*Coastal Sea Level Trends from a Joint Use of Satellite Radar Altimetry, GPS and Tide… DOI: http://dx.doi.org/10.5772/intechopen.98243*

were attained by a different methodology in calculating the rates of absolute and relative sea level change rates and their formal errors. Moreover, in the years following 2010 the VLM rates at the five common TG locations have remained substantially unmodified with respect to the Wöppelmann and Marcos' results. As a final step regarding VLM, we have calculated the root mean square difference (RMSD) of the VLM calculated with GPS and those calculated with the classical and the new (LIPWC-COV) approaches and found that the second one is lower: Classic approach: 1.84 mm y<sup>1</sup> ; LIPWC-COV approach: 1.34 mm y<sup>1</sup> .

The discrepancy observed between this study and that of Wöppelmann and Marcos can largely be ascribed to the different periods covered by the altimetry datasets (C3S and SLCCI datasets cover time periods respectively 44% and 23% longer than the study of Wöppelmann and Marcos). Other factors that may contribute to explain the difference between the results of the two studies are the processing of the altimetry data and the inclusion of the VEPTF TG in this study. The rates of absolute sea level change at the TGs, calculated as the sum of relative sea level change and VLMs derived in this study with the LIPWC-COV approach, for the whole period covered by the TG record, are reported in **Table 6**.

The uncertainty of the sample mean (last row of **Table 6**) was obtained as standard error of the sample mean, considering the rates as random and independent variables. The absolute sea level change rates vary in a very narrow interval, 2.33–2.71, with a sample mean of 2.43 mm y<sup>1</sup> . The standard deviation of the sample is much lower than the precision of each individual determination of SL change rate at the TGs. As pointed out by Wöppelmann and Marcos [50], such a low dispersion is unlikely to be determined from estimates of independent random variables: it is instead the evidence of the high performance of LIPWC method for determining accurate VLM rates from TG and altimetry differenced time series. The ASLR rates calculated by altimetry in 1993–2018 and through the LIPWC-COV technique (1974–2018) are shown in **Figure 6**.

Clearly, the ASLR values calculated for the longer period are smaller than those calculated in the shorter one, but the modulation of the rate from TG to TG is apparently reflected in the LIPWC-COV approach. As already noted, the errors associated to the ASLR rates derived in the LIPWC-COV are also smaller, thanks to the introduction of the constraints on the relative vertical land motion between paired TGs. The mean value of the ASLR calculated for the Adriatic Sea with the LIPWC-COV approach, is in general agreement with both regional studies on the Mediterranean Sea (0.7 0.2 mm y<sup>1</sup> (1945–2000) [65]; 1.60 0.35 mm y<sup>1</sup> (1992–2010) [50]; 2.44 0.5 mm y<sup>1</sup> (1993–2012) [66]; 2.87 0.33 mm y<sup>1</sup> (1992–2016) [67]),


#### **Table 6.**

*ASLR from TG records over whole period 1974–2018, corrected for VLM estimated with the LIPWC-COV approach. All data are in mm y<sup>1</sup> .*

**Figure 5**, where the plot of the LIPWC-COV solution follows the general form of

*VLM results using C3S altimetry dataset (1993–2018). Location in column 1; ASLR in column 2; RSLR in the altimetry era (1993-2018) in column 3; RSLR 1974-2018 in column 4; VLM calculated with the classical approach in column 5, as in column 4 of Table 3, and with the LIPWC-COV in column 6. Columns 7 reports*

\_ *ζ* **(mm y**�**<sup>1</sup> )**

VENEZIA 3.36 � 1.45 5.15 � 1.73 3.26 � 0.73 �1.79 � 0.65 �0.93 � 0.39 �1.59 � 0.65 VEPTF 3.38 � 1.46 5.50 � 1.73 3.78 � 0.73 �2.12 � 0.67 �1.41 � 0.47 — TRIESTE 3.75 � 1.58 3.56 � 1.66 2.30 � 0.67 0.18 � 0.60 0.42 � 0.33 �0.25 � 0.52 ROVINJ 3.33 � 1.58 1.03 � 1.85 1.36 � 0.71 2.30 � 1.06 0.93 � 0.37 �1.51 � 1.03 SPLIT 3.60 � 1.36 2.92 � 1.65 2.20 � 0.66 0.68 � 0.63 0.37 � 0.33 0.10 � 0.64 DUBROVNIK 3.34 � 1.22 3.79 � 1.48 2.69 � 0.58 �0.45 � 0.55 �0.41 � 0.46 �1.83 � 0.70

ð Þ *g*\_ � *s*\_ **(mm y**�**<sup>1</sup> )** **LIPWC-COV (mm y**�**<sup>1</sup> )**

**GPS (mm y**�**<sup>1</sup> )**

In **Figure 5** are reported also the VLM values measured by the CGPS stations of PSAL (VENEZIA), TRIE (TRIESTE) and SPLT (SPLIT), and the values of VLM estimated by Wöppelmann and Marcos [50] with the LIPWC technique without the

In the classical approach, as there is no optimization of errors as in the LIPWC technique, we see a wide spread of the VLM values. This is particularly evident for

Wöppelmann and Marcos [50] presents much lower standard errors than LIPWC-COV solution described in this study. We presume that such low standard errors

*Scatterplots of VLM values derived with the classical* ð Þ g\_ � s\_ *and the LIPWC-COV approaches using the C3S altimetry dataset (period 1993–2018). GPS estimates (in black) are also reported. Results from the study of Wöppelmann and Marcos (W&M) for the period 1992–2010 are shown in green for comparison. The zero*

, while the LIPWC-

. The LIPWC solution proposed by

the classical solution, but with a reduced spread.

COV approach calculates it as less than 1 mm y�<sup>1</sup>

ROVINJ TG, whose ð Þ *<sup>g</sup>*\_ � *<sup>s</sup>*\_ estimates reach more than 2 mm y�<sup>1</sup>

*the VLM values directly detected by the GPS stations associated with three TGs.*

change of variable.

**Table 5.**

**Figure 5.**

**110**

*level is drawn in black. (adapted from [58]).*

**Location** *g*\_

**(mm y**�**<sup>1</sup> )**

*s*\_ **(mm y**�**<sup>1</sup> )**

*Geodetic Sciences - Theory, Applications and Recent Developments*

#### **Figure 6.**

*Absolute sea level change rates as calculated by altimetry 1993–2018, and by the LIPWC-COV approach integrating data from TGs in 1974–2018.*

and at global scale (2.0 � 0.3 mm y�<sup>1</sup> (1971–2010) [68]; 3.0 � 0.7 mm y�<sup>1</sup> (1993– 2010) [69]; 2.8 � 0.5 mm y�<sup>1</sup> (1993–2010) [68]; 3.1 mm y�<sup>1</sup> (1992–2018) [32]).

Among the ASLR altimetry rates associated with the six TGs in the Adriatic Sea, those for TRIESTE are very different in the C3S and SLCCI dataset. In order to investigate such a large difference (0.02 � 0.71 mm y�<sup>1</sup> SLCCI; �1.07 � 0.74 mm y�<sup>1</sup> C3S; see **Table 4**) the SLCCI-AT X-TRACK/ALES 20 Hz along-track coastal altimetry dataset has been used.

The analysis focuses on the descending track 196 of the Jason-1 (2001–2013) and Jason-2 (2008–2019) altimetry missions and covers the period 2002–2016 with 532 cycles (from 22 Jan 2002 to 23 Jun 2016 at 350 m resolution along-track) with MSSH computed using cycles from 1 to 517. The position of the track 196 and the geographical setting are shown in **Figure 7**.

Altimetry data at the 71 observation points of track 196 are compared to 10<sup>0</sup> interval RSL observations of the TRIESTE TG. The TG time series did not undergo any filtering or processing, and the astronomical tide and Dynamic Atmospheric Correction (DAC) corrections are not applied to the altimetry time series.

The goal of the investigation is to explore the possible causes of the different ASLR rates obtained by the two gridded altimetry datasets near Trieste, to look for clues directly into the original along track data from the Jason missions, reprocessed with advanced and coastal specific re-tracking (ALES) and improved coastal processing (X-TRACK). We also want to ascertain the suitability of the new SLCCI-AT record in long term coastal sea level monitoring. We concentrated on the altimeter track 196 of the SLCCI-AT dataset, which first crosses Marano Lagoon and a 0.5 km wide sandbar before entering the Gulf of Trieste from north, near Grado, and then flies over Umag and the full extent of the Istria peninsula. The retrieval is particularly problematic in the gulf area due to the complex morphology of the land. Moreover, some data loss could be due to sea-to-land and land-to-sea crossings that might influence the behavior of the on-board tracker. Operational altimetry products do not provide data over this section of the Gulf of Trieste, while the SLCCI-AT dataset provides 71 points along track, most of which yield over 70% of valid data (blue box in **Figure 8**). The most improvement is near the Istrian peninsula with more than 90% of data recovered. The valid data percentages decrease abruptly over a distance ranging 5 km from the coast. The reduced performance over the

lagoon and islet (almost all data have been rejected) is probably related to the data corruption in the land-sea-transition. Note that at 1 Hz, any coastal altimetry along track product would give no more than 3–4 points along this 24 km long stretch of

*J1 + J2 track 196 geographical settings. Left: Percentage of valid data along the track. Right: Correlation with TRIESTE tide gauge. Adapted from the coastal sea level project of the ESA climate change initiative (SL\_CCI bridging phase) document "Part II: Validation Results" (http://www.esa-sealevel-cci.org/webfm\_send/588).*

*Gulf of Trieste. The positions of the SLCCI along-track 20 Hz altimetric product version 1.0 samples of the descending track 196 (white circles), and the TRIESTE TG stations (red triangle). Umag in Croatia and*

*Coastal Sea Level Trends from a Joint Use of Satellite Radar Altimetry, GPS and Tide…*

The data accuracy can be assessed in more detail comparing the altimeterderived 20-Hz SLA with corresponding tide gauge sea level measurements. It should be noted that the TRIESTE TG is located in the harbor, and therefore it does not measure exactly the same ocean dynamics as the altimeter flying offshore. Nonetheless, the Pearson's linear correlation coefficient of most of the 71 points along the section of track 196 facing the Trieste harbor exceeds 0.9 (red box in **Figure 8**). The RMS difference between altimetry observations and tide gauge measurements of instantaneous sea level is almost constant along the track 196

section in the Gulf of Trieste, and around 10 cm (not shown).

track 196.

**113**

**Figure 8.**

**Figure 7.**

*Grado in Italy are also shown (green circles).*

*DOI: http://dx.doi.org/10.5772/intechopen.98243*

*Coastal Sea Level Trends from a Joint Use of Satellite Radar Altimetry, GPS and Tide… DOI: http://dx.doi.org/10.5772/intechopen.98243*

**Figure 7.**

and at global scale (2.0 � 0.3 mm y�<sup>1</sup> (1971–2010) [68]; 3.0 � 0.7 mm y�<sup>1</sup> (1993– 2010) [69]; 2.8 � 0.5 mm y�<sup>1</sup> (1993–2010) [68]; 3.1 mm y�<sup>1</sup> (1992–2018) [32]).

*Absolute sea level change rates as calculated by altimetry 1993–2018, and by the LIPWC-COV approach*

*Geodetic Sciences - Theory, Applications and Recent Developments*

those for TRIESTE are very different in the C3S and SLCCI dataset. In order to investigate such a large difference (0.02 � 0.71 mm y�<sup>1</sup> SLCCI; �1.07 � 0.74 mm y�<sup>1</sup> C3S; see **Table 4**) the SLCCI-AT X-TRACK/ALES 20 Hz along-track coastal

Jason-2 (2008–2019) altimetry missions and covers the period 2002–2016 with 532 cycles (from 22 Jan 2002 to 23 Jun 2016 at 350 m resolution along-track) with MSSH computed using cycles from 1 to 517. The position of the track 196 and the

Altimetry data at the 71 observation points of track 196 are compared to 10<sup>0</sup> interval RSL observations of the TRIESTE TG. The TG time series did not undergo any filtering or processing, and the astronomical tide and Dynamic Atmospheric Correction (DAC) corrections are not applied to the altimetry time series.

The goal of the investigation is to explore the possible causes of the different ASLR rates obtained by the two gridded altimetry datasets near Trieste, to look for clues directly into the original along track data from the Jason missions, reprocessed with advanced and coastal specific re-tracking (ALES) and improved coastal processing (X-TRACK). We also want to ascertain the suitability of the new SLCCI-AT record in long term coastal sea level monitoring. We concentrated on the altimeter track 196 of the SLCCI-AT dataset, which first crosses Marano Lagoon and a 0.5 km wide sandbar before entering the Gulf of Trieste from north, near Grado, and then flies over Umag and the full extent of the Istria peninsula. The retrieval is particularly problematic in the gulf area due to the complex morphology of the land. Moreover, some data loss could be due to sea-to-land and land-to-sea crossings that might influence the behavior of the on-board tracker. Operational altimetry products do not provide data over this section of the Gulf of Trieste, while the SLCCI-AT dataset provides 71 points along track, most of which yield over 70% of valid data (blue box in **Figure 8**). The most improvement is near the Istrian peninsula with more than 90% of data recovered. The valid data percentages decrease abruptly over a distance ranging 5 km from the coast. The reduced performance over the

altimetry dataset has been used.

*integrating data from TGs in 1974–2018.*

**Figure 6.**

**112**

geographical setting are shown in **Figure 7**.

Among the ASLR altimetry rates associated with the six TGs in the Adriatic Sea,

The analysis focuses on the descending track 196 of the Jason-1 (2001–2013) and

*Gulf of Trieste. The positions of the SLCCI along-track 20 Hz altimetric product version 1.0 samples of the descending track 196 (white circles), and the TRIESTE TG stations (red triangle). Umag in Croatia and Grado in Italy are also shown (green circles).*

**Figure 8.**

*J1 + J2 track 196 geographical settings. Left: Percentage of valid data along the track. Right: Correlation with TRIESTE tide gauge. Adapted from the coastal sea level project of the ESA climate change initiative (SL\_CCI bridging phase) document "Part II: Validation Results" (http://www.esa-sealevel-cci.org/webfm\_send/588).*

lagoon and islet (almost all data have been rejected) is probably related to the data corruption in the land-sea-transition. Note that at 1 Hz, any coastal altimetry along track product would give no more than 3–4 points along this 24 km long stretch of track 196.

The data accuracy can be assessed in more detail comparing the altimeterderived 20-Hz SLA with corresponding tide gauge sea level measurements. It should be noted that the TRIESTE TG is located in the harbor, and therefore it does not measure exactly the same ocean dynamics as the altimeter flying offshore. Nonetheless, the Pearson's linear correlation coefficient of most of the 71 points along the section of track 196 facing the Trieste harbor exceeds 0.9 (red box in **Figure 8**). The RMS difference between altimetry observations and tide gauge measurements of instantaneous sea level is almost constant along the track 196 section in the Gulf of Trieste, and around 10 cm (not shown).

From the time series of SLCCI-AT SLA at each data point of the track 196 facing Trieste, we have calculated the slopes of the fitting lines, gradually growing the confidence interval from 68–95%, and performed Mann-Kendall statistical significance tests [70, 71] modified for autocorrelated data [72] on all the 71 fitting lines. The Mann-Kendall test is commonly employed to detect monotonic trends in time series. The null hypothesis is that the data come from a population with independent realizations and are identically distributed. The alternative hypothesis is that the data follow a monotonic trend. In **Figure 9** the results of such calculation are reported for a preliminary version of the SLCCI-AT dataset at 20 Hz. The black diamonds mark the acceptance or rejection of the null hypothesis following this scheme:


Already with a 68% confidence interval the null hypothesis (the trend is not statistically significant) is rejected in less than 24% of the data points. With a 95% confidence interval only for four fitting lines out of 71 the null hypothesis is rejected. In both cases the errors associated to the slopes are higher than the slopes themselves.

A similar analysis replicated on the final version of the SLCCI-AT dataset, published at the end of the SLCCI project, gave better results. **Figure 10** reports the representation of the statistical characteristics of the slopes derived from the last version of the data of the SLCCI-AT X-TRACK/ALES SLA 20 Hz, with 95% confidence interval. The left panel shows slopes and associated errors at every data point latitude (low latitudes are near Umag, high latitudes near Grado); different colors indicate the statistical significance of the Mann-Kendall test (blue: significant; orange: not significant). The right panel shows the box and whisker plots of the two distributions (left: not significant; right: significant). The number of statistically significant slopes is much higher in the final version of the dataset, even if the variability is still rather high and difficult to explain because of the limited spatial variability along the track. Slopes are higher towards north (Grado), and lower near Umag. Considering only the statistically significant slopes in the SLCCI-AT dataset, their sample mean and standard deviation result to be 3.40 1.01 mm y<sup>1</sup> (Feb-2002 – Jun-2016) which is not far from the trends we have found in the Adriatic Sea at all the tide gauges. We recalculated the altimetry trends near TRIESTE in the

SLCCI and C3S gridded products. The altimetry ASLR trends found so far in the

3.40 1.01 3.66 3.97 5.07 3.64

*SLCCI SLA 20 Hz. Left: Slopes and slope errors of the lines fitting every data point of the track 196 in the Gulf of Trieste. Blue: Statistically significant slopes according to the Mann-Kendall test. Orange: Slopes not significant. Right: Box and whisker plots for the statistically significant and non-significant slopes. Red: Median value. Box: Upper and lower quartiles. Whiskers: Highest and lowest observations. Adapted from the coastal sea level project of the ESA climate change initiative (SL\_CCI bridging phase) document "Part II: Validation*

*Coastal Sea Level Trends from a Joint Use of Satellite Radar Altimetry, GPS and Tide…*

*g*\_ **TRIESTE 2002–2016**

**SLCCI (mm y<sup>1</sup> )**

*Trends for Trieste in February 2002 – June 2016 from SLCCI project and C3S altimetry. Column 1: SLCCI-AT along track 20 Hz product. Column 2: SLCCI gridded product. Column 3: C3S gridded product.*

**C3S (mm y<sup>1</sup> )**

The trends calculated with the SLCCI dataset (along track and gridded) are in good agreement, apart from the different errors affecting the two results, due to the different methods used to calculate them. The C3S trend is instead higher than the other two. We believe that the difference between the results is to be ascribed to the different methodologies used in the two products. In any case the difference between the SLCCI and the C3S results is not yet explained by this further analysis, and the Gulf of Trieste remains a controversial place for the derivation of climatologically relevant oceanic variables from altimetry, because of the proximity of the land and the geometry of the surrounding coastline, and the very short time cover-

The sea level is a key variable of the climate system. Tide gauges measuring sea level variability are in operation since the 1900s. Satellite-based observations of sea level changes are more recent. Nevertheless, they play a crucial role in understanding the future coastal sea level changes. Advance in the processing of satellite radar altimetry have expanded the utility of this data set for climate-related studies and extended the potential exploitation in the coastal zone. The joint usage of the two different measuring systems (in situ and satellite) has two challenges. First how the two data sets can be consistently and systematically used in synergy to address that objective of estimating robust coastal sea level trends. Second how using high-rate (i.e. 20 Hz) altimeter measurements with a coastal-oriented processing could improve the satellite-based trend estimates with respect to the standard (1 Hz) data,

analysis are summarized in **Table 7**.

*Results" (http://www.esa-sealevel-cci.org/webfm\_send/588).*

*DOI: http://dx.doi.org/10.5772/intechopen.98243*

**Figure 10.**

**SLCCI-AT (mm y<sup>1</sup> )**

**Table 7.**

age of the altimetric datasets.

**8. Summary and prospects**

especially near coast.

**115**

#### **Figure 9.**

*Slopes and slop errors of the lines fitting the time series of along track SLA at every data point of track 196 near Trieste. Also plotted the Mann-Kendall test results. Black diamonds: 1 - rejection of the null hypothesis (the sample has no trend); 0 - no rejection. Green line: p\_value. Adapted from the coastal sea level project of the ESA climate change initiative (SL\_CCI bridging phase) document "Part II: Validation Results" (http://www.esa-sea level-cci.org/webfm\_send/588).*

*Coastal Sea Level Trends from a Joint Use of Satellite Radar Altimetry, GPS and Tide… DOI: http://dx.doi.org/10.5772/intechopen.98243*

#### **Figure 10.**

From the time series of SLCCI-AT SLA at each data point of the track 196 facing Trieste, we have calculated the slopes of the fitting lines, gradually growing the confidence interval from 68–95%, and performed Mann-Kendall statistical significance tests [70, 71] modified for autocorrelated data [72] on all the 71 fitting lines. The Mann-Kendall test is commonly employed to detect monotonic trends in time series. The null hypothesis is that the data come from a population with independent realizations and are identically distributed. The alternative hypothesis is that the data follow a monotonic trend. In **Figure 9** the results of such calculation are reported for a preliminary version of the SLCCI-AT dataset at 20 Hz. The black diamonds mark the acceptance or rejection of the null hypothesis following this

• 1 - the null hypothesis "the sample has no trend" is rejected.

*Geodetic Sciences - Theory, Applications and Recent Developments*

• 0 - the null hypothesis "the sample has no trend" cannot be rejected.

Already with a 68% confidence interval the null hypothesis (the trend is not statistically significant) is rejected in less than 24% of the data points. With a 95% confidence interval only for four fitting lines out of 71 the null hypothesis is rejected. In both cases the errors associated to the slopes are higher than the slopes

A similar analysis replicated on the final version of the SLCCI-AT dataset, published at the end of the SLCCI project, gave better results. **Figure 10** reports the representation of the statistical characteristics of the slopes derived from the last version of the data of the SLCCI-AT X-TRACK/ALES SLA 20 Hz, with 95% confidence interval. The left panel shows slopes and associated errors at every data point latitude (low latitudes are near Umag, high latitudes near Grado); different colors indicate the statistical significance of the Mann-Kendall test (blue: significant; orange: not significant). The right panel shows the box and whisker plots of the two distributions (left: not significant; right: significant). The number of statistically significant slopes is much higher in the final version of the dataset, even if the variability is still rather high and difficult to explain because of the limited spatial variability along the track. Slopes are higher towards north (Grado), and lower near Umag. Considering only the statistically significant slopes in the SLCCI-AT dataset, their sample mean and standard deviation result to be 3.40 1.01 mm y<sup>1</sup> (Feb-2002 – Jun-2016) which is not far from the trends we have found in the Adriatic Sea at all the tide gauges. We recalculated the altimetry trends near TRIESTE in the

*Slopes and slop errors of the lines fitting the time series of along track SLA at every data point of track 196 near Trieste. Also plotted the Mann-Kendall test results. Black diamonds: 1 - rejection of the null hypothesis (the sample has no trend); 0 - no rejection. Green line: p\_value. Adapted from the coastal sea level project of the ESA climate change initiative (SL\_CCI bridging phase) document "Part II: Validation Results" (http://www.esa-sea*

scheme:

themselves.

**Figure 9.**

**114**

*level-cci.org/webfm\_send/588).*

*SLCCI SLA 20 Hz. Left: Slopes and slope errors of the lines fitting every data point of the track 196 in the Gulf of Trieste. Blue: Statistically significant slopes according to the Mann-Kendall test. Orange: Slopes not significant. Right: Box and whisker plots for the statistically significant and non-significant slopes. Red: Median value. Box: Upper and lower quartiles. Whiskers: Highest and lowest observations. Adapted from the coastal sea level project of the ESA climate change initiative (SL\_CCI bridging phase) document "Part II: Validation Results" (http://www.esa-sealevel-cci.org/webfm\_send/588).*


#### **Table 7.**

*Trends for Trieste in February 2002 – June 2016 from SLCCI project and C3S altimetry. Column 1: SLCCI-AT along track 20 Hz product. Column 2: SLCCI gridded product. Column 3: C3S gridded product.*

SLCCI and C3S gridded products. The altimetry ASLR trends found so far in the analysis are summarized in **Table 7**.

The trends calculated with the SLCCI dataset (along track and gridded) are in good agreement, apart from the different errors affecting the two results, due to the different methods used to calculate them. The C3S trend is instead higher than the other two. We believe that the difference between the results is to be ascribed to the different methodologies used in the two products. In any case the difference between the SLCCI and the C3S results is not yet explained by this further analysis, and the Gulf of Trieste remains a controversial place for the derivation of climatologically relevant oceanic variables from altimetry, because of the proximity of the land and the geometry of the surrounding coastline, and the very short time coverage of the altimetric datasets.
