ESA Earth Home Missions Data Products Resources Applications
    02-Sep-2014
EO Data Access
How to Apply
How to Access
3rd ERS SYMPOSIUM Florence 97 - Abstracts and Papers
Determination of Sea Ice and Ocean Wave Parameters using Complex ERS SAR Data
Estimation of wind, wave and ice paramters at th
Services
Site Map
Frequently asked questions
Glossary
Credits
Terms of use
Contact us
Search


 
 
 

Estimation of wind, wave and ice parameters at the ice boundary by using active microwave systems of the ERS satellites

S.Lehner, J.Schulz-St., R. Bamler   DLR-DFD, D82230 Wessling
lehner@dfd.dlr.de
 

Abstract

The presented work is a progress report on the ERS AO 2 Project 'Swell travelling into Sea Ice'. In this project synthetic aperture radar images are used to study the wave dynamics at the ice boundary. Two-dimensional ocean wave spectra are derived from the intensity images using nonlinear inversion methods. From the scattering of the waves at the ice boundary sea ice parameters are determined. In additon the phase information of the single look complex SAR images is used to generate different looks seperated by about half a second. These can be used to detect the dynamics of the sea surface that takes place during this time interval. In the case of ocean waves the phase speed of the waves is derived. Using the theoretical dispersion relationship this can give additional information on currents or ice parameters.
Using the SCAT CMOD4 algorithm on recalibrated SAR images wind speed at the ice boundary is derived and compared to measurements using the azimuthal cross correlation algorithm and to ground truth.
A first digital elevation model of the sea surface derived from X-SAR data taken during a single pass interferometric airborne campaign in February 1997 is presented.

Keywords: Ocean Waves, Sea Ice, Cross Spectra

Introduction

The ice boundary is an important part of the global climate system. As relevant meteorological parameters like windspeed, ice thickness and ocean wave spectra are difficult to obtain by in situ measurments, satellite observations are of great importance. Space born SAR is of special interest in this context because of its all weather imaging capabilities. In this study we investigate how the different state of the art wind speed and ocean wave algorithms can be used to study the meteorological conditions in the marginal ice zone.

It is assumed that the radar backscatter of ocean waves is dominated by Bragg scattering. As the roughness of the sea surface is influenced by wind speed and the roughness of ice is related to the ice type, calibrated SAR images can be used to derive wind and ice parameters [1],[2],[3],[4].
Windspeed at the ice boundary is derived from the normalized radar backscatter cross section (NCRS) of recalibrated SAR images, using the ESA CMOD4 algorithm [5]. This algorithm was originally developed for the ERS scatterometer (SCAT). As the SCAT and SAR on board ERS-1/2 operate at the same frequency, CMOD4 can be applied to the SAR. Using full swath 100x100 km ERS SAR images mesoscale windfields at a very high spatial resolution of about 100 m can be derived.
In contrast to the SCAT the SAR collects data though only from one antenna, therefore, the wind direction is needed as further input to derive the wind speed by the CMOD4. Usually, SAR images show distinct features like wind streaks or shadowing behind coasts from which the wind direction can be estimated to high accuracy. The results of the CMOD4 wind measurements are compared to the results using the Ifremer azimuthal correlation algorithm (IAC) on complex SAR data [6],[7], where the azimuthal cutoff is used to determine windspeed. Complex SLC data are used to estimate propagation direction and phase speed of ocean waves by cross spectral techniques.

Representative Scenario

FIGURE 1: SAR image of the sea ice boundary near Greenland and two subscenes showing ocean waves in open water (lower part) and in ice (upper part)

 

To demonstrate the different methods a 40 by 150 km SLC (complex) quarter scene of a full swath ERS-1 SAR image, taken at 23.01.92, 23:32 UTC, near the south east coast of greenland (Orbit 2736, Frames 1341, 1359) is studied (see figure 1). The lower left part shows open water, further north to the coast different ice types can be detected, the black part referring to grease ice, the brighter part to pancake ice. In the water at the lower left part of the image a 500m wave travelling in azimuth direction and a weaker 200m wave in range direction can be seen. The long ocean waves can be seen travelling into the ice until they are completely damped out in the upper part of the image.

The weather map showed, that the long azimuthally travelling waves were generated the day before by a strong polar low with windspeeds up to 25 m/sec. At the time the SAR image was taken the windspeed has slowed down to less than 4 m/sec allowing even azimuthally travelling waves to be imaged by the SAR system without being blurred out by velocity bunching effects, that are mainly due to the orbital velocities of the short wave wind sea.

Calibration

FIGURE 2 : Range dependent radar cross section for open water (lower left part) and different ice types

Originally for calibration the grey levels of SAR images were to be converted into NCRS values in a straightforward manner, just using a single calibration constant K. Due to a problem of saturation of the ERS analog to digital converter, the calibration becomes more complicated and a special recalibration algorithm [8],[9] has to be used, taking the replica power and a power loss term into account.

The IRP/RRP factor is part of the internal sensor calibrations and the power loss factor is needed to correct for the saturation of the analog to digital (ADC) converter, which occurs for bright areas (like fast ice or water even in medium wind conditions). For ERS-2 SAR the gain settings were reduced and therefore the problem of power loss is less severe. The recalibration works well for homogeneous SAR images but problems occur at sharp boundaries like e.g. icebergs in sea ice [4],[10].

Figure 2 shows the SAR image of figure 1 squeezed in azimuth and the average value of NCRS over three different areas of this image. In the pancake sea ice region the NCRS value is fairly constant over range at about -8 dB. In the grease ice region the NCRS drops to -20 dB, ocean waves can no longer be detected. In the water the drop off in backscatter from near range to far range is used to estimate the windspeed. As power loss plays an important role only for NCRS values above -5 dB, in this case recalibration is relevant only for the water areas of the image.

Wind Speeds

FIGURE 3: Radar cross section for different wind speeds and incidence angles

The backscatter from the rough ocean surface for moderate incidence angles of 20 deg to 60 deg is modelled by resonant Bragg scattering [11]. The backscatter signal is caused by the water wave component which is in resonance with the incidence radiation. The resonant water wave number k_w is related to the electromagnetic wave number k_el of the radar according to

where alpha is the local incidence angle of the radar beam. In case of ERS SAR, operating at C-band with incidence angles between 20 deg and 26 deg the range of scattering wavelenghts extends from 8.2 cm to 6.5 cm. Therefore the NRCS can be used to evaluate parameters, which influence the small scale roughness, like the wind speed.

Figure 3 shows the expected backscatter values using CMOD4 for the ERS SAR incidence angles when the wind is blowing towards the antenna for windspeeds from 2 to 26 m/sec in steps of 4 m/sec. The dashed and dotted lines show the respective backscatter values for ERS-1/2 SAR if recalibration were not applied. The open water area in the lower left part of figure 1 shows NCRS values from -4 dB to -7 dB for incidence angles between near range and about 21 degrees. For wind blowing towards the antenna this corresponds to a windspeed of about 4 m/sec using CMOD4. The corresponding ground truth data from the European Weather Bulletin confirm very low windspeeds of 4 m/sec for this area.

FIGURE 4: Azimuthal autocorrelation function for 7 subscenes of the sea surface

Another algorithm to estimate windspeeds was developed using the width of the peak of the autocorrelation function (ACF) in azimuth direction as described in [6]. This width can be deduced either from the ACF of the PRI image or the azimuthal cross correlation between different looks of the SAR SLC image. In this algorithm a Gauss function is fitted to the peak of the ACF and the width of this function is determined. The width is then used to determine the local windspeed. Windspeed measured by this algorithm showed good correlation to windspeed by scatterometer for a large dataset of imagettes [7],resulting in a linear relationship between windspeed U and width of the ACF. The imagette dataset was taken though in the open ocean with very few imagettes showing azimuthally travelling waves or surface features that will affect the azimuthal cutoff. In the case shown here however, in the presence of long azimuthally travelling waves, the mean orbital velocity is not dominated by local windspeed and the algorithm fails completly. Figure 4 shows the ACF for 7 subscenes of the SAR image from the sea surface. It can be seen that the ACF is dominated by the long wave field of the 500m wave on the image. It can be shown though, that the width of the ACF reduces for the waves in the sea ice.This is due to the much lower speeds of the backscattering ice floes, in comparison to the speed of the modulating Bragg scattering ripple waves in the water area.

Cross Spectra of Ocean Waves

It is well known that the SAR imaging of ocean waves is a strongly nonlinear process, so that the wave structures seen in SAR images cannot be interpreted in a straightforward manner. In the two scale modell [12] it is assumed, that the scattering mechanism is dominated by Bragg scattering where the long ocean waves are imaged due to the modulation of the short Bragg waves. Apart from this modulation, motion effects play an important role, resulting in the strongly nonlinear velocity bunching mechanism.

Due to the 180 degree directional ambiguity and the azimuthal cut off, which is caused by the velocity bunching, the mapping relation from ocean wave spectrum to SAR image spectrum cannot be inverted without using some a priori knowledge e.g. a wave model.

In the case of ERS-SAR every point scatterer at the sea surface is illuminated for about 0.7 seconds. By using the phase information of the complex SLC data it is possible to select subintervalls of this integration time, creating different looks of coarser spatial resolution, which are seperated in time as much as 0.5 sec. As the ocean waves , depending on wavelength, only move about the order of one pixel size during this time the shift of the waves can hardly seen by eye. The direction of the wave movement can be analyzed however, by computing the cross spectrum of the two looks .

Compared to the classical image spectrum the cross spectrum has been shown to have two major advantages [13] :

1) It provides an estimate of the image spectrum that is not biased by pedestal noise.

2) It contains information about the wave propagation direction and the phase speed, since it uses looks taken at different times.

While in [13] real and imaginary part of the cross spectra were analysed seperately, we additionally exploit the phase information, which is a direct measure of phase velocity.

Figure 5 shows the absolute value of the cross spectra computed from the water and sea ice subscenes presented in figure 1.

FIGURE 5: Cross spectra of ocean waves in open water (bottom) and ice (top)

In figure 6 the real and imaginary part of the cross spectrum of the open water subscene are presented. The propagation direction of the two wavesystems are indicated by the positive peaks in the antisymmetric imaginary part.

FIGURE 6 : Real and imaginary part of cross spectrum of subscene in the open water

Figure 7 shows a cut through the peak of the azimuthal wave. The dotted lines represent the phase, which follows from wave theory in the case of deep water. It can be seen that the coincidence for wavenumbers carrying high energy, the only area were phase information is meaningful, is quite good.

FIGURE 7: Cut through cross spectrum in azimuthal direction

FIGURE 8 : Real and imaginary part of cross spectrum of subscene in ice

Figure 8 shows the cross spectrum of the subscene in ice. It can be seen that the range travelling wave of the open water subscene is damped out in the ice. Additionally a short azimuthally travelling wavesystem can be observed. The propagation direction of the two azimuthally travelling waves is again given by the positive peaks in the imaginary part.

Outlook

In oceanography SAR was so far mainly used to measure the radar cross section of the sea surface. Now it becomes possible to use SAR interferometry to deduce digital elevation models of the sea surface. Because of the short coherence time of the sea surface this is of course only possible using single pass interferometry, which up to today is only available in aircraft systems. Figure 9 shows a digital elevation model of the sea surface taken in the German Bight on the 6.2.97 at 15:32 . The sea surface was imaged by an airborne X-Band SAR System of AEROSENSING using single pass interferometry. The image shows a range travelling wave with estimated wavelength 20m and a waveheight of 1.4m. These values agree very well with those predicted by the ECMWF model. This dataset gives the unique opportunity to evaluate the SAR imaging mechanism in a straightforward manner. For 1999 a spaceborn X-SAR mission (SRTM) using single pass interferometry is scheduled on the shuttle [14].

FIGURE 9: Digital elevation modell of the sea surface

Conclusions

Data of space born SAR can be used to derive wind, ice and wave parameters at the ice boundary.

When accurately calibrated images are needed, like when using CMOD4 to derive windspeeds, generally the use of ERS-2 data is recommended. For the use of CMOD4 some additional information on wind direction is needed - this can often be deduced from the SAR image itself. The azimuthal correlation algorithm does not need additional information on wind direction, but is highly sensitive to azimuthally travelling ocean waves or surface features affecting the cutoff of the azimuthal correlation. The two algorithms should therefore be used to complement each other.

Using complex SAR data and cross spectral techniques, it is possible to derive additional information like direction and phase speed of the ocean waves.

Single pass SAR interferometry is a very promising technique to derive digital elevation models of the sea surface, which will give further insight into the SAR imaging process of the sea surface.

Acknowledgements: The ERS SAR images were provided by ESA under the project number AO2.D146-2. The project was funded by the German ministry of research BMBF in the joint research project 'sea ice' under contract number 03PL 018 A

References

[1] Alpers, W.R., Bruemmer, B.,1994 :
Atmospheric Boundary Layer Rolls Observed by the Synthetic Aperture Radar Aboard the ERS-1 Satellite, J. Geophys. Res., Vol. 99, pp. 12613--12621
[2] Johannessen, O.M., J.A. Johannessen, A.D. Jenkins, K. Davidson, D.R. Lyzenga, R. Schuchman, P. Samuel,
H.A. Espedal, J. Knulst, E. Dano and M. Reistad, 1996 :
COAST WATCH-95: ERS-1/2 SAR Applications of Mesoscale Upper Ocean and Atmospheric Boundary Layer Processes off the Coast of Norway, IGARSS'96
[3] Wadhams, P., 1973 :
Attenuation of swell by sea ice, J. Geophys. Res.,78, 3552-3563
[4] Lehner,S., Horstmann, J., Koch, W., Rosenthal, W., 1997:
Mesoscale Wind Measurements using recalibrated SAR data, submitted to JGR
[5] Stoffelen, Ad and D. Anderson, 1993:
Characterisation of ERS-1 Scatterometer Measurements and Wind Retrieval, Proc. Second ERS-1 Symposium -- Space at the Service of our Environment, Hamburg, Germany, ESA SP-361, pp. 997--1001
[6] Chapron, B., T. Elfouhaily and V. Kerbaol, 1994:
A SAR Speckle Wind Algorithm, Proceedings of the Second ERS-1 Workshop, IFREMER Brest, BP 70, 29280 Plouzane, France
[7] Kerbaol, V. , B.Chapron, Elfouhaily T. and Garello R.,1996
Fetch and wind dependence of SAR azimuth cutoff and higher order statistics in a mistral wind case, IGARSS'96, Lincoln, Nebraska, USA, May 1996
[8] Meadows, P.J. and P.A. Wright, 1994:
ERS-1 SAR Analogue to Digital Convertor Saturation, Proceedings of the CEOS SAR Calibration Workshop, Ann Arbor Michigan, pp. 24--37
[9] Laur, H., P. Bally, P. Meadows, J. Sanchez, B. Schaettler and E. Lopinto, 1996 :
Derivation of the Backscattering Coefficient sigma_0 in ESA ERS-1/2.SAR.PRI Data Products, Issue 2, Rev.1, ESA
[10] Rosenthal, W., S. Lehner, J. Horstmann and W. Koch, 1995:
Wind measurements using ERS-1 SAR, Proc. of the Second ERS Applications Workshop, London, UK, pp. 355--358
[11] Wright, J., 1968:
A new modell for sea clutter, IEEE Trans. Antennas and Propag., AP-16, 217-223
[12] Hasselmann, K, Hasselmann,S., 1991
On the nonlinear mapping of an ocean wave spectrum into a SAR image spectrum, J. Geophysi. Res. C96, 10713-10729
[13] Engen, G., Johnson, H., 1995
SAR-Ocean wave inversion using image cross spectra}, IEEE Trans. Geosci. and Remote Sens., vol. 33, no. 4, pp. 1047-1056
[14] Bamler,R., Eineder,M., Breit, H., 1996:
The X-SAR single -pass interferometer on SRTM: expected performance and processing concept, Proc. of the EUSAR conference

Keywords: ESA European Space Agency - Agence spatiale europeenne, observation de la terre, earth observation, satellite remote sensing, teledetection, geophysique, altimetrie, radar, chimique atmospherique, geophysics, altimetry, radar, atmospheric chemistry