This article was downloaded by: [University Of Maryland]

On: 15 October 2014, At: 02:42

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Remote Sensing Letters

Publication details, including instructions for authors and
subscription information:
[Link]

Improved determination of Indian

Ocean surface currents using satellite
data
a a a a
Rajesh Sikhakolli , Rashmi Sharma , Raj Kumar , B. S. Gohil ,
a b a
Abhijit Sarkar , K. V. S. R. Prasad & Sujit Basu
a
Atmospheric and Oceanic Sciences Group, Space Applications
Centre (SAC), ISRO , Ahmedabad , 380015 , India
b
Department of Meteorology and Oceanography , Andhra
University , Visakhapatnam , 530 003 , India
Published online: 17 Oct 2012.

To cite this article: Rajesh Sikhakolli , Rashmi Sharma , Raj Kumar , B. S. Gohil , Abhijit Sarkar , K.
V. S. R. Prasad & Sujit Basu (2013) Improved determination of Indian Ocean surface currents using
satellite data, Remote Sensing Letters, 4:4, 335-343, DOI: 10.1080/2150704X.2012.730643

To link to this article: [Link]

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
 Conditions of access and use can be found at [Link]
and-conditions
Downloaded by [University Of Maryland] at 02:42 15 October 2014
 Remote Sensing Letters
Vol. 4, No. 4, April 2013, 335–343

Improved determination of Indian Ocean surface currents using

satellite data

RAJESH SIKHAKOLLI*†, RASHMI SHARMA†, RAJ KUMAR†,

B. S. GOHIL†, ABHIJIT SARKAR†, K. V. S. R. PRASAD‡ and SUJIT BASU†
†Atmospheric and Oceanic Sciences Group, Space Applications Centre (SAC), ISRO,
Ahmedabad 380015, India
‡Department of Meteorology and Oceanography, Andhra University, Visakhapatnam
530 003, India
Downloaded by [University Of Maryland] at 02:42 15 October 2014

(Received 13 May 2012; in final form 12 September 2012)

A diagnostic algorithm for determining the upper layer current in the tropical
Indian Ocean (TIO) is presented in this study. The algorithm is based on quasi–
linear and steady-state physics. The current in the upper layer consists of three
components, namely geostrophic component, wind-driven (Ekman) component
and buoyancy component. Each of these components is estimated from differ-
ent satellite data. The geostrophic component is estimated from merged altimeter
product. The Ekman component is estimated from ocean surface winds derived
from Oceansat-2 Scatterometer (OSCAT), whereas the buoyancy component is
estimated from sea surface temperature (SST) derived from a combination of
infrared and microwave radiometers. The product derived for the year 2010 is com-
pared with a similar product, known as Ocean Surface Current Analysis Real-Time
(OSCAR), computed by the same algorithm, but with somewhat different data sets.
Afterwards, the product is compared with currents measured by several buoys at
different locations in the TIO and it is observed that the current variability is better
captured by the data generated using the present algorithm.

1. Introduction
Monitoring the variations of ocean surface currents is extremely important for climate
studies, since these currents are major transporters of heat and salt. Surface current
in the Indian Ocean displays a unique feature of annual reversal in response to annu-
ally reversing monsoon winds (Shenoi et al. 1999). Traditionally, ocean currents have
been measured by ships, moored current meters and drifting buoys. However, the
situation in the Indian Ocean is worse than in the Atlantic and Pacific because of
the relative sparsity of in situ data. Although many ocean surface parameters, such
as sea level anomaly (SLA), sea surface wind (SSW) and sea surface temperature
(SST), are being routinely measured by satellite-borne sensors, there is no sensor yet
to measure sea surface currents. Fortunately, Bonjean and Lagerloef (2002) devised an
algorithm for determining this important variable from all the three parameters men-
tioned previously, namely the SLA, SSW and SST. This product is known as Ocean
Surface Current Analysis Real-time (OSCAR). The method for deriving the surface

*Corresponding author. Email: rajeshsikhakolli@[Link]

Remote Sensing Letters
ISSN 2150-704X print/ISSN 2150-7058 online © 2012 Taylor & Francis
[Link]
[Link]
 336 R. Sikhakolli et al.

current is based on the resolution of quasi-steady quasi-linear momentum equations,

neglecting local acceleration. Equatorial velocities are obtained by solving a weak for-
mulation of the momentum equations using a basis set of orthogonal polynomials.
Validation and error analysis of the OSCAR product was carried out in the tropical
Pacific Ocean (Johnson et al. 2007). A detailed validation of these OSCAR currents
in the tropical Indian Ocean (TIO) with Research Moored Array for African–Asian–
Australian Monsoon Analysis and Prediction (RAMA) buoy data has been carried
out by Sikhakolli et al. (2012).
After the end of QuikSCAT mission in November 2009, the OSCAR product is
being generated using National Center for Environmental Prediction (NCEP) model
winds. At the same time, daily analysed SSW from the scatterometer on board
Oceansat-2 satellite launched by the Indian Space Research Organization (ISRO)
has now become available. Moreover, Archiving, Validation and Interpretation of
Downloaded by [University Of Maryland] at 02:42 15 October 2014

Satellite Oceanographic data (AVISO) have recently started generating daily sea sur-
face height (SSH) as a merged product using data from several altimeters on an
experimental basis. Both these facts provide us an opportunity to generate daily
ocean surface currents using scatterometer SSW, altimeter SLA and radiometer SST
analogous to the determination of OSCAR current. As usual, once the currents are
determined, one has to validate the product using in situ and other auxiliary data
sets. This determination and subsequent validation constitute the main body of our
article.

2. Data sets used

The operational wind vector products from Oceansat-2 Scatterometer (OSCAT) are
regularly being generated on 0.5◦ × 0.5◦ swath grid using the algorithms developed by
Gohil et al. (2008, 2010) and are made available on ISRO’s website, ‘[Link]’.
OSCAT daily gridded (0.5◦ × 0.5◦ grid) analysed winds are also generated using Level
2B (L2B) swath wind data (Chakraborty and Kumar 2013) and are operationally
available on ISRO’s website, ‘[Link]’. In this study, these analysed winds
during the year 2010 for the TIO region are used. The global gridded daily SLA as
a merged product of several altimeters for the year 2010 has been obtained from
AVISO FTP site ([Link]). The SLA data are at a resolution of
0.33◦ × 0.33◦ . Apart from SLA, the climatology of mean dynamic height (Rio et al.
2009) at a resolution of 0.5◦ × 0.5◦ has also been used. This is required for produc-
ing daily SSH data for the TIO. The daily SST analysis used in this study is from
[Link] The spatial resolution of the data is
0.25◦ × 0.25◦ grid. For this study, all the input data sets of SSW, SSH and SST are
brought to the uniform resolution of 0.5◦ × 0.5◦ using interpolation and averaging.
Apart from the three sets of satellite data used for determining Space Applications
Centre (SAC) current, we have also used certain validation data sets. The first such
data are the OSCAR currents for the year 2010. The data are available at 5-day inter-
vals and at 1◦ × 1◦ spatial resolution. In the TIO, there are several RAMA buoys,
which measure surface currents (McPhaden et al. 2009). In this study, the time series
data of currents from 12 RAMA buoys for the year 2010 have been used.

3. Methodology
The determination of ocean surface current is based on the solution of the equations
of quasi-linear and steady flow in a surface layer where the horizontal velocity
 Ocean surface current in the tropical Indian Ocean 337

U(x,y,z,t) = (u(x,y,z,t), v(x,y,z,t)) is a function of the horizontal coordinates x, y, the

vertical coordinate z and time t. The vertical turbulent mixing is characterized by an
eddy viscosity A uniform with depth. The vertical shear U z reaches zero at a constant
scaling depth of z = −H.
Using complex notation U(x,y,z,t) = u + iv and ∇ = ∂/∂x + i∂/∂y, the basic
equations are as follows:

if U = −(1/ρ)∇p + AU z (1a)

(1/ρ)pz = −g + ∇θ (1b)

∇θ = gχT ∇(SST) (1c)

Downloaded by [University Of Maryland] at 02:42 15 October 2014

where f is Coriolis parameter, p is pressure, Pz is the vertical derivative of pressure, and

θ is temperature, with −H ≤ z ≤ 0, and subject to the following boundary conditions:

U z (z = 0) = τ /A (2a)

U z (z = −H) = 0 (2b)

The characteristic density is ρ = 1025 kg m −3 . The acceleration due to gravity

g = 9.8 m s−2 and the coefficient of thermal expansion χ T = 3 × 10–4 K−1 . The vector
field τ = τ x + iτ y represents the surface wind stress divided by ρ, whereas τ x and τ y
are the zonal and meridional components of τ. H has been chosen to be 70 m. The
parameter A is chosen by the empirical formulation due to Santiago-Mandujano and
Firing (1990).
Since the solution method is the same as the one presented in Bonjean and Lagerloef
(2002), we refer the readers to the above-mentioned reference for the elaborate math-
ematics involved. However, we would like to point out the fact that the data sets used
in the present case are quite different and most importantly the use of daily fields
permitted us to determine the Indian Ocean surface currents at high temporal resolu-
tion (in fact, daily) unlike the OSCAR currents which are at 5-day intervals. Although
OSCAR data sets are still being provided, the winds used for determining the cur-
rents are from NCEP, unlike the present case where we have used winds measured by
OSCAT.

4. Results and discussion

Ocean surface currents generated by our algorithm will be hereafter referred as SAC
currents. In figure 1 (left panel), we show the snapshots of SAC currents for two repre-
sentative dates in January and July 2010 corresponding to the northeast and southwest
monsoon seasons. For the purpose of comparison, we also show side by side (right
panel of figure 1) the OSCAR currents for the same dates. First thing to be noted is
the similarity of the two products. They look strikingly similar, which is not surprising,
since both the products are generated using same mathematical algorithm and from
similar data sets. One can clearly notice the major ocean surface currents in the TIO
which are very well captured by both the products. Here, by major currents we mean
 338 R. Sikhakolli et al.

(a) (b)
30° N 30° N
m s–1
20° N 20° N

10° N 10° N
2.0

EQ EQ
1.8

10° S 10° S 1.6

20° S 20° S 1.4

30° S 30° S 1.2

40° E 50° E 60° E 70° E 80° E 90° E 100° E110° E 40° E 50° E 60° E 70° E 80° E 90° E 100° E 110° E
Downloaded by [University Of Maryland] at 02:42 15 October 2014

(c) (d) 1.0

30° N 30° N

0.8
20° N 20° N

0.6
10° N 10° N

0.4
EQ EQ

0.2
10° S 10° S

0.0
20° S 20° S

30° S 30° S
40° E 50° E 60° E 70° E 80° E 90° E 100° E110° E 40° E 50° E 60° E 70° E 80° E 90° E 100° E110° E

Figure 1. Snapshot of ocean surface currents from SAC algorithm (left panel) and the corre-
sponding currents from OSCAR product (right panel) for two representative dates in January
(a) and (b) and July (c) and (d) in the year 2010.

the northeast monsoon current in January and the southwest monsoon current and the
south equatorial current in July 2010. However, a closer inspection reveals that there
are subtle differences. This again is to be expected, since the data sets used to generate
the currents, although similar, are not strictly identical. Before 2010, the OSCAR cur-
rents were generated using QuikSCAT winds. Presently, NCEP model winds are being
used to determine OSCAR currents. Since winds are the main driving force for the
Ekman component of OSCAR currents, this difference is to be expected. Apart from
that, the SSTs used in both the products are also different. Moreover, the OSCAR
currents are slightly less in intensity compared to SAC currents because the SAC cur-
rents are generated daily and the OSCAR product represents the mean current during
a 5-day period.
In figure 2, we show the density scatter plot between SAC and OSCAR products
for the study period. For this purpose, the SAC current is also brought to 5-daytime
intervals and to 1◦ × 1◦ grids like OSCAR product by using running averages in
time and spatial collocation. As shown in the figure, the match between both the
currents is quite well with a small bias of 0.04 m s−1 for zonal component and no
significant bias for the meridional component. The root mean square errors (RMSEs)
 Ocean surface current in the tropical Indian Ocean 339

(a) (b)
No. of points: 821,293 No. of points: 821,293
X-Mean: –0.04 100 0.4 X-Mean: –0.01 100
0.4
Y-Mean: –0.01 Y-Mean: –0.02
90 90
Bias: –0.03 0.3 Bias: 0.00
0.3
RMSE: 0.16 RMSE: 0.15 80

OSCAR meridional current (m s–1)

80
Corrleation: 0.68
OSCAR zonal current (m s–1)

0.2 Corrleation: 0.84 0.2

70 70

0.1 0.1 60
60

0 50 0 50

40 –0.1 40
–0.1
30 30
–0.2 –0.2
20 20
–0.3 –0.3
10 10
–0.4 1 –0.4 1
Downloaded by [University Of Maryland] at 02:42 15 October 2014

–0.4 –0.3 –0.2 –0.1 0 0.1 0.2 0.3 0.4 –0.4 –0.3 –0.2 –0.1 0 0.1 0.2 0.3 0.4
SAC zonal current (ms–1) SAC meridional current (m s–1)

Figure 2. Density scatter (peak normalized) plot between SAC algorithm-derived currents and
OSCAR product in the TIO for the period 1 January 2010 to 31 December 2010: (a) for zonal
current (m s−1 ) and (b) for meridional current (m s−1 ).

are 0.16 m s−1 and 0.17 m s−1 and the correlation coefficients are 0.81 and 0.58 for
zonal and meridional currents, respectively.
However, as is the practice for validating any satellite-derived product, this one is
to be validated by comparing with in situ data prior to its actual use in any physi-
cal oceanographic applications. We have performed such a comparison using all the
available RAMA buoys in our region of study (12 in number). However, for the econ-
omy of space, we are showing the result for only one representative buoy in figure 3,
after setting aside the results for all the buoys in a tabular form (table 1). The repre-
sentative buoy is located at 15◦ N and 90◦ E. This buoy happens to be located in the
central Bay of Bengal region. There are many high-frequency fluctuations in the buoy
data. SAC and OSCAR products, obtained as solutions of a differential equation, are
not expected to exhibit these high-frequency variabilities and are obviously smoother
(barring of course, the small-scale fluctuations caused by errors in the data sets used
to generate the actual current product from the solution). Another possible reason for
the absence of the high-frequency variabilities is the neglect of the time-derivative term
while deriving these products.
Nevertheless, it is gratifying to note that the match between buoy and SAC cur-
rents is somewhat better than that between buoy and OSCAR. One possible reason
could be the use of scatterometer-observed winds in SAC product, whereas the other
reason could be the use of more frequent (daily, in fact) satellite products. In order
to further strengthen our results, in figure 4, we show a similar comparison by
averaging RAMA buoy and SAC currents similar to the averaging used in OSCAR
(10-day running average at every 5-day interval). It is evident from the figure that
the match between SAC current and buoy current is still better than that between
OSCAR current and buoy current. Moreover, we also observed a slight phase shift
for the meridional component of OSCAR current in comparison with that of buoy
current.
In figure 5, we show the statistics of comparison between SAC and RAMA buoy
measured daily currents in the TIO as a scatter plot, which shows a correlation of
0.7 for the zonal component and a correlation of 0.32 for the meridional component.
 340 R. Sikhakolli et al.

(a)
0.60

0.40
Zonal current (m s–1)

0.20

0.00

–0.20

–0.40
Downloaded by [University Of Maryland] at 02:42 15 October 2014

–0.60
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
2010
(b)
0.80

0.60

0.40
Meridional current (m s–1)

0.20

0.00

–0.20

–0.40

–0.60

–0.80
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
2010

Figure 3. Comparison of ocean surface current time series at a RAMA buoy location (15◦ N
and 90◦ E) in the Bay of Bengal for the year 2010: (a) for zonal component and (b) for merid-
ional component. The black line represents the RAMA buoy data and the red and green lines
represent the SAC and OSCAR currents, respectively.

The poorer match between the meridional components of SAC and buoy currents
is caused by the fact that in the equatorial buoys there is practically no correlation
between the two currents (table 1). This mismatch has vastly contributed to the overall
poor correlation. This mismatch for such products is, in fact, generic in nature and the
OSCAR product is also plagued by this mismatch, see, e.g. Johnson et al. (2007) for the
mismatch in the tropical Pacific Ocean and Sikhakolli et al. (2012) for the mismatch
in the TIO. The possible reasons for this meridional mismatch are described in those
studies. In particular, in the Indian Ocean this is caused primarily by the neglect of the
local acceleration term in the momentum balance.
 Ocean surface current in the tropical Indian Ocean 341

Table1. Statistics of comparison between SAC-derived and RAMA buoy-measured currents

during the year 2010.

Latitude Longitude Number of points Bias (m s−1 ) RMSE (m s−1 ) Correlation (r)

0.00 80.50 256 0.26 (0.04) 0.55 (0.42) 0.56 (0.15)

0.00 90.00 76 0.24 (0.10) 0.47 (0.39) 0.35 (−0.01)
1.50 80.50 69 0.28 (0.30) 0.27 (0.39) 0.74 (0.17)
1.50 90.00 202 0.11 (−0.06) 0.38 (0.38) 0.49 (0.07)
12.00 90.00 63 −0.13 (−0.01) 0.11 (0.12) 0.78 (0.71)
–12.00 67.00 36 0.06 (−0.01) 0.12 (0.13) 0.60 (0.75)
–12.00 80.50 173 −0.10 (0.02) 0.10 (0.10) 0.69 (0.56)
15.00 90.00 196 0.03 (−0.02) 0.12 (0.13) 0.79 (0.90)
–4.00 80.50 103 0.01 (0.03) 0.21 (0.20) 0.41 (0.63)
8.00 90.00 275 0.02 (0.01) 0.18 (0.22) 0.67 (0.31)
–8.00 55.00 175 −0.06 (0.04) 0.16 (0.13) 0.56 (0.57)
Downloaded by [University Of Maryland] at 02:42 15 October 2014

–8.00 80.50 148 −0.13 (0.03) 0.16 (0.13) 0.59 (0.12)

Note: RMSE denotes root mean square error.

(a)

0.40
Zonal current (m s–1)

0.20

0.00

–0.20

–0.40

JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
2010
(b)
0.60

0.40
Meridional current (m s–1)

0.20

0.00

–0.20

–0.40

–0.60

JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
2010

Figure 4. Same as in figure 3 except that the RAMA buoy-observed and SAC-derived currents
have been averaged to 5-day intervals.
 342 R. Sikhakolli et al.

(a) (b)
2.00 1.00
No. of points: 1772 No. of points: 1772
X-Mean: 0.04 X-Mean: 0.00
Y-Mean: 0.08 Y-Mean: 0.03
1.50

SAC meridional current (m s–1)

Bias: 0.05 Bias: 0.02
SAC zonal current (m s–1)

RMSE: 0.32 0.50 RMSE: 0.27

Correlation: 0.70 Correlation: 0.32
1.00

0.50 0.00

0.00
–0.50
–0.50

–1.00 –1.00
Downloaded by [University Of Maryland] at 02:42 15 October 2014

–1.00 –0.50 0.00 0.50 1.00 1.50 2.00 –1.00 –0.50 0.00 0.50 1.00
RAMA buoy zonal current (m s–1) RAMA buoy meridional current (m s–1)

Figure 5. Scatter plot between SAC algorithm-derived and RAMA buoy-measured currents
in the TIO during 2010: (a) for zonal component and (b) for meridional component.

5. Summary
Following Bonjean and Lagerloef (2002), a diagnostic algorithm has been formulated
for computing the upper layer current in the TIO. The steady current is derived as the
sum of geostrophic current, wind-driven current and buoyancy current. Each of these
components is derived using one particular type of satellite product. The geostrophic
component is derived using merged altimeter data and mean dynamic height. The
wind-driven component is derived from OSCAT winds, whereas the buoyancy com-
ponent is from SST derived from satellite radiometers. The currents generated for the
year 2010 are compared with a similar kind of product, known as OSCAR. The two
products appear to be qualitatively similar. A detailed quantitative validation is carried
out by comparing the product with currents from several RAMA buoys in the TIO
and it has been found that the variabilities are better captured by the present prod-
uct than by the existing OSCAR product. However, a definitive conclusion regarding
the superiority (or, otherwise) of this product requires more extensive validation of
the product over a longer time period. Such a study is planned to be taken up in
future.

Acknowledgements
We thank the director, SAC, and the deputy director, Earth, Ocean, Atmosphere,
Planetary Sciences and Applications Area, SAC, for motivation. We would like to
place on record our sincere gratitude to the group director, Atmospheric and Oceanic
Sciences Group, for encouragement. We also thank Mr. Abhishek Chakraborty for
the daily gridded OSCAT wind product; NOAA, USA, for the OSCAR product; and
AVISO for the altimeter product.

References
BONJEAN, F. and LAGERLOEF, G.S.E., 2002, Diagnostic model and analysis of the surface
currents in the tropical Pacific Ocean. Journal of Physical Oceanography, 32,
pp. 2938–2954.
 Ocean surface current in the tropical Indian Ocean 343

CHAKRABORTY, A. and KUMAR, R., 2013, Generation and evaluation of analysed wind vectors
over the global ocean. Remote Sensing Letters, 4, pp. 114–122.
GOHIL, B.S., SARKAR, A. and AGARWAL, V.K., 2008, A new algorithm for wind vector retrieval
from scatterometer. IEEE Geoscience Remote Sensing Letters, 5, pp. 387–391.
GOHIL, B.S., SHARMA, P., SIKHAKOLLI, R. and SARKAR, A., 2010, Directional stability and
conservation of scattering (DiSCS)-based directional-ambiguity removal algorithm for
improving wind-fields from scatterometer: a QuikSCAT example. IEEE Geoscience
Remote Sensing Letters, 7, pp. 592–595.
JOHNSON, E.S., BONJEAN, F., LAGERLOEF, G.S.E., GUNN, J.T. and MITCHUM, G.T., 2007,
Validation and error analysis of OSCAR sea surface currents. Journal of Atmospheric
and Oceanic Technology, 24, pp. 688–701.
MCPHADEN, M.J., MEYERS, G., ANDO, K., MASUMOTO, Y., MURTY, V.S.N.,
RAVICHANDRAN, M., SYAMSUDIN, F., VIALARD, J., YU, L. and YU, W., 2009,
RAMA: the research moored array for African-Asian-Australian monsoon analysis
Downloaded by [University Of Maryland] at 02:42 15 October 2014

and prediction. Bulletin of American Meteorological Society, 90, pp. 459–480.

RIO, M-H., SCHAEFFER, P., MOREAUX, V., LEMOINE, J-M., BRONNER, E., 2009, A new Mean
Dynamic Topography computed over the global ocean from GRACE data, altimetry
and in-situ measurements. In OceanObs09 Symposium, 21–25 September 2009, Venice.
SANTIAGO-MANDUJANO, F. and FIRING, E., 1990, Mixed-layer shear generated by wind stress
in the central equatorial Pacific. Journal of Physical Oceanography, 20, pp. 1576–1582.
SHENOI, S.S.C., SAJI, P.K. and ALMEIDA, A.M., 1999, Near-surface circulation and kinetic
energy in the tropical Indian Ocean derived from Lagrangian drifters. Journal of Marine
Research, 57, pp. 885–907.
SIKHAKOLLI, R., SHARMA, R., BASU, S., GOHIL, B.S., SARKAR, A. and PRASAD, K.V.S.R., 2012,
Evaluation of OSCAR ocean surface current product in the tropical Indian Ocean using
in situ data. Journal of Earth System Science (in press).