View source for Use of ground based radar in hydrography

Friedwart Ziemer, Klaus-Werner Gurgel, Marius Cysewski and Stylianous Flampouris:
==Ground based radar in hydrography, basics and examples==

===1. Introduction===
Why ground based radar?

Radar remote sensing in Earth observation does not only cover the global and regional survey of geophysical parameters by satellite- or airborne radars, it also includes local observations by ground based radar techniques. Due to their continuous measurement capabilities, the information provided by ground based instruments installed at the coast or on offshore platforms does not suffer from the episodic character of the satellite- or airborne radar observations. In contrast to the satellite systems, the incident angle of ground based radars is only a few degrees and near to zero (grazing) at far ranges. In this case, the working range of the radar strongly depends on the operating frequency. Microwave radars are limited by line-of-sight propagation to the horizon, while decameter wave radars, also known as High Frequency (HF) radars, make use of ground wave propagation far behind the horizon. Due to the large available bandwidth in the microwave frequency range, the spatial resolution can be down to 5 m at 5 km working range, while HF radars provide about 1.5 km at up to 200 km working range. By installing a microwave radar onboard a ship, the observation area can be extended along the ship track to a regional scale. 

===2. Principals ===

====2.a) Principals of microwave radar application in hydrography====

In general radar remote sensing of the sea surface provides a broad variety of observations, which have been discussed intensively over the past decades. While single radar images give the intensity distribution of the backscattered radar power attained from a kind of instantaneously frozen surface, the surface dynamic can be observed by tracking features in subsequent radar measurements. The scattering mechanism is known as Braggscatter that interprets the backwards directed part of the radar power induced by constructive interference of the electromagnetic wave with the structures of the entire radar footprint (see figure 1). The spatial and time variability of the radar signature underlie the interaction of the sea surface with wind, waves, currents and the hydrosphere modulating as well radar intensity as well as radar frequency Doppler shifts. A profound overview on the use of radar in general and in its use in geophysics is given by Skolnik, (1990). 

[[Image:Radargeometrie.png|thumb|above|Figure 1]]

By acquiring radar intensity or radar Doppler shift an image of the related processes may be composed on a regional to local scale. By knowing the physics behind these modulations the steering hydrodynamic processes may be assessed. In this article we will focus on the use of ground based X-band radar as tool for hydrographic observations. In coastal management the regional and local survey by ground based radar provides a valuable completion to the in-situ observations. Information on local wind, wave and current conditions as well as indirect parameters like the local bathymetry, and even the coverage by ice, spills or slicks may be deduced routinely from radar products.
Although the microwave signal does not enter into the water, as almost 100% are reflected or scattered directly at the surface, it is possible to deduce additional information on the hydrodynamic from below the surface. For example the propagation of waves in shallow water allows deducing the local water depths out of the local wave propagation. Basing on this relation it is possible to compose maps of the water depth (bathymetry), which can be used in monitoring the sand transport [[3a. Microwave Radar: Bathymetric survey and current field observation using local gravity wave dispersion]]. Another example is the extrapolation of surface features down to the sea bottom as it is used by the operational model Vogelzang et al, (1997).

''Acquisition of radar cross section'' 
The amount of radar signal intensity that is detected at the receiving antenna is called radar cross section. The amplitude and phase of the cross section are instantaneously and locally modulated in time and space by hydrodynamic effects. The most important influence to the radar cross section of an ocean surface element is the mean amplitude of the surface roughness, which is instantaneously steered by the local wind impact. The radars under discussion here are operated synchronously with a precise navigation to compose the cross section values into a geo-coded radar map. While a radar pulse sweeps along a radial line, the received radar cross section is decomposed into a series of radial bins. The radar return is allocated to corresponding scattering surface elements by measuring the microwave runtime and the antenna azimuth. The focusing in azimuth depends on the length of the antenna and the in range on the duration of the transmitted radar signal (see: Figure 1). Typical values for the range resolution are 5 m x 5 m or 10 m x 10 m and for the antenna opening angle about 1°. Whereas the return of each individual radial beam is acquired during nanoseconds the acquisition of the total image takes some seconds or even minutes depending on the acquisition mode. The image composition under the use of a rotating antenna takes about 2 seconds and the product is a matrix in polar coordinates (range and azimuth).

''Acquisition of radar Doppler shift''
For the detection of Doppler shifts incurred to the microwave signal by the motion of the surface elements an acquisition time of at least some 100 milliseconds is necessary. During that time the antenna must be kept directed towards the individual surface elements. The longer the Doppler frequency information is integrated the more increases its significance. As the frequency shift is sensitive on the instantaneous local wind friction the knowledge on the wind impact is crucial. We demonstrate an application of the Doppler measurements in chapter [[3b. Microwave Radar:  Mapping by Radar Doppler Current Profiler (RDCP)]].

====2.b) HF Radar: Principals of radar application in hydrography====
HF radars are operated in the 3-30 MHz frequency range, where the electromagnetic waves can travel far behind the horizon by ground wave propagation along a good conducting layer (salty ocean water) or due to reflections by the ionoshpere (sky wave propagation). Ground wave systems are installed at the coast close to the water to ensure a good coupling to the ocean and achieve working ranges up to 200 nautical miles, while sky wave radars are normally installed far off the coast, e.g. the JINDALEE radar in the centre of Australia http://www.dsto.defence.gov.au/page/3984/ and may reach thousands of nautical miles range. Only a few sky wave systems exist world wide, which in most cases are not used for oceanographic applications. The University of Hamburg recently developed a ground wave HF radar for coastal applications called WEllen RAdar (WERA), Gurgel et al. (1999). Fig. 2 shows a typical example of an antenna installation set up for a WERA radar.
[[Image: ANT_PET_1_150DPI.png|thumb|above|Figure 2]]

The development of HF radars in hydrography started in 1955, when Crombie discovered that electromagnetic waves in the HF band are interacting with the ocean surface due to Bragg resonant scattering. The first radar for mapping of ocean surface currents was described by Barrick et al in (1977). A typical backscatter spectrum from a selected patch of the ocean at a given range and angle is shown in Fig 3: Two strong first-order Bragg peaks can be identified, which are generated by a resonant backscattering of the electromagnetic wave due to ocean waves of half the electromagnetic wavelength, e.g. 10 m long electromagnetic waves are scattered back by 5 m long ocean waves.
[[Image: spectrum.png|thumb|above|Figure 3]]
The phase speed ''C'' of these resonant ocean waves is given by the dispersion relationship.
<math>C = \sqrt{\frac{g \lambda}{2 \pi} \tanh{\frac{2 \pi h}{\lambda}} }</math>

Where ''g''  is the acceleration due to gravity, ''lambda'' is the ocean wave length, and ''h'' is the water depth. The shallow water term,<math>\tanh(\frac{2 \pi h}{\lambda})</math> , approximates to 1 and can be neglected,if the water depth is larger than the ocean wave length. As there are always ocean waves travelling towards and away from the radar, two strong first-order Bragg peaks can be observed. By changing the radar operating frequency, a specific ocean wavelength can be selected. Around the two first-order peaks, second-order side bands can be observed, which contain information on the ocean wave directional spectrum.

===3. Applications and examples===

====3.a) Microwave Radar: Bathymetric survey and current field observation using local gravity wave dispersion====

Special challenge in coastal protection management is provided by the management of sandy coastlines. Here men interact with the dynamic processes of the ocean with the shore line and undertake measures for their stabilization by measures like “beach nourishment” [http://www.encora.eu/coastalwiki/Image:Theme05_40.png]]. The interference of men with nature must be attended by intense observations. The increase of sand to the coastal system means the survey of the sand’s residence on the one side but on the other side the observation of the forces causing erosion, transport and deposition of sand. For this it is not enough to acquire time series of physical parameters at single observation points. It longs for area mapping observations to identify the response within the system and to assign the process to their origins, either occasional forcing by storms or to continuous forcing as e.g.: by the tidal cycle. We will point out in the following the methods of radar hydrography allowing area covering and continuous survey of the forcing as well as the response of the bathymetry. The area of investigation is List West at the north end of the Island Sylt in the German bight.

[[Image:Dispersion.png|thumb|Figure 4]]

In the current paragraph we present the assessment of the bathymetry and the current field by imaging the wave backscatter using marine radar. Time series of these images are inversed locally where the linear wave theory is valid and the surface wave dispersion holds. The algorithm used is known as “Dispersive Surface Classificator” and the commercial acronym is DiSC. The method has been developed at the Radar Hydrography Department of GKSS http://www.gkss.de and is licensed as commercial product by Vision 2 Technology GmbH http://www.v2t.de , partner of the Geesthachter Innovations- und Technologie-Zentrum (GITZ). The method is based on the linear wave theory that allows by tracking of wave crest in space and time to determine the parameters water depth and current vector. The dispersion relation of sea-surface waves is derived from the Eulerian equations of motion, the continuity equation and the dynamic and kinematic boundary conditions at the sea surface and the sea floor. A detailed description of the derivation of the dispersion relation is given in Senet et al., (2001). The analysis of the image sequences of the inhomogeneous wave field provides a set of physical parameters on a local spatial scale. The basic idea of the method is that in shallow waters the waves vary their propagation (phase speed and wave number) locally and thus impresses the local bathymetry into the image series. The same mechanism acts via the local current that may be assessed by the DiSC as well. 

[[Image:Bathymetries.png|thumb|Figure 5]]
The water depth and the current vector are free parameters influencing the shape of the dispersion shell in the wavenumber -frequency domain (see figure 5). It must be mentioned here that for this application the radar has not be calibrated, as the wave height is not influencing the wave dispersion. From the three dimensional radar observation the three dimensional power spectra in wavenumber -frequency domain is calculated by Fourier Transformation. Next the shape of the actual effective dispersion is deduced by fitting the detected power values to a plane approximating the dispersion. Deviations from the undisturbed dispersion are used to determine the local water depth and current vector, Senet et al., (2007). The local results are composed to spatial hydrographic-parameter maps.

The method is illustrated at the figure 6 and the main steps of analysis are the following: 
[[Image:DiSC_flowchart.jpg|thumb|Figure 6]]

*preservation of the complex-valued 3D FFT image spectrum,
*filtering techniques of complex spectrum to separate the wave signal from noise, in contrast to the global method that the power spectrum is filtered
*directional and dispersion separation of the complex-valued spectrum into spectral bins at 2D wavenumber planes of constant frequency, 
*2D inverse Fast Fourier Transformation(2D FFT<sup>-1</sup>) of the spectral bins, yielding complex-valued, one-component spatial maps in the spatio-frequency domain, 
*calculation of spatial maps of local wavenumbers from the one-component images of constant frequency, 
*composition of the one-component local wavenumber maps of constant frequency to local 3D spectra and 
*calculation of spatial hydrographic-parameter maps from the local 3Dspectra.

The result of the DiSC is the instantaneous local depth, the bathymetry, and the estimation of the current field. Figure 2a and 2b illustrate the bathymetries averaged over a tidal cycle each before and after storm (wind conditions 8-9 Bft.). The two maps have common reference, corrected by gauge measurements. 

[[Image:3Tide_cropped.gif|thumb|Figure 7]]

The difference of the sediment volume between the beginning and end of the storm, which under conditions could be considered as the sediment net increase of the area of investigation, is estimated in 50.000m3, with ±10% of accuracy for each grid cell, if we consider the accuracy of the method is approximately 90% per cell as it is given in the Flampouris et al., (2007). In general the assumption of the constant difference between two periods, of the mean sea level is not strong enough and decreases the accuracy of the calculation.

In Fig. 7, is illustrated an example of the current field in the same area as above, the wind conditions are approximately 5 Beaufort and the date of the data acquisition is the 12th July of 2001. The falsely too high current values close to the shore and over the shoal at the north west corner of the observation area indicate grid cells for which the linear dispersion is not valid.

====3.b) Microwave Radar:  Mapping by Radar Doppler Current Profiler (RDCP)====
The radar Doppler shift from the sea surface can be used for the detection of the sea surface currents. With a coherent radar system it is possible to measure the scatterers speed directly. The ground based coherent X-band radar system can be mounted either onshore or onboard a ship to observe sea surface scatterer velocities with high resolution. The antenna view direction and data acquisition is computer controlled; adjusting the antenna to fixed azimuthal angles while acquiring data is possible. Three different operating modi are possible and necessary for different applications: rotating antenna (also with different rotating speeds), stepping (changing view directions by set observing time) and the third mode is to work with fix view angle all the time. At a shore based permanent radar station the stepping mode is used for current and small scaled feature measurements like convergences/divergences or small eddies. Using the fixed view direction and a second radar antenna a full current vector can be achieved by moving ship and the covered area increases versus a shore set up. This method by scanning of horizontal current profiles we called Radar Doppler Current Profile (RDCP). The transmitters are synchronized to acquire the two orthogonal components of the sea surface current during a single ships passing.

[[Image:RDCP_Scheme.png|thumb|Figure 8]]

The Figure above shows the plan view of a ship with a mounted RDCP. In addition to the radars a precise navigation is needed to track the accurate position and the instantaneous radar view directions. ADCP measurements are acquired to validate the surface current observations and to extrapolate the current field in the vertical direction. To avoid interferences the alternating operation is controlled by the master radar (1) that interlaces every second pulse to the slave radar (2). By this each radar transmits and receives 1000 coherent pulses per second. The scan angle between the two antenna directions is 90°, with one antenna looking 45° ahead and the second looking 45° aft. By the Doppler relation we calculate the radial velocities from the backscattered signal for each range bin (length ~ 7.5m). Integrating the radar observations over a second the radial speed results in an accuracy of 1.5 cm/s. These values have to be corrected by the local impact of the actual wind friction and ship movement. An automatic quality control rejects routinely faulty data. For offshore application a motion sensor will be integrated in the system to correct the instantaneous antenna movements due to wave impact. The post processing procedure is to compose the full surface current vector by merging the two components into a geo-coded grid with the grid mash up to 10 m. 

[[Image:rdcp_result_new.png|thumb|Figure 9]]

The figures above 9. a) and b) show the speed and direction of the surface current field. An eddy with a diameter of about one nautical mile shows the outflow directed westward on the northern side of the gully and a counter current directed eastward on the south side. This is overlaid by a current modulation due to the change in the cross section over the under sea sand dunes, which interact with the ebb current in a way that we observe acceleration over the crest of the dunes and slowing down where the cross section is widening. For the comparison with the satellite SAR data it is of high interest to see that the directionality within the current field is in interaction with the dunes as well. It is evident that the current direction rotates into the main gully direction (clockwise), where the current speeds up above narrowing cross sections. The direction rotates anti clockwise in cross gully orientation slowing down currents above widening cross sections.

====3.c) HF Radar:  Measurement of surface current maps====

If an ocean surface current is transporting the Bragg resonant waves (cf. the two first-oder Bragg peaks in Fig. 3)
the first-order Bragg peaks are shifted in frequency, away from the value given by the dispersion relationship (Equ.1). This shift moves both peaks by the same amount and to the same direction and its amount can be used to calculate the radial component of the surface current. The surface current value measured represents the average from the surface down to about 16% of the ocean wave length, which is e.g.  the top 160 cm of 
the ocean at 15 MHz radar frequency.

Because one HF radar measures the radial component of the current, a second HF radar some 10 km apart is required to calculate the 2-dimensional surface current from the two radial components. Fig. 10 shows an example of a surface current field measured off Brest, France.

[[Image: Example_currents.png|thumb|Figure 10]]

As the 2-dimensional surface current is calculated from two radial components, the resulting accuracy depends on the errors of the radial components itself and on the angle between them. An angle of 90° between the radial components leads to the lowest errors, while angles around 0° or 180° dramatically increase the error orthogonal to the radial components. This degrade of accuracy is also known from the Global Positioning System (GPS) and is called Geometrical Dillution Of Precision (GDOP). To measure the accuracy, several comparisons between currents measured by current meters or ADCPs and HF radar have been made. The total RMS difference between these instruments is in the range 5-15 cm/s Essen  et al. (2000), Graber et al (1997). Note, that both systems by principle can not measure the same: The HF radar averages the current over an area of about 1 km<sup>2</sup>, while the in situ instruments always provide point measurements. Especially in areas with strong current shear due to topography or geostrophic effects, large deviations can be observed between the instruments.

====3.d) HF Radar:  Measurement of ocean wave spectra maps====

The second-order sidebands seen in Fig. 3 contain information about the ocean wave directional spectrum. Note, that these sidebands may be hidden by stronger first-order Bragg peaks shifted by a different current velocity 
from other directions in case of non-directive receiving antennas.

There are two kinds of algorithms to derive the ocean wave spectra from the second-order sidebands: Most are based on the inversion of the Barrick-Weber equations Barrick  et al (1977), e.g. the method developed by Wyatt (2000), which allows resolving bi-modal sea states. This theory reaches it's limitations, when the length of the 
Bragg-resonant ocean wave raises to the order of the significant waveheight. The other kind of algorithms is based on an empirical approach, finding calibration factors by linear regression. The algorithm described by Gurgel  et al. (2006) falls into this category. Fig. 11 shows an example of a wave field measured off Brest, France.

Due to problems to reliably identify the first-order Bragg peaks when the significant waveheight reaches the length of the Bragg-resonant ocean waves, both algorithms show saturation effects: At 27 MHz radar frequency, the wave measurement saturates at Hs ≈ 7.5 m, at 12.5 MHz, the limit is at Hs ≈ 16 m. During high sea state conditions, a lower radar frequency has some advantages, however, the lowest detectable sea state is higher in this case.

There are several papers discussing the accuracy of HF radar wave measurements. An extensive comparison has been published by Wyatt et al. (2003) within the EuroROSE (European Radar Ocean Sensing, EU Mast-3, CT98-0168) project.

[[Image: Example_waves.png|thumb|Figure 11]]

====3.e) HF Radar:  Measurement of wind direction maps====

By exploiting the ratio of the two first-order Bragg peaks in the backscatter Doppler spectrum (cf. Fig. 3), the direction of the Bragg-resonant ocean waves can be calculated. As these are quite short and refer to the wind-induced part of the ocean wave spectrum, they give a good indication on the wind direction. The wind-induced ocean waves follow changes in the wind direction with some 10-15  min.  delay. Fig. 12 shows an example of the wind direction measured off Brest, France.

[[Image: Example_wind.png|thumb|Figure 12]]

The algorithm to calculate the wind direction is to fit the ratio of the amplitudes of the two first-order Bragg peaks to a <math>\cos^s(0.5 \theta)</math> or <math>sech^s(1.0 \theta)</math> angular distribution model Gurgel et al. (2006). As this distribution model has a symmetric shape, the solution is ambiguous if one Bragg line ratio is available. A second Bragg peak ratio from a different direction, provided by a second HF radar, solves this ambiguity. A second radar site is required to measure ocean currents, anyway.

Wind speed measurements bt HF radar are still under investigation. The problem is, that with increasing winds the wave energy of the Bragg-scattering waves inceases less and less and nearly reaches a saturation. Instead of further inceasing the waveheight, the engergy is tranfered to longer ocean waves by nonlinear wave-wave interaction. Exploiting the level of the first-oder Bragg peaks is not sufficient isn this case, and the complete ocean wave spectrum must be considered.

===4. References===

Alpers, W., & K. Hasselmann (1982) Spectral signal to clutter and thermal noise properties of ocean wave imaging synthetic aperture radars. ''Source:'' Int. J. Remote Sensing, 3, 423-446

Barrick,D.E., Evans,M.W., Weber,B.L. (1977) Ocean surface current mapped by radar. “Source:” Science 198:138--144

Braun, N, F. Ziemer, & A. Bezuglov (2007) Sea-Surface Current Features Observed by Doppler Radar. ''Source:'' accepted by IEEE Transactions on Geoscience and Remote Sensing.

Chowdhury, M. (2007) Assessment of water flow and the impact on sediment motion in a tidal channel of north Sylt basing on radar observation. ''Source:'' Coastal Research Laboratory, Christian Albrecht University of Kiel. 

Crombie,D.D. (1955) Doppler spectrum of sea echo at 13.56 Mc/s. “Source:” Nature 175:681--682

Cysewski, M. (2003) Radarscanning in der Hydrographie. ''Source:''  Diplomarbeit GKSS 2003/26. Geesthacht.

Essen,H.-H., Gurgel,K.-W., Schlick,T. (2000) On the accuracy of current measurements by means of HF radar. “Source:” IEEE J.~Oceanic Engineering 25(4):472--480

Flampouris, S. (2006) Investigation of correlations between radar deduced bathymetries due to the outer impact of a storm in the area “Salzsand”. ''Source:''  M.Sc. Thesis; GKSS 2006/16. Geesthacht.

Flampouris, S., Ziemer F. & Seemann J. (2007) Survey of Bathymetry and Current Fields by Radar Image Series Acquired by Shore Based X-Band Radar. ''Source:''  eProceeding IGARSS 2007, Barcelona 23-27 July 2007.

Graber, H.C., Haus, B.K., Chapman, R.D., Shay, L.K. (1997) HF radar comparisons with moored estimates of current speed and direction: Expected differences and implications. “Source:” JGR 102-C8: 18,749--18,766 

Gurgel,K.-W., Essen,H.-H., Kingsley,S.P. (1999) HF radars: Physical limitations and recent developments. “Source:” Coastal Engineering 37(3-4):201—218

Gurgel,K.-W., Antonischki,A., Essen,H.-H., Schlick,T. (1999) Wellen Radar (WERA), a new ground-wave based HF~radar for ocean remote sensing. “Source:” Coastal Engineering 37(3-4):219—234

Gurgel,K.-W., Essen,H.-H., Schlick,T. (2003) The use of HF radar networks within operational forecasting systems of coastal regions. “Source:” In: Building the European Capacity in Operational Oceanography. 3rd International EuroGOOS Conference, Proceedings 245--250.  Elsevier, ISBN 0 444 51550 X

Nieto, J.C. & C. G. Soares  (2000) Analysis of directional wave fileds unsing X-band navigation radar. ''Source:'' Elsevier Science Coastal Engineering V. 40, pp 375-391

Plant, W.J., Keller, W.C. & K. Hayes (2005) Measurement of river surface currents with coherent microwave systems. ''Source:'' IEEE Trans. Geosci. Remote Sens., 43, pp. 1242-1257.

Senet, C.M., Seemann, J. & F. Ziemer (2001) The Near-Surface Current Velocity Determined from Image Sequences of the Sea Surface. ''Source:'' IEEE Transactions on Geoscience and Remote Sensing, Vol. 39, No. 3, pp. 492-505.

Senet, C.M., Seemann, J., Flampouris, S. & Ziemer, F. (2007) Determination of Bathymetric and Current Maps by the Method DiSC Based on the Analysis of Nautical X–Band Radar-Image Sequences of the Sea Surface. ''Source:'' accepted by IEEE Transactions on Geoscience and Remote Sensing.

Shearman,E.D.R. (1987) Chapter~5: Over-the-horizon Radar. In: Modern Radar Techniques, “Source:” Editor Scanlan,M.J.B. 200--240. COLLINS, 8 Grafton Street, London W1

Vogelzang, J., Wensink, G.J., Calkoen, C.J. & M.W.E. Van der Kooij (1997) Mapping submarine sandwaves with multi-band imaging radar 2. Experimental results and model comparison. ''Source:'' Journal of Geophysical Research, 102, pp. 1183-1192

Wyatt,L.R. (2000) Limits to the inversion of HF radar backscatter for ocean wave measurement. “Source:” J. Atmospheric and Oceanic Technology 17:1651--1665

Ziemer, F. & Cysewski, M. (2006) High Resolution Sea Surface Maps Produced by Scanning with Ground Based Doppler Radar. ''Source:''  eProceeding IGARSS 2006, [http://ieeexplore.ieee.org/Xplore IEEE] Denver, Colorado 31  July – 04 August 2006.

'''Links'''

Department Radarhydrographie [http://www.gkss.de] 

V2T – Vision to Technology GmbH [www.v2t.de]

Uni Hamburg HF Radar [http://ifmaxp1.ifm.uni-hamburg.de/]