=Paper=
{{Paper
|id=Vol-2858/short9
|storemode=property
|title=Estimation of water quality parameters from space images
|pdfUrl=https://ceur-ws.org/Vol-2858/short9.pdf
|volume=Vol-2858
|authors=Anastasia Popova
|dblpUrl=https://dblp.org/rec/conf/aicts/Popova20
}}
==Estimation of water quality parameters from space images==
https://ceur-ws.org/Vol-2858/short9.pdf
Estimation of water quality parameters from space images
A K Popova
Matrosov Institute for System Dynamics and Control Theory SB RAS, 134
Lermontova Street, Irkutsk, 664033, Russia
chudnenko@icc.ru
Abstract. Water quality affects many human activities. Remote sensing is efficient and
economical instrument for water monitoring. The paper investigates the problem of choosing
an algorithm for Chl-a concentration determination. In this study, we made calculations for
Multispectral Instrument (MSI) on Sentinel-2 for Lake Baikal by different empirical algorithms
and C2RCC tool. We choose 3 band combination that have high correlation with in situ data of
Chl-a. Resultant distribution map display spatial dynamics of Chl-a in the lake. Our research is
intended to help environmental scientist to assess pollution level of the Lake Baikal and
interpret the ecological meaning of results.
1. Introduction
In the traditional approach, water quality is estimated using water samples from different depths. The
samples are analyzed in a laboratory to determine the physical and chemical properties of the water.
There are also automatic stations and autonomous underwater vehicles capable of determining water
parameters. Such methods are accurate but require investment of time, money, trained personnel, and
laboratory equipment. In some areas, estimation of water quality is complicated by low transport
accessibility.
These problems can be eliminated by using remote sensing, which provide spatial measurements in
near real time in almost any area. With remote sensing we can detect, track and map water pollutants -
oil and chemical spills, algal blooms and high concentrations of suspended solids. Satellite data has a
wide temporal coverage and allows to estimate the quality over an extended water area. Traditional
measurements are carried out from once a month to once every five years, and remote sensing have a
frequency of survey from one day. Spatial coverage plays an important role in determining pollution
in a reservoir as a whole, since one or more sampling points do not always accurately describe the
state of the entire object [1-2].
2. Materials and Methods
Water quality can be determined by its optical properties, which depend on many factors, for example,
on the amount of suspended organic and inorganic particles and dissolved substances. The most
commonly used indicators for quality assessment are the concentration of chlorophyll-a (Chl-a) and
the total suspended matter (TSM), the absorption coefficient of colored dissolved organic matter
(CDOM), and the transparency of water measured by Secchi disk (SD). Lakes are optically more
complex and diverse than sea or ocean waters, therefore these indicators can have large temporal and
_____________
Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0
International (CC BY 4.0).
spatial variations, depending on changes in weather, biological composition and physical properties of
the lake [3-4].
Phytoplankton forms the basis of the aquatic food web and it is the main driver of biogeochemical
processes in the ocean. In inland waters, phytoplankton biomass is an important parameter of water
quality, since the abundance of algae can potentially indicate the degree of eutrophication in a
particular water body. The main phytoplankton pigment is Chl-a, one of the most well-known
parameters of water quality, therefore it is used as a simple substitute for phytoplankton biomass all
over the world [5].
Lakes are increasingly subject to phytoplankton blooms due to eutrophication caused by
cyanobacteria, which produce potentially lethal toxins that pose health risks. Also, phytoplankton
bloom is one of the main factors affecting water transparency, and there is a negative correlation
between these two parameters. Consequently, Chl-a is an important parameter for tracking
eutrophication, lake ecological status and health risks from cyanobacterial blooms [6].
TSM determines the organic and mineral suspended solid particles that can come from rivers or
rise from the bottom: living and dead phytoplankton, humic substances, clay minerals, detritus, etc.
CDOM is a mixture of organic molecules produced by the decomposition of terrestrial vegetation,
higher aquatic plants, phytoplankton, or bacteria. SD is an estimation of the water transparency with a
Secchi disk, which is defined as the reflection of light from disk surface. All these parameters can be
determined with satellite images.
Phytoplankton is an effective light absorber in blue and red bands due to the presence of
intracellular pigments. Therefore, remote sensing of phytoplankton is usually based on the detection of
its absorption of light. Currently, there are no generally accepted algorithms for accurately obtaining
Chl-a in inland waters based on the reflection coefficient of remote sensing. Unlike the open oceans,
inland waters are influenced by terrestrial materials such as humus and iron-rich minerals, which
contribute significantly to the absorption and scattering of light in water. As a result, the standard blue-
to-green bands ratio algorithms that are acceptable for open ocean are not suitable for inland waters.
The near infrared (NIR) bands are also used to determine Chl-a. The advantage of longer
wavelengths is to reduce the effect of absorption of light by non-algae particulate matter and CDOM,
as their absorption coefficients decrease exponentially with wavelength [7-8].
The choice of the Chl-a estimation algorithm depends on the optical properties of the water. Most
remote sensing algorithms for assessing Chl-a concentration are based on the principles of changing
light absorption depending on the amount of algae pigment. The calculations use the following
properties of the reflection coefficients:
the first peak of strong absorption in the blue region with a wavelength between 440 and
510 nm;
minimum absorption in the green region between 550 and 555 nm;
a second absorption peak in the red region between 670 and 675 nm;
minimum absorption in the NIR region between 685 and 710 nm [9].
Algorithms based on two or three spectral band ratios are more preferable, since they help to
reduce the influence of light, atmosphere and water surface [10].
Currently, water color data is provided by a number of satellite sensors: the geostationary ocean
color imager (GOCI), the moderate resolution spectroradiometer (MODIS), the color ocean and land
imaging sensor (OLCI), the Landsat operational land imager (OLI), and the Sentinel-2 multispectral
instrument (MSI). All these sensors have suitable spectral and spatial resolution, capable of detecting
pollutants and assessing water quality parameters. MSI and OLCI sensors are more often used in
research now.
In the work we used Sentinel-2 MSI images to determine the amount of chlorophyll in Lake Baikal.
Sentinel-2 was launched under the European Space Agency Copernicus program in 2015 (Sentinel-
2A) and 2017 (Sentinel-2B). Both carry a MultiSpectral Instrument (MSI) with a multispectral imager
with 13 spectral bands (Table 1), covering channels from blue to short wave infrared (SWIR) with a
resolution of 10 to 60 m. Provides global coverage every 5 days.
Table 1. Spectral bands for the Sentinel-2 sensors.
Central
Band wavelength
(nm)
Band 1 – Coastal aerosol 442
Band 2 – Blue 492
Band 3 – Green 559
Band 4 – Red 665
Band 5 – Vegetation red edge 705
Band 6 – Vegetation red edge 740
Band 7 – Vegetation red edge 782
Band 8 – NIR 832
Band 8A – Narrow NIR 864
Band 9 – Water vapour 945
Band 10 – SWIR – Cirrus 1373
Band 11 – SWIR 1613
Band 12 – SWIR 2202
S2 MSI has high potential for monitoring Chl-a in coastal and inland waters due to the red channel
at around 705 nm (band 5) and near the second absorption peak of Chl-a in the red band (band 4, 665
nm).
3. Results and Discussion
We took data of the Chl-a concentration near the village Bolshie Koty (105°04′ E, 51°53′ N) from
open sources as in situ data [11]. This is data from an automatic station for measuring water quality
parameters; it collects data obtained from sensors of dissolved oxygen, hydrogen index (pH), redox
potential, chlorophyll-a and other indicators of the lake Baikal. Measurement results are available on
the server [14].
We downloaded Sentinel-2 MSI images for the southern end of Lake Baikal for July 2017 from
Copernicus Open Access Hub. The calculations were performed in the SNAP. SNAP is an open-
source software with toolboxes for processing of Earth Observation data from Sentinel satellites.
Satellite images were previously preprocessed: all MSI bands were previously resampled to 20 m
and passed atmospheric correction by the algorithm Case-2 Regional CoastColour (C2RCC).
Atmospheric correction is an essential procedure as for 90% of the signal that reaches the image
sensor is affected by the absorption and scattering by aerosols and different particles in the atmosphere
(water vapor, carbon dioxide, ozone). The retrieval of water constituents, or its optical properties, is
achieved by inversion of the water leaving reflectance spectrum, measured at top of atmosphere and
thus requiring a correction for atmospheric effects [12].
The C2RCC processor relies on a large database of simulated water leaving reflectances and related
top-of-atmosphere radiances. Neural networks are trained in order to perform the inversion of
spectrum for the atmospheric correction, i.e. the determination of the water leaving radiance from the
top of atmosphere radiances, as well as the retrieval of inherent optical properties of the water body.
Also, C2RCC gives outputs results of chl-a and TSM and provides the possibility to add additional
background information such as salinity, elevation, ozone, temperature, and air pressure [13].
Based on the literature overview, 13 empirical algorithms were selected. There are the various
combinations of the ratios of two and three S2 MSI bands. In table 2 we present the results of
calculations by empirical algorithms and by built-in function of C2RCC. Column “Range of values”
presents min and max values obtained in calculations for the whole image, column “Value at Bolshie
Koty” presents the value at the point of in situ data.
The mean absolute percentage error (MAPE) is a measure of prediction accuracy of a forecasting
method in statistics, for example in trend estimation, also used as a loss function for regression
problems in machine learning. It usually expresses the accuracy as a ratio defined by the formula:
𝑛
1 𝑥 − 𝑦𝑖
М = ∑| |
𝑛 𝑥
𝑖=1
where x is the actual value and y is the forecast value. Last column in table 2, “M” is MAPE for each
algorithm. The larger M value mean the larger error of algorithm.
Table 2. Comparison of results.
Value at
Empirical algorithms Range of values Bolshie M
Koty
C2RCC MSI 0,009…28,87 0,54 0,5
R665/R705 0,4…1,258 1,11 0,0277
R705/R740 0,26…1,38 0,94 0,129
R665/R740 0,12…1,582 0,98 0,0925
R740/R665 0,63…8,2 1,04 0,037
log(R705/R665) -0,23…0,885 -0,11 1,101
R740/R705-R705/R665 -0,32…2,157 0,22 0,796
R560/R665 0,67…1,983 1,59 0,472
R783/R665 0,45…10,97 0,80 0,259
(R490-R665)/R560 -0,548…1,064 0,84 0,22
R705/R665 0,596…10,2 0,94 0,129
(R490-R665)/(R560-R665) -142,9…195,03 2,19 1,027
(R665-1-R705-1)*R740 -0,17…4,82 -0,10 1,09
The best result with smallest error was obtained for the ratios of bands R665/R705, in second place –
R740/R665, and third is R665/R740. All these algorithms use red and vegetation red edge bands with 665,
705, 740 nm wavelength. At the same time C2RCC showed values far from in situ as well as some
other algorithms, although they were good in research on European lakes [3, 6].
Maps of chlorophyll distribution in the South Baikal and Angara river were constructed in SNAP.
Figure 1 shows result of the calculations for some algorithms. Blue color matches the low values and
dark red – high values for calculated chl-a.
Figure 1. Maps of chlorophyll distribution in Sourth Baikal calculated by empirical algorithms.
4. Conclusion
Remote sensing is an effective method for water bodies monitoring. It helps to manage rivers, lakes
and marine coastal area and support decision making by quickly measurement of various water quality
parameters. Real-time or near-real-time measurements of water pollutants and toxins at various spatial
scales are necessary to monitor and manage environmental impacts and understand the processes
governing their spatial distribution.
The purpose of the study is to explore algorithms for chlorophyll-a concentration calculations for
Lake Baikal. It did not include an assessment of the pollution level of the lake and the ecological
meaning of the results. We hope that the chosen algorithms will help environmental scientists to
estimate these parameters.
In the future, we plan to expand the studies with calculations on the images from Sentinel-3 with
the OLCI sensor, which has more bands in the red and NIR spectrum, and to calculate other water
quality parameters for Lake Baikal.
Acknowledgments
The results were obtained within the framework of the State Assignment of the Ministry of Education
and Science of the Russian Federation for the project "Methods and technologies of cloud-based
service-oriented platform for collecting, storing and processing large volumes of multi-format
interdisciplinary data and knowledge based upon the use of artificial intelligence, model-guided
approach and machine learning" (state registration number 121030500071-2). Some results were
obtained using the facilities of the Centre of collective usage "Integrated information network of
Irkutsk scientific educational complex".
References
[1] Hafeez S, Sing Wong M, Abbas S, Y. T. Kwok C, Nichol J, Ho Lee K, Tang D and Pun L 2019
Detection and Monitoring of Marine Pollution Using Remote Sensing Technologies Monit.
Mar. Pollut.
[2] Topp S N, Pavelsky T M, Jensen D, Simard M and Ross M R V 2020 Research trends in the use
of remote sensing for inland water quality science: Moving towards multidisciplinary
applications Water (Switzerland) 12 1–40
[3] Soomets T, Uudeberg K, Jakovels D, Brauns A, Zagars M and Kutser T 2020 Validation and
Comparison of Water Quality Products in Baltic Lakes Using Sentinel-2 MSI and Sentinel-3
OLCI Data Sensors 20 742
[4] Toming K, Kutser T, Uiboupin R, Arikas A, Vahter K and Paavel B 2017 Mapping water
quality parameters with Sentinel-3 Ocean and Land Colour Instrument imagery in the Baltic
Sea Remote Sens. 9, 10
[5] Alikas K, Kangro K and Reinart A 2010 Detecting cyanobacterial blooms in large North
European lakes using the maximum chlorophyll index Oceanologia 52 237–57
[6] Ansper A and Alikas K 2019 Retrieval of chlorophyll a from Sentinel-2 MSI data for the
European Union water framework directive reporting purposes Remote Sens. 2019 11, 64
[7] Zheng G and DiGiacomo P M 2017 Remote sensing of chlorophyll-a in coastal waters based on
the light absorption coefficient of phytoplankton Remote Sens. Environ. 201 331–41
[8] Filipponi F 2018 River Color Monitoring Using Optical Satellite Data Proceedings 2 569
[9] Ha N T T, Thao N T P, Koike K and Nhuan M T 2017 Selecting the best band ratio to estimate
chlorophyll-a concentration in a tropical freshwater lake using sentinel 2A images from a
case study of Lake Ba Be (Northern Vietnam) ISPRS Int. J. Geo-Information 6 290
[10] Conopio M, Japor R K, Blanco A C and Tamondong A M 2019 Estimation of chlorophyll-a
concentration in Laguna de Bay using Sentinel-3 satellite data Int. Arch. Photogramm.
Remote Sens. Spat. Inf. Sci. - ISPRS Arch. 42 125–32
[11] Shimaraeva, S.V., Pislegina, E.V., Krashchuk, L.S. et al. Dynamics of chlorophyll a
concentration in the South Baikal pelagic during the direct temperature stratification period.
Inland Water Biol 10, 59–63 (2017). https://doi.org/10.1134/S1995082917010163
[12] Pereira-Sandoval M, Ruescas A, Urrego P, Ruiz-Verdú A, Delegido J, Tenjo C, Soria-Perpinyà
X, Vicente E, Soria J and Moreno J 2019 Evaluation of atmospheric correction algorithms
over Spanish inland waters for sentinel-2 multi spectral imagery data Remote Sens. 11 1–23
[13] Brockmann C, Doerffer R, Peters M, Stelzer K, Embacher S and Ruescas A 2018 Evolution of
the C2RCC neural network for Sentinel 2 and 3 for the retrieval of ocean colour products in
normal and extreme optically complex waters J. Mater. Process. Technol. 1 1–8
[14] Data server of Baikal water parametrs. Income accessed online on 22th December 2020 via
https://hlserver.lin.irk.ru/shs/rinko/