EN中文

Journal of Geo-information Science >

2016 , Vol. 18 >Issue 11: 1544 - 1550

DOI: https://doi.org/10.3724/SP.J.1047.2016.01544

Orginal Article

Wind Field Retrieval of South China Sea Based on Gaussian-FFT Method

  • TIAN Siyu ,
  • HUANG Xiaoxia , * ,
  • LI Hongga ,
  • WANG Hao ,
  • LI Xia ,
  • CHENG Peng
Expand
  • Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100101, China
*Corresponding author: HUANG Xiaoxia, E-mail:

Received date: 2015-12-07

  Request revised date: 2016-05-05

  Online published: 2016-11-20

Copyright

《地球信息科学学报》编辑部 所有
Fold

Abstract

Wind speed is a basic parameter of oceanography. It plays an important role in the interaction between ocean and atmosphere. Therefore, it is significant and necessary to obtain the wind datasets over the sea surface. However, due to the large area and complex condition, it is usually difficult to get the wind field data of South China to satisfy different demands in time. Conventional approaches, such as placing observation stations or buoys, are not only expensive but also dependent on the weather condition. Therefore it is urgently necessary to find other ways to get the wind datasets timely. ENSISAT ASAR, an all-weather and all-time microwave radar sensor, could collect the real-time and dynamic information over sea surface, which provides a new approach for researchers to acquire wind field datasets over sea surface, especially for the waters with complicated conditions, such as South China Sea. In this paper, the Gaussian-FFT method is firstly applied to retrieve the wind field of South China Sea based on ASAR image. At first, the FFT spectrum of ASAR image is acquired with the FFT algorithm. Secondly, a “cigar-shaped” two-dimensional (2-D) Gaussian function is fitted to the FFT spectrum to find the direction of wind streaks and further to obtain the wind direction which is perpendicular to it. In this experiment, the wind direction acquired from the ASAR image by the Gaussian-FFT algorithm also has a 180 ambiguity in direction. To resolve the 180 ambiguity, CCMP wind field datasets are taken into consideration to act as the wind field references. Besides, the wind direction computed with the Gaussian-FFT method is compared with the wind direction obtained by the Peak-FFT method. Then, the optimal wind direction (Gaussian-FFT wind direction) is input into the CMOD4 and CMOD5 models to compute the wind speed values respectively. Through comparing the wind field retrieval results with the CCMP datasets, we proved that it is valid to retrieve wind direction from ASAR image with Gaussian-FFT algorithm and it is achievable to obtain wind speed value over South China Sea with CMOD4 model. The approach used to obtain the wind field in this paper is of great significance to provide guidance to the wind field inversion in other waters of South China Sea, especially in areas that are lack of field observations. In addition, it is also critical for other researches whose specialties are related to oceanography, as this approach could offer vital wind parameters to these researches.

Key words： South China Sea wind retrieval; ENVISAT ASAR; Gaussian-FFT; CMOD4 model; CMOD5 model

Cite this article

TIAN Siyu , HUANG Xiaoxia , LI Hongga , WANG Hao , LI Xia , CHENG Peng . Wind Field Retrieval of South China Sea Based on Gaussian-FFT Method[J]. Journal of Geo-information Science, 2016 , 18(11) : 1544 -1550 . DOI: 10.3724/SP.J.1047.2016.01544

1 引言

海面风场对了解海洋中的物理过程以及海洋与大气之间的相互作用至关重要。而南海因其复杂的大气和海面状况,仅依靠传统的观测方法（如岸基观测站、船只和浮标等方法）通常很难获取海面风场的时间和空间序列,并且传统观测方法通常观测点少、费用昂贵、依赖天气气候。合成孔径雷达（Synthetic Aperture Radar,SAR）以其全天候、全天时、高空间分辨率的成像特点,为反演海面风场,尤其是面积宽阔且海面状况复杂的南海风场,提供了一种新的途径。海面风通过风应力对海面作用产生风生表面波,这些表面波直接或间接地改变了海面粗糙度,SAR通过海面成像记录海面粗糙度信息,进而进行海面风场反演。
利用SAR数据反演海面风场包括反演海面风向和风速2部分。反演海面风向的方法主要是通过SAR影像上与风向垂直的风条纹方向,获得具有180º方向模糊的海面风向。通常借助二维傅里叶变换（FFT方法）[1-2]、小波分析[3]、基于不同空间尺度的局部梯度估计[4-5]以及检测最大方差方向[6]等方法获取风条纹方向。这些方法中,FFT方法计算SAR图像谱准确且速度快。获得傅里叶图谱后,可通过以下2种方法获取风条纹方向：① 通过谱峰值连线获取风条纹方向;② 采用高斯曲线拟合图像傅里叶谱进而得到风条纹方向。采用这2种方法获得的风向,都具有180º方向模糊,需要参考其他辅助风场数据或者利用影像自身信息去除方向模糊,以获得海面风向。
由于卫星传感器波段不同,反演风速方法又可分为基于C波段、基于L波段以及基于X波段等不同的反演方法。基于C波段的反演方法有CMOD4[7],CMOD5、CMOD-IFR2,统称为CMOD[8-12]系列。王珂等[13]基于ENVISAT/ASAR IMP格式数据提出海面风速分段反演算法对青岛海岸附近的风场进行反演,为近岸海面风场反演提供了指导。此外,利用SAR数据还可估算恶劣天气下的海面风场。周旋等[14]利用台风“艾利”、“卡努”和“奥菲利娅”的星载SAR数据估算台风最大风速,估算结果与最佳路径数据基本一致,为台风监测提供了新的技术途径。针对L波段波长较长,其GMF（Geophysical Model Function）对中等风速及较大入射角敏感程度低于C波段GMF的特点,CIsoguchi等[15]将基于C波段的GMF算法进行了改进,得到适合L波段SAR数据反演风场的GMF。张毅[16]通过2次相反飞行方向获取的L波段机载SAR影像去除风向反演中出现的180º模糊,为去除方向模糊提供一种新的方法。基于X波段数据的风速反演模型有XMOD1[17]、XMOD2[18]、X-PR模型[19]。XMOD1用于SIR-X-SAR数据反演风场。XMOD2是使用TerraSAR-X和实测浮标数据开发的适合于X波段数据反演风场的非线性GMF。Shao等[19]通过对C波段的2个PR模型参数（Elfouhaily和Thompson）进行改进,得到2个适用于X波段的模型。
南海是中国面积最大的海域,获取南海风场对了解南海海洋过程具有重要意义。本文基于已获取的ENVISAT ASAR数据,将结合高斯曲线-FFT风向反演方法和CMOD4模型风速方法应用于南海风场反演,并通过与峰值-FFT风向反演方法和CMOD5模型风速反演方法的结果对比,验证了该方法在南海风场反演中的有效性,为南海风场反演研究提供一定参考。

2 实验数据及海面风场反演实验

2.1 实验数据

2.1.1 ENVISAT ASAR 影像
本文实验数据为ENVISAT卫星ASAR传感器于2006年4月16日14:13（UTC）获得的IMP模式数据,研究区域范围为：东经115°~116.5°,北纬21°~22.5°（图1中黑色方框）。影像是C波段（5.33 GHZ）,VV极化数据,方位向和距离向的像元分辨率均为12.5 m。进行风场反演前,先利用NEST软件对获取影像进行预处理,包括辐射校正、LEE滤波处理、地理投影等。
Fig.1 Preprocessed SAR image

图1 预处理后的SAR 影像

2.1.2 CCMP风场数据
CCMP（Cross-Calibrated Multi-Platform）风场资料从ESE（NASA Earth Science）获得,时间分辨率为6 h,空间分辨率为0.25°×0.25°。CCMP数据是通过变分同化方法融合了SSM/I、TMI、AMSR-E、QuikSCAT、ADEOS-II多源卫星数据及船舶、浮标实测等数据资料得到的海面风场数据。该数据已广泛应用于海洋风浪及台风研究中[20-21],其中也包括验证反演风场[22]。根据本次实验研究区域范围及ASAR影像的获取时间,截取CCMP数据在2006年4月16日12时（图2（a））和18时（图2（b））的数据,范围为东经115.125°~116.30°,北纬20.875°~22.125°。CCMP数据只在每天4个特定时刻（0:00、6:00、12:00、18:00）（UTC）有数值模拟数据,其他时刻风场数据只能通过插值获得,这并不能满足风场获取的时效性要求,因而有必要利用遥感影像反演海面风场进而获得海面风场信息。
Fig.2 CCMP wind data at UTC 12:00 and UTC 18:00 on April 16, 2006

图2 CCMP 风场2006年4月16日12:00和18:00时（UTC）数据

2.2 海面风场反演实验

2.2.1 风向反演方法
（1）FFT变换获取傅里叶谱
SAR影像因风浪等影响通常存在明暗相间的条纹,由风作用产生的明暗条纹称为风条纹,风条纹方向就是海面风向的垂直方向。SAR影像中风条纹的方向可从影像的二维波数谱中获得。由于波浪作用也能在SAR影像上产生明暗相间的条纹,故在提取风条纹方向前需先对影像进行滤波处理以去除波浪作用产生的条纹。海面波浪条纹尺度通常为几百米的数量级,风条纹的尺度通常为几公里的数量级,因而风条纹的信息主要集中在图像谱零频中心位置附近,大约距离零频中心2~8个像元。文中采用滤波的方法去除部分波浪条纹后,对滤波后的影像进行FFT获得SAR影像的二维波数谱。利用FFT获得SAR影像的傅立叶谱公式如式(1)所示。
Y l , m = ∑ j = 1 N ∑ k = 1 N X j , k e - 2 πi ( jl + km ) / N （1）
式中：Y为图像低波数谱;X为图像灰度值;l, m=1,2,3,…,N。
（2）二维高斯曲线拟合FFT谱
在反演实验中,SAR影像上风条纹特征并不明显,利用FFT方法获得图像的傅里叶谱后,通过分析发现,傅里叶谱上的双峰值现象并不突出（这与风条纹不明显现象一致）,而中心的拖尾现象明显,此时采用二维高斯曲线拟合FFT谱提取风条纹方向。高斯曲线拟合二维FFT谱可以提取这种拖尾现象的分布信息,包括拖尾现象的主轴及次轴方向信息,其中主轴方向与风条纹方向相同,次轴方向与风向相同。利用高斯曲线拟合能更全面有效地拟合影像的波谱信息,提取波谱信息中包含的风条纹方向信息。高斯曲线的拟合公式如式（2）-（5）所示。
F x , y = A 0 + A 1 e - U / 2 （2）
U = x ′ / a 2 + y ′ / b 2 （3）
x ′ = x - h cosT - ( y - k ) sinT （4）
y ′ = x - h sinT - ( y - k ) cosT （5）
式中： F x , y 为对应的 x , y 位置上的FFT谱值; A 0 和 A 1 分别是高斯曲线的常数项和比例项; a 和 b 分别是高斯曲线的在 x ′ 方向和 y ′ 方向的轴长; h , k 是高斯曲线的中心;T为二维高斯曲线长轴从X轴方向逆时针旋转的角度,该值与风条纹的方向相同或相反,与风向相垂直。获取风条纹方向后,可得到具有180º方向模糊的风向,然后参考CCMP数据风向去除方向模糊,得到海面风向。
2.2.2 风速大小反演原理
本文实验分别采用CMOD4模型和CMOD5模型进行海面风速反演。这2种模型均属于经验模式：假定雷达后向散射截面与风速的幂次方成正比,并以现场测量值为基础确定后向散射截面与入射角、风速和风向等因子的函数关系,是针对已有C波段VV极化微波散射计改进的经验公式。CMOD4[23]和CMOD5[24]分别如式(6)、(7)所示。
σ 0 = b 0 1 + b 1 cos ( ϕ ) + b 3 tan h b 2 cos 2 ϕ 1.6 （6）
σ 0 = B 0 1 + B 1 cos ( ϕ ) + B 2 cos 2 ϕ 1.6 (7)
式中： σ 0 是后向散射系数;b0、b1、b2、B1、B2、B3均是风速大小 ν 和入射角 θ 的函数; ϕ 是相对风向,即风向与天线方位角之差。从CMOD4模型公式可看出,对于C波段VV极化数据,后向散射系数是入射角、风速大小以及相对风向3个参数的函数。当已知入射角和相对风向时,便可求出海面风场的风速大小。

3 南海风场反演实例结果及分析

3.1 高斯曲线-FFT方法反演南海实验区风向

利用高斯曲线-FFT反演海面风向的主要步骤如图3所示。首先在ENVI软件中,将像元分辨降低到100 m。接着通过IDL编程实现该影像的风向反演。将分辨率为100 m的遥感影像分成每幅250像元×250像元的子图像（本文实验区内有6个子图像,子图像按照经度从低到高,纬度从高到低依次编号1到6,如图4所示）,对每幅子图像都进行中值滤波,去除部分与风速无关的低频率波。对经过上述处理的子图像进行FFT（快速傅里叶）变换,得到每幅子图像的傅里叶谱。从傅里叶图谱中可看出,子图像中拖尾现象明显,而峰值现象并不明显（如子图像1与子图像4中,中心有个亮斑,尺度大于2个像元）,故适合采用高斯曲线拟合傅里叶谱。图4的红色椭圆代表拟合的高斯曲线,红色实线方向代表风条纹方向,红色虚线方向代表风场方向或者风场方向的逆方向。
Fig.3 Flowchart of wind direction retrieval

图3 风向反演流程图

参考CCMP风场数据,将6幅子图的中心位置进行空间和时间插值,得到子图像中心在ASAR影像获取时刻的CCMP风场矢量插值数据（表1）。对比CCMP风场,去除风向模糊,得到每个子图像的风向,如表1中FFT风向。在实验中利用传统峰值连线方法计算FFT谱方向,与高斯曲线拟合方法进行对比。风向的起始方向是正北方向,顺时针旋转为正（本文所有风向方向皆以此为标准）。
Tab.1 ASAR sub-images’ wind direction retrieval results with respect to Gaussian-FFT method and PEAK-FFT method (°)

表1 ASAR子图像的高斯曲线拟合-FFT方法和峰值-FFT方法反演风向值(°)

子图名 高斯曲线拟合-FFT风向 峰值-FFT风向 CCMP风向 高斯曲线拟合-FFT风向与CCMP风向差值 峰值-FFT风向与CCMP风向差值
1 218.32 213.69 227.28 8.96 13.59
2 222.01 225.00 226.19 4.18 1.19
3 213.00 209.75 224.99 11.99 15.24
4 222.82 215.54 227.11 4.29 11.57
5 214.46 198.44 226.00 11.54 27.56
6 200.93 191.31 224.75 23.82 33.44
均值 10.80 17.10
方差 7.23 11.63
通过比较表1中FFT风向、峰值-FFT风向及CCMP方向发现,高斯曲线拟合-FFT风向与CCMP风向有10.8°左右的差值,方差约7.23°,均比峰值-FFT小,这说明高斯曲线拟合的方向与CCMP风向更接近,也说明在风条纹不明显时,利用结合高斯曲线拟合的FFT方法反演南海海面风向优于峰值-FFT方法,这与Gerling[25]得出的结论一致。因此,在反演南海海面风速时可采用高斯曲线拟合-FFT风向进行海面风速反演。
Fig.4 Fourier spectrum of sub-images and the related Gaussian curves

图4 子图像傅里叶谱及相应高斯曲线图

3.2 CMOD4模型和CMOD5模型反演风速大小

获取风向后,利用IDL语言编程实现CMOD4和CMOD5模型,将校正的后向散射值 σ 0 、高斯曲线拟合-FFT风向值及雷达入射角输入CMOD4模型和CMOD5模型,得到相应风速值（表2）。
Tab.2 Wind speed retrieval by CMOD4 model and CMOD5 model based on the wind direction by Gaussian-FFT

表2 高斯曲线拟合-FFT风向反演的CMOD4和CMOD5风速

子图名 CCMP风向/° CCMP风速值/(m/s) 高斯曲线拟合-FFT风向/° CMOD4风速值/(m/s) CMOD4风速值与CCMP风速值差/(m/s) CMOD5风速值/(m/s) CMOD5风速值与CCMP风速值差/(m/s)
1 227.28 11.29 218.32 14.46 3.17 19.25 7.96
2 226.19 11.57 222.01 14.19 2.62 17.57 6.00
3 224.99 11.77 213.00 15.85 4.08 18.53 6.76
4 227.11 11.21 222.82 13.16 1.95 17.13 5.92
5 226.00 11.44 214.46 14.15 2.71 18.05 6.61
6 224.75 11.61 200.93 17.50 5.89 21.38 9.77
均值 3.40 7.17
方差 1.41 1.47
将风速计算结果分别与CCMP风速值进行对比,发现CMOD4风速与CCMP风速差值的均值约为3.40 m/s,方差为1.41 m/s,均小于CMOD5模型相应值。从对比结果看,CMOD4风速值与CCMP风速值更接近。利用高斯曲线拟合-FFT风向值和CMOD4模型反演风速值,得到每个子图像的风场矢量（图5）。
Fig.5 Wind retrieval result of ASAR image

图5 ASAR影像风场反演结果图

4 结论

本文利用ENVISAT ASAR影像数据反演了南海部分区域的海面风场。在风向反演过程中,通过对比高斯曲线拟合-FFT方法和峰值-FFT方法反演结果,发现在风条纹现象不明显时,高斯曲线拟合-FFT风向与CCMP风向更接近。将高斯曲线拟合-FFT风向分别输入CMOD4模型和CMOD5模型得到风速大小值。通过2种反演风速值与CCMP风速值对比,发现CMOD4模型反演结果与CCMP风速值更接近。由此可看出,利用结合高斯曲线拟合的FFT方法反演南海海面风向是有效的。这对于反演南海大面积海域风场信息,具有重要的理论和实际意义。在后续工作中,将对南海其他部分区域进行风场反演实验,以不断改进反演南海大面积海面风场的方法。

The authors have declared that no competing interests exist.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Lehner S, Horstmann J, Koch W.Mesoscale wind measurements using recalibrated ERS SAR images[J]. Journal of Geophysical Research, 1998,103(C4):7847-7856.The precision images (PRI) of the synthetic aperture radars (SAR) on board the European Remote Sensing Satellites ERS-1 and ERS-2 are used to derive mesoscale wind fields over the ocean. For calculation of the wind speed the C-band model (CMOD4) is used, which was originally developed by Stoffelen and Anderson [1993] for the European Space Agency (ESA) to derive wind fields from measurements of the wind scatterometer (SCAT). In the case of the ERS-1/2 SAR the CMOD4 is used to compute the wind speed from the normalized radar backscatter cross section (NRCS) and the incidence angle of the radar beam, both computed from the SAR.PRI data. The third input variable is the wind direction, which is estimated from the wind streaks in the images or from ground truth measurements. The SAR data are affected by a power loss, caused by saturation of the analog to digital converter (ADC) of the SAR. Therefore the images have to be recalibrated. Errors in the derived wind speed are mainly due to ADC saturation and uncertainties of the input wind direction. These errors are estimated for various wind conditions. Mesoscale wind fields computed from ERS-1/2 SAR.PRI images taken between the Shetland Islands and the west coast of Norway are compared to ground truth measurements and modeled wind fields from the German weather service (DWD). Wind fields of the nonhydrostatic mesoscale model Geesthacht simulation model of the atmosphere (GESIMA) are compared to the derived wind field of the ERS-1 SAR.PRI image at the island R眉gen in the Baltic Sea.

DOI

[2]
Wackerman C C.2000, Estimating wind vectors from RADARSAT synthetic aperture radar imagery[R]. Ann Arbor, MI: Veridian ERIM International, 2000.

[3]
Du Y, Vachon P W, Wolf J.Wind direction estimation from SAR images of the ocean using wavelet analysis[J]. Canadian Journal of Remote Sensing, 2002,28:498-509.We present a method for the automatic estimation of wind directions from synthetic aperture radar (SAR) images of the ocean. The method is based on a wavelet analysis and assumes that the wind direction aligns with boundary-layer atmospheric roll vortices, which often appear as streaks at kilometre scales in SAR images of the ocean, and measures the orientation of the streaks. Unlike estimation methods that use the discrete Fourier transform (DFT), the streaks in SAR images are described quantitatively as a natural output of this method. Furthermore, more optimal wind directions are obtained by comparing the directional orientation of the streaks at different spatial scales. Sub-scenes in which the streaks are too weak to determine wind direction do not return a direction, as governed by a user-selected threshold. Wind directions for these sub-scenes are based on those in neighbouring sub-scenes by using an adaptive smoothing technique. Quality control involves tuning the threshold level. We apply the method to two examples of RADARSAT-1 SAR images. The results are compared with those of a DFT-based wind direction analysis, and it is shown that a robust wind direction field is obtained. Mesoscale wind structures can be described by using a finer computing grid. The estimated wind directions still include a 180掳 direction ambiguity.

DOI

[4]
Horstmann J, Lehner S, Kock W, et al. Computation of wind vectors over the ocean using spaceborne synthetic aperture radar[J]. The Johns Hopkins University Technical Digest, 2000,21(1):100-107.The high resolution and large coveeage of satellite synthetic aperture radar (SAR) offer a unique opportunity to derive mesoscale wind fields over the ocean surface and to investigate their spatial variation. For this purpose, different algorithms were developed and tested using SAR images from two European Remote Sensing satellites. In this article, the methods for deriving wind fields from SAR data are introduced. The wind directions are extracted from wind-induced streaks visible on most SAR images. Wind speeds are derived from the normalized radar cross section by applying an empirical C-band model. The different sources of error in wind retrieval that must be considered are discussed with respect to Radarsat/ScanSAR data. Furthermore, SAR-retrieved ocean surface winds are used to investigate the spatial variation of winds at different scales.

DOI

[5]
Koch W.Directional analysis of SAR images aiming at wind direction[J]. IEEE Transactions on Geoscience and Remote Sensing, 2004,42:702-710.were tested. This paper describes the local gradients method including the filtering of nonwind generated image features and gives some application examples.

DOI

[6]
Wackerman C C, Horstmann J, Koch W.Operational estimation of coastal wind vectors from RADARSAT SAR imagery[C]. Proceedings of International Geoscience and Remote Sensing Symposium, 2003.

[7]
Alpers W, Henninus I.A theory of the imaging mechanism of underwater bottom topography by real and synthetic aperture radar[J]. Journal of Geophysical Research, 1994,89(C6):10529-10546.A simple theoretical model of the imaging mechanism of underwater bottom topography in tidal channels by real and by synthetic aperture radar (SAR) is presented. The imaging is attributed to surface effects induced by current variations over bottom topography. The current modulates the short-scale surface roughness, which in turn gives rise to changes in radar reflectivity. The bottom topography-current interaction is described by the continuity equation, and the current-short surface wave interaction is described by weak hydrodynamic interaction theory in the relaxation time approximation. This theory contains only one free parameter, which is the relaxation time. It is shown that in the case of tidal flow over large-scale bottom topographic features, e.g., over sandbanks, the radar cross-section modulation is proportional to the product of the relaxation time and the gradient of the surface current velocity, which is proportional to the slope of the water depth divided by the square of the depth. To first order, this modulation is independent of wind direction. In the case of SAR imaging, in addition to the above mentioned hydrodynamic modulation, phase modulation or velocity bunching also contributes to the imaging. However, in general, the phase modulation is small in comparison to the hydrodynamic modulation. The theory is confronted with experimental data which show that to first order our theory is capable of explaining basic features of the radar imaging mechanism of underwater bottom topography in tidal channels. In order to explain the large observed modulation of radar reflectivity we are compelled to assume a large relaxation time, which for Seasat SAR Bragg waves (wavelength 34 cm) is of the order of 30-40 s, corresponding to 60-80 wave periods.

DOI

[8]
Wackerman C C, Rufenach R, Johannessen J, et al. Wind vector retrieval using ERS-I 1 synthetic aperture radar imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 1996,34(6):1343-1352.An automated algorithm intended for operational use is developed and tested for estimating wind speed and direction using ERS-1 SAR imagery. The wind direction comes from the orientation of low frequency, linear signatures in the SAR imagery that the authors believe are manifestations of roll vortices within the planetary boundary layer. The wind direction thus has inherently a 180° ambiguity since only a single SAR image is used. Wind speed is estimated by using a new algorithm that utilizes both the estimated wind direction and σ values to invert radar cross section models. The authors show that: 1) on average the direction of the roll vortices signatures is approximately 11° to the right of the surface wind direction and can be used to estimate the surface wind direction to within ±19° and 2) utilizing these estimated wind directions from the SAR imagery subsequently improves wind speed estimation, generating errors of approximately ±1.2 m/s, for ERS-1 SAR data collected during the Norwegian Continental Shelf Experiment in 1991

DOI

[9]
Vachon P W, Dobson F W.Validation of wind vector retrieval from ERS-1 SAR images over the ocean[J] . The Global Atmosphere-Ocean System, 1996,5(2):177-187.Wind vector retrieval from synthetic aperture radar (SAR) images over the ocean is shown to be feasible. A set of 16 radiometrically calibrated ERS-1 SAR images, including analog to digital converter saturation power loss correction, are compared with two ERS-1 scatterometer wind retrieval models driven by accurate in situ wind vector mesurements. It is shown that, if the wind direction is known and by using the CMOD4 or CMOD5 wind retrieval models, the agreement is to within ± 1 dB, representing a wind speed extraction error of ± 15 m/s, for the 3 to 12 m/s wind speed conditions encountered. This result validates the feasibility of extracting wind speed from a suitably-calibrated ERS-1 SAR image using a scatterometer wind retrieval model We demonstrate that the wind direction may be deduced from the low-wavenumber portion of the SAR image spectrum for the conditions represented in our data set. It is shown that agreement in wind direction is to within ±24°, thus presenting the opportunity for extracting the full wind vector from SAR images on a routine basis.

[10]
Fetterer F, Gineris D, Wackerman C C.Validating a scatterometer wind algorithm for ERS-1 SAR[J]. IEEE Transactions on Geoscience and Remote Sensing, 1998,36(2):479-492.The ocean surface wind field is observed from space operationally using scatterometry. The European Space Agency's (ESAs) ERS-1 satellite scatterometer routinely produces a wind product that is assimilated into forecast models. Scatterometry cannot give accurate wind estimates close to land, however, because the field of view of a spaceborne scatterometer is on the order of 50 km. Side lobe contamination, due to the large contrast in backscatter between land and water, compounds the problem. Synthetic aperture radar (SAR) can provide wind speed and direction estimates on a finer scale, so that high-resolution wind fields can be constructed near shore. An algorithm has been developed that uses the spectral expression of wind in SAR imagery to estimate wind direction and calibrated backscatter to estimate wind strength. Three versions, based on C-band scatterometer algorithms, are evaluated for accuracy in potential operational use. Algorithm estimates are compared with wind measurements from buoys in the Gulf of Alaska, Bering Strait, and off the Pacific Northwest coast by using a data set of 61 near-coincident buoy and ERS-1 SAR observations. Representative figures for the accuracy of the algorithm are 卤2 m/s for wind speed and 卤37掳 for wind direction at a 25-km spatial resolution

DOI

[11]
Kim D J, Moon W M, Nam S H.Evaluation of ENVISAT ASAR data for measurement of surface wind field over the Korean cast coast[C]. IEEE International Geoscience and Remote Sensing Symposium, 2003.

[12]
Nie C, Long D G.RADARSAT ScanSAR wind retrieval and rain effects on ScanSAR measurements under hurricane conditions[C]. IEEE International Geoscience and Remote Sensing Symposium, 2008.

[13]
王珂,洪峻,张问一,等.SAR 反演邻近岸海面风场方法[J].电子与信息学报,2013,35(8):1800-1805.该文对SAR反演邻近岸海面风场的有关问题进行了深入研究。 首先提出了邻近岸海面风向估计方法，在最小距离准则下利用邻近海域的风向估计所需的邻近岸海面风向。然后给出了使用 ENVISAT/ASAR 的 IM 成像模式PRI数据反演邻近岸海面风速的方法，比较了地球物理模型函数(Geophysical Model Function, GMF)模型性能，提出了海面风速分段反演算法。它们组成了完整的SAR反演邻近岸海面风场方法。通过实验、比较，验证了上述方法的有效性和合理性。

DOI

[ Wang K, Hong J, Zhang W Y, et al. Method of SAR retrieving ocean surface wind in near shore[J]. Journal of Electronics & Information Technology, 2013,35(8):1800-1805. ]

[14]
周旋,杨晓峰,李紫薇,等.基于星载SAR数据的台风参数估计及风场构建[J].中国科学:地球科学,2014,44(2):355-366.传统的星载SAR数据海面风场反演方法是利用海面风场与雷达后向散射系数之间的经验关系即CMOD5模式函数求解海面风场.但在台风条件下,由于降雨对雷达信号的影响及高风速条件下CMOD5模式函数的停滞效应,海面风场的反演精度迅速下降.针对降雨对雷达信号的影响,本文基于星载SAR卫星平台未搭载降雨测量载荷的特点,将多时次的静止气象卫星红外云图用于推导台风云系的运动矢量,并由该运动矢量及非同步观测降雨数据估算星载SAR数据过境时的降雨强度.最后,利用订正模型和降雨强度数据进行降雨订正.针对高风速条件下CMOD5模式函数的停滞效应,本文基于台风的SAR图像特征和改进的HOLLAND台风模型,提出了台风参数估计及风场构建方案.首先,利用基于小波分析的风向提取算法提取台风风场的海面风向信息,并通过地球物理模式函数和风向信息反演海面风速.然后,根据台风眼的SAR图像特征计算台风中心位置和最大风速半径,并将其代入改进的HOLLAND台风模型.最后,利用中低风速数据通过最小二乘法拟合台风中心气压和最大风速,并将台风风向、中心位置、最大风速半径、中心气压和最大风速等参量代入改进的HOLLAND模型构建台风海面风场.为了验证方案的精度,选择台风"艾利"、"卡努"和"奥菲利娅"的星载SAR数据进行试验,并利用美国联合台风预警中心和飓风研究中心的最佳路径数据和风场数据进行精度检验.结果表明,本文利用星载SAR数据估算的台风中心位置、中心气压、最大风速与最佳路径数据基本一致,构建的海面风场精度较高,其中,海面风速的均方差为1.4 m s-1,风向的均方差为2.1°,为台风监测提供了新的技术途径.

[ Zhou X, Yang X F, Li Z W, et al. Estimation of tropical cyclone parameters and wind fields from SAR images[J]. Science China: Earth Sciences, 2014,44(2):355-366. ]

[15]
Isoguchi O, Shimada M.An L band ocean geophysical model function derived from PALSAR[J]. IEEE Transactions on Geoscience and Remote Sensing, 2009,47(7):1925-1936.This paper examines L-band normalized radar cross section (NRCS) dependence on ocean surface wind. More than 90 000 match-ups, each consisting of the L-band HH NRCS, incidence angles, wind speeds, and wind directions, were collected from the Phased-Array L-Band Synthetic Aperture Radar (PALSAR) and scatterometer wind vectors. Based on the match-ups, the L-band HH NRCS dependence on incidence angle and wind vector is modeled for 0-20-m/s wind speeds and 17deg-43deg incidence angles. The derived relation indicates that the wind sensitivity of the L-band NRCS is less than that of the C-band at moderate winds and large incidence angles, whereas comparable at stronger winds ((>10 m/s) and small incidence angles. The upwind-crosswind difference is amplified in the 10-15-m/s range followed by an almost zero amplitude from 4 to 8 m/s, which represents a clear phase shift with the C-band VV and Ku-band HH models. Wind speeds are then estimated from the match-ups, based on the derived model function. A comparison with the reference scatterometer winds reveals a 0.05-m/s bias and a 1.85-m/s root mean square error, where crosswind data give rise to large errors due to low wind sensitivity at wind speeds of around 10 m/s, particularly at large incidence angles. The L-band NRCS behavior in strong winds (>20 m/s), at which the C-band is saturated, was not determined in the current model due to the limited number of the match-ups.

DOI

[16]
张毅. L 波段 SAR 海面风场反演研究[D].北京: 中国科学院研究生院, 2007.

[ Zhang Y.Study of ocean surface wind field retrievals from L-band SAR[D]. Beijing: Institute of Electronics, Chinese Academy of Science, 2007.

[17]
Ren Y Z, Lehner S, Brusch S, et al. An algorithm for the retrieval of sea surface wind fields using X band TerraSAR-X data[J]. International Journal of Remote Sensing, 2012,33(23):7310-7336.TerraSAR-X (TS-X) is a new, fully polarized X-band synthetic aperture radar (SAR) satellite, which is a successor of the Spaceborne Imaging Radar X-band Synthetic Aperture Radar (SIR-X-SAR) and the SRTM. TS-X has provided high-quality image products over land and oceans for scientific and commercial users since its launch in June 2007. In this article, a new geophysical model function (GMF) is presented to retrieve sea surface wind speeds at a height of 10聽m () based on TS-X data obtained with VV polarization in the ScanSAR, StripMap and Spotlight modes. The X-band GMF was validated by comparing the retrieved wind speeds from the TS-X data with observations, the high-resolution limited area model (HIRLAM) and QuikSCAT scatterometer measurements. The bias and root mean square (RMS) values were 0.03 and 2.33聽m s, respectively, when compared with the co-located wind measurements derived from QuikSCAT. To apply the newly developed GMF to the TS-X data obtained in HH polarization, we analysed the C-band SAR polarization models and extended them to the X-band SAR data. The sea surface wind speeds were retrieved using the X-band GMF from pairs of TS-X images obtained in dual-polarization mode (i.e. VV and HH). The retrieved results were also validated by comparing with QuikSCAT measurements and the results of the German Weather Service (DWD) atmospheric model. The obtained RMS was 2.50聽m swhen compared with the co-located wind measurements derived from the QuikSCAT, and the absolute error was 2.24聽m swhen compared with DWD results.

DOI

[18]
Li X M, Lehner S. Sea surface wind field by TerraSAR-X and Tandem-X data: algorithm development[J]. IEEE Transactions on Geoscience and Remote Sensing, 2012, in press.Ministry of Science and Technology of China (MOST); European Space Agency (ESA)

[19]
Shao W Z, Li X M, Lehner S, et al. Development of polarization ratio model for sea surface wind field retrieval from TerraSAR-X HH polarization data[J]. International Journal of Remote Sensing, 2014,35(11-12):4046-4063.In this article, the polarization ratio (PR) of TerraSAR-X (TS-X) vertical-vertical (VV) and horizontal-horizontal (HH) polarization data acquired over the ocean is investigated. Similar to the PR of C-band synthetic aperture radar (SAR), the PR of X-band SAR data also shows significant dependence on incidence angle. The normalized radar cross-section (NRCS) in VV polarization data is generally larger than that in HH polarization for incidence angles above 23掳. Based on the analysis, two PR models proposed for C-band SAR were retuned using TS-X dual-polarization data. A new PR model, called X-PR hereafter, is proposed as well to convert the NRCS of TS-X in HH polarization to that in VV polarization. By using the developed geophysical model functions of XMOD1 and XMOD2 and the tuned PR models, the sea surface field is retrieved from the TS-X data in HH polarization. The comparisons with in situ buoy measurements show that the combination of XMOD2 and X-PR models yields a good retrieval with a root mean square error (RMSE) of 2.03 m s

DOI

[20]
唐建,史剑,李训强,等.基于台风风场模型的台风浪数值模拟[J].海洋湖沼通报,2013(2):24-30.以CCMP风场资料为背景风场，结合藤田风场模型、Myers风场模型、Jelesnianski风场模型、Holland风场模型，分别构建全新的海面风场。基于非结构三角网格，采用第三代近岸海浪模式SWAN对2011年第5号强热带风暴“米雷”产生的台风浪进行数值模拟。比较SWAN模式模拟的结果和浮标实测数据，发现风场模型能够有效提高台风中心附近3至5倍台风最大风速半径范围内风场和台风浪有效波高的模拟精度。对比4种风场模型对应的台风浪模拟结果，发现Holland风场模型模拟的有效波高与浮标实测值最接近。

[ Tang J, Shi J, Li X Q, et al. Numerical simulation of typhoon waves with typhoon wind model[J]. Transaction of Oceanology and Limnology, 2013,2:24-30. ]

[21]
张鹏,陈晓玲,陆建忠,等.基于CCMP卫星遥感海面风场数据的渤海风浪模拟研究[J].海洋通报,2011,30(3):266-271.CCMP（Cross Calibrated Multi-Platform）风场数据是一种具有较高的时间、空间分辨率和全球海洋覆盖能力的新型卫星遥感资源。在充分分析CCMP海面风场数据可靠性 的基础上,以该卫星遥感海面风场数据为强迫输入项,运用第三代浅水波浪模式SWAN对渤海一次风浪过程进行了模拟,将模拟的结果与T/P、Jason卫星 高度计观测得到的有效浪高数据进行比较分析,发现两者相关性达到0.78,模拟结果平均偏高0.3 m。试验表明CCMP卫星遥感风场数据能满足海洋浪高预报需求,能在海洋数值预报和海洋环境研究中发挥重要作用。

DOI

[ Zhang P, Chen X L, Lu J Z, et al. Research on wave simulation of Bohai Sea based on the CCMP remotely sensed sea winds[J]. Marine Science Bulletin, 2011,30(3):266-271. ]

[22]
张康宇. 基于ASAR近海风场反演方法研究[D].杭州:浙江大学,2015.

[ Zhang K.Measurements of offshore ocean surface wind using ASAR[D]. Hangzhou: Zhejiang University, 2015. ]

[23]
Stoffelen A, Anderson D.Scatterometer data interpretation: estimation and validation of the transfer function CMOD4[J]. Journal of Geophysical Research: Oceans, 1997,102(C3):5767-5780.In this paper we estimate the 18 coefficients of the CMOD4 σ 0 -to-wind transfer function using a maximum likelihood estimation (MLE) method in order to improve the prelaunch function. We show that a MLE method has to be used with caution when dealing with a nonlinear relationship or with measurement errors that depend on the measured values. In the transfer function estimation it is crucial to use the components of the wind, rather than wind speed and direction, to use σ 0 in logarithmic units rather than physical ones, and to use well-sampled input data. In Stoffelen and Anderson [1997a] we showed that the triplets of measured backscatter are very coherent and, when plotted in a three-dimensional measurement space, they lie on a well-defined conical surface. Here we propose a strategy for validation of a transfer function, the first step of which is to test the ability of a transfer function to fit this conical surface. We derive an objective measure to compute the average fit of the transfer function surface to the distribution of measured σ 0 triplets. The transfer function CMOD4, derived in the first part of this paper, is shown to fit the cone surface to within the observed scatter normal to the cone, i.e., within roughly 0.2 dB, equivalent to a root-mean-square wind vector error of 650.5 m s 611 The second step in the validation strategy is the verification of retrieved scatterometer winds at each position on the cone surface. Scatterometer winds computed from CMOD4 compare better to the European Centre for Medium-Range Weather Forecasts model winds than real-time conventional surface wind data (ship, buoy, or island reports) with the root-mean-square wind vector difference typically 3.0 m s 611 . This surprising result can be explained by the so-called representativeness error. We further show that no significant spatial wind error correlation is present in scatterometer data and therefore conclude that the ERS 1 scatterometer provides winds useful for weather forecasting and climate studies.

DOI

[24]
Hersbach H, Stoffelen A, De Haan S.An improved C-band scatterometer ocean geophysical model function: CMOD5[J]. Journal of Geophysical Research: Oceans, 2007,112(C3):C03006.In this paper CMOD5, a new C-band geophysical model function (GMF), is derived on the basis of measurements from the scatterometer on board of the European Remote Sensing Satellite ERS-2. First-guess winds from the European Centre for Medium-Range Weather Forecasts were used as a reference for the period from August to December 1998, adding up to more than 22,000,000 collocations. CMOD5 corrects some deficiencies of the currently widely used CMOD4 GMF. Linear and higher-order wind speed corrections as computed with a triple collocation method are implemented. Recent measurements of extreme backscatter and wind obtained by aircraft and in situ data are fitted. Also, a more accurate fit of the two-dimensional cone surface in three-dimensional measurement space is established, especially in the regime of strong winds. These improvements result not only in better wind retrievals at high wind speed, but also in a more uniform performance across the ERS scatterometer swath. Moreover, the wind ambiguity problem has been reduced owing to the improved fit of the cone surface, resulting in about 75% skill of the first rank solution for winds above 10 m/s. These improvements aid the general usefulness of retrieved C-band scatterometer winds for climate and weather applications, and the ambiguity removal in dynamical and extreme weather conditions in particular.

DOI

[25]
Gerling T W.Structure of the surface wind field from the SeasatSAR[J]. Journal of Geophysical Research, 1986,91(C2):2308-2320.An intensive analysis of the vector wind field of one Seasat data set: pass 1339 is described. Wind speed and direction signatures are found in Seasat SAR (synthetic aperture radar) images, and the resulting estimates are compared with the Seasat-A scatterometer system (SASS) and simultaneous NOAA P-3 aircraft measurements. A power law is presented to relate SAR-measured backscatter to SASS-estimated wind speed, and the SAR estimates are shown to agree with the SASS estimates to within a standard error of 0.7 m/s over a range of wind speeds from 3 m/s to 13 m/s. The surface expressions of atmospheric roll vortices are apparent in several of the SAR images and may be responsible for the wind-direction signature in these cases. Wind field estimates averaged over regions of variable sizes are possible because of the high resolution of SAR imagery. A method for extracting low wave number directionality and its variability from SAR spectra is described, and SAR direction estimates obtained from spectra of 6.4-km-square images are shown to have a precision of approximately 10掳. Although a comparison data set that could validate these higher resolution estimates is lacking, averages over 40-km-square regions are in good agreement with the other wind field information. The SAR wind direction estimates yield a more complete interpretation in a region with a turning wind field near a front that is ambiguous with only SASS observations. In this region the flow patterns of the high-resolution estimates appear consistent with our knowledge of the overall circulation, considering all the observations. However, in another region the small-scale variability is too large and random to represent real wind variability, although averages derived from these estimates are still reliable.

DOI

Options
Outlines

/

〈 〉