DECEMBER 2015 BRINGAS AND GONI 2253

Early Dynamics of Deep Blue XBT Probes

FRANCIS BRINGAS AND GUSTAVO GONI

National Oceanic and Atmospheric Administration/Atlantic Oceanographic and
Meteorological Laboratory, Miami, Florida

(Manuscript received 12 March 2015, in final form 23 September 2015)

ABSTRACT

Expendable bathythermographs (XBTs) are probes widely used to monitor global ocean heat content,
variability of ocean currents, and meridional heat transports. In the XBT temperature profile, the depth is
estimated from the time of descent in the water using a fall-rate equation. There are two main errors in these
profiles: temperature and depth errors. The reduction of error in the estimates of the depth allows a corre-
sponding reduction in the errors in the computations in which XBTs are used. Two experiments were carried
out to study the effect of the deployment height on the depth estimates of Deep Blue XBT probes. During
these experiments, XBTs were deployed from different heights. The motion of the probes after entering the
water was analyzed to determine the position and the velocity of the probes as a function of time, which was
compared to that obtained using the Hanawa et al. fall-rate equation. Results showed a difference or offset
between the experimentally observed depths and those derived from Hanawa et al. This offset was found to be
linked to the deployment height. To eliminate the offset in the fall-rate equation for XBTs deployed from
different heights, a methodology is proposed here based on the initial velocities of the probes in the water (or
deployment height). Results indicate that the depth estimates in the profiles need to be corrected for an offset,
which in addition to having a launch height dependence is time dependent during the first 1.5 s of descent of
the probe in the water, and constant after that.

1. Introduction z(t) 5 at 2 bt2 , (1)

Expendable bathythermographs (XBTs) are widely
known as the fall-rate equation (FRE), where the co-
used to measure temperature profiles since their intro-
efficients a and b are determined empirically for different
duction several decades ago. Currently, approximately
types of XBT probes. These two coefficients are defined
20 000 XBTs are globally deployed along fixed transects
using a dynamical model describing the movement of the
every year from ships of opportunity and research ves-
XBT in the water as the result of a force balance between
sels. Typically, the deployments are made from different
the hydrodynamic drag, the decreasing buoyancy caused
heights up to approximately 25 m, at ship speeds be-
by the wire loss, and the increase of seawater viscosity
tween 0 and 25 kt (1 kt 5 0.514 m s21). The temperature
with depth (Green 1984; Hallock and Teague 1992,
is determined with a thermistor whose signal is trans-
hereafter HT92). The coefficient a represents the terminal
mitted to the recorder on board the ship through a wire
velocity of the XBT in the seawater, while the coefficient
that spools out of the probe, constituting the main lim-
b represents a deceleration introduced as a correction to
itation to the maximum depth that can be reached. Since
the first term in (1) as a result of the loss of weight of the
the probes do not have pressure sensors, the depth is
probe as the wire is spooled out. Since b
a, according to
determined by a semiempirical relation z(t) of the form
the FRE the XBTs fall at an almost constant velocity
(Green 1984),
throughout their descent in the water. When the vertical
axis is defined as positive downward, both coefficients are
positive. The original values of coefficients a and b used
Corresponding author address: Francis Bringas, National Oce- for the Deep Blue type of XBTs were 6.472 m s21 and
anic and Atmospheric Administration/Atlantic Oceanographic
and Meteorological Laboratory, 4301 Rickenbacker Causeway, 0.002 16 m s22, respectively (NOAA 2014, code 51).
Miami, FL 33149. There are several sources of error that may affect the
E-mail: [email protected] accuracy in temperature and depth of the XBT profile

DOI: 10.1175/JTECH-D-15-0048.1
2254 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 32

(Cowley et al. 2013). The work presented here focuses efforts and methodologies address XBT biases originated
on depth offsets using the XBT FRE estimates. The by the fall-rate coefficient discrepancies, few experiments
XBT fall rate, determined by (1), has been the subject of (e.g., HT92; Xiao et al. 2006) and numerical simulations
numerous studies since the 1980s, which included re- (e.g., Abraham et al. 2012a; Abraham et al. 2012b; Xiao
ports on systematic errors in the depth estimates (Seaver and Zhang 2012) have been conducted to assess the dy-
and Kuleshov 1982; Roemmich and Cornuelle 1987). namic characteristics of the XBT during the first few me-
The impact that these depth errors may have in climate ters of movement in the water. During this short distance
research was later recognized from a comprehensive after the XBT hits the water, the probe may not yet be
analysis carried out using CTD and XBT data in several moving at its terminal velocity, as prescribed by the most
regions (Hanawa et al. 1995, hereafter H95), where new commonly used FREs. Thus, it could be expected that the
coefficients a and b were estimated as first few meters of movement may significantly contribute
to a potential depth offset in the temperature profile as
a 5 6:691 m s21 ; b 5 0:002 25 m s22 . (2) estimated by most FREs.
Because of the extensive use of XBTs in ocean current
New advances in the research of XBT FRE biases in- monitoring (e.g., Carton and Katz 1990; Goni et al. 1996;
dicated that these biases are time dependent (Gouretski Gilson et al. 1998; Arnault et al. 1999, 2011; Phillips and
and Koltermann 2007). This inspired a large number of Rintoul 2002; Rintoul et al. 2002; Ridgway et al. 2008;
studies to determine the cause of these biases, to assess Gourcuff et al. 2011; Goes et al. 2013; Domingues et al.
their values, and to propose correction schemes for XBT 2014; Dong et al. 2009, 2014; Goes et al. 2015), and to
data. These corrections have been obtained using a suite determine the ocean thermal structure and heat trans-
of different methods, including XBT–CTD comparisons port (Smith et al. 1999; Roemmich et al. 2001; Baringer
(Seaver and Kuleshov 1982; Heinmiller et al. 1983; and Garzoli 2007; Abraham et al. 2013; Dong et al.
Hanawa and Yoritaka 1987; Singer 1990; HT92; 2009), as well as their remarkable contribution to the
Thadathil et al. 2002; Reseghetti et al. 2007; Wijffels knowledge we have about the ocean today, it is critical
et al. 2008; Ishii and Kimoto 2009; Levitus et al. 2009; to properly assess the fall-rate equation coefficients to
Gouretski and Reseghetti 2010; DiNezio and Goni 2011; improve historical XBT datasets and to establish stan-
Kizu et al. 2011; Hamon et al. 2012; Cowley et al. 2013; dards for current and future observations.
Cheng et al. 2015) and XBT and altimetry, (DiNezio and This work reports two experiments carried out with the
Goni 2010) among others. In particular, some of these goal to study the main characteristics of the Sippican Deep
studies generally reported the presence of depth offsets Blue XBT rate of descent during the first 11 m after en-
without a clear indication of the physical mechanisms tering the water. The objectives of these experiments were
that caused these offsets. For example, using a statistical 1) to determine the time it takes for the XBTs, which are
analysis of a large number of side-by-side XBT–CTD entering the water at different initial velocities, to accel-
pairs, Cowley et al. (2013) found that in their linear model erate or decelerate to their terminal velocity; 2) to assess
for depth bias, the offset term appears to compensate for whether different deployment heights may create a depth
an acceleration over the upper 100 m. Reseghetti et al. offset; and 3) to evaluate whether the currently used FRE
(2007) also analyzed collocated XBT–CTD pairs, with coefficients also apply to the first 11 m of the descent.
XBTs deployed from different initial heights. While the This work represents the first attempt to experimen-
possible effect of the deployment height and the verti- tally verify the theoretically derived descent rate of XBT
cality of the probe while entering the water were discussed probes in the first meters in the water and to assess the
in their work as possible factors to cause a depth offset in physical mechanisms that create a depth offset in XBT
the XBT profile, their results provided uncertain indica- profiles. This manuscript is organized as follows. The
tions of the influence of the deployment height on the experimental setups and the methodology used to ana-
motion of the probes during the first few meters in the lyze the experimental data are described in section 2.
water. Other factors, such as the response time of the ac- Section 3 contains the results and the discussion, in-
quisition system, were found to have a greater impact on cluding the proposed correction of the FRE estimates
the XBT profiles in the near-surface layer. DiNezio and for deployments made from different heights. The con-
Goni (2010) estimated systematic biases between XBT and cluding remarks are presented in section 4.
Argo observations using satellite altimetry. In this case
depth offsets in the upper meters were found to be differ-
2. Experimental setups
ent among ocean basins, which made the authors unable to
identify systematic errors in the FRE resulting from tran- Two experiments were conducted to study the dy-
sients in the initial descent of the probes. Although many namics of the Sippican Deep Blue XBT probes during
DECEMBER 2015 BRINGAS AND GONI 2255

the first 11 m of descent after the probes enter the water. trajectory z(t) from the recording of the four cameras
The experiments were designed to assess the vertical used. The experimental values of the vertical trajecto-
displacements of the probes as a function of time, z(t), ries were then compared to the H95 FRE and also used
for different deployment heights, and to determine the to validate the HT92 model.
time that it takes the probes to reach their terminal
vertical velocity in the water as a function of the
deployment height. 3. Results
The first experiment was carried out in a swimming
a. Deployment height and the first meters of descent
pool 5 m deep at the University of Miami. A digital
camera with a frame rate of 1/30 s was installed at 1 m The experiments reported here were designed to in-
above the bottom of the pool and a rope with marks vestigate the effect of the deployment height on the fall
every 1 m was placed in the pool and used in the cali- rate during the first meters of the XBT descent in the
bration of the recorded videos to determine the depth of water. During these experiments, XBT probes were
the XBTs. During this experiment, XBTs were deployed deployed in a 5-m-deep pool and in an 11.12-m-deep
with a straight orientation (vertically aligned facing water tank, as described in section 2. The XBTs were
down) from heights of 3 and 5 m above the surface of the released from different heights and with a vertically
water. These experimental vertical displacements were straight orientation, with the objective of determining
compared against those derived from the FRE esti- the deviation of the actual fall in the water from the H95
mates, using the coefficients proposed by H95 in (1) and FRE estimates. In addition, these observations were
(2), which provide the depth of the XBT in meters when used to assess the maximum depth at which these probes
the time is expressed in seconds. reach terminal velocity.
The second experiment was conducted in a water tank The difference between the actual fall and the H95
11.12 m tall at the NOAA National Data Buoy Center FRE estimates are shown in Fig. 1. Results indicate that
laboratory. To obtain a recording of the complete XBT XBTs deployed at the surface of the water (h0 5 0),
trajectory, four digital cameras, each with a frame rate of which also correspond to an initial velocity w0 5 0,
1/30 s, were installed at 1.52-, 4.27-, 7.0-, and 9.75-m moved slower than the FRE estimates after entering the
depth. These depths were selected because of the logistics water. On the other hand, when the height of de-
to work inside the water tank and to ensure overlaps in the ployment was larger than h0 5 9 m, the probes moved
images between the videos produced by adjacent cameras. faster than the FRE estimates in the first upper meters.
A vertical rope with markings every 1 m was placed in the Small differences between the experimental results and
water tank and used in the calibration of the recorded the FRE estimates during the first 11 m were found for
videos to determine the depth of the XBTs. During this probes deployed from heights 2 m # h0 # 4 m. For de-
experiment, XBTs were deployed in a straight orientation ployments up to 20-m height, results show that the
from heights of 0, 2, 4, 9, 14, 20, and 25 m. probes reach terminal velocity within the first 11 m. For
In total, 135 deployments were carried out from nine deployment heights of 25 m, the probes did not reach
different heights. The recordings obtained from both terminal velocity in the first 11 m; however, their speed
experiments were analyzed frame by frame using the of descent indicated that they were approaching termi-
software Tracker (Brown 2012) to obtain the values of nal velocity. Results show that while the vertical dis-
the vertical position as a function of time. All deployments placement of XBTs deployed from heights 2 m # h0 #
were repeated 15 times, and a third-degree polynomial 4 m approximately coincides with the FRE estimates,
was used to estimate the mean z(t) for each deployment displacement of probes deployed outside this range do
height. These fitted polynomials were then used to com- not agree with those obtained from the H95 FRE.
pute the vertical velocity in the water as a function of time Therefore, this result indicates that the depth in profiles
for each deployment height. obtained from deployments made from heights h0 # 2 m
The images analyzed in this study correspond to XBTs and h0 $ 4 m needs to be corrected during the first 2 s
that fall in an orthogonal plane in front of the camera. In after the probe enters the water and until it reaches
the analysis of the images, the first frame where the XBT terminal velocity. If this depth error is not corrected, it
was visible inside the water was used as the initial po- will be carried over through the entire depth of the XBT
sition (t 5 0) of its vertical location. Thus, a linear in- temperature profile.
terpolation of the data from different deployments was The H95 FRE implies an almost constant velocity of
performed in order to obtain results that correspond to the probes during the entire fall in the water with only a
the same reference. In the second experiment, the very small deceleration as a result of the wire payout,
videos had to be further calibrated to obtain a single which is given by the coefficient b in (1). Therefore, the
2256 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 32

FIG. 1. Vertical position as a function of time for Deep Blue XBT probes deployed from different
heights (blue circles), FRE estimates (black line), and third-order polynomial best fit of the experimental
data (red line).

deployment of probes from different heights, which probes in the water, which is a function of the de-
implies different initial velocities of the XBTs in the ployment height. We present here experimental results
water, produced vertical displacements that were dif- on the reduction of the velocity of the probes deployed
ferent from those estimated by the H95 FRE at the 2 m and above the surface after they hit the water. The
beginning of the fall. Results show that after entering the velocity of the probes at the instant of water impact, or
water, the XBTs accelerated (decelerated) from initial velocity of free fall wff is
velocities, which were smaller (larger) than a 5 6.691 m s21,
qffiffiffiffiffiffiffiffiffiffi
until their terminal velocity reached a value wT that was
wff 5 2gh0 , (3)
equal or very close to a. From that instant, the probes
continued to move at a rate close to the one described by
(1) using the coefficients in (2). The results presented where g 5 9:81 m s22 is the gravity acceleration and h0 is
here (Fig. 1) show this behavior for all deployment the height of deployment. The velocity at the instant of
heights under 25 m. For deployment heights of 25 m, water impact is compared here with the initial velocity
the depth of the water tanks used in the experiment of the probe in the water immediately after impact,
did not allow for the XBT probes to fully reach ter- which is estimated using the first two or three frames of
minal velocities. the videos recorded during both experiments. Results
are presented in Table 1 and in Fig. 2. Figure 2 shows
b. The initial velocity of descent
that after the XBTs hit the water, their velocities
The initial velocity of descent of the XBT probes in quickly decrease as a result of the impact, unless they
the water depends on the velocity of impact of the are deployed from h0 5 0. This decrease in velocity
DECEMBER 2015 BRINGAS AND GONI 2257

TABLE 1. Theoretical free fall velocity wff estimated as a function

of the deployment height h0 and observed initial velocity w0 of the
XBTs in the water. The difference Dw 5 wff 2 w0 indicates the loss
of momentum of the probes upon impact in the water.

h0 (m) wff (m s–1) w0 (m s–1) Dw (m s–1)

2 6.26 5.73 6 0.42 0.53 6 0.42
3 7.67 6.40 6 0.55 1.27 6 0.55
4 8.86 6.86 6 0.18 2.00 6 0.18
5 9.90 8.23 6 0.72 1.67 6 0.72
9 13.29 9.78 6 1.40 3.51 6 1.40
14 16.57 12.47 6 1.25 4.10 6 1.25
20 19.81 16.04 6 1.83 3.77 6 1.83
25 22.14 17.05 6 0.54 5.09 6 0.54

(Dw) is proportional to the deployment height h0 . Al-

though the results shown in Fig. 2 do not suggest a
linear relationship between Dw and h0 , we expect a
continuous increase of Dw with the height of de-
ployment in most applications, where XBTs are de- FIG. 2. Difference between the velocity of the probes before and
after impact in the water as a function of the deployment height.
ployed from h0 , 35 m. The observed initial velocities
of the XBTs in the water (Fig. 3) were determined for
deployment heights between 0 and 25 m (an initial ve- this effect is that the XBT depth estimates are shallower
locity w0 5 0 is assumed for h0 5 0) (Table 1). Cur- than the actual depth of the XBTs.
rently, the most commonly used XBT probes, the Deep A correction to the FRE that can account for the
Blues, are manufactured to be deployed from heights deviations from a constant velocity of descent (Fig. 1)
of 3–4 m above the sea surface and have been reported should therefore include the initial velocity in the water.
to produce initial velocities in the water of 6–7 m s21 Since this value cannot be obtained during normal XBT
(NOAA 2014, codes 51 and 52), which is confirmed by deployments, Fig. 3 and (4) provide a method to esti-
our experiments. Therefore, probes deployed from the mate this initial velocity as a function of the deployment
height interval of 3–4 m (gray area in Fig. 3) will not height, a parameter that is measurable.
introduce a depth bias in the FRE of the form (1) and
will not require an offset correction because of de- c. FRE correction as a function of deployment height
ployment height. The fall-rate equation of an XBT can be obtained
To obtain the initial velocity of the XBTs as a function theoretically assuming a balance between the vertical
of the height of deployment, an exponential function of acceleration, the total buoyancy force, and the hydro-
the form dynamic drag, which is considered proportional to the
square of the velocity of descent in the water (Green
w0 (h0 ) 5 phq0 (4)
1984):
was fitted to the experimental data (Fig. 3), obtaining d 2 z g(m 2 mw ) rw CD Aw2
the following values for the parameters: p 5 3:7066 and 5 2 , (5)
dt2 m 2m
q 5 0:4745.
Results obtained here indicate that for a deployment where z is the vertical position (positive downward), t is
height of 1 m, XBTs will start their descent in the water time, g is the acceleration of gravity, m is the mass of the
at 3.7 m s21, which is lower than the XBT terminal ve- XBT probe, mw is the mass of displaced water, rw is the
locity of 6.691 m s21. These probes will accelerate for water density, CD is the drag coefficient, A is the effec-
several meters until they reach their terminal velocity. tive cross section (area) of the probe, and w is the ver-
The consequence of not considering this effect is that the tical velocity (positive downward). Equation (5) can be
XBT depth estimates are deeper than the actual depths integrated by a parameterization of varying magnitudes,
of the XBTs. On the other hand, for a deployment m 2 mw and CD , and by considering a negligible accel-
height of 15 m, XBTs will start their descent at eration, d2 z/dt2 ’ 0, to obtain (Green 1984)
13.4 m s21, which is larger than the terminal velocity, and
will decelerate for several meters until they reach their hw2T t2
z(t) 5 wT t 2 , (6)
terminal velocity. The consequence of not considering 4
2258 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 32

d2 z dw
5 5 b 2 aw2 , (8)
dt2 dt

where
rwC
DA
a5 , (9)
2m
gm0
b5 , (10)
m

and m0 is the weight of the XBT in the water. Since

d2 z/dt2 5 0 when the XBT velocity of descent (w) rea-
ches its terminal velocity (wT ), from (8) it is obtained,
rffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b 2gm0
wT 5 5 . (11)
a r w CD A

Equation (8) can be written in terms of the vertical ve-

FIG. 3. Experimental values of the initial velocity of the XBT in
the water (w0 ) as a function of the deployment height h0 . The black locity and integrated twice to obtain w 5 w(t) and
curve is a nonlinear fit of the function w0 (h0 ) 5 phq0 , with p 5 3:7066 z 5 z(t), using the initial conditions w(0) 5 w0 and
and q 5 0:4745. XBTs deployed from a height within the gray area z(0) 5 0. The result for z(t) is obtained as
will have initial velocity in the water close to the terminal velocity.
8   9
>
> sinh(awT t 1 C1 ) w0 >
>
>
> ln ; . 1 >
>
>
< sinhC1 wT >
=
which is the fall-rate equation in the form of (1), with 1
z1 (t, w0 ) 5 , (12)
a>>   >
>
>
> cosh(awT t 1 C2 ) w0 >
>
wT 5 a, >
: ln ; ,1> ;
coshC2 wT
hw2T
5 b, (7)
4 where C1 5 coth21 (w0 /wT ) and C2 5 tanh21 (w0 /wT ).
Using the experimental results obtained in this work,
and h is a parameter used to account for the linear vari- we can validate the theoretical model above for the
ations assumed for CD and the mass decrease as a result vertical position of the XBT during the first 11 m of
of the wire payout as a function of depth. The nature of descend as given by (12). The differences between the
the parameters mw , CD , and rw in (5) may vary for dif- FRE using the coefficients of H95 and the observed
ferent geographical regions but are assumed to be con- vertical displacements for different deployment heights
stant in this work. Additionally, m, CD , mw , and rw vary are presented in Fig. 4, together with the difference
with depth. The physical meaning of the solution (6) between the fall-rate equation of H95 and the solution
obtained for (5) indicates that the probe is expected to obtained using HT92 [(12)], with wT 5 6:691 m s21;
have a motion in which the acceleration, the left-hand g 5 9:806 65 m s22; a 5 0:166 23 m21, determined from
side of (5), is very small and that the probe falls at an
almost constant wT throughout the entire water column, gm0
including at the beginning of the descent. However, the re- a5 , (13)
mw2T
sults presented in the previous section show that the descent
of an XBT exhibits different initial velocities after entering where m0 /m 5 0:758 88; and the initial velocity of descent
the water, which depend on the deployment height, until it w0 from Table 1.
reaches wT . The differences between the H95 FRE and the ob-
A model proposed by HT92 for the fall of the XBT in served vertical positions, and the H95 FRE and the
the first 10 m in the water is intended to account for the HT92 model, are shown in Fig. 4, using the experimental
difference between the actual velocity of the XBT at the values of w0 from Table 1. Results show that for
beginning of its descent and wT . For z # 10 m (t # 1:5 s), h0 5 4 m, these differences are very small (on average
the contribution of the quadratic term in (5) is very small, 0.04 m 6 0.02 m). These differences are larger for de-
approximately 0.5 cm, and therefore (5) can be simplified by ployment heights below and above 4 m, consistent with
considering that the parameters in this equation are con- the results already presented in Fig. 1. An important
stant while including the vertical acceleration, to obtain result obtained here is that since the differences between
DECEMBER 2015 BRINGAS AND GONI 2259

z(t) 5 at 2 bt2 1 D, (16)

where a, b are defined in (1). Given a FRE in the form

of (1), the new depth can be estimated from (14)
provided that the initial velocity of the probe in the
water is known. In most practical situations and for
Deep Blue XBTs, the initial velocity can be obtained
from (4) if the height of deployment is known. This is a
parameter that can be estimated during an XBT de-
ployment. This result indicates that by using H95 [(1),
(2)], the depth offset will always exist unless the XBT
is deployed from heights between 3 and 4 m (gray area
in Fig. 3).
The methodology proposed in this work to correct
FIG. 4. Difference between the observed depth zExp and the H95
the depth offset caused by different deployment
FRE zH95 (solid symbols) and the depth from the HT92 model zHT92 heights in temperature profiles obtained from Deep
and zH95 (open symbols) for different heights of deployment h0 . Blue XBT probes with an FRE of the type (1) is as
follows:

the observed positions and H95 are very similar to 1) The initial velocity of the XBTs in the water (w0 ) is
those between HT92 and H95, then HT92 provides a estimated using (4).
very good description of the observed descent of the 2) The initial (w0 ) and terminal [wT ; i.e., coefficient a
probes during the first meters. In particular, while dif- in the FRE of the type (1)] velocities of the XBTs in
ferences of as much as 15% can be observed between the water are used to compute z1 (t, w0 ) using (12).
the experimental observations and HT92, both com- 3) Here z1 (t, w0 ) and the coefficient b of the FRE of the
pare very well for the extreme deployment heights of type (1) are used to compute the new depth as a
0 and 20225 m. function of time using
The model in HT92 provides a solution [(12)] for the
z(t) 5 z1 (t, w0 ) 2 bt2 . (17)
depth of the XBT during the first 10–15 m (t # 1:5 s) of
descent as a function of the initial velocity of the probe
As an example, Fig. 5 shows how the proposed meth-
in the water.
odology estimates the depth of an XBT when different
Since (12) was obtained by disregarding the con-
heights of deployments are considered. The tempera-
tribution of the quadratic term at the beginning of the
ture profile in this example (Fig. 5a) was obtained at
descent, a general FRE can be proposed by adding the
location 32.718N, 46.038W on 19 March 2014 from an
quadratic term in (6) to the solution [(12)] to obtain
XBT deployed from a height of approximately 10 m.
(HT92)
This profile is characterized by a mixed layer depth of
approximately 218C to 150 m, a pronounced thermocline
hw2T t2
z(t) 5 z1 (t, w0 ) 2 , (14) between 150 and 200 m, and a smaller monotonic neg-
4
|fflfflfflfflffl{zfflfflfflfflffl} ative temperature gradient between 200 and 800 m.
ffi0for t#1:5s
|fflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl} After the methodology proposed here is applied, a
t.0
depth offset is removed from the original profile. De-
where z1 (t, w0 ) is given by (12). Equation (14) is the pending on the deployment height, different probe
general FRE that should be used to eliminate the depth depths are estimated for one given time. Differences
offset resulting from varying deployment heights. between H95 and the observed depths are closer to zero
It can be shown that for sufficiently large times for h0 5 3 m (Fig. 5b). These differences are negative
(t . 1:5 s), (positive) for deployment heights below (above) 3 m
(Fig. 5b). In this example, the depth offset for a de-
z1 (t) 5 wT t 1 D , (15) ployment height of 10 is 2 m. Also, differences between
the profiles before and after the offset correction
where D is the maximum offset in the temperature (Fig. 5c) for this specific deployment made from a de-
profile resulting from the deployment height. Therefore, ployment height of 10 m show that the temperature
for sufficiently large times (t . 1:5 s) (14) approaches the differences are 1) between 08 and 20.18C within the
form mixed layer; 2) up to 20.68C in the thermocline, at
2260 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 32

approximately 150 m deep; and 3) between 08 and 0.158C

below the thermocline. Depth offsets produced in pro-
file data, such as the one described above, have large
errors in depth estimates at the depth of large temper-
ature gradients. For the case above, the offset produces
an excess of heat above the 188C isotherm of 2.5 kJ cm22.
This indicates that the depth offset may have a notice-
able impact on upper-ocean heat content computations.
Following the results presented above, temperature
profiles obtained using H95 need to be adjusted by con-
verting time into depth, where the depth is determined
using (17), which is a function of w0 , wT , and time; w0 is
determined by (4), which is a function of the deployment
height; and wT is given by the coefficient a in (2).
d. Maximum offset in the FRE estimates
The methodology described in this work estimates the
offset to be removed from the profile computed using
any FRE of the form (1). For times t , 1:5 s, the depth
offset is a function of time and is given by the difference
between the FRE used and (12). For t . 1:5 s the XBT is
already moving at terminal velocity and the offset is
constant. During the transition period until the XBT
reaches wT , this offset is introduced as the FRE (1)
considers that the XBT is moving at terminal velocity
since the beginning of the descent. The results in Fig. 4
can be used to estimate the time at which wT is reached.
This time is when the maximum value of Dz is reached
between the experimental z(t) and the estimates of the
H95 FRE. At this moment, D has been reached and the
XBT is moving with terminal velocity.
The values of D introduced in the FRE as observed in
our experiments are presented in Fig. 6 and Table 2. This
offset depends on the deployment height, being ap-
proximately equal to zero for deployments done from
heights between 3 and 4 m and negative (positive) for
deployments from heights below (above) these values.
Figure 6 shows that during common XBT deployments
from cargo ships, where the height of deployment can be
between 10 and 35 m, a depth offset of more than 3.7 m
may be introduced in the profiles computed using H95
FRE. Table 2 and Fig. 6 are only provided here in order
to illustrate the maximum magnitude of D resulting
from different deployment heights. Since this depth
offset is a function of time, the actual profile depth with
the offset removed should be computed using the
methodology described above [based on Eq. (17)].

FIG. 5. Example of depth offset correction of an XBT profile. 4. Summary and conclusions
(a) Original XBT profile. (b) Depth difference as a function of time
between the original (H95) and the corrected (using HT92) profiles Two experiments were conducted in order to assess
during the first 1.5 s for several deployment heights. (c) Temperature the effect of the deployment height in the fall rate of
difference as a function of depth between the original (H95) and the Deep Blue XBT probes in the water. Experimental
corrected (using HT92) profile for a deployment height of 10 m.
DECEMBER 2015 BRINGAS AND GONI 2261

TABLE 2. Maximum offset D in the H95 FRE observed during this

work for different deployment heights h0 .

h0 (m) D (m)
0 24.05
2 20.44
3 0.03
4 0.40
5 0.72
9 1.65
14 2.47
20 3.22
25 3.73

cargo ships in low-density mode or frequently repeated

mode, deployments are mostly made from typical
heights between 15 and 35 m.
Results presented here are independent of the FRE,
FIG. 6. Maximum offset D observed in the depth of XBT probes XBT recording system, and sea state, among other pa-
when using the H95 FRE without considering a correction for the rameters, because they are based on direct observations
deployment height.
of XBT positions as a function of time. Therefore, the
results confirmed theoretical studies in which there is a depth offset described and assessed here is purely due to
depth offset in the XBT FRE that depends on the height the effect of initial velocities on XBT depth estimates
of deployment. This offset ranges from approximately caused by different launch heights. However, other
24 to 4 m depending on the deployment height. This factors, including the potential effect of varying ship
offset is time dependent during the first 1.5 s of descent speeds, water entry conditions, ocean conditions, ac-
in the water and constant after that. To eliminate the quisition and recording system, water density, temper-
offset during the first 1.5 s of descent, a theoretical model ature, etc., may also contribute to the depth offset and
(HT92) was shown to provide an adequate estimate of will need to be assessed. Our results highlight the need
the actual depth of the XBT. of continuous improvements in observational datasets.
To correct the depth of a Deep Blue XBT profile This work was specifically carried out for Sippican Deep
obtained using a FRE of the type z(t) 5 at 2 bt2 —for Blue XBT probes, and similar experiments, studies, and
example, the H95 equation, (17), can be used with analysis should be performed for other types of probes.
z1 (t, w0 ) given by (12)—a 5 0:166 23 m21, wT 5 a, and Studies such as this one contribute to reducing the error
w0 (h0 ) 5 phq0 , with parameters p 5 3:7066 and q 5 in global ocean heat content estimates by improving the
0:4745 that were obtained empirically on this work quality of historical datasets, especially during the pe-
(sections 3b and 3c). This correction applies to any riod 196922001, when XBTs accounted for most of the
FRE of the type (1) and assumes that the coefficients a temperature profile data.
and b are representative of the correct form of the
XBT FRE after the probe has reached its terminal Acknowledgments. The authors would like to ac-
velocity. knowledge Grant Rawson, Pedro Pena, and Kyle Seaton
We recommend that the correction proposed here be for their work during the execution of the experiments
used to improve the historical XBT database and be presented here, as well as Zach Barton for the help
applied to current observations, provided that the de- provided tracking the movement of the XBTs in the
ployment height is known. Although the deployment videos. We would also like to acknowledge the support
height is not always known, it could be inferred from the of NOAA’s National Marine Fisheries Service and
metadata containing information on the ships from Southeast Fisheries Science Center during these exper-
which the deployments were performed. For research iments. This work was supported by NOAA’s Climate
ships the typical deployment height ranges from 2 to 5 m, Program Office and NOAA AOML.
while for cargo ships it ranges from 10 to 35 m. In ad-
dition, for XBTs deployed using automatic launchers, REFERENCES
which include most XBT deployments in high-density Abraham, J. P., J. M. Gorman, F. Reseghetti, E. M. Sparrow, and
mode since 1997, typical deployment heights are be- W. J. Minkowycz, 2012a: Drag coefficients for rotating
tween 10 and 15 m. For deployments carried out from expendable bathythermographs and the impact of launch
2262 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 32

parameters on depth predictions. Numer. Heat Transfer, 62A, ——, M. Baringer, and G. Goni, 2015: The impact of historical
25–43, doi:10.1080/10407782.2012.672898. biases on the XBT-derived meridional overturning circulation
——, ——, ——, ——, and ——, 2012b: Turbulent and transitional estimates at 348S. Geophys. Res. Lett., 42, 1848–1855,
modeling of drag on oceanographic measurement devices. doi:10.1002/2014GL061802.
Modell. Simul. Eng., 2012, 567864, doi:10.1155/2012/567864. Goni, G., S. Kamholtz, S. Garzoli, and D. Olson, 1996: Dynamics of
——, and Coauthors, 2013: A review of global ocean temperature the Brazil-Malvinas Confluence based on inverted echo
observations: Implications for ocean heat content estimates sounders and altimetry. J. Geophys. Res., 101, 16 273–16 289,
and climate change. Rev. Geophys., 51, 450–483, doi:10.1002/ doi:10.1029/96JC01146.
rog.20022. Gourcuff, C., P. Lherminier, H. Mercier, and P.-Y. LeTraon, 2011:
Arnault, S., B. Bourlès, Y. Gouriou, and R. Chuchla, 1999: In- Altimetry combined with hydrography for ocean transport
tercomparison of the upper layer circulation of the western estimation. J. Atmos. Oceanic Technol., 28, 1324–1337,
equatorial Atlantic Ocean: In situ and satellite data. J. Geophys. doi:10.1175/2011JTECHO818.1.
Res., 104, 21 171–21 194, doi:10.1029/1999JC900124. Gouretski, V., and K. P. Koltermann, 2007: How much is the ocean
——, I. Pujol, and J. L. Melice, 2011: In situ validation of Jason-1 really warming? Geophys. Res. Lett., 34, L01610, doi:10.1029/
and Jason-2 altimetry missions in the tropical Atlantic Ocean. 2006GL027834.
Mar. Geod., 34, 319–339, doi:10.1080/01490419.2011.584833. ——, and F. Reseghetti, 2010: On depth and temperature biases in
Baringer, M. O., and S. L. Garzoli, 2007: Meridional heat transport bathythermograph data: Development of a new correction
determined with expendable bathythermographs—Part I: scheme based on analysis of a global ocean database. Deep-
Error estimates from model and hydrographic data. Deep-Sea Sea Res. I, 57, 812–833, doi:10.1016/j.dsr.2010.03.011.
Res. I, 54, 1390–1401, doi:10.1016/j.dsr.2007.03.011. Green, A. W., 1984: Bulk dynamics of the expendable bathy-
Brown, D., 2012: Tracker: Video analysis and modeling tool for thermograph (XBT). Deep-Sea Res., 31, 415–426, doi:10.1016/
physics education. [Available online at http://www.cabrillo. 0198-0149(84)90093-1.
edu/;dbrown/tracker/.] Hallock, Z. R., and W. J. Teague, 1992: The fall rate of the T-7
Carton, J. A., and E. J. Katz, 1990: Estimates of the zonal slope and XBT. J. Atmos. Oceanic Technol., 9, 470–483, doi:10.1175/
seasonal transport of the Atlantic North Equatorial Coun- 1520-0426(1992)009,0470:TFROTT.2.0.CO;2.
tercurrent. J. Geophys. Res., 95, 3091–3100, doi:10.1029/ Hamon, M., G. Reverdin, and P.-Y. Le Traon, 2012: Empirical
JC095iC03p03091. correction of XBT data. J. Atmos. Oceanic Technol., 29, 960–
Cheng, L., and Coauthors, 2015: XBT science: Assessment of in- 973, doi:10.1175/JTECH-D-11-00129.1.
strumental biases and errors. Bull. Amer. Meteor. Soc., doi:10.1175/ Hanawa, K., and H. Yoritaka, 1987: Detection of systematic errors
BAMS-D-15-00031.1, in press. in XBT data and their correction. J. Oceanogr. Soc. Japan, 43,
Cowley, R., S. Wijffels, L. Cheng, T. Boyer, and S. Kizu, 2013: 68–76, doi:10.1007/BF02110635.
Biases in expendable bathythermograph data: A new view ——, P. Rual, R. Bailey, A. Sy, and M. Szabados, 1995: A new
based on historical side-by-side comparisons. J. Atmos. Oceanic depth time equation for Sippican or TSK T-7, T-6 and T-4
Technol., 30, 1195–1225, doi:10.1175/JTECH-D-12-00127.1. expendable bathythermographs (XBT). Deep-Sea Res. I, 42,
DiNezio, P. N., and G. Goni, 2010: Identifying and estimating 1423–1451, doi:10.1016/0967-0637(95)97154-Z.
biases between XBT and Argo observations using satellite Heinmiller, R. H., C. C. Ebbesmeyer, B. A. Taft, D. B. Olson, and
altimetry. J. Atmos. Oceanic Technol., 27, 226–240, doi:10.1175/ O. P. Nikitin, 1983: Systematic errors in expendable bathy-
2009JTECHO711.1. thermograph (XBT) profiles. Deep-Sea Res., 30A, 1185–1197,
——, and G. J. Goni, 2011: Direct evidence of a changing fall-rate doi:10.1016/0198-0149(83)90096-1.
bias in XBTs manufactured during 1986–2008. J. Atmos. Oceanic Ishii, M., and M. Kimoto, 2009: Reevaluation of historical ocean
Technol., 28, 1569–1578, doi:10.1175/JTECH-D-11-00017.1. heat content variations with time-varying XBT and MBT
Domingues, R., G. J. Goni, S. Swart, and S. Dong, 2014: Wind depth bias corrections. J. Oceanogr., 65, 287–299, doi:10.1007/
forced variability of the Antarctic Circumpolar Current south s10872-009-0027-7.
of Africa between 1993 and 2010. J. Geophys. Res. Oceans, Kizu, S., C. Sukigara, and K. Hanawa, 2011: Comparison of the
119, 1123–1145, doi:10.1002/2013JC008908. fall rate and structure of recent T-7 XBT manufactured by
Dong, S., S. Garzoli, M. Baringer, C. Meinen, and G. Goni, 2009: Sippican and TSK. Ocean Sci., 7, 231–244, doi:10.5194/
Interannual variations in the Atlantic meridional overturning os-7-231-2011.
circulation and its relationship with the net northward heat Levitus, S., J. I. Antonov, T. P. Boyer, R. A. Locarnini, H. E.
transport in the South Atlantic. Geophys. Res. Lett., 36, Garcia, and A. V. Mishonov, 2009: Global ocean heat
L20606, doi:10.1029/2009GL039356. content 1955–2008 in light of recently revealed in-
——, M. O. Baringer, G. J. Goni, C. S. Meinen, and S. L. Garzoli, strumentation problems. Geophys. Res. Lett., 36, L07608,
2014: Seasonal variations in the South Atlantic Meridional doi:10.1029/2008GL037155.
Overturning Circulation from observations and numerical NOAA, 2014: WMO code table 1770. [Available online at https://
models. Geophys. Res. Lett., 41, 4611–4618, doi:10.1002/ www.nodc.noaa.gov/GTSPP/document/codetbls/wmocodes/
2014GL060428. table1770.html.]
Gilson, J., D. Roemmich, B. Cornuelle, and L.-L. Fu, 1998: Re- Phillips, H. E., and S. R. Rintoul, 2002: A mean synoptic view of
lationship of TOPEX/Poseidon altimetric height to steric the sub-Antarctic front south of Australia. J. Phys. Ocean-
height and circulation in the North Pacific. J. Geophys. Res., ogr., 32, 1536–1553, doi:10.1175/1520-0485(2002)032,1536:
103, 27 947–27 965, doi:10.1029/98JC01680. AMSVOT.2.0.CO;2.
Goes, M., G. Goni, V. Hormann, and R. C. Perez, 2013: Variability Reseghetti, F., M. Borghini, and G. M. R. Manzella, 2007: Factors
of the Atlantic off-equatorial eastward currents during 1993– affecting the quality of XBT data—Results of analysis on
2010 using a synthetic method. J. Geophys. Res. Oceans, 118, profiles from the western Mediterranean Sea. Ocean Sci., 3,
3026–3045, doi:10.1002/jgrc.20186. 59–75, doi:10.5194/os-3-59-2007.
DECEMBER 2015 BRINGAS AND GONI 2263

Ridgway, K. R., R. C. Coleman, R. J. Bailey, and P. Sutton, 2008: Singer, J. J., 1990: On the error observed in electronically digitized
Decadal variability of East Australian Current transport in- T-7 XBT data. J. Atmos. Oceanic Technol., 7, 603–611, doi:10.1175/
ferred from repeated high-density XBT transects, a CTD 1520-0426(1990)007,0603:OTEOIE.2.0.CO;2.
survey and satellite altimetry. J. Geophys. Res., 113, C08039, Smith, N. R., and Coauthors, 1999: The role of XBT sampling in the
doi:10.1029/2007JC004664. ocean thermal network. Proc. OceanObs’99, Saint Raphaël,
Rintoul, S. R., S. Sokolov, and J. Church, 2002: A 6 year record of France, Centre National d’Études Spatiales, 1–26.
baroclinic transport variability of the Antarctic Circumpolar Thadathil, P., A. K. Saran, V. V. Gopalakrishna, P. Vethamony,
Current at 1408E derived from expendable bathythermograph N. Araligidad, and R. Bailey, 2002: XBT fall rate in waters of ex-
and altimeter measurements. J. Geophys. Res., 107, 3155, treme temperature: A case study in the Antarctic Ocean. J. Atmos.
doi:10.1029/2001JC000787. Oceanic Technol., 19, 391–396, doi:10.1175/1520-0426-19.3.391.
Roemmich, D., and B. Cornuelle, 1987: Digitization and calibra- Wijffels, S. E., J. Willis, C. M. Domingues, P. Barker, N. J. White,
tion of the expendable bathythermograph. Deep-Sea Res., A. Gronell, K. Ridgway, and J. A. Church, 2008: Changing
34A, 299–307, doi:10.1016/0198-0149(87)90088-4. expendable bathythermograph fall rates and their impact on
——, J. Gilson, B. Cornuelle, and R. Weller, 2001: Mean and time- estimates of thermosteric sea level rise. J. Climate, 21, 5657–
varying meridional transport of heat at the tropical/subtropical 5672, doi:10.1175/2008JCLI2290.1.
boundary in the North Pacific Ocean. J. Geophys. Res., 106, Xiao, H., and X. Zhang, 2012: Numerical investigations of the fall
8957–8970, doi:10.1029/1999JC000150. rate of a sea-monitoring probe. Ocean Eng., 56, 20–27,
Seaver, G. A., and S. Kuleshov, 1982: Experimental and analyti- doi:10.1016/j.oceaneng.2012.08.002.
cal error of expendable bathythermograph. J. Phys. Ocean- ——, C. Liu, and J. Tao, 2006: Numerical simulation and experi-
ogr., 12, 592–600, doi:10.1175/1520-0485(1982)012,0592: ments of a probe descending in the sea. Ocean Eng., 33, 1343–
EAAEOT.2.0.CO;2. 1353, doi:10.1016/j.oceaneng.2005.10.001.