Coastal Sea Level Monitoring in the Mediterranean and Black Seas

. Spanning over a century, a traditional way to monitor sea level variability by tide gauges is – in combination with modern observational techniques like satellite altimetry – an inevitable ingredient in sea level studies over the climate scales and in coastal seas. The development of the instrumentation, remote data acquisition, processing and archiving in last decades allowed for extending the applications towards a variety of users and coastal hazard managers. The Mediterranean and Black 50 seas are an example for such a transition – while having a long tradition for sea level observations with several records spanning over a century, the number of modern tide gauge stations are growing rapidly, with data available both in real-time and as a research product at different time resolutions. As no comprehensive survey of the tide gauge networks has been carried out recently in these basins, the aim of this paper is to map the existing coastal sea level monitoring infrastructures and the respective data availability. The survey encompasses description of major monitoring networks in the Mediterranean and Black 55 seas and their characteristics, including the type of sea level sensors, measuring resolutions, data availability and existence of ancillary measurements, altogether collecting information about 236 presently operational tide gauge stations. The availability of the Mediterranean and Black seas sea level data in the global and European sea level repositories has been also screened and classified following their sampling interval and level of quality-check, pointing to the necessity of harmonization of the data available with different metadata and series at different repositories. Finally, an assessment of the networks’ capabilities 60 for their usage in different sea level applications has been done, with recommendations that might mitigate the bottlenecks and assure further development of the networks in a coordinated way, being that more necessary in the era of the human-induced climate changes and the sea level rise.

All mentioned hazards are present in the Mediterranean and the Black Sea (M/BS hereafter) and pose a threat to densely populated coastal areas, cultural heritage and historical cities lying near the shore (Fig. 1) (Reimann et al., 2018). This is particularly relevant for some regions exposed to substantial sea level variations spanning over a range of frequencies ( Fig.   1), from minutes (like meteotsunamis or tsunamis) through hours, days, weeks (storm surges, or planetary wave forcing) and seasonal oscillations, to interannual variability and decadal trends (Pugh and Woodworth, 2014). 80 As the Mediterranean and Black seas are microtidal basins (Tsimplis et al., 1995), the atmospherically-driven component of sea level (storm surge) is often the most common cause of extreme coastal sea levels. Conjoined with windgenerated waves and/or intense precipitation (also related to increased river discharge), flood risks during a storm surge may lead to devastating flooding events (Bevacqua et al., 2019). According to Cid et al. (2016), Tunisian (Gulf of Gabes), Aegean and Adriatic coasts undergo the highest number of sea level extreme events per year. In the Black Sea, the western coast is the 85 most exposed to storm surges (Bresson et al., 2018). Low elevation areas, deltas and sinking land areas (e.g. Venice Lagoon and Po Delta in Italy, Nile Delta in Egypt, Ebro Delta in Spain) are also subjected to the high flood risks during extreme events (Ferrarin et al., 2021;El-Fishawi, 1989;Hereher, 2015;Grases et al., 2020). Consequently, coastal zone management and protection bodies are carrying out extensive sea level measurements to support coastal flooding forecasts and issue timely alerts to the population, like The Tide Monitoring and Forecast Centre of the City of Venice that maintains a network 90 comprising tens of tide gauges. From time to time, exceptional high sea levels in the Adriatic Sea threaten particularly the city of Venice (acqua alta phenomena), causing severe flooding and disruption of people's lives. As an example, on November 12, 2019, sea level reached 1.89 m (~ 1.3 m surge contribution), the second highest storm surge since 1966 event (1.94 m) (Ferrarin et al., 2021). Occasionally, at other areas, less extreme storm surge events (~ 0.5 m) are also able to cause, in combination with waves, substantial damage to infrastructure, coastal erosion and flooding episodes (e.g. the storm Gloria hitting the Spanish 95 Mediterranean coast in January 2020: Amores et al., 2020;. Tsunamis are amplified coastal sea level oscillations with periods ranging from minutes to hours, mainly generated by strong submarine earthquakes. The convergence of the African and Eurasian plates makes them a likely hazard in the Mediterranean (Tinti and Maramai, 1996;Tinti et al., 2001;Papadopoulos and Papageorgiou, 2012;Papadopoulos et al., 2014, Maramai et al., 2014Samaras et al., 2015), where around 10% of all tsunamis worldwide occur, being particularly 100 destructive in the Hellenic Arc area (Fig. 1b). The earthquakes can reach a magnitude of 7.5 to 8 there, triggering > 5-6 m wave heights. The hazard is lower in the Western MS, but tsunami waves over 1 m can reach most of the M/BS locations (Sørensen et al., 2012;Álvarez-Gómez et al., 2011). A well-known ancient tsunami is the one generated by a strong earthquake (magnitude 8 -8.5) off Crete in 365 A.D., that caused many deaths and damage in the Middle East and all the way up the Adriatic Sea, and the destruction of Alexandria port and library (event t4 in Fig. 1b). Also, in Greece, around 1600 B.C. a giant 105 tsunami triggered by the collapse of Santorini volcano is considered to have caused the end of the Minoan civilization in Crete (event t5 in Figure 1b). Tsunamis can also be generated by landslides and volcanic eruptions, but these events are mostly localised (e.g. Stromboli volcano, Tinti et al., 2005). Recent tsunami events (like the 9 July 1956 Amorgos 12-m high tsunami, Okal et al., 2009, the 21 May 2003 Algerian tsunami reaching the Balearic Islands shores as a 2-m wave, Alasset et al., 2006;https://doi.org/10.5194/os-2021-125 Preprint. Discussion started: 20 January 2022 c Author(s) 2022. CC BY 4.0 License. All the stations transmit 1 min data with 1 min latency to PdE, the National Geographic Institute (National Tsunami Warning System) and the IOC Sea Level Station Monitoring Facility (SLSMF). Automatic quality control and processing is applied every 15 min for integration in the multi-model sea level forecasting system . In addition, 2 Hz raw data are processed every hour to characterize higher frequency sea level oscillations with periods of minutes (García-Valdecasas et al., 2021). Delayed-mode quality-control and processing is performed annually and monthly mean sea levels are 225 sent to the Permanent Service for Mean Sea Level (PSMSL). The data are also available through CMEMS INS TAC (NRT product), EMODnet and GESLA datasets. Mediterranean stations operated by PdE are listed in Table A1.

SOCIB tide gauge network in the Balearic Islands
The SOCIB tide gauge network (Tintoré et al., 2013; was established in 2009 and currently consists of six stations around the Balearic Islands, five of which are located on Mallorca and one on Ibiza. The length of the nearly-continuous tide 230 gauge records varies between 5 and 12 years, with the shortest series dating back to 2016. With the exception of the tide gauge in Sant Antoni (Ibiza) that is a radar gauge, the rest are pressure gauges. The sampling frequency is 1 min for all tide gauges.
Together with sea level, all stations measure atmospheric pressure, sampled every 30 s, and five of them also monitor water https://doi.org/10.5194/os-2021-125 Preprint. Discussion started: 20 January 2022 c Author(s) 2022. CC BY 4.0 License. temperature every minute. All data are freely distributed in near-real time through the SOCIB website and are also available through the CMEMS INS TAC data portal. Sea level observations are referenced to the tide gauge benchmarks whose positions 235 are controlled on a yearly basis through GNSS surveys. Stations operated by SOCIB are listed in Table A2.

Spanish National Geographic Institute tide gauge network
The tide gauge network of the National Geographic Institute of Spain (IGN) has a set of sensors that gather changes and variations of the mean sea level over the time. It started in the 19th century, when three tide gauges were set up in order to determine the national altimetry datum in Alicante, Santander and Cadiz. The purpose was to establish the required 240 infrastructure to start levelling works. The tide gauge network has been extended and its instruments have been improved since then, including recent upgrades from float to radar sensors. Nowadays the National Geographic Institute has ten tide gauges: five are located in Iberian Peninsula, one in Alboran Island and four in the Canary Islands. Only five are on the Spanish Mediterranean coast. All of them have one or two radar sensors and are linked to GNSS permanent stations. A sea level dataset starting in 1870 has been recently published for Alicante, based on data from historic float gauges and modern radar sensors 245 at this harbour (Marcos et al., 2021). The Mediterranean part of the network is complemented by five IDSL in Cartagena, La Mola de Mahon, Ciutadella, Ceuta.
In addition to maintenance works, network management and connection to High Precision Levelling Network (REDNAP), National Geographic Institute is analysing the historical series of its tide gauges. It is also comparing mean sea level changes with GNSS observations among which a new technique is standing out. It is called GNSS reflectometry  and it is still under study. Mediterranean stations operated by IGN are listed in Table A3.

Spanish Institute of Oceanography tide gauge network
The sea level data network operated by the Instituto Español de Oceanografía (IEO) consists of 11 stations, six on the Iberian Peninsula, one at Ceuta on the Northern Africa coast, one in the Balearic Islands and 3 in the Canary Archipelago. For historical operative reasons, most of these locations are those in which IEO local headquarters are located. Each tide gauge station is 255 equipped with two sensors: an analogue float-type tide gauge with digital encoder and a radar-based one.
The analogue network is one of the oldest ones in Spain, with some of the measurements dating back to 1943.
Historical data are made available through SeaDataNet data portal (www.seadatanet.org). Each of the analogue sensors consist of a float gauge, mechanically connected to an analogue-digital encoder which converts the data to an on-site computer. The radar sensors duplicate the measurements and ensures the measurements continuity when one of the two devices fails. 1 min 260 sampled data are locally stored and recovered via modem by the central data centre once a day, where the data quality is assessed and archived.
The routines for data recovery, quality assessment, detection of high frequency events and data representation are currently being updated, and it is expected that the frequency of data availability will reach 1 per minute by the end of the year

Spanish Hydrographic Office
Currently, the Spanish Hydrographic Office (IHM) is making a great effort in the installation of a permanent tide gauge network along the national coast. Usually, the IHM installs tide gauges in the hydrographic works area on a temporary basis to be able to calculate the real or reduced probe data. Once the work is finished, the tide gauge is usually removed.
Since 2021, the IHM has installed permanent tide gauges in different ports on the coast: Rosas, Castellón (in the 270 Mediterranean coast) and Huelva (in the Gulf of Cadiz). The intention is to install 8 new tide gauges during 2022, with the objective to further increase density of tide gauges along the Spanish coast. The equipment consists of acoustic sensors with a frequency of 1 Hz and with spatial positioning in real time through a GNSS. At the moment, the web is under development.
Mediterranean stations operated by IHM are listed in Table A5.

L'Estartit tide gauge (Meteolestartit, in collaboration with ICM/CSIC-Spain) 275
The tide gauge is a part of the Meteorological and Oceanographic Station (Meteolestartit) at the harbour of L'Estartit, a coastal town at the Catalan coast, in the NW Mediterranean. It is a float gauge with an analogical record on paper. Recordings are collected every week and digitised with a 2 h resolution. Paper records are preserved for further detailed analyses, if required in some special circumstances, such as seiches. The position of the tide gauge is georeferenced every 5 years by the Catalan Cartographic Institute and sea level data is backwards linearly corrected for each period. Sea level record collection at this 280 point started in January 1990, as part of Meteolestartit, which started in 1969, as a personal initiative of Josep Pascual with the collaboration with the Marine Sciences Research Institute in Barcelona (ICM/CSIC). Data collected included basic meteorological and oceanographic data. More details can be found in Pascual and Salat (2019) and Salat et al. (2019). Details about L'Estartit tide gauge are in Table A6.

SHOM tide gauges network RONIM, France 285
French Naval Hydrographic and Oceanographic Service (SHOM) has been observing the tides for many years, but the RONIM network, as it exists today, was initiated in 1992, mainly to meet the needs of tide prediction and reduction of bathymetric surveys. It was based on a network which was already in place and which consisted, in 1996, of five tide gauges. It was then densified to reach 23 tide gauges in 2007, including 5 in the Mediterranean (Marseille, Toulon, Nice, Monaco and Ajaccio). transmission. The acquisition rate is 1 Hz, and these data are sent directly to SHOM and tsunami warning service by a VPN link.
The 1 min data are computed at SHOM and transmitted in near-real-time to the data.shom.fr website and to the IOC website with a latency of 5 min. The satellite messages are clocked every 6 min. The data transmitted by satellite are also on 300 the IOC website. As there is a double transmission system, on the IOC site, the stations are in double (ex: Toulon (internet transmission) and Toulon2 (satellite transmission)).
The 10 min data is calculated by the tide gauge data logger. They are retrieved once a day and transmitted to the data.shom.fr website.
A major upgrade of the RONIM network is underway (2021/2022). The data logger and the real time transmission process will be replaced, and two additional sensors (webcam + meteo station) will be added at most of tide gauges. A unique supervision system will be implemented for all tide gauges, allowing for a better assessment of real time recording and transmission issues and an improvement of the network reliability. Mediterranean stations operated by SHOM are listed in 310 Table A7.

ISPRA tide gauge networks along the Italian coast
The Italian Institute for Environmental Protection and Research (ISPRA) comprehensively and systematically provides highresolution estimates for the physical state of the Italian seas as well as real-time monitoring at national and local level. The marine observation system includes two sea level measurement networks: Italian Tide Gauge Network (Rete Mareografica  315 Nazionale -RMN), which continuously monitors the sea level and a number of related meteorological and physical parameters, and the North Adriatic and Venice Lagoon Tide Gauge Network (Rete Mareografica della Laguna di Venezia e del Litorale Nord Adriatico -RMLV) which is used for the real-time storm surge prediction and warning system.
The RMN network is a crucial source of information related to sea level. It provides data useful for analysing sea level variations, predicting storm surges and developing a tsunami early warning system. The RMN consists of 36 measuring 320 stations uniformly distributed along the Italian coast, mainly located within harbours. Some of these stations have been operating since the 1970s. The tidal-wave measurements for the entire network are provided by two different instruments (radar and float) and can be simply configured by a remote command. Each measurement station is also equipped with meteorological sensors: anemometer (wind speed and direction at 10 m above ground level), barometric sensor, multiparametric sensor for humidity and temperature, that are necessary for the real-time evaluation of sea and weather 325 conditions. Four stations (Venezia, Crotone, Gaeta, Carloforte) are collocated with GNSS instruments, to detect the horizontal and vertical displacement of the cabin.
The RMLV network is composed of 26 tide gauge stations equipped for the systematic and widespread measurement of water level and other related parameters, such as wind direction, wind speed, atmospheric pressure, precipitation, and wave-https://doi.org/10.5194/os-2021-125 Preprint. Discussion started: 20 January 2022 c Author(s) 2022. CC BY 4.0 License. heights in the Lagoon of Venice and along the North Adriatic coastline. Lots of the RMLV stations have been operating for 330 several decades and Venezia -Punta della Salute station for more than 150 years. Real-time data represent one of the main utilities, fundamental for prediction and warnings of exceptional or atypical high tides (storm surges). Two RMLV stations (Venezia Punta della Salute and Grado) are collocated with GNSS in order to detect the continuous vertical shift of the local Zero Tide Level which is the reference benchmark for tide measurements in the Lagoon of Venice (ZMPS). The real-time operability of this network is crucial for several purposes such as: analysis and elaboration of data referring in particular to 335 extreme events (storm surges), signalling and forecasting exceptional high tides. Moreover, data from the Venice Lagoon and North Adriatic tide gauge network (RMLV) is an important source for planning Venice defence from the phenomena of high tides and for scientific studies on sea level variations.
Some IDSL stations were also tested, but the only one still operational is in Marina di Teulada.  Table   A8. 345

The Joint Research Center (JRC) IDSL network
Acknowledging the low quantity of sensors deployed on the Mediterranean coasts, the JRC started investigating the adoption of retail technology in 2014, in order to produce a low-cost solution. Since 2014, with the adoption of a few custom components, the reliability of The Inexpensive Device for Sea Level (IDSL) measurements further evolved retaining its best characteristics: low cost (about 2k€), high frequency (5 secs), local intelligence (detection of anomalous sea oscillations 350 through assessing deviation from moving averages).
Funded by the IOC-UNESCO, three campaigns delivered devices throughout the Mediterranean, the Black Sea and the North-East Atlantic. Another collaboration led to providing Indonesia with six IDSL devices. All data collected by the IDSL network are available online through the TAD Server at the JRC web site. Stations operated in Italy by JRC with ISPRA are listed in Table A9. 355

The tide gauge station of Trieste, Molo Sartorio, Italy
The Italian National Research Council (CNR) operates one station through the Trieste branch of the Institute of Marine Sciences (ISMAR). The station, included in the GLOSS Core Network with No. 340, is located at Molo (Pier) Sartorio, in the harbour of Trieste, and it is equipped with two float instruments with digital encoder, and one 50-year old fully analogue tide gauge. Direct sea level measurements are performed at least twice a month to check the instrument stability. The data quality 360 control is performed in delayed mode at least once a year. https://doi.org/10.5194/os-2021-125 Preprint. Discussion started: 20 January 2022 c Author(s) 2022. CC BY 4.0 License.
The tide gauge cabin also hosts two barometers, one of which is digital and one analogue. A GNSS receiver, operated by the University of Bologna, is mounted on top of the building that includes the tide gauge cabin.
The earliest sea level measurements were made in 1859, and since then the tide gauge remained on the same pier; more details can be found in Zerbini et al. (2017). Unfortunately, from 1875 to 1939, only monthly mean sea levels are available 365 (with a few gaps); they can be retrieved from the Permanent Service for Mean Sea Level (PSMSL) data bank. Hourly data are available for 1939-onwards (Raicich, 2019), while 1 min data exist since 2001. Hourly means are available as Fast Delivery data from the University of Hawaii Sea Level Center (UHSLC) since June 2009. The review of pre-1939 data including the digitisation of the available charts is in progress. Details about station Trieste Molo Sartorio are in Table A10.

The mareographic station in Koper, Slovenia 370
Operational sea level monitoring in Slovenia began in 1958 with the construction of the tide gauge station in Koper, for which hourly sea level measurements are available since 1961. The station is operated by the Slovenian Environment Agency and is collocated with the GNSS station. The existing float-type sensor in a stilling well was upgraded in 2005 with an additional radar sea level sensor, having 1 mm accuracy and 10 min sampling time. It provides sea level data, sea temperature data at 1 m of depth, GNSS data and essential meteorological data (air pressure, wind, air temperature, relative humidity, solar 375 irradiance). The quality-control is automatic, while additional manual controlling of sea level measurements is performed weekly at the station location. Details about station Koper are in Table A11.

Croatian tide gauge network
The Croatian tide gauge network is operated by the Andrija Mohorovičić Geophysical Institute (AMGI), Hydrographic Institute of the Republic of Croatia (HHI), Institute of Oceanography and Fisheries (IOF), National Agency for Water 380 Management Hrvatske vode (Croatian Waters, HV), and by the Ruđer Bošković Institute (RBI). The network consists of permanent stations based on float-type technology installed in stilling wells, established during the 20th century, and of radartype stations installed from 2004 onward.
Float-type stations: Systematic measurements of sea level in Croatia started in 1929 when the AMGI (Zagreb) established a tide gauge station in Bakar. In the same year, sea level measurements were initiated in Split Harbor by the HHI, 385 but the station was destroyed by bombing in World War II. During the 1950s four long-term coastal stations were installed: Dubrovnik (1955), Split Harbour (1955 and Rovinj (1955)  them with up to 15 min latency to home institutions. Exceptions are stations Golubinka and Prosika from which data is collected on-site at least once a year. At most stations digital data is automatically quality checked, processed, and displayed 395 (http://geo101.gfz.hr/~bakar/index_files/, https://adriaticsea.hhi.hr, http://vodostaji.voda.hr/) in real-time mode or once a day (Bakar), while analogue data are digitized at hourly resolution, and mostly processed once a year. Maintenance of stations includes regular checks of recorder zero (mainly twice a year), cleaning of the stilling-well and connecting pipe (once in two to five years) and levelling of the contact-point level against tide gauge benchmarks. All stations are levelled towards the national geodetic datum. Most time series have shorter gaps due to various reasons. Monthly and yearly averages from the 400 AMGI, HHI and IOF stations have been sent to the PSMSL since the stations were established.
In addition, the HHI plans to install two additional radar-type tide gauges in the following years in the areas of Šibenik and Mali Lošinj. Stations operated by Croatian institutions are listed in Table A12. 425 https://doi.org/10.5194/os-2021-125 Preprint. Discussion started: 20 January 2022 c Author(s) 2022. CC BY 4.0 License.

Montenegrin tide gauge Network
The Institute of Hydrometeorology and Seismology, Department of Hydrography, is responsible for monitoring and maintenance of the tide gauges installed in the Montenegrin part of the Adriatic Sea. The network includes two permanent tide gauge stations based on float-type technology. The tide station in Bar was established in 1965 and till 1991 was the part of Former Yugoslavia tide gauge network (Slovenia, Croatia, Montenegro). During the 1990s the tide station in Bar was not fully 430 operational for a long time, but it has been restored, and re-connected to the national geodetic network through levelling. The second Montenegrin tide gauge station was established in Kotor in 2010. For both stations the sea level is measured once every 6 min, with GSM-based data retrieval to the central server. Stations operated by Montenegro are listed in Table A13.

Tide monitoring network in Albania
The Institute of Geoscience, Department of Hydrology, is responsible for monitoring all water resources in Albania, including 435 the sea level. The sea level observations are taken manually, two times a day -at 7 am and 7 pm. These data are stored in a book by the observer and sent every month to the Institute of Geoscience (Tirana, Albania). At the centre, this information is archived and not controlled, unless there is a request for this data. Stations operated by Albania are listed in Table A14.

Sea level observations in the Maltese Islands
The routinely collection of sea level data in the Maltese Islands was initiated in May 1993 by the Physical Oceanography Unit 440 (later Physical Oceanography Research Group) using an ENDECO-type 1029/1150 differential pressure gauge in Mellieha Bay on the northern coast of Malta. The station remained in operation until 2001, measuring in delayed mode every 2 min, supplemented with meteorological measurements collected at a nearby station in Ramla tal-Bir overlooking the southern Comino Channel. This endeavour was mainly intended to assess the sea level variability and to study the phenomenology of strong seiches locally known as the 'milghuba' (Drago, 2000(Drago, , 2009. the University of Malta as well as to the JRC TAD server (https://webcritech.jrc.ec.europa.eu/TAD_server/Device/555) with a temporal frequency of 5 s. Each station is equipped with a radar sensor connected to electronics that also measures air temperature and captures visual images of the sea state every 15 min.
In March 2021, a Radac WaveGuide sensor was installed at the tip of the Marsaxlokk breakwater to measure sea 460 level, wave height and wave period. The instrument was procured through the SIMIT-THARSY project, partially funded by the ERDF through the Operational Programme. This radar sea level gauge is capable of measuring water displacement with a resolution of 3 mm at a frequency of 10 Hz. The wave height is measured with an accuracy of 1 cm at 1 min intervals. In front of it, on the other end of the bay, another IDSL was deployed. The operational data from these stations is linked to other  Table A15.

Greek tide gauge network operated by the Hellenic Navy Hydrographic Service
The Hellenic Navy Hydrographic Service (HNHS) monitors sea level variability over the Aegean and the Ionian Seas through a network of 22 tide gauges. All hydrographic stations consist of analogue float type gauges equipped with rotating drums, 470 rotating with speed of 1 cm/hr. These tidal stations are located within port facilities to record continuously sea level and provide these measurements to the relevant port authorities. The tidal stations at Thessaloniki, Kavala and Piraeus are the oldest of the network, operating continuously since 1933.
Since 1990, HNHS upgraded the network by installing digital water level sensors at seven selected stations (Piraeus, Alexandroupolis, Kalamata, Katakolo, Lefkada, Siros and Chios), operating in parallel to the analogue systems. These stations 475 collect, store and transfer water level data to central servers in near-real-time mode. Digital stations collect additionally a series of ancillary parameters, like the sea surface temperature and meteorological data (air temperature, humidity, barometric pressure and wind speed and direction). In parallel, during that period, the network's tidal data, recorded in paper charts, have been digitized into hourly values, and subsequently organized and stored in the HNHS databases. The water level error is estimated to approximately 1 cm (Tsimplis, 1994). 480 Data from four stations (Piraeus, Katakolo, Siros and Kalamata) are visualised on-line through the HNHS web page (https://www.hnhs.gr/en/online-2/tide-graphs). Data from all tidal gauges covering the 1969-2020 period are also transferred to the Permanent Service for Mean Sea level (PSMSL) (https://www.psmsl.org/), supported by the International Oceanographic Commission (Bitharis et al., 2017). Sea level data from these four digital stations are also provided and visualized in real-time mode to the IOC -SLSMF (http://www.ioc-sealevelmonitoring.org/). Stations operated by HNHS are 485 listed in Table A16.  Table   A18. 495

National Institute for Marine Research and Development "Grigore Antipa" coastal sea level monitoring, Romania
Sea level measurements started in Romania in 1856, at the initiative of the European Danube Commission. However, these data have not been preserved, Regular recordings of the sea level started in Romania in 1933, by installing a float gauge.
Presently, two methods of measurements are used in Constanta: a pressure sensor and a float gauge, with a hydrometric sight, 500 at which visual measurements are performed three times a day for data quality purpose. The accuracy of the analogue measurements on a paper chart are estimated to 1 mm. At Sulina and Mangalia, the measurements are made using pressure sensors. The sea level data is transmitted to the server via GPRS/GSM method.
In Sulina, Mangalia and Constanta three IDSL were deployed in the last decade. Presently, only the one in Mangalia is operational. Stations operated by NIMRD "Grigore Antipa" are listed in Table A19. 505

Russian tide gauge network in the Black Sea
Sea level observations on the Russian coast of the Black Sea began in the middle of the 19th century. Systematic measurements of the sea level were initiated in 1873. Since 1944, the sea level network was restored, reconstructed and expanded. Sea level float-type recorders were installed at many tide gauges. Since 1977, all tide gauges have been tuned in a single system of heights (the Baltic Height System). In total, during the years, aat the territory of the USSR on the Black Sea, there were 44 510 tide gauges, the data from which were saved until 1985. For 23 tide gauges, there are long-term digital series of hourly observations of sea level. Today short-period sea level variations on the Russian coast are measured at five tide gauges: Tuapse, Sochi, Sevastopol, Yalta, and Feodosia. The Russian tide gauge network is owned by The Russian Federal Service for Hydrometeorology and Environmental Monitoring (Roshydromet) and is operated by the All-Russian Research Institute of Hydrometeorological Information -World Data Center and data availability. Tuapse tide gauge is co-located with a GNSS. 515 Russian tide gauge stations in the Black Sea are listed in Table A20.

Coastal sea level monitoring in Türkiye
Sea level monitoring activities in Türkiye date back to the mid-1930s, when the first float-type tide gauge was installed at Antalya harbour to determine the national vertical datum. Since then, a considerable number of temporary mechanical and analogue tide gauges had been deployed at different spots. In 1999 the Turkish National Sea Level Monitoring System 520 (TUDES) programme was initiated by the General Directorate of Mapping. Up to 2011, under this programme, a network consisting of 20 digital tide gauges with acoustic sounding tubes, for continuous measurements, was established. Due to the significant maintenance problems related to the acoustic gauges and the strong need for the VLM monitoring, all TUDES stations have been replaced with radar gauges after 2015, and most are GNSS collocated. The data averaged at 30 s and 15 min intervals are transmitted to the data center in Ankara in near-real time through GSM and internet. Real-time and delayed 525 mode quality controls, data analysis, database management, and data distribution activities are performed at the data center.
Further, TUDES is delivering sea level data to the regional networks (e.g. ICG/NEAMTWS). More information about the TUDES can be found at https://tudes.harita.gov.tr/?lang=us.
In addition, three IDSL stations were deployed in Bozcaada, Samsun and Bodrum. TUDES tide gauge stations are listed in Table A21, while in Table A22 are listed JRC IDSL stations in Türkiye operated by KOERI.  (Papazachariou, 2014). 545 Furthermore, studies were carried out to examine the trend of the sea level variations at the Paphos station (Loizides et al, 2010), as well as inter-comparison of the Paphos time series data with satellite altimetry time series (Banks et al. 2003;Papazachariou et al., 2014) https://doi.org/10.5194/os-2021-125 Preprint. Discussion started: 20 January 2022 c Author(s) 2022. CC BY 4.0 License.
The Paphos MedGLOSS/ESEAS tide gauge was equipped with a Paroscientific Digiquartz Intelligent pressure sensor pressure type, while the rest three Cyprus MedGLOSS/ESEAS tide gauges were equipped with AANDERAA pressure sensors. 550 In the frame of the Interreg THALCHOR2 project, a new tides gauge network named PYTHEAS has been deployed in Cyprus during 2018 (Chris et al., 2020). The PYTHEAS tides gauges network consists of four stations (Paralimni and Pomos fisheries shelters and Paphos and Larnaca harbours), all owned by the Cyprus Governmental Department of Lands and Surveys (DLS), while the fifth one (old Limassol harbour) is owned by the Cyprus University of Technology. All the PYTHEAS sea level networks were equipped with radar tide gauge sensors, along with atmospheric sensors, such as pressure, air temperature, 555 humidity, as well as sea temperature. The station at the old Limassol harbour was also set to serve as a GNSS reference. The data from the PYTHEAS tides gauges network are transmitted in real-time to the hydrographic database of the DLS.
In parallel and independent from the PYTHEAS network, an additional acoustic tide gauge station was deployed at the end of March 2018 by the European JRC-Joint Research Center at the Zygi fisheries shelter, next to the location of the Cyprus MedGLOSS/ESEAS Zygi station. The 5 s data from this station are provided in real time also to the IOC-SLSMF 560 (http://www.ioc-sealevelmonitoring.org/station.php?code=zygi1).
In 2019, an offshore station for sea level variation was deployed in the frame of Interreg HERMES project using a bottom mounted AWAC ADCP (Acoustic Current Doppler Profiler) pressure sensor. The ADCP was deployed at the water depth of 40 m in the Larnaca Bay, close to the famous scuba diver shipwreck "Zenobia". The ADCP measures in addition to the sea level variation, sea currents at 20 water depths, waves, sea temperature and suspended particles. The data are transmitted 565 in real time via cable from the bottom mounted ADCP to a connected surface oceanographic buoy, and then via a GPRS to the ORION NGO server for use in the dedicated HERMES web page (Zhuk et al., 2020). Cyprus stations are listed in Table A23.

Egyptian tide gauge network
Six tide gauges were operational along the Egyptian Mediterranean coast. These gauges were deployed at Port Said, Burullus new harbour, Abu-Qir Bay, Alexandria Western Harbour, Sidi Abdel-Rahman and Mersa Matrouh. The periods of data 570 availability are different for each location, with the longest records (30 years) at Alexandria Western Harbour and the shortest records (4 years) at Mersa Matrouh.
In June 2018, in collaboration with the JRC, the National Institute of Oceanographic and Fisheries deployed an IDSL station in the port of Alexandria. Stations of Egyptian tide gauge network are listed in Table A25.

Algerian tide gauges network 575
The Being aware of the significance in providing the national territory with a precise altimetric reference, the INCT put the effort in the installation of automatic tide gauges along the Algerian coasts, first to bring the bathymetric surveys to a stable reference, hydrographic zero or nautical chart zero, and then to predict the tide or define reference levels. In addition, the INCT is in charge for setting up the hydrographic zero as an altimetric reference for water heights, and uses any tidal data as a national referent (official journal): general knowledge of the tide, determination of harmonic constants and extreme levels and 585 tidal prediction.
The installation of six tide gauge stations with automatic acquisitions along the Algerian coasts (Ghazaouet, Oran, Ténès, Algiers, Jijel and Annaba) in the mid-2010s upgraded substantially the network, in particular for monitoring sea level variations and modernization of the national altimetric reference. Currently, the INCT is additionally upgrading the network through setting up permanent GNSS stations in collocation with the tide gauges stations at the ports of Algiers, Jijel, Oran, 590 Annaba, Ghazaouet and Ténès. This approach would constitute an important phase for the creation of a multi-observation observatory for spatial measurements, gravity field, levelling and tide gauge. Upgrade of observatories to have real-time data acquisition is also planned for near future. Stations of Algerian tide gauge network are listed in Table A26.

Summary of the existing coastal sea level monitoring infrastructure
With the aim of facilitating a more general assessment of the coastal sea level monitoring networks individually described in 595 Section 2.1, a brief questionnaire survey was conducted, resulting with information on: i) institution, country, contact name of a network; ii) number of stations; iii) main purpose of the network; iv) funding mechanism; v) data policy; vi) raw time sampling interval; vii) data latency; viii) number of sea level sensors at each station; ix) type of tide gauge; x) levelling strategy; xi) number of tide gauges collocated with a GNSS; xiii) ancillary measurements; xiv) quality control and processing and xv) public data availability. The survey responses are tabulated and may be found in Supplementary material. 600 Based on the responses received, 236 active stations have been identified so far in the M/BS. These are operated by 30 agencies, representing all countries in the region except Morocco, Tunisia (both have sea level networks, e.g. Jabnoun and Harzallah, 2020) and Libya. We could also not get information from the Italian Hydrographic Office, that operates Genoa tide gauge, one of the longest-operating Mediterranean tide gauges. In some countries tide gauges are operated by several institutions (e.g. 6 in Spain, 5 in Croatia), while others have one national network managed by a single agency (e.g. SHOM, 605 in France). ISPRA in Italy runs the largest network in terms of number of stations (62)   answer, depending on technology and time sampling and latency (e.g. the 2 radar sensors installed for storm surge/tsunami warning). Despite more than half of the answers claim open and free access to data (altogether 10 agencies, and 7 agencies through specific data portals), there are also some data policy issues (Fig. 2b): open and free only for research were selected by 6 respondents (+1 for 1-h qc data), while 5 institutions mentioned unspecified issues or restrictions on data access.
With respect to network sustainability related to maintenance strategy and status, most of the respondents confirm no 625 problems of funding now or in the near-future. Problems with maintenance are reported by NIOF in Egypt, and for a couple of stations in Croatia (IOF) and Israel (IOLR). Most serious issues are raised by the Albanian network, which is not being maintained at this moment and there are no plans for funding in a short term. JRC also warns that the network funding is guaranteed now, but it might stop anytime. All the agencies rely on their own resources for in situ maintenance except PdE (Spain), ISPRA (Italy) and Croatian Waters (Croatia), that subcontract this work. 630 Raw time sampling interval is 1 min for most of the networks in the region (12 of the respondents). However, a large range of raw sampling options are provided by the rest of contributors, from 6.7 Hz resampled to 5 s (JRC tsunami stations) to 55 min (Ruder Boskovic Institute), 1 h (Croatian Waters) or 2 h (Meteolestartit). Some agencies are more specific in the answer to this question and report several samplings available depending on data portal or application. Latency of data transmission is claimed to be real time (<=1 min) by 9 of the respondents (IOLR only for the two radar sensors), near-real time 635 (minutes, hours or days) by 15 of them and delayed mode access only by 7 agencies.
One of the most challenging aspects of a tide gauge network is datum stability and link to other official references, especially for long-term and mean sea level studies. According to the responses, the majority of networks perform highprecision connection of the tide gauge benchmark (TGBM) to the national geodetic datum, as the basic levelling strategy. In some cases, they provide the date of last levelling campaigns, but in general there is a lack of information about their frequency, 640 apart from periodic levelling near the TGBM during routine maintenance reported by 4 institutions (usually once per year).

Connection of all stations to a nearby GNSS station is reported by the Spanish IGN and by the Algerian network.
With respect to quality-control and data processing, the response is also diverse, as shown in Fig. 3. In most cases only delayed-mode quality-control is performed by data originators (17 respondents, 6 of which also generate sea level products). In addition to delayed-mode quality-control and products, automatic quality-control in near-real time is also 645 performed nowadays by 7 agencies in the area, while only 2 agencies do not perform quality-control or processing at all.

Data availability in existing international programs
In this chapter, we assess the M/BS sea level data availability. The assessment is primarily based on the public global and European data repositories as of 15 November 2021, which have been established for various applications (from real-time applications to research-quality products) and contain data of different time sampling step (from a minute to a month).

Monthly sea level data repository
The oldest sea level data repository, at which monthly sea level data have been collected for almost a century, is the Permanent Service for Mean Sea Level (PSMSL, www.psmsl.org). The PSMSL was established in 1933 by Joseph Proudman, with an aim of collecting, analysing and interpreting monthly sea level data at the global scale (Woodworth and Player, 1993;Holgate 700 et al., 2013). The collected data is of high quality and freely available, thus providing an excellent dataset for mean sea level studies in this era of climate changes (e.g. Spada and Galassi, 2012). In total, there are 158 M/BS data series with a high reliability (classified as revised local reference (RLR) data) (Fig. 8a), with six records spanning over more than 100 years. length, respectively, and data coverage of 94%, 87% and 99%, respectively). Due to their exceptional length, these series have been thoroughly checked (e.g. Tsimplis and Spencer, 1997;Wöppelmann et al., 2014) and used in numerous long-term sea level assessment studies (e.g. Letetrel et al., 2010;Pashova, 2012).
In the PSMSL repository, there are 38 tide gauge stations with monthly records longer than 50 years, concentrated mostly along the northern coastlines of the M/BS (Fig. 8a). Along the North African coast there are long-term (50+) data from 710 only two stations, Alexandria and Ceuta, making proper quantification of sea level changes along the southern Mediterranean challenging (e.g. Gomis et al., 2012). At a number of sites (24 of them) there are two or more tide gauges collocated, with the long-term data coming from the float-type tide gauges, and with additional sea level series normally coming from the digital instruments (radar, acoustic or pressure tide gauges) and spanning over the last few decades at best. The median data coverage of the M/BS PSMSL series is 92%, while 63 stations have the data coverage higher than 95%. There are 19 sea level records 715 which have 100% data coverage, but all of them are relatively short records (up to 30 years)these series mostly come from digital instruments. Oppositely, there are 4 sea level records with data coverage less than 50%, with 3 of them containing data from the 1960s and 1970s.

Hourly sea level data repositories
At hourly resolutions, some of sea level records coming from the PSMSL tide gauges are also available in the Copernicus 720 Marine Service (CMEMS) In Situ Thematic Assembly Centre (http://www.marineinsitu.eu). In general, the CMEMS is a service providing operational data-driven ocean products, plus systematic information on the ocean and sea-ice state (Le Traon et al., 2019). The CMEMS database contains hourly sea level records from 115 tide gauges in the M/BS (Fig. 8b), with the median length of the series of 2 years. Hourly sea level data records can also be found in the EMODnet Physics in situ observations repository (Novellino et al., 2015, https://www.emodnet-physics.eu/map), which is developed to be a single point 725 of access to near real time and historical ocean data in Europe. The repository contains hourly sea level observations from 119 tide gauges in the these two basins (Fig. 8b). For some stations, the data is available both from recent observations and from historical periods, sometimes of the same length or even longer than the PSMSL records, but mostly covering shorter periods In addition to the quoted centralized data repositories, some of hourly sea level data can also be accessed through SeaDataNet portal (https://www.seadatanet.org), which is a virtual centre that provides different ocean metadata, data and 740 products archived by data providers. However, the access to the sea level archives is not straight-forward. It is mostly distributed towards data providers, sea level data is often combined with data from other observing platforms (e.g. echosounders, synthetic aperture radars)data is also provided in different formats and split in small observing intervalsmaking this portal rather complicated to use. Still, some data not listed in the EMODnet Physics, CMEMS and GESLA dataset may be found there, like sea level records for the Georgian, Ukrainian and Maltese stations. 745 It is a challenge to a researcher to choose the best database out of the three and to locate research-quality data. In various databases, data originating from the same location frequently have different names, identification numbers and metadata. We thus encourage founders of these data repositories to join their data together, and provide one high-quality dataset for further research, with unique data policies allowing for accessibility of the data and following the FAIR (Findable, Accessible, Interoperable and Reusable) principles (Wilkinson et al., 2016). 750

Another important sea level data portal is the UNESCO Intergovernmental Oceanographic Commission Sea Level Station
Monitoring Facility (IOC SLSMF, http://www.ioc-sealevelmonitoring.org), whichcontrary to other sea level data repositorieshas not been developed for collection of the research quality data, but for operational purposes (Aarup et al., 2019). The IOC SLSMF has been developed to deliver the information about the status of tide gauges operating in real time, 755 as well as to visualise the data downstreaming to the service (Flanders Marine Institute (VLIZ), 2021). Further, the service provides high-resolution (mostly with a minute resolution) sea level data at global level, following demands of the operational and early warning systems that emerged after the 2004 Indian Ocean tsunami.
As of 15th of November 2021, the IOC SLSMF provides information and data from 143 tide gauge stations in the Mediterranean and Black Sea, most of which (all but 30 for the preceding week) are operational in real-time with data 760 transmitting latency between 1 and 10 min (Fig. 8c). Ten stations have been operational since 2008, when they were installed

Data-providers' sea level data repositories
Aside from the listed global sea level data repositories and research products, these sea level data is also accessible 785 through websites of data providers (e.g. PdE, SOCIB, SHOM, ISPRA; see more in Section 2). Further, there are sea level data that are not included in listed repositories, like ISPRA Rete Mareografica Laguna di Venezia, containing sea level records from 24 tide gauge stations, but are available through local repositories (https://www.venezia.isprambiente.it). Last but not least, there are more sea level records, in particular at hourly timescale and the raw data, not easily reachable by users. Some of these tide gauge operators provide the access only to recent sea level data, some are visualised through a graphical interface, 790 while others are available on demandoften only for research purposes, and other have even more restrictive data policies.

Assessment of the network to fulfil targeted applications
Based on the information compiled in Sections 2 and 3, an overview of the fit-for-purpose status of the network, to fulfil some 810 specific applications, is presented below. Examples of use of tide gauges in the M/BS are provided, focusing on early warning systems, long-term sea level variability or altimetry calibration, all in order to show the main gaps or lack of information in terms of coastal in-situ measurements.

Tide tables and port operations 815
Historically, the first requests for sea level information came from the needs of navigation near the coasts and access to ports with significant tidal ranges. Tide scales or other visual systems quickly gave way to tide gauges, which become a standard in port operations, in particular at these ports with intense traffic and large tidal range (Lemon, 2003). Although the traffic in the M/BS is substantial, the tides are small and ranging normally to a few tens of centimetres, except in the northern Adriatic and Bay of Gabes (Tsimplis et al., 1995). Nonetheless, several countries produce tide tables to predict the astronomical tide. 820 The tides are normally predicted using harmonic analysis of measured signal. Sea level observations allow for estimation of the harmonic constituents necessary to recompose the tidal signal and provide highly precise estimates of the tidal signal. In the Tide Table, predictions of the times and heights of high and low tides are published for certain points, called reference ports or main ports, for which there are generally long series of observations that allow to obtain an accurate prediction. Each reference port can be linked to one or more secondary ports. For these secondary ports, predictions are 825 obtained by applying time and distance offsets from the reference port.
According to the International Hydrographic Organization (IHO) recommendations, the reference level of the tide/water height observations and predictions for navigators should be the same as the chart datum.
Use of tide gauges for computing tidal harmonic constants and tide predictions, usually done by national hydrographic services, has allowed port authorities along the world to programme their operations well in advance. This is particularly 830 important for those ports that receive large cargo ships with a large draft, for which the time of high or low water is critical. In the M/BS, where the tide is often very small, even negligible, tide tables are only used on those areas with higher tidal ranges. This is the case of the northern Adriatic Sea, where port authorities and pilots in Croatia, Slovenia and Italy use Tide Tables with this purpose. Tide predictions are not used at some other coastlines with smaller tides: as an example, the Spanish Hydrographic Office does not generate Tide Tables for Spanish Mediterranean ports. In these cases, shorter-term sea level 835 forecasts based on storm surge models are the main prediction tool available for harbour operators, as it is the case of Barcelona harbour, that relies on the forecasting system developed by PdE and described in Section 4.1.2. Of course, more precise information can simultaneously be derived from sea level observations transmitted in real time by the tide gauges.
As ships are getting larger and larger, they have more and more drafts, leaving little room for manoeuvre for navigation. The precise knowledge of the water level in real time allows the port authorities to optimize the movement of ships. PdE in Spain has developed a harbour visualization tool essential for managing port operations, which integrates real time tide gauge data with other measurements (e.g. wind) and data from models. Sea level is a key component of a local early warning system for each harbour, which issues alert messages whenever sea level or high-frequency sea level oscillations over predetermined thresholds occur. These thresholds can be configured by harbour operators. In Croatia, the Hydrographic Institute also developed a web application with real time and predicted data, so port authorities and pilots can visually see the 845 difference and adjust their operations accordingly.
Real time tide gauge data are also used by the ports during local bathymetric surveys and dredging activities.
The data collected by the sensors all around the world are used by the JRC Sea Level Database (JRC-SLD, https://webcritech.jrc.ec.europa.eu/) to compute the harmonic coefficient needed to forecast the tide in locations of the IDSL sensors and the surrounding areas. Most of the data collected by the JRC-SLD are analysed with the same software used by 850 the IDSL network to assess the behaviour of the sea level rise and decrease. In case of anomaly, this detection disseminates alerting information.

Storm surges and coastal flooding
Apart from high-frequency sea level disturbances (discussed separately in Section 4.1.3), the Mediterranean coastal floods typically occur due to constructive superposition of tides and meteorologically induced storm surges. High local precipitation, 855 increased river discharge and waves may further worsen coastal flood risks and lead to compound flooding episodes (Bevacqua et al., 2019). All these events pose a threat to cultural heritage, and densely populated coastal communities of the M/BS. Storm surges are most severe along Tunisian, Aegean and Adriatic coasts, at which the highest number of sea level extreme events per year is observed (Cid et al., 2016). Especially critical is the area of the northern Adriatic Sea and Venice Lagoon, exposed to the well-known acqua alta phenomenon. To face these hazards, coastal regions rely on forecasting 860 operational systems , measurement networks and extreme events research activities (Calafat et al., 2014). In the Venice Lagoon, investment in coastal protection through planning and building of flood barriers has been essential.
These challenges are purely operational: short-term day-to-day flood risk needs to be continuously estimated on synoptic timescales using observation-driven numerical forecasting models and other types of early warning systems (Žust et 865 al., 2021;Makris et al., 2021;Ferrarin et al., 2020;Bajo et al., 2019;Mel and Lionello, 2014;Ferrarin et al., 2013;Pérez Gómez et al., 2012. Ensemble modelling allows dealing with uncertainties by generating probabilistic envelopes of possible sea levels (Bernier and Thompson, 2015;Bertotti et al., 2011. Of course, different modelling, ensemble and observation-ingestion paradigms, may be used (Calafat et al. 2014). Panoramically speaking, modelling of sea levels can be classified into numerical physical models on one hand and deep learning approaches on the other. Advantages of 870 the former are typically more extensive spatial coverages -which come at a higher computational cost -and better performance in the extreme tails of sea level distributions. Deep learning systems on the other hand demonstrate rapid progress in their forecasting capabilities at very low computational cost (once trained) (e.g. Žust et al., 2021). Operational forecasting sea level models, producing forecasts on the timescales of hours to days, typically cannot resolve low-frequency sea level variability on the timescales of days to weeks (shown to significantly precondition coastal 875 flooding in some parts of the Mediterranean; e.g. Pasarić et al., 2000;Pasarić and Orlić, 2001;Ferrarin et al., 2021) and are further constrained by the limited sea state knowledge at the time of the simulation. Therefore, sea level observations must somehow be introduced into the model during or after runtime. This can be done in several ways. A simple nudging scheme makes use of near-real time tide gauge data from the last 7 days in the Nivmar storm surge forecasting system run by Ports of Spain since 1998 (Álvarez Fanjul et al., 2001). Another reliable approach to do this is tide gauge data assimilation (e.g. using The Adriatic Sea serves as an example where observation-driven ensemble sea level modelling is absolutely imperative for reliable predictions of cyclone-induced wind driven floods in Venice and other coastal towns along the northern Adriatic coast. Its elongated shape leads to well defined seiche periods and resonant amplification of tides (Medvedev et al., 2020). Its bathymetry on the other hand leads to topographic amplification of sea level signal on the northern Adriatic shelf.
Total sea level signal in the northern Adriatic therefore critically depends on the mutual reinforcement between storm surge, 890 tides and seiches, which in turn depends on the temporal phase difference between peak storm surge, peak tide and peak seiche (see e.g. Cavaleri et al., 2010). High sensitivity to the phase difference between these components is the reason that even minor errors in predicting the storm timing or trajectory may lead to substantial errors in the total sea level forecast (Cavaleri et al., 2020), which makes ensemble modelling a clear advantage.
Focusing on short term forecasting of the Venice Lagoon sea level, ISPRA developed and manages an integrated 895 system, made up of in situ data coming from the RMN and RMLV networks (see Section 2.1.8) and numerical and statistical models. The sea level forecasting system is mainly based on the deterministic hydrodynamic finite elements numerical model SHYFEM, and it provides sea level forecasts up to 96 h depending on the spatial resolution (40 km in the Mediterranean Sea, 2 km in the Adriatic Sea, 100 m in the Venice Lagoon). It uses ECMWF and BOLAM meteorological fields, as input data, and it assimilates the sea level measured by the 36 RMN tide gauges. This integration and the improvement of both in situ 900 observations and modelling system shows a virtuous example of efficiency and functionality to prevent and mitigate the impact of flooding and meteo-marine extreme events on the Italian coastal environment.
Other operational storm surge forecasting systems have been in place in the Mediterranean Sea for the last two decades . These systems are progressively being improved or combined with existing 3D baroclinic models, new higher resolution models, and ensemble and multi-model statistical techniques that provide sea level forecasts 905 with a confidence interval (probabilistic forecast). In the Western MS, Ports of Spain runs a multi-model storm surge forecast A multi-model ensemble forecasting system has also been recently developed for the Adriatic Sea combining 10 models predicting sea level height (either storm surge or total water level) and 9 predicting waves characteristics . 915 On longer timescales, mid-to long-term coastal management and spatial planning indicates a growing need to Several long-term studies (Androulidakis et al., 2015) indicate that different dynamic contributions to the global mean sea level will have to some extent compensat each other, but it can nevertheless be claimed with very high confidence that global mean sea level rise will lead to a substantial overall risk increase in coastal extreme events (Oppenheimer et al., 2019).
Consequently, a well-functioning sea level observation network will be more and more imperative for synoptic coastal flood 925 forecasting and mitigation in the M/BS.

Tsunamis and meteotsunamis
A carefully planned network of real-time accessible tide gauge stations is a must for efficient research, monitoring and issuing of tsunami and meteotsunami early warnings. The meteotsunami network should, in addition, be supplemented with air pressure and wind sensors. 930 Tsunamis, as earthquakes, cannot be predicted. Once they are triggered, mostly by submarine earthquakes, they propagate over thousands of km in the ocean and will reach the coastline in a matter of hours, or even of minutes for those coastal areas which are closer to the tsunami source. As fast detection is essential, tsunami warning systems must rely on realtime seismological networks (real-time information about the earthquake), and real-time information of sea level height oscillations. The latter are provided by shore-based tide gauges and by offshore buoys with bottom pressure sensors 935 (tsunameters, e.g. DART buoys). First alert messages are issued from seismic information. However, assessing the tsunamigenic potential of an earthquake is not easy, so sea level measurements are needed to confirm that a tsunami was generated and to reduce the number of false alarms. Tsunami propagation models are used to forecast the time and amplitude of the wave on arrival at different coastal points, and can also be validated with sea level observations. Adequate communications infrastructure allows issuing correct and timely warnings to local emergency management officials, who can 940 decide to activate their emergency protocols to evacuate low-lying coastal areas in advance of the initial tsunami wave.
When the tsunami of 2004 hit the Indian Ocean causing one of the most devastating disasters of our recent history, only the Pacific Ocean had a tsunami warning system in place. Considering how many lives could have been saved, IOC/UNESCO established several intergovernmental working groups for implementation of regional and national tsunami warning systems in other basins, such as the North-Eastern Atlantic, the Mediterranean and Connected Seas (NEAMTWS) 945 Tsunami Warning System (UNESCO/IOC, 2012a). In 2005, most of the tide gauges in NEAMTWS region were not suitable for tsunami warning. Requirements for this application are less restrictive in terms of accuracy or datum stability than for longterm sea level trends estimates, and consist mainly of improving timeliness and lowering sampling intervals to 1 min or less for adequate measurement of tsunami wave amplitude and arrival time. A basin-wide distribution of stations is needed, with more stations in those areas closer to tsunamigenic sources, as is the case for a significant part of the Mediterranean coast. 950 Following new NEAMTWS requirements for sea level data exchange, many stations have been upgraded in the M/BS region since 2005, and provide today higher-frequency sea level data in real time to regional and national tsunami warning systems. In addition, the last Global Sea Level Observing System (GLOSS) Implementation Plan (UNESCO/IOC, 2012b) suggests that GLOSS Core stations can be configured to support storm surge and tsunami warning systems. This approach has ensured the multi-purpose character of some of the stations, good for network sustainability, but has raised new challenges on 955 standard quality control and data processing techniques that need to be adapted and the development of automatic tools for tsunami detection (Holgate et al., 2008;Beltrami et al., 2011;Pérez Gómez et al., 2013;UNESCO/IOC, 2020). modelling to monitor the existing sea level network capacity for tsunami detection (Schindelé et al., 2008(Schindelé et al., , 2015. As an example, they found that the tsunami generated by an earthquake North of Algeria in 2003 would not be confirmed by the tide gauge network in less than 70 min, when the travel time to the most affected zones was 30-40 min. The tool, that provides guidance for the implementation of additional tide gauges, demonstrated that adding four tide gauges, two in the Balearic Islands, one in Sardinia, and one in Sicily, would reduce tsunami detection time by more than 20 min for sources along the 970 North Algeria and Tunisia shoreline. They also recommend implementation of two tsunameters offshore at specific points in that area, to reduce the detection delay to less than 15-25 min along Tunisian coast. However, the high installation and Meeting in 2019) and in Malta (beginning of November 2021), how interconnecting sensing devices with an alerting system may warn the population promptly and allow a safe evacuation. Possibly, the exercise will be repeated in Indonesia during 2022. The two events differ because Kos was a near shore event and thus required the activation before the official alert from service provider was issued, while Malta was a distant shore event so that the first activation was triggered by the information 980 coming from the Tsunami Service Providers. The system is based on a network including IDSLs and seismometers as sensing devices and a long-range siren with two alerting panels to warn the population. As soon as both IDSLs confirmed the anomalous behaviour of the sea level, the alerting devices started alerting the population automatically, providing indications about the countermeasures to be taken. In the case of Kos this was the first triggering information. Specific signages deployed in the streets provided indications about the evacuation routes and the assembly points. 985 The system allows several degrees of automation, from completely automatic to manual activations only, depending on the standard procedures adopted by the authorities entitled to protect the population. This system is based on the multipurpose platform developed by the JRC and called RIO (Remote InterOperability), that is a complete software platform that allows easy integration of various instruments and analytical computations and activations. The RIO system was also adopted in another project, a low cost GNSS-based buoy that provides sea level data from the open sea. Such a type of device 990 could be therefore naturally integrated in the same network and would provide the alerting information much in advance compared with devices deployed on the coasts: depending on the installation site, the improvement in reaction time could be consistent.
Traditionally, tide gauge stations have not been located at meteotsunami-prone areas, but rather at locations of interest for other sea level processes (e.g. storm surges and sea level rise), and large ports and harbours where sea level data is of 995 utmost importance for safety of navigation or at the well-inhabited coastal towns. Up to the beginning of the 21th century, at the meteotsunami hot spots the measurements have been done only during specific experiments, aimed purely at better understanding of meteotsunamis and other high-frequency sea level phenomena. Such experiments include, but are not limited to: (1) several field experiments with simultaneous air pressure and sea level measurements done during the years 1989-1992 (Monserrat et al., 1991;Rabinovich and Monserrat, 1996-1998(Monserrat et al., 1998Šepić et al., 2009)  for proper monitoring of destructive events, but more needs to be done for an efficient warning system to be implemented, 1010 given that a warning should be issued at least an hour before the meteotsunami occurs at an endangered location. Vilibić et al. (2016) presented a design of a meteotsunami warning system based on simultaneous numerical modelling of synoptic, mesoscale atmospheric and barotropic sea level conditions, as well as a real-time assessment of atmospheric and ocean data measured a few tens up to a couple of hundreds of kilometres away from the most endangered spots, that is at locations where off-shore meteotsunami generation and growth occur. Denamiel et al. (2019Denamiel et al. ( , 2021 further upgraded the 1015 concept suggesting that stochastic approaches are more reliable than deterministic numerical modellingthe prototype of such system, which gives estimates of the sea level exceedance probability during meteotsunamis, was tested for the Croatian coast of the Adriatic Sea (Denamiel et al., 2019). Similar modelling and data assessment strategies have been suggested and tested for the Balearic Islands area (Renault et al., 2011;Šepić et al., 2016a;Romero et al., 2019), where a continuously upgraded meteotsunami warning system, with both deterministic and probabilistic component, has been operational since 1985 (Jansà 1020 andRamis, 2021).
In spite of the great efforts invested in development of the meteotsunami warning systems, the results are still not satisfactory, resulting in a loss of trust in the early warning systems. The forecasts are known to be wrong, especially when it comes to estimating the strength and destructiveness of the event (Jansà and Ramis, 2021). The crucial problem is an intrinsic inability of the atmospheric modelsrelated, among else, to coarse resolutions and inadequate physical parameterization at 1025 the mesoscaleto reproduce exact properties (spatial outreach and rate of air pressure change, speed, pressure and wind spatial gradient) of the fast changing atmospheric disturbances responsible for the meteotsunami generation (Belušić et al., 2007).
Slight changes of any of these properties (e.g. reducing the rate of air pressure change, translating meteotsunamigenic disturbances off their pathways for a few tens of kilometres, changing the propagation speed) may change the modelled sea level response for several times, in particular at the most endangered areas (Vilibić et al., 2008;Orlić et al., 2010;Orfila et al., 1030Orfila et al., 2011Šepić et al., 2016a;Ličer et al., 2017;Mourre et al., 2021).

Long-term variability and sea level trends
In order to estimate the long-term variability and trends of the sea level for M/BS, the PSMSL RLR dataset has been used as the primary source of monthly means (see Section 3.1). The analysis is made on a geographical extraction on M/BS of the PSMSL relative sea level trends product (https://www.psmsl.org/products/trends/). This product fits a model composed of   in PSMSL since 2017, highlighting the importance of updating records within the PSMSL. Although they do not appear yet in the PSMSL RLR dataset, some long-term records can be found from external sources. This is the case for Alicante station in 1065 Spain that has been recently digitized by Marcos et al. (2021), the Venezia station in Italy that has been updated recently (Zanchettin et al., 2021) and the data from Koper in Slovenia that has been provided by the co-authors of this paper. We have decided to add these stations in our analysis (represented in blue in Fig. 9).
If the last 30 years (1990-2019) are considered, the situation is a bit better with 27 stations available. It is noticeable that no long-term records are available for the south of the Mediterranean Sea (North African coast). These long-term records 1070 are the only means to compute robust estimation of sea level rise.

Evolution of extreme sea levels
Long-term changes, at interannual and longer time scales, in extreme sea levels are primarily driven by changes in mean sea 1090 level (MSL, Woodworth et al., 2019). However, variations in extremes unrelated to MSL variability have also been identified in tide gauge records at hourly scale worldwide (Wahl and Chambers 2015;Marcos and Woodworth, 2017) and linked to changes in storminess. In the Mediterranean Sea, long-term multidecadal fluctuations in sea level extremes have been noticed in long tide gauge records, such as those in Trieste since the 1930s (Raicich, 2003) and other stations in the northern Adriatic (Masina and Lamberti, 2013), and in Marseille since the early 20th century (Letetrel et al., 2010;Marcos et al., 2015). These 1095 variations are regionally coherent as they normally originate from large-scale atmospheric forcing (Calafat et al. 2014;Marcos et al., 2015;Lionello et al 2021). Given the variability at such low frequencies, the estimation of linear trends is highly dependent on the selected period. For example, Raicich (2003) identified a decrease in the frequency of extremes in Trieste tide gauge record during the period 1940-2000, with no clear trend in their intensity. In contrast, Masina and Lamberti (2013) identified a small increase in the magnitude of sea level extremes in the northern Adriatic that was associated with an 1100 intensification of the bora wind in the 1990s. In Venice, Lionello et al (2021) reported that the frequency of flooding resulting from storm surges has increased since the mid-20th century, although they linked this effect to relative sea level rise rather than to a sustained trend in storminess. At interannual time scales changes in extreme sea levels are correlated with the North Atlantic Oscillation (Marcos et al., 2009b;Masina and Lamberti, 2013), even after the removal of the yearly MSL signal.

Satellite altimetry calibration 1105
Satellite altimetry is nowadays an inevitable tool for mapping sea level changes over the ocean basins (Ablain et al., 2017).
However, their measurements are biased by a number of processes (Fu and Haines, 2013) and therefore a proper calibration of these data is the prerequisite (Andersen and Cheng, 2013;, in particular when approaching the coastal zone (Vignudelli et al., 2019). The tide gauge data is normally used for satellite altimetry calibration, yet the problem is that the altimetry measurements are not reliable close to land, while large differences in mean sea level may occur at the 1110 coastal distances (e.g. in the last 4-5 km to the coast, as observed for the Senetosa calibration site at Corsica for both TOPEX/Poseidon and Jason altimetry missions, Gouzenes et al., 2020). Indeed, a careful examination should be conducted to verify nonexistence of all potential errors that could explain the increased rate of sea level rise close to the coast -like spurious https://doi.org/10.5194/os-2021-125 Preprint. Discussion started: 20 January 2022 c Author(s) 2022. CC BY 4.0 License. trends in the geophysical corrections, imperfect inter-mission bias estimate, decrease of valid data close to the coast and errors in waveform retracking -before ascribing a finding to the real physical. Here, for the Sanatosa calibration site, it has been 1115 proven that the steric sea level component is responsible for such changes (Dieng et al., 2021).
These problems are even more amplified in enclosed seas, such as the M/BS, where the impact of a basin topography and a complex coastline is more pronounced than at the oceans. Orbit-related sea level errors are an example, as found to be prominent in the Mediterranean (Esselborn et al., 2020). For that reason, calibration of satellite altimeters has a long history there. This particularly applies to some selected tide gauge sites in the Mediterranean locations, like Ibiza (Martinez-Benjamin 1120et al., 2004Frappart et al., 2015) or Corsica (Bonnefond et al., 2003(Bonnefond et al., , 2021Cancet et al., 2013). Recently, advanced learning methods (such as deep learning networks) have been used to improve calibrations of altimeter data (Yang et al., 2021).

Definition of vertical frames of reference
The precise information on vertical references, which strongly relies on tide gauge measurements, is also a major issue, for both ocean and land charts. Vertical reference surfaces can be categorized under three general headings: 1125 • tidal datum, also called chart datum which should, according to the IHO recommendations, correspond to the lowest astronomical sea level (LAT) or an equivalent reference level considered by the hydrographic services as being as close as possible to the LAT; • vertical reference datum which is a surface of zero elevation to which heights of various points at land are referred, being a base for defining height systems (e.g. European Vertical Reference System, EVRF2019) 1130 • ellipsoidal reference datum, which allows to define the ellipsoidal height important for satellite altimetry measurements and GNSS receivers at tide gauges (e.g. Adebisi et al., 2021), such as Global Reference System 1980 (GRS80) or World Geodetic System 1984 (WGS84).
At tide gauges, the referencing of one with respect to another is essential to allow tidal observations to serve all possible applications (e.g. the analysis of the tidal observations allows to define the characteristics of the tide at the tide gauges 1135 (LAT, MSL, HAT...) and consequently to determine the tidal/chart datum.
Traditionally, bathymetric data has been collected and stored relative to a tidal datum and topographic data relative to a geodetic datum. Close to a tide gauge, bathymetric data can be referenced to the chart datum by subtracting the observations of the tide gauge directly or associated with models. One of the most significant challenges in traditional hydrography is establishing the relationship between the instantaneous water surface and chart datum away from water level gauge locations. 1140 In this case, to obtain the chart depths, the vertical positions of the bottom are referred to the so-called Vertical Reference Surface for Hydrography (VRSH), which, following the recommendations of the IHO, is identified with the development of a 3D separation model between Chart Datum (LAT) and the geodetic datum (geoid and/or respect to the ellipsoid).
In modern times, the hydrographic surveying community is using high-accuracy Global Navigation Satellite System In Spain, the orthometric and ellipsoidal heights of the LAT have been established for the Iberian Peninsula and Canary Islands domains from model-reanalysis sea level data fields, which were first validated and then adjusted to 1150 experimental data from 119 tide gauge stations. The sea level height with respect to the geoid was obtained from the modellingreanalysis service provided by the Iberia-Biscay-Ireland Monitoring and Forecasting Center (IBI MFC), in the framework of the EU Copernicus Marine Environment Monitoring System (Sotillo et al., 2015). Away from water level gauge locations the model will be adjusted in the future by GNSS water level buoys, to establish chart datum at offshore locations.

Summary and conclusions 1155
An overview of existing coastal sea level infrastructure in the M/BS has been presented, based on the contribution from 30 institutions/sea level scientists operating tide gauges in the region. These stations are essential to monitor and study sea level variations that pose a hazard in these basins, such as storm surges, tsunamis, meteotsunamis and sea level rise, by providing accurate sea level data at all frequency ranges along the coastline. The initiative gives an insight into the status of the in situ sea level network in both basins and confirms several challenging aspects such as the diversity of national strategies, sea level 1160 technology, funding or data availability, often linked to differences on the primary and evolving objectives of these installations, and their unbalanced spatial distribution. National contacts and relevant basic metadata are provided, as a starting point for improving coordination across the region. In most countries tide gauges are operated by several agencies, usually targeting different purposes. Only 7 of these institutions are in charge of more than 15 stations, while the majority operate a smaller number of stations, sometimes one single station. 1165 We have identified 236 active stations covering nearly all the country's coastlines in the M/BS. Several stations in Morocco and Tunisia could not be added to this inventory, and no information was obtained from Libya. confirming the lack of information along the southern Mediterranean coast from previous initiatives.
There are still many float gauges as the ones installed since the end of the XIXth century for tides and hydrography applications, still being the second most important application, especially along the northern and eastern Adriatic and the Greek 1170 coasts. However, tides in the M/BS are generally small, so tide predictions are not computed nor needed for port operations everywhere. Therefore, an increasing number of stations are becoming multi-purpose, a term we apply here to tide gauges upgraded to be used as well in tsunami, storm surge and other early warning systems. A significant part of the network is based on radar sensors (134, 57% of the stations) providing 1 min time sampling data, with real or near-real time data transmission. to regional or national tsunami warning systems in this region. In addition, lower time samplings have allowed, as for the global network, an improved dataset for understanding and warning meteotsunami events, more frequent than tsunamis at 1180 several spots in the M/BS. While 1 min sampling is sufficient for detection of most of these sea level oscillations, access to even higher frequency raw data from modern sensors would be desirable in the future, for a better characterization of all periods above 30 s. As an example, PdE in Spain has recently developed a tool for operational characterization of 2 Hz raw data from the REDMAR network, that provides these data and derived data products including waves through PdE OpenDap server. In addition, at least 4 more institutions participating in this survey compute wind wave parameters from tide gauge raw 1185 data.
In terms of length of the sea level records, the numbers decrease significantly: only 10 stations have been identified with enough valid data covering the last 100 years (7 with data in the PSMSL), while 27 would have data for the last 30 years (the altimetry period). This reflects the limitations of the network to provide reliable sea level trends and their spatial variability along the coastline. Apart from the smaller number of tide gauges in the past, this is perhaps the most challenging application, 1190 as it requires a precise knowledge of the station history, data archaeology efforts and often the careful combination of data from different technologies/locations inside a harbour. Fortunately, access to VLM information has also improved in recent years, with up to 46 stations now collocated with a permanent GNSS in the M/BS. Some of these stations are contributing, with different time sampling, latency and quality control to one or more of the seven international data integrators or portals described in Section 3, where a detailed assessment of data accessibility in 1195 the M/BS is presented. The most populated portal, PSMSL, contains data from 158 stations in the region, which means that there is still a significant number of M/BS stations focused on local or national services and not included in international programs. This is a problem for basin scale applications and research studies and it is often related to data policy issues: not all tide gauges provide open and free data to users, and some claim data availability only for research applications. It must be emphasized that most of national operators rely on their own personnel and resources, including in situ maintenance as well 1200 as quality control, data processing and product generation, and international programs have traditionally relied on these national efforts. This requires enough funding of the networks from the Member States, not always guaranteed, and it could also explain the lack of access to data, or the delays in updating the time series with recent records.
On the contrary, in other cases, the same sea level time series can be found at different repositories, as well as at national data portals, with different name, metadata, or quality control, which may be confusing for end-users. These problems 1205 are not exclusive to the M/BS, and are partly linked to the lack of unique identifiers and adequate and standard metadata information for tide gauges. Data integrators should collaborate between them and work more closely with original data providers, to ensure interoperability and homogeneous and good quality datasets, according to FAIR principles. These coast. Apart from float, acoustic, pressure and radar sensors, a novel technique based on GNSS Interferometric Reflectometry (GNSS-IR) has recently emerged and revealed its potential as the number of satellite constellations increased (Peng et al., 2021). GNSS receivers have the advantage of providing both sea level and land motion information. Some institutions in the 1215 M/BS (e.g. Spanish Geographic Institute) are already exploring this technique. In the framework of the EuroSea project and the EuroGOOS Tide Gauge Task Team activities, the UK National Oceanographic Center (NOC) has recently developed a global GNSS-IR data portal, hosted at the PSMSL (https://eurosea.eu/new/a-global-sea-level-data-portal-using-globalnavigation-satellite-system-interferometric-reflectometry/). The use of GNSS receivers for sea level is becoming a reality and it is going even further. For many applications, in 1220 situ sea level measurements offshore would be a significant improvement of the coastal sea level network. One example is tsunami warning, where detection of the wave before reaching the coast is a clear advantage. GNSS receivers on buoys (GNSSbuoys) have been used for years in Japan's tsunami warning system. Despite this inventory is not including this type of stations in the M/BS, we know that several countries and agencies are now planning their implementation by adding GNSS receivers to existing or new buoys. These data would be also valuable for calibration of coastal altimetry or, as described in Section 2, 1225 for determination of offshore chart datum.
This survey has also revealed that at least 64% of active stations in the M/BS (150) have some kind of ancillary sensor providing meteorological and/or oceanographic data. Atmospheric pressure and wind are the most traditional and frequent additional parameters, very often with time samplings of 1 min, useful for meteotsunami studies. But a number of stations in the Appendix are in fact multi-sensor platforms providing a large range of parameters, as do meteorological stations (humidity, air 1230 temperature, precipitation, etc) and even ocean data like water temperature, salinity or currents. Several agencies have already or plan to install cameras (e.g. webcams planned by SHOM, in France). Multi-sensor platforms deployment seems a reasonable approach for ensuring sustainability of the networks by expanding even more their range of applications.
In summary, the assessment of the coastal sea level monitoring capacities in the M/BS exposed several important issues that are a prerequisite for making sea level operations and science over a variety of timescales and applications in these enclosed 1235 basins comprehensive: (1) a longevity of calibrated measurements is threatened at some monitoring sites, which may lower the confidence of the sea level rise estimates in the era of climate changethis should be immediately bypassed by putting again in operations all inoperable tide gauges that have multi-decadal time series; (2) the gaps in monitoring systems or their inadequacy that exist in some coastlines (e.g. Libya, Albania) should be bridged, by upgrading the existing stations or installing new ones; (3) the quoted activities should be done through collaboration and knowledge-transfer from more experienced tide 1240 gauge networks, in particular towards North Africa countries, preferably within the umbrella of existing international programmes (IOC, MONGOOS), agencies (e.g. through extending the IDSL network of JRC) or joint projects; (4) the data should be available for research and follow the open science policies, in particular of the FAIR principles (Wilkinson et al., 2016); and (5) the cacophony of sea level data repositories should be minimized, with the clear and unique provision of data for real-time and research purposes through one-stop shop service, including harmonized quality-check procedures. We hope 1245 that the next decade of coastal sea level monitoring will be as dynamic as the last decade, in which substantial progress in