AWS Location: s3://gnss-ro-data

AWS Region: us-east-1

Managing Organization: Atmospheric and Environmental Research, Inc.

Correspondence: Stephen Leroy (sleroy@aer.com) or Amy McVey (amcvey@aer.com)

This Readme presents a brief background to GNSS radio occultation, describes the AWS RO data formats calibratedPhase, refractivityRetrieval, atmosphericRetrieval, defines the path/file naming conventions for the RO data in the AWS Open Data Registry, briefly describes the RO DynamoDB database, and documents license information. There will be multiple versions of the data description document:

• Data description document for version 1.1. This update removes refractivityRetrieval and atmosphericRetrieval files for bad occultations, creates a "setting" variable in those same two file types that defines whether an occultation is rising or setting, and will manifest COSMIC-2, Spire, and GeoOptics radio occultation data.

Access to the RO metadata and data files can be handled seamlessly through the awsgnssroutils package, which now resides in the PyPI open source repository. It can be installed by "pip install awsgnssroutils". The source package can be found in awsgnssroutils.

Tutorial demonstrations, which incorporate use of the awsgnssroutils package, can be found in tutorials.

Finally, the Python software used to ingest RO data, catalogue the metadata, and produce the data files that are contained in the AWS Registry of Open Data RO repository can be found in reformatting_system.

Radio occultation is the limb sounding of planetary atmospheres by measuring the bending of a well-timed, temporally coherent radiation transmitted from a spacecraft outside the atmosphere that then transits the limb of the atmosphere and is received on the other side of the atmosphere. The atmosphere itself bends the signal as it passes through the atmosphere because of vertical gradients of the index of refraction, and the receiver infers the amount of bending by measuring Doppler shifts in the carrier frequency of the transmitted signal. The received radiation field, both the amplitude and the phase, can be inverted for vertical profiles of the index of refraction. Empirical formulas relate the index of refraction to various constituents in the atmosphere, which enable retrieval of temperature, pressure, etc., with the a few additional constraints, such as the hydrostatic equation.

Planetary radio occultation has yielded some of the strongest knowledge of the temperatures and pressures in planetary atmospheres since the 1960's. Radio occultation of the Earth's atmosphere using the already existing satellites of the Global Positioning System as transmitters commenced with the proof-of-concept mission GPS/MET in 1995. Since then, over 20 satellites have obtained radio occultation data using GPS. Modern Earth RO missions use the transmitters of other Global Navigation Satellite Systems (GNSS), such as the Russian GLONASS, the European Galileo, and the Chinese BeiDou. Still others use the transmitters of geo-synchronous regional navigation systems such as the Indian SBAS and the Japanese QZSS. Consequently, Earth radio occultation is now referred to as GNSS RO.

Because GNSS RO is a timing measurement as opposed to a radiance measurement, and because the amplitude and phase of the signal are tracked in the course of a sounding, GNSS RO begets two astonishing properties that set it aside from other remote sensing techniques for temperature and pressure. First, because the international definition of the unit of time can be scaled to an accuracy of one part in 10¹⁵, GNSS RO has the potential to benchmark the state of the climate to great accuracy. This makes GNSS RO a formidable climate monitoring technique. Second, because the phase of the radiation is measured in addition to the amplitude of the transmitted signal, holographic reconstruction methods become possible, and the result is extraordinary vertical resolution. The physical limitation on the vertical resolution is Fraunhofer diffraction, which yields a limit of 10 meters in the vertical. Other practical considerations provide the actual limitations, however --- the data rate, receiver electronic noise, and the horizontal-to-vertical aspect ratio of the atmospheric features being studied --- and the more practical result is that GNSS RO obtained ~100-meter vertical resolution. Such vertical resolution enables a host of interesting atmospheric process studies.

The path and file naming convention follow in the second section. The third section contains descriptions of the file formats available in the GNSS RO repository in the AWS Open Data Registry. A detailed PDF document describes the rationale for the formats together with insights into their utility in addition to their actual contents. The following are all useful to a beginner at GNSS RO:

• A description of GNSS RO;
• A complete error analysis of GNSS RO;
• A description of a typical GNSS RO retrieval system;
• Vertical coordinates in GNSS RO;
• Physical optics processing algorithms;
• Statistical optimization to smooth RO in the upper stratosphere;
• A description of super-refraction;
• Retrieving tropospheric water vapor in GNSS RO; and
• Ionospheric calibration and the removal of ionospheric residual.

The AWS Open Data Registry now makes available all GNSS RO obtained and processed by three independent RO processing centers: the COSMIC DAAC of the University Corporation for Atmospheric Research, the Jet Propulsion Laboratory of the California Institute of Technology, and the Radio Occultation Meteorology Satellite Application Facility (ROM SAF) of EUMETSAT. One other independent processing center is considering participation: the Wegener Center of the University of Graz. The data are available at three levels: calibrated satellites data (level 1b), retrievals of bending angle and refractivity (level 2a), and retrievals of atmospheric temperature, pressure, and water vapor (level 2b).

All RO data is provided by at least three independent processing centers, but those centers did not process RO data from all GNSS RO missions, and each center implemented different versions of quality control. Consequently, there will be enormous overlap in RO sounding data between the different centers, but the overlaps are not supersets or subsets of one another. It will not be unusual at all to find multiple retrievals of the same RO sounding, nor will it be unusual to find only version even if multiple retrieval centers set out to process those soundings.

In all files, time is provided in units of GPS seconds, which is the number of seconds elapsed since 00:00 UTC on 6 January 1980. It does not add in leap seconds; thus, as of July of 2021, GPS time appears to lead UTC time by 18 seconds. Also, every file contains metadata on the processing center that provisioned the data, a pointer to the relevant data use license, and a list of references in the form of digital object identifiers that document the processing system used to generate the data in each file.

Calibrated GNSS RO data are provided as calibratedPhase files. The format is NetCDF4, one occultation sounding per file. The contents of each file includes the variables time (which is the independent coordinate), snr, excessPhase, positionLEO, and positionGNSS for an arbitrary number of signals tracked. Other variables define the 3-digit RINEX 3 observation codes that identify the signals (see section 5.1 of the RINEX 3.04 documentation for example), whether or not the navigation data bits have been removed when unwinding the excess phase, and the range and phase models used in open loop tracking. Each variable is well described in its variable attributes. Global attributes provided necessary metadata on the occultation sounding, including three-digit identifier of the transmitting GNSS, the names of the mission and the satellite carrying the receiver, the name of a reference GNSS satellite used to calibrate the receiver clock if calibration was performed by single- or double-differencing, the name of the ground station used to calibrate the GNSS transmitter clocks if the calibration was performed by double-differencing. Note that the excessPhase and the positions of the LEO (positionLEO) are given as a functions of time in the files, the time corresponding to the time of reception of the signal by the LEO satellite; however, the positions of the transmitter positionGNSS are specified at the times the received signals were transmitted by the GNSS satellite. The GNSS positions were computed by interpolating the GNSS positions backward in time from the receive time by an amount corresponding by the light travel time between the transmitter and the receiver. This is done because it is the positions of the transmitter at transmit time that directly enters into the RO retrieval process.

Retrievals of profiles of bending angle, microwave refractivity, and "dry" pressure are provided in refractivityRetrieval files. A dry atmospheric retrieval is a retrieval of pressure and temperature that is obtained when the contribution water vapor to microwave refractivity is considered non-existent. This is a good approximation in the upper troposphere and stratosphere, but not so in the lower troposphere where water vapor contributes approximately 10% of the refractivity. Dry retrievals are generated because RO does not provide enough information by itself to distinguish between the contributions of water vapor and the "dry" atmospheric constituents (nitrogen, oxygen, carbon dioxide) to microwave refractivity.

The variables include impactParameter (which is the independent coordinate for bending angle retrieval), bendingAngle, combinedBendingAngle (which is ionosphere-corrected) for bending angle retrievals; altitude, geopotential, longitude, latitude, refractivity and dryPressure for dry atmospheric retrievals. Dry temperature can be computed by dividing dry pressure by refractivity and multiplyting by 0.776 N-units K/Pa. Other useful auxiliary data define a reference geolocation for the sounding, the shape of the mean sea level geoid, and the effective center of curvature for the sounding.

Note: Geopotential can be converted to geopotential height by dividing by the appropriate constant value of gravitational acceleration at the Earth's surface. In most cases, that constant is taken as the World Meteorological Organization standard of 9.80665 J kg^-1 m^-1. Some atmospheric models and processing algorithms for other atmospheric datasets have implemented different values for this constant. Their documentation must be searched for the value it incorporated in processing or execution and then used in the conversion to geopotential height when the data of this archive is compared to the atmospheric model output or atmospheric dataset, whether it be satellite or in-situ.

Retrievals of profiles of temperature, pressure, and water vapor are provided in atmosphericRetrieval files. In each case, auxiliary information on temperature and/or water vapor has been used to disentangle the contributions of water vapor and the dry atmosphere to microwave refractivity. The auxiliary data usually comes from the forecasts or analyses of a numerical weather prediction model or atmospheric reanalysis. These profiles are usually of much coarser vertical resolution than the refractivityRetrieval files because the atmospheric model data used as an auxiliary input is much coarser in vertical resolution than RO is capable of.

The variables include geopotential (the natural independent coordinate of RO atmospheric retrieval), refractivity, temperature, pressure, and waterVaporPressure. Other data provide geolocation information for the sounding and indicate whether whether or not super-refraction is present in the sounding and if a special algorithm was used to account for the influence of super-refraction in the retrieval.

The path and file naming convention for all of the RO data is contributed/version/center/mission/filetype/yyyy/mm/dd/filetype_mission_center_version_occid.nc. The various mnemonics in this path are defined in the following table:

The occultation identifier occid is defined as ttt-leo-yyyymmddhhnn in which ttt is the three-digit RINEX standard identifier of the transmitting GNSS satellite and leo is the name of the low-Earth orbiting receiving satellite. The remaining symbols denote the time of the occultation. Note that an occultation sounding is uniquely identified by the transmitter, receiver, and time of the occultation, and the precision of the time needs to be no better than a few minutes.

The names of the GNSS RO missions and LEO/receiving satellites are given in the following table.

Multiple satellites are listed for each mission if the mission consisted of multiple satellites (such as COSMIC-1 and COSMIC-2) or the same program deployed a series of similar environmental sounding satellites (such as Metop). The exception is GPS/MET, which is just one satellite. GPS/MET is a special case because it obtained RO soundings in two different modes: one when the GPS anti-spoofing encryption on L2 was turned off ("gpsmet") and one when the GPS anti-spoofing encryption on L2 was turned on ("gpsmetas"). The anti-spoofing encryption greatly inhibited the ability to track the L2 signal during RO soundings, and consequently special retrieval algorithms were applied to process these data at all. Since GPS/MET, flight GNSS RO receivers were enabled with greatly improved technology to track GPS L2 even when encrypted by anti-spoofing. Preently, most GPS satellites transmit a civil signal at L2 that requires no specialized tracking techniques.

The following table defines the names of the contributing RO processing centers.

An application programming interface (API) for Python is provided that simplifies the querying, subsetting, and downloading of RO data according to metadata such as geolocation, date-time range, local (solar) time range, RO mission or receiver satellite, transmitting constellation or satellite, occultation geometry (rising v. setting), and availability of RO data files. The API is available through PyPI and thus can be installed by a pip install. The source code is provided here in the directory awsgnssroutils.

A set of five tutorial demonstrations is provided in the tutorials directory. The first demonstrates the use of the Python API, the others are representative of common research activities that involve RO data. The tutorials specialize in research using RO data in the AWS Registry of Open Data repository. The same directory also contains material used in two public workshops.

The format definitions are the outcome of consultations of an international team of GNSS RO retreival scientists and experts. They were drawn largely from the International Radio Occultation Working Group (IROWG), from public and private concerns, from universities, government laboratories, and satellite agencies.

The data use licenses for the various contributing centers are

• The COSMIC Project Office at UCAR, a creative commons license,
• The NASA Jet Propulsion Laboratory, California Institute of Technology, a creative commons license, and
• The ROM SAF and EUMETSAT, the EUMETSAT data use license.

The repository of GNSS RO data in the AWS Registry of Open Data was assembled and continues to be maintained by scientists and software engineers at Atmospheric and Environmental Research, Inc. Funding for this effort was provided by the NASA Advancing Collaborative Connections for Earth System Science (ACCESS) Program 2019, grant 80NSSC21M0052.

Last update: 30 November 2023