See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/359047710

Evaluation of surface water quality of Ukkadam lake in Coimbatore using UAV

and Sentinel-2 multispectral data

Article in International journal of Environmental Science and Technology · March 2022

DOI: 10.1007/s13762-022-04029-7

CITATIONS READS

12 422

3 authors, including:

Brema Jay Dr. G. Jims John Wessley

Karunya University Karunya University
52 PUBLICATIONS 214 CITATIONS 40 PUBLICATIONS 129 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Brema Jay on 21 July 2023.

The user has requested enhancement of the downloaded file.

International Journal of Environmental Science and Technology
https://doi.org/10.1007/s13762-022-04029-7

ORIGINAL PAPER

Evaluation of surface water quality of Ukkadam lake in Coimbatore

using UAV and Sentinel‑2 multispectral data
T. S. Rahul1 · J. Brema1 · G. Jims John Wessley2

Received: 14 September 2021 / Revised: 5 January 2022 / Accepted: 10 February 2022

© The Author(s) under exclusive licence to Iranian Society of Environmentalists (IRSEN) and Science and Research Branch, Islamic Azad University 2022

Abstract
Remote sensing was used as a potential solution for monitoring the surface water quality parameters as an alternative to the
traditional in situ measurements which are time consuming and labour-intensive. While most of the studies are restricted
in just analysing the optical water quality parameters, only few studies have attempted the estimation of non-optical water
quality parameters. In this paper, a correlation was developed between various optical and non-optical water quality param-
eters, thereby establishing an indirect relationship between non-optical parameters and reflectance data based on which
the predictive models were developed. The water body chosen for this present study is the Ukkadam Lake situated in
Coimbatore, Tamilnadu, India (10.9917° North, 76.9722° East). The correlation between reflectance data obtained from
Sentinel-2 and Unmanned aerial vehicle images along with in situ measured data were analysed using stepwise regression
method. Algorithms were developed for assessing the water quality parameters like turbidity, Total suspended solids, Total
organic carbon, Chemical oxygen demand, Biological oxygen demand and Dissolved oxygen that were based on Sentinel-2
with high coefficient of determination (R2). Unmanned aerial vehicle-based Stepwise regression models were employed for
assessing Total suspended solids, Total organic carbon and Chemical oxygen demand. The developed models were validated
with 25% of sample data acquired, and the algorithms showed that multispectral data from Sentinel-2 and RGB data from
Unmanned aerial vehicles can be effectively used to estimate the concentration of various water quality parameters with
reasonable accuracy in case of large water bodies, including the one chosen for this study.

Keywords Accuracy · Stepwise regression · Surface water quality parameters · Total organic carbon · Total suspended
solids · Unmanned aerial vehicle · Water bodies

Introduction Moreover, the climatic changes due to global warming cause

drought, flood and noticeable increase in infectious diseases
Among the earth’s renewable resources, freshwater sys- leading to further degradation in the water quality. Hence,
tems are the most endangered habitats in recent years. More from the start of the twentieth century, surface water quality
than half of the world’s wetlands have disappeared from analysis in water bodies was observed as a crucial measure
the earth’s surface in the past decade with steep decline in with stringent and precise control of industrial discharges
water quality due to rapid urbanization and industrialization. and chemicals through sewage (Bourouhou and Salmoun
2021), (Khaki 2020). Even though the traditional in situ
measurements and laboratory testing offer higher accuracy
Editorial responsibility: J Aravind. in water quality assessment, its feasibility is typically low in
providing constant water quality monitoring database based
* T. S. Rahul on regional scale due to enormous time consumption and
[email protected]
labour intensiveness (Vercruysse et al. 2020). Remote sens-
1
Department of Civil Engineering, Karunya Institute ing, with the integration of remotely sensed data, together
of Technology and Sciences, Karunya Nagar, Coimbatore, with technologies like Geographic information system (GIS)
Tamil Nadu, India and Global positioning system (GPS) was employed as an
2
Department of Aerospace Engineering, Karunya Institute alternative solution for monitoring the surface water quality
of Technology and Sciences, Karunya Nagar, Coimbatore, parameters (SWQPs) and to enhance different management
Tamil Nadu, India

13
Vol.:(0123456789)
 International Journal of Environmental Science and Technology

plans for various natural resources (El Mountassir et al. Another research conducted by Abayazid and El-Adawy
2020). (2019) over the Egyptian Coastal Lake presented with an
However, the application of remote sensing for monitor- approach for deriving the Dissolved oxygen (DO) by com-
ing small water bodies incurs some serious challenges. Due parative analysis of multiple regression algorithms which
to inappropriate satellite sensors, satellite monitoring was were carried out using several combinations of optical
hampered to a greater extent (Tóth et al. 2021). The ocean parameters namely Turbidity, Chlorophyll-a, TSS and
sensors such as medium resolution imaging spectrometer Temperature. The results showed successful statistically
(MERIS) and moderate resolution imaging spectroradi- significant correlation in certain combinations considered,
ometer (MODIS) with frequent re-visit time of 1–3 days, and the optimal results were established based on turbidity
and radiometric resolution of 12 bits were not suitable in and natural logarithm of temperature with high R2 value.
remote monitoring in small water bodies due to their lower In a similar study by Song et al. (2011), over Central
spatial resolution (300–1000 m) (Verpoorter et al. 2014). Indiana Water Supply Reservoirs estimated the Total phos-
This inadequacy was made to overcome till some extent with phorus (TP) using hyperspectral remote sensing owing to
the advent of Sentinel-2 Multispectral instrument (MSI) its close association with Chl-a, Secchi disc transparency
and Landsat 8 Operational land imager (OLI) sensors. The (SDT), turbidity and total suspended matter. Correlation
availability of Unmanned aerial vehicles (UAVs) or drones analysis was performed to determine the sensitive spec-
which can be loaded with multispectral and hyperspectral tral variables including TP, Chl-a, and SDT. The results
sensors is another competitive method of remote sensing indicated that TP had closer association with diagnostic
in small water bodies with high-resolution images captured spectral variables with ­R2 ranging from 0.55 to 0.72.
at very low altitudes over the water bodies (Cillero Castro Sharaf and Zhang (2017) also attempted to develop
et al. 2020). Recent studies showed that, UAVs could be uti- an integrated Landsat 8 band ratios and stepwise regres-
lized in acquiring periodically monitored environmental data sion model to estimate the concentrations of non-optical
by capturing high temporal and spatial resolution images SWQPs such as chemical oxygen demand, biochemical
allowing decision-making in hours rather than weeks with oxygen demand and dissolved oxygen by correlating with
major reduction in manpower and operational cost (Tsouros optical variables such as turbidity and total suspended sol-
et al. 2019). A considerable amount of studies were reported ids. The study was carried out at Saint John River, Canada
on the measurement of water quality using indicators like which finally showed better correlations between Landsat
turbidity, Secchi depth, Coloured dissolved organic mat- 8 surface reflectance and concentrations of SWQPs with a
ter (CDOM) and chlorophyll measured using remote sens- coefficient of determination equal to 0.85.
ing techniques (Papenfus et al. 2020), (Minu et al. 2020), Arango and Nairn (2019) developed statistical water
(Alvado et al. 2021), (Mishra et al. 2019), (Gitelson et al. quality models that possess the capability in reliable esti-
1993), (Vertucci et al. 1989). There are few findings reported mation of optical water quality parameters like Total sus-
where researchers have attempted to estimate the concentra- pended solids (TSS), Chl-a and Secchi disc depth (SDD)
tion of non-optical water quality parameters such as total with respect to non-optical water quality parameters like
nitrogen (Kapalanga et al. 2021), total phosphorus (Xiong Total phosphorous (TP) and Total nitrogen (TN) in oli-
et al. 2019), chemical oxygen demand (Yang et al. 2011), gotrophic and eutrophic systems with the help of remote
etc. sensing images that were collected using a multispectral
However, the concentration of non-optical surface water sensor. The study concluded that, with the employment
quality parameters could be correlated with optical param- of statistical water quality assessment and multiple linear
eters such as total suspended solids that have the potential in regression models, the developed models were effective in
affecting water colour, therefore will be able to be detected predicting the optical and non-optical water quality param-
by the satellite sensors. Some of the recent case studies and eters with increased prediction capability ­R2 greater than
models developed by utilizing these correlations were dis- 0.80.
cussed below. A study carried out by Yu et al. (2016) at Hasab et al. (2020) carried out the assessment of sur-
Bohai Sea, China correlated the dissolved inorganic nitro- face water quality parameters in Al-Hawizeh Marsh, Iraq
gen (DIN), which is considered as a non-optically active on the basis of Remote sensing and GIS-based approaches.
compound in water with optical parameters such as Chlo- The Landsat 8 images were employed for predicting and
rophyll-a (Chl-a) concentration and Total suspended solids assessing the spatial variations together with map dis-
(TSS) values to generate a DIN estimation model. Here, the tribution with respect to salinity, sulphate and calcium
MODIS hyperspectral satellite images were used for the esti- carbonate inside the Al-Hawizeh Marsh for duration of
mation process. Correlation and regression analysis showed two seasons on the basis of mathematical modelling. The
a highly significant positive relationship between DIN con- developed model for evaluating the salinity, sulphate and
centration and various band combinations. calcium carbonate showed the observations to be lowest

13
International Journal of Environmental Science and Technology

during winter season but highest during autumn season. second series of sampling was carried out on 19th March
The observed values of correlation coefficient (­ R2) cor- 2021.
responding to the real data with respect to mathematical
models during the two seasons were 0.95, 0.96 and 0.92,
respectively. Materials and methods
Mihu-Pintilie (2018) carried out a physico-chemical
examination in Cuejdel lake basin that was realized on the Study site and water quality analysis
basis of the monitoring programme that constituted the
analysis of seasonal variations of thirteen leading physico- Ukkadam Lake is situated in Ukkadam, Coimbatore, Tamil-
chemical parameters. Around thirty samples are taken for nadu, South India. It lies between 10.9917° N latitude and
consideration in both spring season and autumn season. Cor- 76.9722° E longitude. The lake occupies an area of around
responding readings were acquired from both in situ aquatic 1.295 ­km2 with an average depth of 5.82 m. The lake con-
ecosystem and from the laboratory. The findings revealed sists of a total volume of 1,982,179.262 ­m3 of water (https://​
that the degree of contamination by phosphorus and nitro- en.​wikip​edia.​org/​wiki/​ukkad​am_​lake).
gen compounds including nitrates, nitrites, phosphates, etc., The canals diverted from Noyyal river supply water to
were extremely lower and the value of nutrients revealed the the Ukkadam Lake. Another lake, Selvachinthamani, located
natural function of the limnosystem. upstream in the north also provides sufficient water to the
It is evident from the study carried out in the Poyang Lake lake. The four sluice gates located at the south side of the
in China using empirical models, Sentinel-2A MSI band 7 is lake release water to external use. The lake becomes pol-
suitable for sediment-laden waters, while Sentinel-2A MSI luted with effluents of sewage released from surrounding
band 4 is suitable for estimating the TSS in clear water (Liu areas and is also encroached by water hyacinth to a consid-
et al. 2017). In a similar study by Toming et al. (2016) and erable area. Some of the recent studies also highlighted the
Caballero et al. (2019), the dependence of the concentra- increasing pollution rate of this lake. A study carried out by
tion of turbidity or suspended sediments on red and near Jeyaraj et al. (2016) revealed that the water quality of this
infrared (NIR) regions was established. In another study by lake was found to be polluted with reference to almost all
Alparslan et al. (2007), the blue, red, green, and NIR bands the water quality parameters. Moreover, the water remains
of Landsat-7 ETM were used in effectively monitoring the as alkaline nature with the total concentration of total dis-
TSS concentrations of the reservoir behind Omerli Dam, solved solids (TDS) and turbidity to exceed the permissible
Ustanbul. Hellweger et al. (2004) conducted a study with limit, thereby concluding with the ultimate root cause of
Landsat-5 TM images using regression models developed to pollution as domestic and industrial discharges. Further-
measure the turbidity levels in the New York Harbour and more, another study carried out by Gowraraju et al. (2014)
found that the turbidity levels at the river run-off areas were estimated the water quality index of this particular lake by
correlated to the red band with coefficient of determination National sanitation foundation—Water quality index (NSF-
(R2) = 0.78. Landsat -7/TM imagery over reservoirs of Shen- WQI) method. Here, all the measured parameters were found
zhen, China showed higher correlation coefficient of 0.626 to be very much higher when compared to the permissible
between chemical oxygen demand (COD) concentration and limit prescribed by the World health organization (WHO)
reflectance values of band 1–3 by multiple linear regression and thereby seems to be unhealthy for public use.
approaches (Wang 2004). In order to de-silt the lake and clear the encroachments,
The present study intends to develop the correlations the Coimbatore corporation unveiled a plan in 2010 and sub-
between in situ data and reflectance data of various water sequently conducted a series of studies. However, the study
quality parameters by regression-based techniques. Based conducted in this paper is the first study intending to develop
on the available literature studies, stepwise regression was image processing algorithms for estimation of water quality
found to be most adequate for this study. The present study parameters based on spectral characteristics of Sentinel-2
aims to (1) substantiate the application of Sentinel and UAV satellites and UAVs in Ukkadam lake. The satellite view of
multispectral characteristics for analysing the water quality, the Ukkadam lake is shown in Fig. 1.
and (2) develop algorithms to estimate the concentration of The study can be basically classified as remote sensing
both optical and non-optical SWQPs using Stepwise regres- analysis and water sampling analysis. The same is repre-
sion (SWR) technique from UAV and Sentinel-2 reflectance sented as a flow diagram in Fig. 2.
data. The water body chosen for this present study is the The operational features of the flow diagram in this study
Ukkadam Lake situated in Coimbatore, Tamilnadu, India could be revealed using the following considerations:
(10.9917° North, 76.9722° East). The water samples were
collected during the months of January and March 2021. The
first in situ sampling was done on 03rd January 2021 and

13
 International Journal of Environmental Science and Technology

Fig. 1  Sentinel-2 image of

Ukkadam Lake

(1) This proposed research methodology could be basically (7) From these images, the reflectance values of various
classified as remote sensing analysis and water sam- water sampling points were extracted using SNAP and
pling analysis. ENVI software.
(2) For remote sensing analysis, the data from two sources (8) The water samples from the Ukkadam Lake were col-
were mainly acquired, which includes Level-1C (L1C) lected at 16 strategic locations on the same day in which
Sentinel-2 satellite images and UAV drone images. the satellite data and the UAV data were accessed and
(3) The 13 bands of Sentinel-2 images were re-sampled by finally tested in the laboratory to obtain values for vari-
nearest neighbour method to bring them to a uniform ous surface water quality parameters.
resolution of 10 m and the area of interest was deter- (9) Finally, the validation of the developed stepwise regres-
mined. sion models is done using 25% of in situ sampling
(4) The subset of study was subsequently corrected to bot- results and accurate spatial distribution maps were
tom of atmospheric (BOA) reflectance using the Case 2 generated accordingly.
regional coast colour (C2RCC) atmospheric correction
processor in SNAP 8.0 software to obtain eight atmos-
pherically corrected angular dependent water leaving Remote sensing analysis
reflectance (rhow) bands.
(5) Similarly, in case of UAV image capturing, the flight For remote sensing analysis, two methods were adopted in
mission was planned using drone deploy software at this study. The Sentinel-2 data were obtained from the Euro-
a flight height of 300 feet so as to obtain 755 images pean space agency (ESA) Copernicus Open Access hub.
which were finally sewed up together to obtain Ortho- The 13 spectral bands of Sentinel-2 range from the Visible
mosaic image of high spatial resolution up to 3 cm. (VNIR) and Near-infrared (NIR) to the Short wave infrared
(6) The geometric corrections, correction for lens distor- (SWIR). The multispectral bands of Sentinel-2 data and its
tions and camera tilts were also carried out with this corresponding spatial resolution are displayed in Table 1.
software to finally attain an image of study area consist-
ing of three bands.

13
International Journal of Environmental Science and Technology

Sentinel-2 MSI Water Sampling UAV: DJI Mavic 2 Pro

Downloading cloud free L1C

images from Copernicus Hub
UAV flight: multispectral
images by FC 220 camera

Resampling bands to 10 m
resolution & creating subset of Simultaneous
the study area
In-situ water

Sample collection
Georeferencing
& Mosaicking
Atmospheric with Drone
Correction by deploy
C2RCC
Processor

Testing of
Samples for
Various
SWQPs

8 rhow Bands 3 Bands

Extract water leaving Extract reflectance data from

reflectance data for various Preparation of data for Sampling Points by Envi
sample points using SNAP calibration & validation Software
software

Model Calibration & algorithm development for various SWQPs

Model Validation

Study of spatial distribution of WQ parameters

Fig. 2  Flow chart of methodology in the present study

13
 International Journal of Environmental Science and Technology

Table 1  Sentinel-2 band details processor in SNAP 8.0 software. It is a full spectrum ver-
Band number Central wavelength (nm) Spatial sion using a set of neural networks which were trained
Resolution for simulated top-of-atmosphere reflectance (https://​sea-
(m) das.​gsfc.​nasa.​gov/​help/8.​0.0/​c2rcc/​C2RCC Tool. html).
C2RCC performs best for water bodies, achieving highest
1 443 60
coefficient of determination and lowest root mean square
2 490 10
deviation (RMSD) when compared with other processors
3 560 10
like Acolite, l2gen, sen2cor, etc., (Warren et al. 2019).
4 665 10
Finally, the atmospherically corrected angular-dependent
5 705 20
water leaving reflectance (rhow) for various bands in the
6 740 20
area of interest was extracted. The extraction of pixel
7 783 20
information corresponding to the sampling points was
8 842 10
done using ENVI and SNAP software. The atmospheri-
8a 865 20
cally corrected image after C2RCC atmospheric process-
9 945 60
ing is shown in Fig. 3.
10 1375 60
11 1610 20
12 2190 20

Unmanned aerial vehicle imagery analysis

Pre‑processing of Sentinel‑2 satellite imagery For unmanned aerial vehicle (UAV) image capturing, a
commercially available DJI Mavic 2 Pro Quadcopter drone
The Sentinel-2 imageries that are ortho-rectified, geo- with FC220 RGB camera was used. The flight mission was
located and radiometrically calibrated top-of-atmosphere planned using drone deploy software at a flight height of
(TOA) reflectance in Universal transverse mercator (UTM) 300 feet to obtain the images of high spatial resolution
projection with the WGS84 datum with less cloud cover up to 3 cm. The images were captured on the same day in
were selected for this analysis. Since the 13 bands of Sen- which the Sentinel satellite passes over the study site, and
tinel-2 images were not of the same resolution, the images the corresponding imageries were chosen for this analysis.
were re-sampled by nearest neighbour method to bring them On the day of analysis, a total number of 755 images were
to a uniform resolution of 10 m and the area of interest is captured using the automated flight planning platform. The
determined. equivalent features of DJI Mavic 2 Pro Quadcopter drone
The varying atmospheric conditions and degradation used for this study are detailed in Table 2.
of accuracy due to surface reflection adversely affect the
remote estimation of lake water quality (Koponen et al.
2002). Hence, the S2 MSI images downloaded from Pre‑processing of Unmanned aerial vehicle imagery
Copernicus Hub-ESA, were subsequently corrected to
Bottom of atmospheric (BOA) reflectance using the Case The images captured using the UAV are combined together
2 regional coast colour (C2RCC) atmospheric correction to obtain an ortho mosaic map using the Drone deploy

Table 2  Features of DJI Mavic 2 Pro Quadcopter drone

S. No UAV Specifications
1 Model DJI Mavic Pro Platinum (M1X)

2 Maximum speed 40 mph

3 Maximum flight time 30 min
4 Stabilization 3-axis (pitch, roll, yaw)
5 Camera sensor 1/2.3″ (CMOS)
6 Effective pixels 12.35 M
7 Total pixels 12.71 M
8 Operating temperature 32° F—104° F (0 °C to 40 °C)
9 Navigation GPS & Vision position-based
Fig. 3  Atmospherically corrected image after C2RCC atmospheric
navigation
processing

13
International Journal of Environmental Science and Technology

Fig. 4  Ortho mosaic imagery of

Ukkadam Lake (3 cm resolu-
tion)

Flight Planning: Drone Deploy Image Processing

software. Geometric corrections, correction for lens dis- regression (SWR) technique is the most suitable option for
tortions and camera tilts were also carried out with this this study, as the variables are quantitative, which means the
software. As the flight altitude is 300 feets, the images water quality parameters are measurable and the independent
captured were less influenced by the atmospheric effects. variables (surface reflectance values of bands/band ratios)
Hence, atmospheric correction was not applied to these are not highly correlated with each other.
UAV images, and finally, the raw digital numbers were In this study, the scope of using Sentinel-2/UAV reflec-
extracted using ENVI software for further calculations. tance values in predicting water quality parameters was
The different steps included in ortho mosaic imagery gen- identified using this SWR technique. The sampling points
erating process using Drone deploy software were shown were divided into two datasets: 75% of the total sample
in Fig. 4.

Estimation of concentrations of surface

water quality parameters

Regression analysis is a proven predictive modelling tech-

nique which correlates the reliance of a dependent variable
to a group of independent variables. It is a process to analyse
the dependence of variables that were not treated to have
symmetrical correlation.
Here, the aim is to obtain a prediction of one variable,
given the values of the others (Bartholomew et al. 2010).
Several types of regression techniques are available like
logistic regression linear regression, ridge regression, Fig. 5  Sentinel-2 image of Ukkadam Lake with 16 water sampling
polynomial regression and stepwise regression. Stepwise points

13
 International Journal of Environmental Science and Technology

was used for standardization, while the remaining 25% of were observed in between the time span of field water sam-
samples were implemented to validate the models. The pling and satellite overpassing, which might cause altera-
performance of the developed models was evaluated by tions in the values of water quality parameters.
using regression lines’ equations, ­R 2 and significant P The water samples which were stored in a dark con-
value. The model categories were identified as follows tainer with cooler were subsequently taken to laboratory
(Chang et al. 2001): analysis for the estimation of TDS, Potential of hydrogen
(pH), Electrical conductivity (EC), Biological oxygen
• Category 1: Accurate prediction if 0.80 ≤ R2 ≤ 1.00 demand (BOD), COD, DO, Total organic carbon (TOC),
• Category 2: Satisfactory prediction if 0.50 ≤ R2 < 0.80 TSS, Nitrate and Turbidity. The BOD, COD & TOC tests
• Category 3: Unacceptable prediction if R2 < 0.50 were carried out according to IS3025; Bureau of Indian
Standards. Here, the BOD results were obtained after
3 days (https://​law.​resou​rce.​org/​pub/​in/​bis/​S02/​is.​3025.​
Water sampling and laboratory analysis 44.​1993.​pdf) of incubation procedure & COD by titration
method (https://​law.​resou​rce.​org/​pub/​in/​bis/​S02/​is.​3025.​
The water samples from the Ukkadam Lake were collected 58.​2006.​pdf). Also, the instruments including water ana-
from 16 strategic locations on the same day in which the lyser 371 were used for testing pH, EC, TDS, DO, and
satellite data and the UAV data were accessed from the DR/890 colorimeter was used for analysing TSS and tur-
study site. Figure 5 shows the Sentinel 2 satellite image of bidity in this analysis.
Ukkadam Lake with various water sampling points. These parameters were particularly selected for this
The samples were collected during the months of Janu- study based on the codal provisions and guidelines pub-
ary and March of the year 2021. The first in situ sampling lished by the World Health Organization and Bureau of
was done on 03rd January 2021, and the second series Indian Standards corresponding to water quality assess-
of sampling was carried out on 19th March 2021. The ment. A turbidity meter denotes the turbidity based on
sampling methodology from a water body mainly depends the amount of light scattered by the particles in the water
upon various factors like size of the water body, depth of column. Testing for COD helps to quantify the amount of
water column and the flow rate (water quality sampling organics present in a water sample. A strong oxidant was
techniques (usgs.gov)). The water bodies which are shal- used to react with all the organic matters in the sample.
low and well mixed, only requires surface (0–1 m) water BOD is the measure of dissolved oxygen required by the
sampling (field sampling programme (water quality (gov. aerobic organisms present in the water body for the break-
au)). Here, in this lake, as most of the water sampling posi- down of organic materials present in 5 days at a tempera-
tions were shallow in depth, the lake water will be in well ture of 20 °C. Also, the level of dissolved oxygen (DO),
mixed-up condition. From the boat, the sampler bottle was which is an indicator of free oxygen present in the water
dipped by hands to just below the water surface (typically samples, was estimated. TOC, which indicates the amount
0.30 m–0.50 m depth) and by holding the sampler bottle of carbon found in an organic compound is mainly a non-
in downstream position. Moreover, as the hands were cov- specific indicator of water quality, was measured with a
ered with a plastic disposable glove, the contaminations total organic carbon analyser.
from surface films were avoided. Whatever vessel was
being used for water sample collection, it should possess
the ability to face the water currents, therefore, the water
Table 3  SWQP results from the study
samples were taken from the front of the vessel, preferably
while moving slowly in forward direction to minimize the SWQP January and March 2021
contamination caused by the boat itself (Apte et al. 1998). Mean Max Min Std. Dev
In this study, all the water sampling points were taken
Turbidity (NTU) 55.08 150 6.2 62.13
only from the clear surface water area, and also the por-
TSS (mg/L) 357.92 624 200 162.09
tions covered by aquatic vegetation were not taken into
DO (mg/L) 4.49 5.8 3 0.79
consideration as the vegetation portion will restrict the
BOD (mg/L) 46.42 92 17 23.23
true reflectance of the water. Furthermore, the sampling
COD (mg/L) 94.25 150 48 27.90
positions were precisely determined by a hand-held GPS
Nitrate (mg/L) 19.73 27 12 5.18
receiver, obtaining the exact co-ordinate values. The sam-
TOC (mg/L) 34.50 56 23 9.26
pling points were mainly selected along the vicinity of the
TDS (mg/L) 889.58 1,196 600 194.02
outfall as all the untreated household as well as industrial
pH 8.50 8.80 8 0.24
sewages might accumulate around these locations making
EC (µs/cm) 724.08 757 700 18.79
it a pollutant hotspot. Moreover, no weather perturbations

13
International Journal of Environmental Science and Technology

Results and discussion corresponds to the values of various surface water quality
parameters (dependent variable) and X axis represents the
The spectral data from Sentinel-2 satellite and UAV were reflectance values of the bands which showed the highest
collected from the test site and processed to extract the correlation with the parameter (independent variable). Here,
concentrations of both optical and non-optical SWQPs. the regression line drawn after plotting the scatter diagram
The key findings of this study were discussed in the sub- represents the relationship or correlation between the two
sequent sections. variables. The direction of the slope of the line specifies the
correlation as positive or negative; an upward sloping line
Concentration of surface water quality parameters indicates the positive correlation and a downward slope rep-
resents the negative correlation between the datasets. And
The summary of water analysis results corresponding to the finally, the equation given in each case represents the equa-
in situ samples which were collected from 16 points syn- tion of the line plotted.
chronous with the UAV flights and Satellite overpass are Figure 6 shows the correlation between sentinel-2 rhow
given in Table 3. Here, even though 10 parameters were bands/band ratios with in situ testing results of various water
tested and analysed, due to low correlation and coefficient quality parameters. All the water quality parameters cho-
of determination values, the parameters such as TDS, pH sen in this study have shown significant correlations with
and EC were eliminated from further algorithm develop- coefficient of determination (R2) greater than 0.5, and with
ment stages. p values less than 0.05. Sentinel-2 bands such as coastal
aerosol (B1), blue (B2), red (B4), red edge (B5, B6, B7) and
Correlation of optical and non‑optical surface water NIR (B8A) have contributed in the development of algo-
quality parameters rithm models for various surface water quality parameters.
Furthermore, the band ratios were much helpful in model
The non-optical parameters which were considered in this development due to its ability to enhance the spectral con-
study showed positive correlation with the optical param- trast between various targets. C2RCC processor generates
eters. Most of the non-optical parameters like DO, BOD and rhow bands, i.e. atmospherically corrected angular depend-
COD showed a high level of correlation with turbidity which ent water leaving reflectance which plays an integral part
is an optical parameter, while TOC correlates with TSS as in developing algorithms for water quality parameters with
shown in Table 4. direct band correlations.
Thus, even though it was believed that non-optical param- It is evident from Fig. 6e that TSS shows strong correla-
eters do not directly affect the light signals measured by tion with red edge bands and NIR bands with R2 value of
satellite sensors, it could be seen that the optical parameters 0.727 and p value of 0.001, respectively. The optical param-
affect the light reflectance. An indirect relationship can thus eter turbidity shows good correlation with blue, coastal aero-
be established between non-optical water quality parameters sol and red edge bands with R2 value of 0.777 and p value of
and light reflectance data, based on which the estimation 0.015, respectively. These results are found to be matching
models were developed. with the earlier literature reported by Moore (1980), Pavel-
sky and Smith (2009), and Gholizadeh et al. (2016), where
Evaluation of relationship between Sentinel‑2 it was seen that the reflectance in the red region increases
and surface water quality parameters with increase in turbidity of water. Non-optical water qual-
ity parameters like BOD, COD and DO showed strong cor-
A scatter plot diagram was used to summarize the rela- relations with blue band (B2) with high R2 values. This is
tionship between two sets of data. The Y axis of the graph in agreement with the earlier studies by Mushtaq and Nee
(2016) that the non-optical water quality parameters like

Table 4  Correlation matrix Parameter Optical Non-Optical

of optical and non-optical
parameters Turbidity TSS DO BOD COD TOC

Optical Turbidity 1 0.403 − 0.879 0.951 0.821 0.430

TSS 0.403 1 − 0.395 0.395 0.507 0.834
Non-Optical DO − 0.879 − 0.395 1 − 0.965 − 0.934 − 0.532
BOD 0.951 0.395 − 0.965 1 .908 0.503
COD 0.821 0.507 − 0.934 .908 1 0.66
TOC 0.430 0.834 − 0.532 0.503 0.66 1

13
 International Journal of Environmental Science and Technology

BOD = 119.508-7420.376*B2 (R2 = 0.619; P - Value = 0.004) DO = 1.827+269.029*B2 (R2 = 0.719; P - Value = 0.001)

(d)
(a)
TSS = 8133.146-11002.901*B7/(B6+B8A) (R2 = 0.727; P - Value = 0.001)
COD = 197.342-10503.715*B2 (R2 = 0.854; P - Value < 0.001)

(b) (e)

TOC = -28.793+18.898*B5/B6 (R2 = 0.577; P - Value = 0.007)

TB = 409.842-22011.709*B2-108.005*B2/(B1+B6) (R2 = 0.777; P - Value = 0.015)

(c) (f)

Fig. 6  Sentinel-2 estimation models for a BOD, b COD, c TOC, d DO, e TSS, and fTurbidity using the SWR technique based on calibration
dataset

COD were strongly correlated with the single OLI band 2 spectral resolution (3 bands) of UAV, limited correlations
(Blue) of Landsat 8. were obtained with respect to various SWQPs. TSS showed
strong correlation with red, green, blue bands with R2 value
Evaluation of relationship between surface water equals to 0.779 and significance (p) value of 0.002. Similarly,
quality parameters using dataset from Unmanned the COD and TOC also showed good correlations with red
aerial vehicle and blue bands with R2 value of 0.556 and an optimal p value.
Only a few studies were carried out in analysing the
Figure 7 shows the correlation between various bands of SWQPs using UAV with a majority of research works
UAV with in situ testing results of SWQPs. Due to the low that were conducted over optical water quality parameter

13
International Journal of Environmental Science and Technology

TSS=8535.594*B3/(B1+B2)-3602.206 (R2 = 0.779; P - Value = 0.002) Validation of Sentinel‑2 and Unmanned aerial
vehicle‑based stepwise regression models

The surface reflectance data from Sentinel-2 and UAV

together with 75% of in situ sampling test results were cor-
related to generate various algorithms of SWQPs using step-
wise regression technique. The validation of the developed
SWR models was done using 25% of in situ sampling results.
Using the algorithms, SWQPs were predicted for various
sampling points and validated with the actual laboratory
results.
In case of Sentinel-2 based SWR models, strong correla-
(a) tion was established between the predicted and measured
surface water quality parameter values with coefficient of
COD= -443.832+569.964*B3/B1 (R2 = 0.556; P - Value = 0.021)
determination (R2) more than 0.8. Table 5 shows the vali-
dation data of sentinel-2 band using stepwise regression
models. Table 6 shows the validation of predicted SWQPs
from UAV-based SWR models, and the measured test results
which showed good correlation with R2 > 0.80 thereby vali-
dating this present study.

Spatial distribution maps of Sentinel‑2

and Unmanned aerial vehicle based on stepwise
regression models
(b) With the help of SNAP software, the algorithms derived
from stepwise regression models were applied to every
TOC= 265.732-217.790*B1/B3 (R2 = 0.556; P - Value = 0.021)
pixel of the Sentinel-2 and UAV images to develop accurate
spatial distribution maps for various surface water quality
parameters over Ukkadam Lake.
Figures 8 and 9 show the respective spatial distribution
maps of sentinel-2 and UAV-based SWR models. It was
observed from the above figures that DO varies between
2.09 mg/l and 6.93 mg/l and in most of the points, the DO
value is between 4 and 5 mg/l, from which it was evident
that the aquatic life is partially strained. The BOD and
COD values were observed to be high throughout the lake
which shows clear evidence of industrial pollution. Another
important parameter considered, TOC, was seen to vary
(c) between 20 mg/l and 45 mg/l. As TOC is the sum of Dis-
solved organic carbon (DOC) and Particulate organic carbon
Fig. 7  UAV estimation models for a TSS b COD c TOC (POC) present in the water, its increased value indicates high
rate of organic material decomposition (Richard et al. 2000),
(https://​www.​elgal​abwat​er.​com/​blog/ total-organic-carbon-
studies. The novelty of the present study lies in establish-
toc). It was also seen that, the turbidity and TSS values were
ing significant correlations of algorithmic models with
optimal in some of the sampling points while showing peak
reasonable acceptance limits, proving the capability of
values in those parts of the lake with inlets from industries
data obtained using UAVs in non-optical surface water
and construction sites.
quality determination.

13
 International Journal of Environmental Science and Technology

Table 5  Validation of Sentinel 2 data using stepwise regression models

Algorithms SWQP Point 13 Point 14 Point 15 Point 16 Validation
Correlation
(R2)

119.508–7420.376*B2 BOD Predicted 63.80 74.17 72.81 18.27 0.821

Measured 55.00 62.00 81.00 28.00
197.342–10,503.715*B2 COD Predicted 118.48 133.17 131.24 54.04 0.896
Measured 126.00 115.00 122.00 67.00
1.827 + 269.029*B2 DO Predicted 3.85 3.47 3.52 5.50 0.908
Measured 4.20 3.50 2.90 5.80
-28.793 + 18.898*B5/B6 TOC Predicted 41.30 40.58 36.87 27.77 0.868
Measured 37.00 44.00 32.00 21.00
8133.146–11,002.901*B7/ (B6 + B8A) TSS Predicted 479.23 467.60 404.27 216.16 0.81
Measured 420.00 388.00 448.00 184.00
409.842–22,011.709*B2-108.005*B2/(B1 + B6) Turbidity Predicted 88.12 120.30 127.50 3.21 0.853
Measured 107.00 94.00 153.00 9.75

Table 6  Validation of UAV data Algorithms SWQP Point 13 Point 14 Point 15 Validation
based on stepwise regression Correlation
models (R2)

8535.594*B3/(B1 + B2) TSS Predicted 330 280 510 0.86

− 3602.206
Measured 420 388 448
-443.832 + 569.964*B3/B1 COD Predicted 110 92 112 0.804
Measured 126.00 115.00 122.00
265.732–217.790*B1/B3 TOC Predicted 35.02 29.45 47.94 0.88
Measured 37 44 32

Discussions on the findings containing higher concentration of organic matter resulted

in increased microbiological activity, which led to greater
In this paper, the reflectance values from both Sentinel-2 and oxygen depletion. COD is likewise a measure of organic
UAV were analysed for obtaining the correlations between matter and oxygen demand, where the depletion of oxygen
band values or combination of band values with various sur- demand is through chemical oxidation and non-biological
face water quality parameters. The surface reflectance data activity. TOC on the other hand measures the amount of
generated from Sentinel-2 satellite imagery with C2RCC carbon resulted from the generation of C ­ O2 because of oxi-
atmospheric correction give specific water leaving reflec- dation reaction. On analysing the test results of various water
tance as output. Furthermore, the wide spectral resolution quality parameters as shown in Table 4, it could be clearly
of the satellite imagery has stimulated the formation of algo- inferred that non-optical parameters like BOD and COD can
rithms to predict SWQPs like COD, BOD, DO, TSS, TOC be correlated to optical parameter like turbidity, and TOC
and Turbidity with reasonably increased correlations when is non-optical with TSS in this present study area. This is
compared to UAV imagery. further evident in the algorithms generated from Sentinel-2
In this present study, BOD was tested in laboratory satellite. Here, the water quality parameters like BOD, COD
after the incubation of samples with microbes, measuring and Turbidity can be determined from the blue band of Sen-
the depletion of oxygen directly resulting from biological tinel-2, while the parameters TOC and TSS can be deter-
activity aided by organic matter in the samples. Samples mined from the red edge bands.

13
International Journal of Environmental Science and Technology

(a) Spatial map of DO (b) Spatial map of BOD (c) Spatial map of TOC

(d) Spatial map of Turbidity (e) Spatial map of COD (f) Spatial map of TSS

Fig. 8  Spatial distribution maps of Sentinel-2 based SWR models

(a) Spatial map of COD (b) Spatial map of TOC (c) Spatial map of TSS

Fig. 9  Spatial distribution maps of UAV-based SWR models

Considering most of the practical real-world applications, available and local authorities could use these developed
the traditional in situ measurement and laboratory testing models to retrieve various surface water quality parameters
for analysing the water quality are both time consuming of Ukkadam Lake in periods with weather conditions simi-
and expensive. SWR models from Sentinel/UAV possess lar to those used for the generation of model and with rea-
increased potential to play pivotal role in continuous moni- sonable accuracy together with reduced cost and time. The
toring of water quality and to take appropriate action plans correction tools available and widespread spectral resolu-
accordingly. However, the remote sensing data are readily tion thus play a very important role in the formulation of

13
 International Journal of Environmental Science and Technology

algorithmic models for retrieving the water quality param- Funding This research did not receive any specific grant from funding
eters. The conventional models employing UAV imagery agencies in the public, commercial, or not-for-profit sectors.
even though being more accessible, they face the shortcom-
Availability of data and material Authors are willing to share data and
ing of reduced spectral resolution in UAV. Furthermore, the material according to the relevant needs.
non-availability of a processor and necessity for generating
a specific water reflectance have limited the development of Declarations
predictive algorithms for acquiring additional water quality
parameters. Conflict of interest The authors declare no competing financial, pro-
Conversely, for future research, in order to additionally fessional and personal interests.
validate the applicability of this developed Sentinel-2 and
Consent for publication We declare that we consented for the publica-
UAV-based SWR algorithms, further studies are required tion of this research work.
for the analysis of these selected water quality parameters
at various seasons together with different water sampling
depths as well. The related consequences towards data inter-
operability, machine learning models and data sharing for
reinforcement of interactions amid UAVs and satellites will References
also need to be assessed. Additionally, generating Sentinel-
Abayazid H, El-Adawy A (2019) Assessment of a non-optical water
2-based water quality index models have to be explored quality property using space-based imagery in Egyptian Coastal
which will assist in summarizing the complicated water Lake. J Water Resour Prot 11:713–727. https://​doi.​org/​10.​4236/​
quality data into simplified numerical values. jwarp.​2019.​116042
Alparslan E, Aydöner C, Tufekci V, Tüfekci H (2007) Water qual-
ity assessment at Ömerli Dam using remote sensing techniques.
Environ Monit Assess 135:391–398. https://​doi.​org/​10.​1007/​
Conclusion s10661-​007-​9658-6
Alvado B, Sòria-Perpinyà X, Vicente E, Delegido J, Urrego P, Ruíz-
Several interesting findings related to the novel approach Verdú A, Soria JM, Moreno J (2021) Estimating organic and
inorganic part of suspended solids from sentinel 2 in different
for monitoring the surface water quality using UAV and inland waters. Water 13:2453. https://​doi.​org/​10.​3390/​w1318​
Sentinel-2 multispectral data were summarized in this 2453
paper. The water leaving surface reflectance data gen- Apte SC, Batley GE, Szymczak R, Rendell PS, Lee R, Waite TD (1998)
erated from Sentinel-2 imagery using C2RCC processor Baseline trace metal concentrations in New South Wales coastal
waters. Mar Freshw Res 49:203–214. https://​doi.​org/​10.​1071/​
were used for analysing the water quality parameters. The MF961​21
wider spectral resolution of Sentinel-2 has enabled the Arango JG, Nairn RW (2019) Prediction of optical and non-optical
formulation of algorithms to predict SWQPs including water quality parameters in oligotrophic and eutrophic aquatic sys-
COD, BOD, DO, TSS, TOC and Turbidity with better cor- tems using a small unmanned aerial system. Drones 4:1. https://​
doi.​org/​10.​3390/​drone​s4010​001
relations (R2 > 0.5 and p < 0.05). Using UAV for the esti- Bartholomew DJ (2010) Analysis and interpretation of multivariate
mation of water quality was also explored and algorithms data. Int Encycl Educ. https://d​ oi.o​ rg/1​ 0.1​ 016/b​ 978-0-0​ 8-0​ 44894-​
were generated for optical and non-optical parameters like 7.​01303-8
TSS, COD and TOC with good correlations. Here, 75% of Bourouhou I, Salmoun F (2021) Sea water quality monitoring using
remote sensing techniques: a case study in Tangier-Ksar Sghir
the sample data were used for generating the models, and coastline. Environ Monit Assess 193:557. https://d​ oi.o​ rg/1​ 0.1​ 007/​
the models so developed were validated using the remain- s10661-​021-​09242-4
ing 25% data. The SWR models from both Sentinel-2 and Caballero I, Stumpf R, Meredith A (2019) Preliminary assessment
UAV imagery on validation resulted in R2 > 0.80 demon- of turbidity and chlorophyll impact on bathymetry derived from
Sentinel-2A and Sentinel-3A satellites in South Florida. Remote
strating increased reliability and accuracy of the models Sensing 11:645. https://​doi.​org/​10.​3390/​rs110​60645
in predicting various water quality parameters. Chang CW, Laird DA, Mausbach MJ, Hurburg CR (2001) Near-infra-
red reflectance spectroscopy–principal components regression
Acknowledgements The authors would like to thank the authorities of analyses of soil properties. Soil Sci Soc Am J 65:480. https://​doi.​
Coimbatore Municipal Corporation for their support in the fieldwork org/​10.​2136/​sssaj​2001.​65248​0x
activities and data collection. Cillero Castro C, Domínguez Gómez JA, Delgado Martín J, Hinojo
Sánchez BA, Cereijo Arango JL, Cheda Tuya FA, Díaz-Varela R
Author’s contribution All authors equally participated in overseeing (2020) An UAV and satellite multispectral data approach to moni-
the laboratory work, data analysis, and manuscript writing and review. tor water quality in small reservoirs. Remote Sensing 12:1514.
https://​doi.​org/​10.​3390/​rs120​91514

13
International Journal of Environmental Science and Technology

El Mountassir O, Bahir M, Ouazar D, Ouhamdouch S, Chehbouni satellite remote sensing. Sci Rep 9:18310. https://d​ oi.o​ rg/1​ 0.1​ 038/​
A, Ouarani M (2020) The use of GIS and water quality index s41598-​019-​54453-y
to assess groundwater quality of krimat aquifer (Essaouira; Moore GK (1980) Satellite remote sensing of water turbidity/Sonde de
Morocco). SN Applied Sciences 2:871. https://​doi.​org/​10.​1007/​ télémesure par satellite de la turbidité de l’eau. Hydrol Sci Bull
s42452-​020-​2653-z 25:407–421. https://​doi.​org/​10.​1080/​02626​66800​94919​50
Field sampling program (waterquality.gov.au) Mushtaq F, Nee Lala MG (2016) Remote estimation of water quality
Gholizadeh M, Melesse A, Reddi L (2016) A Comprehensive review parameters of Himalayan lake (Kashmir) using Landsat 8 OLI
on water quality parameters estimation using remote sensing imagery. Geocarto Int 32:274–285
techniques. Sensors 16:1298. https://​d oi.​o rg/​1 0.​3 390/​s 1608​ Papenfus M, Schaeffer B, Pollard AI, Loftin K (2020) Exploring the
1298 potential value of satellite remote sensing to monitor chlorophyll-
Gitelson A, Garbuzov G, Szilagyi F, Mittenzwey KH, Karnieli A, Kai- a for US lakes and reservoirs. Environ Monit Assess 192:808.
ser A (1993) Quantitative remote sensing methods for real-time https://​doi.​org/​10.​1007/​s10661-​020-​08631-5
monitoring of inland waters quality. Int J Remote Sens 14:1269– Pavelsky TM, Smith LC (2009) Remote sensing of suspended sediment
1295. https://​doi.​org/​10.​1080/​01431​16930​89539​56 concentration, flow velocity, and Lake Recharge in the Peace-
Gowraraju N, Mayandi J, Polavakkalipalayam NM (2014) Water qual- Athabasca Delta. Can Water Resour Res. https://​doi.​org/​10.​1029/​
ity index of Ukkadam pond connected to River Noyyal, Coim- 2008w​r0074​24
batore, Tamil Nadu, India. Malaya J Bio Sci 1:100–103 Richard S, Seri CY, France VW, Bruno C (2000) Microbial decomposi-
Hasab HA, Jawad HA, Dibs H, Hussain HM, Al-Ansari N (2020) Eval- tion of large organic particles in the northwestern Mediterranean
uation of water quality parameters in Marshes Zone Southern of Sea: an experimental approach. Mar Ecol Prog Ser 198:61–72.
Iraq based on remote sensing and GIS techniques. Water Air Soil https://​doi.​org/​10.​3354/​meps1​98061
Pollut. https://​doi.​org/​10.​1007/​s11270-​020-​04531-z Sharaf El Din E, Zhang Y (2017) Estimation of both optical and
Hellweger FL, Schlosser P, Lall U, Weissel JK (2004) Use of satellite non-optical surface water quality parameters using Landsat 8
imagery for water quality studies in New York Harbor. Estuar OLI imagery and statistical techniques. J Appl Remote Sens
Coast Shelf Sci 61:437–448. https://​doi.​org/​10.​1016/j.​ecss.​2004.​ 11:046008. https://​doi.​org/​10.​1117/1.​JRS.​11.​046008
06.​019 Song K, Li L, Li S, Tedesco L, Hall B, Li L (2011) Hyperspectral
https://​en.​wikip​edia.​org/​wiki/​Ukkad​am_​Lake remote sensing of total phosphorus (TP) in three central indiana
https://​law.​resou​rce.​org/​pub/​in/​bis/​S02/​is.​3025.​44.​1993.​pdf (Assessed water supply reservoirs. Water Air Soil Pollut 223:1481–1502.
21 September 2021) https://​doi.​org/​10.​1007/​s11270-​011-​0959-6
https://​law.​resou​rce.​org/​pub/​in/​bis/​S02/​is.​3025.​58.​2006.​pdf (Assessed Toming K, Kutser T, Laas A, Sepp M, Paavel B, Nõges T (2016) First
11 October 2021) experiences in mapping lake water quality parameters with Sen-
https://​seadas.​gsfc.​nasa.​gov/​help/8.​0.0/​c2rcc/​C2RCC​Tool.​html tinel-2 MSI imagery. Remote Sens 8:640. https://d​ oi.o​ rg/1​ 0.3​ 390/​
https://​www.​elgal​abwat​er.​com/​blog/​total-​organ​ic-​carbon-​toc rs808​0640
Jeyaraj M, Ramakrishan K, Jai Anandhi A, Arunachalam S, Tóth VZ, Ladányi M, Jung A (2021) Adaptation and validation of a
Magudeswaran P (2016) Investigation of physico-chemical and sentinel-based chlorophyll-a retrieval software for the Central
biological characteristics of various lake water in Coimbatore European Freshwater Lake, Balaton. PFG – Journal of Photo-
District, Tamilnadu, India. Orient J Chem 32:2087–2094 grammetry, Remote Sensing and Geoinformation Science. https://​
Kapalanga TS, Hoko Z, Gumindoga W, Chikwiramakomo L (2021) doi.​org/​10.​1007/​s41064-​021-​00160-1
Remote-sensing-based algorithms for water quality monitoring in Tsouros DC, Bibi S, Sarigiannidis PG (2019) A review on UAV-based
Olushandja Dam, north-central Namibia. Water Supply 21:1878– applications for precision agriculture. Information 10:349. https://​
1894. https://​doi.​org/​10.​2166/​ws.​2020.​290 doi.​org/​10.​3390/​info1​01103​49
Khaki M (2020) Satellite remote sensing in hydrological data Vercruysse K, Grabowski RC, Hess T, Lexartza-Artza I (2020)
assimilation. Springer, Cham. https:// ​ d oi. ​ o rg/ ​ 1 0. ​ 1 007/​ Linking temporal scales of suspended sediment transport
978-3-​030-​37375-7 in rivers: towards improving transferability of prediction.
Koponen S (2002) Lake water quality classification with airborne J Soils Sediments 20:4144–4159. https://​d oi.​o rg/​1 0.​1 007/​
hyperspectral spectrometer and simulated MERIS data. Remote s11368-​020-​02673-5
Sens Environ 79:51–59. https://​doi.​org/​10.​1016/​s0034-​4257(01)​ Verpoorter C, Kutser T, Seekell DA, Tranvik LJ (2014) A global
00238-3 inventory of lakes based on high-resolution satellite imagery.
Liu H, Li Q, Shi T, Hu S, Wu G, Zhou Q (2017) Application of Sentinel Geophys Res Lett 41:6396–6402. https://​doi.​org/​10.​1002/​2014g​
2 MSI images to retrieve suspended particulate matter concen- l0606​41
trations in Poyang Lake. Remote Sens 9:761. https://​doi.​org/​10.​ Vertucci FA, Likens GE (1989) Spectral reflectance and water quality
3390/​rs907​0761 of Adirondack mountain region lakes. Limnol Oceanogr 34:1656–
Mihu-Pintilie A (2018) The seasonal variation of physical-chemi- 1672. https://​doi.​org/​10.​4319/​lo.​1989.​34.8.​1656
cal parameters and water quality assessment (WQI). In: Natu- Wang Y (2004) Water quality change in reservoirs of Shenzhen, China:
ral Dam Lake Cuejdel in the Stânişoarei Mountains, East- detection using LANDSAT/TM data. Sci Total Environ. https://​
ern Carpathians. Springer, Cham. https://​d oi.​o rg/​1 0.​1 007/​ doi.​org/​10.​1016/​s0048-​9697(04)​00100-7
978-3-​319-​77213-4_7 Warren MA, Simis SGH, Martinez-Vicente V, Poser K, Bresciani M,
Minu P, Souda VP, Baliarsingh SK, Dwivedi RM, Ali Y, Ashraf Alikas K, Ansper A (2019) Assessment of atmospheric correction
PM (2020) Assessing temporal variation of coloured dissolved algorithms for the Sentinel-2a multispectral imager over coastal
organic matter in the coastal waters of South Eastern Arabian and inland waters. Remote Sens Environ 225:267–289. https://d​ oi.​
Sea. Acta Oceanol Sin 39:102–109. https://​doi.​org/​10.​1007/​ org/​10.​1016/j.​rse.​2019.​03.​018
s13131-​020-​1534-z Water quality sampling techniques (usgs.gov)
Mishra S, Stumpf RP, Schaeffer BA, Werdell PJ, Loftin KA, Meredith Xiong J, Lin C, Ma R, Cao Z (2019) Remote sensing estimation of
A (2019) Measurement of cyanobacterial bloom magnitude using lake total phosphorus concentration based on MODIS: A case

13
 International Journal of Environmental Science and Technology

study of Lake Hongze. Remote Sens 11:2068. https://​doi.​org/​10.​ Yu X, Yi H, Liu X, Wang Y, Liu X, Zhang H (2016) Remote-sensing
3390/​rs111​72068 estimation of dissolved inorganic nitrogen concentration in the
Yang B, Liu Y, Ou F, Yuan M (2011) Temporal and spatial analysis of Bohai Sea using band combinations derived from MODIS data.
COD concentration in East Dongting Lake by using of remotely Int J Remote Sens 37:327–340. https://d​ oi.o​ rg/1​ 0.1​ 080/0​ 14311​ 61.​
sensed data. Procedia Environ Sci 10:2703–2708. https://​doi.​org/​ 2015.​11255​55
10.​1016/j.​proenv.​2011.​09.​420

13

View publication stats