First spaceborne demonstration of BeiDou-3 signals for GNSS reflectometry from CYGNSS constellation

Weiqiang LI, Estel CARDELLACH, Serni RIBO´ , Antonio RIUS,Bo ZHOU

a Institute of Space Sciences (ICE, CSIC), Barcelona 08193, Spain

b Institut d’Estudis Espacials de Catalunya (IEEC), Barcelona 08034, Spain

c Shanghai Spaceflight Institute of TT&C and Telecommunications, Shanghai 201109, China

KEYWORDS Global Navigation Satellite System (GNSS);BeiDou-3;Small satellites;Bistatic radar;GNSS Reflectometry(GNSS-R);CYGNSS;Tropical cyclone;Flood inundation

Abstract The full constellation of Chinese Global Navigation Satellite System (GNSS) BeiDou-3 has been deployed completely and started fully operational service.In addition to providing global Positioning, Navigation and Timing (PNT) services, the BeiDou-3 satellites transmissions can also be used as the sources of illumination for Earth Observation (EO) with a bistatic radar configuration.This innovative EO concept,known as GNSS reflectometry(GNSS-R),allows to measure the Earth surface characteristics at high resolution via the reflected L-band radar signals collected by a constellation of small, low cost and low Earth orbiting satellites. For the first time in orbit, earth reflected BeiDou-3 signal has been detected from the limited sets of raw data collected by the NASA’s Cyclone GNSS (CYGNSS) constellation. The feasibility of spaceborne BeiDou-3 reflections on two typical applications, including sea surface wind and flooding inundation detection,has been demonstrated. The methodology and results give new strength to the prospect of new spaceborne GNSS-R instruments and missions, which can make multi-GNSS reflectometry observations available to better capture rapidly changing weather systems at better spatio-temporal scales.

1. Introduction

The rise of ‘‘NewSpace” concept has been leveraging new Earth Observation (EO) missions based on constellations of small, low-cost satellites, which can capture rapidly changing weather systems such as hurricanes, extreme rain events, and flood inundation.The use of reflected Global Navigation Satellite System(GNSS)signals for sensing the Earth’s surface is an innovative concept with a broad spectrum of geophysical applications.This new radar remote sensing technique,known as GNSS Reflectometry (GNSS-R),requires only the receiver part of the radar, which can significantly reduce the size,power and cost of the sensor, and enables the deployment of constellations of small satellites. In addition,it takes measurements along several reflection tracks from different GNSS transmitters in parallel, and can provide exceptional spatial sampling capability and rapid revisit time over the Earth surface.These unique features of this concept suggest a new paradigm in spaceborne Earth environmental monitoring.

After early conceptual ground-based and airborne experiments,GNSS-R concept was demonstrated at spacecraft altitude onboard the UK Disaster Monitoring Constellation(DMC) satelliteand TechDemoSat-1 satellite,from which different geophysical applications have been explored over ocean, land, and cryosphere.Following the success of these explorations, different spaceborne missions have been proposed by using this innovative concept.Among others,a constellation of 8 micro-satellites for NASA’s Cyclone Global Navigation Satellite System (CYGNSS) mission was launched in 2016 with the specific aim of measuring extreme wind events in and near the inner core of tropical cyclones.Being the first pathfinder mission using the GNSS-R technique, CYGNSS observations have been also used to demonstrate some other scientific utilities, such as lake water level,flooded areasor soil moisture.

Meanwhile, continuous evolution of GNSS constellations provides world-wide users improved Positioning, Navigation and Timing (PNT) services by combining the observations from more visible GNSS satellites.These new GNSS constellations also increase the number of available sources of signal of opportunity, and thus can further improve by folds the spatial–temporal sampling efficiency of the GNSS-R systems.The BeiDou global navigation satellite system (BeiDou-3,BDS-3) has started initial global PNT service since the end of 2018 and full global PVT services since 31 July 2020 with advanced features.With the launch of the final BeiDou-3 satellite in June 2020, BeiDou-3 constellation consists of 30 navigation satellites in different orbits, including 24 Medium Earth orbit (MEO) satellites, 3 Inclined Geostationary Orbit(IGSO) satellites and 3 Geostationary Orbit (GEO) satellites.By using reflected signals from both GPS and BeiDou-3 satellites, the number of GNSS-R measurements can be improved at least by a factor of two.Such improvement can be well justified by the orbit simulation results shown Fig.1,in which the number of GNSS-R observations from GPS and BeiDou-3 satellites are significantly improved compared to the GPS only case. The maximum incidence angle is set to 40°. Fig. 1 also highlights a more significant coverage improvement over the Asian Pacific region, which is mainly due to the contribution from the BeiDou-3 GEO and IGSO satellites. In principle,reflected signals from BeiDou-3 can also be used for different GNSS-R applications by nature. In fact, the use of reflected signals from BeiDou satellites have been successfully demonstrated at low receiver altitude for different applications,however these previous works only use the reflected signals from BeiDou-2 satellites and collected with ground-based setups. Due to the difference in signal structure and transmitted power,the question still remains whether reflected signal from BeiDou-3 satellites can be detected at low orbit altitude.

Along with the initial operation of BeiDou-3,the CYGNSS satellites have been scheduled to collect the raw Intermediate Frequency (IF) data occasionally for the exploration of innovative GNSS-R applications.These raw IF datasets are unprocessed signal samples before any demodulation and down-linked directly to the ground station,and thus also include the direct and reflected BeiDou-3 signals within the receiver bandwidth (see Fig. 2(a)). By processing these raw IF data, Earth reflected BeiDou-3 signals have been detected for the first time in low earth orbit,and the potential of spaceborne BeiDou-3 reflectometry are demonstrated experimentally. The rest of this paper is organized as follows: the main data set and the data processing strategy are introduced in Section 2;two typical GNSS-R applications, i.e. sea surface wind retrieval and flooding inundation detection, are demonstrated with BeiDou-3 reflections in Section 3; finally, the conclusion and prospects are provided in Section 4.

Fig. 1 Global distributions of the number of 1 s GNSS-R observations per cell of 0.25°×0.25° (about 25 km×25 km) provided by a constellation of 8 GNSS-R satellite in 600 km Sun-synchronous orbit, accumulated over a period of 3 days.

2. Data sets and data processing

2.1. Description of data sets

CYGNSS constellation is designed to observe near surface wind speed over the ocean, and its main data product (Level 1 data) is the Delay-Doppler Maps (DDMs) of the forward scattered power of the GPS signal.In addition to the Level 1 data,the CYGNSS instruments also record the raw IF signal samples of the direct and reflected signals occasionally to explore other possible applications.Each raw IF data consists of raw signal sample streams (30–60 s) received by the zenith and nadir antennas. The bandwidth of the raw signal is ～2.5 MHz centered at the IF of ～3.8 MHz. As the BeiDou-3 B1C signal shares the same center frequency with GPS L1 C/A signal(1575.42 MHz),its direct and reflected signals within the instrument bandwidth were also recorded. A conceptual sketch of raw IF data collection with reflected BeiDou-3 signals is presented in Fig. 2(a) (BeiDou-3 satellite image credit:China Academy of Space Technology,CYGNSS spacecraft image credit: University of Michigan).

The CYGNSS mission has collected more than 200 raw IF collections over a period of more than 2 years since 2017.Aiming to different applications,these raw IF collections have been scheduled mainly around the US coast (the Gulf of Mexico,the Caribbean Sea and the western Atlantic Ocean)and Asian Pacific regions. Each collection contains reflections from different GNSS constellations occurring simultaneously in different directions with respect to the receiver, making a total of more than 2000 reflections. There are more than 500 tracks reflections from the BeiDou-3 satellites, which are shown in Fig.2(b) with different colors indicating the mean SNR.Each BeiDou-3 reflection is around 210–420 km along the track of the Specular Point (SP) over the Earth surface depending on the duration of its corresponding raw IF collection.It is noted that CYGNSS operates continuously over both ocean and lands, which makes it possible to demonstrate different scientific utilities using BeiDou-3 over from different types of natural covers, as presented in Section 3.

Fig. 2 GNSS-R raw IF data collection and data processing chain for spaceborne BeiDou-3 reflectometry.

2.2. Data processing

The processing of the raw IF data is based on a softwaredefined radio BeiDou-3 receiver extended for reflectometry,which consists of a closed-loop processing of the direct signal(S（t）)and an Open Loop(OL)processing of the reflected signals (S（t）).The ancillary information, including the positions, velocities and timing information of the GNSS transmitters and CYGNSS spacecrafts are extracted from the BeiDou-3 precise ephemeris and the CYGNSS raw IF and Level 1 metadata. A brief block diagram of the CYGNSS raw data processing is presented in Fig. 2(c). The codes offset(φ),carrier frequency(f)of the direct BeiDou-3 B1C signal are estimated with signal acquisition/tracking through a Delay-Locked Loop (DLL) and Phase-Locked Loop(PLL).It is noted that the BeiDou-3 B1C signal consists of a data component and a pilot component modulated by the navigation data bits (D) and the secondary code (C),respectively, which are also extracted from the direct signal tracking loops.

3. Results and analyses

Since the BeiDou-3 reflections can only be obtained from limited sets of CYGNSS raw IF data collected occasionally, we cannot consider to demonstrate operational GNSS-R applications, which generally relies on large number of accumulated observations. Instead, we focus on exploring the feasibility of spaceborne BeiDou-3 reflections on two typical applications,i.e. ocean surface wind measurement and flooding inundation detection.

3.1. Preliminary measurement of ocean surface wind

Being successfully demonstrated with UK TechDemoSat-1 and NASA’s CYGNSS mission,GNSS-R ocean scatterometry has risen with great potential to become an operational application. Moreover, its high spatial–temporal sampling capability and the penetration of L-band signal through heavy rain make GNSS-R wind speed measurements unique in monitoring extreme wind events near the center of tropical cyclones.To demonstrate the ocean scatterometry capability of reflected BeiDou-3 signal, 35 tracks of BeiDou-3 reflections obtained over the ocean are used for analyses. Similar to the CYGNSS baseline wind speed retrieval method,the fundamental measurement from the BeiDou-3 reflected signal is sea surface normalized bistatic radar cross section (NBRCS or σ), which is linked to the surface roughness and, indirectly, to the nearsurface winds.Due to the lack of information about the transmitted power and transmitting antenna gain of the BeiDou-3 satellites,the power of the direct signal received with the zenith antenna is used as the reference.And the NBRCS is computed from the power DDM of the reflected B1C signal through an inversion of the Zavorotny-Voronovich (Z-V) modelas

The sea surface σcomputed with Eq.(1)from the BeiDou-3 DDM measurements are then compared to collocated wind speed at 10 m above the sea surface (u) from the climate reanalyses data of the European Centre for Medium-Range Weather Forecasts (ECMWF ERA-5).It is noted that two tracks of BeiDou-3 reflections overpass over hurricane (as show in Fig. 3(c) and (d)) are excluded from this comparison,which will be analyzed later independently due to different sea surface response to wind stress in these cases. As shown in Fig. 3(a), the BeiDou-3 GNSS-R σmeasurements decrease rapidly with increasing wind speed. This behavior is well captured by a 2nd degree polynomial function of 1/usimilar to the CYGNSS standard Geophysical Model Function(GMF)as

Fig. 3 Sea surface wind speed retrieval using BeiDou-3 reflections.

in which the coefficients（a，a，a）are（13.2，79.67，20.12）according to the ordinary least squares fit solution.With this preliminary GFM, Fig. 3(b) presents the BeiDou-3 GNSS-R derived wind speed, and its error distribution corresponding to the ECMWF wind speed, which yields bias and Root-Mean-Square Error(RMSE)values of 0.07 m/s and 2.04 m/s,respectively.It is noted that these results cannot fully represent the achievable wind speed retrieval performance using reflected BeiDou-3 signals.The wind speed retrieval performance can be improved significantly if the BeiDou-3 σcan be properly calibrated with the information of transmitted power of the BeiDou-3 B1C signal and the receiver radiometry calibration information.

There are two tracks of BeiDou-3 reflections crossover Hurricane Florence (Category 4 on the Saffir-Simpson scale)in 2018, which provide unique opportunities to investigate the capability of BeiDou-3 reflections on high wind speed measurement through heavy precipitation. Figs. 3(c) and(d)show the interception between BeiDou-3 reflection tracks and Hurricane Florence in North West Atlantic on 13 and 15 September 2018, noted as track 1 and track 2, respectively. The distance between Track 1 and the center of Hurricane Florence is ～60 km with the mean SNR of the DDM being -6.3 dB,and the distance between Track 1 and the center of Hurricane Florence is ～190 km with the mean SNR of the DDM being 1.3 dB. The geocolor images derived from the GOES-16 satellite data are also shown as the backgrounds to indicate the location and extension of the hurricane. To show the sensitivity of reflected BeiDou-3 signal to sea surface wind in hurricane conditions, Figs. 3(e) and (f) shows the along-track evolution of BeiDou-3 σthrough the hurricane, from which it can be clearly seen that the high values of σaway from the hurricane center and the decrease as the center of cyclone is approached.Figs. 3(e) and (f) also present the along-track collocated wind speed from ECMWF ERA-5, which show good correlations with the BeiDou-3 σmeasurements with correlation coefficients of 0.62 and 0.89 for track 1 and 2,respectively, corresponding to RMSEs of 2.5 m/s and 1.5 m/s, respectively. Notably, the sensitivity of σto the sea surface wind speed at high wind speed is also determined with the linear regression, resulting in a sensitivity factor (∂σ/∂u) of-0.20. This sensitivity factor is consistent with the CYGNSS GMF sensitivity at high wind speed (-0.188),which further consolidates the effects of sea surface wind on BeiDou-3 reflection measurements. The difference between the sensitivity factors for the BeiDou-3 measurements and the CYGNSS standard GMF may be due to different sea state conditions in both cases. These results confirm that it is still possible to acquire BeiDou-3 reflections at spaceborne altitude even in extreme wind condition when the reflected signals are very weak. More comprehensive conclusions on how well spaceborne BeiDou-3 reflections can measure the intensity of typhoons/hurricanes would require dedicated spaceborne data collections with current and future GNSS-R missions, e.g.NASA’s CYGNSS and China’s BuFeng-1 constellations.

3.2. Mapping flooding inundation

The mapping of inundated areas during flooding events is increasingly necessary to gain insight about both causes and remedies.Rapidly monitoring of water extent during largescale flooding is crucial for emergency response to identify the area affected and to evaluate damage.In addition,detailed mapping of inundated areas over a long-time span can help scientists to build floodplain hydrodynamic model for hazard risk prediction.Previous airborne and spaceborne GNSS-R data has clearly shown the sensitivity of reflected GNSS signal power to the presence of surface water,which can be applied for inland water/wetland detection.Such information could provide a compromise between that provided by traditional microwave remote sensing satellites to quantify flood inundation dynamics at appropriate scales. Nevertheless, current approaches are mainly based on the standard Level 1 data,i.e.the power of the reflected signal with a couple of Hz resolution,which require a large amount of accumulated data to get better spatial resolution.In addition to the power of the reflected signal,the presence of liquid water over the land surface can significantly shift the diffuse reflection limit and improves the coherence of GNSS-R observations, which potentially can be used for identification of the inundated areas.

Here, we demonstrate the use of high sampling rate complex DDM for flood inundation detection along limited tracks of BeiDou-3 reflections over the 2019 Mississippi River flooding.The flooding along the Mississippi River valleys during the spring 2019 has been found to be the longest-lasting flood since the 1927 Great Flood. There are 4 tracks of BeiDou-3 reflection crossing extensively over the Lower Mississippi River and adjacent watersheds between 23 and 27 March 2019 (see Fig. 4(b)). In addition, the other 3 tracks of BeiDou-3 reflection are selected as the reference, which were collected over the same region during non-flooding conditions between 13 and 21 October 2018 (see Fig. 4(a)). The complex DDM are generated from these raw IF tracks following the procedure in Section 2.It is noted that the complex waveforms are generated before any incoherent processing,and thus keep both the amplitude and phase information. The main observable used in this analysis is the Coherent Coefficient (CoC) of two successive complex DDMs at the peak pixel as

By using a fixed Nof 10, the mean CoC along each BeiDou-3 reflection track are shown in Figs.4(a) and (b) for non-flooding and flooding conditions,respectively.The optical images from the Sentinel-2 satellite collected over the same region are also presented inFigs.4(a)and(b),with green areas indicating significant vegetation growth and blue areas for open or standing water. Overall, the mean CoC is generally elevated (i.e. R>0.9 in Tracks 4–7) over the inundated areas, where flooding extent is clearly indicated by the Sentinel-2 images (darker). By contrast, the mean CoC shows significant degradation in Fig.4(a)over non-inundated vegetation and bare soil regions(i.e.R<0.2 in Tracks 1–2),which indicates that the forward scattered signals from these regions are dominated mainly by incoherent scattering. It is remarkable that the bare soil/non-inundated vegetation generally shows higher mean CoC in flooding condition (i.e. 0.3–0.6)than that in non-flooding condition (～0.2), which is likely due to high soil moisture or saturated soil in flooding condition.

To better characterize the capability of reflected BeiDou-3 signal on inundation area characterization, two sub-regions with in Fig. 4(b) are selected and shown in detail in Figs. 4(c) and (d), from which two important aspects of the BeiDou-3 reflection CoC measurement are highlighted.

(1) The mean CoC measurements can characterize the inundated regions with high spatial resolution,which is clearly justified by the consistence between the along track evaluation of the Rvalue and the transition between flooded and non-flooded land at small scales.For example,the pixels along the BeiDou-3 reflection tracks show low Rvalues over the small regions(10 s to 100 s meters along the reflection tracks)indicated by the white arrows in Figs. 4(c) and (d), which correspond to the noninundated land indicated by the Sentinel-2 images. It is noted that each pixel along the BeiDou-3 reflection tracks corresponds to a ground distance of ～140 m with N=10 of 2 ms coherent integrated complex waveforms, which represents the spatial sampling resolution of the CoC measurements from BeiDou-3 reflections.This resolution can be adapted to characterize flood inundation dynamics at appropriate scales by using different number coherent integration.

Fig. 4 Inundated areas detection over the 2019 Mississippi River Flooding using reflected signals from BeiDou-3.

(2) The mean CoC measurements from BeiDou-3 reflection are still sensitive to surface water even when the water is obscured by dense vegetation,thanks to the penetration of L-band signal into the forest canopy.The two ground track segments of BeiDou-3 reflections pass across a densely vegetated inundated area. This area includes the Panther Swamp National Wildlife Refugeand the Delta National Forest,where the above ground biomass (AGB) can be as high as 300 Mg/ha as shown in Fig. 4(e).Despite the vegetation attenuation(～0.05 dB/(Mg/ha)),the forward scattered BeiDou-3 signal still shows high coherence (R>0.85) in Fig. 4(c) and (d) over these regions, implying its capability on detection inundation beneath vegetation canopy. It is noted that the full potential of using BeiDou-3 and other GNSS signals on sensing vegetated inundated area needs to be further investigated, which relies on more raw IF data collections over such regions, such as the Amazon basin.

4. Conclusions and prospects

Continuous evolution of GNSS constellations can increase the number of available sources of signal of opportunity for earth observation. As one of the main GNSS constellations,BeiDou-3 can significantly improve the number of GNSS-R observations,and thus the spatio-temporal sampling capability of further spaceborne GNSS-R systems. There has been no dedicated GNSS-R instrument capable of generating BeiDou-3 reflectometry measurements in orbit. By using the raw IF samples collected by the CYGNSS mission, the earth reflected BeiDou-3 signal has been detected for the first time,and the feasibility of spaceborne BeiDou-3 reflections on geophysical applications has been demonstrated. From these results, the following aspects are remarkable:

Further spaceborne GNSS-R instruments can take the advanced features of BeiDou-3 signals to improve the quality of the GNSS-R observations. Firstly, the larger bandwidth and sharper auto-correlation function allow one to use longer coherent integration time to improve the SNR of the GNSS-R observations.Secondly,it is suggested to combine the data and pilot components in B1C signal coherently to further improve the SNR. Finally, there are two open service signals at BeiDou-3 B1 band with 14.322 MHz apart,i.e.the B1C signal and B1I signal. Combing these two signals by using different strategies would allow new types of GNSS-R measurements to explore new GNSS-R applications. However, the GNSS-R instrument onboard the CYGNSS satellites can only cover the B1C signal,and the combination of different signal components at BeiDou-3 B1 band will be investigated in further work, which relies on the acquisition of BeiDou-3 signal with wider bandwidth covering both B1I and B1C signals.

It is demonstrated that sea surface wind speed can be retrieved from BeiDou-3 reflections with similar performance as that obtained with GPS signal.The wind speed retrieval performance can be further improved in future GNSS-R missions compatible with BeiDou-3 and with a throughout radiometric calibration. More importantly, the BeiDou-3 reflections acquired in extreme wind condition is still detectable and show sensitivity to in situ wind speed measurements. Nevertheless,more work is needed to understand the real achievable performance of BeiDou-3 reflections on ocean surface wind measurements, especially in high wind condition, with different instrumental configurations.

The capability of using high resolution GNSS-R measurements, i.e. the complex DDM, for flooding inundation mapping has been also demonstrated with reflected BeiDou-3 signals. It is shown that the inundated regions can be characterized with high spatial resolution (10 s to 100 s meters along the reflection track).In addition,the BeiDou-3 reflection measurements shows potential to detect surface water beneath thick vegetation canopy(up to 300 Mg/ha),which can provide a complement to the existing observational networks.

These results presented in this paper give new strength to the prospect of GNSS-R concept in the multi-GNSS era,and suggest the potential of building spaceborne GNSS-R constellations with improved effectiveness to understand rapidly changing geophysical phenomena at better spatio-temporal scales.With the accumulation of more spaceborne raw IF data collected by different GNSS-R missions, a thorough understanding of reflected signals and their observable from BeiDou-3 can be further investigated,which is of great importance in exploiting the potential of future spaceborne GNSS-R missions.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported in part by the Spanish Ministry of Economy and Competitiveness and EU/FEDER (ESP2015-70014-C2-2-R), the International Science and Technology Cooperation Projects of Shanghai (No. 17220730600), and the ESA-MOST China Dragon5 Program (ID. 58070). Weiqiang Li and Estel Cardellach are the members of the Cyclone Global Navigation Satellite System mission’s extended science team. The authors are grateful to the CYGNSS team for providing the CYGNSS L0 raw data, and making the L1 data publicly available through the NASA EOSDIS Physical Oceanography Distributed Active Archive Center(PO.DAAC)(http://dx.doi.org/10.5067/CYGNS-L1X21). The authors would like to thank Dr. Rengui Ruan from Xi’an Research Institute of Surveying and Mapping for providing the BeiDou-3 precision orbit. The Sentinel-2 optical images are available at the Copernicus Open Access Hub website(https://scihub.copernicus.eu/). The GOES-16 satellite image are provided by the National Oceanic and Atmospheric Administration (NOAA) through the Amazon Web Services(Amazon Resource Name: arn:aws:s3:::noaa-goes16).