- 1 -
EM Technical Note
Date: January 7th, 2000
Author: Erik Hammerstad
Subject: Backscattering and Seabed Image Reflectivity
1. Introduction
The Simrad EM multibeam echo sounders all have beam backscattering strengths and optionally seabed
image reflectivity as part of their data output. These data may be used for bottom classification, provided
that how the data is collected and processed is clearly defined. This note will describe how this is done in
the Simrad multibeam echo sounders. It will also show what corrections may be made in postprocessing
these data to remove the bathymetry dependent part of the data. Finally a comparison will be made with
ordinary sidescan sonars and some other multibeam echo sounders having imagery output.
2. Theory
The echo level, EL, of the signal backscattered from the bottom, may be derived from the sonar
equation:
Here SL is the multibeam echo sounder's source level, 2TL is the two-way transmission loss, and BTS
the bottom target strength. The transmission loss consists of two parts, one due to spherical spreading of
the signal, the other due to absorption loss in the water:
Here R is the range and α the absorption coefficient in dB/m.
The bottom target strength will depend both on the reflective property of the seabed, but also on the
extent of the bottom which contributes to the backscattered signal at any time. It is therefore usual to
define a bottom backscattering coefficient, BS, given in dB/m
2
, as the characterizing quantity for the
bottom reflectivity. The backscattering area will be bounded by the beam geometry, as defined by θ
x
and
θ
y
, at normal incidence (0° incidence angle or 90° grazing angle) while in other directions it will be
bounded by the alongtrack beamwidth, θ
x
, and the transmit pulse length, τ.
EL = SL - 2TL+ BTS
2TL = 2 R+ 40 R
α
log
BTS = BS + 10
R
for = 0
xy
2
log
θθϕ
°
BTS = BS + 10
c
2
R for > 0
x
log
sin
τ
ϕ
θ
ϕ
°
- 2 -
How the backscattering coefficient varies with incidence angle, ϕ, is of course an important part of
seabed characterization and in determining the type of material which is on the seabed surface.
The receivers of the multibeams have limited dynamic range and a time variable gain (TVG) is therefore
run during the ping to avoid overload or having the echo return buried in noise. The TVG must be
predicted before reception, and must be devised so that the average signal level in the receiver is at an
optimum level so as to be able to cater for random variations in bottom reflectivity. The limiting factor
here is the A/D converters which with 12 bit as in most systems have 66 dB dynamic range. An
additional reason for running such a TVG is that it will flatten the beam sample amplitudes. This is
beneficial for bottom detection, but also important for display of the seabed image, where one is
primarily interested in reflectivity contrasts, which resolvability is strongly limited by the number of
colors or gray shadings available (or discernable) in today's printers.
When the EM 12 was designed in 1990, an investigation of the literature was done to get an idea of how
backscattering coefficients varied with incidence angle. Unfortunately, most of the reported results dealt
with low grazing angles, i.e. outside the region of interest. The conclusion that was drawn was that for
incidence angles larger than about 25° a good approximation for most conditions would be to assume
that a uniform flat bottom is characterized by a mean backscattering coefficient, BS
O
, and that angular
variation is given by Lambert's law, i.e.:
A paper (by Gensane in IEEE JOE Jan. 89) described measured backscattering coefficients versus
grazing angle also gave data near normal incidence. The paper did show deviations from the Lambert's
law, very little though for incidence angles from 40° to 80° but somewhat larger between 25° and 40°
(but the data given in Urick's standard underwater acoustics book fits Lambert's law well also in this
region). For smaller incidence angles, a reasonable fit to the data could be achieved by the simple
scheme of assuming that the backscattering coefficient changes linearly with incidence angle from BS
N
at
0° to BS
O
at 25°. Thus, after recognizing that an incidence angle of 25° is equivalent to about R = 1.1R
I
,
and replacing the trigonometric functions by the equivalent expressions in R and R
I
, the full model is:
From later literature no reasons for changing the above model has been found, with one exception. The
crossover angle between the two regions has been shown to be quite variable depending on material
type, and can be anywhere in the 5-30° region.
3. Implementation
BS =
BS
+20 ( )
O
log cos
ϕ
BTS =
BS
+ 3.162 R /R-1(BS-
BS
)-5 (R/
R
)
[(R /
R
)
-1]
+10
c
2
R for R<R<1.1
R
N
I
ON
I
2
I
2
x
II
log
log
τ
θ
BTS =
BS
-5 (R/
R
)
[(R /
R
)
-1]+10
c
2
R for R 1.1
R
O
I
2
I
2
x
I
log log
τ
θ
≥
BTS =
BS
+10
R
forR
R
Nxy
2
I
log
θ
θ
≤
- 3 -
The TVG law run in the Simrad multibeam echo sounders is based upon the above model with a fixed
crossover angle between the two regions of 25°.
1) Based on previous pings, the range to normal incidence, R
I
, and the backscattering coefficients at
normal and oblique incidence, BS
N
and BSO, are estimated.
2) The fixed gain is set to provide an echo level maximizing dynamic range.
3) The gain is varied in time or equivalently range according to the above model.
In the gain setting, nominal values are used for source level and receiver sensitivity. After beamforming,
the sample amplitudes are corrected for processing gain, and beam pointing angle dependent variations
in source level and receiver sensitivity (for the deep sea systems corrections are also applied for the
different frequencies used in different transmit sectors). Finally the used BS
O
is subtracted which gives
correctly scaled backscattering strengths in the area outside of normal incidence, but still with the ususal
normal incidence peak flattened to the same level as outside.
After bottom detection further corrections are applied to take into account any errors in the estimate of
the range to normal incidence and any lack of gain at the extreme ranges or too much gain applied at the
lesser ranges (the latter may occur due to limitations in the dynamic range possible in the TVG circuitry).
The data which are provided in the seabed image datagrams are picked from the beam amplitude samples
in such a way that when fitted together the total array of samples represent a continuous set along the
bottom with a fixed interval in range according to the range sampling rate of the multibeam echo sounder
and the mode it is used in. Due care is taken for beams with a lesser detected range than its neighbor
closer to the nadir beam, and to some extent data is picked to avoid holes due to beams lacking valid
detections. The sample corresponding to the detected range in a beam is identified in the datragram to
allow correct placements of the imagery samples on the bottom. The used absorbtion coefficient, R
I
, BSN
and BS
O
are stored with the data to allow a user to fully take into account the model used and apply any
corrections required. Furthermore on the latest systems (excepting the EM 12, EM 950 and EM 1000)
the data are corrected according to an operator settable angle for where the crossover from the normal
incidence to Lambert’s law region is to take place (this angle is also logged with the data).
The BS values given in the depth datagrams are an average value of the sample amplitude values. Short
averaging lengths are used and the maximum average level within a beam is chosen to represent the
beam BS. However when the echo in a beam is very short the maximum sample amplitude is chosen
(usually near normal incidence in shallow waters). In contrast to what is done in the seabed imagery
datagrams the backscattering strengths around normal incidence are corrected for the TVG law used in
this region, and the effect of the Lambert’s law assumption taken out. Thus the BS values in the depth
datagrams are correctly scaled at all angles.
The result of the implementation is that the measured seabed image amplitudes are “correct”, i.e. they are
the seabed's backscattering coefficients, or at least these are recoverable in postprocessing. An inherent
uncertainty in the values due to variation in transducer sensitivities may be estimated to be typically ±1
dB, but this may be larger on a specific system, and for a specific sample at least ±3 dB (especially in the
EM 950/1000 due to transmit pattern ripple and less overlap between receive beams, which again can be
corrected in postprocessing).
4. Postprocessing
- 4 -
For seabed classification it seems reasonable to assume that the beam backscatter coefficients will be
sufficient, especially when equidistant beam spacing is used. As all incident angle dependence and
backscattering level assumptions have been removed from this data, the only parameter used in the realtime calculations which may need to be corrected is the scattering area which will not be correct if the
bottom is not really flat. The areas used are:
A = ψTψ
R
R2 around normal incidence
A = ½cτ ψTR/sinφ elsewhere
Here c is the sound speed, τ the pulse length, and ψT and ψ
R
are the transmit and receive
beamwidths respectively. The first equation for the area is valid until the bottom incidence
angle is larger than the largest of the following two angles, given by:
cos
φ
τ
L
c
D
1
1
1
2
=+
−
sin
φ
ττ
L
RR
D
c
D
c
2
2
1=− +
+
ΨΨ
The above equations assume the bottom to be flat, and may require a slight revision when the bottom has
a significant slope in any direction, i.e.
For the seabed imaging data there are a number of things that can be done with depending on what one is
looking for. Some possible corrections are the removal of the Lambert's law, corrections for actual
bottom slope, calculation of true backscatter coefficients (i.e. reinsertion of angular dependence),
application of a different model for backscatter variation with incidence angle (such as using Gesand's
suggestion to model incidence variation by γ
0
- 10 log(sin
2
ϕtgϕ) for incidence angles larger than 30°, γ
0
being chosen to give a smooth transition to the model). Note that as many corrections are small and vary
little with angle, perhaps except near normal incidence, quite often they need only be calculated and
applied beam by beam.
0)(
2
Rc
10=
A
y
y
x
x
r
≠
ϕ
ϕ
ϕ
θ
τ
sincos
log
r
xy
2
xy
A
=10Rlog
cos cos
θθ
ϕϕ
- 5 -
It should be emphasized that the seabed image data are not corrected for beam pattern variations in the
receiver beams. This may be required where the beams are not strongly overlapping, such as in the
middle of the swath when running equidistant beamspacing when the number of beams is low such as on
the EM 1000. It may also be required when the coverage of a beam has been extended due to a failed
detection in a neighboring beam, or in the extreme outer part of the outermost beams. The beam pattern
variation with angle is given in most textbooks on sonar technology or radio antennas or radar, and is
generally in the form of a sinc function (sinx/x), but note that the pattern is not symmetrical with angle
when appreciable beam steering has been applied such as on the EM 12.
5. Comparisons
The major differences between the seabed imagery derived from a Simrad multibeam echo sounder and
an ordinary sidescan sonar are:
• The sidescan sonar does not measure calibrated backscattering strength, but only amplitudes with an
unknown absolute level and only compensated by a 20/30/40 logR TVG.
• The sidescan sonar data may be slant range corrected, but only presuming a flat bottom and cannot
thus be correctly scaled taking into account actual variation in bottom slope.
• The sidescan sonar is usually towed closed to the bottom and most data is collected at low incidence
angles for shadow detection purposes. In contrast, most of the multibeam data is collected at higher
incidence angles for classification purposes.
Finally it must be observed that the resolution of a sidescan may be better as it is cheaper to provide a
sidescan sonar with narrower alongtrack beamwidth and higher range sampling rates than a multibeam
echo sounder, allthough the specifications of the newer Simrad systems are quite good in this regard.
Other multibeam echo sounders (Seabeam, Atlas) are also to some extent capable of providing imagery
data. However, as far as we know, the logged data of these systems are not such that they allow the
postprocessing to absolute levels, resolution or correct geographical scaling as for the Simrad systems.
The records provided seem to be of the same data as would usually go to a greyscale recorder of an
ordinary sidescan sonar. Thus the number of points is limited, the scaling is not correct, and there seems
to be limited knowledge provided with repsect to how the data is acquired and what the internal
processing has done with the data.
Lambert's law removal can be done by adding 20 log(R/R
I
) to all data points. Correcting for erroneous
RI (such as could occur when the central beams do not have valid detections) would involve
calculating and applying the difference between the applied TVG and the correct TVG. Correcting for
non-flat bottom involves removal of used Lambert's law and replacing it by 20 log(sine of 2D
incidence angle) plus the same area correction as described above for the beam backscatter
coefficients. Use of a different model would likewise involve removal of used Lambert's law,
correction for area, and subtraction of model's backscatter using the actually encountered 2D
incidence angle. To obtain true backscatter coefficients involves removal of used Lambert's law, area
correction, and then correction for used backscatter coefficient in the flattening process near normal
incidence.
Loading...