improved vessel squat modeling for hydrographic and navigation applications using kinematic gnss...
TRANSCRIPT
ORIGINAL ARTICLE
Improved vessel squat modeling for hydrographic and navigationapplications using kinematic GNSS positioning
Alireza A. Ardalan • Mohammad-Hadi Rezvani
Received: 27 June 2012 / Accepted: 25 April 2013
� Springer-Verlag Berlin Heidelberg 2013
Abstract The squat phenomenon, that is, the sinkage of a
vessel due to its motion can affect the safety of navigation
and reduce the accuracy of hydrographic bathymetry.
Therefore, it is necessary to model and predict the squat of
vessels as a function of cruise speed. We present a Global
Navigation Satellite Systems–based squat modeling
method for both hydrographic and navigation applications.
For implementation of the proposed method, onboard GPS
antennae configurations are offered to model bow squat for
full-form ships such as supertankers or ore–bulk–oil car-
riers as well as stern squat for fine-form vessels such as
passenger liners or container ships. In the proposed meth-
odology, the onboard GPS observations are used to deter-
mine cruise ground speed, heave, attitude, and controlling
the quality of kinematic positioning via fixed baselines.
The vessel squat is computed from ellipsoidal height dif-
ferences of the onboard antennae with respect to a refer-
ence state, after removal of all disturbing effects due to
roll, pitch, heave, tide, vessel load, and geoidal height
variations. The final products of the proposed approach are
the analytical squat models usable for hydrographic and
navigation applications. As the case study, the method is
applied to a survey vessel in the offshore waters of Kish
harbor. Numerical results indicate that the experimental
precision of the derived analytical squat models is in the
range of 0.003–0.028 m. The computed navigation squat of
the test vessel at a speed of 12.64 knots is 30 % of the
vessel draft and about twice its hydrographic squat.
Although the field test was performed on a survey vessel,
the method can be applied to any ship at any waterway.
The proposed method can address the inevitable demand of
reliable squat models for delicate hydrographic projects
and high-speed marine traffic.
Keywords Ship squat modeling � GNSS � Hydrographic
surveying � Safety of navigation � Bathymetry � Marine
traffic
Introduction
The vessel squat, that is, the vertical drop caused by water
pressure fall beneath the keel of a moving vessel, can
increase the draft and the risk of grounding by reducing the
under keel clearance (UKC). The squat effect in hydro-
graphic surveying can also reduce the accuracy of
bathymetry if it is not considered and removed from the
observations. The sinkage of a vessel due to the squat can
be accompanied by changes in pitch and roll, which makes
the maximum squat to appear at either port or starboard
sides of bow or stern of the vessel depending on its
structural characteristics. Generally, full-form ships such as
supertankers or ore–bulk–oil carriers, which are technically
characterized by the block coefficient Cb [ 0.7, have bow
squat, while fine-form vessels such as passenger liners or
container ships (Cb \ 0.7) experience stern squat (Derrett
1999). Figure 1 schematically illustrates UKC, draft, stern
squat, and bow squat.
The squat in confined channels depends on the breadth
and depth of the channels, in addition to cruise speed
(Briggs 2006; Maynord and Briggs 2006; Delefortrie
A. A. Ardalan (&) � M.-H. Rezvani
Department of Surveying and Geomatics Engineering,
Center of Excellence in Geomatics Engineering and Disaster
Prevention, College of Engineering, University of Tehran,
P. O. Box 11155-4563, Tehran, Iran
e-mail: [email protected]
M.-H. Rezvani
e-mail: [email protected]
123
GPS Solut
DOI 10.1007/s10291-013-0326-8
2010). However, in unrestricted waterways or deep chan-
nels, the squat is mainly driven by cruise speed (Varyani
2006). According to Feng et al. (1996), squat is a signifi-
cant component of the UKC allowance. Therefore, the
modern demand for high-speed and safe marine transport,
especially in confined channels and restricted waterways,
requires prediction of the squat as a function of cruise
speed (Feng and O’Mahony 1999; EL-Kader et al. 2003).
Moreover, the development of reliable bathymetry for
delicate hydrographic operations, such as pipe laying,
drilling rig installation, and dredging, requires squat
correction. As Global Navigation Satellite Systems
(GNSS) like GPS became available, the geodetic commu-
nity began to explore GPS applications in squat modeling.
Generally, GNSS-based squat models rely on ellipsoidal
height differences of the onboard antenna with respect to a
reference motionless state while tidal changes are removed
(Gourlay and Cray 2009). In this respect, in addition to a
nearby base GPS antenna onshore, onboard GPS antennae
configuration at bow, port, and starboard of bridge wings is
used for bow squat determinations and UKC predictions of
full-form vessels (Maynord and Briggs 2006; Giorgi et al.
2010). The non-parallelism of geoid and reference ellip-
soidal surface as well as the pitch-derived sinkage must
also be considered for correct squat modeling (Zilkoski
et al. 1999; Harting et al. 2009).
We differentiate between squat modeling for hydro-
graphic and navigation applications. For the safety of
navigation, the modeled squat must encompass the UKC
variations resulting from roll, pitch, and the sinkage caused
by cruise speed. Therefore, the maximum squat, which
results in the minimum UKC, must be determined by
monitoring the squat at port and starboard sides of bow or
stern of a vessel. This in turn requires proper configuration
of the onboard GPS antennae. In contrast, for hydrographic
surveying, since bathymetry is usually corrected for the roll
and pitch effects, the computed squat must only include
vertical motion of the echo sounder transducer due to
cruise speed.
For squat modeling, we offer correction formulae which
consider the lever-arm coordinates of the onboard GPS
antennae with respect to the center of gravity (CG) of the
vessel, roll, pitch, heave, tide, vessel load, and geoidal
height variations. The roll, pitch, and heave information is
needed to remove the rotational and vertical disturbances
from ellipsoidal heights of the onboard GPS antennae
proportional to their lever-arms. The tidal information
during the field operations is required for the removal of
instantaneous sea-level variations from the ellipsoidal
heights. The vessel load variations, which are mainly
caused by the fuel consumption, must be observed in order
to remove the corresponding effects. Finally, the geoidal
height variations are needed to account for the effect of
non-parallelism of reference ellipsoidal surface and geoid.
The proposed method in this study can be explained as
follows: (1) Installation of three GPS antennae onboard,
and at least a base antenna at the wharf near to the area of
the field operations. The onboard GPS antennae must be
installed at the following locations depending on the type
of the ship: bow-port, bow-starboard, and stern-port/stern-
starboard for full-form ships, and stern-port, stern-star-
board, and bow-port/bow-starboard for fine-form vessels.
In the methodology, the onboard GPS observations provide
attitude, heave, and cruise ground speed information. In
(a)
(b)
UKC
UKC Squat
Sea surface
Sea surface
Seabed
Seabed
Draft
Squat
Sea surface
Seabed
Draft
Draft
UKC
Draft forward
Draft forward
Draft forward
Draft aft
Draft aft
Draft aft
UKC forward UKC aft
UKC aft
UKC aft
UKC forward
UKC forward
(c)
Fig. 1 The squat effect of a vessel on its UKC and draft. a Vessel at
rest, b in motion with stern squat for a fine-form vessel, and c in
motion with bow squat for a full-form ship
GPS Solut
123
addition, the fixed baselines between the onboard antennae
are used to control the quality of kinematic positioning. (2)
Starting GPS observations at the wharf before departure to
resolve the ambiguity and to develop the initial motionless
reference state of the squat modeling. (3) Departure from
the wharf and sailing with stepwise increasing cruise speed,
that is, keeping the speed constant for a while and then
increasing the speed in incremental steps while GPS
observations are continued. (4) Computation of the vessel
squat at 1 s interval with respect to the reference state
while considering the aforementioned corrections. (5)
Application of the least squares (LS) to the computed squat
values during the constant cruise speeds in order to take
advantage of the redundant values. (6) Developing the
analytical squat models as a function of cruise speed for
hydrographic and navigation applications by fitting suitable
functions to the corresponding LS-derived squat values
based on the robust M-estimations.
In the next section, the mathematical setup for squat
modeling based on the proposed method will be presented.
Then, the experimental setup for practical implementation
of the method will be offered in order to prepare for the
presentation of the results and discussions about the field
experiment in the subsequent sections. The last section is
devoted to the conclusions and final remarks.
Mathematical setup
Since the squat is the vertical displacement of a moving
vessel, it should be observable by kinematic GPS posi-
tioning with respect to the motionless reference state of the
vessel. According to Fig. 2, the difference of the ellipsoidal
height h(ti) of an onboard GPS antenna at epoch ti with
respect to the ellipsoidal height h(t0) at the reference epoch
t0 can provide us with the squat of the vessel, once all
disturbing effects have been removed. The disturbing
effects are due to attitude, heave, tide, vessel load, and
geoidal height variations between the two epochs. There-
fore, if S(ti) is the squat associated with the epoch ti when
the vessel is sailing at the cruise speed vðtiÞ, we have
SðtiÞ ¼ hðt0Þ � hðtiÞ þ dhAðtiÞ þ dhHðtiÞ þ dhTðtiÞþ dhLðtiÞ þ dhNðtiÞ ð1Þ
where dhAðtiÞ, dhHðtiÞ, dhTðtiÞ, dhLðtiÞ, and dhNðtiÞ are the
corrections due to attitude, heave, tide, vessel load, and
geoidal height variations between the two epochs. Note that
in (1), we have followed the convention of positive squat
‘‘down’’ (Briggs 2006).
According to Melachroinos et al. (2011), the effect of
attitude variations on the ellipsoidal height of the onboard
GPS antenna dhAðtiÞ can be computed by knowing the roll,
pitch, and the offset of the antenna from the CG of the
vessel (Fig. 3). Let the attitude parameters, which are the
rotation angles between the so-called body frame (b-frame)
and local-level frame (ll-frame), populate the rotation
Fig. 2 The ellipsoidal height difference of the onboard GPS antenna
between two epochs t0 and ti, which is equal to the squat at epoch tiplus the disturbing effects on the GPS antenna height at epoch ti with
respect to epoch t0
(a) (b)
Fig. 3 The ll- and b-frames, lever-arm coordinates {bkx , bk
y , bkz } of antenna k, and the proposed GPS antennae configurations for a fine-form
vessels, and b full-form ships
GPS Solut
123
matrix Cllb , then the transformation from the b-frame to the
ll-frame is given by
rll ¼ Cllbrb ð2Þ
where rll and rb are the position vectors of the onboard
GPS antenna in the respective frames.
As illustrated by Fig. 3, the b-frame is the coordinate
system fixed to the vessel with the center at the CG, x-axis
toward starboard, y-axis toward bow, and z-axis along the
vertical axis of the vessel that is positive upward. The ll-
frame is centered at the CG, with x-axis toward geodetic
east, y-axis toward geodetic north, and z-axis along the
ellipsoidal normal that is positive upward (Fig. 3). The
vector of attitude variations drllðtiÞ at epoch ti with respect
to the epoch t0 can be computed from (2) by
drllðtiÞ ¼ ðCllbðtiÞ � Cll
bðt0ÞÞrb ð3Þ
where the z component of drllðtiÞ, the so-called attitude
induced heave dhAðtiÞ, is as follows
dhAðtiÞ ¼ � sinðrðtiÞÞ cosðpðtiÞÞbx þ sinðpðtiÞÞby
�
þ cosðrðtiÞÞ cosðpðtiÞÞbzg � � sinðrðt0ÞÞ cosðpðt0ÞÞbxfþ sinðpðt0ÞÞby þ cosðrðt0ÞÞ cosðpðt0ÞÞbz
�
ð4Þ
where {rðt0Þ, pðt0Þ}, and {rðtiÞ, pðtiÞ} are the measured roll
and pitch angles at epochs t0 and ti, and the symbols bx, by,
and bz denote the lever-arm coordinates of the onboard
GPS antenna with respect to the b-frame.
For the roll and pitch determination by GPS, the fol-
lowing iterative procedure can be used. According to the
GPS antennae configurations shown in Fig. 3, for fine- and
full-form ships, the following relation holds
r ¼ �sin�1 dh13
l13
� �� r0 ð5Þ
where r is the roll angle, dh13 the ellipsoidal height
difference between antennae 1 and 3, l13 the spatial
distance between the two antennae, and r0 is the roll offset.
Next, after removing the effect of the roll angle from the
ellipsoidal height difference dh12 between antennae 1 and
2, the pitch angle p can be computed for both fine- and full-
form ships as
dh012 ¼ dh12 þ ðb1z � b2
z ÞðcosðrÞ � 1Þ � ðb1x � b2
xÞ sinðrÞ� �
ð6Þ
p ¼sin�1 dh0
12
l12
� �� p0 for fine-form vessels
�sin�1 dh012
l12
� �� p0 for full-form ships
8<
:ð7Þ
where l12 is the spatial distance between antennae 1 and 2,
dh012 the roll-removed height difference between the
antennae, and p0 is the pitch offset. Then, the effect of
the pitch angle on the ellipsoidal height difference of
antennae 1 and 3 can be removed by
dh013 ¼ dh13 þ ðb1z � b3
z ÞðcosðpÞ � 1Þ þ ðb1y � b3
yÞ sinðpÞn o
ð8Þ
Having derived the pitch-removed dh013, the procedure
must be started from (5) by substituting dh13 by dh013 until
the differences between the computed roll and pitch
angles in two consecutive steps become negligible. This
completes the iterative algorithm to determine the vessel
attitude.
The next correction dhHðtiÞ appearing in (1) is due to
heave at epoch ti with respect to epoch t0. Since GPS
antennae, like all other onboard sensors, is fixed to the
vessel, the GPS-derived heave is along the z-axis of the
b-frame (IHO 2005, p. 195). However, what we need is the
heave effect with respect to the ll-frame, therefore, dhHðtiÞcan be derived by using (2) as
dhHðtiÞ ¼ cosðrðt0ÞÞ cosðpðt0ÞÞHðt0Þf g� cosðrðtiÞÞ cosðpðtiÞÞHðtiÞf g
ð9Þ
where the heaves Hðt0Þ and HðtiÞ are determined via the
difference of the instantaneous ellipsoidal heights of the
onboard antenna from a time-averaged height in a moving
window with a certain time span, for example, 25 s
(Godhavn 2000).
Next, the tidal correction dhTðtiÞ at epoch ti with respect
to epoch t0 can be derived as
dhTðtiÞ ¼ TðtiÞ � Tðt0Þ ð10Þ
where Tðt0Þ and TðtiÞ are the tide gauge observations at
epochs t0 and ti at a nearby tidal station. Note that in (10),
we assume that the selected tidal station is representative of
the sea-level variations for the whole area of field opera-
tions. If this is not the case, then the co-tidal and co-range
corrections must be applied.
The next correction appearing in (1) is the vessel load
variations dhLðtiÞ at epoch ti with respect to epoch t0.
Therefore, by measuring water-level marks Lðt0Þ and LðtnÞat the beginning (t0) and the end (tn) of the field operations,
dhLðtiÞ can be derived using a linear interpolation as
dhLðtiÞ ¼ti � t0tn � t0
� �LðtnÞ � Lðt0Þ ð11Þ
The last correction dhNðtiÞ in (1) corresponds to the
variations of the geoidal height of the onboard GPS
antenna due to movements of the vessel. A global
geopotential model can be used to derive the geoidal
heights at epochs t0 and ti, that is, N0 and Ni, and from that
dhNðtiÞ can be computed by
dhNðtiÞ ¼ Ni � N0 ð12Þ
GPS Solut
123
So far there has been no differentiation between squat
determination for hydrographic and navigation applications.
Indeed, (1) corresponds to squat modeling for hydrographic
applications since the roll and pitch effects are removed.
However, for navigation applications, such effects must be
retained in order to arrive at the maximum squat which
results in the minimum UKC. Consequently, for navigation
applications, the squat computation formula for the cruise
speed vðtiÞ at the antenna location can be written as
SðtiÞ ¼ hðt0Þ � hðtiÞ þ dhHðtiÞ þ dhTðtiÞ þ dhLðtiÞþ dhNðtiÞ ð13Þ
where all the notations are the same as (1).
Since the field operations must be performed with a
stepwise increase in cruise speed, while the speed is kept
constant for a short time during each step, there will be
redundant squat values at each speed which allows the LS-
derived squat values. After that, the analytical squat models
for hydrographic and navigation applications can be
derived by fitting suitable functions to the corresponding
LS-derived squat values.
Experimental setup
We now present the details of practical implementation of
the proposed squat modeling using a test field in the off-
shore waters of Kish harbor and using the fine-form survey
vessel of National Geographic Organization (NGO) of Iran
with 27 m length, 6-m beam, and 1-m draft. To develop the
squat models for both navigation and hydrographic appli-
cations, three GPS antennae were installed onboard at the
bow-port, stern-port, and stern-starboard locations. In order
to increase the accuracy of GPS attitude determination, the
onboard antennae were mounted on outstretched arms from
both sides of the vessel as schematically shown in Fig. 3a.
In this way, the distances between the antennae along the
transverse and longitudinal axes of the vessel became
approximately 15 m.
The lever-arm coordinates of the onboard GPS antennae
and the echo sounder transducer were determined by sur-
veying measurements when the vessel was on a dry dock
for repair. Table 1 offers the measured lever-arm coordi-
nates of the aforementioned sensors with respect to the
b-frame.
Additionally, a base GPS antenna was established at the
wharf close to the test field for the relative kinematic
positioning. The field operations were performed in a calm
day to avoid excessive motions of the vessel and related
disturbances on the kinematic positioning. The GPS
observations were started about 20 min before departure
when the vessel was moored at the wharf and continued
throughout the sailing.
The sea-level information during the field operations
was taken from a nearby coastal tide gauge station. The
water-level marks were measured from four corners of the
deck at the beginning and the end of the field operations to
record the vessel load variations. Table 2 presents the main
features of the experimental setup.
Results and discussions
After the field operations, the Bernese GPS Software 5.0
(Dach et al. 2007) was used for precise point positioning as
well as carrier phase double-difference processing, in order
to obtain accurate kinematic coordinates of the onboard
GPS antennae. Among the processing settings, the 5-s
satellite clock corrections obtained from the Center of
Orbit Determination in Europe (CODE) (ftp://unibe.ch/
aiub/CODE/) and the 3� cutoff angle are worth mentioning.
The position information of the onboard GPS antennae was
used to determine the cruise ground speed, attitude, and
heave, which are needed for the squat modeling, as well as
controlling the quality of kinematic positioning through the
fixed baselines. Figure 4 shows the results of the quality
control with some outstanding disturbances. It is found that
those disturbances are related to the turns of the vessel,
which are accompanied by sudden speed reductions.
Table 1 The lever-arm coordinates of the onboard sensors
Sensors bx (m) by (m) bz (m)
Ant. 1 at stern-port -8.719 -3.985 4.650
Ant. 2 at bow-port -8.553 10.412 4.673
Ant. 3 at stern-starboard 6.325 -3.834 4.562
Echo sounder transducer 0.000 3.065 -0.965
Table 2 The main features of the experimental setup
Items Descriptions
Onboard GPS receivers Dual-frequency JAVAD
(TPS LEGACY)
Onboard GPS antennae Choke ring JAVAD
(JPSREGANT_DD_I)
Base GPS receiver LEICA GX1220
Base GPS antenna LEIAX1202
Range of the cruise ground speed 0.00–12.67 knots
Range of the water depth 22.08–44.89 m
Sampling rate of the GPS
observations
1 s
Sampling rate of the tide
gauge observations
10 min
Duration of the field operations 50 min
Maximum tracked satellites 11
GPS Solut
123
Moreover, we found that the biggest disturbance in Fig. 4
at about 5.3 9 104 s GPS time is connected to the U-turn
of the vessel. Further investigations revealed that such
disturbances could be resulted when wrong or no antenna
phase center variations (PCV) correction is applied, and/or
orientations of the onboard GPS antennae change (Kouba
J., personal communication, Geodetic Survey Division,
National Resources Canada). Both were the case in the
field operations.
Using the derived formulae, we computed the roll, pitch,
and heave of the vessel. Figures 5 and 6 show the results.
As Fig. 5 shows, the pitch angle increases as the vessel
speed increases, and its fluctuations are caused by the
sudden speed reductions during the vessel turns. Figure 5
also indicates that the U-turn has changed the trend of the
roll angle. According to Fig. 6, the range of the heave
variations are about 2 cm. Moreover, the turns of the vessel
have caused the heave jumps, and the maximum one is in
relation to the U-turn.
Next, the corrections due to the attitude, heave, tide, vessel
load, and geoidal height variations were computed. Due to
small extent of the test field, no corrections from the co-tidal
and co-range charts were applied to the coastal tide gauge
observations. The geoidal height variations within the test
field were computed from EGM2008 global geopotential
model (Pavlis et al. 2008). Table 3 offers a statistical sum-
mary of all applied corrections. According to this table, the
attitude and geoidal height variations were the dominant
disturbing effects. As the table shows, the biggest attitude
effects are related to the bow-port antenna, and the smallest
ones to the stern-port antenna, as the result of their lever-arms
given in Table 1. From Table 3, we can also conclude that
the load variations are the least among the disturbing effects.
For hydrographic applications, the squat values at the
echo sounder location were derived from the computed
squats at the onboard GPS antennae by knowing the lever-
arm coordinates. To determine the maximum squat for
navigation applications, the squats at the stern-port and
stern-starboard of the vessel were derived and compared
with each other. Figure 7 shows the derived squat values as
a function of GPS time at the transducer, stern-port, and
stern-starboard locations. According to the figure, the
maximum squat happened at the stern-starboard of the
vessel; therefore, it is the one that must be considered for
the navigation applications. We also plotted the GPS-
derived ground speed of the vessel as a function of GPS
5.15 5.2 5.25 5.3 5.35 5.4x 10
4
−0.02
−0.01
0
0.01
0.02
GPS time (s)
Bas
elin
e di
ffere
nce
(m)
Mean = 0.0005 m, RMS = 0.0083 m
5.15 5.2 5.25 5.3 5.35 5.4x 10
4
−0.02
−0.01
0
0.01
0.02
Bas
elin
e di
ffere
nce
(m)
GPS time (s)
Mean = −0.0049 m, RMS = 0.0194 m
5.15 5.2 5.25 5.3 5.35 5.4x 10
4
−0.02
−0.01
0
0.01
0.02
GPS time (s)
Bas
elin
e di
ffere
nce
(m)
Mean = 0.0014 m, RMS = 0.0054 m(a)
(b)
(c)
Fig. 4 Comparison of the GPS and geodetic-derived distances
between the onboard antennae as a function of GPS time. a Baseline
2–3, b baseline 1–3, and c baseline 1–2
5.15 5.2 5.25 5.3 5.35 5.4x 104
−0.5
0
0.5
1
1.5
GPS time (s)
Atti
tude
ang
le (
deg)
Roll anglePitch angle
Fig. 5 The GPS-derived roll and pitch angles of the vessel as a
function of GPS time
GPS Solut
123
time in Fig. 8. The sudden speed changes during the turns
are evident in the figure. Comparison of the Figs. 5, 7, and
8 indicates the strong correlation of the squat with the pitch
and vessel speed, and its weak correlation with the roll.
The redundant squat values during the constant cruise
speed allowed LS estimation of the squats at different
speeds and their precisions by knowing the error budget of
the input data given in Table 4. Figure 9 shows the LS-
derived squat values as a function of cruise speed at the
transducer and stern-starboard locations. As can be seen in
the figure, the hydrographic squat values are considerably
less than the navigation ones, which supports the need of
having separate squat models for the two applications.
A summary of statistical information of the estimated
precisions of the LS-derived squat values at different speeds
for hydrographic and navigation applications is given in
Table 5. According to the table, the average precisions of
the LS-derived squat values for hydrographic and navigation
applications are 0.029 and 0.010 m, respectively.
According to Fig. 9, the LS-derived squat values are
contaminated with outliers. To suppress the effect of the
outliers in the squat models, the Tukey M-estimator with
bisquare weight function based on the iteratively
reweighted least squares (IRWLS) was used (Marrona et al.
2006) and a variety of fitting functions were tested. The
exponential functions were found the best based on the
RMS’s of fit among the tested ones. The parameters of
the exponential functions which present the analytical
squat models for hydrographic and navigation applications
are given in Table 6. Figures 10 and 11 show the analytical
squat models as the robust fit to the corresponding
LS-derived squat values. As can be seen from the figures,
the analytical squat models, while following the trend of
squat variations versus cruise speed, are free from the
effects of the outliers thanks to the robust M-estimation.
Having the estimated precisions of the parameters of the
analytical squat models and the GPS-derived cruise ground
speed, application of the error propagation law enabled us
5.15 5.2 5.25 5.3 5.35 5.4x 10
4
−0.06
−0.04
−0.02
0
0.02
0.04
0.06
0.08
GPS time (s)
Hea
ve (
m)
Fig. 6 The GPS-derived heave of the vessel as a function of GPS
time
Table 3 A statistical summary of the corrections for attitude dhAðtiÞ,heave dhHðtiÞ, tide dhT ðtiÞ, vessel load dhLðtiÞ, and geoidal height
dhNðtiÞ variations in the case study
Corrections Antennae Min.
(m)
Max.
(m)
Mean
(m)
STD
(m)
dhAðtiÞ Bow-port -0.013 0.373 0.116 0.085
Stern-port -0.125 0.107 -0.017 0.059
Stern-starboard -0.175 0.013 -0.050 0.043
dhHðtiÞ All -0.033 0.028 0.000 0.004
dhT ðtiÞ All 0.001 0.030 0.026 0.008
dhLðtiÞ All 0.001 0.010 0.005 0.003
dhNðtiÞ All -0.002 0.313 0.122 0.102
5.15 5.2 5.25 5.3 5.35 5.4x 10
4
0
0.05
0.1
0.15
0.2
0.25
0.3
GPS time (s)
Squ
at (
m)
Stern−portStern−starboardTransducer
Fig. 7 The squat values as a function of GPS time at the transducer,
stern-port, and stern-starboard locations
5.15 5.2 5.25 5.3 5.35 5.4x 10
4
0
2
4
6
8
10
12
GPS time (s)
Cru
ise
grou
nd s
peed
(kn
ots)
Fig. 8 The GPS-derived ground speed of the vessel as a function of
GPS time
GPS Solut
123
to compute the precisions of the derived models, which are
summarized in Table 7. According to the table, the mean
precisions of the hydrographic and navigation squat models
are 0.009 and 0.012 m, respectively. Note that the
improved precision of the hydrographic squat model over
the LS-derived squat values, compare Tables 5 and 7, is
due to application of the robust M-estimation and reduction
in the outliers effect.
In order to see the significance of the squat modeling
for hydrographic applications, we refer to the IHO stan-
dards (IHO 2008) for bathymetry accuracy. By consid-
ering the average depth of the test field, that is, 33.5 m,
the required bathymetry accuracies based on the men-
tioned standards are 0.35, 0.66, 1.26, and 1.26 m for
special, 1st a, 1st b, and 2nd orders projects, respectively.
If we assume that the speed of the vessel during hydro-
graphic bathymetry is 12.5 knots, then in reference to
Fig. 10, the associated squat value at the transducer
location will be 0.14 m. Considering the other sources of
the bathymetry errors such as sound speed, heave, and the
orientation, it will be quite unlikely that the required
accuracy for the mentioned orders, especially the special
and 1st a orders, be obtained if the squat effect is not
considered.
For navigation applications, consider the vessel of the
case study which experienced at a speed of 12.64 knots a
squat of 30 cm, which is 30 % of its 1-m draft. This shows
how the squat can affect the draft and UKC of a vessel
0 2 4 6 8 10 12
0
0.05
0.1
0.15
0.2
0.25
0.3
Cruise ground speed (knots)
Squ
at (
m)
Stern−starboardTransducer
Fig. 9 The LS-derived squat values as a function of cruise speed at
the transducer (for hydrographic applications) and the stern-starboard
(for navigation applications)
Table 5 A summary statistical information of the precisions of the
LS-derived squat values for hydrographic and navigation applications
Measures Precisions of LS-derived squats
Hydrographic Navigation
Min. (m) 0.005 0.003
Max. (m) 0.075 0.024
Mean (m) 0.029 0.010
Table 6 Parameters of the exponential functions SðvÞ ¼ a exp ðbvÞ þc as the analytical squat models versus the cruise speed v ðknots) for
hydrographic and navigation applications
Parameters Squat models
Hydrographic Navigation
a ðm) 0.012±0.001 0.025 ± 0.002
b ð1=knots) 0.202 ± 0.007 0.197 ± 0.005
c (m) -0.013 ± 0.002 -0.023 ± 0.003
RMS of fit (m) 0.006 0.008
Table 4 Error budget of the input data for the LS estimation of the
squats at different speeds
Input data Error budget
vðtiÞ 0.013 knots
hðt0Þ 0.001 m
hðtiÞ 0.007 m
dhNðtiÞ 0.011 m
Hðt0Þ, HðtiÞ 0.010 m
Tðt0Þ, TðtiÞ 0.010 m
Lðt0Þ, LðtnÞ 0.010 m
rðt0Þ, rðtiÞ 0.039�pðt0Þ, pðtiÞ 0.041�bx, by, bz 0.005 m
0 2 4 6 8 10 12
−0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
Cruise ground speed (knots)
Squ
at (
m)
Fig. 10 The analytical squat model (dark blue line) for hydrographic
applications derived as the result of the robust M-estimation, and the
LS-derived squat values (light blue dots) at the transducer location
GPS Solut
123
which are among major concerns for the safety of navi-
gation, especially at high speeds in confined channels.
Conclusions and remarks
A GNSS-based squat modeling method is formulated and
presented for hydrographic and navigation applications.
Removal of the disturbing effects from ellipsoidal heights
of the onboard GPS antennae with respect to a reference
state provides the basis of the method. Suitable GPS
antennae configurations onboard full-form ships and fine-
form vessels are proposed to model bow and stern squats,
respectively. Performance of the method was tested by the
field operations. According to the numerical results, the
navigation squats were about twice the hydrographic ones,
which demonstrate the importance of differentiating
between hydrographic and navigation applications for
squat modeling. Though the method is implemented on a
survey vessel in open waters, it can be applied to any type
of ship at any waterway. A navigation squat model can be
integrated with the nautical charts for real-time UKC
determinations to assist the safety of marine traffic at high
speeds. Moreover, a hydrographic squat model can provide
real-time squat corrections for accurate bathymetry.
Acknowledgments The authors would like to thank National
Geographic Organization (NGO) of Iran for their cooperation and
support during the field operations. The help from Eng. N. Abdi for
data processing using the Bernese GPS software is gratefully
acknowledged. Last but not least, the authors would like to express
their gratitude to two anonymous reviewers for their constructive
comments and suggested corrections.
References
Briggs M (2006) Ship squat predictions for ship/tow simulator.
Coastal and hydraulics engineering technical note CHETN-I-72,
U.S. Army Engineer Research and Development Center, Vicks-
burg, MS
Dach R, Hugentobler U, Fridez P, Meindl M (2007) Bernese GPS
software. Version 5.0, Astronomical Institute, University of
Bern, Bern
Delefortrie G (2010) Squat prediction in muddy navigation areas.
Ocean Eng 37(16):1464–1476
Derrett DR (1999) Ship stability for masters and mates. Revised by
Barrass CB, 5th edn, Butterworth-Heinemann, ISBN: 0-7506-
4101-0
El-Kader FA, EL-Soud MSA, EL-Serafy K, Hassan EA (2003) An
integrated navigation system for Suez Canal (SCINS). J Navig
56(2):241–255
Feng Y, O’Mahony S (1999) Measuring ship squat, trim, and under-
keel clearance using on-the-fly kinematic GPS vertical solutions.
Navigation 46(2):109–117
Feng Y, Kubic K, O’Mahony S (1996) On-the-fly GPS kinematic
positioning for measuring squat and trim of large ships. In:
Proceedings of ION GPS, The Institute of Navigation, Kansas
City, MO, 9:367–373
Giorgi G, Gourlay TP, Teunissen PJG, Huisman L, Klaka K (2010)
Carrier phase ambiguity resolution for ship attitude determina-
tion and dynamic draught. In: XXIV FIG International Congress,
Sydney, Australia
Godhavn, J-M (2000) High quality heave measurements based on
GPS RTK and accelerometer technology. Ocean Conf Record
(IEEE) 1:309–314
Gourlay TP, Cray WG (2009) Ship under-keel clearance monitoring
using RTK GPS. In: Proceedings of Coasts and Ports,
Wellington
Harting A, Laupichler A, Reinking J (2009) Considerations on the
squat of unevenly trimmed ships. Ocean Eng 36(2):193–201
IHO (2005) Manual on hydrography, 1st edn. Publication M-13,
International Hydrographic Bureau, Monaco
IHO (2008) IHO standards for hydrographic surveys. 5th edn,
Special Publication No. 44, International Hydrographic Bureau,
Monaco
Marrona RA, Martin DR, Yohai VJ (2006) Robust statistics. Theory
and methods. Wiley, New York, ISBN: 0-470-01092-4
Maynord ST, Briggs MJ (2006) Ship squat measurements using GPS
at Charleston harbor. In: Proceedings of the international
conference on civil engineering in the oceans, Oceans VI:
449–463
Melachroinos SA, Tchalla M, Biancale R, Menard Y (2011)
Removing attitude-related variations in the line-of-sight for
kinematic GPS positioning. GPS Solut 15(3):275–285
Pavlis NK, Holmes SA, Kenyon SC, Factor JK (2008) An earth
gravitational model to degree 2160: EGM2008. In: The General
Assembly of the European Geosciences Union, Vienna, Austria
Varyani KS (2006) Squat effects on high speed craft in restricted
waterways. Ocean Eng 33(3–4):365–381
0 2 4 6 8 10 12
0
0.05
0.1
0.15
0.2
0.25
0.3
Cruise ground speed (knots)
Squ
at (
m)
Fig. 11 The analytical squat model (red line) for the navigation
applications derived as the result of the robust M-estimation, and the
LS-derived squat values (pink dots) at the stern-starboard location
Table 7 Precision of the analytical squat models for hydrographic
and navigation applications
Measures Model precisions
Hydrographic Navigation
Min. (m) 0.003 0.004
Max. (m) 0.022 0.028
Mean (m) 0.009 0.012
GPS Solut
123
Zilkoski DB, D’Onofrio JD, Fury RJ, Smith CL, Huff LC, Gallagher
BJ (1999) Centimeter-level positioning of a U.S. coast guard
buoy tender. GPS Solut 3(2):53–65
Author Biographies
Alireza A. Ardalan received a
M.Sc. degree in geodesy from
K.N. Toosi University Tehran,
Iran, and M.Sc. and Ph.D.
degrees from Stuttgart Univer-
sity Stuttgart, Germany. He is a
full professor of geodesy at the
Department of Surveying and
Geomatics Engineering, College
of Engineering, University of
Tehran. His research interests
include satellite positioning,
engineering geodesy, hydrogra-
phy, and gravity field modeling.
Mohammad-Hadi Rezvanireceived a M.Sc. degree in
hydrography from University of
Tehran, Iran. He is a research
fellow at the Department of Sur-
veying and Geomatics Engineer-
ing, College of Engineering,
University of Tehran. His
research interests include hydrog-
raphy, multibeam sonar calibra-
tion, and GNSS applications.
GPS Solut
123