Regional Studies in Marine Science 75 (2024) 103556

Contents lists available at ScienceDirect

Regional Studies in Marine Science

journal homepage: www.elsevier.com/locate/rsma

The assessment of chlorophyll-a retrieval algorithm and its spatial-temporal

distribution using sentinel-2 MSI off the Banjir Kanal Timur River,
Semarang, Indonesia
Lilik Maslukah a, *, Anindya Wirasatriya a, Yusuf Jati Wijaya a, Dwi Haryo Ismunarti a,
Rikha Widiaratih a, Heru Nur Krisna b
a
Department of Oceanography, Faculty of Fisheries and Marine Sciences, Universitas Diponegoro, Semarang, 50275, Indonesia
b
Master Program, Marine Sciences, Faculty of Fisheries and Marine Sciences, Universitas Diponegoro, Semarang, 50275, Indonesia

A R T I C L E I N F O A B S T R A C T

Keywords: In aquatic ecology, Chlorophyll-a (Chl-a) plays a pivotal role in biogeochemical cycles, as well as a key indicator
Coastal algorithm for the trophic state in the environment. Coastal area of Semarang is known as eutrophic waters. Thus continuous
Chlorophyll-a monitoring of Chl-a will be beneficial for the management of the impact of the trophic dynamic in the coastal
Visible-NIR band
waters of Semarang. Taking the benefit of the continuous monitoring of satellite measurement, the present study
Sentinel-2 MSI
provides the assessment of Chl-a off the Banjir Kanal Timur (BKT) River, Semarang, using the proper remote
sensing approach. In this study, algorithms were generated according to bands sensitive to Chl-a scattering
properties, using multiple regression between in-situ Chl-a and Bottom of Atmospheric (BoA) reflectance from
Sentinel-2 with 10 m resolution. A total of 43 water samples were collected to build the algorithm and 43 of the
water samples were used for validation during the passing time of Sentinel-2 at the equator i.e., on 21 August
2021 at 9.00–11.00 local time. The results show that the performance of the algorithm generated in this study
has high accuracy as denoted by the low RMSE, and MAPE, respectively. For the seasonal variation, the highest
(lowest) concentration of Chl-a is observed in December (September), with values reaching 11.92 µg/L (7.06 µg/
L). The seasonal fluctuation of Chl-a depends on air temperature and sunshine duration. High (low) air tem­
perature and long (short) duration of solar exposure result in increased (decreased) Chl-a off the BKT River.
Conversely, the elevated wind velocity only impacts the rise of Chl-a in the region far from the river estuary,
which is associated with the mixing process. BKT waters are categorized as eutrophic conditions, with a con­
centration exceeding 6 µg/L. These algorithms can provide valuable insights into the health and productivity of
coastal ecosystems, aiding in their monitoring and management.

1. Introduction Tran et al., 2023).

The determination of Chl-a distribution within a given area is typi­
Phytoplankton holds significant ecological importance within the cally achieved through the implementation of field sampling techniques,
marine ecosystem owing to its capacity for photosynthesis, a metabolic which are subsequently complemented by laboratory-based analyses
process that facilitates the synthesis of organic compounds utilizing light (Abbas et al., 2019; Maslukah et al., 2019). Nevertheless, this approach
energy and carbon dioxide (Rost et al., 2008; Abbas et al., 2019; Aranha proves to be ineffective and inefficient due to its prolonged duration,
et al., 2022). Phytoplankton possesses chlorophyll-a (Chl-a) within its elevated expenses, and restricted geographical reach (Duan et al., 2010;
cellular structure and plays a crucial role in ecological systems, serving Ogashawara et al., 2021; Aranha et al., 2022). To address these con­
as a fundamental component of the aquatic life cycle and serving as the cerns, several remote sensing techniques have been devised to monitor
primary source of energy within oceanic food webs. Chl-a is a crucial Chl-a (Ouma et al., 2020;). Remote sensing methods show promise for
factor in assessing water quality due to its regulatory role in biological monitoring Chl-a levels in a given area due to their extensive coverage
processes occurring within coastal waters (Ogashawara et al., 2021; and ability to provide periodic observations (Maslukah et al., 2019;

* Corresponding author.
E-mail address: [email protected] (L. Maslukah).

https://doi.org/10.1016/j.rsma.2024.103556
Received 11 December 2023; Received in revised form 19 April 2024; Accepted 27 April 2024
Available online 5 May 2024
2352-4855/© 2024 Elsevier B.V. All rights reserved.
L. Maslukah et al. Regional Studies in Marine Science 75 (2024) 103556

Ouma et al., 2020). The utilization of ocean color data within the visible visible NIR band (Ouma et al., 2020), making it particularly useful for
and near-infrared (VNIR) wavelength bands facilitates the estimation of monitoring and evaluating Chl-a levels in complex coastal waters (Gohin
Chl-a levels (Ouma et al. 2020). The acquisition of Chl-a data from these et al., 2020).
satellites necessitates the utilization of an algorithm that is heavily The Banjir Kanal Timur (BKT) River is one of Semarang’s flood
reliant on the band and spectral resolution capabilities offered by the control systems located in the eastern part of the city (Fig. 1). This river
satellite (Ogashawara et al., 2021). has a length of ± 14.50 km. The BKT River is the unification of several
The quantification of Chl-a in reservoirs and lakes has been exten­ rivers and canals such as the Penggaron River through the Pucang
sively studied, and multiple algorithms have been devised for this pur­ Gading outlet, Kedung Mundu River drainage outlet, Candi River, Bajak
pose. These algorithms encompass different approaches, such as River, Kartini pump drainage and large rice field / Sambirejo pump
utilizing two bands (red and near-infrared) as demonstrated by O’Reilly drainage and empties into the Java Sea. The upstream flow have a major
et al. (1998) and Tran et al. (2023), as well as employing a single band, influence on the input of organic and non-organic materials from soil
as explored by Bramich et al. (2020). The phenomenon above is simi­ erosion and solid and liquid waste, which can affect water quality.
larly observed in marine regions, where algorithms are employed to During the rainy season, sediment from land enters the marine waters at
calculate the ratio between blue and green color bands (Tilstone et al., a higher rate and causes the estuary to become more turbid. The results
2021; Tran et al., 2023). In the expanse of open ocean waters, the of research by Utama et al. (2021) explained that the suspended solids in
presence of colored dissolved organic matter (CDOM) and suspended front of the BKT estuary reached 290 mg/L in the rainy season and
particulate matter is limited to relatively small quantities (Nelson and 60 mg/L in the dry season (Hutasuhut et al., 2022). Apart from high
Siegel, 2002). Additionally, the concentration of Chl-a in these waters is turbidity, the front of the river mouth is also a eutrophic area with an
comparatively lower than that observed in coastal regions. Estuarine average range of chlorophyll-a of 12.67 µg/L in the dry season (Maslu­
regions, characterized by terrestrial influence, exhibit heightened opti­ kah et al., 2019) and 18.06 µg/L in the wet season (Maslukah et al.,
cal complexity as a result of the absorption caused by CDOM and 2020). Based on the classification of Hakanson & Bryan (2008), BKT
non-algal particles (Ouma et al., 2020; Tran et al., 2023). The optical waters are in eutrophic conditions. Thus, temporal and continuous
properties of water are significantly influenced by the presence of turbid monitoring of Chl-a is required to manage the impact of the trophic
coastal waters characterized by high concentrations of CDOM and sus­ dynamic in the coastal waters of Semarang.
pended non-phytoplanktonic particles (Lewis & Arrigo, 2020). Those The previous monitoring of Chl-a using Sentinel-2 in front of the BKT
mentioned above intricate aquatic conditions diminish the efficacy of Semarang estuary was conducted by Subiyanto et al. (2018) following
the band ratio algorithm in estimating Chl-a concentration. This is the algorithm used by Nuriya et al. (2016) in the Madura Strait. In the
because the signal of the blue band declines below the detection algorithm, the development band is the ratio of the red and NIR (Near
threshold as a result of substantial absorption by phytoplankton Infra Red) bands to the green band. However, this research is still
(Soja-Wozniak et al., 2020). questionable because it is without validation of field data. The use of the
The utilization of ocean color remote sensing in turbid regions (case NIR band for coastal waters reduces the absorption of colored dissolved
2) presents significant challenges due to the unique characteristics organic matter (Gilerson et al., 2010). It was further explained by Yoon
exhibited by each coastal area. Hence, an algorithm formulated for a et al. (2019) that the empirical algorithm of blue and green bands is
specific domain may lack generalizability to other domains. The more suitable for oligotrophic and mesotrophic coastal waters. Given
Sentinel-2 satellite has a superior spatial resolution of 10 m for the the absence of a Chl-a estimation algorithm for Sentinel-2, which

Fig. 1. The research site is located in the vicinity of the BKT river estuary (blue line) in Semarang, Indonesia. Red star is a meteorological station used for pre­
cipitation data retrieval.

2
L. Maslukah et al. Regional Studies in Marine Science 75 (2024) 103556

characterizes the foreshore waters of the BKT river and accommodates m., aligning with the specific date and time of the Sentinel-2 data
the existence of previous studies regarding the bands, this study aims to acquisition over Semarang City (refer to Fig. 3). It is worth mentioning
develop an algorithm for estimating Chl-a levels in Sentinel-2 imagery. that the sampling area under consideration exhibited characteristics of
This study examined the response of each spectral band with 10 m eastern monsoon conditions. Analysis of meteorological data revealed
spatial resolution to field Chl-a concentrations. In addition, statistical that the recorded wind speed ranged from 2.19 to 2.40 m per second.
analyses and validation techniques were performed to determine the Water samples of 500 mL were collected from the near surface, with
reliability of the algorithm in estimating chlorophyll-a concentrations, a depth range of ±30 cm. The spectrometric method was employed to
which have never been previously investigated. The resulting algorithm conduct the chlorophyll analysis, utilizing three wavelengths (trichro­
was further utilized on other images to capture the spatial-temporal matic). The suspension was extracted using a solvent consisting of 90 %
fluctuations of Chl-a around the BKT River. This study is the first acetone. The filtration process employed filter paper made from cellu­
investigation into the seasonal fluctuations of Chl-a levels in the BKT lose nitrate with a pore size of 0.45 μm, specifically sourced from Mil­
River, using the self-built Sentinel-2 algorithm. It is expected that this lipore Merck. The Chl-a concentration in water samples was determined
algorithm can estimate Chl-a concentrations in other eutrophic and using the calculation formula prescribed by the American Public Health
shallow waters. Association (APHA) method, as described in the studies conducted by
Maslukah et al. (2019) and Maslukah et al. (2021).
2. Data and methods In order to refine the Chl-a algorithm, a dataset was generated
comprising of match-up data. This dataset included in-situ Chl-a mea­
2.1. Satellite data surements as the ground truth, along with corresponding satellite ob­
servations obtained from collocated locations. In order to ensure the
The satellite data employed in this study consisted of the Sentinel-2 representation of the sampling distribution, the match-up positions for
image, which encompassed the coastal region of Semarang City. The algorithm tuning and validation were evenly distributed across the study
Sentinel-2 image has undergone geometric and radiometric corrections, area. Out of the total 100 match-ups, we allocated 50 for the purpose of
resulting in reflectance values derived from the Bottom of the Atmo­ tuning and the remaining 50 for validation. In both subsets, the exclu­
sphere (BOA). This image is equipped with twelve multi-spectral bands, sion of outlier data was implemented. The figure provided, Fig. 2, dis­
spanning a range of wavelengths from 443 nm to 22,202 nm (Gatti & plays the positions of the stations used for tuning and validation
Bertolini, 2015). In the present investigation, a novel approach was purposes. The initial analysis involved evaluating the correlation (r)
devised to assess Chl-a concentration. This approach integrates elements between the overall concentration data and the visible-NIR band’s Rrs.
from the methodology proposed by Moutzouris-Sidiris and Topouzelis Subsequently, a total of 43 match-ups were employed to develop the
(2021), which employs the blue to green band ratio, as well as the green Chl-a algorithm for Sentinel-2A, while an additional 43 match-ups were
and red band ratio introduced by Ha et al. (2017). Additionally, insights utilized for the purpose of validation. Table 1 displays the statistical data
from Chen et al. (2017) regarding the NIR band and Aranha et al. (2021) pertaining to in-situ Chl-a for the purposes of algorithm tuning and
concerning the visible band were incorporated into the combined validation.
method. This research investigates the application of visible (blue,
green, red)-NIR bands, with a spatial resolution of 10 m, for the analysis
2.4. The algorithm assessment
of Chl-a levels in the BKT River. An analysis was conducted to establish a
Chl-a algorithm for the region adjacent to the BKT River by comparing
The accuracy test of the algorithms developed in this research is
the visible and NIR bands with in-situ Chl-a measurements. Therefore, in
based on three statistical techniques, namely the coefficient of deter­
order to derive the seasonal fluctuations of Chl-a, our algorithm was
mination (R2) as reported by Shaik et al. (2021) and Moutzouris-Sidiris
implemented on Sentinel-2 images that accurately represent each month
and Topouzelis et al. (2021), the root mean square error (RMSE), the
of the year 2020.
mean absolute percentage error (MAPE), and the bias, as discussed by
Watanabe et al. (2018) and Ciancia et al. (2020). The mathematical
2.2. Meteorological data equations employed for the evaluation of algorithms are derived from
formulas 1, 2, 3, and 4, correspondingly.
The meteorological data regarding daily rainfall and wind speed

⎢ ⎥2
⎢ ⎥
⎢ ∑N ∑N ∑ N ⎥
⎢ N x (Chlai , meas) xChlai,retr ) − Chlai,meas x Chlai,retr ⎥
⎢ ⎥
R2 = ⎢
⎢√ i=1 i=1 i=1
̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
⌊ ⌋ [ ]
⎥
⎥ (1)
⎣ ∑N ∑N ( )2 ∑N ( )2 ∑N ⎦
N x (Chlai , meas)2 − ( Chlai, meas N x Chlai,retr − ( Chlai,retr )2
i=1 i=1 i=1 i=1

were collected from the Achmad Yani International Airport’s Meteoro­

logical Station (red star at Fig. 1). The data collected on a daily basis √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√N (
were subsequently aggregated to form monthly averages. The data can √∑
√
)2
be accessed through the official website of the Indonesian Meteorology, √i=1 Chlai , meas − Chlai,retr
RMSE = (2)
Climatology, and Geophysics Agency (BMKG) at https://dataonline. N
bmkg.go.id/home. ⃒( )⃒
N ⃒
100% ∑ Chlai , meas − Chlai,retr ⃒
MAPE = x (3)
2.3. Sample collection and analysis N i=1
Chlai,meas

A total of 100 field samples were collected. The sampling procedure

took place on August 21, 2021, between the hours of 08:00 and 11:00 a.

3
L. Maslukah et al. Regional Studies in Marine Science 75 (2024) 103556

Fig. 2. Location of sampling for match-up model ( ) dan validation ( ). The red rectangle ( ) denotes (inshore and offshore) the regions in closest proximity to and
furthest distance from the river’s mouth.

1 ∑
N
( )
Table 1 BIAS = x Chlai,retr − Chlai,meas (4)
Statistics on in situ Chl-a used for algorithm tuning and validation. N i=1

Data n average (µg/L) Min (µg/L) Max (µg/L) St. Dev (µg/L)
where Chlai,meas and Chlai,retr refer to the measured (in-situ) and
Tunning 43 9.49 2.15 17.27 3.47 retrieved (estimated) Chl-a respectively; and N is the total number of
Validation 43 8.75 3.5 14.52 2.90
samples.

3. Results

3.1. Generating Chl-a algorithm

In order to generate the Chl-a algorithm, the initial step involves

assessing the correlation between the visible-NIR (blue/B2, green/B3,
red/B4, and NIR/B8) spectral bands and Chl-a observations. The
resulting coefficients obtained are 0.69, 0.69, 0.65, and 0.65, respec­
tively (as shown in Fig. 3). Fig. 3 demonstrates that all Rrs exhibit sig­
nificant correlations with the observed Chl-a levels. To develop the new
algorithm, we utilized Rrs NIR, red, green, and blue wavelengths to
estimate the concentration of Chl-a in the BKT offshore region. Ac­
cording to Aranha et al. (2022), bodies of water characterized by a
prevalence of phytoplankton exhibit two distinct peaks of absorption in
the electromagnetic spectrum. Specifically, one peak is observed in the
blue region at approximately 440 nm, while the other peak is located in
the red region at around 670 nm. The green band, characterized by a
wavelength range of 560 nm, exhibits a high reflectance of Chl-a owing
to its low absorption coefficient. Furthermore, the existence of phyto­
plankton is distinguished by its maximum reflection at 700 nm, specif­
ically within the near-infrared spectrum (Matthews et al., 2012). Given
Fig. 3. Scatter plot of observed Chl-a versus remote sensing reflectance of the
visible-NIR. the intricate nature of the region under investigation, the study incor­
porated an examination of the utilization of four bands on Sentinel-2
satellites, each possessing an equivalent spatial resolution of 10 m.
A model was constructed employing multiple regression analysis to

4
L. Maslukah et al. Regional Studies in Marine Science 75 (2024) 103556

Fig. 7 illustrates that, on the whole, the patterns exhibit similarity,

except for February. The region adjacent to the estuary exhibits the
highest amount in February, whereas the vicinity near the estuary
demonstrates the lowest value. This variation is associated with the
recorded monthly mean wind. The region adjacent to the estuary is
characterized by a moderate degree of sediment input. The presence of
sunlight reaching the depths of the water facilitates the flourishing of
phytoplankton in these regions, as opposed to the adjacent coastal
waters.

3.2. Seasonal dynamics of Chl-a

The island of Java is part of the maritime continent, so its climate is

influenced by the Asian (Australian) monsoon which brings humid (dry)
air, and causes the rainy season (Alifdini et al., 2021). The rainy season
occurs from December to February, while the dry season is from June to
August. From March to May and from September to November are
known as the first and second transition seasons, respectively. Fig. 8
depicts the temporal fluctuations of Chl-a levels in the BKT River. The
images depicting the rainy season, first transition, dry season, and sec­
ond transition are observed during the months of January and February,
April and May, July and August, and October and November,
respectively.
Fig. 4. Validation of Chl-a from Eq. (5) with observed Chl-a (observation). According to the data presented in Fig. 9, it is evident that the peak
concentration of Chl-a is observed during the rainy season. During the
derive estimations of Chl-a levels in the BKT River. The calculated co­ rainy season, a notable elevation in the concentration of Chl-a, reaching
efficient of determination (R2) is 0.45, The statistical significance of this up to 12 µg/L (approximately mg/m3), was observed in the vicinity of
relationship is supported by a p-value of less than 0.05. The empirical the BKT river. However, this concentration then decreased in the sub­
model produced between the observation and reflectance data is as sequent season. In the second transition, the presence of Chl-a origi­
follows Eq. 5.(Fig. 4) nating from the BKT Semarang River has resulted in a rise of 0.42 µg/L.
According to the findings of Maslukah et al. (2021), the Chl-a concen­
Chl − a = 10(2.743− 0.725∗(log(B2∗c))− 0.625∗(log(B3∗c))+1.623∗(log(B4∗c))+0.809∗(log(B8∗c))
(5)
trations in Semarang waters were observed to be higher during the rainy
season (February) compared to the dry season (July), as indicated by the
Where c: 0.036/B8; B2; B3; B4; B8 (reflectant blue, green, red and NIR).
in situ (field measurement) data. Therefore, to examine potential factors
The performance of the algorithm model in Eq. 5 was then tested
contributing to the temporal fluctuations of Chl-a, we conducted a
statistically, which resulted in RMSE, Bias and MAPE values of 2.47 μg/
comparative analysis between Chl-a levels and various meteorological
L, − 0.63, and 23.52, respectively (Fig. 5). The negative bias indicates
variables including precipitation, wind speed, average air temperature,
that estimated Chl-a underestimates the observed Chl-a. Since the RMSE
and duration of solar irradiation. These data were obtained from a
and MAPE are quite small, the Chl-a off the BKT River estimated from
neighboring meteorological station located at Ahmad Yani airport in
the obtained algorithm has a good accuracy. Thus, Eq. (5) can be used to
Semarang. The variation in Chl-a levels is influenced by terrestrial fac­
estimate TSS off the BKT River.
tors, particularly precipitation. Increased and prolonged precipitation
Hence, the Eq. (5) can be employed for the estimation of Chl-a levels
can lead to higher runoff discharge in rivers, resulting in the trans­
off BKT River. Figs. 5 and 6 depict the spatial and temporal distribution,
portation of additional nutrients from terrestrial areas. This influx of
respectively.
nutrients promotes the growth of phytoplankton (Maslukah et al.,
The highest concentration was observed in December, reaching a
2021). According to Hutasuhut et al. (2021), the velocity of wind can
value of 11.92 µg/L, while the lowest concentration was recorded in
induce water currents that have the potential to initiate resuspension
September, with Chl-a concentration of 7.06 µg/L. The observed pattern
processes, leading to the liberation of nutrients from the sedimentary
in this study diverges from the findings of Wirasatriya et al. (2019), who
layer. Additional climatological factors, such as air temperature and
utilized MODIS imagery to examine the area adjacent to the northern
duration of irradiation, play a role in influencing the variability of Chl-a
coast of Java. Wirasatriya et al. (2019) reported that the pattern reached
in the BKT River. The findings pertaining to the relationships among
its maximum in June and January. It is important to note that the pre­
climatological parameters are depicted in Fig. 6 and summarized in
vious study analyzed data over a longer time span of 10 years, encom­
Table 2.
passed a significantly larger regional area, and employed a lower image
The data presented in Table 2 demonstrates that air temperature and
resolution of 4 km. Shabrina et al. (2018) also demonstrated distinct
solar radiation have a negative correlation with Chl-a concentration.
variations in seasonal dynamics during the summer period (March-­
The findings of this study diverge from those presented by Aranha et al.
August 2015 and 2016), with the highest concentration of Chl-a
(2021) in a tropical reservoir, wherein they observed a decline in Chl-a
observed in June. The finding suggests that the temporal patterns of
concentrations during periods of minimal solar radiation. According to a
each region exhibit distinct variations.
study conducted by Shabrina et al. (2018), the concentration of Chl-a in
High Total Suspended Solids (TSS) concentrations in estuaries cause
the northern waters of Central Java during the rainy season is influenced
disturbances in the photosynthesis process, which results in low
by the levels of sunlight and temperature. Nevertheless, the degree of
phytoplankton biomass. The results of Wirasatriya et al. (2023)
correlation was not evaluated in this study.
explained that the river is the main contributor of TSS and wind affects
It was observed that wind had a positive impact, although the cor­
its distribution pattern, which will influence the temporal dynamics of
relation was not statistically significant (p>0.05). Wirasatriya et al.
Chl-a. To investigate this, in this study, we divided two areas, near and
(2021) discovered a noteworthy pattern within the waters of South
off the river, which are given a red box in Fig. 1, and the time series are
Sulawesi. In the northern region, it was observed that the wind exerted a
shown in Fig. 7 & Fig. 8 and the correlation results are shown in Table 2.
significant impact on the rise of Chl-a concentrations. However, in the

5
L. Maslukah et al. Regional Studies in Marine Science 75 (2024) 103556

Fig. 5. Distribution of Chl-a (a) 3 January 2021 (b) 27 February 2021 (c) 14 March (d) 23 April 2021 (e) 13 May 2021 (f). June 22, 2021 (g). 27 July 2021 (h) 21
August 2021 (I) 10 September 2021 (j) 10 October 2021 (k). November 4, 2021 (l). December 14, 2021 estimated from Sentinel-2 imagery using Eq. (5). The BKT
Semarang estuary, depicted with a red box.

southern region, a distinct pattern emerged, wherein elevated levels of influence on the growth and development of phytoplankton. The esti­
Chl-a were observed under conditions characterized by weakened wind mation of phytoplankton biomass can be derived by assessing the con­
speeds. This is related to sediment resuspension and influences the centration of Chl-a, a predominant green pigment within phytoplankton,
proliferation of phytoplankton. However, this analysis contains no field as indicated by Munandar et al. (2023). The presence of a significant
measurement data, which is a limitation. The responses of the optical amount of Chl-a serves as an indicator that can be utilized to assess al­
properties captured by the imagery are substantially related to the water terations in water quality (Ciancia et al., 2020; Shaik et al., 2021;
conditions, which may impact the analysis of the chlorophyll data Saberioon et al., 2020). Hence, the assessment of environmental
captured by the imagery. In turbid regions, Maslukah et al. (2022) degradation can be conducted by observing alterations in the concen­
discovered that Chl-a concentrations in coastal waters derived from tration of Chl-a. Long-term field measurements are characterized by
imagery data were underestimated. high costs, significant time requirements, and substantial financial in­
vestments (Ogashawara et al., 2021; Aranha et al., 2022). The applica­
4. Discussion tion of remote sensing presents a viable approach for addressing this
issue.
Prior research has indicated that the waters adjacent to the Banjir The BKT river estuary exhibits a relatively constrained spatial extent
Kanal Timut (BKT) exhibit elevated levels of Chl-a concentrations in comparison to the expansive waters of the open ocean. Utilizing high-
(Maslukah et al., 2019; Maslukah et al., 2021), indicating a state of resolution satellites is a more advantageous approach for monitoring the
eutrophication in terms of trophic status. Moreover, the escalation of variability of Chl-a. Hence, the focus of this research lies in the utiliza­
industrial activities and population density has led to a consequential tion of the Sentinel-2 MSI, renowned for its frequent revisits and
impact on the quality of estuarine waters. This is primarily due to the possession of a red edge spectral band at a wavelength of 705 nm. Ac­
substantial influx of nutrients from land, resulting in an adverse cording to Warren et al. (2019), this particular sensor demonstrates

6
L. Maslukah et al. Regional Studies in Marine Science 75 (2024) 103556

Fig. 8. Same as Fig. 6 but for offshore Chl-a, inshore Chl-a and precipitation.
The position of offshore and inshore areas is depicted in Fig. 2.

indicating a slight underestimation of the Chl-a estimate compared to

the observed Chl-a level. The study conducted by Moutzouris-Sidiris and
Topouzelis (2021) investigated the application of different algorithms in
case 1 and 2 waters within the Mediterranean Sea. The findings revealed
the presence of diverse bias values, both positive and negative. In a study
conducted by Maslukah et al. (2022), Sentinel-3 imagery was utilized to
assess the Chl-a concentration in the western coastal waters of Semarang
Bay. The researchers found that the Chl-a concentration derived from
Fig. 6. Temporal variation of Chl-a with wind speed (a) and precipitation (b)
the imagery exhibited a significant disparity compared to the observa­
based on Sentinel imagery on 3 January 2021, 27 February 2021, 14 March
2021, 23 April 2021, 13 May 2021, 22 June 2021, 27 July 2021, 21 August 202,
tion data. Specifically, the imagery displayed a bias value of 3.43,
10 September 2021, 10 October 2021, 4 November 2021, 14 December 2021 indicating a consistent overestimation of Chl-a levels. The observed
for whole study area (Fig. 2). range of Chl-a concentration in the imagery spanned from 0.70 to
9.07 mg/m3. Nevertheless, the RMSE about the estimated Chl-a con­
centration is recorded as 2.47 mg/m3, while the MAPE stands at
23.52 %. This finding suggests that the algorithm used to estimate the
Chl-a levels from the BKT River is still yielding relatively accurate re­
sults. The algorithm under consideration exhibits superior performance
compared to the algorithm proposed by Aranha et al. (2021), as the
latter demonstrated a relative mean error of 28 %. In contrast to the
findings of Maslukah et al. (2022), which reported a significantly lower
percentage of relative error at 1.19 %. The study conducted by Maslukah
et al. (2022) exhibits several limitations. Firstly, the sample size is
relatively small, consisting of only nine participants. Consequently, this
limited sample may introduce potential biases in the findings. Addi­
tionally, the resulting bias in the study, measured at 3.5 mg/m3, is
relatively high.
Based on Fig. 4, the MAPE value generated in this study has a high
error. This can illustrate that the use of three of the visible and NIR bands
(B8) at 10 m resolution, cannot be used as an accurate estimation of Chl-
a. Further development of algorithms related to the mixture of all bands
both visible and NIR and the use of band ratio is recommended, as has
Fig. 7. Same as Fig. 6 but for offshore Chl-a, inshore Chl-a and wind speed. The been performed by Saberioon et al. (2020) with a combination of all
position of offshore and inshore areas is depicted in Fig. 2. bands, Shaik et al. (2021) with visible and NIR bands, Katlane et al.
(2020) with single band (red), and Bramich et al. (2021) with a single
suitability to monitor water quality in both coastal and inland waters. band (NIR) and a combination of green and SWIR (Ouma et al., 2020).
The reflectance values utilized in this study were obtained from the The algorithm obtained is subsequently utilized to estimate the
BOA, which had undergone calibration using the Sen2Core. The algo­ temporal and spatial variations. The present study demonstrates that
rithm used in this study, we developed the empirical method as in the temporal fluctuations exhibit a discernible seasonal pattern, wherein
previous study by combining four bands with a resolution of 10 m (Eq. alterations in precipitation levels lead to subsequent modifications in
5). Each of these bands is multiplied by the ratio of the Near-Infrared river discharge. The westerly monsoon prevailing in Indonesia induces
(NIR) band (Rrs 705). Warren et al. (2019) conducted a study in elevated wind velocities, thereby influencing the process of sediment
which they examined three separate spectral ratios: Rrs444/Rrs560, resuspension. River discharge and resuspension are two mechanisms
Rrs490/Rrs560, and Rrs704/Rrs665. However, their findings did not that have the potential to impact nutrient inputs into the water column,
yield satisfactory results for both marine and inland areas. consequently influencing the growth of phytoplankton. The biomass of
The algorithm developed through the empirical model in this study phytoplankton can be assessed by measuring Chl-a concentrations.
(Eq. 5) was then validated to the observed Chl-a concentration, as The determination of water’s trophic status can be achieved by
illustrated in Fig. 4. The bias was achieved in this study at − 0.63 mg/m3, assessing the concentration of Chl-a. According to Hakanson and Bryann

7
L. Maslukah et al. Regional Studies in Marine Science 75 (2024) 103556

Table 2
The Pearson correlations of Chl-a with climatological data.
Chl-a Chl-a far the monthly of wind the weekly of wind the weekly of rain the monthly of rain Temperature of_ air sunshine duration
near (hour)

Chl-a .786** .797** .298 .344 .022 .128 -.682* -.803**

Chl-a inshore 1 .319 -.113 -.191 -.016 .031 -.325 -.646*
Chl-a .319 1 .590* .715** -.128 .141 -.833** -.753**
offshore

*. Correlation is significant at the 0.05 level (2-tailed).

**. Correlation is significant at the 0.01 level (2-tailed).

RMSE, MAPE, and bias of 2.44 µg/L, 23.52 %, and − 0.63 µg/L,
respectively. The study revealed that there was a greater degree of
spatial variation in Chl-a concentrations towards the mouth of the river
compared to the offshore region. This study reveals that the temporal
fluctuations in the area of the outer estuary are primarily driven by
variations in wind speed rather than precipitation-related factors. The
other climatological factors that exerted a significant influence were the
air temperature and the duration of solar radiation (SS), exhibiting a
negative correlation.

CRediT authorship contribution statement

Rikha Widiaratih: Visualization, Methodology. Heru Nur Krisna:

Fig. 9. Seasonal variation of Chl-a in 2021 during the rainy season (January Visualization, Formal analysis. Lilik Maslukah: Writing – original draft,
and February), first transition (April and May), dry season (July and August), Methodology, Formal analysis, Conceptualization. Anindya Wirasa­
and second transition (October and November). triya: Supervision, Conceptualization. Yusuf Jati Wijaya: Writing –
review & editing. Dwi Haryo Ismunarti: Software, Formal analysis.
(2008), oligotrophy is characterized by Chl-a concentrations below
2 mg/m3 in aquatic environments. Mesotrophic conditions are indicated
Declaration of Competing Interest
by Chl-a concentrations ranging from 2 to 6 mg/m3, while eutrophic
conditions are associated with concentrations between 6 and 20 mg/m3.
The authors declare the following financial interests/personal re­
Hypertrophic conditions, on the other hand, are defined by Chl-a con­
lationships which may be considered as potential competing interests:
centrations exceeding 20 mg/m3. According to the classification uti­
Lilik Maslukah reports financial support was provided by Diponegoro
lized, the coastal waters of BKT fall within the eutrophic classification.
University. Other authors declare that they have no known competing
According to the findings of Maslukah et al. (2021), the trophic condi­
financial interests or personal relationships that could have appeared to
tions of the Semarang coastal waters, which encompass the BKB and BKT
influence the work reported in this paper.
estuaries, were determined based on field observations of Chl-a con­
centration data and chlorophyll analysis from images. Based on empir­
Data availability
ical data, coastal waters can be classified as eutrophic or mesotrophic
based on the levels of Chl-a derived from satellite imagery. However, the
Data will be made available on request.
findings of this study indicate that both samples exhibited eutrophic
conditions. The issue at hand is closely tied to the limited spatial reso­
lution of MODIS, which must not be conflated with observational data Acknowledgments
obtained from a localized region. According to the findings of Maslukah
et al. (2019), the study elucidated that the concentration of Chl-a in the This research was funded by a grant from FPIK, Diponegoro Uni­
vicinity of the BKT estuary exhibits significant fluctuations within a versity under the scheme International Publication Research with con­
range of approximately ± 500 m, with variations in concentration tract number: 55/UN7.F10/PP/II/2024. The author would also like to
ranging from 1 to 10 mg/m3. The MODIS image data is acquired by thank the reviewers who helped improve the quality and clarity of this
sampling each pixel at a spatial resolution of 4 ×4 km. manuscript significantly.

5. Conclusion References

Abbas, M.M., Melesse, A.M., Scinto, L.J., Rehage, J.S., 2019. Satellite estimation of Chl-a
In this study, the seasonal spatial and temporal distribution of near- using moderate resolution imaging spectroradiometer (MODIS) sensor in shallow
surface chl-a concentrations in Semarang coastal waters in 2020 is coastal. Water bodies Valid. Improv. Water 11, 1621. https://doi.org/10.3390/
described using visible and NIR band reflectance data from Sentinel-2 w11081621.
Alifdini, I., Shimada, T., Wirasatriya, A., 2021. Seasonal distribution and variability of
imagery. This study includes the use of a new empirical model based surface winds in the Indonesian seas using scatterometer and reanalysis data. Int. J.
on Sentinel-2 observed in 2020 and taken from Semarang coastal waters Climatol. 41 (10), 4825–4843. https://doi.org/10.1002/joc.7101.
specifically (BKT river front). In addition, the relationship between Chl-a Aranha, T.R.B.T., Martinez, J.-M., Souza, E.P., Barros, M.U.G., Martins, E.S.P.R., 2022.
Remote analysis of the Chl-a concentration using sentinel-2 MSI images in a semiarid
concentration and climatological factors such as rainfall, length of environment in northeastern Brazil. Water 14, 451. https://doi.org/10.3390/
sunshine wind and tidal conditions was studied. Therefore, we tuned the w14030451.
algorithm based on the blue, green, red, and Nir bands and derived the Bramich, J., Bolch, C.J.S., Fischer, A., 2020. Improved red-edge Chl-a detection for
Sentinel-2. Ecol. Indic. 140 (106876), 1–9. https://doi.org/10.1016/j.
empirical method as follows: Chl-a ¼
ecolind.2020.106876.
10(2.743− 0.725(logB2∗c)− 0.625(logB3∗c)+1.623(logB3)+0.809(logB8∗c) where c=0.036/B8. Chen, J., Zhu, W., Tian, Y.Q., Yu, Q., Zheng, Y., Huang, L., 2017. Remote estimation of
The accuracy of the empirical method can be reflected in the values of colored dissolved organic matter and chlorophyll-a in Lake Huron using Sentinel-2
measurements. J. Appl. Remote Sens. 11 (3), 36007.

8
L. Maslukah et al. Regional Studies in Marine Science 75 (2024) 103556

Ciancia, E., Campanelli, A., Lacava, T., Palombo, A., Pascucci, S., Pergola, N., Pignatti, S., Ogashawara, I., Kiel, C., Jechow, A., Kohnert, K., Ruhtz, T., Grossart, H.P., Hölker, F.,
Satriano, V., Tramutoli, V., 2020. Modeling and multi-temporal characterization of Nejstgaard, J.C., Berger, S.A., Wollrab, S., 2021. The Use of Sentinel-2 for Chl-a
total suspended matter by the combined use of sentinel 2-MSI and Landsat 8-OLI spatial dynamics assessment: a comparative study on different lakes in Northern
data: the Pertusillo lake case study (Italy). Remote Sens. 12, 2147. https://doi.org/ Germany. Remote Sens. 13 (8), 1–26. https://doi.org/10.3390/rs13081542.
10.3390/rs12132147. Ouma, Y.O., Noor, K., Herbert, K., 2020. Modelling reservoir Chl-a, TSS, and turbidity
Duan, H., Ma, R., Xu, J., Zhang, Y., Zhang, B., 2010. Comparison of different semi- using Sentinel-2A MSI and Landsat-8 OLI satellite sensors with empirical
empirical algorithms to estimate Chl-a concentration in inland lake water. Environ. multivariate regression. J. Sens. 2020 (8858408), 1–21. https://doi.org/10.1155/
Monit. Assess. 170, 231–244 ([CrossRef] [PubMed]). 2020/8858408.
Gatti, A., & Bertolini, A. (2015). Sentinel-2 Products Specification Document. 〈https:// Rost, B., Zondervan, I., Wolf-Gladrow, D., 2008. Sensitivity of phytoplankton to future
sentinel.esa.int/documents/247904/685211/〉Sentinel-2-Product-Specifications changes in ocean carbonate chemistry: current knowledge, contradictions and
Document. research directions. Inter. Sci. Publ. 373, 227–237. https://doi.org/10.3354/
Gilerson, A.A., Gitelson, A.A., Zhou, J., Gurlin, D., Moses, W., Loannou, L., Ahmed, S.A., meps07776.
2010. Algorithms for remote estimation of chlorophyll-a in coastal and inland waters Saberioon, M., Brom, J., Nedbalc, V., Souček, P., Cίsař, P., 2020. Chl-a and total
using red and near infrared bands. Opt. Express 18, 24109–24125. suspended solids retrieval and mapping using Sentinel-2A and machine learning for
Gohin, F., Bryère, P., Lefebvre, A., Sauriau, P.-G., Savoye, N., Vantrepotte, V., Bozec, Y., inland waters. Ecol. Indic. 113, 106236 https://doi.org/10.1016/j.ecolind.
Cariou, T., Conan, P., Coudray, S., Courtay, G., Francoise, S., Goffart, A., Fariñas, T. Shabrina, B., Maslukah, L., Wulandari, S.Y., 2018. Chl-a distribution and its relation with
H., Lemoine, M., Piraud, A., Raimbault, P., Retho, M., 2020. Satellite and in situ current pattern in Northern Waters of Central Java. Omni Akuatika 14 (1), 69–76.
monitoring of chl-a, turbidity, and total suspended matter in coastal waters: Shaik, I., Mohammad, S., Nagamani, P.V., Begum, S.K., Kayet, N., Varaprasad, D., 2021.
experience of the year 2017 along the French Coasts. J. Mar. Sci. Eng. 8 (665) Assessment of Chl-a retrieval algorithms over Kakinada and Yanam turbid coastal
https://doi.org/10.3390/jmse8090665. waters along east coast of India using Sentinel-3A OLCI and Sentinel-2A MSI sensors.
Ha, N.T.T., Thao, N.T.P., Koike, K., & Nhua, M.T. (2017). International Journal of Geo- Remote Sens. Appl. Soc. Environ. 24 (2021), 100644.
Information 6(290):1-15. 10.3390/ijgi6090290. Soja-Woźniak, M., Laiolo, L., Baird, M., Matear, R., Clementson, L., Schroeder, T.,
Hakanson, L., Bryann, A.C., 2008. Eutrophication in the Baltic Sea Present Situation, Doblin, M.A., Suthers, I.M., 2020. Effect of phytoplankton community size structure
Nutrient Transport Processes, Remedial Strategies. Springer-Verlag, Berlin on remote-sensing reflectance and chlorophyll a products. J. Mar. Syst. 211
Heidelberg, p. 261. (103400) https://doi.org/10.1016/j.jmarsys.2020.103400.
Hutasuhut, A., Ismanto, A., Rochaddi, B., Maslukah, L., Widiaratih, R., 2022. Sediment Subiyanto, S., 2018. Remote sensing and water quality indicators in the west flood canal
suspension distribution models in east canal flood estuary waters, Semarang, Central Semarang city: spatio-temporal structures of lansat-8 derived Chl-a and total
Java, Indonesia. Trop. Aquat. Soil Pollut. 2 (2), 76–89. https://doi.org/10.53623/ suspended slids. IOP Conf. Ser. Earth Environ. Sci. 98, 1–10.
tasp.v2i2.93. Tilstone, G.H., Pardo, S., Dall’Olmo, G., Brewin, R.J.W., Nencioli, F., Dessailly, D.,
Katlane, R., Du-pouy, C., El Kilani, B., Berges, J.C., 2020. Estimation of chlorophyll and Kwiatkowska, E., Casal, T., Donlon, C., 2021. Performance of ocean colour
turbidity using sentinel 2A and EO1 data in kneiss archipelago Gulf of Gabes, chlorophyll a algorithms for Sentinel-3 OLCI, MODIS-Aqua and Suomi-VIIRS in
Tunisia. Int. J. Geosci. 11, 708–728. https://doi.org/10.4236/ijg.2020.1110035. open-ocean waters of the Atlantic. Remote Sens. Environ. 260 (2021), 11244.
Lewis, K.M., Arrigo, K.R., 2020. Ocean color algorithms for estimating chlorophylla, https://doi.org/10.1016/j.rse.2021.112444.
CDOM absorption, and particle backscatteringin the Arctic Ocean. JGR Oceans 125 Tran, M.D., Vantrepotte, V., Loisel, H., Oliveira, E.N., Tran, K.T., Jorge, D., Mériaux, X.,
(6), e2019JC015706. https://doi.org/10.1029/2019JC015706. Paranhos, R., 2023. Band ratios combination for estimating Chl-a from Sentinel-2
Maslukah, L., Zainuri, M., Wirasatriya, A., Salma, U., 2019. Spatial distribution of Chl-a and Sentinel-3 in coastal waters. Remote Sens. 2023 15, 1653. https://doi.org/
and its relationship with dissolved inorganic phosphate influenced by rivers in the 10.3390/rs15061653.
North Coast of Java. J. Ecol. Eng. 20 (7), 18–25. https://doi.org/10.12911/ Utama, I.M.R.P., Maslukah, L., Wulandari, S.Y., 2021. Distribution of suspended solids
22998993/108700. and phosphate materials in the waters of Semarang, Central Java. J. Mar. Res. 10 (1),
Maslukah, L., Zainuri, M., Wirasatriya, A., Maisyarah, S., 2020. The Relationship among 89–96. https://doi.org/10.14710/jmr.v10i1.28533 (in Bahasa).
dissolved inorganic phosphate, particulate inorganic phosphate, and Chl-a in Warren, M.A., Simisa, S.G.H., Martinez-Vicentea, V., Poserb, K., Brescianic, M.,
different seasons in the coastal seas of Semarang and Jepara. J. Ecol. Eng. 21 (3), Alikasd, K., Spyrakose, E., Giardinoc, C., Ansper, A., 2019. Assessment of
135–142. https://doi.org/10.12911/22998993/118287. atmospheric correction algorithms for the Sentinel-2A multispectral imager over
Maslukah, L., Setiawan, R.Y., Nurdin, N., Zainuri, M., Wirastriya, A., Helmi, M., 2021. coastal and inland waters. Remote Sens. Environ. 225, 267–289. https://doi.org/
Estimation of Chl-a phytoplankton in the coastal waters of Semarang and Jepara for 10.1016/j.rse.2019.03.018.
monitoring the eutrophication process using MODIS-AQUA imagery and Watanabe, F.S.Y., Alcântara, E., Stech, J.L., 2018. High performance of Chl-a prediction
conventional methods. J. Ecol. Eng. 22 (1), 51–59. https://doi.org/10.12911/ algorithms based on simulated OLCI Sentinel-3A bands in cyanobacteria-dominated
22998993/108700. inland waters. Adv. Space Res. 62, 265–273. https://doi.org/10.1016/j.
Matthews, M.W., Bernard, S., Robertson, L., 2012. An algorithm for detecting trophic asr.2018.04.024.
status (Chl-a), cyanobacterial dominance, surface scums and floating vegetation in Wirasatriya, A., Prasetyawan, I.B., Triyono, C.D., Muslim, Maslukah, L., 2019. Effect of
inland and coastal waters. Remote Sens. Environ. 2012 (124), 637–652. https://doi. ENSO on the variability of SST and chlorophyll a in Java Sea. IOP Conf. Ser. Earth
org/10.1016/j.rse.2012.05.032. Environ. Sci. 116, 012063 https://doi.org/10.1088/1755-1315/116/1/012063.
Moutzouris-Sidiris, I., Topouzelis, K., 2021. Assessment of chlorophyll - a concentration Wirasatriya, A., Susanto, R.D., Setiawan, J.D., Ramdani, F., Iskandar, I., Jalil, R.,
from Sentinel-3 satellite images at The Mediterranean Sea using CMEMS open source Puryajati, A.D., Kunarso, K., Maslukah, L., 2021. High chlorophyll-a areas along the
in situ data. Open Geosci. 13 (1), 85–97. https://doi.org/10.1515/geo-2020-0204. western coast of south Sulawesiindonesia during the rainy season Revealed by
Munandar, B., Wirasatriya, A., Sugianto, D.N., Susanto, R.D., Purwandana, A., Kunarso, satellite data. Remote Sens. 13 (4833), 1–12. https://doi.org/10.3390/rs13234833.
2023. Distinct mechanisms of Chl-a blooms occur in the Northern Maluku Sea and Wirasatriya, A., Maslukah, L., Indrayanti, E., Yusuf, M., Milenia, A.P., Adam, A.A.,
Sulu Sill revealed by satellite data. Dyn. Atmos. Oceans 102, 101360. https://doi. Helmi, M., 2023. Seasonal variability of total suspended sediment off the Banjir
org/10.1016/j.dynatmoce.2023.101360. Kanal Barat River, Semarang, Indonesia estimated from Sentinel-2 images. Reg. Stud.
Nelson, N.B., Siegel, D.A., 2002. Chromophoric DOM in the Open Ocean. In Mar. Sci., 102735 https://doi.org/10.1016/j.rsma.2022.102735.
Biogeochemistry of Marine Dissolved Organic Matter. Elsevier, pp. 547–578. Yoon, J.-E., Lim, J.-H., Son, S., Youn, S.-H., Oh, H.-J., Hwang, J.-D., Kwon, J.-I., Kim, S.-
Nuriya, H., Hidayah, Z., Nugraha, W.A., 2016. The assessment of chlorophyll-a S., Kim, I.-N., 2019. Assessment of satellite-based chlorophyll-a algorithms in
concentration with Landsat Etm-7 images processing and laboratory test in the eutrophic Korean coastal waters: Jinhae bay case study. Front. Mar. Sci. 6, 359.
Western Madura Sea. J. Kelaut. 3 (1), 60–65.