Copernicus POD Service Operational Experience

Transcription

1 SpaceOps Conferences May 2016, Daejeon, Korea SpaceOps 2016 Conference / Copernicus POD Service Operational Experience Jaime Fernández 1, Diego Escobar 2 and Francisco Ayuga 3 GMV AD., Isaac Newton 11, Tres Cantos, Spain and Pierre Féménias 4 ESA/ESRIN, Via Galileo Galilei, I Frascati, Italy ANX CPOD DORIS EGP ESA IPF LRR PDGS POD SLA SLR Currently the European Space Agency (ESA) is deploying an operational system for routine Earth Observation named Copernicus, a joint initiative of the European Commission and the European Space Agency, designed to support a sustainable European information network by monitoring, recording and analyzing environmental data and events around the globe. The Copernicus program will consist of six different families of satellites being the first three missions, Sentinel -1, -2 & -3, the subject of this paper. GMV is currently operating the Copernicus POD (CPOD) Service, in charge of providing accurate orbits and attitude products for the Sentinel-1, -2, and -3 missions as an operational service. This service is operated by GMV routinely and continuously on a 7x24 basis for a global community of users. Although being operated as an external service, the Copernicus POD Service is part of the Payload Data Ground Segment (PDGS) of the Sentinel missions and it is subject to a Service Level Agreement (SLA) that monitors the quality of the service provided. This paper describes the architecture of the Copernicus POD Service and the operational experience obtained during the routine operation of Sentinel-1A, Sentinel-2A, and Sentinel- 3A. Nomenclature = Ascending Node = Copernicus POD = Doppler Orbitography and Radio-positioning Integrated by Satellite = External GPS Provider = European Space Agency = Instrument Processing Facility = Laser Retro-Reflector = Payload Data Ground Segment = Precise Orbit Determination = Service Level Agreement = Satellite Laser Ranging 1 POD Expert & Project Manager, Flight Dynamics & Operations, jfernandez@gmv.com 2 Section Head & Technical Leader, Flight Dynamics & Operations, descobar@gmv.com 3 Project Engineer, Flight Dynamics & Operations, fjayuga@gmv.com 4 Earth Observation Application Engineer & Technical Officer, Earth Observation Ground Segment Department, pierre.femenias@esa.int 1 Copyright 2016 by GMV & ESA. Published by the, Inc., with permission.

2 I. Introduction RECISE Orbit Determination (POD) has traditionally been an area of expertise reserved to research institutes P and space agencies. The reason is the detailed and complex modeling needed to achieve the required accuracies (up to 1 cm radial error) and state-of-the-art tracking techniques involved: mainly Global Navigation Satellite Systems such as GPS, Satellite Laser Ranging (SLR) and Doppler Orbitography and Radio-positioning Integrated by Satellite (DORIS). Very few companies have so far developed proper skills in this rather scientific field. GMV has been involved in POD activities since the early 90 s providing support to the POD operations of the ERS missions at the European Space Operations Centre (ESOC) in Darmstadt, Germany. Since then GMV has provided uninterruptedly expert support to ESA in this field for other missions including the estimation of the precise orbits and clocks of the GPS, GLONASS and GALILEO; other examples are the ENVISAT, ERS-1 and ERS-2. Furthermore, GMV has developed the operational POD system for EUMETSAT s MetOp satellites, ESA s SWARM mission and the Spanish PAZ satellite. Currently the European Space Agency (ESA) is deploying an operational system for routine Earth Observation named Copernicus, a joint initiative of the European Commission and the European Space Agency, designed to support a sustainable European information network by monitoring, recording and analyzing environmental data and events around the globe. The Copernicus program will consist of six different families of satellites being the first three missions Sentinel -1, -2, and -3. These missions have very demanding requirements in terms of orbital accuracy and timeliness, with requirements of 8-10 cm in less than 30 minutes for Near Real Time applications to 2-3 cm in less than one month for Non-time Critical applications. GMV developed and it is currently operating the Copernicus POD Service, in charge of providing accurate orbit and attitude products for the Sentinel -1, -2, and -3 missions as an operational service. This service is operated by GMV routinely and continuously on a 7x24 basis for a global community of users from its premises in Tres Cantos, near Madrid, Spain. Although being operated as an external service, the Copernicus POD Service is part of the Payload Data Ground Segment (PDGS) of the Sentinel missions and it is subject to a Service Level Agreement (SLA) that monitors the quality of the service provided. This paper describes the architecture of the Copernicus POD Service and the operational experience obtained during the routine operation of Sentinel-1A (launched on April 2014) and Sentinel-2A (launched on June 2015) and Sentinel-3A (launched on February 2016). During the commissioning phase, the Copernicus POD Service carries out a number of Calibration and Validation activities to prepare the system for the Routine Operation Phase. After the commissioning, the mission begin the Routine Operation Phase (ROP). During this phase, the Copernicus POD Service run completely automatically and it is monitored using NAGIOS to monitor the health of the overall system, and Quality Reports to monitor the quality of the products generated. This paper will present the architecture that support the service together with the activities performed during the commissioning phase of Sentinel-1A (finished on September 2014), Sentinel-2A (finished on September 2015) and Sentinel-3A (to be carry out between February and July, 2016). In particular, the complexity of the Cal/Val activities of Sentinel-3A, which carries a DORIS receiver and a LRR in addition to a GPS receiver, will be presented. Finally the activities performed during the ROP will be described. II. The Copernicus program The Copernicus 1 program is a joint initiative of the European Commission (EC) and the European Space Agency (ESA), designed to support a sustainable European information network by monitoring, recording and analyzing environmental data and events around the globe. ESA is developing a new family of satellites called Sentinels 2 specifically for the operational needs of the Copernicus program. Each Sentinel mission is based on a constellation of two satellites to fulfill revisit and coverage requirements, providing robust datasets for Copernicus Services. These missions carry a range of technologies, such as radar and multi-spectral imaging instruments for land, ocean and atmospheric monitoring. 1) Sentinel-1 is a polar-orbiting, all-weather, day-and-night radar imaging mission for land and ocean services. The first Sentinel-1 satellite, Sentinel-1A, was launched on a Soyuz rocket from Europe's Spaceport in French Guiana on 3 April Sentinel-1B will be launched in April ) Sentinel-2 is a polar-orbiting, multispectral high-resolution imaging mission for land monitoring to provide, for example, imagery of vegetation, soil and water cover, inland waterways and coastal areas. 2

3 Sentinel-2 can also deliver information for emergency services. Sentinel-2A was launched on 23 June 2015 and Sentinel-2B will follow in the second half of ) Sentinel-3 is a multi-instrument mission to measure sea-surface topography, sea- and land-surface temperature, ocean color and land color with high-end accuracy and reliability. The mission will support ocean forecasting systems, as well as environmental and climate monitoring. Sentinel-3A was launched on 16 February Sentinel-3B is scheduled for launch in ) Sentinel-4 is a payload devoted to atmospheric monitoring that will be embarked upon a Meteosat Third Generation-Sounder (MTG-S) satellite in geostationary orbit. MTG-S is scheduled for launch in ) Sentinel-5 is a payload that will monitor the atmosphere from polar orbit aboard a MetOp Second Generation satellite. MetOp-SG-A is scheduled for launch in ) Sentinel-5 Precursor satellite mission is being developed to reduce data gaps between Envisat, in particular the Sciamachy instrument, and the launch of Sentinel-5. This mission will be dedicated to atmospheric monitoring. Sentinel-5P is scheduled for launch in the second half of ) Sentinel-6 carries a radar altimeter to measure global sea-surface height, primarily for operational oceanography and for climate studies. Sentinel-6A is scheduled for launch in The Copernicus POD Service provides orbital products for the Sentinel-1, -2 and -3 missions. A. The Sentinel-1, -2 and -3 missions The Sentinel-1 mission is the European Radar Observatory that continues the C-band SAR Earth Observation heritage of ESA's ERS-1, ERS-2 and ENVISAT. It includes C-band imaging operating in four exclusive imaging modes with different resolution (down to 5 m) and coverage (up to 400 km). It provides dual polarization capability, very short revisit times and rapid product delivery. Synthetic Aperture Radar (SAR) has the advantage of operating at wavelengths not impeded by cloud cover or a lack of illumination and can acquire data over a site during day or night time under all weather conditions. The mission is composed of a constellation of two satellites, Sentinel-1A and Sentinel-1B, in a near-polar, sun-synchronous orbit with a 12 days repeat cycle. Both satellites share the same orbital plane with a 180 orbital phasing difference. Sentinel-1 satellites carry two dual-frequency GNSS receivers for the purpose of POD. The receivers provide observations with a rate of 0.1Hz, in addition to the navigation solution computed on-board. Only one receiver provides observations at any time, switching to the other in case of failure. Finally the AOCS provides the quaternions representing the attitude of the satellite. Figure 1. Sentinel-1 (credit: ESA) The Sentinel-2 mission will ensure the future continuity of high-resolution optical missions such as Landsat and SPOT for multi-spectral imaging for land mapping. The mission will monitor variability in land surface conditions, and its wide swath width and high revisit time will support monitoring of changes to vegetation within the growing season. Vegetation, soil and coastal areas are among the monitoring objectives. The mission is composed of a constellation of two satellites, Sentinel-2A and Sentinel-2B, in a near-polar, sun-synchronous orbit with a 10 days repeat cycle. The Mean Local Solar Time (MLST) at the descending node is 10:30 (am). Both satellites share the same orbital plane with a 180 orbital phasing difference. Sentinel-2 satellites also contain two dual-frequency GNSS receivers, with similar characteristics to those onboard Sentinel-1, for the purpose of POD. Also the AOCS provides the quaternions representing the attitude of the satellite. 3

4 Figure 2. Sentinel-2 (credit: ESA) The Sentinel-3 mission will ensure the future continuity of medium resolution sensors like MERIS, (A)ATSR and VGT, as well as for the Altimetry System (RA-2, MWR, DORIS) on-board ENVISAT and on-board the ERS platforms (RA, MWR, PRARE) for global ocean and land monitoring. The primary objective is marine observation, and it will study sea-surface topography, sea and land surface temperature, ocean and land color with high accuracy and reliability to support ocean forecasting systems, environmental monitoring and climate monitoring. The mission is composed of a constellation of two satellites, Sentinel-3A and Sentinel-3B, in a near-polar, sun-synchronous orbit with a 27 days repeat cycle. The Mean Local Solar Time (MLST) at the descending node is 10:00 (am). Both satellite share the same orbital plane with a 180 orbital phasing difference. Sentinel-3 satellites carry two dual-frequency GNSS receivers, a DORIS receiver and a Laser Retro-Reflector (LRR) for the purpose of POD. The GNSS receivers provide observations with a rate of 1Hz. Only one receiver provides observations at any time, switching to the other in case of failure. Finally the AOCS provides the quaternions representing the attitude of the satellite. Figure 3. Sentinel-3 (credit: ESA) Tab. 1 below shows additional information of the satellites. Sentinel-1 Sentinel-2 Sentinel-3 Altitude 639 km 786 km km Inclination 98.18º 98.58º 98.65º Mass 2300 kg 1140 kg 1250 kg Table 1. Altitude, Inclination, and Mass of Sentinel-1, -2 and -3 missions B. POD accuracy and timeliness requirements Detailed requirements for these missions can be found in Ref. 3 to 8. All orbital products generated by the Copernicus POD Service will be mainly GNSS-based, as each Sentinel mission carries two dual frequency GNSS receivers. Additionally, Sentinel-3 mission contains a DORIS receiver 4

5 and a Laser Retro-Reflector (LRR); the corresponding observations will be also used together with the GNSS observations. A wide variety of orbit determination products are required by the different Sentinel missions. Each of these products is defined by a different timeliness or latency constraint: 1) Near Real Time (NRT, 30 minutes to 3 hours latency), for all missions. 2) Short-Time Critical (STC, one day latency), for Sentinel-3 only. 3) Non-Time Critical (NTC, less than 1 month latency), for Sentinel-1, and Sentinel-3 only. 4) And, additionally, Reprocessing campaigns (REP) to obtain a homogeneous set of precise orbits and auxiliary products for the complete mission lifetime based on new physical models, strategies or algorithms. In the following Tab. 2, these accuracy requirements are summarized for all required products: Mission Orbit Product Product Category Latency Requirement Position Accuracy (RMS orbit) S-1 NRT Restituted Within 3 hours (from the reception of GNSS data) < 10 cm (2D) S-2 NRT Predicted 90 minutes before Sentinel-2 A/B ANX is crossed < 10 m (2D) 3-sigma S-2 NRT Restituted Within 30 minutes (from the reception of GNSS data) < 3 m (3D) 3-sigma S-3 NRT Restituted Within 30 minutes (from the 10 cm threshold with a target of reception of GNSS data) 8cm (Radial) S-3 STC Preliminary Within 1,5 days (36h) 4 cm threshold with a target of 3cm (Radial) S-1 NTC Precise Within 20 days 5 cm (3D) S-3 NTC Precise Within 28 days (i.e. 2 days before 3 cm threshold with a target of NTC science product delivery) 2cm (Radial) Table 2. Summary of timeliness and accuracy requirements III. The Copernicus POD Service The Copernicus PDGSs are responsible for the operational generation and dissemination of the mission products. These products are then distributed to the Copernicus Service Segment for value-adding processing and provision of operational services. The Copernicus Service Segment concept is a conceptual envelope drawn around all valueadded processing done on Copernicus satellite data, with clearly identified, reliable and validated end products. This Service Segment generates higher-level products to provide the modelers, the value-adders and the final users with the required information. The Copernicus POD Service is part of the PDGS Ground Segment of the Sentinel missions. It is a complex system where several partners are involved and many elements interact to obtain the required orbital accuracies in a timely manner. The POD Service is composed of two elements, the main one, called offline POD, is located at GMV premises in Tres Cantos, near Madrid (Spain), while the other element is an Instrument Processing Facility (IPF), in charge of computing the Sentinel-3 NRT products, located in the two Sentinel-3 PDGS centers, the so-called Marine Center, located in EUMETSAT, Darmstadt (Germany) and the Land Center, located in Svalbard (Norway). Although the IPF is not operated directly by GMV, it is also under GMV s responsibility. In the following, we will concentrate on the characteristics of the Copernicus POD Service, also briefly describing the commonalities and exploited synergies with the NRT POD IPF for Sentinel-3 mentioned above. A. Partners The Copernicus POD Service has been developed and it is operated by GMV, but it interacts with different entities, both public and private, that act as clients, users and subcontractors. Following are the current main members of the Copernicus POD Service: - ESA ESRIN (European Space Research Institute). This center leads the development of the different PGDSs, and in particular, the Sentinel-1 and -2 PDGS are located here. - ESA ESOC (European Space Operation Center). This center hosts the Flight Operations Segment (FOS) of Sentinel-1, -2, and -3 missions during the commissioning phase and also during the Routine Operation phase except for Sentinel-3 mission, which is handed-over to EUMETSAT. 5

6 - EUMETSAT (European Organization for the Exploitation of Meteorological Satellites). This center hosts the Flight Operations Segment (FOS) of Sentinel-3 mission during the Routine Operation Phase, and also the so-called Marine Centre PDGS of Sentinel-3. - CNES (Centre National d'études Spatiales) provides accurate orbits and platform data files for the Sentinel-3 mission. CNES has also contributed with the DORIS instrument, and as so, it also provides RINEX DORIS files to the Copernicus POD Service. - GMV Innovating Solutions is the prime of the Copernicus POD Service. It has developed and it is operating the Service from its headquarters in Tres Cantos, near Madrid (Spain). It is responsible also of the overall management and the evolutions of the system. - POSITIM UG provides expertise in the LEO POD field, prototypes improvements in algorithms and manages, on behalf of ESA/GMV, the Quality Working Group. - DLR (Deutsches Zentrum für Luft- und Raumfahrt) provides expertise in the LEO POD and GNSS fields. Additionally, it contributes with orbital products for external validation of products. - TUM (Technische Universität München) provides expertise in the LEO POD field. Additionally, it contributes with orbital products for external validation of products. - AIUB (Astronomisches Institut, Universität Bern) contributes with orbital products for external validation of products. - TU Delft (Technische Universiteit Delft) contributes with orbital products for external validation of products. - VERIPOS is the provider of accurate GPS orbits and clocks in NRT and STC timeliness for it use in the GNSS POD processing. B. Elements The Copernicus POD Service interacts with different entities, internal and external to the Copernicus program. Following is a description of the main elements interacting with the Copernicus POD Service: - Payload Data Ground Segment (PDGS): It is part of the Ground Segment of the Copernicus program. It is in charge of processing the scientific data retrieved by the Sentinel satellites. The Copernicus POD Service is part of the PDGS Segment, in charge of providing orbital products and platform data files, which are used in the different scientific processing done by the PDGS. The PDGS is the main provider of inputs to the Copernicus POD Service (L0 data containing the GNSS measurements, PVt solutions and attitude information) and the unique recipient of the products generated. - Sentinels Flight Operations Segment (FOS): Located at ESOC for Sentinel-1, Sentinel-2 and Sentinel-3 (for Sentinel-3 only during the commissioning phase) and EUMETSAT for Sentinel-3 during the routine operation phase, it is the provider of operational orbit, maneuver information and the satellite mass and center of gravity evolution. - External GNSS data Provider (EGP): A dedicated high accuracy and reliability external GPS data provider (VERIPOS). High rate orbits and clocks are provided with different levels of timeliness and accuracy. In case of unavailability a back-up solution has been put in place based on magicgnss 9, a GMV solution for GNSS accurate products. IGS final orbits and clocks are also retrieved and used in the computation of the NTC products. - International Laser Ranging Service (ILRS): They provide Sentinel-3 SLR data to be used in the POD processing, while the Copernicus POD provides orbit predictions (CPF files) of Sentinel-3 to ILRS. - External Auxiliary providers: Other ancillary data (from several sources): Mainly Earth Orientation Parameters and leap seconds from IERS (International Earth Rotation Service), solar and magnetic activity information for geodesy computations from NOAA and atmospheric gravity models from NASA. - Centre National d'études Spatiales (CNES): They provide Sentinel-3 orbital and attitude products (for comparison purposes), together with DORIS data. Copernicus POD provides GNSS Observation RINEX files to CNES. - External Validation: A number of independent institutions (ESOC, DLR, TUM, AIUB, TU Delft) provide independent orbit solutions for validation purposes to assess the quality of the CPOD products. For Sentinel-3 other centers will provide orbital products, including CNES and EUMETSAT. - CPOD Quality Working Group (CPOD QWG): The main purpose is to monitor the performance of the operational POD products (both the orbit products as well as the input tracking data) and to define potential and future enhancements to the orbit solutions. 6

7 Figure 4. Copernicus POD Service Elements C. Physical Architecture of the Copernicus POD Service From the point of view of the physical architecture, the CPOD system consists of three independent environments: Operational, Validation and Development. These systems are distributed into two independent networks, the CPOD (operational) network and the GMV network. The GMV network contains a system architecture emulating that of the operational network where the Development system is installed. Here CPOD engineer makes the improvements to the system. See Fig. 5 in the left hand side. Figure 5. Physical Architecture of the Copernicus POD Service The Operational and Validation systems are installed inside the CPOD network. The Operational system consists of a redundant ftp service with two machines mounted as a cluster, showing a single virtual IP direction to the external users through a redundant firewall. Two operational machines process the received data independently, one working as the nominal and other as the backup. These two machines store the redounded data into a Storage Area 7

8 Network (SAN) system mounted in RAID 5 to ensure the required low risk levels. The system can be accessed through a desktop machine for upgrades of the system and through a VPN service for operations. The Validation system just consists of a single machine processing the very same real data; this is used for testing changes before deploying into the operational system. See Fig. 5 in the right hand side. D. Functional components The main building blocks or functional components of the CPOD Service are the following: - Storage and Dissemination: An archiving system where the incoming data and the generated products are stored according to the specific requirements for each mission. The stored products are exposed to the users by means of shared storage system of an FTP server. - POD Computational Core: All low level functions associated to the processing of the POD data. The interfaces of the POD computational core to the Data Management System are standard navigation formats (i.e. RINEX, SP3, ANTEX, etc.). Following section provides details on the computational processing chains that are used to process the data. - Quality Check: Verification on the resulting POD products. The type of quality checks implemented depends on type of implemented POD process, NRT, STC or NTC. In the NRT systems this function is limited by the availability of reference data but it is intended to implement sufficient checks to guarantee the outgoing product quality and integrity. As part of the Copernicus POD Service, the NRT product will be quality controlled with respect to the non-time critical (NTC) solutions and the NTC solutions themselves will undergo internal and external (external validation) checks. - Reporting: Recollection of POD and monitoring information to generate the data required to analyze and assess the POD system performance. - Interface: Pre-processing of the incoming data and the formatting of the outgoing products. - Management System: Manages the incoming and outgoing data and the processes that manipulate the data as part of the POD process. E. Computational Processing Chains The whole POD Computational Core of the CPOD Service (and also the Sentinel-3 NRT POD IPF) is based on the existing ESA s NAPEOS technology. The NAPEOS-based POD processing shares the same functional modules across the different Sentinel missions and orbit solution levels (NRT, STC, and NTC). In this way, the consistency among the different levels of solutions is ensured in what regards processing algorithms, reference frames, dynamic and measurement models, etc. Basically, each sequence involves the following generic steps: 1. Data retrieval from the archive which contains all inputs and products for the complete mission 2. Processing 3. Quality control of products 4. Storage of products in the archive 5. Publication of products in the FTP server F. Quality Control and Reporting System The Quality Control element is in charge of the Operational POD quality control as well as the Long Term Monitoring and Validation of all products. The results of these two analyses must be included in the reports generated by the CPOD Service. The following aspects will be considered as part of the orbit processing chain and the POD quality control activities itself: - Orbit comparisons (e.g. against CNES for S-3) - Analysis of orbits overlaps between consecutive solutions - Analysis of Orbit Determination performance metrics (e.g. phase and code residuals RMS) - Analysis of the external GPS orbits and clocks quality (differences with respect to IGS final solutions) - Analysis of mission constraints (e.g. ground-track). An on-line system to routinely monitor the status of the CPOD Service has been also developed. This system is accessible by a standard web browser via https and it is based on NAGIOS tool. NAGIOS is an open source monitoring and alerting application for servers, switches, applications, and services. 8

9 IV. Operational Experience The operational experience can be divided in three phases: - The first phase corresponds to the rehearsal tests performed prior the launch of the satellite. During these tests, it is important to have realistic data that can be used by the operational system. It has been shown that this is particularly difficult for the POD component. A POD based on GNSS data requires a consistent set of GNSS observables (wrapped into a L0 binary format), accurate GNSS orbits and clocks, together with all the ancillary data like Earth Orientation Parameters (EOP), Solar Activity, etc. - The second phase corresponds to the Commissioning Phase, which span from the launch of the satellite till the beginning of the Routine Operational phase. In the case of Sentinel-1 and -2, the expected duration was 3 months, while for Sentinel-3 is 5 months. During this period of time, a number of tasks are performed to initiate the generation of all the products (orbital and auxiliary files) and to fine-tune the POD algorithms to achieve the best possible solution. - The last phase corresponds to the Routine Operational Phase (ROP), which span from after the Commissioning phase till the end of the mission. During this period of time, the system runs mostly automatically and the tasks of the project are to maintain the system, solve anomalies, evolve the system (including algorithms, hardware and software), manage the Quality Working Group and perform periodic reviews of the Service to monitor the quality of it. Following is a description of the experience in each of these phases with each of the missions launched by the time of writing this paper. A. Prior to the launch Rehearsal tests Prior to the launch of each individual satellite, there are rehearsal tests involving all the ground segment. In particular the Copernicus POD Service interacts mainly with the PDGS, as source of the GNSS data and consumer of the orbital products generated. Additionally, it receives inputs from FOS, EGP, CNES and ILRS (only for Sentinel-3), etc. (see section III for details). From the point of view of the Copernicus POD Service, the tests are split in two categories: 1. Interface test with the external entities (like PDGS, FOS, etc.,) to check the correct reception, and delivery of inputs/outputs, including checking the format of the files. 2. Internal (to the Copernicus POD Service) end-to-end test to check the overall POD processing chain. The end-to-end test is needed to test the continuous generation of all orbital products. In order to do these tests, a complete simulation environment was developed by GMV to simulate all external interfaces, including: - the GNSS observables and quaternions (in the native L0 format), delivered by the PDGS - GPS orbits and clocks (in SP3 and clock RINEX formats) delivered by the External GPS Provider, which operationally is VERIPOS, but on simulation it is based on IGS products. - Products delivered by FOS, like maneuver information, mass history file, location of center of gravity and operational orbits (which includes the orbit number). - DORIS data and CNES orbital and platform data file products (only for Sentinel-3) - SLR data (only for Sentinel-3) - Ancillary inputs like Earth Orientation Parameters, Leap Seconds and Solar activity, in this case the files are retrieved from operational entities (IERS, NASA, NOAA ) With all this information it was possible to simulate all the inputs to the Copernicus POD Service in such a way that a whole end-to-end simulation was carried out. The system was running in our development environment for extensive periods of times (months) to simulate the real behavior of the system. This simulation environment has proven to be very useful to test the whole processing chain and to identify problems before the launch of the satellite. Indeed the simulation environment continues to be used to test improvements before testing it with real data. B. Commissioning Phase After the launch of the satellite, the commissioning phase is used by the Copernicus POD Service for: - Initiation of the routine generation of orbital products and platform data files (with attitude information). For a description of the products, see section II-B). This is done sequentially: o Firstly the decoding of GNSS data is tested and analyzed, to check the correct production of RINEX observation files o Then the FOS inputs are checked, to see if the maneuver information is available 9

10 o Finally the generation of orbital products is activated, starting with the NRT products, then the STC products and finally the NTC products. The main problem faced during the commissioning phase was the correct decoding of the GNSS L0 files which required adaptations to the decoding SW to handle several anomaly circumstances not anticipated during the development phase. With the exception of Sentinel-2A, which required sometime to adapt the GNSS L0 decoding software to a time-stamping issue of the GNSS measurements, the initial generation of products typically begins few days after the initial reception of inputs. - Validation of the accuracy of the orbital products. This is done by direct comparison of the operational orbital products against external solutions provided by the following institutions: AIUB, DLR, ESOC, TU Delft and TUM (see section III-A for a description of them), for all Sentinels, and in addition CNES and EUMETSAT for Sentinel-3. All of them provide orbital solutions computed with other POD software and strategies (e.g. dynamic, reduce-dynamic and kinematic PODs) that allowed to quantify the quality of the orbital products. The analysis done during the commissioning phase of Sentinel-1A showed that the NRT products fulfilled the 10 cm (2D RMS) requirement, but not the 5 cm (3D RMS) of the NTC orbital accuracy. In order to achieve the NTC accuracy, an evolution of the system was done after the Sentinel-1A commissioning phase to implement the IERS2010 standards on the POD SW, and to update the geopotential model used. After these changes the accuracy was easily achieved and currently is below 4 cm (3D RMS). Sentinel-2A has less demanding orbital accuracy requirements, which for NRT restituted orbit means 1 meter (3D RMS), which is easily achieved; indeed the actual accuracy was below 10 cm for the beginning of the mission. The second product, an orbital prediction with an expected accuracy of 3.3 meters (2D RMS), was also easily fulfilled provided there are no maneuvers in the prediction. Finally, at the time of writing this paper, the commissioning phase of Sentinel-3A is on-going and only partial results are available. The accuracy results of the STC products indicates a difference of 2.5 cm in the radial component when compared with a similar product generated by CNES, which indicates a fulfillment of the STC requirement of 3-4 cm in the radial direction. In any case, further analysis will be carried out during the Sentinel-3A commissioning phase to perform a refinement of the orbital generation and a crosscomparison against all external solutions. In the end, the accuracy expected will be similar to those generated for Sentinel-1A and Sentinel-2A. - Fine tuning of the POD processing, in particular the determination arc length, the number of CD (drag) and CPR (empirical accelerations) was analyzed to determine if the a-priori set-up was optimum. Originally the configuration was designed to use 24 hours of data for NRT solutions, and 72 hours of data for STC/NTC solutions. This was later on refined to use 48 hours of data in the STC/NTC solutions. The number of drag coefficients estimated (CD) was originally set-up to 6 every 24 hours, but was later changed to 10 every 24 hours to handle better the drag changes induced by the solar activity. Additionally the number of empirical accelerations was assessed; the number used is 2 every 24 hours. The current POD set-up can be seen here: o o o o Arc length: 48 hours (STC/NTC); 24 hours (NRT) Reference System: Polar motion and UT1: IERS C04 08 Pole model: IERS 2010 Conventions Precession / Nutation: IERS 2010 Conventions Gravity Model: Gravity field (static): EIGEN-6S2.5ext (120x120) Gravity field (time varying): drift / annual / semi-annual piece wise linear terms up to degree/order 50 Solid Earth tides: IERS 2010 Ocean tides: EOT11a (30x30, 99 tidal constituents) Atmospheric gravity: AGRA (20x20) Earth pole tide: IERS 2010 Ocean pole tide: IERS 2010 Third bodies: Sun, Moon, Planets DE405 Surface and empirical Forces: Radiation Pressure model: Box-wing model Earth radiation: NAPEOS model for Albedo and IR Atmospheric density model: msise90 10

11 o o Radiation pressure coefficient: 1 per arc; estimated Drag coefficients: 10 per day 1/rev empiricals: 2 sets per day in along and cross track direction (sine/cosine) Characteristics of the GPS measurements: Relativity: applied Sampling: 10 seconds Observations: iono-free linear combinations of phase and pseudo-range Weight: 0.8 m (pseudo-range) / 10 mm (carrier-phase) Elevation angle cutoff: 7 degrees Antenna phase-center wind-up correction: applied Antenna phase-center variation: applied (after inflight calibration from CP residuals) Receiver clocks: per epoch, every 10 sec Receiver ambiguities: estimated (float) GPS orbits: fixed (IGS finals) GPS clocks: fixed (IGS finals) Maneuvers are calibrated in the POD processing - Analysis of residuals to prepare a Phase Centre Variation (PCV) map in the form of an ANTEX file. Based on the phase residuals of the POD processing, a PCV map is generated for each individual GNSS antenna (there are two antenna in each satellite). The generation of the PCV map was done originally with a limited amount of data (5-10 days), and subsequently was refined later on using more data, and applying recursive estimation until convergence. The following figures shows a representation of the ANTEX files for Sentinel-1A and Sentinel-2A, as estimated from the POD processing. Figure 6. Sentinel-1A ANTEX PCV map based on phase residuals (mm) 11

12 Figure 7. Sentinel-2A ANTEX PCV map based on phase residuals (mm) - Fine tuning of the quality flagging mechanism. Each state-vector of the orbital products is flagged according to a number of criteria: maneuver, data gaps, number of GNSS observations used and rejected and the RMS of the code and phase residuals is used. All this is used to decide whether the state-vector is nominal, or it is degraded for any of the reasons mentioned above. - DORIS/SLR fine tuning: In the case of Sentinel-3 mission, it is necessary to fine tuning the processing of DORIS and SLR observations. This requires firstly to configure correctly the location of the respective stations, followed by the identification of biases (per station, and per system GNSS, SLR, DORIS) to remove them in order to have a three technique solution. C. Routine Operational Phase Sentinel-1A was launched on April 3 rd, 2014 and the commissioning phase spanned until end of September Sentinel-2A was launched on June 23 rd, 2015 and the commissioning phase extended until mid-october Finally Sentinel-3A was launched on February 16 th, 2016 and the commissioning phase will last until July After each commissioning phase, the Copernicus POD Service begins the Routine Operation Phase (ROP) of that satellite, until the end of life of the satellite. During the ROP phase, the main activities are: - The routine generation of orbital and platform data files. The generation of all products is completely automatic. There are basically two kind of products, i) those with a fix generation rate, which are for instance the STC and NTC products, where a single file is generated every day, ii) and those products that are generated after the reception of a new GNSS L0 file downlinked by the satellite after passing from a ground station. In the case of Sentinel-1 and -2, there are three downlink stations: in Svalbard (Norway), which has visibility of the satellite every orbit, in Matera (Italy) and Maspalomas (Spain), both of which have visibility of the satellite in some of the orbits. In the case of Sentinel-3, only Svalbard is used. With this downlink scheme, the Copernicus POD Service receives at least a GNSS L0 file, covering a full orbit, every 100 minutes (the approximate revolution period of the satellites), from which the NRT products are generated. The automatic generation of products is monitored automatically using NAGEOS, which allows to check the availability of input/outputs and check the creation of errors. In case of problems, an alert is generated in the Nagios tool (a webpage) and in addition an is sent to the operators. Depending on the criticality of the problem, the operators could need to apply procedures to solve the problem, including warning the project (ESA) about any unavailability of the system. The current experience with Sentinel-1A and Sentinel-2A has proved the robustness of the system, with no significant downtime. - The routine evaluation of the accuracy of the orbital products. This is done in collaboration with a number of external validation institutions: AIUB, DLR, ESOC, TUM and TU Delft for all missions and CNES and EUMETSAT with Sentinel-3. 12

13 In the case of Sentinel-1A, the institutions compute offline accurate orbits and provide them to GMV for cross-comparison. It has been shown that systematically the accuracy requirements are fulfilled without major problems. Fig. 8 shows the results of the last comparison campaign of January 2016; it shows the 3D RMS per day, during 10 days in January 2016, between the operational Sentinel-1 NTC solution (CPOD) and the daily solutions provided by each institution. Additionally there is a combined solution (COMB) computed as a weighted average of all individual solutions. It can be seen that the differences are systematically below the required 5 cm. Figure 8. Sentinel-1A orbit comparisons (3D RMS; cm) between CPOD and external solutions; red line is threshold Additionally, all NRT and NTC products are compared routinely against an offline ESOC solution (which used the same POD SW, NAPEOS, but of a different version and using different configuration and inputs). The following plots show the differences with respect to ESOC from October 2015 to January 2016, where it can be seen that systematically the differences are well below the threshold. The cases above the threshold are typically due to maneuvers and data gaps: 13

14 Figure 9. Sentinel-1A Restituted Orbital Product vs. ESOC (2D RMS) from 1st October 2015 until 31st January 2016 Figure 10. Sentinel-1A Precise Orbital Product vs. ESOC (3D RMS) from 1st October 2015 to 31st January 2016 In the case of Sentinel-2A, the accuracy results are similar to those of Sentinel-1A. Fig. 11 shows the results of the last comparison campaign of January 2016; it shows the 3D RMS per day, during 10 days in January 2016, between the operational Sentinel-2 NTC solution (CPOD) and the daily solutions provided by each institution. Additionally there is a combined solution (COMB) computed as a weighted average of all individual solution. It can be seen that the differences are systematically below 5 cm, like in the case of Sentinel-1A. 14

15 Figure 11. Sentinel-2A orbit comparisons (3D RMS; cm) between CPOD and external solutions Fig. 12 shows the accuracy of Sentinel-2A NRT product (which has an accuracy requirement of 1 meter 3D RMS) as compared against the NTC internal solution, which has been proved before to have an accuracy better than 5 cm. It can be clearly seen that the typical accuracy is below 10 cm, like with Sentinel-1A. Points above the 10 cm threshold indicated in the plot are due to maneuvers or data gaps. Fig. 13 shows the accuracy of Sentinel-2A predicted product (which has an accuracy requirement of 3.3 meters 2D RMS). Once again the accuracy is typically fulfilled without problems. Points above threshold are due to manoeuvers and data gaps. 15

16 - Evolution of the system. Together with POSITIM and the rest of the POD QWG, long term analysis and improvements are implemented. For instance, during 2015, new ANTEX files for Sentinel-1A and -2A have been created to remove systematic biases between solutions, and new Box-wing models have been developed for Sentinel-1A, -2A and -3A to be used together with the drag and solar radiation force modelling. - Every four months a Regular Service Review is carried out to analyze in detail the quality of the service. This review analyze not only the accuracy of the products, but also the quality of the GPS orbits and clocks provided by VERIPOS. Additionally the evolutions to the system, actions and anomalies are discussed. - Twice per year the Quality Working Group gathers to discuss the quality of the products and the future evolutions to the system in an effort to maintain the system state-of-the-art. - Finally, the Copernicus POD Service is subject to a Service Level Agreement (SLA) that measures the quality of the service with a number of Key Performance Indicators that measure the availability and quality of data, availability of the service (measured as downtimes of the ftp) and responsiveness to the anomalies. Every month a report is sent to ESA for evaluation. V. Conclusion This paper has presented the Copernicus POD Service, in charge of providing orbital products to support the Sentinel-1, -2, and -3 Copernicus missions. It has been presented the requirements on timeliness and accuracy followed by a description of the Service: its partners, elements and architecture. Finally it has been described the operational experience with the first satellites launched, Sentinel-1A, Sentinel- 2A and Sentinel-3A. Future satellites will be launched this and next year to have a complete system with six satellites by end of Acknowledgments The Copernicus POD Service is financed under ESA contract no /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. References S-1 Mission Requirements Document, ES-RS-ESA-SY-0007, Issue S-2 Mission Requirements Document, EOP-SM/1163/MR-dr, Issue

17 5 S-3 Mission Requirements Document, EOP-SMO/1151/MD-md, Issue GMES S-1 System Requirements Document - S1-RS-ESA-SY GMES S-2 System Requirements Document - S2-RS-ESA-SY GMES S-3 System Requirements Document - S3-RS-ESA-SY