GNSS-Reflectometry (GNSS-R) is an innovative remote sensing technique that exploits reflected signals from Global Navigation Satellite Systems (GNSS) to retrieve geophysical information about the Earth surface such as near surface ocean winds. Based on bistatic scattering, navigation signals emitted by GNSS satellites such as GPS and Galileo in Medium Earth Orbit (MEO) are received by a GNSS-R receiver on a Low Earth Orbit (LEO) observatory, that features both a zenith-pointing antenna to receive the direct signal from the GNSS constellation and a nadir-pointing antenna to acquire the earth-reflected signal. The reflected signal originates from a glistening zone on the ocean surface sited around the Specular Point (SP), the geometrical point on the Earth surface where GNSS signals are forward scattered in the specular direction. The direct and reflected signals are then correlated within the receiver for different shifts in time (delay) and frequency (Doppler) relative to the specular point (SP) to produce a so-called Delay Doppler Map (DDM) of forward scattered electromagnetic power from the surface.

A geophysical inversion algorithm denoted Calibrated Bistatic Radar Equation (C-BRE) was developed at the National Oceanography Centre (NOC) for the retrieval of ocean surface winds from spaceborne GNSS-R data. Using global GNSS-R data from the UK TechDemoSat-1 (TDS-1) mission, the C-BRE wind processor leads to retrieved ocean wind speeds that agree with ASCAT scatterometers with an RMSE a ~2m/s in the 0-20m/s range. This study presents some recent refinements introduced in the retrieval of wind speed for TDS-1, including improvements in aspects related to routine data quality control and automatic rejection of anomalous DDM. The TDS-1 quality control system now includes sea-ice flagging based on a novel sea ice detection algorithm based on the shape of the GNSS-R waveform. The introduction of this new sea-ice flagging module has enabled extension of the geograghical coverage of TDS-1 wind speed products to higher latitudes, and provides the C-BRE processor with the ability to produce truly global GNSS-R ocean winds, including over the polar oceans. It was also clear from previous work that some regions of the globe are particularly prone to contamination from man-made radiofrequency interference (RFI). Building on existing RFI detection methods relying on the statistical properties of the DDM noise, an entirely novel approach was designed, developed and validated at the NOC, showing promising results for the automatic detection and removal of RFI-contaminated GNSS-R data. Finally, an improved correction strategy is presented in this paper, to account for the temporal evolution of available direct power output by different GNSS transmitters.

On the other hand, it is well-known that other parameters (sea state, ionosphere, SST, SSS, etc), play an important role on the received GNSS-R signal. More work is underway to quantify the contributions from these effects, in order to mitigate them and further improve the accuracy of GNSS-R wind retrievals.

Area D - Board 362

The NASA CYGNSS Mission [1] is a constellation of 8 microsatellites, launched in December 2016, which collects GPS reflections over both oceans and land, and within a latitudinal range of approximately +-38°.
CYGNSS data are primarily used for wind speed estimation, and wind speed products from CYGNSS are routinely available from https://podaac.jpl.nasa.gov. The methodology to derive winds from CYGNSS data, illustrated in detail in [2], relies on the use of two observables defined from CYGNSS Delay-Doppler Maps, and called Delay/Doppler Map Average (DDMA) and Leading Edge Slope (LES), and on the derivation of geophysical model functions (GMFs) to map each value of the observable into a wind speed estimation. Then, the winds retrieved from DDMA and LES are optimally combined into a Minimum Variance (MV) estimator, which deliver the final wind speed estimate.
The drawback of the current baseline approach is that the GMFs are derived in a very empirical manner, and as such the empirical parameters that determine the GMF derivation need to be re-tuned every time the observable changes, or a new release of CYGNSS data occurs.
Here, we propose a new approach for the derivation of GMF from any GNSS-R observable, entirely based on the statistics of the observable itself. The method has been illustrated for the first time back in 1993 in [3], to derive a wind speed geophysical model function from altimetry data, and it is in fact a method suitable to be applied to a wide range of remote sensing techniques, and as such also to GNSS-R. The approach relies on the assumption that the observable varies monotonically and uniquely with wind speed, and it requires knowledge of the separate, rather than joint, distribution functions of the observable and of the wind speed. The wind speed GMF in particular is solely determined by the Cumulative Distribution Function (CDF) of the observable, and by the inverse CDF for wind speeds [3]. The method provides improved wind speed retrievals, compared to those obtained from the baseline GMF derived as in [2]. The method has been tested using both DDMA and LES, and using wind speeds output from the ECMWF model. It provides unbiased retrieved winds, whose probability density function matches perfectly the pdf of true wind speeds. The 2D distribution of retrieved winds versus wind speeds is much more symmetrical around the 1:1 line, compared to the baseline winds.
The agreement with true winds is particularly good in the low to medium wind speed range, since the method is driven by the statistics of the observable and of the winds, and the low-medium winds constitute the majority of the population. As a consequence of that, the performances are slightly degraded at high winds (i.e. higher than 12 m/s), compared to the baseline approach.
Despite this, the proposed methodology delivers overall improved wind speed estimates compared to the baseline winds, and should be considered as a replacement of the baseline approach, for future releases of CYGNSS wind products.

[1] Ruf, C., R. Atlas, P. Chang, M. Clarizia, J. Garrison, S. Gleason, S. Katzberg, Z. Jelenak, J. Johnson, S. Majumdar, A. O’Brien, D. Posselt, A. Ridley, R. Rose, V. Zavorotny (2015), “New Ocean Winds Satellite Mission to Probe Hurricanes and Tropical Convection,” Bulletin of American Meteorological Society, doi:10.1175/BAMS-D-14- 00218.1.

[2] Ruf, C. and R. Balasubramaniam, “Development of the CYGNSS Geophysical Model Function for Wind Speed”, IEEE JSTARS, May 2018, DOI: 10.1109/JSTARS.2018.2833075.

[3] Freilich, M. and P. Challenor, ”A New Approch for determining Fully Empirical Altimeter Wind Speed Model Functions”, Journal of Geophysical Research, Vol. 99, No. C12, December 1994.

Area D - Board 363

Previous studies, using observations of outgoing longwave radiation and products of operational numerical weather prediction, found that intraseasonal oscillation (ISO) is closely related with the active/break periods in the Indian summer monsoon. Relevant surface parameters, such as wind speed, are obscured by rain. CYGNSS data has sufficient sampling and allows us to examine intraseasonal oscillation of ocean surface wind speed under rainy conditions of the organized convections in the summer monsoon. One of the distinguished characteristics of the summer ISO is northward propagation in the Bay of Bengal, besides the equatorial eastward-propagating Madden and Julian Oscillation. In this study, the capability of CYGNSS measurement in observing the prominent features of ISO in the Indian summer monsoon is demonstrated. Along with other satellite and model output data: precipitation from the Tropical Rainfall Measuring Mission (TRMM) and the Global Precipitation Measurement mission (GPM), sea surface temperature (SST) and computed latent heat flux from the Advanced Microwave Scanning Radiometer -2 (AMSR-2), shortwave radiation from the International Satellite Cloud Climatology Project (ISCCP), and mixed layer depth from Estimating the Circulation and Climate of the Ocean (ECCO), we examined the physical mechanisms/hypotheses associated with ISO in the Indian summer monsoon region.

In the life cycle of ISO, the CYGNSS wind speed shows coherent spatial structures, with positive correlation with rain and negative relation with SST. That leads to stronger/weaker evaporation over colder/warmer water. It is the manifestation of wind-evaporation feedback that sustains the summer ISO. Out-of-phase relation between rain and shortwave radiation and in-phase relation of shortwave radiation and temperature tendency indicate the rain/cloud-radiation-SST feedback in sustaining ISO in the Indian Ocean. Shallower/deeper mixed layer depth linked with warmer/colder water indicates the ocean stratification and hydrodynamics are also important factors for ISO in the Indian summer monsoon. Further studies are needed to better understand the relative importance of various mechanisms under different climate regimes and seasons, and relation with interannual variations.

Area D - Board 364

Wind speed retrieval from satellite Global Navigation Satellite System Reflectometry (GNSS-R) is expected to notably impact weather forecasts in the upcoming future. For such purpose, GNSS-R wind speed should be retrieved with an accuracy equal to or better than 2 m/s in the range 0-20 m/s. To the best of our knowledge this performance is achieved only in the range 0-7 m/s, while in the remaining range, accuracy can be even worse than 2.5 m/s, with a consequent poor impact on data assimilation in Numerical Weather Prediction (NWP) models. The reasons for such poor accuracy are various, among which the inability to properly exploit the entire information content of the delay Doppler map (DDM) plays an important role.
It has been recently demonstrated that satellite DDMs acquired by the UK TechDemo-Sat 1 (TDS-1) mission are affected by some distortions that are strictly correlated to specular point estimation inaccuracies. In this study, it is shown that also the DDMs acquired by the US satellite mission Cyclone GNSS (CYGNSS) are affected by similar distortions.
The main aims of this study are a) demonstrating the basic mechanism that causes such distortions, b) assessing their impact on the DDM observables that are operationally used in the wind speed retrieval algorithms, notably the DDM peak value for TDS-1 and the DDM average (DDMA) and leading edge slope (LES) for CYGNSS and c) analyzing their effects on the stare processing approach for wind speed retrieval.
For such purpose, a set of CYGNSS tracks operated in the so-called “science mode” have been re-compressed in the DDM format by using the Doppler frequency at the specular point estimated position output with a more accurate geoid model (provided in the metadata of the regular CYGNSS files). The science mode consists of recording the entire raw reflected GNSS-R echo, which is routinely discarded in the operational acquisition mode, and therefore not available for reprocessing. This mode is typically switched on for a short time interval (around 60 seconds), and only in some special cases such as extreme events.
The tracks for this study have been acquired over the hurricanes Harvey (25th of August 2017) and Irma (8th of September 2017). These tracks have been collocated with the operational ECMWF forecast model wind field output.
Our preliminary results show that distortions are caused by a DDM sampling problem. Indeed, the on-board algorithm for the estimation of the specular point location, based on the quasi-spherical (QS) approximation of the Earth surface, introduces inaccuracies on the 0-Doppler reference value that may reach few hundred Hz, which amount to a non-negligible fraction of the DDM Doppler bin size (500 Hz). Such distortions cause an asymmetry on the arrival time between the positive and the negative Doppler DDM wave-forms (WFs). Furthermore, it is here demonstrated that such distortions also significantly impact the estimation of the above mentioned observables. A comparison between the regular observables and those obtained by reprocessing the reflected echo with the more accurate 0-Doppler frequency at the specular point is presented and discussed. Finally, the implications of the reported distortions on the stare processing approach used for wind speed retrieval are discussed.

Area D - Board 365

The Cyclone Global Navigation Satellite System (CYGNSS) is a micro-satellite constellation that measures GNSS reflections to estimate ocean wind speed. The fundamental measurement of CYGNSS is a 17 x 11 Delay-Doppler Map (DDM) which has 17 delay bins and 11 Doppler bins, representing the convolution between the total scattered power from the ocean surface and the ambiguity function of the GPS transmitted signal. The baseline wind speed retrieval algorithm of CYGNSS uses only a 3 x 5 window of DDM power bins near the specular point bin (center of reflection) to estimate wind speed, i.e., CYGNSS Level 2 products. Assimilation of CYGNSS Level 2 wind speed may lose much potential information compared to assimilation of the whole 17 x 11 DDM.

Our approach is the direct assimilation of whole DDMs into atmospheric models to utilize all information from the measurement. A GNSS-R DDM forward model is first developed to link the wind speed and the measurement. Taking in a gridded wind field, the forward model can produce a simulated DDM in the same delay-Doppler coordinate of the observed DDM and a Jacobian matrix representing the sensitivity of the DDM power respect to the wind speed in each of the 17 x 11 bins.

Then a two-dimensional variational analysis method (VAM) is used to assimilate the DDMs into a 2-D vector wind field via integration of the forward model and application of the Jacobian to modify the background, or a priori, input vector wind field. In the VAM, the wind speed analysis is optimized by minimizing a cost function including observation, background, and dynamic constraints. In our experiments, the Hurricane Weather Research and Forecasting (HWRF) model supplies the background wind field. Different levels of data quality control are conducted before and during the assimilation. A by-product of the VAM analyses of DDMs is estimated wind speed and vector winds at CYGNSS observation locations. Results of the assimilation of the CYGNSS Level 1 DDM power will be assessed though (a) comparison with CYGNSS Level 2 wind speed (baseline wind speed retrieval), and (b) Observing System Experiments (OSEs) with HWRF that compare the impacts of assimilating baseline retrieved wind speed, VAM-derived wind speed, and VAM-derived wind vectors.

Area D - Board 366

The scientific use of GNSS signals in Earth observation has grown since more than two decades. A recent field of research lies in the application of reflected GNSS signals for remote sensing of the Earth surface and for ocean altimetry. The aspect of reflected signals (reflectometry) is included in the ADVANTAGE project which is an initiative of the German Helmholtz Association. This project investigates challenges in a next generation GNSS putting emphasis on navigation technologies and new prospects for Earth observation. The reflectometry part focuses on the amplitude of the coherently reflected signal with twofold objectives. On the one hand surface parameters, for example over sea ice, shall be estimated from the permittivity response of the amplitude. On the other hand a reliable prediction of the coherent signal amplitude is important for precise altimetric retrievals in space-based scenarios.
In a first step a pilot study has been conducted within the ADVANTAGE project. The amplitude is estimated using GNSS reflection data gathered during ship measurements in Arctic sea ice. Signals at grazing elevation angles (< 30deg) are considered. Power ratios are retrieved from the amplitude estimates at left- and right-handed polarizations. In particular the cross-polar ratio allows to invert sea ice concentration along the ship's track. The retrieved power ratios follow the predicted reflectivity under sea ice conditions. Deviations occur at lowest elevation angles (< 5deg), see Fig. below.

Fig.: Elevation dependent distribution of estimated power ratios from Arctic ship measurements. Predictions for bulk water and bulk multi-year ice are given in solid and dotted red lines, respectively.

In a second step the processor is currently adapted to use satellite data for estimation of signal amplitude and sea ice concentration. A case study investigates example events over Arctic sea ice recorded by the UK Technical Demonstration Satellite TDS-1. The second step also comprises predictions of the coherent reflection amplitude considering opportunities of the next generation GNSS. The influence of geometric parameters is investigated: receiver/transmitter orbit heights, the reflection angle and the required spatial resolution. Furthermore, the influence of surface parameters, particularly permittivity and roughness, are taken into account. The accuracy of predicted reflection amplitudes will be assessed based on the pilot/case studies. The results are important to consider coherent reflectometry for next generation GNSS in the ADVANTAGE project. Furthermore the current investigations contribute to prepare for dedicated satellite measurements. Such measurements are an essential objective of the PRETTY (Passive ReflecTomeTrY) mission and the G-TERN (GNSS Transpolar Earth Reflectometry moNitoring system) concept, both proposed to ESA.

Area D - Board 367

Ground-based GNSS multipath reflectometry is a promising alternative for measuring the sea level anomaly close to coastal GNSS stations, with heights given directly in an ellipsoidal reference frame compared to tide gauges, which are normally given in a local reference frame. In the ESA project, GOCE++Dynamic ocean topography (DYCOT), we processed GNSS data at more than 100 stations globally distributed. For most stations, data is available in the period 2016 to 2018.
To validate the GNSS based time series we compare with SAR mode satellite altimetry from the Sentinel-3 and from the CryoSat-2 missions. For each GNSS station, altimetry data in a box of 20 by 20 km is extracted if available. When deriving the sea level anomaly from both data sets a careful evaluation of the applied corrections must be performed to ensure a qualitative comparison. In addition, the use of consistent tide systems is important so as not to introduce biases.
Time series of the sea level anomaly from both data sets are generated with the R software package "tsHydro" https://github.com/cavios/tshydro. The tsHydro package is based on a state-space model with an error distribution that follows a mixture between the Gaussian and Cauchy distributions allowing the model to be robust against erroneous data. The latter is especially important for altimetry data, where outliers can be present near the coast due to land contamination of the signal. With the modeled time series, we can predict the GNSS sea levels anomaly at the times where the altimetry data is available allowing for an easy comparison
A preliminary evaluation of the comparison with Sentinel-3A altimetry data at 42 stations reveals an overall good agreement at most stations. For summary statistics, we calculated the mean squared Error (MSE) based on both the absolute and relative differences for all the time series. To estimate the relative MSE the mean sea level anomaly was removed from the individual data sets, hence centering the anomaly around zero. The Absolute and relative MSE are in the interval of 0-0.4 m and 0-0.05 m, respectively, indicating that there is a bias between the two data sets. The origin of this bias is still to be investigated.

Area D - Board 368

Many environmental processes that occur over northern land areas and at high elevations are influenced by the seasonal transition between predominantly frozen and non-frozen soil conditions. Therefore, a synoptic monitoring of the freeze/thaw (FT) state (which describes the transition of the landscape between frozen and thawed conditions) can be of fundamental importance for many applications (e.g., those related to climate, hydrological, ecological and biogeochemical processes). Spaceborne remote sensing can help in achieving such objective. However, FT state is a variable highly heterogeneous, both in time and in space. This is due to the contribution of different aspects that influence it, such as: land cover, moisture status, slope, aspect and microclimate (Chew et al., 2017; Derksen et al., 2017). Therefore, its accurate monitoring is not a trivial task. Most of the existing spaceborne remote sensing technologies prevent FT monitoring at high spatio-temporal resolution. Active microwave sensors can provide observations at high spatial resolution (~100 m) and low temporal resolution (> 7 days). Passive microwave sensors, instead, can provide observations at high temporal resolution (1-3 days) but low spatial resolution (> 25 km) (Chew et al., 2017). Therefore, it is worth investigating new remote sensing technologies for improving synoptic FT monitoring performances. Since recent studies have shown the potential of exploiting Global Navigation Satellite System (GNSS) Reflectometry (GNSS-R) for observing FT state (Wu and Jin, 2014; Chew et al., 2017), this work aims at evaluating the potentialities of the GNSS-R data collected by the TechDemoSat-1 (TDS-1) satellite for FT detection. The long-term objective is the assessment of the potential retained by spaceborne GNSS-R for FT monitoring. The idea is to support and foster the development of future GNSS-R satellite missions.

To this aim, two observables derived from the TDS-1 satellite (Level 1B products), i.e., the signal to noise ratio (SNR) and the surface reflectivity, were analyzed. Because of the reduction of the dielectric constant in frozen soil conditions, both SNR and surface reflectivity are expected to decrease when the soil freezes (Chew et al., 2017). The study area selected for the analysis was Siberia (Lat. Min.=45°N; Lat. Max.=80°N; Lon. Min.=50°E; Lon. Max.= 180°E); the time period ranged from April 2015 to July 2017. The experiment was structured as follows. Initially, the study area was divided in selected, homogeneous, regions characterized by the same (macro) land cover classification (Open Shrublands, Savannas, Forest Deciduous, Forest Evergreen, Mixed Forest, Croplands, Grasslands). The MODIS land cover data product (IGBP Land Cover Type Classification) was used for identifying such regions. Internal and open water bodies were masked out, as they can produce anomalously high reflected signals. Since the land cover classification of Siberia is strongly depended on the latitude, each homogeneous land cover region was constrained within a continuous range of latitude. Such ranges were, approximately: 60°N-75°N for Open Shrublands, 63°N-68°N for Savannas, 58°N-65°N for Forest Deciduous, 60°N-65°N for Forest Evergreen, 50°N-63°N for Mixed Forest, 51°N-57°N for Croplands, 45°N-53°N for Grasslands. This division was required to reduce the uncertainty due to different land cover classes and latitudes on the FT detection. Subsequently, the monthly average of the two TDS-1 observables was computed for each land cover class and compared against the average value of the SMAP FT product, thus comparing data by performing a “coarse” spatio-temporal collocation. More specifically, for the comparison it was used the “freeze thaw fraction” of the SMAP L2 Radiometer Half-Orbit 36 km EASE-Grid Soil Moisture product.

Preliminary results showed a strong agreement between the time series trends of the TDS-1 observables and the SMAP FT product, except for the southernmost land cover class (i.e., Grassland) that, instead, showed noisy TDS-1 trends. This can be explained by the less marked discrimination of freeze and thaw conditions in areas located at lower latitudes. Nevertheless, this outcome provided evidences about the strong sensitivity of the GNSS-R data to FT state for different land cover classes; notwithstanding the limited availability of the TDS-1 observations. In fact, during the time period under analysis, the TDS-1 satellite acquired data only for few days every month (only since the end of 2017 TDS-1 is collecting data continuously). This fact reduced the number of the GNSS-R observations that were available for the analysis. Moreover, it must be pointed out that the SMAP data have a spatial resolution of 36 km. The GNSS-R signal over land, instead, is mainly due to the coherent reflection from the surface (at least for flat surface) and it originates mainly from the first Fresnel zone. For a spaceborne receiver, the latter has a dimension of the order of few kilometers. The impact of the different spatial resolution of the TDS-1 and SMAP data when discriminating FT conditions (both at lower and at higher latitudes) has to be better understood; as well as the impact of their different acquisition geometry (i.e., images for SMAP; tracks of specular points for TDS-1) and their different number of observations. Future analyses will be thus carried out to investigate these topics. In this exercise, the computation of the monthly average values over the macro land cover classes was required to allow a “fair” comparison between the TDS-1 and SMAP data, reducing the aforementioned differences between the two datasets. Further analyses will be also carried out in the future by using longer and denser time series of TDS-1 data, that will be compared against the SMAP FT product by performing an accurate temporal and spatial collocation.

References
- Chew, C.; Lowe, S.; Parazoo, N.; Esterhuizen, S.; Oveisgharan, S.; Podest, E.; Zuffada, C.; Freedman, A. SMAP radar receiver measures land surface freeze/thaw state through capture of forward-scattered L-band signals. Remote Sens. Environ. 2017, 198, 333–344, doi:10.1016/j.rse.2017.06.020.
- Derksen, C.; Xu, X.; Scott Dunbar, R.; Colliander, A.; Kim, Y.; Kimball, J. S.; Black, T. A.; Euskirchen, E.; Langlois, A.; Loranty, M. M.; Marsh, P.; Rautiainen, K.; Roy, A.; Royer, A.; Stephens, J. Retrieving landscape freeze/thaw state from Soil Moisture Active Passive (SMAP) radar and radiometer measurements. Remote Sens. Environ. 2017, 194, 48–62, doi:10.1016/j.rse.2017.03.007.
- Wu, X., Jin, S., 2014. Can we monitor the bare soil freeze-thaw process using GNSS-R? A simulation study. Proceedings SPIE 9264, Earth Observing Missions and Sensors: De- velopment, Implementation, and Characterization III. 9264, pp. 1–6.

Area D - Board 369

PRETTY (Passive REflecTometry and dosimeTrY) is a small satellite project of the European Space Agency (ESA). The CubeSat (Figure), which will be built by TU Graz and RUAG Space, is a nanosatellite measuring 30 by 10 by 10 centimetres excluding the foldable solar arrays. PRETTY includes a demonstrator payload for a radiation dosimeter and a passive GNSS based reflectometer (GNSS-R) focusing on low elevation angles whereby the direct and reflected signal will be received via the same antenna. The main scientific goal is the precise altimetric determination of water and ice surfaces using the interferometric GNSS-R (iGNSS-R) phase-delay altimetry approach. This approach is appropriate to be applied aboard small satellites for potential low-cost GNSS remote sensing constellations. It also provides increased spatiotemporal resolution of the altimetric observations of the water and ice surfaces compared to near nadir measurements (group-delay GNSS-R altimetry).

The phase-delay altimetric approach at grazing incidence angles was demonstrated initially using GPS radio occultation data from GPS/MET and CHAMP. These observations were achieved with low-gain antennas and state-of-the-art GPS radio occultation receivers without any specific optimization for GNSS reflectometry. A case study of phase data from CHAMP presented an altimetric retrieval over the Greenland ice sheet and sea ice. The derived surface heights had 0.7 m precision in 0.2 second averaging (1 km resolution). This approach was recently also proposed within new spaceborne GNSS Reflectometry experiments, as, e.g., GEROS-ISS or G-TERN for ocean and ice remote sensing, including surface altimetry for elevation angles at the specular point between 5° and 30°. Both missions also include group-delay near-nadir reflectometry observations. The geometry of these measurements is better suited for altimetry, as it minimizes the impact of any delay uncertainty or miscorrection into the vertical (altimetric) retrieval. However, the scattering of GNSS signals over the sea surface at near-nadir angles of incidence is essentially diffuse, only group-delay measurements can be conducted, of worse precision than phase-delay observations. Phase-delay altimetric simulations have been performed for ocean applications within the GEROS-ISS related scientific ESA study GARCA to optimize the observation and data analysis strategy. The simulation results show that phase-delay altimetric retrievals are sensitive to anomalies of the ocean topography and that an altimetric precision of 10 cm in 1 second observation is possible in this respect. Similar precision was demonstrated with airborne experiments. Recent Observation System Simulation Experiments (OSSE) indicated the large potential of future phase-delay GNSS-R data to significantly improve ocean modelling and forecast systems.

The PRETTY mission and the observation concept will be introduced. We also inform on the recent status of the project.

Figure: The PRETTY satellite for precise phase altimetry (Copyright: TUG).

Area D - Board 370

The use of Global Navigation Satellite Systems (GNSS) such as the Global Positioning System (GPS) as a source of opportunistic signals for bistatic radar observations, also known as GNSS reflectometry (GNSS-R) has become since its foundations a mature technique for remote sensing of the earth surface. With several spaceborne missions currently flying and able to measure GNSS-R signals reflected off the Earth Surface, including the Techdemosat-1 demonstrator, the recently launched dedicated CYGNSS constellation, the field is currently opening up to global systematic observations, and therefore a good understanding of the interactions of GNSS signals with the land surfaces is crucial. For this purpose, the GLObal navigation satellite system Reflectometry INstrument (GLORI) (Motte et al., 2016) was developed at CESBIO and used in a campaign performed in June-July 2015 in order to further investigate the sensitivity and interactions of the GNSS-R signals with land surface. In this paper we are presenting results from this campaign over agricultural and forest areas.
The airborne GNSS-R data were collected with the GLORI Instrument during the GLORI 2015 Campaign. Five scientific flights were performed over the southwest of France on a two weeks span, in order to cover several agricultural and forest areas of interest. The aircraft speed was ~100m/s and the altitude above ground was ~600m. The antenna radiation pattern allowed for measurements ranging between 45 and 90º of elevation. Intensive ground truth measurements were performed in collocation with the flights over agricultural and forest areas. For agricultural areas, the following variables were monitored over twenty agricultural reference fields (crop type, soil roughness, soil moisture, vegetation height, vegetation water content). For forest areas, measurements were performed in collocation with the flights over two zones of interest, Nezer and Tagon-Marcheprime. In these zones more than 100 stands of maritime pine trees were selected. In a dedicated ground truth collection field campaign that took place at the beginning of 2016, the following parameters were measured and/or estimated from allometric equations: Age, Diameter at Breast Height (DBH), Height, Density, Stem Wood Biomass (SWB), Above Ground Biomass (AGB), Above Ground Wood Volume (VAL), Basal Area (G).
For agricultural area, the sensitivity of the GNSS-R measurements to soil moisture content was evaluated through the interpretation of data corresponding to low and high LAI levels, ranging from bare soil to dense cover, and for two elevation intervals (50°-70° and 70°-90°). Three GNSS-R observables (right-left apparent reflectivity, the reflected signal noise ratio SNR, and the polarimetric ratio PR) were found to increase with increasing soil moisture. The sensitivities of reflectivity and reflected SNR to soil moisture were also found to increase with an increasing elevation angle, equal for example to 40 dB/[m3/m3] and 45.7 dB/[m3/m3] for the case of reflectivity and the two elevation intervals (50°-70°) and (70°-90°), respectively. The third observable PR was weakly correlated to the soil moisture. The role of vegetation on the GNSS-R observables dynamic was analyzed using a class of data with approximately the same soil moisture conditions, dry conditions with soil moisture lower than 0.1 m3/m3, revealing a sensitivity to the vegetation LAI, with a clear decrease in all GNSS-R observables with an increasing vegetation LAI. The correlations between the vegetation LAI and both reflectivity and the reflected signal SNR were strongest in the case of low elevation angles. The correlation was very weak for high elevation angles. The sensitivities of these observables were quite similar, equal to -0.91 dB/[m2/m2] and -0.76 dB/[m2/m2] for the case of elevation angle between 50° and 70°, respectively. The PR observable illustrates the high potential for vegetation LAI retrieval, probably due to the high effect of vegetation on GNSS data depolarization.
A simplified tau-omega model was proposed for modeling the reflectivity, observed over reference plots as a function of soil moisture, LAI and elevation angle. Although several different types of plants were considered in this study, the proposed model demonstrated good potential for the simulation of reflectivity data, with RMSE equal to 2.9 dB and R2 equal to 0.64 for GNSS-R data with elevation angles ranging between 50° and 90°. The validation step applied using a 5-fold cross-validation approach illustrated an RMSE equal to 0.06 m3/m3 between retrieved and ground soil moistures values (Zribi et al., 2018).

For forest area, the GNSS-R data recorded over reference stands were used to produce two observables: the apparent right-left reflectivity and the polarimetric reflectivity ratio PR. It is found that these two observables are sensitive to the AGB. The highest sensitivities are observed at high elevation angles. When computed with a 20 ms coherent integration time, these sensitivities are equal to -0.051 dB/[Mg.ha-1] and -0.053 dB/[Mg.ha-1] for reflectivity and PR, respectively. The results presented in this study highlight the potential of GNSS-R for the estimation of biomass in purely resinous plantations such as those of maritime pine forests in the South West of France. For the linear regressions between AGB and GNSS-R observable data, the computed value of RMSE is found to be moderate, and of the same order as that reported in the literature in the case of L-band SAR data, i.e. 29.4 Mg.ha-1 and 37.2 Mg.ha-1 for reflectivity and PR, respectively. The observed influence of integration time on our results confirms previous theoretical studies: in the case of the GLORI airborne campaign configuration, an integration time of 20 ms leads to a higher sensitivity to AGB, a higher AGB correlation coefficient, and a reduced reflectivity RMSE, with respect to the results computed for shorter integration times, i.e. 1 ms and 5 ms, or for longer integration times, i.e. 40 ms and 60ms (Zribi et al., 2019).
Although airborne measurements often vary under different experimental conditions (tree types, crops, topography, etc.), and from one site to another, the present study shows that GNSS-R satellite signals, such as those produced by CYGNSS, can be used to improve the retrieval of soil moisture and biomass at low spatial scales.

References:
Erwan Motte, Mehrez Zribi, Pascal Fanise, José Darrozes, Alejandro Egido, Nicolas Baghdadi, Frédéric Baup, Frédéric Frappart, Pierre-Louis Frison, Dominique Guyon, Thierry Perrin, Jean-Pierre Wigneron, GLORI: A GNSS-R dual polarization Airborne Instrument for Soil Moisture and Vegetation Monitoring, sensors, 2016, 16, 732; doi:10.3390/s16050732.

Mehrez Zribi, Erwan Motte, Nicolas Baghdadi, Frédéric Baup, Sylvia Dayau, Pascal Fanise, Dominique Guyon, Potential Applications of GNSS-R Observations over Agricultural Areas: Results from the GLORI Airborne Campaign, Remote Sensing 10 (8), 1245.

Mehrez Zribi, Dominique Guyon, Erwan Motte, Sylvia Dayau, Jean Pierre Wigneron, Nicolas Baghdadi, Nazzareno Pierdicca, Performance of GNSS-R GLORI data for biomass estimation over the Landes forest, Int J Appl Earth Obs Geoinformation 74, https://doi.org/10.1016/j.jag.2018.09.010, (2019)

Area D - Board 371

Recently Global Navigation Satellite System (GNSS) Reflectometry (GNSS-R) is emerging as an alternative or a complement to classical approaches for Earth remote sensing. GNSS-R techniques can be viewed as a multi-static radar, based on the exploitation of the GNSS signals of opportunity. One of the main advantages that this technique offers, is that with only one passive receiver it is possible to acquire several acquisitions simultaneously, as many as the number of visible GNSS satellites. Initially, GNSS-R was conceived for ocean mesoscale scatterometry and altimetry, but its use has been extended to other fields, such as land applications. Several experiments carried out from aircrafts, towers and balloons, have preliminary demonstrated sensitivity of GNSS-R observations to retrieve land geophysical parameters, such as soil moisture and vegetation biomass.
The European Space Agency is funding a project (ESA/ESTEC CONTRACT n. 4000120299/17/NL/AF/hh) aiming at demonstrating to which extent what we know from models and past ground/airborne experiments on GNSS-R applications for soil moisture and vegetation biomass is: i) still applicable from space, ii) which additional phenomena one should try to model/explain, iii) which final bio-geophysical products one can deliver and how (with recommendations for future systems).
The project involves a large team that is looking at the problem from different perspectives. After the review of the state of the art, TechDemoSat and CYGNSS data have been pre-processed to generate and compare different observables (SNR and surface reflectivity) that have been co-located in space and time with other sources of soil moisture and biomass data, both at global scale and in specific test areas. We have considered SMAP soil moisture and vegetation optical depth (VOD), SMOS soil moisture (different products having different resolution), as well as in situ soil moisture probe networks. As for biomass, in addition to SMAP VOD we have considered global maps available in the literature, forest height data from spaceborne laser (ICESat), as well as comparison with biomass retrievals from L-band SAR data collected by the ALOS-2 mission over specific sites. Some investigations using airborne GNSS-R data were devoted to understanding the peculiarity of spaceborne data that should be considered to assess their real capability for land applications. For instance, the fluctuations of the signals from different platforms have been compared, and the different behaviours (larger fluctuations of satellite data) have been ascribed to surface inhomogeneity, as well as surface topography and signal coherent/incoherent integration and sampling.
Then, as a first step, the sensitivity of the reflections (including specular coherent reflection and to some extend diffuse incoherent scattering) to soil moisture and forest biomass has been considered and compared to predictions carried out by two different simulators based on rigorous electromagnetic modelling frameworks. A number of updates to the previously developed simulators have been implemented to account for some peculiarities of the satellite case, that poses additional problems due to the low magnitude of the reflected signals and their fluctuations. In particular, we have implemented the effect of topography including a digital elevation model and of the surface inhomogeneity within the discriminated area in both simulators and we have compared their outcomes.
Finally, retrieval approaches have been applied to satellite GNSS-R data to asses the expected accuracy through a comparison with the collocated reference data indicated before. Different approaches have been considered, such as simple multivariate linear regressions or more advanced neural network simulators. It was considered that a single observable from the current GNSS-R mission is not capable to retrieve accurately earth surface parameters due to the complexity of the target (combined effect of moisture, roughness, vegetation) so the algorithms take generally advantage from a multi-sensor approach. For instance, the soil moisture retrieval exploits the knowledge of the vegetation conditions (e.g., through the SMAP VOD or other satellite products), or the biomass retrieval can take advantage from a combination of GNSS-R and L-band SAR data.
At the time of the conference the project will be almost completed so that the presentation will give a review of the most important outcomes.

Area D - Board 372