**1. Introduction**

Coastal flooding events, including their occurrence and triggers for the appearance of waves, have been investigated in the past to contribute to a better understanding of flood events in coastal areas as well as efficient risk management. Recent investigations have studied the occurrence of storm and sea wave events [1,2], the numerical modelling of wave propagation in coastal areas [3], flood water movement over land areas [4], the impact of storms on coasts [5] and flood vulnerability [6].

Special attention has been given to the development of an early warning system (EWS) for the timely warning of residents in exposed areas as well. Early warning systems are also an integral part of coastal flood management plans and contribute to the development of long-term management strategies [7]. Existing sea-state monitoring technology, historical databases, numerical forecasting models and computer science have been parts of the operational coastal flood early warning system [8]. A EWS modelling framework based on a Bayesian network has been used to link coastal hazards to their socio-economic and environmental factors [9] and connect available field measurements, data obtained from numerical wave simulations and an empirical wave run-up approach [10]. An EWS that determines the total sea level height by combining predictions of tides and sea level anomalies with wave runup estimates has been presented in [11].

Technological progress in computation and communication sciences in the last decade allows for work with large databases that can store registered meteorological data that affect the occurrence of sea floods. This type of data represents a basis for developing different

**Citation:** Nikoli´c, Ž.; Srzi´c, V.; Lovrinovi´c, I.; Perkovi´c, T.; Šoli´c, P.; Kekez, T. Coastal Flooding Assessment Induced by Barometric Pressure, Wind-Generated Waves and Tidal-Induced Oscillations: Kaštela Bay Real-Time Early Warning System Mobile Application. *Appl. Sci.* **2022**, *12*, 12776. https://doi.org/ 10.3390/app122412776

Academic Editor: Jürgen Reichardt

Received: 31 October 2022 Accepted: 8 December 2022 Published: 13 December 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**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/).

forecasting platforms, which use input data and boundary conditions from global or regional scales and open sea and weather forecast databases, providing wave propagation in the ports [12]. These platforms use a number of effective hydrodynamic numerical models for the simulation of storm surge and the prediction of sea water level [13]. It is important to notice that the sea water level prediction procedure in real time is a time-consuming process which needs significant computational resources, as well as the development and monitoring of a software architecture [14] to model coupling and integration. The models based on machine learning [15] and artificial intelligence provide easy implementation of numerical models with low computational cost, as well as fast training, validation, testing and evaluation [16].

In the present study, coastal flooding caused by the simultaneous combination of several mechanisms, such as tidal-induced sea level oscillations resulting from barometric pressure changes and wind-generated waves, have been investigated in the Kaštela Bay. A special focus has been given to the area of the city of Kaštela, placed at the north part of the bay (Figure 1), with many cultural and historical buildings and/or areas located near the coastline, which are potentially endangered by coastal flooding and subject to significant consequences and damage (Figure 2).

The design of the early warning system for sea flooding risk in the city of Kaštela includes the following components:


Recently, Internet of Things (IoT) systems that monitor in different scenarios to ensure a green, sustainable future have been widely used to create smart environments tailored to particular human needs [17,18]. Affordable equipment, miniaturized in its deployment, is ensured a long lifetime through solar/battery power; the communication ranges are also extremely increased and can be measured in kilometers in urban areas, which is especially pertinent in meteorological scenarios [19,20]. In this study, the established sea level monitoring system uses an IoT-based, solar-powered anemometer (Barani design—Meteo Wind) and solar-powered meteo-station (Barani design Meteo-helix) for monitoring air pressure.

The system is deployed to acquire current meteorological information used as an input for the developed model to estimate the sea level. The system is based on the LoRaWAN IoT radio, which delivers the information to The Things Network (TTN) cloud system, which is then used by the cross-platform web/mobile application called Waves, which is used as an early warning monitoring system. The multilanguage system itself implements: (1) the algorithm that defines sea level based on the tidal, wind speed/direction and air pressure information; (2) information on the shore height at the dedicated measurement and early warning zones; (3) push notification logics that can be separately activated based on user's zone of the interest. monitoring air pressure. The system is deployed to acquire current meteorological information used as an input for the developed model to estimate the sea level. The system is based on the LoRaWAN IoT radio, which delivers the information to The Things Network (TTN) cloud system, which is then used by the cross-platform web/mobile application called Waves, which is used as an early warning monitoring system. The multilanguage system itself implements: (1) the algorithm that defines sea level based on the tidal, wind speed/direction and air pressure information; (2) information on the shore height at the dedicated measurement and early warning zones; (3) push notification logics that can be separately activated based on user's zone of the interest. design—Meteo Wind) and solar-powered meteo-station (Barani design Meteo-helix) for monitoring air pressure. The system is deployed to acquire current meteorological information used as an input for the developed model to estimate the sea level. The system is based on the LoRaWAN IoT radio, which delivers the information to The Things Network (TTN) cloud system, which is then used by the cross-platform web/mobile application called Waves, which is used as an early warning monitoring system. The multilanguage system itself implements: (1) the algorithm that defines sea level based on the tidal, wind speed/direction and air pressure information; (2) information on the shore height at the dedicated measurement and early warning zones; (3) push notification logics that can be separately activated based on user's zone of the interest.

Recently, Internet of Things (IoT) systems that monitor in different scenarios to ensure a green, sustainable future have been widely used to create smart environments tailored to particular human needs [17,18]. Affordable equipment, miniaturized in its deployment, is ensured a long lifetime through solar/battery power; the communication ranges are also extremely increased and can be measured in kilometers in urban areas, which is especially pertinent in meteorological scenarios [19,20]. In this study, the established sea level monitoring system uses an IoT-based, solar-powered anemometer (Barani design—Meteo Wind) and solar-powered meteo-station (Barani design Meteo-helix) for

Recently, Internet of Things (IoT) systems that monitor in different scenarios to ensure a green, sustainable future have been widely used to create smart environments tailored to particular human needs [17,18]. Affordable equipment, miniaturized in its deployment, is ensured a long lifetime through solar/battery power; the communication ranges are also extremely increased and can be measured in kilometers in urban areas, which is especially pertinent in meteorological scenarios [19,20]. In this study, the established sea level monitoring system uses an IoT-based, solar-powered anemometer (Barani

*Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 3 of 20

*Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 3 of 20

(**a**) (**b**)

**Figure 1.** (**a**) Location of the city of Kaštela in Kaštela Bay; (**b**) A view of an old historical core, Kaštel Kambelovac. **Figure 1.** (**a**) Location of the city of Kaštela in Kaštela Bay; (**b**) A view of an old historical core, Kaštel Kambelovac. **Figure 1.** (**a**) Location of the city of Kaštela in Kaštela Bay; (**b**) A view of an old historical core, Kaštel Kambelovac.

**Figure 2.** Coastal flooding events in the city of Kaštela. **Figure 2.** Coastal flooding events in the city of Kaštela. **Figure 2.** Coastal flooding events in the city of Kaštela.

warning system for coastal flooding. It is based on frequently measured meteorological data and sea level predictions integrated on a single platform to provide real-time information about potential risk for the citizens of the city of Kaštela. Novelties of the manuscript are summarized as follows: (i) to our best knowledge, This paper presents a methodological approach for the development of an early warning system for coastal flooding. It is based on frequently measured meteorological data and sea level predictions integrated on a single platform to provide real-time information about potential risk for the citizens of the city of Kaštela. This paper presents a methodological approach for the development of an early warning system for coastal flooding. It is based on frequently measured meteorological data and sea level predictions integrated on a single platform to provide real-time information about potential risk for the citizens of the city of Kaštela.

This paper presents a methodological approach for the development of an early

the site of the application has been used for the very first time to demonstrate the application of the coastal flooding early warning system; (ii) the relevance of the selected site relies on the fact that the area is faced with coastal flooding more than 15 days per year on average, thus increasing the need for early warning system development, (iii) the man-Novelties of the manuscript are summarized as follows: (i) to our best knowledge, the site of the application has been used for the very first time to demonstrate the application of the coastal flooding early warning system; (ii) the relevance of the selected site relies on the fact that the area is faced with coastal flooding more than 15 days per year on average, thus increasing the need for early warning system development, (iii) the man-Novelties of the manuscript are summarized as follows: (i) to our best knowledge, the site of the application has been used for the very first time to demonstrate the application of the coastal flooding early warning system; (ii) the relevance of the selected site relies on the fact that the area is faced with coastal flooding more than 15 days per year on average, thus increasing the need for early warning system development, (iii) the manuscript couples three mechanisms contributing to the coastal flooding vulnerability assessment, those being barometric pressure-induced sea level changes, wind-generated waves and tidal-induced fluctuations, (iv) barometric- and tidal-induced sea level changes are obtained in real time, fully relying on local conditions arising from the observations, (v) wind-generated wave heights are site specific, taking into consideration bathymetric features and coastal structure type and (vi) the whole procedure has been implemented in the form of a real-time mobile application, thus resulting in a reliable tool for the end users.

Compared to relevant publications [16,21], this paper couples three of the mechanisms leading to the sea water level rise and refers to a site-specific area. Although it does not offer general findings, the procedure shows the potential to be applied to other sites all over the Mediterranean basin after prior modifications. The latter refers to bathymetric features, the reflection of the coastline, the determination of incident deep water wave parameters and tidal observations.

The paper is structured as follows: Section 2 is the Methodology and Study area section, which couples the methodology used in the study with representative site related data. Section 3 offers Results from the study, while Section 4 consists of relevant Discussion topics and Section 5 summarizes main conclusion points.

### **2. Methodology and Site Description**

### *2.1. Sea Water Elevation Prediction*

Sea water level is a random process resulting from the simultaneous combination of several mechanisms, three of which are found to be dominant, those being tidal-induced oscillations and oscillations due to barometric pressure changes and wind-generated waves. Although characterized by different time scales, those three mechanisms simultaneously contribute to the absolute sea water elevation definition.

Tidal-induced sea water oscillations are driven by tidal forcing, which is induced by simultaneous inter gravity forces between the Sun, Moon and Earth. Tidal-induced changes are characterized as a mixed semidiurnal type at the location of interest, with main periodic intervals corresponding to both semi diurnal and diurnal ones. Compared to tidal oscillations, sea level changes induced by barometric pressure changes are aperiodic with time scales corresponding to two main factors: (i) daily barometric pressure changes corresponding to daily scale air temperature change and (ii) time scales corresponding to the time necessary for the air mass transfer from different geographic locations to the location of interest to occur. The latter corresponds to time scales usually equal to several hours. Wind-generated waves are characterized by very small time scales, up to 8 s in the area of interest, and these are generated as a result of the air mass kinematic energy transfer to the sea surface.

Previous research [22] has shown that tides characterizing the Adriatic Sea basin consist mainly of seven dominant constituents, of which three are diurnal (O1, P1 and K1) and four are semidiurnal (N2, M2, S2 and K2). In its origin, each constituent represents a sinusoidal function with an associated amplitude, period, and phase, contributing to the full tidal signal. To determine the unknown values of the amplitudes, periods and phases, the original signal is initially transferred from the time to frequency domain by applying the Discrete Fourier Transform (DFT) [22]. DFT results are often plotted as Amplitude Spectral Density (ASD). By normalizing magnitudes with number of samples, the amplitude of each frequency can be easily obtained. Due to the fact that DFT calculations are time demanding [23], Fast Fourier Transform (FFT) [24] incorporated in Python in SciPy library [23] has been used for faster transfer of signals from the time to frequency domain. Based on the determined tidal constituent parameters, the tidal-induced sea level can be simulated using a linear superposition of sine functions corresponding to the number of relevant constituents:

$$h\_l = \sum\_{i=1}^{7} A\_i \times \sin\left(\frac{2\pi t}{t\_{pi}} + \frac{2\pi \varphi\_i}{360}\right), \ t = 0, \ 1, \ \ldots, M \tag{1}$$

where *h<sup>t</sup>* is the simulated sea level [m], *A<sup>i</sup>* is the amplitude [m], *tpi* is the period [h] and *ϕ<sup>i</sup>* phase [◦ ] of *i*-th constituent, *M* is the sample size and *t* is relative time [h].

Simulated sea level *hsea* can be calculated as a superposition of tidal harmonics from Equation (2) by adding the mean sea level value calculated from the observed sea level signal:

$$h\_{sea} = h\_t + h \tag{2}$$

where *h<sup>t</sup>* represents tidally induced sea level oscillations and *h* represents the mean sea level value.

Increasing the barometric pressure by 1 [cm] of the saltwater column leads to a decrease in the sea level by approximately 1 [cm] and inversely, decreasing the barometric pressure by 1 [cm] raises the sea level by 1 [cm] [25]. This effect is called the Invert Barometric effect (IB). The sea level rise caused by the change of the barometric pressure can be calculated as follows:

$$h\_{at} = \frac{-p\_a}{\rho \xi} \tag{3}$$

where *hat* is the sea level change resulting from a change in barometric pressure, *p<sup>a</sup>* is the barometric pressure [Pa] change, *ρ* is the density of sea water [kg/m<sup>3</sup> ] and *g* is the gravitational acceleration [m/s<sup>2</sup> ]. Therefore, the simulated sea level incorporating both tidal oscillations and barometric pressure-induced changes is updated from Equation (2) for the value of *hat*, as shown in Equation (3):

$$h\_{sea} = h\_t + h\_{at} + \overline{h} \tag{4}$$

where the third right hand term (*h*) represents mean sea level values as calculated from the observed signal, *h<sup>t</sup>* represents tidally induced sea level oscillations and *hat* stands for barometric pressure-induced sea level change.

To assess wind-generated wave height in front of the coastline, we start with a determination of deep water wave parameters. For the location of interest and relevant incident wave directions, fetch length has been assessed by applying the Saville method [26]. In order to determine deep water wave parameter values, data sets from the Section 2.3 climatological station have been used. Fully developed sea conditions have been checked by applying the Wilson criteria [27] prior to the determination of wave parameters. After defining relevant fetch length for a given wind duration and wind velocity, the Groen−Dorrestein nomogram is used to determine deep water wave parameters [28].

Wave transform analysis has been performed by applying SMS CGWAVE software [29]. The wave phenomena that can be simulated by CGWAVE are: bathymetric refraction, diffraction by structures (e.g., breakwaters) and bathymetry, reflection (from structures, natural boundaries (seawalls, coastlines, etc.) and bed slopes), friction, wave breaking and floating (fixed) docks influence the wave field. The model is based on the use of a triangular finite element formulation to solve the two-dimensional elliptic mild slope equation, with grid sizes varying throughout the domain based on the local wavelength [30]. The grid can be efficiently generated using the SMS graphical interface when a bathymetry file is provided. The model allows one to specify the desired reflection properties along the coastline and other internal boundaries. While the basic equation is intended for monochromatic waves, irregular (i.e., spectral) wave conditions are simulated in CGWAVE through a linear superposition of monochromatic simulations [31,32].

The procedure or the algorithm for the absolute sea water elevation assessment is based on three steps which are presented above and summarized as: (i) sea level change caused by tidal forcing prediction, (ii) sea level change caused by a barotrophic pressure forcing assessment and (iii) assessment of the wind-generated wave height in front of the coastline. Sea water elevation assessment as a consequence of simultaneous effects for those three mechanisms is assessed as follows:


$$h\_{\text{sea}} = h\_t + h\_{at} + h\_{S(d=4\text{m})} \tag{5}$$

which offers an easy-to-implement way to assess absolute sea level, where *h<sup>t</sup>* represents the tidally induced component, *hat* represents the barometric pressure-induced component and *hS*(*d*=4m) stands for the significant wind-generated height as found 4 m offshore. The value of *hsea* is expressed relative to HVRS71 datum [33].

### *2.2. Site Description*

The location of the study area with definition of the tidal gauge and meteorological stations are shown in Figure 3. The meteorological station Split with geographical coordinates *ϕ* = 43◦310 N, λ = 16◦260 E is located northwest of the city Split, on the Marjan hill, at the altitude of 122 m a.s.l. The terrain is a slope towards the sea on the SW-W-NW-N side. To the north is the bay of Kaštela and to the east is the city of Split. At a distance of 7 km toward the north is the mountain Kozjak with the highest peak at 779 m a.s.l. The measuring system Fuess [34] is installed at a 5 m high terrace column on the building roof (12 m above ground). The only obstacles found near the meteorological station building are found to the west and northwest side (trees and terrain elevation overgrown with trees more than 10 m tall). The roughness class of the terrain is 2.5 (z<sup>0</sup> = 0.2) [34]. The wind speed and direction data obtained by the classic Feuss measurement consist of an average hourly wind speed with corresponding wind.

The nearest tidal gauge near Kaštela Bay is located within the Institute of Oceanography and Fisheries (IOR) in Split, at the western border of peninsula Marjan. The time series of measured sea surface elevations over a total duration of five years have been obtained from the Marjan tide gauge for the period from 1 January 2010 to 31 December 2014. The sampling frequency of the tide gauge was set to 1 h for the entire time series. The measured water level elevation is referred to as the HVRS71 vertical datum [33]. Due to the maintenance and malfunction of the tide gauge IOR Marjan, the observed time series are characterized by periods without recorded values of the sea level. For the purpose of this work, a continuous time series is required so the longest continuous time series has been found from 25 January 2010 at 12:00 h to 27 June 2011 at 23:00 h with a total of 12,444 h of data.

Barometric pressure data are obtained from two meteorological stations: Split-Airport and Split-Marjan. The meteorological station Split-Airport is located at 21 m a.s.l. and the barometric pressure values are recorded three times a day at 7 am, 2 pm and 9 pm. The meteorological station Split-Marjan is located at 122 m a.s.l. and has an hourly measurement frequency including barometric pressure values. Logs of barometric pressure were obtained from both meteorological stations for the period from 1 January 2010 to 31 December 2014. Due to the location of Kaštela Bay, the meteorological station Split-Airport was considered as relevant, but due to its low measurement frequency (3 times per day), the data from Split-Marjan was used for further analysis. First, the data from Split-Marjan was compared with the data from Split-Airport and a mean difference of 12.67 hPa with a standard deviation of 0.67 hPa was found. All data from Split-Marjan were corrected for the values of 12.67 hPa. The available data from the Split-Marjan station has a continuous record with no missing data for the entire 5-year period. However, since the sea level signal does not have the same continuity as the barometric pressure, the same period of barometric pressure as sea level is included in the further analysis.

Insight into bathymetric features emphasize depth values up to 37 m with a pretty uniform decrease towards the shoreline (Figure 4). A bathymetric survey for the purpose of this paper has been performed in April 2021 with a single beam setup. Shoreline

*Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 7 of 20

determination of the reflection coefficient has been done at the site by the inspection of both shoreline type and depth in front (Figure 5). sea level signal does not have the same continuity as the barometric pressure, the same period of barometric pressure as sea level is included in the further analysis.

the barometric pressure values are recorded three times a day at 7 am, 2 pm and 9 pm. The meteorological station Split-Marjan is located at 122 m a.s.l. and has an hourly measurement frequency including barometric pressure values. Logs of barometric pressure were obtained from both meteorological stations for the period from 1 January 2010 to 31 December 2014. Due to the location of Kaštela Bay, the meteorological station Split-Airport was considered as relevant, but due to its low measurement frequency (3 times per day), the data from Split-Marjan was used for further analysis. First, the data from Split-Marjan was compared with the data from Split-Airport and a mean difference of 12.67 hPa with a standard deviation of 0.67 hPa was found. All data from Split-Marjan were corrected for the values of 12.67 hPa. The available data from the Split-Marjan station has a continuous record with no missing data for the entire 5-year period. However, since the

**Figure 3.** Location of the study area with the definition of the tidal gauge and meteorological stations (IOR—Institute of Oceanography and Fisheries). **Figure 3.** Location of the study area with the definition of the tidal gauge and meteorological stations (IOR—Institute of Oceanography and Fisheries). *Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 8 of 20

**Figure 4.** Bathymetric features at the location of interest. ing it perfect for a scenario in which sensor devices collect wind and barometric data and **Figure 4.** Bathymetric features at the location of interest.

**Figure 5.** Shoreline reflection coefficient along the coastline.

*2.3. Data Collection from LoRaWAN-Based Sensor Device* 

LoRaWAN is one of the most widely used Low Power Wide Area technologies that aims to collect and communicate data from an end sensor device at a large distance, mak-

**Figure 4.** Bathymetric features at the location of interest.

**Figure 5.** Shoreline reflection coefficient along the coastline. **Figure 5.** Shoreline reflection coefficient along the coastline.

### *2.3. Data Collection from LoRaWAN-Based Sensor Device 2.3. Data Collection from LoRaWAN-Based Sensor Device*

LoRaWAN is one of the most widely used Low Power Wide Area technologies that aims to collect and communicate data from an end sensor device at a large distance, making it perfect for a scenario in which sensor devices collect wind and barometric data and LoRaWAN is one of the most widely used Low Power Wide Area technologies that aims to collect and communicate data from an end sensor device at a large distance, making it perfect for a scenario in which sensor devices collect wind and barometric data and convey them to a centralized system. LoRaWAN employs a typical star-of-star topology where end devices communicate data in a single hop to one or more gateway devices. These messages are further forwarded to the network and application server for further processing, allowing authorized data to be forwarded to other external services (e.g., using MQTT message forwarding). LoRaWAN allows battery-operated devices to periodically transmit sensor data over large distances, while minimizing consumption during inactive periods. LoRaWAN technology finds its application in smart city/smart agriculture environments where there is no need for real-time (every second) transmission from end devices. During the inactive period, end devices simply cut off the consumption, allowing a battery lifetime up to a couple of years without any external power source.

Figure 6 depicts the architecture of the MeteoHelix weather station of Barani design that utilizes LoRaWAN as a radio technology. MeteoHelix [35] is an automatic all-inone microweather station which is solar powered and can be active for up to 6 months without sun. It measures air temperature to WMO accuracy, air humidity to WMO accuracy with dew and frost point output, barometric pressure and solar irradiation (pyranometer). Another sensor of Barani design was also installed that utilizes LoRaWAN communication— MeteoWind IoT PRO. MeteoWind [36] is used for wind monitoring and employs two sensors: a separate wind vane and anemometer. MeteoWind allows 4+ months of battery life without sun and a maintenance-free service life with long-term measurement stability due to its elliptical cup and metal construction. As depicted in Figure 7, both devices are placed at a 10 m height without any object around within 150 m so that both the wind speed and wind direction are not distorted, while both are located in location of Kaštela Bay. Since both MeteoHelix and MeteoWind IoT PRO employ LoRaWAN communications to convey data over the air to the centralized system, The Things Network as a service provider was used to collect data for further processing. As a LoRaWAN gateway, an indoor Sentrius RG1xx LoRaWAN gateway device placed around 150 m from the sensor devices was employed that forward messages to The Things Network (TTN) cloud infrastructure. Once the message arrives at the gateway, it is forwarded to the TTN Network and Application server. Furthermore, TTN allows message forwarding from TTN infrastructure to our

dedicated Waves Seafront Monitoring app using MQTT protocol, which is described more in detail in the following section. server. Furthermore, TTN allows message forwarding from TTN infrastructure to our dedicated Waves Seafront Monitoring app using MQTT protocol, which is described more in detail in the following section. dedicated Waves Seafront Monitoring app using MQTT protocol, which is described more in detail in the following section.

*Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 9 of 20

*Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 9 of 20

convey them to a centralized system. LoRaWAN employs a typical star-of-star topology where end devices communicate data in a single hop to one or more gateway devices. These messages are further forwarded to the network and application server for further processing, allowing authorized data to be forwarded to other external services (e.g., using MQTT message forwarding). LoRaWAN allows battery-operated devices to periodically transmit sensor data over large distances, while minimizing consumption during inactive periods. LoRaWAN technology finds its application in smart city/smart agriculture environments where there is no need for real-time (every second) transmission from end devices. During the inactive period, end devices simply cut off the consumption, allowing a battery lifetime up to a couple of years without any external power source.

convey them to a centralized system. LoRaWAN employs a typical star-of-star topology where end devices communicate data in a single hop to one or more gateway devices. These messages are further forwarded to the network and application server for further processing, allowing authorized data to be forwarded to other external services (e.g., using MQTT message forwarding). LoRaWAN allows battery-operated devices to periodically transmit sensor data over large distances, while minimizing consumption during inactive periods. LoRaWAN technology finds its application in smart city/smart agriculture environments where there is no need for real-time (every second) transmission from end devices. During the inactive period, end devices simply cut off the consumption, allowing a battery lifetime up to a couple of years without any external power source.

Figure 6 depicts the architecture of the MeteoHelix weather station of Barani design that utilizes LoRaWAN as a radio technology. MeteoHelix [35] is an automatic all-in-one microweather station which is solar powered and can be active for up to 6 months without sun. It measures air temperature to WMO accuracy, air humidity to WMO accuracy with dew and frost point output, barometric pressure and solar irradiation (pyranometer). Another sensor of Barani design was also installed that utilizes LoRaWAN communication— MeteoWind IoT PRO. MeteoWind [36] is used for wind monitoring and employs two sensors: a separate wind vane and anemometer. MeteoWind allows 4+ months of battery life without sun and a maintenance-free service life with long-term measurement stability due to its elliptical cup and metal construction. As depicted in Figure 7, both devices are placed at a 10 m height without any object around within 150 m so that both the wind speed and wind direction are not distorted, while both are located in location of Kaštela Bay. Since both MeteoHelix and MeteoWind IoT PRO employ LoRaWAN communications to convey data over the air to the centralized system, The Things Network as a service provider was used to collect data for further processing. As a LoRaWAN gateway, an indoor Sentrius RG1xx LoRaWAN gateway device placed around 150 m from the sensor devices was employed that forward messages to The Things Network (TTN) cloud infrastructure. Once the message arrives at the gateway, it is forwarded to the TTN Network and Application

Figure 6 depicts the architecture of the MeteoHelix weather station of Barani design that utilizes LoRaWAN as a radio technology. MeteoHelix [35] is an automatic all-in-one microweather station which is solar powered and can be active for up to 6 months without sun. It measures air temperature to WMO accuracy, air humidity to WMO accuracy with dew and frost point output, barometric pressure and solar irradiation (pyranometer). Another sensor of Barani design was also installed that utilizes LoRaWAN communication— MeteoWind IoT PRO. MeteoWind [36] is used for wind monitoring and employs two sensors: a separate wind vane and anemometer. MeteoWind allows 4+ months of battery life without sun and a maintenance-free service life with long-term measurement stability due to its elliptical cup and metal construction. As depicted in Figure 7, both devices are placed at a 10 m height without any object around within 150 m so that both the wind speed and wind direction are not distorted, while both are located in location of Kaštela Bay. Since both MeteoHelix and MeteoWind IoT PRO employ LoRaWAN communications to convey data over the air to the centralized system, The Things Network as a service provider was used to collect data for further processing. As a LoRaWAN gateway, an indoor Sentrius RG1xx LoRaWAN gateway device placed around 150 m from the sensor devices was employed that forward messages to The Things Network (TTN) cloud infrastructure. Once the message arrives at the gateway, it is forwarded to the TTN Network and Application server. Furthermore, TTN allows message forwarding from TTN infrastructure to our

**Figure 6.** Architecture of LoRaWAN-enabled MeteoHelix and MeteoWind devices from Barani De-**Figure 6.** Architecture of LoRaWAN-enabled MeteoHelix and MeteoWind devices from Barani Design. **Figure 6.** Architecture of LoRaWAN-enabled MeteoHelix and MeteoWind devices from Barani Design.

**Figure 7.** Installation of MeteoHelix and MeteoWind IoT Pro devices.

### *2.4. Waves Seafront Monitoring Architecture—Overview and Functionalities* **Figure 7.** Installation of MeteoHelix and MeteoWind IoT Pro devices.

Waves Seafront Monitoring is a cross-platform mobile application that allows users to access and interactively view the current sea level, discretized by coastal segments with respect to coastal height. As shown in Figure 8a, it comprises a cloud and client side. At the cloud side, the Google Firebase server component executes an evaluation algorithm based on data from the sensor it communicates with, serves as a server for these results and performs user authentication. Sensor data comprising air pressure, wind speed and direction are sent to The Things Network (TTN) cloud via the LoRaWAN communication channel, which is forwarded via MQTT protocol to the TTN microservice. Once the packet with sensor data arrives, the TTN microservice captures and stores the LoRaWAN uplink data into the database and forwards the data through the PMO algorithm to estimate sea level according to the data arriving from LoRaWAN sensors (wind speed and direction, barometric pressure and sea tide level). As can be seen, an alarm notification can be sent to the application if the sea level exceeds a predefined level. *2.4. Waves Seafront Monitoring Architecture—Overview and Functionalities*  Waves Seafront Monitoring is a cross-platform mobile application that allows users to access and interactively view the current sea level, discretized by coastal segments with respect to coastal height. As shown in Figure 8a, it comprises a cloud and client side. At the cloud side, the Google Firebase server component executes an evaluation algorithm based on data from the sensor it communicates with, serves as a server for these results and performs user authentication. Sensor data comprising air pressure, wind speed and direction are sent to The Things Network (TTN) cloud via the LoRaWAN communication channel, which is forwarded via MQTT protocol to the TTN microservice. Once the packet with sensor data arrives, the TTN microservice captures and stores the LoRaWAN uplink data into the database and forwards the data through the PMO algorithm to estimate sea level according to the data arriving from LoRaWAN sensors (wind speed and direction, barometric pressure and sea tide level). As can be seen, an alarm notification can be sent to the application if the sea level exceeds a predefined level.

**Figure 8.** (**a**) High-level overview of the implementation of the solution, (**b**) Front-end interface of the mobile application. 2. Log out button from the application (Figure 8b, number 4). **Figure 8.** (**a**) High-level overview of the implementation of the solution, (**b**) Front-end interface of the mobile application.

The front-end of the application is comprised of the following elements, as depicted

tion of registration if he does not have an account by clicking on the appropriate

2. Prominent logo—the project was realized within the Interreg PMO GATE col-

B. Registration Interface—the registration interface is analogous to the login interface, only it allows the user to create an account to access the application by entering login information. To simplify logging into the system, there is no special validation of user

C. Map interface (Figure 8b) —the central part of the application, which contains the

adjusts the interface language (Figure 8b, number 3).

1. At the top is a drop-down menu where the user has a choice between three languages: Croatian, English and Italian. By clicking on an option, the interface text

laboration (Figure 8b, number 2).

link (Figure 8b, number 1),

accounts, such as email confirmation.

following basic functionalities:

in Figure 8b:

The front-end of the application is comprised of the following elements, as depicted in Figure 8b:

	- 1. Form—the user enters his login details (e-mail and password) and has the option of registration if he does not have an account by clicking on the appropriate link (Figure 8b, number 1),
	- 2. Prominent logo—the project was realized within the Interreg PMO GATE collaboration (Figure 8b, number 2).
	- 1. At the top is a drop-down menu where the user has a choice between three languages: Croatian, English and Italian. By clicking on an option, the interface text adjusts the interface language (Figure 8b, number 3).
	- 2. Log out button from the application (Figure 8b, number 4).
	- 3. Values read from the sensor are updated over time depending on how often the data from the TTN arrives. From top to bottom, sea level is the sum of altitudes caused by sea tide, wind speed and direction and barometric pressure (Figure 8b, number 5). The total amount is calculated based on the algorithm submitted by the Client.
	- 4. Google maps with plotted polygons that correspond to discretized segments of the coast according to their heights (Figure 8b, number 6). The color of the zone corresponds to the early warning status: green—safe, i.e., the sea level is below the coast level; yellow—warning, i.e., the sea can exceed 20 cm above the height of the shore; orange—dangerous, i.e., the sea can rise up to 50 cm above the height of the shore; red—flooded, i.e., the sea exceeds 50 cm above the height of the coast. By clicking on an individual zone, the cards will position themselves next to the corresponding zone and the map will be centered on the selected zone. Based on the coastal height data, the monitoring area is divided into 10 zones.
	- 5. Zone maps showing the names and photos of coastal zones. Here, the user can see exactly the height of this segment of the coast, the estimated sea level in relation to the zone and, consequently, the situation in that zone, which is coded in colors analogous to the zones on the map (Figure 8b, number 7). Each zone information contains the sea level with respect to the coastal height, where the number with the minus sign shows how much the sea level is below the coastal height. Once the number becomes positive, the zones change color since this result corresponds with estimated flood. By moving the tabs left or right, the user can focus on a specific zone, and the map will center on that zone.
	- 6. Notification button (Figure 8b, number 8). By clicking on this button individually for each zone, the user can indicate whether he wants to receive notifications when the situation in a particular zone changes.

### **3. Results**

### *3.1. Sea Level Determination Based on the Tidal Fluctuations*

To enable the assessment of tidal-induced sea level oscillations, a total of 12,444 h of observed time series data was used to perform DFT and obtain an amplitude spectrum (Figure 9). Due to the nature of the observed signal, inspection of the amplitude spectrum offers the presence of the trend, visible within the bins corresponding to the lowest frequencies. In total, seven tidal harmonics has been identified as dominant, thus ensuring the tidal-induced sea level oscillation characterization.

3. Values read from the sensor are updated over time depending on how often the data from the TTN arrives. From top to bottom, sea level is the sum of altitudes caused by sea tide, wind speed and direction and barometric pressure (Figure 8b, number 5). The total amount is calculated based on the algorithm submitted

4. Google maps with plotted polygons that correspond to discretized segments of the coast according to their heights (Figure 8b, number 6). The color of the zone corresponds to the early warning status: green—safe, i.e., the sea level is below the coast level; yellow—warning, i.e., the sea can exceed 20 cm above the height of the shore; orange—dangerous, i.e., the sea can rise up to 50 cm above the height of the shore; red—flooded, i.e., the sea exceeds 50 cm above the height of the coast. By clicking on an individual zone, the cards will position themselves next to the corresponding zone and the map will be centered on the selected zone. Based on the coastal height data, the monitoring area is divided into 10

5. Zone maps showing the names and photos of coastal zones. Here, the user can see exactly the height of this segment of the coast, the estimated sea level in relation to the zone and, consequently, the situation in that zone, which is coded in colors analogous to the zones on the map (Figure 8b, number 7). Each zone information contains the sea level with respect to the coastal height, where the number with the minus sign shows how much the sea level is below the coastal height. Once the number becomes positive, the zones change color since this result corresponds with estimated flood. By moving the tabs left or right, the user

can focus on a specific zone, and the map will center on that zone.

when the situation in a particular zone changes.

*3.1. Sea Level Determination Based on the Tidal Fluctuations* 

the tidal-induced sea level oscillation characterization.

6. Notification button (Figure 8b, number 8). By clicking on this button individually for each zone, the user can indicate whether he wants to receive notifications

To enable the assessment of tidal-induced sea level oscillations, a total of 12,444 h of observed time series data was used to perform DFT and obtain an amplitude spectrum (Figure 9). Due to the nature of the observed signal, inspection of the amplitude spectrum offers the presence of the trend, visible within the bins corresponding to the lowest frequencies. In total, seven tidal harmonics has been identified as dominant, thus ensuring

by the Client.

zones.

**3. Results** 

**Figure 9.** Amplitude Spectral Density for observed sea level reduced for the mean value. **Figure 9.** Amplitude Spectral Density for observed sea level reduced for the mean value.

While tidal amplitudes and periods were determined directly from the spectrum shown in Figure 9, phases were determined from the complex component of each of the While tidal amplitudes and periods were determined directly from the spectrum shown in Figure 9, phases were determined from the complex component of each of the seven tidal harmonics. All characteristic values for the relevant tidal constituents are shown in Table 1. The quality of the determined harmonic values were checked by the application of the correlation coefficient when applied to both the simulated and observed time series.


**Table 1.** Tidal constituents with corresponding values of amplitude, period and phase.

Both the observed and simulated sea level signals, one for each 1000 h of the available time series, are compared by using Pearson correlation coefficient and root mean square error. When simulating the sea level by using Equation (2), the simulated signal discovers the absence of the trend within, which is incorporated in the next step as a response to longterm barometric pressure changes. The residual was determined as a difference between the observed and simulated signals. Both signals are shown relative to the HVRS71 vertical datum, as explained in the Methodology section. The same procedure was repeated for each 1000 h data of the total 12,444 h data representing the total sample. Both Pearson correlation coefficient and root mean square error values are shown in Table 2.

**Table 2.** Pearson coefficient and root mean square values obtained between the measured and simulated sea water elevation by Equation (2).


For the entire observed and simulated set of 12,444 h data, the root mean square value equals to 0.136 [m], while Pearson's correlation coefficient equals 0.578 [-]. The high value of RMSE and low value of the correlation coefficient indicate the sea level fluctuations are not only subjected to tidal variations, but also to other factors.

### *3.2. Sea Level Determination Based on the Tidal Fluctuations and Barometric Pressure Changes*

Figure 10 shows the effect of barometric pressure on the simulated signal with seven frequencies. The simulated sea level signal has a similar trend to the observed sea level signal with a residual remaining between the simulated and observed signal. When the sea level signal is simulated using the tidal components and the inverted barometric effect, the RMSE decreases to 0.1057 [m] and the Pearson correlation coefficient increases to 0.777 [-], further highlighting the effect of barometric pressure and its contribution to the observed sea level definition at the location of interest (Table 3). The decrease in the residual values when both the barometric pressure and tidal effects are involved in the procedure for the sea level determination implies the relevance of those two mechanisms in the vertical movement of sea surface elevation. *Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 13 of 20

**Figure 10.** (**a**) Observed and simulated sea level by Equation (4) for the first 1000 h of observed sea level signals, (**b**) Observed barometric pressure, (**c**) Residual between simulated and observed sea level signals*.*  **Figure 10.** (**a**) Observed and simulated sea level by Equation (4) for the first 1000 h of observed sea level signals, (**b**) Observed barometric pressure, (**c**) Residual between simulated and observed sea level signals.

**Table 3***.* Pearson coefficient and root mean square values obtained between the measured and simulated sea water elevation by Equation (4). **Table 3.** Pearson coefficient and root mean square values obtained between the measured and simulated sea water elevation by Equation (4).

Following the knowledge from the location of interest, two relevant wave incident directions have been identified, SE and SSW. The reason for the selection of those two incident directions relies on three facts: (i) longest possible fetch lengths compared to other incident directions, (ii) longest wind duration and wind velocities and (iii) relative position of the shoreline almost perpendicular to the incident directions. Table 4 offers the highest significant wave height equal to 2.78 m with a corresponding period of 4.76 s. For

**V (ms−1) t (h) Fmin (km) tmin (h) FEFF (km) FMJ (km) HS (m) TS (s) L0 (m) 1.5** 13 43.36 8.65 24.8 24.8 0 **3.3** 13 71.25 6.02 24.8 24.8 0 **5.4** 13 97.17 4.8 24.8 24.8 0.5 2.6 10.55 **7.9** 13 123.48 4.03 24.8 24.8 0.85 3.15 15.49 **10.7** 17 215.89 3.5 24.8 24.8 1.21 3.65 20.8 **13.8** 15 213.49 3.12 24.8 24.8 1.7 4 24.98 **17.1** 11 159.77 2.82 24.8 24.8 2.19 4.35 29.54 **20.7** 4 45.07 2.59 24.8 24.8 2.78 4.76 35.38


*3.3. Wind-Generated Waves* 

and 4.15 s.

**Table 4.** Deepwater wave parameters—incident direction SE*.* 

### *3.3. Wind-Generated Waves*

Following the knowledge from the location of interest, two relevant wave incident directions have been identified, SE and SSW. The reason for the selection of those two incident directions relies on three facts: (i) longest possible fetch lengths compared to other incident directions, (ii) longest wind duration and wind velocities and (iii) relative position of the shoreline almost perpendicular to the incident directions. Table 4 offers the highest significant wave height equal to 2.78 m with a corresponding period of 4.76 s. For the SSW incident direction, Table 5 shows the highest wave parameters equal to 2.35 m and 4.15 s.


**Table 4.** Deepwater wave parameters—incident direction SE.

**Table 5.** Deepwater wave parameters—incident direction SSW.


The procedure for wave height determination at the distance of 4 m offshore starts with known wind velocity and duration. For the selected incident wave direction and fetch length, deep wave parameters have been selected from Tables 4 and 5. For the purpose of wave height determination in front of the shoreline, the JONSWAP spectrum is used to incorporate a real wave into the wave transform model. Spectrum parameters have been set up as shown in Table 6, where the obtained deep water wave values correspond to a 100-year return period.

**Table 6.** Deepwater wave parameters—JONSWAP spectrum parameters.


Wave height spatial distribution and direction for the SE incident wave direction is presented in Figure 11. Insight into the inner zone of interest discovers the presence of refraction as a main wave transform mechanism. The insight into significant wave height changes at 4 m offshore is shown in Figures 12 and 13. The same has been shown for the SSW incident direction in Figures 12 and 13. By comparing modeled wave heights, it is obvious the SE incident direction results in more significant wave height values.

**Table 5.** Deepwater wave parameters*—*incident direction SSW*.* 

**Table 6***.* Deepwater wave parameters—JONSWAP spectrum parameters*.* 

**Incident Direction Hs (m) T (s) γ nn**  SE (165°) 3.05 6.40 3.30 4.00 SSW (202.5°) 2.60 5.90 3.30 4.00

obvious the SE incident direction results in more significant wave height values.

spond to a 100-year return period.

**V (ms−1) t (h) Fmin (km) tmin (h) FEFF (km) FMJ (km) HS (m) TS (s) L0 (m) 1.5** 5 11.71 6.68 17.4 11.71 0 **3.3** 5 19.24 4.65 17.4 17.4 0 **5.4** 5 26.24 3.7 17.4 17.4 0.5 2.5 9.76 **7.9** 5 33.35 3.11 17.4 17.4 0.77 2.95 13.59 **10.7** 5 40.37 2.7 17.4 17.4 1.05 3.15 15.49 **13.8** 5 34.91 2.41 17.4 17.4 1.41 3.6 20.23 **17.1** 5 26.94 2.18 17.4 17.4 1.92 3.98 24.73 **20.7** 5 6.75 2 17.4 6.95 2.35 4.15 26.89

The procedure for wave height determination at the distance of 4 m offshore starts with known wind velocity and duration. For the selected incident wave direction and fetch length, deep wave parameters have been selected from Tables 4 and 5. For the purpose of wave height determination in front of the shoreline, the JONSWAP spectrum is used to incorporate a real wave into the wave transform model. Spectrum parameters have been set up as shown in Table 6, where the obtained deep water wave values corre-

Wave height spatial distribution and direction for the SE incident wave direction is presented in Figure 11. Insight into the inner zone of interest discovers the presence of refraction as a main wave transform mechanism. The insight into significant wave height changes at 4 m offshore is shown in Figures 12 and 13. The same has been shown for the SSW incident direction in Figures 12 and 13. By comparing modeled wave heights, it is

**Figure 11.** Wave heights and directions for the SE incident direction—local study area. **Figure 11.** Wave heights and directions for the SE incident direction—local study area*.* 

**Figure 12***.* Wave heights and directions for the SSW incident direction—local study area*.*  **Figure 12.** Wave heights and directions for the SSW incident direction—local study area.

The monitoring system has been installed and put into operation at the Kaštela Bay test site to provide real-time measurement of wind speed, wind direction and barometric pressure. The observed data is used for the real-time prediction of sea level based on the developed algorithm for the estimation of sea level heights and information obtained from the monitoring system, which has been integrated into the algorithm. The real-time prediction of sea level values is obtained based on tidal fluctuations, barometric pressure val-

Resulting sea level values are compared with coastline elevation in order to identify the potential danger of flooding and issue on-time warnings. Figure 14 shows wind characteristics and barometric pressure as observed during a period of 7 days during February 2022. The data containing barometric pressure along with wind direction and speed were collected into a database for the application. The data were used as an input for an algo-

*3.4. Real-Time Prediction of Sea Water Elevation and Warning* 

ues and wind characteristics.

rithm to estimate sea level height.

**Figure 12***.* Wave heights and directions for the SSW incident direction—local study area*.* 

**Figure 13***.* Modeled wave heights along the control line—incident directions: SE and SSW. **Figure 13.** Modeled wave heights along the control line—incident directions: SE and SSW.

**Figure 11.** Wave heights and directions for the SE incident direction—local study area*.* 

### *3.4. Real-Time Prediction of Sea Water Elevation and Warning*

*3.4. Real-Time Prediction of Sea Water Elevation and Warning*  The monitoring system has been installed and put into operation at the Kaštela Bay test site to provide real-time measurement of wind speed, wind direction and barometric pressure. The observed data is used for the real-time prediction of sea level based on the developed algorithm for the estimation of sea level heights and information obtained from the monitoring system, which has been integrated into the algorithm. The real-time prediction of sea level values is obtained based on tidal fluctuations, barometric pressure val-The monitoring system has been installed and put into operation at the Kaštela Bay test site to provide real-time measurement of wind speed, wind direction and barometric pressure. The observed data is used for the real-time prediction of sea level based on the developed algorithm for the estimation of sea level heights and information obtained from the monitoring system, which has been integrated into the algorithm. The real-time prediction of sea level values is obtained based on tidal fluctuations, barometric pressure values and wind characteristics.

ues and wind characteristics. Resulting sea level values are compared with coastline elevation in order to identify the potential danger of flooding and issue on-time warnings. Figure 14 shows wind characteristics and barometric pressure as observed during a period of 7 days during February 2022. The data containing barometric pressure along with wind direction and speed were collected into a database for the application. The data were used as an input for an algo-Resulting sea level values are compared with coastline elevation in order to identify the potential danger of flooding and issue on-time warnings. Figure 14 shows wind characteristics and barometric pressure as observed during a period of 7 days during February 2022. The data containing barometric pressure along with wind direction and speed were collected into a database for the application. The data were used as an input for an algorithm to estimate sea level height. *Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 16 of 20

**Figure 14.** LoRaWAN-obtained observed time series: (**a**) wind speed, (**b**) wind direction and (**c**) barometric pressure. **Figure 14.** LoRaWAN-obtained observed time series: (**a**) wind speed, (**b**) wind direction and (**c**) barometric pressure.

Figure 15 shows sea water elevation as a consequence of each of the components (tidal effect, barometric pressure and wind-generated waves) as well as absolute sea water elevation as a superposition of the abovementioned components for the period from 15 December 2021–7 February 2022. The results are given for one randomly selected zone along the study area to demonstrate the capacity of the proposed procedure to assess the sea water elevation. The estimated water surface elevation is referred to the HVRS71 vertical reference datum determined on the basis of mean sea level, as stated in the Method-Figure 15 shows sea water elevation as a consequence of each of the components (tidal effect, barometric pressure and wind-generated waves) as well as absolute sea water elevation as a superposition of the abovementioned components for the period from 15 December 2021–7 February 2022. The results are given for one randomly selected zone along the study area to demonstrate the capacity of the proposed procedure to assess the sea water elevation. The estimated water surface elevation is referred to the HVRS71 vertical reference datum determined on the basis of mean sea level, as stated in the Methodology section.

ology section.

**Figure 14.** LoRaWAN-obtained observed time series: (**a**) wind speed, (**b**) wind direction and (**c**)

Figure 15 shows sea water elevation as a consequence of each of the components (tidal effect, barometric pressure and wind-generated waves) as well as absolute sea water elevation as a superposition of the abovementioned components for the period from 15 December 2021–7 February 2022. The results are given for one randomly selected zone along the study area to demonstrate the capacity of the proposed procedure to assess the sea water elevation. The estimated water surface elevation is referred to the HVRS71 vertical reference datum determined on the basis of mean sea level, as stated in the Method-

**Figure 15.** Calculated sea level height during the period from 15 December 2021–7 February 2022 caused by: (**a**) tidal effect, (**b**) barometric pressure-induced sea level changes, (**c**) wind-generated wave height and (**d**) absolute sea level height.

If the resulting sea level value estimated from the algorithm exceeds the coastline elevation, the warning is immediately issued by the mobile application to alert the endangered population. Flood hazard warnings in the mobile application are primarily focused on citizens' safety and their classification is based on the particular threat level. Threat levels are based on tests of the instability of a human body in floodwater and different stability thresholds [37]. Following the results of the tests, the first threat level, issuing a yellow alert, is defined for flood water depth on the coastline up to 20 cm, which is mostly threatening for children. The second level of threat corresponds to an orange warning, which is now becoming dangerous for adults and is issued when the water depth is between 20 cm and 50 cm. Finally, an exceedance of flood water depth over 50 cm on the coastline corresponds to the third threat level. In this case, severe injuries can be caused by flood water and the red warning for the population is issued.

### **4. Discussion**

barometric pressure.

ology section.

In this paper, a procedure to assess sea water elevation caused by coastal flooding at Kaštela Bay is presented, together with the monitoring system infrastructure scheme and sea water elevation prediction module. In its origin, the procedure relies on the realtime monitoring IoT-based system for sea level-relevant parameter monitoring to capture wind-generated wave heights and barometric pressure-induced sea level changes.

The latter offers the possibility to determine wind-generated wave heights in front of the coast, with an incorporation of shoreline reflection features as assessed by the sea water depth and type of structure found at the site, and long- and short-term changes in the mean sea water level. Tidal-induced sea level changes are forecasted based on the data series observation from the nearest relevant tidal gauge.

After wind speed/direction and barometric pressure sensors are installed at the site to enable real-time monitoring, a LoRaWAN-based IoT monitoring system was employed, along with the dedicated cross-platform mobile application, which was used as an early

warning notification system. Therein, registered end users can set early warning notifications based on their zone of interest.

As stated, the origin of the presented approach relies on the capability of the monitoring system to observe the wind speed/direction and barometric pressure at the site. The selection of the location of the monitoring infrastructure is performed at the location of interest to neglect spatial variations in barometric pressure variations and offer reliable input for barometric pressure-induced sea level changes.

Relevant mechanisms acting to assess the sea water elevation are selected based on the phenomena characterizing the study area and fit suitable for the pre-assumed conditions mostly found on site. The effect of barometric pressure changes and their influence on mean sea water elevation refers to the sea in a calm state and static conditions when no significant changes in the barometric pressure field are observed. In case this requirement is not met, such as when the storm is moving, a difference in predicted sea level changes compared to real ones can be obtained.

Our results demonstrate the capability of the proposed methodology to enable a realistic and reliable forecast of the sea water elevation by taking into consideration local conditions: (i) wind information, (ii) barometric pressure, (iii) tidal forcing, (iv) water depth and (v) coastal structure or type information.

When coupled with the shoreline's absolute height, one is enabled to assess flooding risk or the vulnerability of the coast to sea level-induced flooding. The developed monitoring system, coupled with data collection from LoRaWAN-based sensor device and assessment procedure, presents a unique and robust approach to be used along different coastal areas of the Adriatic Sea.

Apart from this, different coastal structures found along the specific site can easily be taken into consideration by assessing the reflectance features from the structure type and bathymetric features in front of the coastline.

### **5. Conclusions**

This paper presents a procedure to assess the coastal flooding induced by the sea level in the Kaštela study area. In its origin, it consisted of three modes: (i) the LoRaWAN-based sensor device monitoring infrastructure located at the site to capture relevant parameters in real time, (ii) a step-by-step procedure to assess the sea water elevation and (iii) Waves Seafront Monitoring cross-platform mobile application to enable the real-time insight into coastal flooding occurrence. The main conclusions obtained from the study can be summarized as:


**Author Contributions:** Conceptualization, Ž.N., V.S. and P.Š.; Data curation, V.S., I.L., T.P., P.Š. and T.K.; Funding acquisition, Ž.N.; Investigation, Ž.N., V.S., I.L., T.P., P.Š. and T.K.; Methodology, Ž.N., V.S., P.Š. and T.K.; Software, V.S., I.L. and T.P.; Supervision, Ž.N., V.S. and P.Š.; Validation, Ž.N., V.S., I.L., T.P. and P.Š.; Visualization, I.L. and T.K.; Writing—original draft, Ž.N., V.S., I.L., T.P., P.Š. and T.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the EUROPEAN UNION, Programme Interreg Italy-Croatia, the project "Preventing, managing and overcoming natural-hazards risks to mitigate economic and social impact"—PMO-GATE ID 10046122. The research is also partially supported through project KK. 01.1.1.02.0027, co-financed by the CROATIAN GOVERNMENT and the EUROPEAN UNION through the European Regional Development Fund—the Competitiveness and Cohesion Operational Programme and the project "Internet of Things: Research and Applications", UIP-2017-05-4206, financed by the Croatian Science Foundation.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**

