Content uploaded by Prem Kumar
Author content
All content in this area was uploaded by Prem Kumar on Mar 01, 2021
Content may be subject to copyright.
Mapping of Total Suspended Matter based on Sentinel-2 data on
the Hooghly River, India
Indian Journal of Ecology (2021) 48(1): 159-165
Manuscript Number: 3188
NAAS Rating: 4.96
Abstract: This study aims to develop a regional algorithm for monitoring and retrieval of total suspended matter concentration (C ) using in-
TSM
situ C data, in-situ remote sensing reflectance (R ) data and Sentinel 2 MSI data on Hooghly River. The field measurements were carried for
TSM rs
20 stations using Satlantic Hyperspectral Ocean Colour Radiometer (HyperOCR) and water samples were also collected at a depth of 0.5m
from April to May 2018. The concentration of total suspended matter varies from 132 to 540 mg L . The calibration was done for 70% of the data
-1
between in-situ C and in-situ remote sensing reflectance and the remaining 30% of the data is used for in-situ validation. The in-situ
TSM
validation result shows the band combination of B7/B2 has higher fitting. For satellite validation, Sentinel 2 Multispectral Instrument (MSI)
satellite data applies to the in-situ validation models to the retrieval of C , which also confirms the band ratio of B7/B2 gives a good correlation
TSM
with R =0.75. The study shows, the applicability of Sentinel 2 MSI data for retrieval of C in Hooghly River, Kolkata. The Sentinel 2 MSI B7/B2
2
TSM
is highly recommended for mapping a higher concentration of suspended matter in the study area, respectively.
Keywords: Total suspended matter, Hyperspectral radiometer, Regional algorithm, Sentinel 2 MSI data, Hooghly river
R. Premkumar, R. Venkatachalapathy and S. Visweswaran
Department of Physics, Annamalai University, Annamalai Nagar-608 002, India
E-mail: premkumar140994@gmail.com
In optical remote sensing, the water is classified into the
case I and case II water. The case I water is dominant of
phytoplankton, which is specifically for the open ocean. The
case II water is significantly dominated by the chlorophyll-a
concentration (Chl-a), total suspended matter concentration
(C ) and coloured dissolved organic matter (CDOM).
TSM
Mostly inland water bodies, estuary and its coastal areas
come under case II water. The importance of understanding
the optical properties of Case II water is necessary for
monitoring and protecting the aquatic environment (Avinash
et al 2012). Among all optically active substance, the
concentration of total suspended matter plays a major role
because of its impact in the light transmission, transportation
of minerals and nutrients, contamination and deposition of
sediments in water bodies.
In specific at case II water, anthropogenic and man-
made activities may dominate the C distribution pattern.
TSM
The C values are spatially heterogeneous and temporally
TSM
dynamic in all water bodies. Therefore, obtaining the result of
C at high spatiotemporal level helps to understand the
TSM
driving force of the water bodies and further for maintaining
the aquatic ecosystem. Even through field and lab-based
measurements can provide more accurate results however, it
is costly and may not possible for more temporal and spatial
analysis. In opposite to that, the satellite remote sensing is
used to provide an economic approach for monitoring the
water constituents (Merina 2017). Bhaskaran et al (2014)
encountered a high sedimentation issue that requires
intermi ttent support by the method of dredging.
Ramakrishnan et al (2013) results show the increase of
sediment concentration will increase the water-leaving
radiance in the Gulf of Cambay. In a satellite-based ocean
colour remote sensing, Sentinel-2 MSI data is used in optical
studies of water because of its least spatial resolution.
Sentinel-2 MSI satellite around the sun-synchronous polar
orbit can provide a better spatial resolution of 10, 20 and 60m
and a temporal resolution of 2 to 5 days. Sentinel-2 MSI
satellite data set is employed to retrieve the total suspended
matter concentration value in a turbid estuary (Liu et al 2017).
Dornhofer et al (2016) have processed the Sentinel-2
MSI satellite data for the analysis of optically active
substances over the inland water shows the acceptable
result. Gernez et al (2015) retrieved the C from Sentinel -2
TSM
MSI data in the turbid estuary region. Manzo et al (2015)
results show the accuracy of Sentinel 2 MSI satellite data for
obtaining the optically active substances in the Italian Lakes
by using the bio-optical model. Kutser et al (2016) used the
Sentinel-2 MSI data to retrieve C and it is observed to be
TSM
more beneficial than Landsat 8 OLI for black lakes. The
universal algorithm for the retrieval of the C will not be
TSM
applicable for all study areas (Pradhan et al 2005). So the
regional based algorithm is required for a more accurate
model. To develop the regional algorithm for retrieval of C ,
TSM
the empirical algorithm was created between band ratio of in-
situ Remote sensing reflectance (R ) and in-situ C .
rs TSM
The objective of this present study in the Hooghly River,
Kolkata includes prediction the spectral signature and
measure the concentration of C in the in-situ water s TSM
sample as well as comparison and validation of in-situ C s TSM
data with in-situ R and satellite R data.
rs rs
MATERIAL AND METHODS
Study area: The Hooghly river is situated roughly between
the latitudes 22.02°–21.90° N and longitudes 88.07°-88.01°E
in the state of West Bengal, India which is passing through 2
major cites called Kolkata and Howrah then finally joined in
the Bay of Bengal. The widening of the river increases
gradually to downstream from Haldia port which creates
“Hooghly Estuary”. In the mainstream of the Hooghly river,
two important ports were located. One is Kolkata port
constructed in 1870, which is a major port in India on the
riverine of about 150 km of length from Coastal region.
Another one is Haldia Port which is constructed during 1977
at Haldia to reduce the ship traffic in Kolkata port. Haldia port
is now facing the problem of dredging and siltation for a long
time. The sediment deposition occurs majorly in downstream
of the Hooghly estuary because of a decrease in flow velocity
of the water. To operate the Haldia port, continuous
monitoring of C is important which will also helpful in timely
TSM
the maintenance of dredging and siltation for a long time
(Prasad and Singh 2010).
In-situ measurement: The field survey was carried out from
April to May 2018 (pre-monsoon period). Twenty samplings
were fixed along the estuarine region of the Hooghly River.
Each sampling location reached by using a global positioning
system (Garmin Ltd., USA). To minimise the error in the in-
situ data collection, the boat was switched off to avoid the
turbulence in sampling point and then measurements were
done in the shallow free area of the boat. The surface water
samples were collected in 0.5m depth using the Niskin water
sampler. To measure the spectral signature of water,
Satlantic HyperOCR sensors are used. The sensor in the
instrument used to measure the water-leaving radiance (L )
w
and downwelling irradiance (E ) which gives the optical
d
properties of the water. The Satlantic HyperOCR radiometer
has a spectral range of 350 nm to 800 nm with the spectral
data collection interval of 3.3 nm. All the collected data stored
in the computer with dot raw extension (RAW). SatCon is a
software developed by the Sea-Bird Scientific Company
(http://www.seabird.com) for processing radiometer data
(RAW). The final product of the processed data will provide
the water-leaving radiance (L ) and downwelling radiance
W
(E ). The in-situ remote sensing reflectance (R ) data is
d rs
calculated by the ratio of water-leaving radiance (L ) and
W
downwelling radiance (E ) (Minu et al 2016).
d
Laboratory analysis: To measure the concentration of total
suspended matter (C ), all the collected water samples
TSM
were filtered through pre-dried and weighted Whatman GF/F
glass fiber filter paper of 0.45 m pore size and 47 mm μ
diameter (Strickland et al 1972). The filter papers were kept
at 60 C for 6 hours in the hot-air oven and reweighted again at
o
room temperature. The concentration of total suspended
matter is obtained by dividing the difference value between
the final weight (g) and initial weight (g) by the volume of
sample (ml).
Satellite data acquisition and processing: The Sentinel 2
MSI have a wide view of 290 km and it contains a total
number of 13 spectral bands which has the range from visible
to shortwave infrared (SWIR) region, and it has a spatial
resolution of 60m (3 bands), 20m (6 bands), 10m (4 bands).
Sentinel 2 MSI L1C images captured on 15 May 2018 were
downloaded from the European Space Agency website
(https://scihub.copernicus.eu/dhus/). Sentinel Application
Platform (SNAP) is open-source software which helps to
process the Sentinel products such as opening and exploring
Sentinel data, masking, band visualization, atmospheric
correction etc. All satellite data processing was done by using
SNAP software. The image correction for atmospheric effect
(iCOR) is a tool in the SNAP software which is helping for
atmospheric correction in the Sentinel 2 MSI data over the
land and water (Keukelaere et al 2018). After the atmospheric
correction, satellite remote sensing reflectance (R ) data can
rs
be easily retrieved from the satellite imagery by marking the
sampling location (View-Tool-Windows-Pin manager) in the
SNAP software (Malenovský et al 2012). In satellite
validation, retrieved remote sensing reflectance data will be
applied to the newly developed model for spatial and
temporal mapping of C .
TSM
Model development: To develop the model, 70% of the in-
situ data were used for calibration and 30% of the data is
used for validation. For satellite validation of the model,
Sentinel 2 MSI data were applied to the developed model.
The total dataset is sorted from low to high based on C
TSM
values. The systematic sampling is the statistical method that
allows the researcher to select 70% of the data for calibration
and 30% of the data for validation in equal distribution
manner. The spectral signature of R is plotted and in-situ rs
analyzed. Among the different regression analysis, the
polynomial regression analysis gives a good correlation
between in-situ C and in-situ R (other regression results
TSM rs
are not shown). There are different band ratio model was
developed between in-situ C and in-situ R Out of that, the
TSM rs.
)1........(
)(
)(
)(
Ed
Lw
Rrs
CTSM (mg ) = l-1 Final weight (g) – Initial weight (g)
Volume of sample (ml)* (2)
160 R. Premkumar, R. Venkatachalapathy and S. Visweswaran
top 5 well-correlated models were selected for further
analysis of in-situ and satellite data validation.
In-situ and Satellite validation: The in-situ validation
datasets are applied to calibrated models for estimation of
C . The retrieved C is compared with measured in-situ
TSM TSM
C data. The statistical analysis such as root mean square
TSM
error (RMSE), mean absolute percentage error (MAPE) and
the coefficient of determination (R ) values were calculated
2
between the measured and estimated values to select the
best fitting and validation accuracy. The Sentinel 2 R is
rs
applied to the newly developed models to retrieve CTSM
values. The retrieved C values used to provide the spatial
TSM
distribution of total suspended matter concentration in the
study area (Sravanthi et al 2013).
RESULTS AND DISCUSSION
In-situ & satellite data: The in-situ remote sensing
reflectance spectra show that the study area is dominant of
TSM concentration. The peak near 780nm indicates the
dominance of C in the spectral signature of water sample
TSM
(Fig. 2). The increase of C will increase the remote sensing
TSM
reflectance and vice versa. There are two C dominant
TSM
peaks near 705 nm and 780nm. All the dataset are
2/1
1
2
)(/1
n
i
actual (3)forecastNRMSE
Datasets No. of total data Min. C (mg L )
TSM
-1 Max. C ( mg L )
TSM
-1 Average C ( mg L )
TSM
-1 S.D C ( mg L )
TSM
-1 C.V (%)
Calibration 14 183 461 313 95 30.4
in-situ alidationv 6 137 540 314 193 61.5
Satellite validation 18 222 343 269 42 15.5
Table 1. Statistical analysis of calibration and in-situ & satellite validation data sets of TSM concentration
S.D – tandard deviation, C.V – Coefficient of VariatioS n
Band Wave length (nm) Resolution (m)
B1-Coastal aerosol 443 60
B2-Blue 490 10
B3-Green 560 10
B4-Red 665 10
b5-red edge 705 20
B6-Red edge 740 20
B7-Red edge 783 20
Table 2. Sentinel 2 MSI bands along with central wavelength
and resolution
Fig. 1. Spectral signature of in-situ water sample from
Satlantic HyperOCR Radiometer
statistically derived (Table 1).
Model development: TSM
The C retrieval models were
developed by the polynomial regression analysis between in-
situ C and in-situ R (Fig. 3). The band combination of
TSM rs
B3/B7 shows a high correlation with the coefficient of
determination (R = 0.74) and followed by B7/B3 have a R
2 2
value of 0.73. In the best 5 models, the least correlation
occurred in the ratio of B7/B4 with R =0.69 (Fig.3e).
2
In-situ data validation: in-situFor validation, the remaining
30% of the R data is applied to the newly developed in-situ rs
model (Fig. 4). Validation dataset shows the band
combination of B7/B2 has the highest correlation of 79% and
next to that B7/B3 show R = 0.74. The band ratio of B2/B7
2
model gives the poor correlation in the validation graph with
R =0.06. Among all models, band combination of B7/B2 were
2
acceptable for retrieval of C values in the Hooghly River
TSM
(Fig. 4b). To evaluate the accuracy of the developed model,
statistical analysis such as coefficient of determination (R ),
2
root mean square error (RMSE), mean absolute percentage
error (MAPE) were calculated (Siswanto et al 2011).
MAPE= (1/N)* (Σ actual-forecast ctual )*100 …(4)| |/|a |
Satellite data validation: For the satellite validation, scatter
plot between C values derived from the Sentinel 2 band
TSM
ratio (a) B3/B7 (b) B7/B2 (c) B2/B7 (d) B7/B3 (e) B7/B4
against measured C (Fig. 5). The band ratio of B7/B2
TSM
retrieved model is well correlated with measured TSM
concentration with 75% of accuracy. Among the five models,
the band ratio of (a) B3/B7 (b) B7/B2 (d) B7/B3 can be
suitable for monitoring TSM concentration in the Hooghly
161Total Suspended Matter Concentration
Fig. 3. Development of band ratio algorithm between in-situ remote sensing reflectance and in-situ C (a) B3/B7 (b) B7/B2 (C)
TSM
B2/B7 (d) B7/B3 (e) B7/B4
Fig. 4. Scatter plot between estimated and measured C for in-situ validation datasets (a) B3/B7 (b) B7/B2 (C) B2/B7 (d)
TSM
B7/B3 (e) B7/B4
162 R. Premkumar, R. Venkatachalapathy and S. Visweswaran
River. Similar to validation, band ratio of B7/B2 based in-situ
model gives better results in the estimation of C from
TSM
satellite data. The band ratio of B2/B7 model has performed
very low. The C is higher in the northern region and its
TSM
concentration decrease gradually when it comes to the south
part of the Hooghly river (Fig. 6). After the satellite validation,
retrieved TSM concentration shows the clear spatiotemporal
distribution of TSM concentration throughout the Hooghly
River.
The study area belonging to the high suspended
sediment concentration which is recorded in the previous
study by Pitchaikani et al (2019). In the present study the
concentration of total suspended matter in the downstream of
estuarine were 221 to 256 mg l , whereas, near to upstream -1
or port region, a higher concentration is observed in the range
of 280 to 343 mg l . The higher concentration of suspended -1
matter is due to the geographical structure of the area and
gradually decreased in terms of depth and width of the
estuarine region from mouth to upstream. In the study area,
the width of the upstream was 3 km and gradually increased
to 13 km in the mouth region. The study area is located in the
international ship route and it is India's longest river line port.
Due to the turbulence of ship movement, waves and tide
action, sediments are re-suspended in the surface water
which leads to water appear to be in brown colour and the
region became higher sediment concentration. Models
Fig. 5. Scatter plot of satellite derived C against measured CTSMa) B3/B7 (b) B7/B2 (c) B2/B7 (d) B7/B3 (e) B7/B4
TSM
Fig. 6. t s mConcentration of otal uspended atter (C )
TSM
retrieved from Sentinel-2 MSI satellite for the band
combination of B7/B2
163Total Suspended Matter Concentration
based on Sentinel band B1-B5 gives poor correlation. As
compare to higher wavelength in Sentinel-2 band, the lower
wavelength is affected by strong backscattering and
absorption properties of water.
Therefore shorter wavelength bands are not suitable for
retrieval of higher C . The calibration results show that band
TSM
combination of B2/B7 has poor fitting accuracy. This explains
the band ratio of shorter wavelength may be contributed by
phytoplankton absorption at 665 nm, Colour Dissolved
Organic matter observation at 440 nm. However, the lower
concentration of suspended matter can be well correlated
with Sentinel-2 bands (B1-B5) in case 2 water. The higher
band ratio is not possible due to limitation in spectral in-situ
values. The band combination of B7/B2 is recommended for
retrieval of higher TSM concentration. To achieve the better
result, the time difference between the sample collection and
Sentinel 2 satellite pass should be within 3 hours and also
applying the best and adequate atmospheric correction
model of Sentinel 2 data (iCOR) need to be concentrated.
Finally, the Sentinel 2 MSI data shows the high load of TSM
concentration (C ) in the northern part of the study and it's
TSM
gradually decreased when it reaches the near to the Bay of
Bengal.
ACKNOWLEDGEMENT
The research work was funded by the DST, Government
of India under the Network Programme on Imaging
Spectroscopy and Applications (BDID/01/23/2014 – HSRS).
The authors thank to Prof. Bhabani S. Das, Department of
Agricultural and Food Engineering, IIT Kharagpur and his
team Mr.Sourav Roy and Ms. Ojha Suchitra Rani for their
constant support during the field survey.
CONCLUSIONS
The present study monitors the total suspended matter
concentration in the Hooghly River, Kolkata using in-situ
C , in-situ R data and satellite R data. This study helps to
TSM rs rs
identify the spectral signature of TSM concentration with the
help Satlantic HyperOCR radiometer data. The remote
sensing reflectance data is derived from Satlantic
hyperspectral radiometer gives the spectral signature of
TSM concentration. The band combination of B7/B2 in the
in-situ validation data shows the higher validation accuracy
and also the same for satellite validation result shows higher
accuracy. The newly developed regional algorithm for
retrieval of TSM concentration was working well. The
validation result also proved that B7/B2 region belongs to the
higher concentration of total suspended matter. The
developed regional algorithm will help monitor TSM
concentration in the study area.
REFERENCES
Avinash K, Jena B, Vinaya M, Jayappa K, Narayana and Bhat HG
2012. Regionally tuned algorithm to study the seasonal
variation of suspended sediment concentration using IRS-P4
Ocean Colour Monitor data. The Egyptian Journal of Remote
Sensing and Space Science 15(1): 67-81.
Bhaskaran PK, Mangalagiri S and Bonthu S 2014. Dredging
Maintenance Plan for the Kolkata Port, India. Current Science
107: 1125-1136.
Dörnhöfer K, Göritz A, Gege P, Pflug B and Oppelt N 2016. Water
constituents and water depth retrieval from Sentinel-2A-A first
evaluation in an oligotrophic lake. (11): 941. Remote Sensing 8
Gernez P, Lafon V, Lerouxel A, Curti C, Lubac B, Cerisie S and Barillé
L 2015. Toward Sentinel-2 high resolution remote sensing of
suspended particulate matter in very turbid waters: SPOT4
(Take5) Experiment in the Loire and Gironde Estuaries. Remote
Sensing 7(8): 9507-9528.
Keukelaere LD, Sterckx S, Adriaensen S, Knaeps E, Reusen I,
Giardino C and Vaiciute D 2018. Atmospheric correction of
Landsat-8/OLI and Sentinel-2/MSI data using iCOR algorithm:
validation for coastal and inland waters. European Journal of
Remote Sensing 51(1): 525-542.
Kutser T, Paavel B, Verpoorter C, Ligi M, Soomets T, Toming K and
Casal G 2016. Remote sensing of black lakes and using 810 nm
reflectance peak for retrieving water quality parameters of
optically complex waters. (6): 497. Remote Sensing 8
Liu H, Li Q, Shi T, Hu S, Wu G, Zhou Q 2017.Application of Sentinel 2
MSI Images to Retrieve Suspended Particulate Matter
Concentrations in Poyang Lake. : 761.Remote Sensing 9
Malenovský Z, Rott H, Cihlar J, Schaepman M E, García Santos G,
Fernandes R and Berger M 2012. Sentinels for science:
Potential of Sentinel-1, -2, and -3 missions for scientific
observations of ocean, cryosphere, and land. Remote Sensing
of Environment 120: 91-101.
Manzo C, Bresciani M, Giardino C, Braga F and Bassani C 2015.
Sensitivity analysis of a bio-optical model for Italian lakes
focused on Landsat-8, Sentinel-2 and Sentinel-3. European
Journal of Remote Sensing 48(1): 17-32.
Merina RN 2017. Evaluation of sediment using GIS approach in
southern Tamilnadu, India. 179-Indian Journal of Ecology 44:
184.
Minu P, Lotliker AA, Shaju S, Ashraf PM, Kumar TS and
Meenakumari B 2016. Performance of operational satellite bio-
optical algorithms in different water types in the southeastern
Arabian Sea. (4): 317-326. Oceanologia 58
Pitchaikani JS,Ramakrishna R, Bhaskaran PK, Ilangovan D and
Rajawat AS 2019. Development of regional algorithm to
estimate suspended sediment concentration (SSC) based on
the remotely sensed reflectance and field observations for the
hooghly estuary and West Bengal coastal waters. Journal of the
Indian Society of Remote Sensing 47(1): 177-183.
Pradhan Y, Thomaskutty AV, Rajawat AS and Nayak S2005.
Improved regional algorithm to retrieve total suspended
particulate matter using IRS-P4 ocean colour monitor data.
Journal of Optics A: Pure and Applied Optics 7(7): 343-349
Prasad AK and Singh RP 2010. Chlorophyll, calcite and suspended
sediment concentrations in the Bay of Bengal and the Arabian
Sea at the river mouths. (1): Advances in Space Research 45
61-69.
Ramakrishnan D, Bharti R and Das M 2013. A technique for
estimation of suspended sediment concentration in very high
turbid coastal waters: An investigation from Gulf of Cambay,
India. Marine Geology346: 256-261.
Saraswathy R and Pandian PK 2018. Estimation and mapping
chlorophyll-a concentration in Pulicat Lagoon, South India
Using Sentinel 2AR. (3): 533-537. Indian Journal of Ecology 45
164 R. Premkumar, R. Venkatachalapathy and S. Visweswaran
Siswanto E, Tang J, Yamaguchi H, Ahn YH, Ishizaka J, Yoo S and
Kawamura H 2011. Empirical ocean-color algorithms to retrieve
chlorophyll-a, total suspended matter, and colored dissolved
organic matter absorption coefficient in the Yellow and East
China Seas. (5): 627-650. Journal of Oceanography 67
Sravanthi N, Ramana I, Yunus Ali P, Ashraf M, Ali M and Narayana A
2013. An algorithm for estimating suspended sediment
concentrations in the coastal waters of india using remotely
sensed reflectance and its application to coastal environments.
International Journal of Environmental Research 7(4): 841-
850.
Strickland JDH, Parsons TR and Strickland JDH 1972. A Practical
Handbook of Seawater Analysis. Ottawa: Fisheries Research
Board of Canada.
Received 18 September, 2020; Accepted 05 January, 2021
165Total Suspended Matter Concentration