WO2022005619A2 - Ocean surface wind direction retrieval from reflected radio signals on space-borne platforms - Google Patents

Abstract

A sequence of two or more consecutive DDMs is received. The sequence is calculated over two or more corresponding times from a digital DLOS IF signal, a digital reflected IF signal and known locations of one or more antennas of a receiver, a transmitter, and a specular point (SP). The sequence can be received in real-time or some time after acquisition. A feature vector (x) is calculated that includes two or more wind direction-dependent features of the digital reflected IF signal using the sequence of DDMs. A wind direction ϕ is calculated at the SP from the calculated x and a machine learning model (f(x)) that relates x to ϕ (ϕ = f(x)). The machine learning model f(x) is trained using one or more sets of previously measured data that include measured ϕ values for corresponding sequences of calculated DDMs used to calculate corresponding x values.

Description

OCEAN SURFACE WIND DIRECTION RETRIEVAL FROM REFLECTED RADIO SIGNALS ON SPACE-BORNE PLATFORMS

GOVERNMENT INTEREST

[0001] This invention was made with government support under grant number N68335-19- C-0577 awarded by the U.S. Naval Research Laboratory. The government has certain rights in the invention.

RELATED APPLICATIONS

[0002] This application claims the benefit of U.S. Provisional Patent Application Serial No. 63/026,293, filed on May 18, 2020, the content of which is incorporated by reference herein in its entirety.

TECHNICAL FIELD

[0003] The teachings herein relate to operating a receiver to estimate ocean surface wind direction from a radio frequency (RF) carrier signal that is reflected from the ocean surface. More particularly, the teachings herein relate to systems and methods for estimating wind direction (f) of any wind-driven water surface on the earth from a reflected RF carrier signal. The systems and methods disclosed herein can be performed in conjunction with a processor, controller, microcontroller, or computer system, such as the computer system of Figure 1.

BACKGROUND

[0004] Radio signals transmitted from satellites can be used as a means of opportunistically sensing the signal propagation environment. For example, signals reflected from the Earth’s surface contain information about surface-related geophysical parameters. Estimating such parameters is of significant scientific and commercial interest. Signals that have been leveraged for this purpose include Global Navigation Satellite System (GNSS) signals (e.g., global positioning system (GPS), global navigation satellite system (GLONASS), Galileo, and BeiDou) and communications satellite signals.

[0005] One such application for estimating geophysical parameters is GNSS-reflectometry (GNSS-R), which can be used for remote sensing of ocean surface winds. GNSS receivers on board satellites (e.g., U.K. disaster monitoring constellation (UK-DMC), technology demonstration satellite (TDS-1), and cyclone GNSS (CYGNSS)) have been used to estimate ocean surface wind speed. These systems provide an alternative to scatterometers and radiometers that is low size, weight, power, and cost. The L-band frequency GNSS signals are attenuated little by precipitation, so GNSS-R is well-suited for measuring ocean winds within tropical cyclones and other high-precipitation scenarios.

[0006] However, spaceborne GNSS-R sensors are currently unable to estimate the ocean surface wind direction (i.e., the full vector wind, when combined with wind speed). Doing so can boost the scientific and commercial utility of the GNSS-R system. For example, the wind direction is necessary to identify when a closed circulation of winds forms, marking the start of a tropical cyclone. In general, the vector wind is needed to calculate the wind stress on the ocean surface, which is one of the main contributors to overall ocean circulation.

[0007] Remote sensing using GNSS-R is a form of bistatic radar. In CYGNSS, the transmitter is one of about 30 GPS satellites and the receiver is one of eight CYGNSS spacecraft. The scattered GPS signal power received by the CYGNSS spacecraft comes from the so-called glistening zone, a region of the surface surrounding the specular point. For spaceborne geometry, the size of the glistening zone is on the order of 100 km. The location of the specular point, and thus the location of the measurement being made, is determined by the positions of the transmitter and receiver. In this bistatic geometry, the received signal has been forward-scattered from the surface. In the ocean, higher wind speeds produce increased surface roughness, decreasing the amount of forward-scattered power. This is in contrast to monostatic radar systems, like the advanced scatterometer (ASCAT) and other scatterometers, for which the backscattered radar pulse increases in power for higher levels of surface roughness.

[0008] Figure 2 is an exemplary diagram 200 showing components of a GNSS-R system, upon which embodiments of the present teachings may be implemented. Direct line-of-sight (DLOS) GNSS signals and reflected GNSS signals from earth surface 201 are received by a GNSS- R receiver mounted on low earth orbit (LEO) satellite platform 220. Signal transmitters 210, 211 , 212, 213, and 214 represent the GNSS satellites. The receiver of LEO satellite-based GNSS-R receiver platform 220 includes, for example, two antennas to receive the DLOS signal and the reflected signal, respectively.

[0009] GNSS signals 230, 231, 232, 233, and 234 travel directly from the GNSS satellite transmitters 210, 211, 212, 213, and 214, respectively, to LEO satellite-based GNSS-R receiver platform 220. GNSS signal 230A, for example, travels from the GNSS satellite transmitter 210 to ocean surface 202 on earth 201. GNSS signal 230B results from GNSS signal 230A after being reflected by ocean surface 202 at point 240. GNSS signal 230B travels from ocean surface 202 to LEO satellite-based GNSS-R receiver platform 220. Specular point (SP) 240 represents the location where GNSS signal 230Ais reflected.

[0010] As described above, the current operational GNSSs include GPS, the Galileo navigation system, GLONASS, the BeiDou navigation satellite system, and other regional satellite navigation systems. Signal transmitters 210, 211, 212, 213, and 214 are designed to broadcast radio signals at certain frequencies. For example, current operational GPS satellites broadcast three civil signals simultaneously, i.e., L1C/A, L2C, and L5, at 1575.42 MHz, 1227.6 MHz, and 1176.45 MHz bands, respectively.

[0011] The receiver of platform 220 usually has two or more antennas, a zenith-looking antenna to receive the DLOS GNSS signals 230, 231, 232, 233, and 234 and one or several nadir looking or horizontal-looking antennas to receive GNSS signal 230B and other reflected signals from earth 201.

[0012] The receiver of platform 220 processes GNSS signals usually at two or more frequencies, for example, GPS LI and L2. This means the GNSS signals 230, 230A, 230B, 231, 232, 233, and 234 contain signal components at two or more frequencies. The DLOS signals 230, 231, 232, 233, and 234 are used for the precise orbit determination (POD) of LEO satellite-based platform 220.

[0013] Figure 3 is an exemplary block diagram 300 of a GNSS-R receiver, upon which embodiments of the present teachings may be implemented. Antenna system 301 represents a multi -frequency antenna adapted to signal frequencies, such as GPS LI and L2, with right-hand circular polarization (RHCP). Antenna system 302 represents a multi -frequency antenna adapted to signal frequencies, such as GPS LI and L2, with left-hand circular polarization (LHCP). Antenna system 302 may also be a phased array antenna.

[0014] RF front-end 310 is configured to perform signal conditioning and down-conversions, where the signal spectrum is moved from RF to an intermediate frequency (IF) or a baseband frequency. RF front-end 310 may include one or more signal down-converters (not shown) that can be configured to multiple frequency signals driven by a common local oscillator (not shown). The analog multi -frequency outputs from RF front-end 310 can be digitized and quantized in analog-to-digital converter (ADC) 320.

[0015] The output from ADC 320, i.e ., the digitalized IF or baseband signals, is input to IF signal processing system 330, which is used to estimate the signal parameters of the input IF signal, decode the navigation data bits, and compute receiver position, velocity, and time (PVT) solutions.

[0016] The output from IF signal processing system 330, i.e., signal parameter estimations of both DLOS and reflected signals, the PVT of the receiver platform, and the orbit parameters of the transmitter platform is input to the scientific parameters retrieval module 340.

[0017] Scientific parameters retrieval module 340 is used to retrieve scientific parameters, such as the delay-Doppler map (DDM). The DDM is a measurement of the received scattered signal power as a function of path delay t and Doppler frequency f_{D .} Similar to acquisition in conventional GNSS signal processing, the DDM is produced by correlating the received signal with a local replica across a range of delay and Doppler offsets. DDMs can be modeled using a bistatic radar equation for GNSS-R. The DDM model for fully diffuse scattering from a rough surface is

[0018] Here, A is the glistening zone; p is the position vector of a point on the surface; 71 is the coherent integration time (usually 1 ms); PT is the transmit power; l is the carrier wavelength; GR and GT are the receiver and transmitter antenna gains; RR and RT are the receiver and transmitter ranges. The term c² is the Woodward Ambiguity Function (WAF), which describes the correlation response of the receiver. The influence of the surface characteristics, such as the wind speed v and direction f , is encapsulated in s, the normalized bistatic radar cross-section (NBRCS). A commonly-used model for the NBRCS is

where ¾ is the Fresnel reflection coefficient and q = [qx,q ,qz]^Ti s the scattering vector, with q_± = [ί/ ,ί/n]⁷ These coordinates are given in a local tangent plane coordinate system: the z-axis aligns with local vertical and the x and y axes span the local tangent plane. The scattering vector q is the orientation that a facet on the surface must take to allow specular scattering from the transmitter to the receiver. The wind dependence shows up in the probability density function (PDF) of surface slopes: P{-qi/q₂) = P(s), where 5 ^::::

is the slope vector. A bivariate Gaussian distribution is often used for the slope PDF, with upwind and cross-wind variance chosen according to an empirical model. Because the Gaussian distribution is symmetric about its principal axes, this results in a 180^° ambiguity in the wind direction. This ambiguity can be eschewed by instead using the Gram-Charlier distribution, which introduces skewness in the upwind direction.

[0019] More simply, a plurality of DDMs is measured using retrieval module 340. Each DDM is a function of NBRCS, which, in turn, is a function of a PDF, P(s), that can be used to determine the wind direction. As a result, if the PDF, P(s ), can be determined from the plurality of DDMs, the wind direction can be determined.

[0020] Figure 4 is an exemplary diagram 400 showing the scattering of a satellite signal from an SP and from a point near the SP on a smooth surface. Transmitter satellite 410 transmits a signal that is reflected at SP 401 to receiver satellite 420. The signal is also reflected at point 402 near SP 401. However, the signal reflected from point 402 is not received by receiver satellite 420, because point 402 and SP 401 both lie on a smooth surface.

[0021] Figure 5 is an exemplary diagram 500 showing the scattering of satellite signals from an SP and from multiple points near the SP on a rough surface. Transmitter satellite 510 transmits a signal that is reflected at SP 501 to receiver satellite 520. Due to rough surface 530, points 502, 503, and 504 also reflect the signal to receiver satellite 520. Points close enough to SP 501 that reflect the signal received from transmitter satellite 510 to receiver satellite 520 reside in glistening zone 540. Figure 5 shows that scattering in the glistening zone can provide information about rough surface 530.

[0022] Figure 6 is an exemplary diagram 600 showing the scattering of a satellite signal from an SP and from a point near the SP on a rough surface. Transmitter satellite 610 transmits a signal that is reflected at SP 601 to receiver satellite 620. The signal is also reflected at point 602 near SP 601. The signal reflected from point 602 is also received by receiver satellite 620 due to the slope of the surface at point 602. Figure 6 shows that the surface slope is the parameter that determines whether or not points within the glistening zone of an SP reflect the transmitted signal to receiver satellite 620. Likewise, the slopes of points within the glistening zone of an SP on a rough surface can be determined from the reflected signal received by receiver satellite 620.

[0023] Figure 7 is an exemplary diagram 700 showing the relationship between a wind- driven rough ocean surface and a PDF. Plot 710 of diagram 700 shows ocean wave heights 711 plotted as a function of distance. Ocean wave heights 711 are produced by wind direction 715. Ocean wave heights 711 include positive slopes 712 more often than negative slopes 713. However, negative slopes 713 tend to be steeper than positive slopes 712. Plot 710 shows how the slopes of ocean wave heights 711 are dependent on wind direction 715.

[0024] Plot 720 of diagram 700 shows probability density 721 as a function of the slope of ocean wave heights 711. PDF 721 includes tail 722 that extends in a more negative slope, while peak 723 is at a positive slope. Figure 7 shows how a one-dimensional wind direction produces an asymmetric probability density that is a function of the slopes of ocean wave heights 711. In other words, Figure 7 shows how wind direction can be determined by measuring the slopes of ocean wave heights 711 from scattered signals.

[0025] Figure 8 is an exemplary diagram 800 showing how a two-dimensional (2-D) PDF of ocean wave slopes maps to a contour plot showing wind direction. Plot 810 is a 2-D PDF of surface slopes s_x and s_y. Plot 820 is a contour plot showing the probability densities of plot 810 as contour lines. Using contour plot 820, peak 821 of the 2-D PDF is shown 140° from north. As a result, contour plot 820 reveals that the wind direction, f, is 140° from north, which is in the positive direction of s_y. In other words, from the contour lines of plot 820, the positive slopes of ocean waves are most probable 140° from north.

[0026] Figure 9 is an exemplary diagram 900 showing how a 2-D PDF contour plot maps to scattered power that, in turn, maps to a DDM. Contour plot 910 can be converted to scattered power plot 920. For example, Equation (2) above relates PDF, P(s), to NBRCS. In turn, scattered power plot 920 can be converted to DDM 930. For example, Equation (1) above relates NBRCS to DDM( x,/_D). Likewise, from measured DDM 930, scattered power plot 920 can be determined. In turn, contour plot 910 and then wind direction can be determined from scattered power plot 920. In other words, Figures 8-9 show graphically how wind direction is obtained from a DDM. More specifically, the relationship between contour plot 910 and DDM 930 is described in relation to Figure 10.

[0027] Figure 10 is an exemplary diagram 1000 showing how scattered power in the spatial domain relates to scattered power in the delay -Doppler domain. Plot 1010 depicts the scattered power in the spatial domain and plot 1020 depicts the scattered power in the delay -Doppler domain. The scattered power in the spatial domain (1010) is highest near SP 1011 (the center of the plot) and decreases with distance from SP 1011. Lines 1012 are isolines of path delay (iso-delay) and lines 1013 are isolines of Doppler frequency (iso-Doppler), with the labeled values representing delay and Doppler offsets relative to SP 1011. Plot 1020 shows how the power is mapped to the delay -Doppler domain. Each rectangular delay -Doppler pixel contains the signal power scattered from the region(s) of the surface enclosed by the corresponding iso-delay and iso-Doppler lines. The final DDM also includes the effect of the WAF, not illustrated here, which smears out the scattered power in the delay-Doppler domain so that each pixel also contains contributions from neighboring pixels.

[0028] Except along the ambiguity-free line 1014, there are two distinct surface regions (1015 and 1016) that map to the same delay-Doppler pixel. This two-to-one mapping, which is referred to as the “mapping ambiguity,” has important implications on retrieval algorithms. Another important aspect of the mapping is that the surface area of each region corresponding to a delay-Doppler pixel is not constant — the surface area decreases with distance from SP 1011. The per-pixel physical scattering area must be considered to understand the spatial resolution of a DDM-based measurement. The spatial resolution is also impacted by non-coherent integration time (1 second for CYGNSS, during which SP 1011 moves about 7 km) and the effect of the WAF. These factors, along with the physical scattering area, can be modeled by an effective scattering area. Finally, some pixels in plot 1020, such as those at delays less than zero chips, do not contain physical scattering; however, in an actual DDM, the smearing effect of the WAF means that these pixels contain contributions from neighboring pixels that do contain physical scattering. These non-physical pixels can provide useful information in retrievals.

[0029] Most wind speed and wind direction retrieval algorithms to date are based on processing DDMs. Wind direction retrieval has seen a significant variety of approaches. It has been demonstrated using GNSS-R measurements on airborne platforms. For example, the wind direction has been retrieved with a 20^°RMSE by using an observable related to the asymmetry of the DDM. However, due to the mapping ambiguity and the upwind/downwind ambiguity of the Gaussian PDF for surface slopes, each single-DDM retrieval produces four possible wind directions. Resolving this fourfold ambiguity requires measurements from multiple SPs. It is important to note that this type of ambiguity resolution is not feasible in a spaceborne geometry because SPs corresponding to different GNSS transmitters are spaced hundreds of kilometers apart.

[0030] For a receiver in a spaceborne geometry, wind direction retrieval has proven to be even more difficult. One modeling study has found that the impact of wind direction on purely specular scattering is almost negligible, and that the impact on slightly non-specular scattering (such as the 15 pixel portion of the DDM used in the CYGNSS wind speed retrieval) is likely too small to be of practical use. This study suggests that portions of the DDM farther from the SP should be used for wind direction retrieval.

[0031] One approach is to simply use a larger DDM that includes more pixels that contain contributions from non-specular scattering. In another study, a sensitivity to wind direction was found in simulated DDMs; however, a 200 km wide region of scattering was used to simulate each DDM, resulting in a spatial resolution too coarse to be of practical use. In the first attempted wind direction retrieval from spaceborne measurements, the effective spatial resolution of these retrievals was extremely coarse (larger than 200 km) because the measurements were non- coherently accumulated over 18 seconds (more than 100 km of SP motion) and the DDMs had a large delay and Doppler range. The wind direction retrievals had an RMSE of 30^°, but the statistical significance of this figure is questionable given the small sample size (3 retrievals). More recently, a DDM deconvolution approach has been suggested for wind direction estimation. Again, though, spatial resolution is a concern because of the size of the region considered.

[0032] Another approach has been to produce DDMs in a completely non-specular geometry (i.e., the DDM does not contain the SP). The feasibility of this approach was explored and a retrieval algorithm for simulated DDMs was demonstrated. The results are promising, although the wind direction retrieval relies on a measurement of the corresponding wind speed. The near- specular DDM can be used to estimate the wind speed, but the footprint of this measurement is not exactly coincident with the footprint of the non-specular DDM; the amount of error introduced by the heterogeneity of the surface wind field is unclear.

[0033] Each of the above-mentioned methods is so far unable to resolve wind direction ambiguities from a single measurement. Even more significantly, at least in the context of current practical applicability, is that none of the methods can be applied to CYGNSS LI data. The CYGNSS DDMs have a delay range of [-1, 3] chips and a Doppler range of [-2500, 2500] Hz, and thus lack the non-specular pixels that each of these methods relies on. Furthermore, spatial resolution is a major challenge — to provide useful measurements, a wind direction retrieval algorithm should have a spatial resolution as close to 25 km (the resolution of ASCAT and other scatterometers) as possible.

[0034] As a result, additional systems and methods are needed to resolve wind direction ambiguities from a single measurement.

SUMMARY

[0035] Areceiver, method, and computer program product are disclosed for calculating wind direction (f) of a wind-driven water surface on the earth from a reflected RF carrier signal. The receiver includes one or more antennas, RF front-end circuitry, an ADC, and a processor.

[0036] The one or more antennas receive a DLOS RF signal component and an RF signal component of an RF carrier signal. The RF signal component is reflected from an SP (SP) on a wind-driven water surface of the earth. The RF carrier signal is transmitted from a transmitter located above the surface of the earth.

[0037] The RF front-end circuitry down-converts the DLOS RF signal component to a DLOS IF signal. The RF front-end circuitry also down-converts the RF signal component that is reflected from the SP to a reflected IF signal.

[0038] The ADC converts the DLOS IF signal to a digital DLOS IF signal. The ADC also converts reflected the IF signal to a digital reflected IF signal.

[0039] The processor generates a sequence of two or more consecutive DDMs calculated over two or more corresponding times from the digital DLOS IF signal, the digital reflected IF signal, and known locations of the one or more antennas, the transmitter, and the SP. The processor calculates a feature vector (JC) that includes two or more wind direction-dependent features of the digital reflected IF signal using the sequence of DDMs. Finally, the processor calculates a wind direction f at the SP from the calculated feature vector x and a machine learning model (bc )) that relates v to f (f= /(A)) The machine learning model {fix)) is trained using one or more sets of previously measured data that include measured f values for corresponding sequences of calculated DDMs used to calculate corresponding values.

[0040] While multiple embodiments are disclosed, still other embodiments of the present technology will become apparent to those skilled in the art from the following detailed description, which shows and describes illustrative embodiments of the technology. As will be realized, the technology is capable of modifications in various aspects, all without departing from the scope of the present technology. Accordingly, the drawings and detailed description are to be regarded as illustrative in nature and not restrictive.

[0041] The phrases “in various embodiments,” “in some embodiments,” “according to some embodiments,” “in the embodiments shown,” “in other embodiments,” and the like generally mean the particular feature, structure, or characteristic following the phrase is included in at least one implementation of the present technology and may be included in more than one implementation. In addition, such phrases do not necessarily refer to the same embodiments or different embodiments.

BRIEF DESCRIPTION OF THE DRAWINGS AND APPENDICES

[0042] The skilled artisan will understand that the drawings, described below, are for illustration purposes only. The drawings are not intended to limit the scope of the present teachings in any way.

[0043] Figure 1 is a block diagram that illustrates a computer system, upon which embodiments of the present teachings may be implemented.

[0044] Figure 2 is an exemplary diagram showing components of a GNSS-R system, upon which embodiments of the present teachings may be implemented.

[0045] Figure 3 is an exemplary block diagram of a GNSS-R receiver, upon which embodiments of the present teachings may be implemented.

[0046] Figure 4 is an exemplary diagram showing the scattering of a satellite signal from an SP and from a point near the SP on a smooth surface.

[0047] Figure 5 is an exemplary diagram showing the scattering of satellite signals from an SP and from multiple points near the SP on a rough surface.

[0048] Figure 6 is an exemplary diagram showing the scattering of a satellite signal from an SP and from a point near the SP on a rough surface.

[0049] Figure 7 is an exemplary diagram showing the relationship between a wind-driven rough ocean surface and a PDF.

[0050] Figure 8 is an exemplary diagram showing how a two-dimensional (2-D) PDF of ocean wave slopes maps to a contour plot showing wind direction.

[0051] Figure 9 is an exemplary diagram showing how a 2-D PDF contour plot maps to scattered power that, in turn, maps to a DDM.

[0052] Figure 10 is an exemplary diagram showing how scattered power in the spatial domain relates to scattered power in the delay -Doppler domain.

[0053] Figure 11 is an exemplary diagram showing the scattering of an RF signal at a stare point, P, before the satellites move and the corresponding DDM and contour plots, in accordance with various embodiments.

[0054] Figure 12 is an exemplary diagram showing how stare point P of Figure 11 moves within the DDM and contour plots as the satellites of Figure 11 move over time, in accordance with various embodiments.

[0055] Figure 13 is an exemplary contour plot showing the movement of stare point P of Figures 11 and 12 across the PDF for seven different time values from epoch -3 to epoch 3, in accordance with various embodiments.

[0056] Figure 14 is an exemplary diagram showing ambiguous stare points relative to unambiguous stare point P of Figures 11-13 and how these ambiguous stare points correspond to positions within DDM and contour plots, in accordance with various embodiments.

[0057] Figure 15 is an exemplary contour plot showing the movement of stare point P of Figures 11-13 and the ambiguous stare points of Figure 14 across the PDF for seven different time values from epoch -3 to epoch 3, in accordance with various embodiments. [0058] Figure 16 is an exemplary block diagram showing a system for training, testing, using, and storing a machine learning model, fix), in accordance with various embodiments.

[0059] Figure 17 is an exemplary diagram of a receiver for calculating wind direction (f) of a wind-driven water surface on the earth from a reflected RF carrier signal, in accordance with various embodiments.

[0060] Figure 18 is an exemplary flowchart showing a method for calculating wind direction f of a wind-driven water surface on the earth from a reflected RF carrier signal, in accordance with various embodiments.

[0061] Figure 19 is a schematic diagram of a system that includes one or more distinct software modules that perform a method for calculating wind direction f of a wind-driven water surface on the earth from a reflected RF carrier signal, in accordance with various embodiments.

[0062] Appendix 1 is an exemplary paper describing systems and methods for retrieval of ocean surface wind direction from CYGNSS measurements.

[0063] Appendix 2 is an exemplary presentation describing systems and methods for retrieval of ocean surface wind direction from CYGNSS measurements.

[0064] The drawings have not necessarily been drawn to scale. Similarly, some components and/or operations may be separated into different blocks or combined into a single block for the purposes of discussion of some of the embodiments of the present technology. Moreover, while the technology is amenable to various modifications and alternative forms, specific embodiments have been shown by way of example in the drawings and are described in detail below. The intention, however, is not to limit the technology to the particular embodiments described. On the contrary, the technology is intended to cover all modifications, equivalents, and alternatives falling within the scope of the technology as defined by the appended claims.

DESCRIPTION OF VARIOUS EMBODIMENTS

COMPUTER-IMPLEMENTED SYSTEM

[0065] Figure 1 is a block diagram that illustrates a computer system 100, upon which embodiments of the present teachings may be implemented. Computer system 100 includes a bus 102 or other communication mechanism for communicating information, and a processor 104 coupled with bus 102 for processing information. Computer system 100 also includes a memory 106, which can be a random-access memory (RAM) or other dynamic storage device, coupled to bus 102 for storing instructions to be executed by processor 104. Memory 106 also may be used for storing temporary variables or other intermediate information during execution of instructions to be executed by processor 104. Computer system 100 further includes a read-only memory (ROM) 108 or other static storage device coupled to bus 102 for storing static information and instructions for processor 104. A storage device 110, such as a magnetic disk or optical disk, is provided and coupled to bus 102 for storing information and instructions.

[0066] Computer system 100 may be coupled via bus 102 to a display 112, such as a cathode ray tube (CRT) or liquid crystal display (LCD), for displaying information to a computer user. An input device 114, including alphanumeric and other keys, is coupled to bus 102 for communicating information and command selections to processor 104. Another type of user input device is cursor control 116, such as a mouse, a trackball or cursor direction keys for communicating direction information and command selections to processor 104 and for controlling cursor movement on display 112. This input device typically has two degrees of freedom in two axes, a first axis (i.e., x) and a second axis (i.e., y), that allows the device to specify positions in a plane.

[0067] A computer system 100 can perform the present teachings. Consistent with certain implementations of the present teachings, results are provided by computer system 100 in response to processor 104 executing one or more sequences of one or more instructions contained in memory 106. Such instructions may be read into memory 106 from another computer-readable medium, such as storage device 110. Execution of the sequences of instructions contained in memory 106 causes processor 104 to perform the process described herein. Alternatively, hard wired circuitry may be used in place of or in combination with software instructions to implement the present teachings. Thus, implementations of the present teachings are not limited to any specific combination of hardware circuitry and software.

[0068] In various embodiments, computer system 100 can be connected to one or more other computer systems, like computer system 100, across a network to form a networked system. The network can include a private network or a public network such as the Internet. In the networked system, one or more computer systems can store and serve the data to other computer systems. The one or more computer systems that store and serve the data can be referred to as servers or the cloud, in a cloud computing scenario. The one or more computer systems can include one or more web servers, for example. The other computer systems that send and receive data to and from the servers or the cloud can be referred to as client or cloud devices, for example.

[0069] The terms “computer-readable medium” or “computer program product” as used herein refer to any media that participates in providing instructions to processor 104 for execution. Such a medium may take many forms, including but not limited to, non-volatile media, volatile media, and transmission media. Non-volatile media includes, for example, optical or magnetic disks, such as storage device 110. Volatile media includes dynamic memory, such as memory 106. Transmission media includes coaxial cables, copper wire, and fiber optics, including the wires that comprise bus 102.

[0070] Common forms of computer-readable media or computer program products include, for example, a floppy disk, a flexible disk, hard disk, magnetic tape, or any other magnetic medium, a CD-ROM, digital video disc (DVD), a Blu-ray Disc, any other optical medium, a thumb drive, a memory card, a RAM, PROM, and EPROM, a FLASH-EPROM, any other memory chip or cartridge, or any other tangible medium from which a computer can read.

[0071] Various forms of computer-readable media or computer program products may be involved in carrying one or more sequences of one or more instructions to processor 104 for execution. For example, the instructions may initially be carried on the magnetic disk of a remote computer. The remote computer can load the instructions into its dynamic memory and send the instructions over a telephone line using a modem. A modem local to computer system 100 can receive the data on the telephone line and use an infra-red transmitter to convert the data to an infra-red signal. An infra-red detector coupled to bus 102 can receive the data carried in the infra red signal and place the data on bus 102. Bus 102 carries the data to memory 106, from which processor 104 retrieves and executes the instructions. The instructions received by memory 106 may optionally be stored on storage device 110 either before or after execution by processor 104.

[0072] In accordance with various embodiments, instructions configured to be executed by a processor to perform a method are stored on a computer-readable medium. The computer- readable medium can be a device that stores digital information. For example, a computer- readable medium or a computer program product includes a compact disc read-only memory (CD- ROM) as is known in the art for storing software. The computer-readable medium or computer program product is accessed by a processor suitable for executing instructions configured to be executed.

[0073] The following descriptions of various implementations of the present teachings have been presented for purposes of illustration and description. It is not exhaustive and does not limit the present teachings to the precise form disclosed. Modifications and variations are possible in light of the above teachings or may be acquired from practicing the present teachings. Additionally, the described implementation includes software but the present teachings may be implemented as a combination of hardware and software or in hardware alone. The present teachings may be implemented with both object-oriented and non-object-oriented programming systems.

OCEAN SURFACE WIND DIRECTION MEASUREMENT

[0074] Embodiments of systems and methods for calculating wind direction f of a wind- driven water surface on the earth from a reflected RF carrier signal are provided herein, which includes the accompanying Appendix 1 and Appendix 2. In this detailed description, for purposes of explanation, numerous specific details are set forth to provide a thorough understanding of embodiments of the present invention. One skilled in the art will appreciate, however, that embodiments of the present invention may be practiced without these specific details. In other instances, structures and devices are shown in block diagram form. Furthermore, one skilled in the art can readily appreciate that the specific sequences in which methods are presented and performed are illustrative and it is contemplated that the sequences can be varied and remain within the spirit and scope of embodiments of the present invention.

[0075] Appendix 1 is an exemplary paper describing systems and methods for retrieval of ocean surface wind direction from CYGNSS measurements.

[0076] Appendix 2 is an exemplary presentation describing systems and methods for retrieval of ocean surface wind direction from CYGNSS measurements.

[0077] As described above, radio signals reflected from the earth’s surface can be used to derive the properties of the reflection surface. Scientific parameters retrieved from these reflected radio signals include DDMs. Most wind speed and wind direction retrieval algorithms to date are based on processing DDMs. Wind direction retrieval has seen a significant variety of approaches to this processing. None of these approaches, however, is so far able to resolve wind direction ambiguities from a single measurement. Even more significantly, at least in the context of current practical applicability, is that none of the methods can be applied to CYGNSS LI data. As a result, additional systems and methods are needed to resolve wind direction ambiguities from a single measurement.

[0078] In various embodiments, stare processing of DDMs and a machine learning model for NBRCS are used to resolve wind direction ambiguities from a single measurement. As described above, scatterometric measurements made using GNSS-R receivers rely on the DDM, a measure of the received scattered signal power as a function of path delay and Doppler frequency. In a manner similar to acquisition in conventional GNSS signal processing, the DDM is generated by cross-correlating the received signal with a local replica across a range of delay and Doppler offsets. The DDM depends on the physical surface scattering information, which is determined mainly by the wind speed and direction, as well as: gain and range terms for the transmitter (GNSS satellite) and GNSS-R receiver; the signal structure and the corresponding correlation response of the receiver; and the effect of satellite motion that occurs during the duration of non-coherent integration.

[0079] Furthermore, there is an ambiguous mapping that relates spatial surface scattering to the DDM because there are generally two distinct points on the surface that have the same delay and Doppler value. The size of the DDM used determines the spatial resolution of the measurement. For example, the CYGNSS wind speed retrieval algorithm uses only a small portion of the DDM near the SP (the center of the “glistening zone” from which the signal is scattered towards the receiver) to maintain a 25 km spatial resolution. Studies show that sensitivity to wind direction is only seen in off-specular regions of the DDM, so estimating wind direction from a single DDM while maintaining a reasonable spatial resolution is likely infeasible.

[0080] In various embodiments, a method called stare processing is used to combine information from consecutive DDMs that are typically processed independently. A set of fixed points on the surface is followed as it propagates through delay -Doppler space, providing multiple look angles and probing the wind direction-sensitive periphery of the DDM while maintaining spatial resolution. Stare processing has previously been applied to sequences of simulated DDMs to perform vector wind retrieval. The wind speed and direction were estimated by fitting surface scattering parameters determined through a physical model to the surface scattering parameters sampled from the DDMs.

[0081] For several reasons, this physical model-based approach is difficult to extend to real data. First, the method requires that the model of physical surface scattering accurately represent the true scattering behavior. The model dependence on wind direction is weak and difficult to validate. Second, it is challenging to accurately incorporate more complex effects of how the surface scattering maps to the DDM and the impact of the correlation response of the receiver.

[0082] Machine learning is a powerful tool for learning complex patterns in large amounts of data, and thus provides an opportunity to develop a GNSS-R wind direction retrieval algorithm that does not rely on a physical model. The arrival of full-scale GNSS-R missions with the launch of CYGNSS at the end of 2016 means that the volume of DDM data is sufficient to follow this approach.

[0083] In various embodiments, a wind direction retrieval algorithm uses supervised learning and is enabled by applying stare processing to sequences of DDMs to choose input features that emphasize the wind direction-dependent characteristics of the measurement. This is the first time that wind direction retrieval from real data using spacebome GNSS-R has been demonstrated. The algorithm can be immediately applied to the more than three years of CYGNSS data to produce previously unseen observations of wind direction.

Stare Processing

[0084] In stare processing, a sequence of consecutive DDMs is processed together by following a fixed point or set of fixed points on the surface as it propagates through delay -Doppler space. Doing so allows the wind direction-sensitive periphery of the DDM to be sampled without degrading the spatial resolution.

[0085] Figure 11 is an exemplary diagram 1100 showing the scattering of an RF signal at a stare point, P, before the satellites move and the corresponding DDM and contour plots, in accordance with various embodiments. In Figure 11, at epoch 0, an RF signal sent from transmitting satellite 1110 is scattered from stare point P 1101 to receiving satellite 1120. At epoch 0, stare point P 1101 coincides with the SP. DDM plot 1130 shows the corresponding location of P 1101 at epoch 0. Similarly, contour plot 1140 shows the corresponding location of P 1101 at epoch 0.

[0086] Figure 12 is an exemplary diagram 1200 showing how stare point P of Figure 11 moves within the DDM and contour plots as the satellites of Figure 11 move over time, in accordance with various embodiments. In Figure 12, at epoch 1, transmitting satellite 1110 and receiving satellite 1120 have moved. The radio signal sent from transmitting satellite 1110 and scattered from stare point P 1101 to receiving satellite 1120 is now scattered at a different angle. DDM plot 1130 now shows a different corresponding location of P 1101 at epoch 1. Similarly, contour plot 1140 shows a different corresponding location of P 1101 across the PDF at epoch 1. Note that, at epoch 1, stare point P 1101 and specular point SP 1102 no longer coincide. In other words, stare point P 1101 is only the same as SP 1102 at epoch 0 (this is how stare point P 1101 is defined). At other epochs, SP 1102 is in a different location because the positions of the satellites have changed. SP 1102 moves in physical space from epoch to epoch; stare point P 1101 is fixed (hence the name “stare point”).

[0087] Figure 13 is an exemplary contour plot 1300 showing the movement of stare point P of Figures 11 and 12 across the PDF for seven different time values from epoch -3 to epoch 3, in accordance with various embodiments. Figure 13 shows how stare point P 1101 moves across the PDF as the satellites move. The new positions and line-of-sight velocities of the satellites also result in different delay and Doppler values. Now, samples taken from the corresponding region of the DDM for each epoch can be related to the value of the slope PDF for the new slope. The slope PDF is probed in this way as the satellites continue orbiting, providing a new look angle for each epoch. In other words, the position of the PDF can be determined from the movement of P 1101 across the PDF due to stare processing. As described above, the wind direction can be found from the position of the PDF.

[0088] In various embodiments, stare processing is extended to accommodate multiple stare points as shown in Figure 10. The technique is called ambiguous stare processing in reference to the mapping ambiguity. Except along the ambiguity free line, there is actually a pair of stare points that map into the same delay -Doppler pixel as shown in Figure 10.

[0089] Figure 14 is an exemplary diagram 1400 showing ambiguous stare points relative to unambiguous stare point P of Figures 11-13 and how these ambiguous stare points correspond to positions within DDM and contour plots, in accordance with various embodiments. Contour plot 1410 shows ambiguous stare points A_x, A₂, A₃ , A₄ , , A'₄, A₂' , A'₃ , and A₄' relative to stare point P. The ambiguous stare points are ambiguous in that either point of the pair of points A_t or A'i can map into the same delay-Doppler pixel. DDM plot 1420 shows the pixels corresponding to ambiguous stare points A_x, A₂ , A₃ , and A₄ , or A'₄, A₂ , A₃ , and A ₄. Contour plot 1430 shows the locations of ambiguous stare points A_x, A₂ , A₃ , A₄ , A₄' , A₂ , A₃, and A₄ across the PDF at epoch 0. Again, using stare processing, ambiguous stare points A_x, A₂ , A₃ , and A₄ , orri^, A₂, rib and ^4 can be moved across the PDF.

[0090] Figure 15 is an exemplary contour plot 1500 showing the movement of stare point/¹ of Figures 11-13 and the ambiguous stare points of Figure 14 across the PDF for seven different time values from epoch -3 to epoch 3, in accordance with various embodiments. Figure 15 shows that incorporating the ambiguous stare points into stare processing significantly increases the number of points that can be used to determine the position of the PDF. In other words, incorporating the ambiguous stare points provides a greater diversity in surface slope sampling.

Machine Learning

[0091] In various embodiments, wind direction retrieval is performed using sequences of two or more consecutive DDMs collected by a CYGNSS spacecraft. A large training dataset is generated from DDM sequences that have been collocated with ASCAT vector wind measurements. The ASCAT wind direction is used as ground truth, for example. A neural network is trained using a subset of these collocated sequences to find the relationship y = fix) between a feature vector*, extracted from the CYGNSS data (DDMs and relevant geometrical parameters) via stare processing, and an output y, related to the measured wind direction f provided by ASCAT. The trained model is then be applied to new data. The model is then applied to DDM sequences that are not necessarily collocated with ASCAT, enabling a completely new source of global wind direction measurements.

[0092] Figure 16 is an exemplary block diagram 1600 showing a system for training, testing, storing, and using a machine learning model, fix), in accordance with various embodiments. Pathways 1601 show the movement of collocated measurements through the system. Pathways 1602 show how the collocated measurements are used for training and validating machine learning model 1640. Pathways 1603 show how the machine learning model is stored. Pathways 1604 show how testing data moves through the system. Finally, pathways 1605 show how applied machine learning model 1655 is used to determine the wind direction from new data that is not necessarily collocated with ASCAT.

[0093] Dataset 1610 of collocated sequences is used to train and test the machine learning model 16xx. The collocated sequences consist of DDM sequences collected by a spaceborne GNSS-R receiver and a reference measurement of wind direction, for example. Selection criteria 1615 for the DDM sequences can be established based on algorithm performance goals. For example, selecting only sequences that meet a specified signal -to-noise ratio (SNR) threshold reduces error but also reduces the total number of measurements. The reference measurement of wind direction can be remotely sensed (e.g., from a scatterometer) or in situ (e.g., from a buoy), for example.

[0094] Dataset 1620 is new data that is processed using applied machine learning model 1655. Both Dataset 1610 and dataset 1620 undergo feature extraction 1630. Given the low sensitivity of the measurements to wind direction, especially within the delay and Doppler range available in the CYGNSS DDMs, the success of the retrieval algorithm relies on a carefully- designed feature extraction 1630. The features or parameters extracted include DDM NBRCS values, angle of incidence, SP azimuth, SP speed, and latitude and longitude. After extracting these parameters from the CYGNSS data and applying any necessary encoding, they are concatenated together to form the feature vector v, which is the input to the machine learning model 1640.

[0095] During feature extraction 1630, the normalized bistatic radar cross-section (NBRCS) values are sampled from each DDM in the sequence via stare processing, for example. The satellite geometry associated with each DDM (described by the angle of incidence, SP azimuth, and SP speed) is also included in the feature vector. The retrieval performance can be improved by including additional features. For example, date and time information can be included as features to learn how wind directions tend to vary temporally. In various embodiments, any information that is independently available to the GNSS-R sensor (i.e., not obtained from the reference) can be used as a feature.

[0096] Quality control 1635 is used to remove any measurements that are too small or too large. For example, DDM NBRCS values less than zero or greater than 300 are removed.

[0097] Split dataset function 1636 is used to split collocated dataset 1610 into training, validation, and testing subsets. For example, 60% of dataset 1610 can be allocated for training, 20% for validation, and 20% for testing.

[0098] Before training or validation, one or more features of collocated dataset 1610 are encoded using encoding function 1637. For example, DDM NBRCS values follow an approximately log-normal distribution, the features are encoded as log{ 1 + NBRCS).

[0099] Also, before training or validation, the feature vector, , is standardized along the sample dimension using fit and scaler function 1638. For example, the feature vector can be standardized by removing the mean and scaling to unit variance.

[0100] As described above, machine learning model 1640 is generated or trained using a subset of dataset 1610 of collocated sequences to find the relationship y = f(x) between a feature vector*, extracted from the CYGNSS data (DDMs and relevant geometrical parameters) via stare processing, and an output y, related to the measured wind direction f provided by ASCAT. Figure 16 depicts a neural network for machine learning model 1640, but it is also possible to develop a retrieval algorithm using any suitable supervised learning method. Once the algorithm has been trained, it can be applied to other candidate sequences, providing a completely new source of wind direction observations.

[0101] When machine learning model 1640 is validated, it is saved as stored machine learning model 1650. Stored machine learning model 1650 is the model that results in the smallest mean squared error in the validation dataset (typically within the last 100 epochs of the training period), for example.

[0102] Stored machine learning model 1650 is applied to new or testing data as applied machine learning model 1655. Again, applied machine learning model 1655, fix), produces two- dimensional label vector, y, for example, according to the relationship y = /(*)

[0103] Decode function 1660 scales and decodes label vector, y. Label vector y is scaled and decoded to wind direction angle, f, for example. Wind direction f 1670 is produced for new data. For testing data, error analysis information 1680 is produced by comparing the wind direction calculated from the testing data with previously obtained collocated wind direction f measurements.

Receiver for calculating wind direction rib) from a reflected carrier signal

[0104] Figure 17 is an exemplary diagram 1700 of a receiver for calculating wind direction (f) of a wind-driven water surface on the earth from a reflected RF carrier signal, in accordance with various embodiments. Receiver 1710 includes one or more antennas 1711, RF front-end circuitry 1712, ADC 1713, and processor 1714.

[0105] One or more antennas 1711 receive DLOS RF signal component 1721 and RF signal component 1722 of RF carrier signal 1725. RF signal component 1722 is reflected from SP 1730 on wind-driven water surface 1740 of earth 1750. RF carrier signal 1725 is transmitted from transmitter 1720 located above surface 1740 of earth 1750.

[0106] RF front-end circuitry 1712 down-converts DLOS RF signal component 1721 to DLOS IF signal 1761. RF front-end circuitry 1712 also down-converts RF signal component 1722 that is reflected from point 1730 to reflected IF signal 1762.

[0107] ADC 1713 converts DLOS IF signal 1761 to digital DLOS IF signal 1771. ADC 1713 also converts reflected IF signal 1762 to digital reflected IF signal 1772.

[0108] Processor 1714 is used to receive signals, process signals, produce data, or provide control instructions. Processor 1714 can be part of receiver 1710, as shown in Figure 17, or can be a separate device, for example. Processor 1714 can be, but is not limited to, a controller, a computer, a microprocessor, the computer system of Figure 1, or any device capable of sending, receiving, and processing signals and data.

[0109] Processor 1714 generates a sequence of two or more consecutive DDMs 1780 calculated over two or more corresponding times from digital DLOS IF signal 1771, digital reflected IF signal 1772 and known locations of one or more antennas 1711, transmitter 1720, and point 1730. Processor 1714 calculates a feature vector 1785 (A) that includes two or more wind direction-dependent features of digital reflected IF 1772 signal using sequence of DDMs 1780. Finally, processor 1714 calculates a wind direction f 1795 at point 1730 from calculated 1785 and a machine learning model 1790 (fix )) that relates 1785 to wind direction f 1795 (F= :)). Machine learning model 1790 (bc)) is trained using one or more sets of previously measured data that include measured wind direction f values for corresponding sequences of calculated DDMs used to calculate corresponding feature vector* values. Note that digital DLOS IF signal 1771 is used indirectly in DDM generation. For example, digital DLOS IF signal 1771 is used to compute positions of transmitter 1720 and receiver 1710, from which point 1730 is computed via geometry. Knowing the position (and the corresponding delay/Doppler) of point 1730 is necessary to select the window over which the DDM is generated (for CYGNSS, [-1,3] chips and [-2500, 2500] Hz relative to the SP).

[0110] In various embodiments, one or more antennas 1711 can include a first antenna to receive DLOS RF signal component 1721 and a second antenna to receive RF signal component 1722 that is reflected from point 1730, as shown in Figure 17.

[0111] In various embodiments, the two or more wind direction-dependent features comprise one or more of an NBRCS, an angle of incidence, an azimuth of point 1730, a speed of point 1730, or a latitude and a longitude.

[0112] In various embodiments, the NBRCS is calculated from sequence of DDMs 1780 using stare processing. One or more unambiguous points are used in the stare processing, for example. In various embodiments, one or more pairs of ambiguous points are also used in the stare processing.

[0113] In various embodiments, machine learning model 1790 {fix)) is a neural network model.

[0114] In various embodiments, RF carrier signal 1725 is a GNSS carrier signal and the locations of one or more antennas 1711, transmitter 1720, and point 1730 are determined from information carried by the GNSS carrier signal.

[0115] In various embodiments, RF carrier signal 1725 is a carrier signal of a communications system and the locations of the one or more antennas, the transmitter, and the point are determined from information transmitted separately from ground stations of the communications system.

[0116] In various embodiments, processor 1714 calculates f 1795 in real-time. Method for calculating wind direction (f) from a reflected carrier signal

[0117] Figure 18 is an exemplary flowchart showing a method 1800 for calculating wind direction f of a wind-driven water surface on the earth from a reflected RF carrier signal, in accordance with various embodiments.

[0118] In step 1810 of method 1800, a sequence of two or more consecutive DDMs calculated over two or more corresponding times from a digital DLOS IF signal, a digital reflected IF signal and known locations of one or more antennas of a receiver, a transmitter, and an SP is received using a processor. The one or more antennas receive a DLOS RF signal component and an RF signal component that is reflected from the SP on a wind-driven water surface on the earth of an RF carrier signal transmitted from the transmitter located above the surface of the earth. RF front-end circuitry down-converts the DLOS RF signal component to a DLOS IF signal and down- converts the RF signal component that is reflected from the SP to a reflected IF signal. An ADC converts the DLOS IF signal to the digital DLOS IF signal and converts the reflected IF signal to the digital reflected IF signal. In various embodiments, the sequence is received in real-time. In various alternative embodiments, the sequence is received some time after acquisition.

[0119] In step 1820, a feature vector (A) is calculated that includes two or more wind direction-dependent features of the digital reflected IF signal using the sequence of DDMs using the processor.

[0120] In step 1830, a wind direction f is calculated at the SP from the calculated v and a machine learning model (fix)) that relates v to f (f = fix)) using the processor. The machine learning model fix) is trained using one or more sets of previously measured data that include measured wind direction f values for corresponding sequences of calculated DDMs used to calculate corresponding feature vector values.

Computer program product for calculating wind direction (f) from a reflected carrier signal

[0121] In various embodiments, computer program products include a tangible computer- readable storage medium whose contents include a program with instructions being executed on a processor so as to perform a method for calculating wind direction f of a wind-driven water surface on the earth from a reflected RF carrier signal. This method is performed by a system that includes one or more distinct software modules.

[0122] Figure 19 is a schematic diagram of a system 1900 that includes one or more distinct software modules that perform a method for calculating wind direction f of a wind-driven water surface on the earth from a reflected RF carrier signal, in accordance with various embodiments. System 1900 includes data acquisition module 1910 and analysis module 1920.

[0123] Data acquisition module 1910 receives a sequence of two or more consecutive DDMs calculated over two or more corresponding times from a digital DLOS IF signal, a digital reflected IF signal and known locations of one or more antennas of a receiver, a transmitter, and an SR The one or more antennas receive a DLOS RF signal component and an RF signal component that is reflected from the SP on a wind-driven water surface on the earth of an RF carrier signal transmitted from the transmitter located above the surface of the earth. RF front- end circuitry down-converts the DLOS RF signal component to a DLOS IF signal and down- converts the RF signal component that is reflected from the SP to a reflected IF signal. An ADC converts the DLOS IF signal to the digital DLOS IF signal and converts the reflected IF signal to the digital reflected IF signal.

[0124] Analysis module 1920 calculates feature vector (A) that includes two or more wind direction-dependent features of the digital reflected IF signal using the sequence of DDMs. Analysis module 1920 calculates a wind direction f at the SP from the calculated v and a machine learning model (bc)) that relates v to f (f= /(n)). The machine learning odel fix) is trained using one or more sets of previously measured data that include measured wind direction f values for corresponding sequences of calculated DDMs used to calculate corresponding feature vector x values.

[0125] Unless the context clearly requires otherwise, throughout the description and the claims, the words “comprise,” “comprising,” and the like are to be construed in an inclusive sense, as opposed to an exclusive or exhaustive sense; that is to say, in the sense of “including, but not limited to.” As used herein, the terms “connected,” “coupled,” or any variant thereof means any connection or coupling, either direct or indirect, between two or more elements; the coupling or connection between the elements can be physical, logical, or a combination thereof. Additionally, the words “herein,” “above,” “below,” and words of similar import, when used in this application, refer to this application as a whole and not to any particular portions of this application. Where the context permits, words in the above Detailed Description using the singular or plural number may also include the plural or singular number respectively. The word “or,” in reference to a list of two or more items, covers all of the following interpretations of the word: any of the items in the list, all of the items in the list, and any combination of the items in the list.

[0126] The above Detailed Description of examples of the technology is not intended to be exhaustive or to limit the technology to the precise form disclosed above. While specific examples for the technology are described above for illustrative purposes, various equivalent modifications are possible within the scope of the technology, as those skilled in the relevant art will recognize. For example, while processes or blocks are presented in a given order, alternative implementations may perform routines having steps, or employ systems having blocks, in a different order, and some processes or blocks may be deleted, moved, added, subdivided, combined, and/or modified to provide alternative or sub-combinations. Each of these processes or blocks may be implemented in a variety of different ways. Also, while processes or blocks are at times shown as being performed in series, these processes or blocks may instead be performed or implemented in parallel, or may be performed at different times. Further any specific numbers noted herein are only examples: alternative implementations may employ differing values or ranges.

[0127] The teachings of the technology provided herein can be applied to other systems, not necessarily the system described above. The elements and acts of the various examples described above can be combined to provide further implementations of the technology. Some alternative implementations of the technology may include not only additional elements to those implementations noted above, but also may include fewer elements.

[0128] These and other changes can be made to the technology in light of the above Detailed Description. While the above description describes certain examples of the technology, and describes the best mode contemplated, no matter how detailed the above appears in text, the technology can be practiced in many ways. Details of the system may vary considerably in its specific implementation, while still being encompassed by the technology disclosed herein. As noted above, particular terminology used when describing certain features or aspects of the technology should not be taken to imply that the terminology is being redefined herein to be restricted to any specific characteristics, features, or aspects of the technology with which that terminology is associated. In general, the terms used in the following claims should not be construed to limit the technology to the specific examples disclosed in the specification, unless the above Detailed Description section explicitly defines such terms. Accordingly, the actual scope of the technology encompasses not only the disclosed examples, but also all equivalent ways of practicing or implementing the technology under the claims.

[0129] To reduce the number of claims, certain aspects of the technology are presented below in certain claim forms, but the applicant contemplates the various aspects of the technology in any number of claim forms. For example, while only one aspect of the technology is recited as a computer-readable medium claim, other aspects may likewise be embodied as a computer- readable medium claim, or in other forms, such as being embodied in a means-plus-function claim. Any claims intended to be treated under 35 U.S.C. § 112(f) will begin with the words “means for,” but use of the term “for” in any other context is not intended to invoke treatment under 35 U.S.C. § 112(f). Accordingly, the applicant reserves the right to pursue additional claims after filing this application to pursue such additional claim forms, in either this application or in a continuing application.

Claims

WHAT IS CLAIMED IS:

1. A receiver for calculating wind direction (f) of a wind-driven water surface on the earth from a reflected radio frequency (RF) carrier signal, comprising: one or more antennas that receive a direct line-of-sight (DLOS) RF signal component and an RF signal component that is reflected from a specular point (SP) on a wind-driven water surface on the earth of an RF carrier signal transmitted from a transmitter located above the surface of the earth;

RF front-end circuitry that down-converts the DLOS RF signal component to a DLOS intermediate frequency (IF) signal and down-converts the RF signal component that is reflected from the SP to a reflected IF signal; an analog-to-digital converter (ADC) that converts the DLOS IF signal to a digital DLOS IF signal and converts the reflected IF signal to a digital reflected IF signal; and a processor that generates a sequence of two or more consecutive delay -Doppler maps (DDMs) calculated over two or more corresponding times from the digital DLOS IF signal, the digital reflected IF signal and known locations of the one or more antennas, the transmitter, and the SP, calculates a feature vector (v) that includes two or more wind direction-dependent features of the digital reflected IF signal using the sequence of DDMs, and calculates a f at the SP from the calculated v and a machine learning model (f(x)) that relates to f (f= /(n)), wherein the /(v) is trained using one or more sets of previously measured data that include measured f values for corresponding sequences of calculated DDMs used to calculate corresponding v values.

2. The receiver of claim 1, wherein the one or more antennas comprise a first antenna to receive the DLOS RF signal component and a second antenna to receive the RF signal component that is reflected from the SP.

3. The receiver of claim 1, wherein the two or more wind direction-dependent features comprise a normalized bistatic radar cross-section (NBRCS).

4. The receiver of claim 3, wherein the NBRCS is calculated from the sequence of DDMs using stare processing.

5. The receiver of claim 4, wherein one or more unambiguous points are used in the stare processing.

6. The receiver of claim 5, wherein one or more pairs of ambiguous points are used in the stare processing.

7. The receiver of claim 1, wherein the two or more wind direction-dependent features comprise an angle of incidence.

8. The receiver of claim 1, wherein the two or more wind direction-dependent features comprise an SP azimuth.

9. The receiver of claim 1, wherein the two or more wind direction-dependent features comprise an SP speed.

10. The receiver of claim 1, wherein the two or more wind direction-dependent features comprise latitude and longitude.

11. The receiver of claim 1, wherein the x) is a neural network model.

12. The receiver of claim 1, wherein the RF carrier signal comprises a global navigation satellite system (GNSS) carrier signal and the locations of the one or more antennas, the transmitter, and the SP are determined from information carried by the GNSS carrier signal.

13. The receiver of claim 1, wherein the RF carrier signal comprises a carrier signal of a communications system and the locations of the one or more antennas, the transmitter, and the SP are determined from information transmitted separately from ground stations of the communications system.

14. A method for calculating wind direction (f) of a wind-driven water surface on the earth from a reflected radio frequency (RF) carrier signal, comprising: receiving a sequence of two or more consecutive delay -Doppler maps (DDMs) calculated over two or more corresponding times from a digital direct line-of-sight (DLOS) intermediate frequency (IF) signal, a digital reflected IF signal and known locations of one or more antennas of a receiver, a transmitter, and a specular point (SP) using a processor, wherein the one or more antennas receive a DLOS RF signal component and an RF signal component that is reflected from the SP on a wind-driven water surface on the earth of an RF carrier signal transmitted from the transmitter located above the surface of the earth, wherein RF front-end circuitry down-converts the DLOS RF signal component to a DLOS IF signal and down-converts the RF signal component that is reflected from the SP to a reflected IF signal, and wherein an analog-to-digital converter (ADC) converts the DLOS IF signal to the digital DLOS IF signal and converts the reflected IF signal to the digital reflected IF signal; calculating a feature vector (v) that includes two or more wind direction-dependent features of the digital reflected IF signal using the sequence of DDMs using the processor; and calculating a f at the SP from the calculated v and a machine learning model (f(x)) that relates to f (f=/(n)) using the processor, wherein the/(v) is trained using one or more sets of previously measured data that include measured f values for corresponding sequences of calculated DDMs used to calculate corresponding v values.

15. A computer program product, comprising a non-transitory and tangible computer-readable storage medium whose contents include a program with instructions being executed on a processor to perform a method for calculating wind direction (f) of a wind-driven water surface on the earth from a reflected radio frequency (RF) carrier signal, the method comprising: providing a system, wherein the system comprises one or more distinct software modules, and wherein the distinct software modules comprise a data acquisition module and an analysis module; receiving a sequence of two or more consecutive delay -Doppler maps (DDMs) calculated over two or more corresponding times from a digital direct line-of-sight (DLOS) intermediate frequency (IF) signal, a digital reflected IF signal and known locations of one or more antennas of a receiver, a transmitter, and a specular point (SP) using the data acquisition module, wherein the one or more antennas receive a DLOS RF signal component and an RF signal component that is reflected from the SP on a wind-driven water surface on the earth of an RF carrier signal transmitted from the transmitter located above the surface of the earth, wherein RF front-end circuitry down-converts the DLOS RF signal component to a DLOS IF signal and down-converts the RF signal component that is reflected from the SP to a reflected IF signal, and wherein an analog-to-digital converter (ADC) converts the DLOS IF signal to the digital DLOS IF signal and converts the reflected IF signal to the digital reflected IF signal; calculating a feature vector (v) that includes two or more wind direction-dependent features of the digital reflected IF signal using the sequence of DDMs using the analysis module; and calculating a f at the SP from the calculated v and a machine learning model (f(x)) that relates to f (f= /(n)) using the analysis module, wherein the /(v) is trained using one or more sets of previously measured data that include measured f values for corresponding sequences of calculated DDMs used to calculate corresponding v values.