stabilise in particular the cross-track orbit component. The second main conclusion is that the orbit modelling, for Sentinel-1 in particular the SRP modelling, is of utmost importance to get reliable antenna offset estimates. Finally, resulting antenna offset estimates for S-1A and S-1B lead to consistent displacement vectors from COM to the main GPS ARP of the individual satellites and to consistent estimated orbit parameters in the orbit determination results.

The Sentinel-1 GNSS data are available on the Copernicus Open Access Hub (

HP designed the study, carried out the data processing, interpreted the results and wrote the manuscript. JF contributed to discussions and to the final manuscript. PF contributed to the final manuscript.

The authors declare that they have no conflict of interest.

This article is part of the special issue “European Geosciences Union General Assembly 2019, EGU Geodesy Division Sessions G1.1, G2.4, G2.6, G3.1, G4.4, and G5.2”. It is a result of the EGU General Assembly 2019, Vienna, Austria, 7–12 April 2019.

The Copernicus POD Service is financed under ESA contract no. 4000108273/13/1-NB, which is gratefully acknowledged. The work performed in the frame of this contract is carried out with funding by the European Union. The views expressed herein can in no way be taken to reflect the official opinion of either the European Union or the European Space Agency. The authors also acknowledge the work done by the joint CNES/CLS analysis center of the IGS, which provides the GPS orbit, clock and bias products used for the ambiguity-fixed Sentinel-1 orbit solutions in this study. The provision of combined GPS orbit and clock products by the IGS is also greatly appreciated.

This paper was edited by Adrian Jaeggi and reviewed by two anonymous referees.