CYGNSS was launched on December 15, 2016 at 13:27:21 UTC and today marks the completion of its third year on orbit. Much has happened in that time and the future looks very promising, with all eight microsatellites healthy and operating continuously in their nominal science data-taking mode. Year 1 was spent on engineering commissioning and initial ocean wind measurements. In year 2, the ocean wind retrievals were refined and characterized and investigations of observations over land began in earnest. This past year has seen progress on several fronts. An important refinement to the engineering calibration has been developed. Ocean wind measurements in tropical cyclones have been successfully introduced into numerical hurricane forecast models. And the ability to image inland waterbodies with high resolution has been leveraged to demonstrate two new measurement capabilities.
CYGNSS measures the strength of GPS signals reflected by the Earth surface, from which properties of the surface are derived. To do this accurately requires knowledge of the signal strength transmitted by the GPS satellite. The original approach to calibration assumed each GPS satellite had a unique transmit power level and that it did not vary in time. The unique power levels were estimated using measurements averaged over several months. Subsequent analysis has revealed that GPS transmit power can vary significantly, depending on the particular satellite and on its position in orbit. In order to compensate for the changes, the direct signal GPS receivers on all eight CYGNSS satellites were reprogrammed in 2018 to act as power meters which can monitor the variations in transmit power [Wang et al., 2019. A Real-Time EIRP Level 1 Calibration Algorithm for the CYGNSS Mission using the Zenith Measurements. Proc. IGARSS 2019]. The impact of this change is illustrated in the following figure. The top panel shows the geographical distribution of the average error in CYGNSS wind speed for a single GPS transmitter (SVN 63) assuming its transmit power is constant. The center panel shows the actual GPS transmit power as measured by the new on-board monitoring system. Clear localized differences in the measured power are evident which correlate with the errors in wind speed. The bottom panel shows the distribution of average wind speed error after the monitored power is incorporated into the calibration. The discrepancies have been largely removed.
Over the past year, CYGNSS ocean wind measurements have been added to the HWRF numerical forecast model used by the National Hurricane Center and compared to the standard operational forecast, which does not yet use CYGNSS. To forecast a hurricane, HWRF is first initialized with satellite, airborne, and ground based measurements. The storm is then allowed to develop in software, guided by underlying physical principles of thermodynamics, radiation and mass and energy conservation. An example of this is shown in the following figure, which presents three versions of the horizontal wind speed at 3 km altitude for Hurricane Harvey on 25 Aug 2018 at 06Z. The left panel is the 24 hr forecast predicted on 24 Aug by the operational version of HWRF without CYGNSS data. The center panel is the HWRF prediction with CYGNSS winds included. Note the azimuthal shift in predicted peak winds into the northeast quadrant when CYGNSS winds are included. The right panel shows measurements made by NOAA’s airborne doppler radar of the actual winds at 3 km. The predicted azimuthal shift is confirmed, suggesting that the storm’s development is being more accurately modeled [Cui et al, 2019. A Preliminary Impact Study of CYGNSS Ocean Surface Wind Speeds on Numerical Simulations of Hurricanes. Geophys. Res. Ltrs.].
More accurate modeling of storm development should help forecast the intensity of its surface winds, and this has been demonstrated by another study using HWRF [Annane et al., 2018. Impact of CYGNSS Data on Tropical Cyclone Analysis and Forecasts Using the Operational HWRF. 33rd AMS Conf. Hurricanes Trop. Meteo.]. In this case, a 120 hr forecast of minimum sea level pressure (MSLP) in the eye of Hurricane Michael was conducted on 8 Oct 2018 at 00Z without and with CYGNSS winds included as one of the inputs. The left panel in the following figure shows the true (best track) MSLP in black and the operational forecast (without CYGNSS) in red. The rapid drop in MSLP that occurred between ~30-60 hr is not well forecast. The right panel shows the HWRF forecast with CYGNSS winds included and can be seen to more accurately predict the rapid intensification of the storm.
CYGNSS measurements over land are able to provide high resolution tracks across inland waterbodies when the surface is calm enough to support coherent specular scattering. A new land/water mask has been developed which leverages this capability as well as the ability of CYGNSS to penetrate through clouds, rain and vegetation canopies. As a result, dynamic changes in waterbodies (e.g. due to flooding, seasons or human development) can be resolved better than existing water masks that rely on optical satellite imagers. [Gerlein-Safdi and Ruf, 2019. A CYGNSS-Based Algorithm for the Detection of Inland Waterbodies, Geophys. Res. Ltrs.]. An example of this new capability is shown in the seasonal watermasks of the Okavango Delta in Botswana derived from CYGNSS overpasses in 2018. The left panel, for Dec/Jan/Feb, shows the typical river boundaries that are present during the dry season. The mask in the right panel, for Jun/Jul/Aug, captures the widespread flooding which occurs during the rainy season.
CYGNSS overpasses of rivers also produce high resolution tracks that can be used to measure a river’s width and infer its streamflow rate. This has been successfully demonstrated using overpasses of the Pascagoula River in Mississippi during a major flood event in April 2019 [Warnock and Ruf, 2019. Response to Variations in River Flowrate by a Spaceborne GNSS-R River Width Estimator, Remote Sens.]. Streamflow measurements by a USGS river gauge made before, during and after the event are shown in the upper panel below, together with markers at the times of each of five overpasses by CYGNSS. The lower panel shows the Associated GNSS-R Width (AGW) of the river derived from each CYGNSS overpass, together with the Pascagoula’s streamflow rate at the time.
The two are highly correlated, suggesting that another new type of CYGNSS data product may be possible over inland waterbodies.
The constellation of eight CYGNSS microsatellites reached a milestone today, completing its second year on-orbit. The first year had focused primarily on engineering commissioning, calibration and early validation of the science data products. The second year was spent continuing to refine and improve the quality of the science data products, while also applying the measurements to a number of scientific investigations. Some examples of the refined wind speed measurements are shown below for three overpasses of Hurricane Maria, on 23 Sep 2017 at 00:18 (top), 23 Sep 2017 at 00:23 (mid), and 24 Sep 2017 at 00:18 (bot).
Coincident hurricane overpasses by CYGNSS and NOAA hurricane hunter aircraft
In the figure, the green trace shows the measurements made by CYGNSS while the blue trace shows measurements made by SFMR instruments on the NOAA hurricane hunter aircraft that flew into the storms while CYGNSS was passing overhead. The CYGNSS and SFMR measurements can be seen to agree quite well. Quantitative results of the calibration and validation of CYGNSS science data products are described in a Special Issue of the IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.
Measurements by CYGNSS of Typhoon Trami in Sep 2018, shown below, illustrate its ability to track the evolution of a storm over its full life cycle. In the figure, the lower left image is a composite over the period 20-30 Sep 2018 of all wind speed measurements made at one phase of the constellation’s orbit. The upper right insert shows all measurements made on 28 Sep 2018, with the measurements re-gridded to a storm-centric coordinate system to illustrate the wind speed structure in the inner core region.
CYGNSS wind speed measurements of Typhoon Trami
Science investigations have expanded beyond the primary mission objective to measure wind speed in tropical cyclones. The measurements are being introduced into numerical weather prediction models to assess their impact on the skill in forecasting hurricane track, intensity and structure. Early results look quite promising, with clear signs of a positive impact on forecast skill. The results will be presented in a series of talks at the annual meeting of the American Meteorological Society in January 2019.
In parallel with the activity over ocean, new science investigations are examining data over land as well. CYGNSS measurements have been found to be sensitive to the surface soil moisture, and to the presence of inland water bodies. Good examples of this are documented in recent publications. Kim and Lakshmi (2018) [doi:10.1029/2018GL078923] estimate soil moisture at numerous sites in the continental U.S. over a 12 month period and find close agreement with independent measurements made by ground sensors and another satellite. Jensen et al. (2018) [doi:10.3390/RS10091431] demonstrate the ability of CYGNSS to detect and map the extent of flood inundation under dense forest canopies. Chew et al. (2018) [doi:10.1038/S41598-018-27673-3] use this capability to produce time lapse images of flooding in and around Houston and Havana after landfalls by Hurricanes Maria and Irma, respectively. These results demonstrate the wide range of future applications of CYGNSS data as the mission moves forward into its third year.
As we head into the 2018 Atlantic hurricane season, now is a good time to reflect on the accomplishments achieved by CYGNSS since its launch in December 2016. Early mission operations focused on engineering commissioning of the satellites and of the constellation as a whole. One achievement in particular is noteworthy. The satellites have no active means of propulsion, yet their relative spacing is important for achieving the required spatial and temporal sampling. The desired spacing is achieved by individually adjusting a spacecraft’s orientation and, as a result, the atmospheric drag it experiences. This technique is referred to as “differential drag”. An increase in drag will lower a satellite’s altitude, thereby changing its orbital velocity. We adjust the distance between spacecraft by adjusting their relative velocities. This is a new way of managing the spacing between a constellation of satellites, and one that can be significantly less risky and lower in cost than using traditional active propulsion. As a result, we were able to afford more satellites for the same price, which ultimately led to better, more frequent, sampling of short lived, extreme weather events like tropical cyclones.
Here is a figure, provided by CYGNSS team member Kyle Nave of ADS, illustrating the change in relative speed between two of the CYGNSS spacecraft that occurred the first time a differential drag maneuver was performed, on February 23, 2017.
The orbital phase rate between the two spacecraft is shown before, during and after the higher of the two had its orientation changed to maximize atmospheric drag. Phase rate measures how quickly the angle between two satellites changes. By increasing the drag on the higher one, it lowers to an altitude and orbital velocity closer to the lower one, thus reducing the phase rate. This was an important first confirmation of our ability to perform the maneuver. Since then, there have been many more drag maneuvers. Five of the eight satellites are now properly positioned relative to one another at a common altitude, and the remaining three are expected to have their drag maneuvers completed later this year.
The primary science objective of the CYGNSS mission is measurement of near surface wind speed over the ocean in and near the inner core of tropical cyclones. In an earlier NASA blog, (15 Dec 2017), I reported on our measurements of Hurricane Maria made in September 2017. Since that time, we have been examining the quality of our measurements both within and away from major storms. Measurements at ocean wind speeds below 20 m/s (44 mph) were found to have an RMS uncertainty of 1.4 m/s (3 mph). Measurements of storm force winds during the 2017 Atlantic hurricane season were found to have an uncertainty of 17% of the wind speed. The analysis that produced these results is reported in Ruf et al. (2018). DOI: 10.1109/JSTARS.2018.2825948.
CYGNSS operates continuously, over both ocean and land, and the land data have been another focus of recent investigations. The quality of some of those measurements, in particular regarding its spatial resolution, has come as something of a pleasant surprise. Here is one example of CYGNSS land imagery, of the Amazon River basin in South America, provided by Dr. Clara Chew of UCAR.
In the image, inland water bodies are prominently visible. This includes not only the major arms of the Amazon River but also its quite narrow minor tributaries. Careful examination of this and similar CYGNSS images suggests that the spatial resolution is markedly better here than it is over typical open ocean areas. The explanation lies in a transition of the electromagnetic scattering from an incoherent, rough surface regime over ocean to a largely coherent, near specular regime over inland waters. The fact that coherently scattered signals have inherently better spatial resolution is a well known phenomenon. What was unexpected is the widespread, global extent to which land surface conditions support coherent scattering. It requires the height of the surface roughness to be significantly below the wavelength of the radiowave signal, which in our case is 19 cm. This is apparently a ubiquitous property of wetland regions. It is a very fortuitous property for us, as it should enable an entirely new direction in scientific applications of CYGNSS measurements over land. NASA has recently added new investigators to the CYGNSS team specifically to study these new and exciting land applications.
A recent article summarizing these and other CYGNSS achievements, as well as some of the future applications of its measurements, is available at <www.nature.com/articles/s41598-018-27127-4>. The mission has demonstrated that smaller, more cost-efficient satellites are able to make important contributions to the advancement of science. In the months and years ahead, CYGNSS will hopefully be able to demonstrate that those advances can lead to practical scientific applications, such as extreme weather monitoring and prediction, that will benefit humankind.
CYGNSS was launched into low Earth orbit on December 15, 2016 at 08:37 EST and today is its first anniversary. The mission has had a very busy first year on orbit, transitioning from an early engineering commissioning phase into the science observing phase in time for the very active 2017 Atlantic hurricane season. The mission was supported during the hurricane season by the NOAA Airborne Operations Center (AOC), which operates a fleet of P-3 “hurricane hunter” airplanes that make reconnaissance flights into tropical storms and hurricanes to observe wind speed and other weather conditions first hand. We worked closely with AOC to coordinate many of their flight campaigns with overpasses of the storms by CYGNSS. They were able to time many of their eyewall penetrations to align closely in both time and space with our overpasses, which helps us train and evaluate our own wind speed measurements As a result, we now have dozens of coincident tracks of wind speed observations through the inner core regions of Hurricanes Harvey, Irma, Jose and Maria. The collaboration with NOAA this summer and fall has been incredibly fruitful, and I and the CYGNSS project team are very grateful for their generous support.
As the 2017 hurricane season winded down, we turned our attention to processing the coincident overpass data and characterizing and evaluating the performance of our wind speed measurements. One example is shown here. On September 24, 2017 at 18:13-18:21 UTC, the CYGNSS FM#2 spacecraft flew across Hurricane Maria.
The red line in the figure shows the track of the specular reflection from transmissions by the GPS PRN#13 satellite. CYGNSS makes its wind measurements along this track. The black line in the figure shows the flight path of the P-3 hurricane hunter that day. A distinctive cloverleaf pattern can be seen that results from the plane making multiple eyewall penetrations. The colored portion of the P-3 flight path is the leg closest in time and space to the CYGNSS specular point track. The color scale represents the difference in time between the CYGNSS overpass and the P-3 observing time. With such close coincidence in time and space, we hope and expect that the two measurements of wind speed will be consistent.
The next figure shows the wind speed measured by CYGNSS (blue), measured by the P-3 airplane using its Stepped Frequency Microwave Radiometer (SFMR) wind speed sensor (red), and produced by the ECMWF and GDAS numerical weather prediction (NWP) models (black) along the CYGNSS specular point track.
Away from the storm center at lower wind speeds, CYGNSS and NWP measurements agree well. Near the storm center, CYGNSS responds to the much higher wind speeds. In general, NWP models tend to underestimate peak winds in large storms and this is likely the case here. While NWP models generate winds everywhere, SFMR winds are only available along the portion of the satellite track where the P-3 airplane flew. In the region where coincident measurements were made, CYGNSS and SFMR winds can be seen to agree fairly well. It should be noted that the scatter present in the CYGNSS measurements can be seen to increase as the wind speed increases. This is probably a result of the decrease in GPS signal strength scattered in the specular direction when the sea surface is significantly roughened by high winds. How best to handle this characteristic of the CYGNSS wind speed retrievals will be an important topic of upcoming investigations.
Happy Birthday, CYGNSS!
p.s. and just in time for the first year anniversary of CYGNSS on orbit, new science data files using the latest (v2.0) engineering calibration and science retrieval algorithms have just been posted at the NASA PO.DAAC web site. Access the data by going to
The CYGNSS constellation has been operating in its science data-taking mode continuously since March 2017. The satellite hardware has been performing as designed while we make adjustments to the software on-board and on the ground so we are better able to operate smoothly and autonomously. We also spent much of the summer working on the relative spacing between the satellites, by adjusting their differential drag and, as a result, their relative orbital velocities.
An example of a typical “day in the life” for CYGNSS is shown in this video, which combines together the wind observations made by all eight spacecraft over a 24 hour period on 20 June 2017.
CYNSS wind observations (Level 3 v 1.1) made by the full constellation over a 24 hour period on 20 June 2017.
Several things are noteworthy in the video. Active storm areas appear as the regions with yellow and (especially) red wind speeds. The global coverage is seen to extend between about 38 deg north and 38 deg south latitude. If you focus on any one location within that coverage zone and note when CYGNSS measurements are made there, you can get a feel for the temporal sampling properties. In the Gulf of Mexico, for example, winds are measured during the two time interval 1300-1600 and 2000-2300 UTC. This property, that measurements are made each day during two periods of several hours each, is common to all locations within the coverage zone.
More recently, since the Atlantic hurricane season became especially active with major storms Harvey, Irma, Jose and Maria, we have focused on conducting targeted observations. This consists of predicting when we will pass over an active storm in the next several days, assembling command sequences to activate higher quality data-taking modes while over the storms, uploading those commands to the appropriate spacecraft, then scheduling additional contacts afterwards with our ground stations to downlink the higher volume of data for processing and analysis. In addition, we have been working closely with our colleagues at NOAA who operate the fleet of hurricane hunter aircraft, to coordinate their flights with our overpasses. On a number of occasions, they have been able to time the aircraft penetrations through the storm center so they coincided with CYGNSS overpasses, and to align their flight path so it paralleled the measurement track of the satellite. We are just beginning to evaluate these intercomparison data sets and intend for them to anchor our validation of high wind speed performance.
Our measurements of the major Atlantic hurricanes this season are still preliminary. The algorithms used to convert radar engineering measurements into ocean surface wind speeds have yet to be fully validated at high winds and, when they are, we will likely tweak the algorithms in order to optimize their performance. An example of CYGNSS measurements of the winds in Hurricane Harvey are shown here, taken on the morning of 25 Aug 2017. Harvey made landfall in southeast Texas that evening.
CYGNSS Level 3 gridded surface wind speed data product (v1.1). (top) at 1300-1400 and (bottom) at 1400-1500 UTC on 25 Aug 2017, prior to landfall at ~03:00 UTC on 26 Aug 2017. Hurricane Harvey is centered slightly off-shore, in the region of highest wind speeds shown in red.
CYGNSS makes measurements continuously over both ocean and land. The ocean data are used to estimate surface wind speed. The land data are sensitive to the moisture content of the soil and, in the most extreme circumstances, can be used to detect and image flood waters. This is illustrated in the following series of images of CYGNSS measurements over southeast Texas made shortly before, and then in the days after, Harvey made landfall.
CYGNSS SNR images of southeast Texas before and after Hurricane Harvey landfall. (left) Aug 29 SNR image with coastal flooding circled. (right) Time lapse SNR images, with flooding inundation indicated by large increases in SNR.