The sea level time series of Trieste, Molo Sartorio, Italy (1869–2021)

. The sea level observations carried out at Trieste, Molo Sartorio, from 1869 to 2021 have been revised and updated. Information on the tide gauges and on the geodetic benchmarks on Molo Sartorio during that period have been collected. Basic quality checks have been applied. The hourly data for the 1917–1938 period, digitized from the original charts, have allowed us to build a time series of hourly sea level heights from 1905 to 2021. Gaps of up to 24 h have been ﬁlled by interpolation. The errors affecting the monthly and annual mean sea levels have been estimated. The availability of monthly and annual means prior to 1904 allowed us to build a mean sea level time series spanning 153 years, characterized by linear trends of an observed sea level of 1.36 ± 0.17 mm yr − 1 and of an inverse-barometer-corrected sea level of 1.45 ± 0.13 mm yr − 1 . A signiﬁcant acceleration of 0.008 ± 0.004 mm yr − 2 was estimated from the inverse-barometer-corrected sea level time series. This data set represents the most up-to-date data set of sea level observations and ancillary information relative to the tide-gauge station of Trieste, Molo Sartorio. The data are available through SEANOE (https://doi.org/10.17882/62758, Raicich, 2022).


Introduction
Long time series of environmental observations are fundamental in climate studies.Among them, it is widely recognized that historical sea level records play a key role in the assessment of long-term sea level rise rate and acceleration.Some records date back to the 18th century (e.g.Woodworth, 1999;Wöppelmann et al., 2006;Raicich, 2015), and automatic recordings started in the first decades of the 19th century (Matthäus, 1972).Nevertheless, the number of homogeneous sea level records longer than a century is quite limited, and, moreover, they are unevenly distributed geographically.Therefore, there is an increasing demand for the rescue of historical sea level observations.
This work aimed at recovering and making available the sea level data obtained at Trieste, Molo Sartorio, the information on the instruments used for the observations, and the levelling data of the benchmarks inside and outside the tidegauge hut.The existing data and metadata have been thoroughly revised and previously unpublished data have been added to the existing time series.
Section 2 describes the historical evolution of the station and the elements relevant to sea level observation.The observations are discussed in Sect.3. In Sect. 4 the long-term time series obtained from the revised data set is described, and some basic properties are discussed.Data availability is summarized in Sect. 5. Concluding remarks are presented in Sect.6.

Installations and instruments
The earliest known measurements of sea level height in Trieste were made by the physician Leonardo Vordoni from 1782 to 1794, who was interested in studying the connection of tides and the course of diseases.Although the data set is remarkably long for the period, it mainly has a historical value due to its low quality, as discussed in Raicich (2020).On 23 July 1840, Vincenzo Gallo, the director of the local meteorological observatory, began measuring the sea level height in order to compute the establishment of the port and to make tidal predictions (Gallo, 1840); observations for 21-22 March 1844 only could be recovered (Gallo, 1844).
The systematic observation of sea level started on 16 October 1859, when the first self-recording float tide gauge became operational (Table 1).It was provided with a stilling well opened in the floor of a room in the north-western corner of the Finance Guard building, at the end of Molo Sartorio (Schaub, 1860).Figure 2 displays the tide gauge, seen from the front and from above, and a vertical section of the stilling well, which was connected to the open sea by a siphon (von Chiolich-Löwensberg, 1865, 1866).Around 1860 a vertical tube with a hydrometric scale was fixed to the north-eastern side of the pier to carry out direct sea level measurements for calibration (MGI, 1897).
That setting remained unchanged until 29 November 1924, when the observations were interrupted because the building that hosted the tide gauge was completely restructured.Unfortunately, no temporary tide gauge was put in operation in the meantime, and the observations were resumed on 30 June 1926 in a new tide-gauge hut, built on the same pier approximately 30 m to the east of the previous installation (Fig. 3).The station was provided with a stilling well which communicated with the sea through a 40 cm-long pipe.Since then, the hydrometric tube could be accessed through the floor of the new tide-gauge hut.It still exists, although is it impossible to say whether it is the original one; however, it is no longer used as the calibration measurements are made in the stilling well.
In 1961 the hut was enlarged and a new stilling well was built (Picotti, 1960); this is the present installation.Figure 4 shows the tide-gauge hut in 2001 and four tide gauges, three of which were operational at that time.
Table 1 summarizes the instruments used since 1869 and their main technical characteristics; all the tide gauges are float instruments.The instruments with chart speeds greater than 10 mm h −1 have a 24 h rotation drum, the others a 7 d drum.Unfortunately, little original documentation exists, and the literature is often confusing; therefore, the   wards.These changes are reflected in the complex situations of tide-gauge zeros and benchmarks that are described below.

The tide-gauge zeros
Until 7 December 1910 at 17:00, the tide-gauge zero corresponded to the top edge of the hydrometric scale, which was also known as "the pier edge", "the zero line of the hydrom-eter", or simply "Molo Sartorio".It coincided with benchmark (Höhenmarke, HM) HM 39 of the Austrian Military-Geographic Institute (k.u. k.Militär-geographisch Institut, MGI).The positive versus of the hydrometric scale was downwards (e.g. von Sterneck, 1905).This zero was named the "zero point of Molo Sartorio" for the first time by Anonymous (ca. 1910), and it has generally been known as "Zero Molo Sartorio" (ZMS) starting from Polli (1938)   From 7 December 1910 at 18:00 to 10 December 1910 at 08:00, a temporary zero was defined at 1.64 m below ZMS; the positive versus remained downwards (Osservatorio Marittimo, 1910).On 10 December 1910 at 19:00 a new zero was adopted at 2.16 m below ZMS, corresponding to the lowest height recorded since an unknown date to 1910, namely on 15 January 1907 at 16:00; it was known as "Zero Hopfner" (ZH) (Hopfner, 1913).Since then the positive versus has been upwards.
The zero was changed again probably in August 1919.It was set to 1.50 m below the mean sea level of 1911, also known as the "Hopfner mean sea level", namely 1.123 m above ZH.The zero introduced in 1919 was named "Zero Istituto Talassografico" (ZIT) for the first time by Ferraro (1972) and is the present zero.It corresponds to 2.537 m below the ZMS.
Figure 5 summarizes the scheme of the tide-gauge zeros and the relationships with the tide-gauge contact point (CP) and benchmark (BM).

The Austrian period
The two earliest surveys were performed by the MGI in 1876 and 1884 (MGI, 1885(MGI, , 1896)).Both involved the vertical benchmark in the tide-gauge room, identified as HM 1.The "absolute" height of HM 1, namely 3.352 m, was defined by adding 1.118 m, which was the mean sea level (MSL) relative to the tide-gauge zero, and 2.234 m, which was the height difference between HM 1 and HM 39 obtained in the levelling survey of 1876.The horizontal distance between HM 1 and HM 39 was 52 m along the levelling line (MGI, 1885(MGI, , 1892)).The MSL was probably relative to 1869.During a survey in 1884, HM 1 − HM 39 = 2.2341 m (MGI, 1896).The height of HM 1 was taken as the base of the levelling networks of the Austrian-Hungarian Empire and of several countries that became independent after its dissolution.

The Italian period
The first survey was performed in 1926, when the Italian Military Geographic Institute of Florence (Istituto Geografico Militare Italiano, IGMI) levelled the new benchmarks installed after the new (present) tide-gauge hut was built.On Molo Sartorio the levelling involved the vertical benchmark (caposaldo verticale, CsV) CsV 53A, the horizontal benchmarks (caposaldo orizzontale, CsO) CsO 53, CsO 53A, and CsO 54 and the tide-gauge CP (CP1926).The heights were referred to the zero of the IGMI defined on the basis of the MSL of 1884-1903 at Genoa (IZ1894).CsO 53 was connected with CsO 52, located 879 m away.Subsequently, the survey involved CsO/CsV 53A, 42 m away from CsO 53, and CsO 54, 26 m away from CsO 53A (IGMI, 1926).The ZMS was the only surviving Austrian benchmark, and its height was connected to the others in 1927 (Spinello, 1927).
The IGMI carried out another survey in 1956 involving CsO 39 and CsO 39 ; the latter was the former CsO 54, which had been renamed.A new zero of the IGMI levelling network was adopted, based on the Genoa MSL of 1937-1946 (IZ1942) (Salvioni, 1957).
After the works on the tide-gauge hut in 1961, a new tide-gauge CP was installed in 1965 (CP1965), known as "Piastrina Mareografica" (PM), which is currently used for the direct calibration measurements.
Other levelling surveys were carried out by IGMI in 1977, involving CsO 39a, CsO 39c, CsV 39, CP1965 (IGMI, 1977), and CsO 39 (Lama and Corsini, 2000), and in 1989, involving CsO 39a, CsO 39c, CsV 39, and CP1965 (Lama and Corsini, 2000).Both CsO 39 and CP1965 are 6 m away from CsO 39 (formerly 54).CsV 39 is just above CP1965.The most recent survey dates back to 2002.It was required because new benchmarks were installed before the existing benchmarks became unusable due to the restoration works of the building near the tide-gauge hut.The levelling involved CsO 39a, CsO 39c, CsO 39 , which is the new name of CsO 39 , and CP1965 (IGMI, 2009).

Merging the benchmark heights
Overall, three national reference frameworks have been used to refer to the benchmark heights during the period of sea level observations, namely Austrian Zero AZ1869 and the two Italian zeros IZ1894 and IZ1942.In order to obtain a homogeneous time series of height data, the relationships behttps://doi.org/10.5194/essd-15-1749-2023 Earth Syst.Sci.Data, 15, 1749-1763, 2023 tween those zeros should be known.Unfortunately, it was not possible to find unambiguous information about them in the literature or in public archives.Therefore, they were estimated as explained in Appendix A. Table 2a summarizes the original and normalized heights of the benchmarks on Molo Sartorio obtained during the national levelling surveys.The normalization was made using Eqs.(A16) and (A17).Moreover, a composite time series of benchmark heights relative to IZ1942 (Table 2b) was obtained by merging those of CsO 54/39 /39 for 1926-2002 with those for 1876-1884 relative to the virtual benchmark VZMS (Virtual Zero Molo Sartorio), defined as VZMS = ZMS − 0.0109 m. (1) Equation ( 1) is based on the height difference between ZMS and CsO 54 measured by the Hydrographic Office of the Water Magistrate of Venice (Ufficio Idrografico del Magistrato alle Acque, UIMA) in 1926 (Spinello, 1927) as well as the composite time series.Figure 6 displays the heights of the benchmarks involved in the composite time series in the respective reference systems.
The heights of 1977 are slightly higher than in 1956 and 2002, probably as an effect of the ground deformation induced by the earthquake of May 1976 in the Friuli region (Talamo et al., 1978), with the epicentre at about 90 km from Trieste and M w = 6.4 (Finetti et al., 1979).The heights of 1989 are anomalous, as they are about 7 cm higher than in both 1977 and 2002 for reasons that could not be discovered; they were not taken into account.
The linear trend of the composite time series of tide-gauge BM heights is 0.07 ± 0.06 mm yr −1 , significant at p = 0.02 and corresponding to a total height variation of +9 ± 7 mm from 1876 to 2002.If the height of 1977 is not included in the analysis, the trend becomes 0.05 ± 0.04 mm yr −1 , significant at p = 0.01 and corresponding to a total height variation of +6 ± 5 mm in 126 years.In this work the errors correspond to 5 % significance.The result is consistent with the known relative stability of Trieste compared to the other coastal areas of the northern Adriatic (Carbognin and Taroni, 1996;Carbognin et al., 2004).

The observations
The data from 1859 to 1904 could only be found in the literature, except for a few original charts covering about 10 d of August 1864 (CNA, 1864).The monthly and annual MSLs for 1875-1904 are also summarized and discussed in Raicich (2007).
From 1905 onwards the data could be retrieved from original documents, namely tabulations of hourly heights for 1905-1911and 1913-1914and charts from 1917onwards. Polli (1938) ) stated that manuscript hourly heights from 1912 to March 1915 did exist and reported the monthly and annual means of 1912; the monthly and annual means of 1915 appeared for the first time in Polli (1970).Moreover, according to the Osservatorio Marittimo (1916Marittimo ( , 1917)), the observations were regularly carried out in 1916 too, but they are missing.Polli (1947) provided annual means for 1890-1904, but the data source was not quoted, and it is unclear whether they came from observations or were estimated.Moreover, the values of 1901-1904 are different from those in von Sterneck (1905), who explicitly reported that he obtained the data from the Maritime Observatory (Osservatorio Marittimo).Therefore, we considered Polli's data suspect and discarded them.

The data since 1905
The complete time series of hourly data from 1905 to 2021 obtained in this work is included in the data set for 1869-2021 available in Raicich (2022).The data for 1905-1914 have been digitized from the original tabulations.Due to the absence of original charts for a direct check, only evident mistakes could be corrected.The hourly heights from 1917 to 1938 have been digitized from the original charts; previously, only high waters (HWs) and low waters (LWs) were available.The hourly data from 1939 onwards, which were already available, have also been thoroughly revised.
In principle, the hourly heights are "instantaneous"; that is, the curves were not filtered before digitization.However, in the case of oscillations of periods shorter than approximately 10-15 min, the persons in charge of data digitization used to smooth the curve graphically and digitize the smoothed values.
Gaps no longer than 24 h were filled by interpolation according to UNESCO/IOC (1994, 2020b), which uses linear interpolation of the residual sea level, obtained after subtraction of the astronomical tide from the observations.In the case of failure of the main tide gauge, the sea level heights were generally taken from the charts of auxiliary instruments characterized by a 7 d rotation drum, such as the R 225 and the Richard (Table 1).This allowed one to digitize the HWs and LWs but made it difficult to extract the hourly values, which, in fact, were not usually reported.In such cases, we obtained the hourly values by means of cubic spline interpolation of the HW and LW data.In order to treat the HWs and LWs as true local extremes, two auxiliary data were in-troduced, 1 min before and 1 min after each extreme, respectively; they are 1 mm lower/higher than the corresponding HW/LW.This procedure ensures that the estimated values do not overshoot/undershoot the observed local extremes.We stress that interpolation aims at obtaining reasonable hourly values for the estimate of daily MSL and the subsequent calculation of monthly and annual means, not at estimating the missing data.
The only major gaps are in 1912 and 1915-1916, due to missing observations as explained above, and from 29 November 1924 to 30 June 1926, when the tide gauge was dismantled and reinstalled in the new hut.The other gaps that occurred for different reasons and that could not be filled are summarized in Appendix B.
From 1905 to 2021, except for the December 1924-June 1926 period in which the tide gauge was not operational, 1 011 736 hourly values are potentially available.The number of those estimated by interpolation is 3531, corresponding to 0.3 %, while 30883, i.e. 3.1 %, are missing.At least one auxiliary tide gauge became available in 1927 (Table 1), and this has allowed one to reduce the missing hourly data to 0.1 % since then.
From the hourly data, daily MSLs were estimated by means of a Doodson X0 filter.Monthly MSLs were computed when at least 50 % of the daily values were available.As a result, the monthly MSL could not be determined for the following months: January-December 1912, January 1915-December 1916, and December 1924-June 1926.Annual (calendar) MSLs were computed with at least 11 monthly means: therefore, they could not be determined for 1912for , 1915for -1916for , 1925for , and 1926for . For 1912for and 1915, we adopted the monthly and annual MSLs available from the literature (Polli, 1938(Polli, , 1970)).

The assessment of errors
It is not easy to associate errors with the hourly heights digitized from charts: therefore, we only aim at estimating representative values.
Because the sea level height is defined on the basis of the vertical distance between the tidal curve and a baseline drawn on the chart, we must take into account the accuracies of the positions of those lines.The baseline of the Fuess-Seibt tide gauge charts was identified by a horizontal line drawn a posteriori 8 mm above the bottom edge of the chart.The Ott-Büsum tide gauge provided both the tidal curve and the baseline.At least before June 1961 the curve of Fuess-Seibt was originally marked by a metal tip on special coated paper.The curve was very thin and often faint: therefore, the persons in charge of data digitization used to trace it with a coloured pencil or ink; that practice might have introduced errors, but it made it possible to distinguish each day in case of overlapping curves, as the paper was generally changed every 2 or 3 d.In any case, the line thickness of each line is about 0.5 mm.As the reduction ratio is 1/10, it is realistic to assohttps://doi.org/10.5194/essd-15-1749-2023 Earth Syst.Sci.Data, 15, 1749-1763, 2023 ciate a 1 cm accuracy with the individual digitized heights.We also recall that, because the charts before 1917 are not available, we could not verify the accuracy of the 1905-1914 data.
The uncertainty associated with the interpolation of gaps using cubic splines was assessed empirically as follows.As HW and LW data are available for 1917-2021, we estimated a time series of hourly heights for that period using cubic splines (as in Sect.3.1).The root-mean-square difference between estimated and observed hourly values is about 5 cm, which was assumed to be the representative uncertainty in the individual hourly height estimated with splines.The uncertainty related to the interpolation based on de-tided residuals is more difficult to assess, because the procedure is intrinsically more complex, as it involves the estimate of the astronomical tide, the subtraction from the observations, and the interpolation on a specific time interval.Because we only aim at representative errors, in this case we also assumed a 5 cm uncertainty in the individual estimated hourly height.The errors on the daily, monthly, and annual MSLs were estimated on the basis of the actual number of interpolated hourly data involved.
As a result, the daily MSL is affected by an error between 0.15 cm (no interpolated hourly data) and 0.77 cm (24 interpolated hourly data).Figure 7 displays the monthly errors and percentages of valid days until 1975; afterwards, there are no missing data.With regard to the 1905-2021 period, the MSL of a month with no missing days is affected by an error of 0.03 cm; the largest monthly error is 0.09 cm, due to missing days and/or daily means estimated from interpolated hourly values (Fig. 7a).The errors in annual MSLs are always 0.01 cm (to centimetre precision).It is reasonable to also adopt these errors for the monthly and annual MSLs of  5).The labels on the abscissa axis in panel (a) refer to the beginning of the year.1901-1904, 1912, and 1915, for which the numbers of valid days are unknown.
The original monthly means of the 1875-1889 period are mean tide levels (MTLs).MSLs were estimated using average monthly corrections obtained by comparing the MTLs and MSLs of 1917-2021 (Table 3), thanks to the availability of HWs and LWs during that period.The typical error in monthly MTLs is around 0.10 cm (between 0.09 and 0.11 cm), depending on data availability (Fig. 7b).The errors in the estimated monthly MSLs range approximately between 0.4 and 0.6 cm and are mostly determined by the error in the corrections (Fig. 7a).The errors in the annual MSLs for 1875-1884 are 0.14 cm, including the corrections from MTL to MSL (Table 3); this value is also representative of the error for 1885-1889.The time series reflects the behaviour of sea level variability common to the Mediterranean Sea stations, coherent with the global sea level rise, except for a period of stability approximately between the 1960s and the early 1990s (Tsimplis and Baker, 2000;Marcos and Tsimplis, 2008;Gomis et al., 2012;Zerbini et al., 2017).

The long-term mean sea level time series
The digitization of the 1917-1938 hourly data and the revision of the whole time series mostly led to minor differences from the previously known monthly means, but there are some exceptions.Appendix C summarizes the main differences between the values obtained in this work and those in the PSMSL database used for reference.
There is an issue with the annual mean of 1869.The MSL reported in Lorenz et al. (1873) corresponds to 1.407 m above ZIT, and this has been verified to be correct using the data therein.On the other hand, the MSL of 1869, used to define the zero of the Austrian levelling network, is 1.118 m below ZMS (MGI, 1885(MGI, , 1892;;see Sect. 2.3.1),i.e. 1.419 m above ZIT.The reason for the difference is unknown, but it could be in the conversion from the Viennese foot (the original unit) to the SI units and subsequent rounding or truncations.
In practice, the MSL in Lorenz et al. (1873) is homogenous with the rest of the sea level data that are referred to as ZIT, while the MSL in MGI (1885MGI ( , 1892) ) represents the height of ZMS in the Austrian levelling reference system and is involved in the time series of the benchmark heights shown in Fig. 5.A problem might arise only if the two reference systems interacted with each other, which is not the case in this work.
The linear trend computed with all the annual MSLs from 1869 to 2021 (Fig. 8b) is 1.36 ± 0.17 mm yr −1 .Taking into account only the period covered by hourly data , the trend is the same, namely 1.36 ± 0.19 mm yr −1 .Note that in the fit the annual MSLs have been weighted with the inverse of the respective errors (see Sect. 3.2).
Among the factors that affect the sea level variability, there is the inverted barometer (IB), which consists of an inverse relationship between variations of atmospheric pressure and sea level.In equilibrium conditions, 1 hPa of atmospheric pressure increase approximately corresponds to 1 cm of sea level decrease, and vice versa.According to Raicich and Colucci (2021a), the atmospheric pressure at Trieste exhibits a significant linear trend during the last 150 years, namely 0.5 ± 0.2 hPa per century.The standard IB correction (−1 cm hPa −1 ) was applied to the MSL time series using the pressure data in Raicich and Colucci (2021b).As a result, for the IB-corrected MSL, we obtain linear trends of 1.45 ± 0.13 mm yr −1 for 1869-2021 and 1.46 ± 0.15 mm yr −1 for 1905-2021.
The linear trends estimated here are slightly lower than the global value of 1.73 ± 0.44 mm yr −1 in Fox-Kemper et al. ( 2021) for 1901-2018 but are within the interval defined by the uncertainties.
The sea level acceleration was estimated as twice the coefficient of the quadratic term of the secondorder polynomial fit.If the observed sea level is taken into account, for 1869-2021 the acceleration is 0.006 ± 0.005 mm yr −2 (p = 0.23 significance), and for 1905-2021 it is 0.008 ± 0.006 mm yr −2 (p = 0.22).From the IB-corrected sea level, the acceleration is 0.008 ± 0.004 mm yr −2 (p = 0.05 significance) for 1869- To assess how the steady sea level period affects the trend and acceleration estimates, we analysed an IB-corrected time series from which the annual means from 1967 to 1995 were removed.As a result, the linear trends are 1.55 ± 0.14 mm yr −1 for 1869-2021 and 1.56 ± 0.16 mm yr −1 for 1905-2021, while the accelerations are 0.004 ± 0.005 mm yr −2 for 1869-2021 and 0.004 ± 0.0006 mm yr −2 for 1905-2021, respectively.In both cases accelerations are statistically not significant at p = 0.05.

Data availability
The hourly sea level data and the derived monthly and annual mean sea levels used in this work are available from SEANOE (https://doi.org/10.17882/62758,Raicich, 2022).

Summary and conclusions
We have revised and updated the information about the sea level observations carried out at Trieste, Molo Sartorio, from 1869 to 2021, using the material in the archive of CNR, Institute of Marine Sciences (ISMAR), Trieste, and published and unpublished documents from other sources.
We could identify the tide gauges used for the observations, and we could recover the heights of the geodetic benchmarks on Molo Sartorio since the late 19th century.
The digitization of hourly data from 1905 to 1938 from the original tabulations or charts allowed one to extend the available time series from 1905 to 2021.The quality control provided information on data gaps, most of which were filled by interpolation in order to obtain reasonable daily sea level https://doi.org/10.5194/essd-15-1749-2023 Earth Syst.Sci.Data, 15, 1749-1763, 2023 Marittimo, 1877-1890), and March 1884, which was previously reported erroneously (Raicich, 2007).
Besides the method of computing the MSL, differences come from the interpolation of gaps.
4. Monthly and annual MSLs are now available for 1917-1938, while, previously, they were only available for 1920-1922.The PSMSL database does not include the data for 1917, 1918, and 1923and July-December 1926; instead, it includes December 1924, which does not exist.
5. The monthly MSLs of November-December 1922 and January-November 1924 were substantially corrected after revising the relationship between the zero level of the charts and the tide-gauge zero.
Competing interests.The author has declared that there are no competing interests.
Disclaimer.Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Figure 2 .
Figure 2. The tide gauge installed in 1859.View of the instrument from the front (528a) and from above (528b); vertical section of the north-western end of the building hosting the tide gauge and the stilling well (528c).Mean high and low waters are indicated by "höchster Wasserstand" and "kleinste Ebbe", respectively.The figures are adapted from von Chiolich-Löwensberg (1865, 1866).

Figure 3 .
Figure 3. (a) Map of Molo Sartorio before and (b) after the construction works of 1925-1926.(c) Image of the new tide-gauge hut probably in the early 1930s."Thg.1859 1925" ("Tide gauge 1859 1925") indicates the position of the old tide gauge (a) and "Thg.1926 óta" ("Tide gauge from 1926") that of the new tide-gauge hut (a, b)."Vízmérce" indicates the hydrometer (a, b).North is upwards.Adapted from Bendefy (1958), who credited the photograph to Silvio Polli (Geophysical Institute of Trieste).(Note: the scale unit in panel (a) should be "m".)

Figure 4 .Figure 5 .
Figure 4. (a) The tide-gauge hut in 2001; since 2004, only the wall indicated by the arrow has been visible (see the inset), because the hut is enclosed in the main building, which was enlarged.(b) The inside of the tide-gauge hut in 2001.Four instruments are shown: Ott Thalimedes (1st) (Ott T.), Ott Büsum (Ott B.), Thalassia, and Fuess.PM is the tide-gauge contact point (TG CP), and CsV 39 is a vertical benchmark no longer present.(Photographs by CNR-ISMAR, Trieste.)

Figure 6 .
Figure 6.Heights of the benchmarks used to build the composite time series; the year of the survey is in brackets.The zeros of the levelling networks are shown: Austrian Zero of 1869 (AZ1869), Italian Zero of 1894 (IZ1894), and Italian Zero of 1942 (IZ1942).The differences between the zeros have been estimated as discussed in Appendix A.

Figure 7 .
Figure 7. (a) Monthly errors (cm) for 1875-1975; both the errors in MTL and MSL are shown for 1875-1889.(b) Monthly percentages of valid days for 1875-1975.After 1975 the error is always 0.03 cm, and there are no missing days.The black segments indicate that the errors and percentages of valid days are unknown.The labels on the abscissa axis refer to the beginning of the year.

Figure 8 .
Figure 8.(a) Monthly MSL and (b) annual MSL.The data are expressed in centimetres relative to the TG Zero (ZIT; see Fig. 5).The labels on the abscissa axis in panel (a) refer to the beginning of the year.

Figure 8
Figure 8 displays the time series of monthly MSL (a) and annual MSL (b) relative to ZIT.The time series reflects the behaviour of sea level variability common to the Mediterranean Sea stations, coherent with the global sea level rise, except for a period of stability approximately between the 1960s and the early 1990s(Tsimplis and Baker, 2000;Marcos and Tsimplis, 2008;Gomis et al., 2012;Zerbini et al., 2017).The digitization of the 1917-1938 hourly data and the revision of the whole time series mostly led to minor differences from the previously known monthly means, but there are some exceptions.Appendix C summarizes the main differences between the values obtained in this work and those in the PSMSL database used for reference.

Figure C1 .
Figure C1.Differences between the monthly MSLs obtained in this work and those in the PSMSL database (cm).The blue dots highlight differences greater than 1 cm in absolute value.The labels on the abscissa axes refer to the beginning of the year.

Table 1 .
. Instruments used in the tide-gauge station and sources of their technical characteristics.R is the reduction ratio; V is the paper speed.A dash (-) indicates that the model is unknown.The information in square brackets is uncertain.

Table 2 .
(a) Heights (m) of the benchmarks near the tide gauge at Molo Sartorio relative to IZ1942.The measurements in the original reference system, namely AZ1869 for 1876 and 1884 and IZ1894 for 1926, are reported in parentheses.(b) Composite time series of heights (m) of VZMS = ZMS − 0.0109 m and CsO 54/39 /39 relative to IZ1942.TG BM is the tide-gauge benchmark, and TG CP is the tide-gauge contact point.See the text for other details on denominations and normalizations.

Table 3 .
Average monthly and annual differences (cm) betweenMTL and MSL from 1917-2021 data.