ArticlePDF Available

Post-Nourishment Changes of an Artificial Gravel Pocket Beach Using UAV Imagery

Authors:

Abstract and Figures

Monitoring and analysis of changes in the volume and area of nourished beaches is crucial to inform any beach renourishment programme. The aim of this study is to utilise UAV surveys and SfM photogrammetry to assess the beach nourishment performance of an artificial gravel beach exposed to a range of external forcing, including storms. The paper presents results from nineteen UAV surveys conducted between January 2020 and January 2021 at Ploče, an artificial beach in Rijeka (Croatia). The beach was nourished twice and eleven storm events, ranging from weak to strong, were recorded during this period. The Agisoft Metashape software was used to obtain point clouds and digital elevation models (DEM) from UAV images; Matlab and CloudCompare were used for further analysis of the DEMs. The accuracy and precision of the DEMs was assessed and uncertainty levels of ±5 cm were applied to all derived DEMs. The study provides new insights into the response of the emerged part of the beach to storms. Predictably, the largest changes were recorded after the first storm following beach nourishment. The longshore variability in the beach response to storms was identified from full 3D point clouds. Most of the lost sediment was from the east side of the beach, while the rest of the beach aligned with the predominant wave direction through cross-shore and longshore processes. Offshore/onshore sediment exchange between the lower and upper beach face on the western side manifested itself in beach profile steepening and berm formations. Overall, changes in beach volume and area were small, indicating that this artificial beach is relatively stable. The embayed layout following the natural coastal configuration appears to be effective in retaining nourished sediment on the beach. This work highlights the need to consider pocket embayed beaches in three dimensions, as traditional transect studies can overlook the three-dimensional behaviour. This study also highlighted the wider potential of UAVs and SfM for studies of high-resolution elevation changes on natural and artificial beaches, as well as for coastal monitoring of beach nourishment.
Content may be subject to copyright.


Citation: Tadi´c, A.; Ruži´c, I.;
Krvavica, N.; Ili´c, S.
Post-Nourishment Changes of an
Artificial Gravel Pocket Beach Using
UAV Imagery. J. Mar. Sci. Eng. 2022,
10, 358. https://doi.org/10.3390/
jmse10030358
Academic Editors: Duccio Bertoni
and Alessandro Pozzebon
Received: 31 December 2021
Accepted: 28 February 2022
Published: 3 March 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Journal of
Marine Science
and Engineering
Article
Post-Nourishment Changes of an Artificial Gravel Pocket Beach
Using UAV Imagery
Andrea Tadi´c 1, Igor Ruži´c 1, * , Nino Krvavica 1and Suzana Ili´c 2
1Faculty of Civil Engineering, University of Rijeka, 51000 Rijeka, Croatia; andrea.tadic@gradri.uniri.hr (A.T.);
nino.krvavica@uniri.hr (N.K.)
2Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YW, UK; s.ilic@lancaster.ac.uk
*Correspondence: iruzic@uniri.hr
Abstract:
Monitoring and analysis of changes in the volume and area of nourished beaches is crucial
to inform any beach renourishment programme. The aim of this study is to utilise UAV surveys
and SfM photogrammetry to assess the beach nourishment performance of an artificial gravel beach
exposed to a range of external forcing, including storms. The paper presents results from nineteen
UAV surveys conducted between January 2020 and January 2021 at Ploˇce, an artificial beach in
Rijeka (Croatia). The beach was nourished twice and eleven storm events, ranging from weak to
strong, were recorded during this period. The Agisoft Metashape software was used to obtain point
clouds and digital elevation models (DEM) from UAV images; Matlab and CloudCompare were
used for further analysis of the DEMs. The accuracy and precision of the DEMs was assessed and
uncertainty levels of
±
5 cm were applied to all derived DEMs. The study provides new insights
into the response of the emerged part of the beach to storms. Predictably, the largest changes were
recorded after the first storm following beach nourishment. The longshore variability in the beach
response to storms was identified from full 3D point clouds. Most of the lost sediment was from
the east side of the beach, while the rest of the beach aligned with the predominant wave direction
through cross-shore and longshore processes. Offshore/onshore sediment exchange between the
lower and upper beach face on the western side manifested itself in beach profile steepening and
berm formations. Overall, changes in beach volume and area were small, indicating that this artificial
beach is relatively stable. The embayed layout following the natural coastal configuration appears
to be effective in retaining nourished sediment on the beach. This work highlights the need to
consider pocket embayed beaches in three dimensions, as traditional transect studies can overlook
the three-dimensional behaviour. This study also highlighted the wider potential of UAVs and SfM
for studies of high-resolution elevation changes on natural and artificial beaches, as well as for coastal
monitoring of beach nourishment.
Keywords:
unmanned aerial vehicle (UAV); structure-from-motion (SfM) photogrammetry; beach
nourishment; gravel pocket beach (GPB); point cloud; beach changes
1. Introduction
Artificial beaches, defined as beaches where ‘the shoreline has been moved seaward to
a point never been reached naturally before or during the last century’ [
1
] have been built
all along the world’s coastlines. Many are built for recreational purposes (e.g., [
1
3
]) but
also for protection from coastal erosion [4,5] and flooding [1].
Understanding the processes that affect their stability and performance in protecting
the coastline is important for coastal practitioners involved in design and maintenance of
these beaches. Monitoring provides vital information on beach responses to a range of
environmental conditions.
There are many factors that affect longevity of beach nourishments such as exposure
to waves and in particular storm waves, their shape and other geometric parameters such
J. Mar. Sci. Eng. 2022,10, 358. https://doi.org/10.3390/jmse10030358 https://www.mdpi.com/journal/jmse
J. Mar. Sci. Eng. 2022,10, 358 2 of 24
as length [
6
,
7
], width and slope, and quality, density and grain size of fill material. Their
stability is often enhanced by groynes, artificial headlands [
8
] and submerged sills [
9
,
10
].
As these beaches are built in environments where coastal erosion is dominant or local
environmental conditions do not support the formation of natural beaches, they require
regular maintenance and sediment replenishment. In turn, longevity of replenishment
is linked to the number of storms and other factors described above as well as placing
nourished material on the higher part of the beach and the longshore sediment spreading
rate [
11
]. Previous studies found that initial losses are large [
12
,
13
], which is according
to the theory of beach nourishment due to large morphodynamic adjustments in initial
stages of beach nourishment [
7
]. This is followed by relatively moderate changes and the
beach profile equilibration by wave driven longshore and cross-shore sediment transport.
Liu et al. [14]
found that equilibration of beach profiles on a micro-tidal beach is dependent
on storms and how soon after the nourishment they occur.
In addition to loss of material by coastal processes, the abrasion of gravel sediments
can account for a significant share of the total beach sediment loss. The intensity of
sediment abrasion depends on local conditions, such as wave action, type and shape
of sediment, proportion of time beach been exposed to wave action, etc [
15
18
]. The
theoretical expression for the evaluation of abrasion of gravel sediments is not defined.
Further studies based on in situ measurements are needed to incorporate abrasion into the
sediment budget [17].
Performance of nourished projects has been assessed by using beach volumes and
volume losses [
12
,
14
] but also via assessing other parameters such as beach width [
12
]
and beach orientation [
8
,
19
]. Simple models have been developed to predict the perfor-
mance (e.g., [
20
]) but these are mostly based on monitoring of sand nourishment projects.
Studies of performance of gravel nourishment are rarer (e.g., [
19
,
21
24
]) and based on
monitoring of pre- and post-nourishment beach elevations or shoreline positions.
Monitoring and analysis of changes in the volume and width of nourished beaches is
crucial to inform any beach re-nourishment programme. A range of newly developed re-
mote sensing technologies and techniques have been applied to monitor evolution of beach
nourishment over shorter and longer time-scales. Video monitoring [
8
,
12
,
13
,
19
] provides
information at very high temporal and spatial resolution but it is limited to assessment of
beach planform changes in micro-tidal areas. Although, detecting 3D changes and calcula-
tion of beach volumes is possible for intertidal area in macro-tidal environments. Advances
in satellite imagery [
25
,
26
] provide images taken at shorter time intervals
(e.g., several days)
and of higher spatial resolution (e.g., 10 m), which can then be used for assessment of
shoreline changes. The advantage is that these images cover large world areas and are
freely available (e.g., Sentinel 2) and available in developing countries. However, only
shoreline changes and slopes are assessed from images and the spatial resolution is still
too large for assessing changes on small beaches [
27
,
28
]. Airborne and terrestrial LiDAR
images provide 3D point clouds from which detailed Digital Elevation Models (DEMs) are
developed. These still provide the most valuable information, which can be used to analyse
a range of parameters such as beach volume, slopes, changes in beach elevation, width and
shoreline and other geomorphological features [
29
], but they are relatively expensive and
logistically demanding. In geomorphological studies of the last decades, structure-from-
motion (SfM) photogrammetry is increasingly applied. Image acquisition using unmanned
aerial vehicles (UAVs) and the structure-from-motion (SfM) photogrammetric approach
for image processing have enabled the derivation of high-resolution 3D point clouds and
digital elevation models (DEMs). This low cost method is based on the processing of three-
dimensional clouds generated from a series of photographs [
30
,
31
]. These measurements
are fast, relatively inexpensive, and simple in operation and application. It has been shown
that in inaccessible locations, the use of SfM photogrammetry has significant advantages
over the use of a laser scanner. New developments in SfM photogrammetry have led to an
increase in the accuracy of the point clouds at comparable level with LiDAR’s [
30
,
32
,
33
].
It has been applied in many beach morphological studies in last decade [
34
38
]. How-
J. Mar. Sci. Eng. 2022,10, 358 3 of 24
ever, monitoring of larger scale features is more challenging. The performance of SfM
photogrammetry for monitoring topographic changes in coastal environment is difficult
to control as it depends on many factors such as tides and waves, lighting conditions,
camera position, surface reflectivity, texture and roughness, all of which will affect accuracy
of measurements [
33
]. This requires extra caution when using SfM photogrammetry for
monitoring. James et al. [
32
,
39
] suggested that defining a 3D precision map is imperative
to reduce or account for the uncertainties in the changes detected.
The motivation for this study is to develop a robust, inexpensive and easy to apply
coastal monitoring system based on UAVs and SfM photogrammetry, which would accom-
pany the re-nourishment of natural and artificial beaches in Croatia. Many artificial gravel
beaches were built along the East Adriatic coast (EAC) for the purpose of tourism in last
decades and they are increasingly becoming a useful form of soft coastal defence. They
are designed to mimic natural gravel pocket beaches but their stability is dependent on
regular, mostly annual, replenishment [
40
]. However, volumetric losses and time taken
for these losses to occur is not monitored and hence not much is known on stability and
longevity of these beaches. Besides, few studies have been conducted on artificial gravel
beaches (e.g., [
4
,
41
,
42
]) despite increased research on gravel beaches (e.g., [
43
]). So far
SfM photogrammetry and UAVs have been used in EAC to monitor natural gravel pocket
beaches (GPBs) [3,36,44,45], coastal cliffs stability and coastal vulnerability [4650].
Having detailed point cloud and DEMs rather than just beach profiles has been shown
to be advantageous when analysing changes on GPBs in Croatia, due to the dynamic
changes in the beach orientation and hence longshore variation in the beach slope and the
beach width [
3
]. DEMs were derived using SfM photogrammetry and images obtained
by a hand-held camera [
3
] and UAVs [
44
]. The study presented here is an extension of
this work.
The aim of this particular study is to utilise UAV surveys and SfM photogrammetry
to assess beach nourishment performance of the Ploˇce beach in storm conditions. The
objectives are to evaluate accuracy of UAV surveys, which can then be applied in estimating
different beach parameters and to assess beach nourishment performance in terms of beach
volumes and other parameters—such as changes in the beach elevation, the shoreline
position and the beach areas.
The study contributes to an overall aim of the Beachex project [
51
], in examining the
mechanisms of erosion and restoration of gravel beaches and providing technical support
for long-term beach nourishment in Croatia. The ultimate aim is to better understand the
response of artificial beaches to the variation in external forcing. Field studies are taking
place on the Ploˇce beach in Rijeka (Figure 1), an artificial urban gravel beach constructed
in 2011. The beach is subject to strong erosion, which takes place every autumn/winter
season, and a beach nourishment takes place before every summer season.
This paper presents results from nineteen UAV surveys conducted during 2020 to
assess the post-nourishment changes at an artificial beach Ploˇce, resulting from changes in
the local wave climate.
The paper is organised as follows: the beach site at Ploˇce is described in Section 2,
followed by a description of the data acquisition, the DEMs, the evaluation of the DEMs’
errors and the methods used for the data analysis. The results of the analysis of beach
volumes, beach areas, shoreline position and elevation changes are presented in Section 4.
These are linked to storm wave energy. The paper concludes with a discussion of the ability
of UAV technology to monitor beach nourishment as well as the performance of beach
nourishment in the first phase after nourishment and under the influence of storm waves.
J. Mar. Sci. Eng. 2022,10, 358 4 of 24
J. Mar. Sci. Eng. 2022, 9, x FOR PEER REVIEW 4 of 25
Figure 1. Location of the case study.
2. Study Site
The Ploče beach is the largest beach complex in the city of Rijeka, with a total area of
14,000 m
2
. The beach is located at the western outskirt, 6 km from the city centre (Figure
1). This artificial beach was built in 2011 as a part of the Rijeka swimming pool complex.
The new beach is constructed across two existing natural embayments. It is protected by
three groynes, which are extended from existing headlands. The middle groyne separates
it into two pocket beaches—the eastern and the western beach, both about 130 m wide
(Figure 2). A submerged breakwater about 2 m wide was built at the steep seabed (Figure
2) with the crest at 2.5 m depth. At the landward side, the beach is backed up by a prom-
enade. Once these structures were built, the existing beaches were extended by adding
material on the beach. The material used for the construction of the beach is gravel with
D
50
of 32 mm and supplied from the river Drava in northern Croatia.
Before the summer bathing season, the city of Rijeka replenishes the beach to com-
pensate for material lost during winter season. The main purpose is to maintain the beach
area used for sunbathing. The material is placed on the eastern side, while the sediment
accumulated on the top of the western side of the beach is distributed across the beach.
As a result the beach profiles are flattened. Table 1 shows the amount and size of sediment,
as well as the price spent on replenishment in recent years. It is interesting to note that
recent amount of replenished sediments reduced and that sediment is sourced from a
river in the northern Croatia.
Table 1. Annual cost of renourishment for the Ploče beach.
Year
Cost of Renourish-
ment, before GDP
[EUR]
Annual Amount
[m
3
]
Origin of
Material Grain Size [mm]
2020 (summer) Unknown 100
The Drava
River
32
2020 5800.00 150 32
2019 2700.00 50 32
2018 3100.00 92 32
2017 4800.00 145 32
2016 2300.00 118 32
2015 6700.00 150 32
Figure 1. Location of the case study.
2. Study Site
The Ploˇce beach is the largest beach complex in the city of Rijeka, with a total area of
14,000 m
2
. The beach is located at the western outskirt, 6 km from the city centre (Figure 1).
This artificial beach was built in 2011 as a part of the Rijeka swimming pool complex. The
new beach is constructed across two existing natural embayments. It is protected by three
groynes, which are extended from existing headlands. The middle groyne separates it into
two pocket beaches—the eastern and the western beach, both about 130 m wide (Figure 2).
A submerged breakwater about 2 m wide was built at the steep seabed (Figure 2) with the
crest at 2.5 m depth. At the landward side, the beach is backed up by a promenade. Once
these structures were built, the existing beaches were extended by adding material on the
beach. The material used for the construction of the beach is gravel with D
50
of 32 mm and
supplied from the river Drava in northern Croatia.
Before the summer bathing season, the city of Rijeka replenishes the beach to com-
pensate for material lost during winter season. The main purpose is to maintain the beach
area used for sunbathing. The material is placed on the eastern side, while the sediment
accumulated on the top of the western side of the beach is distributed across the beach. As
a result the beach profiles are flattened. Table 1shows the amount and size of sediment, as
well as the price spent on replenishment in recent years. It is interesting to note that recent
amount of replenished sediments reduced and that sediment is sourced from a river in the
northern Croatia.
The beach was renourished with approximately 150 m
3
of material (D
50
= 32 mm) on
31 January 2020, for purposes of this study. The beach was re-nourished again just before
the summer season in 2020 with approximately 100 m
3
of material. Hence, performance of
these two nourishments will be examined. The focus of this study is on the western beach
and in the rest of the paper it will be referred as the beach.
The tidal oscillations in the Adriatic Sea are small. The mean daily sea amplitude is
0.35 m measured at the tide gauge station in Bakar about 15 km east of the site, the only tide
gauge station in the Kvarner Bay [
53
]. The region is also known for flooding events or so
called acqua alta events. The high sea level during these events is formed by a constructive
superposition of tide, storm surge, seiche, and other low-frequency fluctuations [
54
]. The
storm surge is the predominant component, the range of which depends on the location
J. Mar. Sci. Eng. 2022,10, 358 5 of 24
of low-pressure and the strength and direction of the wind (mainly Scirocco). The highest
measured water level was 1.27 m on 1 November 2018 [55].
J. Mar. Sci. Eng. 2022, 9, x FOR PEER REVIEW 5 of 25
Figure 2. Study site (top) before the construction [52] and (bottom) after construction.
The beach was renourished with approximately 150 m3 of material (D50 = 32 mm) on
31 January 2020, for purposes of this study. The beach was re-nourished again just before
the summer season in 2020 with approximately 100 m3 of material. Hence, performance
of these two nourishments will be examined. The focus of this study is on the western
beach and in the rest of the paper it will be referred as the beach.
The tidal oscillations in the Adriatic Sea are small. The mean daily sea amplitude is
0.35 m measured at the tide gauge station in Bakar about 15 km east of the site, the only
tide gauge station in the Kvarner Bay [53]. The region is also known for flooding events
or so called acqua alta events. The high sea level during these events is formed by a con-
structive superposition of tide, storm surge, seiche, and other low-frequency fluctuations
[54]. The storm surge is the predominant component, the range of which depends on the
location of low-pressure and the strength and direction of the wind (mainly Scirocco). The
highest measured water level was 1.27 m on 1 November 2018 [55].
While wave measurements have been carried in the open sea [56,57], measurements
in coastal waters in the Adriatic Sea are rare and intermittent [58,59]. The wind waves are
dominant in the Kvarner Bay as in the rest of the Adriatic Sea. Weak and moderate winds
with frequent periods of calm conditions prevail in the Kvarner region, while storm winds
with speeds of over 30 m/s are rare and short-lived [60,61]. Due to relatively limited
fetches and the sheltering effect of islands, the wave heights and wave periods are rela-
tively small. The maximum significant wave height of 2.06 m was recorded at a wave buoy
deployed offshore of Rijeka, between 2009 and 2011 [59].
Figure 2. Study site (top) before the construction [52] and (bottom) after construction.
Table 1. Annual cost of renourishment for the Ploˇce beach.
Year
Cost of
Renourishment,
before GDP [EUR]
Annual
Amount [m3]
Origin of
Material
Grain Size
[mm]
2020 (summer) Unknown 100
The Drava
River
32
2020 5800.00 150 32
2019 2700.00 50 32
2018 3100.00 92 32
2017 4800.00 145 32
2016 2300.00 118 32
2015 6700.00 150 32
While wave measurements have been carried in the open sea [
56
,
57
], measurements
in coastal waters in the Adriatic Sea are rare and intermittent [
58
,
59
]. The wind waves are
dominant in the Kvarner Bay as in the rest of the Adriatic Sea. Weak and moderate winds
J. Mar. Sci. Eng. 2022,10, 358 6 of 24
with frequent periods of calm conditions prevail in the Kvarner region, while storm winds
with speeds of over 30 m/s are rare and short-lived [
60
,
61
]. Due to relatively limited fetches
and the sheltering effect of islands, the wave heights and wave periods are relatively small.
The maximum significant wave height of 2.06 m was recorded at a wave buoy deployed
offshore of Rijeka, between 2009 and 2011 [59].
3. Methodology
The Ploˇce beach was surveyed 19 times between 17 January 2020 and 26 February 2021,
outside the summer season and mainly after storm events. The waves were measured
offshore between 1 January 2020 and 30 January 2021 (see Section 3.4).
Prior to the first survey, twelve ground control points were marked along stable
parts of the beach, namely the promenade and groynes (see Figure 3) by the Geodetic
institute Croatia (GZR). Coordinates for the GCPs in the Croatian Terrestrial Reference
System HTRS96 (EPSG: 3765) were obtained with a total station Leica FlexLine TS06plus
(Leica Geosystems AG, Heerbrugg, Switzerland), which provides the high accuracy of
around 1.5 mm [
62
]. In April 2020, additional 16 GCPs were marked and old GCPs were
surveyed again to check whether they remained stable. The GCPs are distributed around
the entire analysed beach area to avoid extrapolation in the area of the interest. Once they
were marked (Figure 3), the aerial surveys using Unmanned Aerial Vehicles (UAVs) were
relatively quick. Usually it would take about an hour to complete the beach survey.
J. Mar. Sci. Eng. 2022, 9, x FOR PEER REVIEW 6 of 25
3. Methodology
The Ploče beach was surveyed 19 times between 17 January 2020 and 26 February
2021, outside the summer season and mainly after storm events. The waves were meas-
ured offshore between 1 January 2020 and 30 January 2021 (see Section 3.4).
Prior to the first survey, twelve ground control points were marked along stable parts
of the beach, namely the promenade and groynes (see Figure 3) by the Geodetic institute
Croatia (GZR). Coordinates for the GCPs in the Croatian Terrestrial Reference System
HTRS96 (EPSG: 3765) were obtained with a total station Leica FlexLine TS06plus (Leica
Geosystems AG, Heerbrugg, Switzerland), which provides the high accuracy of around
1.5 mm [62]. In April 2020, additional 16 GCPs were marked and old GCPs were surveyed
again to check whether they remained stable. The GCPs are distributed around the entire
analysed beach area to avoid extrapolation in the area of the interest. Once they were
marked (Figure 3), the aerial surveys using Unmanned Aerial Vehicles (UAVs) were rel-
atively quick. Usually it would take about an hour to complete the beach survey.
(a) (b)
Figure 3. (a) Study site with marked locations of ground control points (K1, K2, K4–K20, P1, P2
and OT1–OT4); (b) GCP markers
The first and the second survey (UAV_01 and UAV_02) were carried out before and
a few days after the aforementioned nourishment of the beach, respectively. The eleventh
survey was conducted after the second nourishment before the 2020 summer season. The
surveys were conducted during calm periods after the storms and were planned accord-
ing to the weather forecast of the Croatian Meteorological and Hydrological Service [63]
and observed weather and sea conditions.
3.1. Image Acquisition and Processing
The surveys were carried out by two institutionsGeodetic Institute Rijeka (GZR)
and the Faculty of Civil Engineering in Rijeka (GradRi). GZR used a Matrice 200 un-
manned aerial vehicle (UAV, DJI, Shenzhen, China) with a camera ILCE-7M2 (Sony, To-
kyo, Japan) and GradRi used a Phantom 4 Pro UAV (DJI, Shenzhen, China) for the image
acquisition. Mainly the camera at nadir was used to capture emerged beach and also shal-
low water. About 1/4 of the GradRi images were taken from different camera angles to
obtain a complex emerged beach from lower flight heights. The overlap between images
was kept constant, and more than nine overlapping images were used to cover the whole
area. The number of images ranged from 208 to 667 and the flight altitude from 19.2 to 29
Figure 3.
(
a
) Study site with marked locations of ground control points (K1, K2, K4–K20, P1, P2 and
OT1–OT4); (b) GCP markers.
The first and the second survey (UAV_01 and UAV_02) were carried out before and a
few days after the aforementioned nourishment of the beach, respectively. The eleventh
survey was conducted after the second nourishment before the 2020 summer season. The
surveys were conducted during calm periods after the storms and were planned according
to the weather forecast of the Croatian Meteorological and Hydrological Service [
63
] and
observed weather and sea conditions.
3.1. Image Acquisition and Processing
The surveys were carried out by two institutions—Geodetic Institute Rijeka (GZR) and
the Faculty of Civil Engineering in Rijeka (GradRi). GZR used a Matrice 200 unmanned
aerial vehicle (UAV, DJI, Shenzhen, China) with a camera ILCE-7M2 (Sony, Tokyo, Japan)
J. Mar. Sci. Eng. 2022,10, 358 7 of 24
and GradRi used a Phantom 4 Pro UAV (DJI, Shenzhen, China) for the image acquisition.
Mainly the camera at nadir was used to capture emerged beach and also shallow water.
About 1/4 of the GradRi images were taken from different camera angles to obtain a
complex emerged beach from lower flight heights. The overlap between images was kept
constant, and more than nine overlapping images were used to cover the whole area. The
number of images ranged from 208 to 667 and the flight altitude
from 19.2 to 29 m
, and
100–103 m
for GradRi and GZR surveys respectively. Using cameras of different resolution
and taking images from the different flight heights resulted in a range of ground resolution
from 5.1 to 19.7 mm/pixel (see Table 2). Both institutions used Agisoft Metashape Profes-
sional (formerly Agisoft PhotoScan, Agisoft LLC, St. Petersburg, Russia) to generate point
clouds from the captured photos. The workflow is described in several papers (e.g., [
64
,
65
]).
The reprojection errors (distance between the calculated position of the point and its po-
sition marked on the images) after processing of images from both instruments are very
similar, allowing comparison of point clouds. Table 2summarises all surveys and gives
details of the reprojection errors. The spacing between cloud points was 20 mm. Point
clouds were subsampled to 50 mm, to speed up the Matlab analysis.
3.2. Assessment of Errors
The first assessment of the errors was based on Metashape’s calculation of the root-
mean square error (RMSE) on GCPs. Check points were not used due to the insufficient
number of GCPs and their coverage with beach sediments at first, and between 6th and
9th survey respectively. However, the RMSE does not distinguish between systematic
(accuracy) and random (precision) errors and it is recommended to use an additional
approach [
66
]. The second method used was comparison of point clouds on the fixed
part of the scene using the freely available Multiscale Model to Model Cloud Comparison
(M3C2) plugin in the CloudCompare software [
67
,
68
]. In addition to these quantitative
analyses, beach profiles that included a stable promenade were qualitatively examined for
any outliers.
Agisoft Metashape calculates the RMSE for X, Y and Z coordinates, the planimetric
RMSE (XY) and the total RMSE for the GCPs as well as for some additional independent
control points (Figure 3). The equation bellow is given for the total RMSE and the other
RMSEs are calculated in a similar way:
Total =v
u
u
t
n
i=1
(Xi,meas Xi)2+(Yi,meas Yi)2+(Zi,meas Zi)2
n(1)
where: X
i,meas
—measured value for X coordinate for i GCP; X
i
—value for X coordinate for i
GCP; Y
i,meas
—measured value for Y coordinate for i GCP; Y
i
—value for Y coordinate for i GCP;
Zi,meas—measured value for Z coordinate for i GCP; Zi—value for Z coordinate for i GCP.
Calculated RMSEs for the GCPs and for all surveys are given in the Table 3. The
RMSEs for X and Y are below 8 mm while the planimetric (XY) RMSEs are below 10 mm.
The vertical RMSE is on average lower than 10 mm, with the largest total RMSE being
13.4 mm. These values can be used as indicative metrics of survey performance and surveys
two, ten and nineteen had the largest vertical RMSE, though still around 10 mm.
Next, the point clouds of all surveys (UAV_01–UAV_19) were compared on the fixed
parts of the scene—along concrete promenade, paths and walls and stairs
(Figure 3)
using the
Multiscale Model to Model Cloud Comparison (M3C2) plugin in the
CloudCompare [67,68]
.
The plugin allows direct comparison of point clouds and detects changes with very little
manual work. The algorithm is described in several papers (e.g., [
32
,
68
72
]), but the working
principle is to find the most appropriate normal direction for points and calculate the distances
between point clouds along a cylinder of a given diameter (the projection scale), projected
along the normal.
J. Mar. Sci. Eng. 2022,10, 358 8 of 24
Table 2. Survey data from Agisoft Metashape.
Date Survey Label Surveyed by Coverage Area
(km2)
Number of
Aligned Images
Flying Altitude
(m)
Ground Res.
(mm/pix) Tie Points Projections Reprojection
Error (pix)
17 January 2020 UAV_01 GZR 10.065 355 102.0 19.0 115,521 963,731 0.897
7 February 2020 UAV_02 GZR 10.072 467 103.0 18.8 233,105 1,445,253 0.767
12 February 2020 UAV_03 GZR 10.063 321 101.5 19.7 268,994 1,663,459 0.693
2 March 2020 UAV_04 GradRi 20.009 602 19.2 4.7 309,557 2,183,095 0.404
3 March 2020 UAV_05 GradRi 20.013 366 28.0 7.1 206,361 1,153,749 0.526
10 March 2020 UAV_06 GradRi 20.016 388 29.0 7.4 207,342 1,155,296 0.508
20 March 2020 UAV_07 GradRi 20.021 613 31.3 7.9 294,521 1,465,854 0.629
29 March 2020 UAV_08 GradRi 20.016 362 27.4 7.0 171,709 1,320,331 0.631
28 April 2020 UAV_09 GradRi 20.025 382 34.7 8.9 239,525 1,301,030 0.708
5 May 2020 UAV_10 GradRi 20.023 383 26.2 6.6 275,779 1,381,719 0.744
1 October 2020 UAV_11 GradRi 20.018 535 28.1 7.5 300,541 1,906,445 0.772
6 October 2020 UAV_12 GradRi 20.020 492 28.6 7.2 264,272 1,717,020 0.826
13 October 2020 UAV_13 GradRi 20.018 538 28.5 7.1 328,053 1,875,383 0.790
2 November 2020 UAV_14 GZR 10.080 208 103.0 19.5 305,479 1,015,135 0.733
24 November 2020 UAV_15 GradRi 20.008 278 20.5 5.1 154,838 961,184 0.515
10 December 2020 UAV_16 GradRi 20.017 600 24.3 6.1 280,890 1,840,167 0.315
14 December 2020 UAV_17 GZR 10.060 269 100.0 19.1 335,580 1,231,250 0.724
14 January 2021 UAV_18 GradRi 20.016 560 26.7 6.6 308,839 1,653,808 0.318
26 February 2021 UAV_19 GZR 10.060 290 101.0 19.6 381,158 1,469,166 0.791
1Geodetic Institute Rijeka. 2Faculty of Civil Engineering in Rijeka.
J. Mar. Sci. Eng. 2022,10, 358 9 of 24
Table 3. RMSE of point clouds on GCPs—results from Agisoft Metashape.
Survey Label No. of GCPs RMSE [mm]
X Y Z XY Total
UAV_01 6 2.3 0.7 4.8 2.4 5.3
UAV_02 7 4.3 1.3 10.2 4.5 11.2
UAV_03 6 1.0 4.7 6.2 4.8 7.8
UAV_04 8 3.1 5.2 7.7 6.0 9.8
UAV_05 10 8.3 6.2 5.5 10.3 11.7
UAV_06 6 3.3 3.4 2.1 4.8 5.2
UAV_07 5 3.1 3.2 2.1 4.4 4.9
UAV_08 4 0.7 0.6 0.6 0.9 1.1
UAV_09 4 1.1 0.8 0.9 1.3 1.6
UAV_10 9 3.2 3.4 12.5 4.6 13.4
UAV_11 21 4.0 5.1 9.7 6.5 11.6
UAV_12 23 4.9 4.4 3.5 6.6 7.5
UAV_13 23 3.5 4.7 5.1 5.9 7.8
UAV_14 20 7.4 7.2 5.2 10.3 11.6
UAV_15 8 3.7 3.1 5.0 4.8 6.9
UAV_16 7 2.6 3.4 3.9 4.3 5.8
UAV_17 21 2.3 0.7 4.8 2.4 5.3
UAV_18 13 7.4 5.5 2.3 9.3 9.5
UAV_19 20 4.3 1.3 10.2 4.5 11.2
The normal scale was defined as 25 times the average local roughness calculated for a
point cloud by the CloudCompare, in accordance with the work of
Ferrer-González et al. [71]
.
In our case, this value was 0.5 m and diameter of the search cylinder was 0.2 m. The mean and
standard deviation values for the M3C2 distances were calculated and these are associated
with the accuracy and the precision respectively.
Only a few comparison results are given for illustration purposes here; between 10th
and 4th and 19th–14th surveys (Figure 4). The mean value of M3C2 distances for these
cases were 0.0 and
0.04 m, respectively. The spatial distribution of the M3C2 calculated
distances found the accuracies to be mostly between
0.05 and 0.05 m, and the results are
shown in Figure 4. Larger values for the distance are only found in places where vegetation,
fence or sediment from the beach is present or where an occasional pedestrian was ‘caught’
during the survey.
Based on the analysis of ‘changes’ on the fixed parts, level of detection is defined
as
±
0.05 m. Hence, the changes above 0.05 m and bellow
0.05 m can be accurately
detected from the surveys.
3.3. Assessment of Beach Changes
Emerged beach changes were assessed using a few parameters: the beach elevation,
the shoreline position, beach profiles and volumes. Beach elevation changes were calculated
with the same M3C2 plugin in CloudCompare described above, comparing two successive
surveys. The Matlab code was written to extract the position of the shoreline and cross-
shore transects shown in Figure 5from original point clouds.
Beach elevation changes smaller than the uncertainty levels (
±
0.05 m) were disre-
garded [32].
Beach volume calculations were derived from DEMs, which were generated in Matlab
by linear interpolation from available point clouds with a resolution of 0.05 m. Volumes
were calculated for the emerged beach area marked in Figure 5, integrating the sediment
volume for each DEM cell according to reference level set at 0 m.
J. Mar. Sci. Eng. 2022,10, 358 10 of 24
J. Mar. Sci. Eng. 2022, 9, x FOR PEER REVIEW 9 of 25
[67,68]. The plugin allows direct comparison of point clouds and detects changes with
very little manual work. The algorithm is described in several papers (e.g., [32,6872]), but
the working principle is to find the most appropriate normal direction for points and cal-
culate the distances between point clouds along a cylinder of a given diameter (the pro-
jection scale), projected along the normal.
The normal scale was defined as 25 times the average local roughness calculated for
a point cloud by the CloudCompare, in accordance with the work of Ferrer-González et
al. [71]. In our case, this value was 0.5 m and diameter of the search cylinder was 0.2 m.
The mean and standard deviation values for the M3C2 distances were calculated and
these are associated with the accuracy and the precision respectively.
Only a few comparison results are given for illustration purposes here; between 10th
and 4th and 19th14th surveys (Figure 4). The mean value of M3C2 distances for these
cases were 0.0 and 0.04 m, respectively. The spatial distribution of the M3C2 calculated
distances found the accuracies to be mostly between 0.05 and 0.05 m, and the results are
shown in Figure 4. Larger values for the distance are only found in places where vegeta-
tion, fence or sediment from the beach is present or where an occasional pedestrian was
‘caught’ during the survey.
Based on the analysis of ‘changes’ on the fixed parts, level of detection is defined as
±0.05 m. Hence, the changes above 0.05 m and bellow 0.05 m can be accurately detected
from the surveys.
(b) (c)
(a) (d) (e)
Figure 4. Vertical accuracy; (a) M3C2 distances on concrete promenade; (b,c) inserts of M3C2 from
UAV_10-4; (d,e) inserts of M3C2 from UAV_19-14.
3.3. Assessment of Beach Changes
Emerged beach changes were assessed using a few parameters: the beach elevation,
the shoreline position, beach profiles and volumes. Beach elevation changes were calcu-
lated with the same M3C2 plugin in CloudCompare described above, comparing two suc-
cessive surveys. The Matlab code was written to extract the position of the shoreline and
cross-shore transects shown in Figure 5 from original point clouds.
Beach elevation changes smaller than the uncertainty levels 0.05 m) were disre-
garded [32].
Beach volume calculations were derived from DEMs, which were generated in
Matlab by linear interpolation from available point clouds with a resolution of 0.05 m.
Figure 4.
Vertical accuracy; (
a
) M3C2 distances on concrete promenade; (
b
,
c
) inserts of M3C2 from
UAV_10-4; (d,e) inserts of M3C2 from UAV_19-14.
Figure 5.
Location of cross-shore transects marked 10–110; blue polygon shows the extent of the area
for beach area and volume calculations.
For volume analyses the error was taken into account. The possible volumetric errors
(Err) were defined as follows:
Err =AUAV·Herr, (2)
where A
UAV
is the emerged beach area at the moment of surveying and H
err
is previously
defined elevation error (±0.05 cm).
J. Mar. Sci. Eng. 2022,10, 358 11 of 24
3.4. Wave Measurements
Waves are measured by the Hydrographic Institute of the Republic of Croatia with the
Datawell wave rider (type MKIII) in the immediate vicinity of the Ploˇce beach (
4519.5880N
;
14
23.738
0
E, WGS 84, free movement within the radius of the position), offshore of port
Rijeka [
73
]. The standard wave statistics is provided every 30 min between 1 January 2020
and 30 January 2021. Additionally, an ADCP was deployed offshore of the beach from
30 January 2021 to spring 2021 but its data were not used in this study. The wave rose in
Figure 6summarizes the wave heights and directions recorded by the wave buoy between
1 January 2020 and the end of January 2021. For the most of time, calm conditions and
waves up to 0.4 m height were recorded. Figure 7shows that the events with higher wave
heights take place before April and after September with exception of two short lasting
storm during the summer. It is also evident from Figure 7that the largest recorded waves
are from S to SSE directions.
J. Mar. Sci. Eng. 2022, 9, x FOR PEER REVIEW 11 of 25
Figure 6. Wave rose for all wave measurements.
Figure 7. Identified storm events. A dashed line represents the stop of the wave buoy measurement
and the start of the ADCP data (30 January 2021).
Storm events are identified by using a threshold of the 99.5th percentile of the signif-
icant wave height data [74] recorded at the wave buoy, which is 1.26m. A minimum du-
ration of storm was set to 1 h [75] and the time between storms to 10 h. For the significant
wave height threshold of 1.26 m and set duration conditions, eleven storms were identi-
fied. This includes weak and strong storms [76,77]. A storm in March (1 March) and two
in December (5 and 28 December) lasted for more than 4, 19 and 8 h respectively.
The energy peak for all 11 storm events was calculated using the following expres-
sion:
E = H
max2
T
p
, (3)
where H
max
represents the maximum significant wave height of the storm and T
p
is the
wave period at the storm peak [19,78,79].
The storm power index [80] was also calculated:
P
s
= H
max2
D, (4)
Figure 6. Wave rose for all wave measurements.
J. Mar. Sci. Eng. 2022, 9, x FOR PEER REVIEW 11 of 25
Figure 6. Wave rose for all wave measurements.
Figure 7. Identified storm events. A dashed line represents the stop of the wave buoy measurement
and the start of the ADCP data (30 January 2021).
Storm events are identified by using a threshold of the 99.5th percentile of the signif-
icant wave height data [74] recorded at the wave buoy, which is 1.26m. A minimum du-
ration of storm was set to 1 h [75] and the time between storms to 10 h. For the significant
wave height threshold of 1.26 m and set duration conditions, eleven storms were identi-
fied. This includes weak and strong storms [76,77]. A storm in March (1 March) and two
in December (5 and 28 December) lasted for more than 4, 19 and 8 h respectively.
The energy peak for all 11 storm events was calculated using the following expres-
sion:
E = H
max2
T
p
, (3)
where H
max
represents the maximum significant wave height of the storm and T
p
is the
wave period at the storm peak [19,78,79].
The storm power index [80] was also calculated:
P
s
= H
max2
D, (4)
Figure 7.
Identified storm events. A dashed line represents the stop of the wave buoy measurement
and the start of the ADCP data (30 January 2021).
J. Mar. Sci. Eng. 2022,10, 358 12 of 24
Storm events are identified by using a threshold of the 99.5th percentile of the sig-
nificant wave height data [
74
] recorded at the wave buoy, which is 1.26 m. A minimum
duration of storm was set to 1 h [
75
] and the time between storms to 10 h. For the significant
wave height threshold of 1.26 m and set duration conditions, eleven storms were identified.
This includes weak and strong storms [
76
,
77
]. A storm in March (1 March) and two in
December (5 and 28 December) lasted for more than 4, 19 and 8 h respectively.
The energy peak for all 11 storm events was calculated using the following expression:
E=Hmax 2Tp, (3)
where H
max
represents the maximum significant wave height of the storm and T
p
is the
wave period at the storm peak [19,78,79].
The storm power index [80] was also calculated:
Ps= Hmax2D, (4)
where D is the duration of “storm conditions” in hours.
Table 4gives a summary of the energy peaks and the storm power indices for
all 11 storms. The storms with the largest intensities (8 and 10) took place on 5 and
28 December 2020
. The third largest (4) was recorded on 6 March 2020. These are also
marked in Figure 7. The wave direction was between 160–180
during all storms except
during storms 5, 6, 7. During latter, the wave direction was around 190
. During storm 10,
the direction changed from 180–205
. The peak periods were in range from 4.4–6 s. The
wave steepness for the peak conditions for all those storms was close to 0.04 except for the
storm on 6 March for which the wave steepness was 0.05. During this storm, the wave
height increased much more than the wave period (T
p
= 5 s), resulting in larger steepness.
Table 4. Energy peaks and storm power indices for 11 defined storm events.
Storm
Event Start of Storm End of Storm Hs,max [m] Tp[s] Dirp[]Energy Peak
[m2s]
Storm Power
Index [m2h]
1 1 March 2020 00:34 1 March 2020 04:09 1.4 4.4 170.2 8.2 5.3
2 1 March 2020 14:29 1 March 2020 20:32 1.5 4.6 170.2 9.8 9.4
3 3 March 2020 00:15 3 March 2020 02:28 1.6 4.6 182.8 11.3 5.5
4 6 March 2020 05:46 6 March 2020 08:41 2.1 5.0 177.2 21.6 12.6
5 5 June 2020 00:21 5 June 2020 01:53 1.4 5.0 194.1 9.9 3.1
6 25 September 2020 01:18 25 September 2020 03:41 1.5 4.6 189.8 10.3 5.4
7 3 October 2020 15:01 3 October 2020 16:18 1.5 4.6 191.3 10.7 3.0
8 5 December 2020 00:53 5 December 2020 22:58 1.8 5.6 168.8 18.1 62.7
9 6 December 2020 08:31 6 December 2020 09:43 1.6 4.6 165.9 11.1 2.9
10 28 December 2020 10:43 28 December 2020 20:08 2.2 5.9 180.0 27.7 39.7
11 23 January 2021 00:03 23 January 2021 01:35 1.5 4.6 163.1 10.2 3.4
4. Results
4.1. Beach Shoreline Changes
Shoreline changes were analyzed at 0 m above sea level and they are given in the
Figure 8. Shorelines derived from surveys after storms and nourishments are shown in
colour, while the rest are in grey.
Between the 2nd and 3rd survey, (UAV_02 and UAV_03), the shoreline moved onshore
due to first storm after nourishment. It also rotated slightly west (clockwise) on the western
part of the beach. The western end of the shoreline moved about 1.8 m inland, while it
stayed at approximately the same position on the other side. After the first wave energetic
event, the beach stabilize, There was no further rotationand the results generally show
pronounced onshore sediment migrations.
J. Mar. Sci. Eng. 2022,10, 358 13 of 24
J. Mar. Sci. Eng. 2022, 9, x FOR PEER REVIEW 12 of 25
where D is the duration of “storm conditions” in hours.
Table 4 gives a summary of the energy peaks and the storm power indices for all 11
storms. The storms with the largest intensities (8 and 10) took place on 5 and 28 December
2020. The third largest (4) was recorded on 6 March 2020. These are also marked in Figure
7. The wave direction was between 160180° during all storms except during storms 5, 6,
7. During latter, the wave direction was around 190°. During storm 10, the direction
changed from 180–205°. The peak periods were in range from 4.4–6 s. The wave steepness
for the peak conditions for all those storms was close to 0.04 except for the storm on 6th
March for which the wave steepness was 0.05. During this storm, the wave height in-
creased much more than the wave period (Tp = 5 s), resulting in larger steepness.
Table 4. Energy peaks and storm power indices for 11 defined storm events.
Storm
Event Start of Storm End of Storm Hs,max
[m] Tp [s] Dirp] Energy Peak
[m2 s]
Storm Power
Index [m2 h]
1 1 March 2020 00:34 1 March 2020 04:09 1.4 4.4 170.2 8.2 5.3
2 1 March 2020 14:29 1 March 2020 20:32 1.5 4.6 170.2 9.8 9.4
3 3 March 2020 00:15 3 March 2020 02:28 1.6 4.6 182.8 11.3 5.5
4 6 March 2020 05:46 6 March 2020 08:41 2.1 5.0 177.2 21.6 12.6
5 5 June 2020 00:21 5 June 2020 01:53 1.4 5.0 194.1 9.9 3.1
6 25 September 2020 01:18 25 September 2020 03:41 1.5 4.6 189.8 10.3 5.4
7 3 October 2020 15:01 3 October 2020 16:18 1.5 4.6 191.3 10.7 3.0
8 5 December 2020 00:53 5 December 2020 22:58 1.8 5.6 168.8 18.1 62.7
9 6 December 2020 08:31 6 December 2020 09:43 1.6 4.6 165.9 11.1 2.9
10 28 December 2020 10:43 28 December 2020 20:08 2.2 5.9 180.0 27.7 39.7
11 23 January 2021 00:03 23 January 2021 01:35 1.5 4.6 163.1 10.2 3.4
4. Results
4.1. Beach Shoreline Changes
Shoreline changes were analyzed at 0 m above sea level and they are given in the
Figure 8. Shorelines derived from surveys after storms and nourishments are shown in
colour, while the rest are in grey.
Figure 8. Shoreline changes.
Between the 2nd and 3rd survey, (UAV_02 and UAV_03), the shoreline moved on-
shore due to first storm after nourishment. It also rotated slightly west (clockwise) on the
western part of the beach. The western end of the shoreline moved about 1.8 m inland,
while it stayed at approximately the same position on the other side. After the first wave
Figure 8. Shoreline changes.
Prior to survey UAV_04, three ‘pockets’ formed due to the fresh water springs out-
bursts through the beach body on the western part of the beach, causing the local erosion
and onshore shoreline move by an average of 3.4 m (transects 20–30, Figure 8). Beach
recovered by infilling of eroded zones between 6th and 7th survey. Shoreline moved
back offshore for almost the same distance (3.47 m). A similar process occurred prior to
survey UAV_16, except that this time the spring outbursts took place in the central part
of the beach (transects 40–60, Figure 8). This time the shoreline moved offshore due to
formation of sediment lobes at fresh water outbursts, with the greatest distances being
approximately 2.5 and 3.6 m. In the period up to the 17th survey, the beach stabilized again
under smaller waves.
On average, the shoreline changes are between
±
40 cm, which are smaller than
potential errors. However, the post-storm and post-nourishment shoreline changes can be
detected with higher certainty.
4.2. Beach Elevation Changes
Figure 9shows beach elevation changes greater than 0.05 m. Eroded areas are repre-
sented with shades of blue and green, while the sediment deposition is shown in shades of
orange and red.
The beach was nourished between the first two surveys (UAV_02–UAV_01). The beach
elevation changes show that, the material was pushed from the higher parts of the beach
towards the toe of the beach on the west end. On the eastern side, the material was added
to the beach and the positive changes were recorded, shown in red.
The first significant elevation changes took place between the third and fourth survey
caused by the first storm event on 1 March 2020. The measured significant wave height was
up to 1.47 m. A berm formed on the higher parts of the beach as a result of the wave activity.
The wave action was accompanied by heavy rainfall on 1 and 2 March 2020. Overall
46.3 mm of rain fell in just over 16 h. The promenade on the beach was flooded and there
was the fresh water spring outburst through the beach body (Figure 10). Resulting changes
were accretion, namely berms, on the upper beach on the western side, and erosion on the
eastern side. In addition, the areas of concentrated erosion (1.2 m deep), associated with
the fresh water outburst can be seen on the western side of the beach. They are marked in
blue in Figure 9(UAV_04–UAV_03).
Survey UAV_05 was conducted the day after the flood. The wave height increased
(H
s,max
= 1.56 m) and the fresh water was still flowing through the beach. Formation of the
berm higher up on the beach can be seen in Figure 9.
J. Mar. Sci. Eng. 2022,10, 358 14 of 24
J. Mar. Sci. Eng. 2022, 9, x FOR PEER REVIEW 14 of 25
Figure 9. Beach elevation changes (emerged) between successive surveys.
The beach was nourished between the first two surveys (UAV_02–UAV_01). The
beach elevation changes show that, the material was pushed from the higher parts of the
beach towards the toe of the beach on the west end. On the eastern side, the material was
added to the beach and the positive changes were recorded, shown in red.
The first significant elevation changes took place between the third and fourth survey
caused by the first storm event on 1 March 2020. The measured significant wave height
was up to 1.47 m. A berm formed on the higher parts of the beach as a result of the wave
activity. The wave action was accompanied by heavy rainfall on 1 and 2 March 2020. Over-
all 46.3 mm of rain fell in just over 16 h. The promenade on the beach was flooded and
there was the fresh water spring outburst through the beach body (Figure 10). Resulting
Figure 9. Beach elevation changes (emerged) between successive surveys.
J. Mar. Sci. Eng. 2022,10, 358 15 of 24
J. Mar. Sci. Eng. 2022, 9, x FOR PEER REVIEW 15 of 25
changes were accretion, namely berms, on the upper beach on the western side, and ero-
sion on the eastern side. In addition, the areas of concentrated erosion (1.2 m deep), asso-
ciated with the fresh water outburst can be seen on the western side of the beach. They
are marked in blue in Figure 9 (UAV_04–UAV_03).
Figure 10. Flooding and outbursts on the beach, 2 March 2020.
Survey UAV_05 was conducted the day after the flood. The wave height increased
(Hs,max = 1.56 m) and the fresh water was still flowing through the beach. Formation of the
berm higher up on the beach can be seen in Figure 9.
In the period between surveys UAV_05 and UAV_06 great changes were also ob-
served, affecting the whole beach. During this period, another storm formed on 6 March.
Incoming wave direction was 177° and the significant wave height was 2.08 m. On the
same day, measured tidal elevation was 0.86 m. As result of higher water level and waves
of higher wave heights, the berm (about 40 cm high) formed more backshore than in pre-
vious surveys. Due to the high sea level, some material was shifted from the beach to the
promenade. The erosional ‘pockets’ caused by the fresh water spring in previous events
were still present.
Subsequent surveys do not show any significant changes until the summer when the
beach was prepared for bathing season. Comparison of surveys UAV_10 and UAV_11
shows results of typical nourishment done during summer, redistributing deposited ma-
terial from the top to the lower parts of the beach and flattening.
The next storm event on 3 October 2020, between surveys UAV_12 and UAV_11
caused erosion around the base of the groyne on the eastern side and accretion on the
opposite end of the beach.
The next three surveys did not record significant changes on the beach. The sixteenth
survey conducted on 10 December 2020 shows the scale of damage caused by storm events
on 5 and 6 December, when the maximum significant wave height was 1.8 m. Because of
the heavy rain the water started to flow through the beach body again, causing concen-
trated erosion in the middle part of the cell. The material from lower beach was pushed
to the upper beach by waves on the western part of the beach. The site was flooded again
Figure 10. Flooding and outbursts on the beach, 2 March 2020.
In the period between surveys UAV_05 and UAV_06 great changes were also observed,
affecting the whole beach. During this period, another storm formed on 6 March. Incoming
wave direction was 177
and the significant wave height was 2.08 m. On the same day,
measured tidal elevation was 0.86 m. As result of higher water level and waves of higher
wave heights, the berm (about 40 cm high) formed more backshore than in previous surveys.
Due to the high sea level, some material was shifted from the beach to the promenade. The
erosional ‘pockets’ caused by the fresh water spring in previous events were still present.
Subsequent surveys do not show any significant changes until the summer when the
beach was prepared for bathing season. Comparison of surveys UAV_10 and UAV_11
shows results of typical nourishment done during summer, redistributing deposited mate-
rial from the top to the lower parts of the beach and flattening.
The next storm event on 3 October 2020, between surveys UAV_12 and UAV_11 caused
erosion around the base of the groyne on the eastern side and accretion on the opposite
end of the beach.
The next three surveys did not record significant changes on the beach. The sixteenth
survey conducted on 10 December 2020 shows the scale of damage caused by storm events
on 5 and 6 December, when the maximum significant wave height was 1.8 m. Because of the
heavy rain the water started to flow through the beach body again, causing concentrated
erosion in the middle part of the cell. The material from lower beach was pushed to the
upper beach by waves on the western part of the beach. The site was flooded again and
some material ended up on the promenade. Only four days later the beach was surveyed
again (UAV_17). Because there was no wave activity in this period (Figure 7), the changes
were mainly found at the lower parts of the beach.
Between the 18th and 17th survey (UAV_18 and UAV_17), the beach was exposed to
the storm with the largest significant wave height of 2.16 m from the southern direction
(Figure 7). However, changes shown in Figure 9are not the result of this storm, as prior
to survey UAV_18, the promenade was cleaned and the beach flattened. This explains
negative elevation changes on the west end of the beach (Figure 9), and the filled ‘pockets’
in the centre.
J. Mar. Sci. Eng. 2022,10, 358 16 of 24
Wave activity after UAV_18 caused erosion again around the groyne on the eastern
side and accretion of the upper western part of the beach. The observed elevation changes
between surveys indicate presence of longshore and cross-shore sediment transport.
Cross-shore changes are evident in cross-shore transects. Figure 11 shows elevation
changes at both ends of the beach and in the middle, at 20, 50 and 90 m (see Figure 5).
J. Mar. Sci. Eng. 2022, 9, x FOR PEER REVIEW 17 of 25
Figure 11. Elevation changes in transects at the west (top), in the middle (middle) and at the east
end of the beach (bottom).
4.3. Beach Volume Changes versus Wave Energy Flux
The results of the volume changes, including error bars, are given in Figure 12 top.
The beach area was also calculated and shown in the Figure 12 bottom. The average area
is about 1900 m
2
and it changed only slightly from 1841 m
2
in survey UAV_14 to 2049 m
2
in UAV_11. The largest area was recorded just after the summer season after the beach
was renourished and flattened.
Emerged beach volumes range from 2085 ± 47 m
3
(UAV_01) to 2345 ± 51 m
3
(UAV_11). The significance of the error estimates is further discussed in Section 5.1.
Although the changes occurring are smaller than acquired accuracy, a certain trend
is noticeable. Increases in volume and beach area recorded by surveys UAV_02 and
UAV_11 correspond to renourishments and flattening of the beach. There is fluctuation in
beach volumes after the first nourishment, which took place in January. The beach volume
remained stable for longer period of time after the second beach nourishment, which took
place at the end of May.
Figure 11.
Elevation changes in transects at the west (
top
), in the middle (
middle
) and at the east end
of the beach (bottom).
Changes in the transects 20 and 50 show that the berms (0.2 m high) form under the
influence of waves, immediately after the nourishment (UAV_03 and UAV_12)). They
stayed at almost the same places until the stronger storms hit the beach causing them to
move further onshore (UAV_04, UAV_06 and UAV_16). In the middle transect, the berm
moved (horizontally) for 3.7 m due to steepening of the beach profile by the 4th survey.
The material from the lower beach face was moved by wave run-up into a berm on the
upper beach. The berm stayed at more or less the same location and was approximately
the same size, until the wave conditions changed. At the time of the 6th survey (after
storms 3 and 4
), it moved further onshore for additional 5.8 m. The storm on 3 October,
which took place after the second beach nourishment and flattening (UAV_12), formed the
new berm. It moved 1.45 m, and stayed there until the storms on 5 and 6 December, when
it moved further offshore for 2.6 m. The strongest waves coming from the south quadrant
J. Mar. Sci. Eng. 2022,10, 358 17 of 24
are often accompanied by higher sea levels [
55
], causing the berm to form on the higher
parts of the beach.
The same pattern of berm formation is observed at the western end (transect 20),
while on the eastern side (transect 90) there is no berm. Because of erosion, material added
through nourishments is very quickly removed, leaving the rubble underlayer exposed on
this side of the beach.
4.3. Beach Volume Changes versus Wave Energy Flux
The results of the volume changes, including error bars, are given in Figure 12 top.
The beach area was also calculated and shown in the Figure 12 bottom. The average area is
about 1900 m
2
and it changed only slightly from 1841 m
2
in survey UAV_14 to 2049 m
2
in
UAV_11. The largest area was recorded just after the summer season after the beach was
renourished and flattened.
J. Mar. Sci. Eng. 2022, 9, x FOR PEER REVIEW 18 of 25
Figure 12. Beach volumes (top) and beach areas (bottom) during surveying.
Decreases in areas following these surveys are a result of wave influence on a flat
beach. Under wave activity, the beach face becomes steeper—the material is pushed on-
shore and a berm is formed. This is especially a case with surveys UAV_11-14, when the
beach area continued to decrease under waves from the south, while the volume remained
almost the same. This indicates that the beach sediment was redistributed on the emerged
beach rather than lost offshore or onshore.
The peak storm wave energy varied between 8 and 27 m2s, whereas the storm inten-
sity index varied between 3 and 62 m2h. The latter is due to the duration of storms, which
changes with seasons. The changes in the beach volumes and the beach area are relatively
small and independent of these two wave energy parameters. However, the largest
changes in beach elevation and in beach profiles are associated with storms, once the
beach reaches post-nourishment adjustment.
Next, the offshore wave steepness parameter was examined. The offshore wave
steepness (Ho/Lo), was found as potentially useful parameter for explaining offshore ver-
sus onshore sediment transport on gravel beaches (see e.g., [43]). A threshold value of 0.01
is used to distinguish onshore (<0.01) and offshore sediment transport (>0.01). For the peak
storm waves in this study, the offshore wave steepness is between 0.04 and 0.05 and this
would indicate presence of offshore sediment transport. Figure 10 shows that changes in
beach elevations vary along the beach and there are several profiles (e.g., between surveys
3 & 4) where sediment loss is larger than sediment gain. However, the survey would need
to extend offshore to examine whether these losses are due to offshore sediment transport.
Not only cross-shore but also longshore sediment transport and groundwater dynamics
affect the sediment gains and losses on these profiles.
5. Discussion
5.1. Use of UAVs for Monitoring of Beach Changes
This study demonstrated the effectiveness of UAVs and SfM photogrammetry for
repeating monitoring of beach changes, which can inform coastal management. Measure-
ments were performed quickly once fixed GCPs were set up. They are particularly suitable
for targeted surveys of beaches, which experience noticeable changes only after rare high
wave energy events. Using full 3D data (point clouds) rather than other 2D parameters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
1900
2000
2100
2200
2300
2400
2500
Beach Volume [m
3
]
Beach Vol
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
1800
1850
1900
1950
2000
2050
UAV Survey [1]
Beach Area [m
2
]
Beac h area
Figure 12. Beach volumes (top) and beach areas (bottom) during surveying.
Emerged beach volumes range from 2085
±
47 m
3
(UAV_01) to 2345
±
51 m
3
(UAV_11).
The significance of the error estimates is further discussed in Section 5.1.
Although the changes occurring are smaller than acquired accuracy, a certain trend is
noticeable. Increases in volume and beach area recorded by surveys UAV_02 and UAV_11
correspond to renourishments and flattening of the beach. There is fluctuation in beach
volumes after the first nourishment, which took place in January. The beach volume
remained stable for longer period of time after the second beach nourishment, which took
place at the end of May.
Decreases in areas following these surveys are a result of wave influence on a flat beach.
Under wave activity, the beach face becomes steeper—the material is pushed onshore and
a berm is formed. This is especially a case with surveys UAV_11-14, when the beach area
continued to decrease under waves from the south, while the volume remained almost
the same. This indicates that the beach sediment was redistributed on the emerged beach
rather than lost offshore or onshore.
The peak storm wave energy varied between 8 and 27 m
2
s, whereas the storm intensity
index varied between 3 and 62 m
2
h. The latter is due to the duration of storms, which
changes with seasons. The changes in the beach volumes and the beach area are relatively
small and independent of these two wave energy parameters. However, the largest changes
J. Mar. Sci. Eng. 2022,10, 358 18 of 24
in beach elevation and in beach profiles are associated with storms, once the beach reaches
post-nourishment adjustment.
Next, the offshore wave steepness parameter was examined. The offshore wave
steepness (H
o
/L
o
), was found as potentially useful parameter for explaining offshore
versus onshore sediment transport on gravel beaches (see e.g., [
43
]). A threshold value of
0.01 is used to distinguish onshore (<0.01) and offshore sediment transport (>0.01). For the
peak storm waves in this study, the offshore wave steepness is between 0.04 and 0.05 and
this would indicate presence of offshore sediment transport. Figure 10 shows that changes
in beach elevations vary along the beach and there are several profiles (e.g., between
surveys 3 & 4) where sediment loss is larger than sediment gain. However, the survey
would need to extend offshore to examine whether these losses are due to offshore sediment
transport. Not only cross-shore but also longshore sediment transport and groundwater
dynamics affect the sediment gains and losses on these profiles.
5. Discussion
5.1. Use of UAVs for Monitoring of Beach Changes
This study demonstrated the effectiveness of UAVs and SfM photogrammetry for
repeating monitoring of beach changes, which can inform coastal management. Measure-
ments were performed quickly once fixed GCPs were set up. They are particularly suitable
for targeted surveys of beaches, which experience noticeable changes only after rare high
wave energy events. Using full 3D data (point clouds) rather than other 2D parameters
such as shoreline position and beach width, is useful for studies of gravel beaches, where
changes are spatially and temporally dynamic. For example, changes in the beach elevation
due to the formation or erosion of berms and the resulting beach steepening are prominent
after storms and important for beach safety [17].
The accuracy and precision of surveys needs to be taken into account when estimating
the beach parameters (e.g., [
39
]). These depend on many factors such as the resolution of
camera used, flight height and angle of the survey, and other environmental conditions
such as light and wind. In this study, two different camera models were used, different
flight heights and different GCP layouts. Hence, it was even more important to perform
a detailed analysis of the point clouds to establish horizontal and vertical errors for each
survey. Accuracy was tested by comparing point clouds from the stable region around
the beach, while precision was tested using ground control points spread over the stable
region. The upper part of the beach profiles derived from point clouds, which include a
stable promenade, were also qualitatively compared for any outliers. Horizontal RMSEs
for all surveys were less than 7 mm and vertical errors varied up to 10 mm. The largest
errors were associated with surveys taken during unfavourable environmental conditions
(light or presence of water on the beach and promenade). However, this was the case in
only two or three of the surveys and it did not affect the quality of images. The results
of the accuracy and precision analysis suggest that there is no significant difference be-
tween the point clouds delivered from consumer grade (DJI phantom 4) and professional
UAV (DJI Matrice 200, Sony ILCE-7M2). Inferior performance of consumer grade UAV is
compensated with lower flight altitude and a larger number of images. The conservative
uncertainty level values of
±
5 cm were used for estimating the beach parameters from
point clouds derived from both sets of cameras.
The inaccuracy of surveys was significantly lower than elevation changes measured
after storm events. However, as the overall changes in beach volumes, as well as amount of
material that is supplied to the beach each summer, are relatively small (50–150 m
3
), the
estimated total volume changes cannot be accurately determined. If the losses or nourished
material is evenly spread over the beach surface, it would result in thickness of 2.5 cm for
50 m
3
of material. Hence, the accuracy of surveys needs to be less than 2 cm. Instead, DEMs
changes that account for accuracy thresholds can be used to estimate volume changes
between surveys. This would result in consideration of smaller area of the beach and
lead to smaller errors. SfM photogrammetry with UAVs proved itself to be very useful
J. Mar. Sci. Eng. 2022,10, 358 19 of 24
for analyzing and describing changes that occurred on the emerged part of the beach.
However, it was observed that there is an offshore/onshore exchange of material from
submerged part of the beach. The future work should include reduction of vertical error
and underwater surveys.
5.2. Post-Nourishment Changes
The project provided unique opportunity for evaluating behaviour and longevity of
beach nourishment as result of storms. Longevity is defined here in a loose sense, meaning
time before significant sediment loss occur. As part of the Beachex project, the Ploˇce beach
was nourished at the end of January 2020 and it was exposed to a number of high energy
events, including four storm events defined as events with H
s
> 1.26 m and duration of at
least 1 h. The beach volumes seem to be quite stable, which is evident from analysed beach
volumes (Figure 12). A positive trend of the beach area for which uncertainty of estimates
are negligible, has been observed between first nourishment and second nourishment.
Some fluctuations of the beach area were observed after storm events, but overall the beach
area increased until May. The local council re-nourished the beach again in May to extend
the bathing area. By the end of the surveys in February 2021, the beach area had decreased
as a result of at least five storm events. It is worth mentioning that these changes represent
less than 6% of the average beach area. Most of the sediment has been lost from the eastern
part of the beach. This is evident from the elevation changes both in plan and through
transects (Figures 9and 11).
The nourishment itself involves flattening of steep beach profiles on the western side
and adding material on the eastern side, aligning it with WNW—SSE directions. The
main storm events are driven from the direction window SSE—S and after first storm,
the western part of the beach rotate slightly to align with dominant wave direction. This
was also observed on artificial