ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-3-2022

XXIV ISPRS Congress (2022 edition), 6–11 June 2022, Nice, France

COMBINED FLOODING AND WATER QUALITY MONITORING DURING SHORT

EXTREME EVENTS USING SENTINEL 2: THE CASE STUDY OF GLORIA STORM IN
EBRO DELTA

E. Angelats 1 *, J. Soriano-González 1,2, M. Fernández-Tejedor 2, C. Alcaraz 2

1 Centre Tecnològic de Telecomunicacions de Catalunya (CTTC/CERCA), Av. Carl Friedrich Gauss, 7. Building B4, 08860,
Castelldefels, Spain - (eduard.angelats, jesus.soriano)@cttc.cat
2 IRTA Marine and Continental Waters, Carretera Poble Nou Km 5.5, 43540, la Ràpita, Tarragona, Spain - (margarita.fernandez,

carles.alcaraz)@irta.cat

KEY WORDS: Disaster management, Monitoring, Flooding, Water Quality, Change detection, Sentinel-2.

ABSTRACT:

Short extreme events have significant impact on landscape and ecosystems in low-lying and exposed areas such as deltaic systems. In
this context, this paper proposes a combined methodology for the mapping and monitoring of the flooding and water quality dynamics
of coastal areas under extreme storms from Sentinel 2 imagery. The proposed methodology has been applied in a coastal bay of the
Ebro Delta (Catalonia, NE Spain) to evaluate jointly the impact of Gloria storm (January 2020) in land-flooding and water quality. The
experimental results show that the Gloria storm had a strong morphological impact and altered the water quality (chl-a) dynamics. The
results show a recovery in terms of water quality after some weeks but in contrast the coastal morphology did not show the same degree
of resilience. This paper is the first step of an overall goal that is to set the bases in a long term, for a workflow for rapid response and
continuous monitoring of storm effects in coastal areas and/or highly valuable ecosystems such as the Ebro Delta.

1. INTRODUCTION affects most of the Ebro Delta surface and induces morphological
disturbance in the coastal front and bays (Jiménez et al., 2011).
The nature of coastal areas makes them highly dynamic and Furthermore, increased discharge of the Ebro River, land run-off,
susceptible to extreme variations under anomalous weather and sediment suspension and transport in coastal waters suppose
circumstances. Particularly, low-lying areas such as deltaic a direct impact on the aquatic ecosystems and related economic
systems are highly vulnerable to extreme storms. The impact of activities such as fisheries or aquaculture (Berdalet et al., 2020).
the storms, and associated hazards (e.g. wave-driven, sea-level
rise hazards) induces morphodynamic processes such as beach This, together with the expected exacerbation of these trends
erosion or overwash (Jiménez et al., 2012) and flooding episodes under a climate change scenario has led to several studies
that may affect large extensions where multiple human uses assessing the effect of these events in the area. Some studies have
coexist (e.g. urban areas, agri-food industry, ecosystem services). been focused on reporting and documenting the events with
In addition to the physical impacts in the terrestrial ecosystem, social media and media pictures, visual interpretation of airborne
significant rainfall and strong winds related to extreme storms imagery, and reporting of the consequences (Jiménez et al., 2020,
increase run-off and turbulence in coastal waters, thus with an González et al., 2020). Alternatively, most scientific research
expected impact on water quality and aquatic ecosystems efforts have been focused on the investigation of flooding risk
(Hernández et al., 2020). A better understanding of the relation and coverage, and morphodynamics through numerical model
between storm events and landscape/ecosystem development is simulations under storm and climate change scenarios, and do not
needed. This is a key step for improving territorial and include reports on water quality dynamics (Amores et al., 2020,
emergency management strategies on storm-induced hazards Garcia et al., 2013, Alvarado-Aguilar et al., 2012). All this
assessment, preparedness, response, and environmental information is a highly valuable source of knowledge for
resilience monitoring. assessing the consequences of storm-induced hazards in the delta
contributing to improving risk, land, and emergency management
The Ebro Delta is an excellent case of study because supports strategies. However, there is still a gap for improvement in terms
important terrestrial and aquatic ecosystems, as well as economic of observation-driven methods for the monitoring of storm-
activities sensitive to extreme storms. It is a subsident area with induced hazards, complementing and/or enhancing the current
an unprotected coastline enclosing two coastal bays with high approaches in the area. Mainly, on the assessment of flooding
ecological and socio-economical value. The importance of storm events and associated morphodynamic changes under high-
hazards in the Ebro Delta in terms of magnitude and recurrency intensity storms over large areas, and the monitoring of water
has been increasing since the 90s with most hazardous storms quality parameters (e.g. Total Suspended Solids (TSS) or
characterized by low-pressure systems off the Ebro Delta coast chlorophyll-a concentration (chl-a)), when access to in situ
and eastern wave storms, usually accompanied by storm surges measurements may be limited. From this perspective, spaceborne
and heavy rainfall. Such events lead to a compound flooding that multispectral remote sensing is a unique source of data as it

* Corresponding author

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper.
https://doi.org/10.5194/isprs-annals-V-3-2022-361-2022 | © Author(s) 2022. CC BY 4.0 License. 361
 ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-3-2022
XXIV ISPRS Congress (2022 edition), 6–11 June 2022, Nice, France

provides a synoptic view over large areas and frequent systematic overall goal that is to set the bases in a long term, for a workflow
observations. for rapid response and continuous monitoring of storm
emergencies in low deltaic and highly vulnerable areas such as
Among remote sensing platforms, the Sentinel-2 multispectral the Ebro Delta.
imagery satellites (S2) operated by the European Spatial Agency
(ESA) represent a breakthrough in terms of flooding and water 2. MATERIALS AND METHODS
quality monitoring. The S2 MultiSpectral Imagery (MSI) covers
the VIS-NIR and SWIR regions of the spectrum at 10, 20, or 60 2.1 Study area
m spatial resolution (depending on the spectral band) with a
frequency of 5 days, improving temporal and spatial resolutions
of prior similar platforms such as Landsat. Multispectral data or The Ebro Delta (Catalonia, NE Spain), with a total extension of
optical imagery can be used for both flood and water quality 32.500 ha is one of the largest wetlands in the Western
mapping (Caballero et al., 2020, Sobel et al., 2020, Goffi et al., Mediterranean region (Figure 1). It supports highly dynamic and
2020, Cavallo et al., 2021). productive ecosystems with great ecological and economical
value. Several protected areas (e.g. RAMSAR sites, UNESCO
Detection of flooding and land surface water change based on biosphere reserve, the Natura 2000 network) coexist with a
multispectral satellite imagery generally relies on the use of number of anthropic activities. Agriculture is the core of the
water-related indicators proposed as combinations in the VIS, economy with ~ 65 % devoted to rice production (Genua-Olmedo
NIR, or SWIR spectral regions in form of spectral index (Pekel et al., 2016). Inside Ebro Delta bays, the production of bivalves
et al., 2014). Overall, combinations of VIS/SWIR and constitutes also a major economic activity, although fisheries and
NIR/SWIR spectral bands provide for a clear diagnostic of water exploitation of saltworks are also part of the diverse economic
surfaces, showing positive values for flood conditions and activities in the area (Figure 1).
negative values for soil (both wet and dry) (Boschetti et al.,
2014). However, as the selected index response varies as a
function of the study area, water characteristics, environmental
and atmospheric conditions, the combination of multi-source
data is recommended for improving classification accuracy
(Acharya et al., 2018). Multiparameter classification methods
aim to reinforce the characterization by exploiting redundancy or
complementarity provided by multiple spectral features (Goffi et
al., 2020). Therefore, the integration of different band
combinations has been reported to improve the accuracy of land
cover mapping (including flooding and classification of dry and
mixed surfaces) through different segmentation techniques (e.g.
Cavallo et al., 2021, Tavus et al., 2020).

Extensive research can be found related to the retrieval of water

quality parameters such as TSS or chl-a from remote sensing
reflectances (Rrs). The common approaches rely on empirical
and/or bio-optical algorithms that exploit spectral signatures of
different water constituents, along with different band
combinations depending on the water optical properties (Moore,
et al., 2014, Caballero, et al., 2020). All of them use a set of in-
situ values to calibrate and validate the different models. The
performance of the different algorithms depends mainly on the
type of water and the accuracy of the atmospheric correction, but
also on the dynamic range of concentrations of water quality
parameters within the area of study. Some research has been done
in relation to the monitoring of the impact of short extreme events
on water quality. (Sobel et al., 2020) used S2 imagery to evaluate
the impact of several Hurricanes in the Gulf of Mexico in terms
of TSS concentration. Several studies suggest that water quality
recovers relatively fast from severe events (Chen et al., 2017,
Huang et al., 2011) while others show long-term impacts (Beaver
et al., 2013, Wetz and Yoskiwitz, 2013).

This study presents a methodology for the monitoring of the

flooding and water quality (i.e. chl-a) dynamics from S2 imagery
under extreme storms in coastal areas. In particular, the proposed
methodology has been applied in the bays of the Ebro Delta
(Catalonia, NE Spain) to jointly evaluate the impact of land- Figure 1. Summarized depiction of the study area
flooding and water quality in Alfacs Bay during and after the
Gloria storm (January 2020). The study aims to demonstrate how From a topographic point of view, Ebro Delta is a low-lying
S2 imagery together with appropriate modelling can be used as a coastal area with sandy, silty and sandy, and fine soils. However,
tool for assessing the ecosystem response of a coastal area in the an abrupt decrease in Ebro River sediments discharge and the
aftermath of an extreme climatic event, through an integrated accumulative subsidence have reversed the growing trend of the
approach. Moreover, this study presents the first steps of an Ebro Delta surface during the last century. The Ebro Delta

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper.
https://doi.org/10.5194/isprs-annals-V-3-2022-361-2022 | © Author(s) 2022. CC BY 4.0 License. 362
 ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-3-2022
XXIV ISPRS Congress (2022 edition), 6–11 June 2022, Nice, France

currently presents a wave-dominated coast with strong reshaping for defining water (positive NDWIXU ) and non-water classes
processes (Sayol and Marcos, 2018), highly exposed to storms (negative NDWIXU) resulting in binary maps. This method aimed
and extreme events, subjected to massive flooding, beach to first differentiate 4 classes which may be water or non-water
erosion, and overwash episodes (Jiménez et al., 2020). From the and then solve the uncertainty of single images classification by
perspective of water quality, it highlights the importance of the time-series analysis of different classes accounting for different
semi-enclosed bays formed by the northern and southern hemi- types of water and surface state. It was implemented using a
deltas (Figure 1). Both coastal bays have a relatively small customized script developed for R software v3.6 (R Core Team,
surface, shallow waters (0-7 m), and narrow mouths connecting 2017).
them to the Mediterranean Sea. The bays harbour important
populations of seagrass and benthic organisms, as well as 2.4 Chl-a mapping
fisheries and shellfish farming activities, which are sensitive to
water quality deterioration due to storm episodes. The chl-a maps during the Gloria storm are part of a longer time
series (October 1st 2019, to September 30th 2021) in the area of
The results presented in this paper are focused on Alfacs Bay Ebro Delta (Figure 1 top). This subsection summarizes the
(Figure 1), the southern coastal bay where many of the threats methodology used to generate this time series of chl-a maps.
driven by storms in the Ebro Delta are observed. Alfacs bay
concentrates aquaculture, an extremely vulnerable piece of coast All available S2 L1C orbit R051 images during the two-year
heavily affected by erosion (Trabucador barrier), and La Banya period of interest were downloaded from the Copernicus Services
spit, a special Protection Area for birds including a dune system Data Hub (https://cophub.copernicus.eu/dhus/#/home). After
that is part of Ebro Delta Natural Park. image-by-image visual checking for clouds and shadows over the
Ebro Delta coastal waters, the valid images (Nimages = 210) were
2.2 Gloria storm pre-processed. This step included the resampling of the bands to
20 m pixel, the subset to the extent of the area of interest, and the
Gloria storm was a marine storm affecting the Ebro Delta region atmospheric correction specific for coastal waters, all
in January 2020 (19.01.20 – 23.01.20). It has been classified as implemented through the Graph Processing Tool of SNAP v8.0.
one of the most intense among the events in the region during the The atmospheric correction was done with the latest version of
last decades with immediate economic losses of several millions the Case 2 Regional Coast Colour processor (Doerffer, et al.,
of euros. During the event, the situation was dominated by a 2007, Brockman et al., 2016), here referred as to C2XC. It was
southern deep low-pressure system that generated sea level applied on all valid S2 images, with a fine-tuning of daily
surface elevation (up to 1 m) and strong easterly winds which auxiliary data (pressure, ozone, altitude, salinity, and water
induced wind waves (Amores et.al, 2020) with high significant temperature). For land/water segmentation, the valid pixel
height (maximum Hs > 7m) and heavy rainfall on land expression was set as a threshold on band B11 of S2 L1C images.
(maximum cumulative precipitation > 130 mm). The Gloria The threshold was defined independently for each image with the
storm had devasting effects on urban areas, agriculture fields, the histogram-based triangle method. The output of C2XC included
natural environment, and different crucial infrastructures such as Rrs at VIS-NIR S2 bands.
roads, becoming the most hazardous event in Ebro Delta in the
last decades. Field measurements (N = 146) were conducted along 26 dates, at
different locations including waters in Alfacs bay, Fangar bay,
2.3 Flooding mapping and the surrounding exterior Mediterranean sea. The
measurements corresponded to integrated water column samples,
A set of 11 S2A/B L2A (atmospherically corrected surface taken within a window of ±3 hours of S2 pass. For chl-a
reflectance) cloud-free images were visually filtered and determination, water samples were filtered through 0.4-0.6 µm
downloaded from Copernicus Open Acces Hub GF/F glass fiber filters, extracted using standard methods (Shoaf
(https://scihub.copernicus.eu/) accounting for dates pre-, during, and Lium, 1976), and calculated with Jeffrey and Humphrey
and post-Gloria storm. All S2 L2A images were subset to the (1975) equations.
Ebro Delta area (Figure 1) and resampled to 20 m pixel using the
Sentinel Application Platform (SNAP v8.0). The same tool was Then, the match-up exercise between field-measured chl-a data
then used for deriving two variants of Normalized Difference and C2XC Rrs was done using a quality-check procedure based
Water Index (NDWI), namely, NDWIGAO (Gao et al., 1996) and on 3x3 pixel-window centered at field measurements. The quality
NDWIXU (Xu et al., 2006). The NDWIGAO combines NIR (B8 check included the removal of pixels flagged inherently by
band centered at 832 nm) and SWIR (B11 band centered at 1614) C2XC, an additional outlier filtering using boxplot analysis and,
bands of S2 (eq. 1), while the NDWIXU applies a VIS/SWIR the removal of negative Rrs pixels. Only pixel windows with
approach combining Green (B3 band centered at 560 nm) and more than 4 pixels remaining were accepted for posterior
SWIR (B11) S2 bands (eq. 2). analysis. After this step and the rejection of cloud-affected S2
images, the number of valid field chl-a data was 73, ranging
between 0.1 – 8 mg/m3, along 14 different dates.
𝐵8 −𝐵11
NDWIGAO = (1)
𝐵8 +𝐵11
For chl-a estimation, a set of 11 band combinations (2 to 4
𝐵3 −𝐵11 spectral bands) in form of spectral indices (SI) were computed
NDWIXU = (2) for all valid match-ups. These SI included bands B1 (centered at
𝐵3 +𝐵11
443 nm) to B6 (centered at 740 nm) and exploit specific chl-a
These indexes were used later as input in an unsupervised k- absorption peaks in Blue and Red spectral regions, and/or use
means clustering to generate 4 classes for each image. Then, the Red Edge information for mitigating the effect of absorption by
mean and standard deviation of pixels within each class and SI (4 non-algal particles and yellow substances. For instance, common
classes × day-1 × number of dates; N = 44) were extracted and ocean colour algorithms based on Blue to Green ratios (O'Reilly
reclassified into two classes by means, again, of unsupervised k- and Werdell, 2019), the three-band model of Gitelson et al.
means clustering. The NDWIXU cluster centers were then used (2011), and the Normalized Difference Chlorophyll Index

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper.
https://doi.org/10.5194/isprs-annals-V-3-2022-361-2022 | © Author(s) 2022. CC BY 4.0 License. 363
 ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-3-2022
XXIV ISPRS Congress (2022 edition), 6–11 June 2022, Nice, France

(Mishra and Mishra, 2012) were included among the tested SI. great concentrations of TSS is expected. Empirical algorithms, as
Chl-a was modelled with all of them. A 70 % of chl-a data was the one used, can perform well only inside their range of
used for model calibration (cal) and 30 % for validation (val). derivation and for the area for which they were derived. They are
The cal/val datasets were generated randomly. Models were more limited in their ability to discriminate between non-unique
developed for raw and log-transformed (when possible) Rrs and signals from parameters that may be covariant with chl-a, for
chl-a data (i.e. raw-raw, log-raw, log-log). Linear, linear instance, TSS (Matthews et al. 2012), which increases
piecewise (1 breakpoint), polynomial (2nd, 3rd, and 4th order), particularly during storm events. This could not be validated
logarithmic, power, and exponential models were tested. All the because no chl-a nor TSS data were available in such cases, but
process was repeated through 100 iterations, with varying cal/val it was observed in RGB composites that after strong winds or
datasets. For each model (Band Combination × Type of Fitting × storm events, the water of the bays turns darker brownish or
Iteration), performance was evaluated by means of the following whitish, suggesting higher TSS (e.g. in Figure 3 and 4B). In these
statistics: the Mean Average Error (MAE), the Root Mean cases, the selected model will tend to overestimate chl-a since the
Squared Error (RMSE), the Average Percentage Difference reflectance on B5 increases with great TSS, overlapping the
(APD), the BIAS (eq.2-5) and the Pearson’s r (r). The statistics absorption of chl-a in the Blue (Soriano et al., 2019) In these
were computed including calibration and validation datasets. cases, other empirical or semi-analytical algorithms may be more
accurate. However, it is important to highlight that the proposed
To generate the chl-a maps, all flagged pixels (C2XC flagging) methodology for chl-a retrieval is easy to replicate, and
or pixels with negative Rrs in any of the spectral bands of each straightforward using several software tools such as SNAP,
valid S2 image were removed. A mask was applied for avoiding which may be important for a wide part of the scientific
most shallow waters and the presence of seagrass, which make community related with ecosystems monitoring and modelling
the retrieval of chl-a more uncertain. Then, the model with the response but with few or no remote sensing skills.
best performance was applied to all valid images and pixels.
Those with negative chl-a value (residual) were set to 0 mg/m3 Nevertheless, further research should move towards a multi-
and, pixels with chl-a > 30 mg/m3 were removed (noData) given algorithm combination, suitable across different types of water
the higher uncertainty in so high concentrations. Pixels optical properties. The development of a blending algorithm
corresponding to the mussel rafts were masked out with a rafts’ requires accurately defining the optical water type (Moore et al.,
shapefile (Figure 1). 2014) accounting for the spectral shape, magnitude, and
distinctive Rrs spectral features. Furthermore, a greater amount
1
𝑀𝐴𝐸 = ∑𝑁
𝑖=1 |𝑀𝑖 − 𝑂𝑖 | (2) of field data is needed, covering all different scenarios, which
𝑁 were not represented in the current chl-a dataset.

𝑁
(𝑀𝑖 − 𝑂𝑖 )2
𝑅𝑀𝑆𝐸 = √∑ (3)
𝑖=1 𝑁

𝑁
|𝑀𝑖 −𝑂𝑖 |
𝐴𝑃𝐷 = ∑ ∗ 100 (4)
𝑖=1 𝑂𝑖

1
𝐵𝐼𝐴𝑆 = ∑𝑁
𝑖=1(𝑀𝑖 − 𝑂𝑖 ) (5)
𝑁

Where 𝑀𝑖 is the estimated value and 𝑂𝑖 the observed (i.e. field

measured) value.

3. RESULTS AND DISCUSSION

3.1 Chl-a estimates

Figure 2. Linear regression of estimated and measured chl-a.
Overall, models that performed better were related to Rrs bands
combinations including B2 (centered at 490 nm) or B3, and B4
(centered at 665 nm) or B5 (centered at 705 nm). The best model
3.2 Flooding mapping
for chl-a estimation was found with a simple B5/B2 band ratio
fitted with a 2nd polynomial function (eq. 6). The performance Systematic validation of flooding maps was not conducted, albeit
metrics of the model were MAE = 0.60 mg/m3, APD = 36.22 %, a visual comparison of the resulting maps showed good
RMSE = 0.89 mg/m3, BIAS = 0.05 mg/m3 and, r = 0.89. agreement with what can be interpreted from the 10 m RGB
composites of S2 in Alfacs bay (e.g. in Figure 3). The method
allowed us to assess the evolution of the main morphological
𝐵5 𝐵
phenomena that occurred during the Gloria storm in the bay, the
𝑐ℎ𝑙 − 𝑎 = −0.62 + 10.88 ∗ + 8.07 ∗ ( 5)2 (6) breaching of the Trabucador barrier. From visual interpretation,
𝐵2 𝐵2
the main limitations of the method arise from; i) small features
(e.g. shellfish rafts, borders between dry and flooded surfaces)
Despite the implemented model achieved good agreement within which may be misclassified due to the pixel mixing with the
the full range of derivation ~ 0 - 8 mg/m3 (Figure 2), flooded surrounding environment; ii) increasing uncertainty in
overestimation of retrieved chl-a concentrations in scenarios with presence of emerged or submerged vegetation and continuous

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper.
https://doi.org/10.5194/isprs-annals-V-3-2022-361-2022 | © Author(s) 2022. CC BY 4.0 License. 364
 ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-3-2022
XXIV ISPRS Congress (2022 edition), 6–11 June 2022, Nice, France

succession of small emerged and submerged patches (Ozesmi difference between the 15th and 20th days after the storm and the
and Bauer, 2002). image of 5 days before the Gloria storm.

Nevertheless, it is important to remark that the spectral In La Banya spit (Figure 1), the most important effects of Gloria
distribution before and after an extreme even could be quite storm were observed on the eastern side, at the dune fields. In this
different. Additional research is needed for further validating the area, the barchan dunes in the shoreline side showed stability,
accuracy of the proposed methodology. For instance, the with no flooding detected (Figure 5). The inner dunes, however,
minimum number of images to be used must be defined. Also,
showed a more dynamic behavior, alternating flooding, and
the 4 initial k-means classes may be a constraint in more
heterogeneous environments, needing to explore the increasing drying of the northern and southern faces during the period of
of clusters number. Regarding the improvement in the detection study (Figure 5A-B). Where flooding is more intense, the inner
of small features, suitability for resampling to 10 m instead of 20 parts of La Banya spit were subjected to loss of volume
m pixel can be studied, albeit it will strongly depend on the native (Rodríguez-Santalla et al., 2021). The drying out of some of these
resolution of S2 bands. In addition, information on the sea level parts may be explained by the accretion due to sand blown from
simultaneously to the S2 pass must be considered in deeper the beach and dunes and trapped by vegetation growing on the
analysis.
foredune. This phenomenon is observed in Figure 5C, with most
of the dried area concentrated behind barchan dunes. This
highlights the importance of the barchan dunes for maintaining
the shoreline during a storm (Rodríguez-Santalla et al., 2021).

Figure 3. Delineation of emerged land and water surface two

weeks after Gloria storm (05.02.20) with the proposed
methodology (black line) superposed to the S2 RGB composite.
Although that, the proposed method does not depend on large
time series, avoids the difficulty of finding stable thresholds as in
decision trees or rule-based approaches, is fast and unsupervised,
and requires simple computation, allowing for rapid mapping in
the aftermath of a storm, which is a key pillar in the management
of extreme events.

3.3 Chl-a and morphology response to Gloria storm in

Alfacs bay

For assessing the dynamics of water quality and flooding in the

Alfacs Bay, three status maps (Figure 4) and three ‘change maps
(Figure 5) were generated within a period of one month
encompassing the Gloria storm (21.01.20).

The Gloria Storm caused breaching of the Trabucador barrier,

without showing signs of recovery after 20 days (Figure 4B). The
thin line that seems to reconnect the delta with La Banya spit
(Figure 4C and 5B) was the ongoing reconstruction of the road
for in/out of the trucks transporting salt from the saltworks
(Figure 1). La Banya spit remained disconnected from the delta
during all the period shown (Figure 4C and 5C), preventing the
passage of vehicles and the transport of goods to and from the
saltworks. It can be observed in the change maps that the Figure 4. Water quality and flooding mapping in the area of
Trabucador barrier breached from the center and receded back Alfacs Bay before and after Gloria Storm. A) 16.01.20; B)
into the bay. The setback was evident in the extremes of the 05.02.20; C) 10.02.20.
barrier as observed in Figures 5A and 5C corresponding to the
Regarding water quality dynamics associated with the Gloria
storm, water turbulence increased, particularly in the inner north-

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper.
https://doi.org/10.5194/isprs-annals-V-3-2022-361-2022 | © Author(s) 2022. CC BY 4.0 License. 365
 ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-3-2022
XXIV ISPRS Congress (2022 edition), 6–11 June 2022, Nice, France

eastern Alfacs Bay, related to the Trabucador breaching (Fig. 4B) 4. CONCLUSIONS
and the shallow shelf formed behind it. This led to increased chl-
a concentrations, although the model could be overestimating This study presents a methodology for both chl-a (in water) and
flooding mapping (on land) based on parallel processing of
them in this scenario, as high TSS concentration is expected in Sentinel 2 multispectral imagery. The methodology has been
this area behind the Trabucador barrier breaching stretch. In the applied to the assessment of impacts of an extreme coastal storm
last date shown (10.01.20 – Figure 4C), chl-a concentrations affecting the Ebro Delta in January 2020. The Ebro Delta is a
were again closer to pre-storm concentrations (Figure 5C). suitable place to apply the methodology developed here since it
Different from the Trabucador barrier, the water quality showed shares common threats with other coastal regions and, more
signs of recovery in a short term. In the long-term, water quality specifically, with other deltas.
dynamics should be revised again since the breaching of the
Despite constraints related to the validation of flooding maps and
Trabucador barrier has important implications in the the uncertainty of chl-a retrieval under optical water types out of
morphodynamics and hydrodynamics of the bay (Gracia et al., the range of development of the model, the results obtained are
2013). These are major factors for the development of promising. The model presented for chl-a mapping showed an
phytoplankton populations related to the modulation of chl-a accuracy greater than 70 % and, the methodology proposed for
inside Alfacs Bay, with a direct effect on the farming of shellfish mapping of land water surface changes demonstrated great
(Soriano et al., 2019). agreement with visual interpretation from 10 m RGB imagery.
The exposed methodological development allowed assessing
changes both in water and land, through a well-explained
processing chain which can be easily implemented almost with
SNAP open software. Both, flooding delineation and water
quality mapping methodologies can be applied fast, allowing for
rapid evaluation of storm-related perturbances on terrestrial and
aquatic ecosystems, with particular importance in the
management of coastal areas. Sentinel 2 demonstrated great
capabilities for complementing current methods and supporting
management in diverse environments, through the potential for
conducting studies with integrated approaches, which is currently
a line to follow for achieving the sustainable development goals.

The research presented in this paper is one of the first steps with
Sentinel 2 of a long-term strategy to understand the ecosystem
response in Ebro Delta both in space and time in front of short
extreme events. Further research should include an improvement
of the current water quality modelling aiming to de-coupling TSS
from chl-a to improve chl-a estimation. The flooding mapping
must be validated, assessing the accuracy of the method, and
arising potential constraints or improvements not included in this
preliminary study. The proposed methodology is planned to be
extended to all Ebro Delta region and tested also in other short
extreme events that occurred recently such as Filomena storm
(January 2021).

ACKNOWLEDGEMENTS

This research was partially funded by the project New-TechAqua

(European Union's Programme H2020, GA 862658). J. Soriano-
González held a pre-doctoral grant funded by by Agència de
Gestió d’Ajuts Universitaris I de Recerca (2020FI_B2 00148).

REFERENCES

Acharya, T.D., Subedi, A., Lee, D.H., 2018. Evaluation of water

indices for surface water extraction in a Landsat 8 scene of Nepal.
Sensors 18, 2580. doi.org/10.3390/s18082580.

Alvarado-Aguilar, D., Jiménez, J.A., Nicholls, R.J. Flood hazard

and damage assessment in the Ebro Delta (NW Mediterranean)
to relative sea level rise. 2012. Natural Hazards 62, 1301–1321
(2012). doi.org/10.1007/s11069-012-0149-x.
Figure 5. Changes in water quality and flooding related to the
Gloria storm. Differences between A) 05.02.20 -16.01.20; B) Amores, A., Marcos, M., Carrió, D. S., Gómez-Pujol, L., 2020.
10.02.20 - 05.02.20; C) 10.02.20 – 16-01.20. Coastal impacts of storm gloria (2020) over the northwestern

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper.
https://doi.org/10.5194/isprs-annals-V-3-2022-361-2022 | © Author(s) 2022. CC BY 4.0 License. 366
 ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-3-2022
XXIV ISPRS Congress (2022 edition), 6–11 June 2022, Nice, France

mediterranean. Natural Hazards Earth Systems Science. 20, Ocean: The Azov Sea case study. Enviromental. Research
1955–1968. doi: 10.5194/nhess-20-1955-2020. Letters. 2011, 6, 1–6. doi.org/10.1088/1748-9326/6/2/024023.

Berdalet, E, Marrasé, C., Pelegrí, J.L, 2020. Resumen sobre la Goffi, A., Stroppiana, D., Brivio, P.A., Bordogna, G., Boschetti,
Formación y Consecuencias de la Borrasca Gloria (19-24 enero M., 2020. Towards an automated approach to map flooded areas
2020). Institut de Ciències del Mar, CSIC, 20 de marzo de 2020, from Sentinel-2 MSI data and soft integration of water spectral
38 pp, doi.org/10.20350/digitalCSIC/12496. features. International Journal Applied Earth Observation and
Geoinfo. 84, 101951. doi.org/10.1016/j.jag.2019.101951.
Boschetti, M., Nutini, F., Manfron, G., Brivio, P.A., Nelson, A.,
2014. Comparative analysis of normalised difference spectral González, M. (coord.), 2020. El temporal Gloria (19-
indices derived from MODIS for detecting surface water in 23/01/2020): Els efectes dels processos geològics sobre el
flooded rice cropping systems. PLoS ONE 9(2): e88741. territori, 122 pàg. Col·lecció: Monografies tècniques 8. Institut
doi.org/10.1371/journal.pone.0088741. Cartogràfic I Geològic de Catalunya.
doi.org/10.13140/RG.2.2.16730.36802.
Beaver, J. R., Casamatta, D. A., East, T. L., Havens, K. E., Gracia, V., García, M., Grifoll, M. and Sánchez-Arcilla, A.,
Rodusky, A. J., James, R. T., Tausz, C. E., Buccier, K. M., 2013. 2013. Breaching of a barrier beach under extreme events. The
Extreme weather events influence the phytoplankton community role of morphodynamic simulations. Journal of Coastal
structure in a large lowland subtropical lake (Lake Okeechobee, Research 65(sp1), 951-956, (1 April 2013).
Florida, USA). Hydrobiologia, 709(1), 213–226. doi.org/10.2112/SI65-161.1.
doi.org/10.1007/s10750-013-1451- 7.
Hernández, W.J., Ortiz-Rosa, S., Armstrong, R.A., Geiger, E.F.,
Brockmann, C., Doerffer, R., Peters, M., Stelzer, K., Embacher, Eakin, C.M., Warner, R.A., 2020. Quantifying the Effects of
S., Ruescas, A., 2016. Evolution of the C2RCC neural network Hurricanes Irma and Maria on Coastal Water Quality in Puerto
for Sentinel 2 and 3 for the retrieval of ocean colour products in Rico using Moderate Resolution Satellite Sensors. Remote
normal and extreme optically complex waters. Proceedings of the Sensing. 2020, 12, 964. doi.org/10.3390/rs12060964.
Living Planet Symposium 2016, Prague, Czech
Republic,Volume ESA SP, 2016. Hinkel, J., Lincke, D., Vafeidis, AT., Perrette, M., Nicholls, RJ.,
Tol, R. S.J., Marzeion, B., Ionescu, C., Levermann, 2014. Coastal
Caballero, I., Fernández, R., Escalante, O.M., Mamán, L., flood damage and adaptation costs under 21st century sea-level
Navarro, G., 2020. New capabilities of Sentinel-2A/B satellites rise. Proceedings of the National Academy of Sciences 2014; 111:
combined with in situ data for monitoring small harmful algal 3292–3297. doi.org/10.1073/pnas.1222469111.
blooms in complex coastal waters. Science Reports, 10, 1–14.
doi.org/10.1038/s41598-020-65600-1. Huang, W., Mukherjee, D., Chen, S., 2011. Assessment of
Hurricane Ivan impact on chlorophyll-a in Pensacola Bay by
Chen, Y., Cebrian, J., Lehrter, J., Christiaen, B., Stutes, J., Goff, MODIS 250m remote sensing. Marine Pollution Bulletin, 62 (3),
J., 2017. Storms do not alter long-term watershed development 490–498. doi.org/10.1016/j.marpolbul.2010.12.010.
influences on coastal water quality. Marine Pollution Bulletin,
122(1–2), 207. doi.org/10.1016/j. marpolbul.2017.06.038. Jeffrey, S.W., Humphrey, G.F., 1975. New Spectrophotometric
Equations for Determining Chlorophylls a, b, C1 and C2 in
Cavallo, C., Papa, M.N., Gargiulo, M., Palau-Salvador, G., Higher Plants, Algae and Natural Phytoplankton. Biochem.
Vezza, P., Ruello, G., 2021. Continuous Monitoring of the Physiol. Pflanz. 167, 191–194. doi.org/10.1016/S0015-
Flooding Dynamics in the Albufera Wetland (Spain) by Landsat- 3796(17)30778-3.
8 and Sentinel-2 Datasets. Remote Sensing. 2021, 13, 3525.
doi.org/10.3390/rs13173525 Jiménez, J.A., Gracia, V., Valdemoro, H.I., Mendoza, E.T.,
Sánchez-Arcilla, A., 2011. Managing erosion-induced problems
Doerffer, R., Schiller, H, 2007. The MERIS Case 2 water in NW Mediterranean urban beaches. Ocean & Coastal
algorithm. International Journal of Remote Sensing, 28. Management 54 (12), 907-918.
doi.org/10.1080/01431160600821127. doi.org/10.1016/j.ocecoaman.2011.05.003.

Fernández-Montblanc, T., Gómez-Enri, J., Ciavola, P., 2020. The Jiménez, J. A., Sancho-García, A., Bosom, E., Valdemoro, H. I.,
Role of Mean Sea Level Annual Cycle on Extreme Water Levels Guillén, J., 2012. Storm-induced damages along the Catalan
Along European Coastline. Remote Sensing. 2020, 12, 3419. coast(NW Mediterranean) during the period 1958–2008.
doi.org/10.3390/rs12203419. Geomorphology,143–144,24–33.
doi.org/10.1016/j.geomorph.2011.07.034.
Gao, B-C, 1996. NDWI - A Normalized Difference Water Index
for Remote Sensing of Vegetation Liquid Water from Space. Jimenez, J.A, 2020. Dinàmica litoral, efectes dels temporals i
Remote Sensing of Environment 7212 (April): 257–66. comportament de les platges. Chapter of: "Sobre el temporal
doi.org/10.1016/S0034-4257(96)00067-3. Gloria (19-23.01.20), els seus efectes sobre el país i el que se'n
deriva: Report de Resposta Ràpida (R3)". Institut d'Estudis
Genua-Olmedo, A., Alcaraz, C., Caiola, N., and Ibáñez, C., 2016. Catalans, 2020, p. 55-72.
Sea Level Rise Impacts on Rice Production: The Ebro Delta as
an Example. Science of the Total Environment 571: 1200–1210. Matthews, M. W., 2011. A Current Review of Empirical
doi.org/10.1016/j.scitotenv.2016.07.136. Procedures of Remote Sensing in Inland and Near-Coastal
Transitional Waters. International Journal Remote Sensing. 32
Gitelson, A., Gao, B.C., Li, R.R., Berdnikov, S., Saprygin, V., (21), 6855–6899. doi.org/10.1080/01431161.2010.512947.
2011. Estimation of chlorophyll-a concentration in productive
turbid waters using a Hyperspectral Imager for the Coastal

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper.
https://doi.org/10.5194/isprs-annals-V-3-2022-361-2022 | © Author(s) 2022. CC BY 4.0 License. 367
 ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-3-2022
XXIV ISPRS Congress (2022 edition), 6–11 June 2022, Nice, France

Mishra, S., Mishra, D.R., 2012. Normalized Difference Inf. Sci. - ISPRS Arch. 43, 641–648. doi.org/10.5194/isprs-
Chlorophyll Index: A novel model for remote estimation of archives-XLIII-B3-2020-641-2020.
chlorophyll-a concentration in turbid productive waters. Remote
Sensing of Environment. 117, 394–406. Xu, H., 2006. Modification of normalised difference water index
doi.org/10.1016/j.rse.2011.10.016. (NDWI) to enhance open water features in remotely sensed
imagery. International Journal of Remote Sensing, 27, 3025-
Moore, T.S., Dowell, M.D., Bradt, S., Ruiz-Verdú, A., 2014. An 3033. dx.doi.org/10.1080/01431160600589179.
optical water type framework for selecting and blending
retrievals from bio-optical algorithms in lakes and coastal waters. Wetz, M. S., Yoskowitz, D. W., 2013. An ‘extreme’ future for
Remote Sensing of Environment. 143, 97–111. estuaries? Effects of extreme climatic events on estuarine water
doi.org/10.1016/j.rse.2013.11.021. quality and ecology. Marine Pollution Bulletin, 69(1), 7–18.
doi.org/10.1016/j.marpolbul.2013.01.020.
O'Reilly J E, Werdell, P J, 2019. Chlorophyll algorithms for
ocean color sensors-OC4, OC5 & OC6. Journal of Remote
sensing of environment, 229: 32-47.
doi.org/10.1016/j.rse.2019.04.021.

Ozesmi, S.L., Bauer, M.E, 2002. Satellite remote sensing of

wetlands. Wetlands Ecology and Management, 10, 381–402.
doi.org/10.1023/A:1020908432489.

Pekel, J.F., Vancutsem, C., Bastin, L., Clerici, M., Vanbogaert,

E., Bartholomé, E., Defourny, P., 2014. A near real-time water
surface detection method based on HSV transformation of
MODIS multi-Spectral time series data. Remote Sensing
Environment. 140, 704–716. doi.org/10.1016/j.rse.2013.10.008.

RCoreTeam, 2017. R: A Language and Environment for

Statistical Computing. Vienna, Austria: R Foundation for
Statistical Computing.

SNAP - ESA Sentinel Application Platform v8.0,

http://step.esa.int.

Rodríguez-Santalla, I., Gomez-Ortiz, D.; Martín-Crespo, T.,

Sánchez, M.J., Montoya-Montes, I., Martín-Velázquez, S.,
Barrio, F., Serra, J., Ramírez-Cuesta, J.M., Gracia, F.J., 2021.
Study and Evolution of the Dune Field of La Banya Spit in Ebro
Delta (Spain) Using LiDAR Data and GPR. Remote Sensing, 13,
802. doi.org/10.3390/rs13040802.

Sayol, J. M., Marcos, M., 2018. Assessing flood risk under sea
level rise and extreme sea levels scenarios. Application to the
Ebro Delta (Spain). Journal of Geophysical Research:
Oceans,123,794–811. doi.org/10.1002/2017JC013355.

Shoaf, W.T., Lium, B.W., 1976. Improved extraction of

chlorophyll a and b from algae using dimethyl sulphoxide.
Limnology and Oceanography. 21, 926-928.
doi.org/10.4319/lo.1976.21.6.0926.

Sobel, R.S., Kiaghadi, A., Rifai, H.S, 2020. Modeling water

quality impacts from hurricanes and extreme weather events in
urban coastal systems using Sentinel-2 spectral data. Environ
Monit Assess 192, 307 (2020). doi.org/10.1007/s10661-020-
08291-5.

Soriano-González, J., Angelats, E., Fernández-Tejedor, M.;

Diogene, J., Alcaraz, C. First Results of Phytoplankton Spatial
Dynamics in Two NW-Mediterranean Bays from Chlorophyll-a
Estimates Using Sentinel 2: Potential Implications for
Aquaculture. Remote Sensing. 2019, 11, 1756.
doi.org/10.3390/rs11151756.

Tavus, B., Kocaman, S., Nefeslioglu, H.A., Gokceoglu, C., 2020.

A fusion approach for flood mapping using sentinel-1 and
sentinel-2 datasets. Int. Arch. Photogramm. Remote Sens. Spat.

This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper.
https://doi.org/10.5194/isprs-annals-V-3-2022-361-2022 | © Author(s) 2022. CC BY 4.0 License. 368