Chinese Journal of Oceanology and Limnology   2017, Vol. 35 issue(1): 61-69     PDF       
http://dx.doi.org/10.1007/s00343-016-5120-3
Institute of Oceanology, Chinese Academy of Sciences
0

Article Information

ZOU Juhong(邹巨洪), GUO Maohua(郭茂华), CUI Songxue(崔松雪), ZHOU Wu(周武)
Airborne test flight of HY-2A satellite microwave scatterometer and data analysis
Chinese Journal of Oceanology and Limnology, 35(1): 61-69
http://dx.doi.org/10.1007/s00343-016-5120-3

Article History

Received Apr. 13, 2015
accepted for publication Sep. 30, 2015
accepted in principle Nov. 20, 2015
Airborne test flight of HY-2A satellite microwave scatterometer and data analysis
ZOU Juhong(邹巨洪)1,2, GUO Maohua(郭茂华)1,2, CUI Songxue(崔松雪)1,2, ZHOU Wu(周武)1,2        
1 National Satellite Ocean Application Service, Beijing 100081, China;
2 Key Laboratory of Space Ocean Remote Sensing and Application, SOA, Beijing 100081, China
ABSTRACT: This paper introduces the background, aim, experimental design, configuration and data processing for an airborne test flight of the HY-2 Microwave scatterometer (HSCAT). The aim was to evaluate HSCAT performance and a developed data processing algorithm for the HSCAT before launch. There were three test flights of the scatterometer, on January 15, 18 and 22, 2010, over the South China Sea near Lingshui, Hainan. The test flights successfully generated simultaneous airborne scatterometer normalized radar cross section (NRCS), ASCAT wind, and ship-borne-measured wind datasets, which were used to analyze HSCAT performance. Azimuthal dependence of the NRCS relative to the wind direction was nearly cos(2w), with NRCS minima at crosswind directions, and maxima near upwind and downwind. The NRCS also showed a small difference between upwind and downwind directions, with upwind crosssections generally larger than those downwind. The dependence of airborne scatterometer NRCS on wind direction and speed showed favorable consistency with the NASA scatterometer geophysical model function (NSCAT GMF), indicating satisfactory HSCAT performance.
Key words: HY-2     scatterometer     airborne test flight     data analysis    
1 INTRODUCTION

Sea surface wind vectors have major effects on marine waves, currents, water mass, and basic parameters of marine dynamics. They are important in studies such as those aimed at improving the accuracy of global atmosphere and marine dynamics prediction (Liu, 2002 ; Veldoe et al., 2002). Further, sea surface wind vectors represent the main effect on navigation, marine operations, and fishery production. They are critical for optimizing navigation routes, guaranteeing sea routes, typhoon avoidance, search and rescue and relief work. Therefore, the measurement of these vectors is very important. With advantages such as large-area synchronous measurement, fast speed, wide coverage, high temporal and spatial resolution, and continuous measurement, satellite remote sensing has become a major approach to the measurement of global wind vectors. In this approach, spaceborne microwave scatterometers have become the most important microwave remote sensor for global marine wind vectors (Liu, 2002), owing to their 24-hour measurement of sea surface wind vectors (speed and direction) under clear-sky and cloudy conditions. Since the United States launched the satellite Seasat in 1978 with the first scatterometer (Grantham et al., 1997), several spaceborne scatterometers such as the European Remote Sensing satellite (ERS1/2) scatterometer (Attema, 1991), NASA scatterometer (NSCAT; Naderi et al., 1991), QuikSCAT, Advanced Earth Observing Satellite 2(ADEOS II)(Wu et al., 2003), and meteorological operational (MetOp) platform (Figa-Saldaña et al., 2002) have been put into operation.

Airborne scatterometer test flights can serve as early feasibility studies of satellite microwave scatterometers, and are effective methods for examining the performance of such scatterometers. Therefore, many countries have studied airborne demonstration systems before developing spaceborne scatterometers (Schroeder et al., 1985). The findings of these studies have been widely used in areas such as the establishment/revision of geophysical models (Wentz, 1997), comparative study of wind estimation algorithms (Chi and Li, 1988), and verification of observation mechanisms (Jones et al., 1977 ; McIntosh and McLaughlin, 1990 ; Li and Neumann, 1991 ; Carswell et al., 1994).

This paper describes the aim, experimental design, measurement system layout, and data analysis results of the test flight for the microwave scatterometer onboard the HY-2A satellite (HSCAT). We present detection results of this Ku-frequency scatterometer from the test flight. A comparative analysis of the airborne scatterometer measurements, ship measurements, and spaceborne scatterometer data describes HSCAT performance and its capacity to detect the sea surface wind field. This provides a reference for the future development of spaceborne microwave scatterometers.

2 EXPERIMENTAL DESIGN

The HY-2A is the first ocean dynamic environmental satellite in China that uses a comprehensive active/ passive microwave remote sensor to measure marine dynamics parameters (Zou et al., 2014). The microwave scatterometer is one of three main payloads of the satellite, and is used for measuring the global ocean wind field. To ensure success, it is necessary to perform inspection and assessment in areas such as operational functions, work modes, instrument operation, and indicators via test flights. Such flights facilitate verification that should be conducted before assessment of the technical status of the entire satellite and ground application system. The main tasks and aim of the experiment are examining major functions and performance indicators of the load and simultaneously providing a relatively authentic data source for development of ground systems. Specifically, these tasks include verification of measurement precision of the target backscattering coefficient (sigma0), multi-azimuth test and verification, measurement and verification of various polarizations, and verification of internal calibration accuracy and dynamic range.

The analyzed flight was largely satellitesynchronous, and was supported by ship, wave buoy, reanalysis fields, and data of monitoring stations and sites. The flight area was the southeastern part of the South China Sea, and the platform was the Y-8 aircraft. Based on characteristics of the HY-2A payload, the aircraft had an external nacelle, and both the non-airtight and airtight cabin were modified. Based on aircraft conditions, there were also adaptationswith the payload of the scatterometer considering factors such as low-altitude flight.

2.1 Instruments and equipment

(1) Main payload - airborne HSCAT

To adapt to the low flight altitude of the aircraft, HSCAT transmit and detection timing had to be tuned. In performing this modification, we do not want to change the sigma0 measurement accuracy.

For the spaceborne HSCAT, a long pulse width is used to achieve high SNR. For the airborne HSCAT, the pulse width must be reduced to adapt to the low flight altitude, so the SNR will be decreased. Sigma0 measurement accuracy is related to the signal noise rate and effective number of independent samples. Therefore, the pulse rate of the airborne HSCAT was increased to augment the effective number of independent samples, so we expected to have nearly the same sigma0 measurement accuracy as the spaceborne HSCAT.

To infer wind speed and direction from sigma0 measurement, multiple azimuth measurements of sigma0 are required. To simplify the design and installation of the airborne HSCAT, it was necessary to change the rotating antenna onboard HY-2A to one fixed to the aircraft. Then, the multiple azimuth measurements were made during a multiple-square flight pattern. Given that sea surface winds change little during measurement, this modification does not induce substantial error. Key parameters of the airborne HSCAT are shown in Table 1.

Table 1 Airborne HSCAT instrument key parameters

(2) Supplemental equipment

Airborne equipment: inertia guidance and GPS were used to provide longitude, latitude, height and attitude (in three dimensions).

ASCAT: The ASCAT (Figa-Saldaña et al., 2002) is one of the payloads onboard MetOp-A, which was launched on 19 October 2006. ASCAT is a fan-beam scatterometer with six fan-beam antennae. The swath is 500 km on both sides, and 25-km resolution wind vectors are available upon wind retrieval.

Shipborne weather station: for stationary measurement, a shipborne weather station was used to measure wind speed and direction at a sampling interval of 1 min. A GPS onboard the ship measured longitude, latitude, speed, and ship course at the same sampling interval. Actual wind speed and direction were estimated based on the ship course, GPSmeasured speed, and wind speed and direction measured by the weather station.

2.2 Installation of airborne HSCAT test-flight equipment

The test flight of HSCAT fixed its incidence angle. Using two standard round feed-source horns with H and V polarization, two antennae were installed in parallel outside the cabin on the plane belly to monitor sea surface at 46.5° offthe normal direction. The shape and position of the antenna are shown in Figs. 1 and 2.

Figure 1 Shape and position of HSCAT antenna
Figure 2 External antenna device of airborne HSCAT
2.3 Flight requirements

Because the rotating antenna onboard the HY-2A was altered to one fixed to the aircraft, the aircraft had to execute a rotated square flight pattern to acquire multiple azimuth measurements of sigma0 that were required to retrieve wind speed and direction. A steady aircraft attitude is not easy to maintain during circular fight, so a multiple-square pattern was chosen (Fig. 3). The corners of four square flight patterns are marked by C1 to C16, and the locations of each corner are listed in Table 2. The aircraft traversed each side of the four squares, centered on a ship and wave buoy located at (111°E, 19°N). Each square had different directions, with angles 0°, 20°, 45° and 65° relative to north. In the aircraft traverse of one side of a square, it made one azimuth measurement of sigma0. Thus, completing the four squares, 16 different azimuth measurements of sigma0 were made. Side lengths of each square are required to be 15-20 km to ensure sufficient independent samples during a single azimuth measurement of sigma0. Flight altitude was from 5 000 to 8000 m. At least three or four test flights were required to ensure that sigma0 with variable wind speeds could be measured, so we could verify HSCAT performance under different wind speeds. Because the antenna was installed on the left side of the aircraft at a 90° angle from its longitudinal axis and an incidence angle of 46.5°, it was necessary to fly counterclockwise to keep the antenna pointing at the ship measurement synchronous point (111°E, 19°N), which was the center of each square.

Figure 3 Pattern of airborne HSCAT test flight
Table 2 Location of each corner for HSCAT flight pattern in Fig. 3
3 DATA ANALYSIS

Three test flights of the HSCAT were conducted on 15, 18 and 22 January 2010, which yielded the sigma0 measurement data. The multiple-square flight pattern was executed on 18 and 22 January, and a straight line pattern on 15 January. Thus, data acquired on the latter date were only used for coarse calibration of sigma0.

3.1 Data preprocessing

In the multiple-square flight, each group of measurements from the flight along one side of the square corresponds to a single azimuth measurement of sigma0. Raw data from HSCAT compose a timeordered product containing the measurement result for each pulse. Each pulse result was grouped into the appropriate side of the square to calculate sigma0 corresponding to that side. Using the radar equation, the signal measured in each pulse was transformed into sigma0. Taking the average signal of each sequence of 2 000 pulses (corresponding to 4 s), we can obtain a group of 4-s average sigma0 for each side of the square. For example, sigma0 results corresponding to the flight from C9 to C12 and C12 to C11 are shown in Fig. 4 ; a 2-dB coarse calibration value has been added to these results. The coarse calibration value was determined based on comparison of a group of sigma0 measurements on 15 January with a predicted sigma0 based on ship measurement and the NSCAT-2 model. Aircraft attitude and background data such as sea surface wind fields are also shown in Fig. 4.

Figure 4 Backscattering coefficient and corresponding information on platform attitude

It is seen from Fig. 4 that aircraft yaw had a relatively large impact on sigma0. Maxima or minima in the sigma0 time sequence typically corresponded to yaw minima or maxima, respectively. This is because change of yaw alters the incidence angle, which greatly impacts the measurement of sigma0. Therefore, we used attitude information from inertial guidance and eliminated measurements accompanying unstable attitudes in subsequent data processing. Specifically, data associated with yaw angle > 0.4° were eliminated. Assuming very stable wind conditions, we used a threshold to further eliminate outliers with relatively large measurement differences. The threshold value was set to 0.8 dB, four times the expected sigma0 measurement accuracy. After applying the above operations to each group of sigma0 measurements, we obtained the data shown in Fig. 5. We see that abnormal values caused by unstable aircraft attitude have been removed.

Figure 5 Sigma0 after data processing
3.2 Primary analysis of data

(1) Second harmonic characteristics of backscattering coefficient with direction

After four complete square flights, a total of 16 datasets were obtained, each of which corresponds to a group of time-ordered pulse measurement results along one of the 16 sides of the four squares in the flight pattern. Each side corresponds to an independent azimuth sigma0 measurement. The radar equation was used to calculate sigma0 for each pulse, and 16 groups of time-ordered sigma0 were acquired. Taking the average of the time-ordered sigma0 for each group, we obtained an average sigma0 result for each group, thereby producing 16 different azimuthmeasurement average sigma0, as shown in Table 3. In this table, each sigma0 measurement is denoted by the side of the square. For example, C1-C4 correspond to measurement during the flight from C1 to C4. VV polarization sigma0 and HH polarization sigma0 correspond to the average sigma0 on each side. The aircraft course corresponds to its flight direction. Average ship wind speed and direction correspond to the average wind vector observed by the shipborne weather station during the measurement period. The relative angle between the antenna and aircraft course is 270°, and the incidence angle is 46.5° for all measurements.

Table 3 Measurements from scatterometer on 18 January

With the data set in Table 3, we can draw a sigma0 vs. wind direction curve with curve fitting, which was shown in Fig. 6. In Fig. 6, the solid line corresponds to HSCAT sigma0, and the dashed line corresponds to NSCAT-2 result. The change of HSCAT sigma0 with radar look direction with respect to wind direction behaved roughly as the cosine of twice the angle of that direction. The HSCAT sigma0 was in good agreement with the fit of the NSCAT-2(Wentz et al., 1998) curve, except for C13-C14 and C5-C8. Average wind directions during the measurement periods of C13-C14 and C5-C8 were 51.7° and 49.8°, while that of square C1-C2-C3-C4 was ~40°. Thus, the average wind direction over C13-C14 and C5-C8 changed substantially over that of square C1-C2-C3- C4. The mismatch of the HSCAT sigma0 with NSCAT-2 model sigma0 was likely induced by a change of wind direction during the measurement period.

Figure 6 Sigma0 vs. wind direction from 18 January measurement data

To minimize the impact of an unstable wind field during the measurement period, data obtained during similar wind fields were used for curve fitting. Therefore, sigma0 measured in squares C1-C2-C3- C4 and C9-C10-C11-C12 was selected, with the results shown in Fig. 7. There is improved consistency between the measured data and curve fit of NSCAT-2 model data. Azimuthal dependence of the normalized radar cross section (NRCS) relative to wind direction was nearly cos (2w), with NRCS minima for crosswind and maxima near upwind and downwind. The NRCS shows a small difference between upwind and downwind directions, with upwind cross-sections generally larger than downwind.

Figure 7 Sigma0 vs. wind direction using symmetry and 18 January measurement data

Using a processing method similar to that for 18 January measurement data, we fit sigma0 with 22 January data, with the results shown in Fig. 8. It is seen that the difference between measured VV polarization sigma0 and HH polarization sigma0 under weak wind speeds is significantly less than that from the NSCAT-2 model. Root mean square errors (RMSEs) of measured data substantially increased under weak speeds. The data failed to show satisfactory second harmonic distribution characteristics under such speeds. This may be attributed to factors such as wind field change during the measurement, flight attitude, and low SNR.

Figure 8 Sigma0 vs. wind direction using symmetry and 22 January measurement data

(2) Wind measurement accuracy

Wind vectors were retrieved from flight data using maximum-likelihood estimation with the NSCAT-2 geophysical model function (Wentz et al., 1998), and wind vector solution whose wind direction was closer to that of the ship-observed direction was chosen as the actual direction. The wind vector retrieval results are shown in Tables 4 and 5. Table 4 lists wind vector inversion results with 18 January measurement data. The time is Beijing time. The wind speed is at 10-m height in m/s, and the oceanographic convention for wind direction was applied. Under this convention, a wind direction of 0° indicates northward flow. The ship location is 111°E, 19°N, and collocated ASCAT data were for 111.253 7°E, 18.888 7°N, with overpass at 10:34:45 Beijing time. Results for Nos. 1 through 4 in both Table 4 and Table 5 correspond to wind vectors retrieved from measurements for squares C1-C1- C3-C4, C13-C14-C15-C16, C9-C10-C11-C12 and C5-C6-C7-C8, respectively. Result No. 5 corresponds to the wind vector retrieved from squares C1-C1-C3-C4 and C9-C10-C11-C12, and No. 6 to that from all four squares. We see that under intermediate wind speeds (~9 m/s), the RMSE for wind speed is 0.4 m/s and 5° for wind direction, compared with ASCAT wind. Comparing this with ship measurement data, the corresponding RMSEs are 0.53 m/s and 10.9°. Table 5 gives wind vector inversion results using 22 January measurement data. The collocated ASCAT data were at 110.938 4°E, 18.470 8°N, and the overpass was at 10:52:03 Beijing time. We see that under weak wind speeds (~5 m/s), the RMSE for wind speed is 0.7 m/s and 15.6° for wind direction. Comparing ship measurement data, the corresponding RMSEs are 1.2 m/s and 13.2°.

Table 4 Wind vector inversion results and accuracy analysis (using 18 January measurement HSCAT, buoy and ASCAT data)
Table 5 Wind vector inversion results and accuracy analysis (using 22 January measurement HSCAT, buoy and ASCAT data)
4 CONCLUSION

Aimed at evaluating the performance of HSCAT before launch, three test flights of the scatterometer were conducted on 15, 18 and 22 January 2010. Although restricted by a lack of absolute calibration for NRCS, the flight attitude, especially yaw, had strong impacts on the stability of the scatterometer measurement data. However, the flights yielded useful simultaneous datasets of airborne scatterometer NRCS, ASCAT wind, and ship wind.

Analyses of these data clearly documented the dependence of NRCS on wind direction and speed. A comparison of airborne scatterometer, ship, and ASCAT wind data revealed that the sigma0 response to wind speed from both polarizations had clear second harmonic distribution characteristics. There was a substantial difference between downwind and upwind directions. The sigma0 of upwind was slightly larger than that of downwind. There was relatively strong consistency with estimates of the NSCAT-2 model. Under intermediate and weak wind speeds, VV polarized sigma0 was larger than HH polarized sigma0. Comparison of measurement data on 22 January (weak wind speed) with those on 18 January (intermediate wind speed) showed that with the same azimuth angle, the measured sigma0 under weak wind speed was smaller than that under intermediate wind speed. This is consistent with the physical fact of sigma0 increase with wind speed. Comparison of wind vector inversion results with satellite data (ASCAT) gave a wind speed RMSE <0.7 m/s. Wind direction RMSE was <15.6°. Comparing ship measurement data, the wind speed RMSE was <1.2 m/s. Wind direction RMSE was <13.2°. According to these results, we conclude that HSCAT performance is satisfactory.

References
Attema E P W, 1991. The active microwave instrument onboard the ERS-1 satellite. Proceedings of the IEEE, 79 (6) : 791 –799. Doi: 10.1109/5.90158
Carswell J R, Carson S C, McIntosh R E, et al, 1994. Airborne scatterometers: investigating ocean backscatter under low and high-wind conditions. Proceedings of the IEEE, 82 (12) : 1835 –1860. Doi: 10.1109/5.338074
Chi C Y, Li F K, 1988. A comparative study of several wind estimation algorithms for spaceborne scatterometers. IEEE Transactions on Geoscience and Remote Sensing, 26 (2) : 115 –121. Doi: 10.1109/36.3011
Figa-Saldaña J, Wilson J J W, Attema E, et al, 2002. The advanced scatterometer (ASCAT) on the meteorological operational (MetOp) platform: a follow on for European wind scatterometers. Canadian Journal of Remote Sensing, 28 (3) : 404 –412. Doi: 10.5589/m02-035
Grantham W, Bracalente E M, Jones W L, et al, 1997. The SeaSat-A satellite scatterometer. IEEE Journal of Oceanic Engineering, 2 (2) : 200 –206.
Jones W, Schroeder L C, Mitchell J, 1977. Aircraft measurements of the microwave scattering signature of the ocean. IEEE Transactions on Antennas and Propagation, 25 (1) : 52 –61. Doi: 10.1109/TAP.1977.1141533
Li F, Neumann G. 1991. Ocean radar backscatter observations at 14 GHz with airborne scatterometers. In: Proceedings of Ocean Technologies and Opportunities in the Pacific for the 90's. IEEE, Honolulu, Hawaii. p.1212-1216.
Liu W T, 2002. Progress in scatterometer application. Journal of Oceanography, 58 (1) : 121 –136. Doi: 10.1023/A:1015832919110
McIntosh R E, McLaughlin D J. 1990. Ocean surface NRCS measurements obtained with an airborne C-band scatterometer and a spinning antenna. In: Proceedings of 10th Annual International Geoscience and Remote Sensing Symposium. IEEE, College Park, MD, USA. 1065p.
Naderi F M, Freilich M H, Long D, 1991. Spaceborne radar measurement of wind velocity over the ocean-an overview of the NSCAT scatterometer system. Proceedings of the IEEE, 79 (6) : 850 –866. Doi: 10.1109/5.90163
Schroeder L C, Schaffner P, Mitchell J, et al, 1985. AAFE RADSCAT 13. 9-GHz measurements and analysis: windspeed signature of the ocean. IEEE Journal of Oceanic Engineering, 10 (4) : 346 –357.
Veldoe C, Bruske K, Kummerow C et al. 2002. The burgeoning role of weather satellites. In: Simpson R ed. Coping with Hurricanes. Chapter Ⅱ. American Geophysical Union, Washington DC. p.217-218.
Wentz F J, Freilich M H, Smith D K. 1998. NSCAT-2 geophysical model function. In: Proceedings of Fall AGU Meeting. San Francisco, CA.
Wentz F J. 1997. A two-scale scattering model with application to the JONSWAP'75 aircraft microwave scatterometer experiment. NASA CR-2919. NASA, London.
Wu C, Liu Y, Kellogg K H, et al, 2003. Design and calibration of the SeaWinds scatterometer. IEEE Transactions on Aerospace and Electronic Systems, 39 (1) : 94 –109. Doi: 10.1109/TAES.2003.1188896
Zou J H, Xie X T, Zhang Y, Lin M S. 2014. Wind retrieval processing for HY-2A microwave scatterometer. In:Proceedings of IEEE International Geoscience and Remote Sensing Symposium. IEEE, Quebec City, QC.p.5160-5163.