Point wise Determination of Vertical Deformation at southern Coastal Areas of Caspian Sea by Combination of Satellite Altimetry and Tide Gauge Observations

In this paper, a combination of satellite altimetry (SA) and tide-gauge (TG) observations is used to determine the pattern of vertical deformation at southern coasts of Caspian Sea. Satellite altimetry measures the variations of sea water surface with a respect to a reference ellipsoid, while tide gauge data gives the sea level relative to a particular ground station; consequently, the difference between these two observations will give the vertical motion at the ground station. However, altimeter satellites still have problems at shallow waters or coastal areas which leads to inaccurate and corrupted sea surface height at tide gauge stations. To minimize this imprecision, two strategies are applied for the selection of the suitable SA points near the TG stations. The First one is to choose those close passes to the TG stations, provided that the SA data is not corrupted (range observation is available). The second strategy is to extrapolate the SA points at TG stations using Spatio-Temporal Kriging method. In both point of views, use is made of re-tracking methods to improve the ranges of SA observations near the coast. Amplitudes of tidal components are removed from both SA and TG observations to derive the residual tide free signals. Afterward, a linear trend is fitted to the residual signals in the sense of least square; the difference between these two, supplies the vertical deformation of the ground station. The results illustrate that the average vertical crustal motion at Anzali, Noshahr and Neka are -1.387, -1.903 and -3.14 mm/year, respectively.


Introduction
Caspian Sea is encircled by five countries, namely Islamic Republic of Iran, Russia, Republic of Azerbaijan, Turkmenistan, and Kazakhstan. It is the largest extant part of the old Tethys Sea; an ancient sea which had extended from North Pole to Indian Ocean in the first three eras of geology. Disconnection of Tethys occurred in the third period due to geological folding and resulted in the formation of famous Caucasus and Minor Asia mountain ranges; consequently, an uplift happened in European continent and Iran Plain. This in turn, eventuated in emergence of Caspian Sea. As shown in Figureure1, the Caspian Sea is situated north of the equator, from 36°33' to 47°57' north latitude and 46°43' to 54°53' east longitude. This geographical range covers an approximate length and width of 1030 and 240 km respectively [Arpe et al., 2000].
The structural tectonic features in the South Caspian Sea region are particularly important for geodynamical studies. These are mainly due to the active tectonics and structure of the South Caspian Sea and its role in the collision between Arabia and Eurasia [Jackson et al., 2002;Brunet et al., 2003]. The South Caspian depression is an autonomous tectonic unit that consists of: the southern part of the Caspian Sea; the West Turkmen depression adjacent to the latter; the northern lowland periphery of the Alborz system; the Lower Kura trough; and the Apsheron-Kobystan trough [Mammadov, 2008]. This interaction between tectonic plates makes the region subject to frequent and strong earthquake. The South Caspian Basin is surrounded by active earthquake belts on all sides.
The southern Caspian Sea stands out as an aseismic block surrounded by belts of intense earthquake activity. These phenomena cause deformations in the crust of the Earth [Jackson et al., 2002].
A large number of geodynamic studies based on crustal deformations are done by using observations that typically derived from absolute gravimetry [Pagiatakis and Salib, 2003], global positioning system (GPS) [Larson and Van Dam, 2000;Johansson et al., 2002;Caccamise et al., 2005;Beavan et al., 2010], satellite laser ranging (SLR) [Noomen et al. 1996], Doppler orbitography and radiopositioning integrated by satellite (DORIS) , interferometric synthetic aperture radar -InSAR [Dalla et al., 2012], and very long baseline interferometry (VLBI) [Tesmer et al., 2009]. Most of the crustal deformation studies are based on Global Navigation Satellite System (GNSS) observations due to high levels of accuracy, moderate associated costs, and broad spatial coverage in a relation to SLR, VLBI and other techniques [Blewitt, 2007].
However, along the South of Caspian Sea, particularly coastal region, a dense, long term time series and homogenous GPS geodynamic network do not currently exist, and GPS geodynamic stations are located far from the coastal region. Consequently, studying geodynamic phenomena in south coastal area of Caspian Sea presents difficulties. Therefore, other alternatives are necessary. SA and TG records provide alternative and independent methods for estimating vertical crustal motion Lin, 2000;Nerem and Mitchum, 2002;Fenoglio-Marc et al., 2004].
Due to strategic and economic importance of coastal areas and their impact on local people's lives, dynamical changes of these areas and their sources have captured the interests of scientific communities in the recent years.
Multitude of factors such as meteorological effects, oceanic influences, tide, climate change and vertical crustal motion control the dynamics of these regions [García et al., 2007;Pirouznia et al., 2017]. Among these factors, the considerable effect that vertical crustal motion (VCM) has on TG measurements, makes it a subject of interest in geophysical studies. For this reason, hitherto, so much effort has been made to estimate this motion [see Clark et al., 2002;Nerem and Mitchum, 2002].
Tide gauges measure the relative sea level changes, which is the variation of the mean sea surface with a respect to the solid Earth. To obtain such measurements, TG's must be referenced to a benchmark on the crust of the Earth.
However, in the most parts of the world, locations of these benchmarks are subjected to the Earth's deformation processes such as earthquakes and their deformation cycles, tectonic plate motion, basin evolution, Glacial Isostatic Adjustment (GIA), and anthropogenic effects such as local subsidence due to water extraction, mining, oil or natural gas drilling. Therefore, the signals of TG records contain both sea level changes and VCMs [Kuo et al., 2008].
Observations of SA consist of (i) the precise satellite location in a relation to a reference ellipsoid, and (ii) the height of satellite over the surface measured using laser or radar pulses which are sent by satellite and bounced back from the target surface. Subtracting (ii) from (i) gives the sea surface height (SSH) relative to the reference ellipsoid at the given location and time. On the other hand, the TG measurement gives the sea level height relative to the solid ground on which the TG is fixed. Hence any vertical ground motion, regardless of its source, will be interpreted as an illusory change in sea level height observed by TG. As the land under the benchmark moves up and down, an apparent fall or rise will be seen in sea level which is due to the tectonic movement of sea floor. In other words, TG measures both pure sea level fluctuations which are variations in the water content and sea floor deformation which is due to tectonic activity. It is clear that the vertical ground motion can be obtained by subtracting these two data types, namely, altimetry and TG [Nerem and Mitchum, 2002;Garcia et al., 2007;Braitenberg et al., 2011]. Shum et al. [2002], used historic TG records and a decade of TOPEX/POSEIDON (T/P) and Geosat altimeter data in a stochastic adjustment to compute the estimation of vertical motion on the Great Lakes regions. In other paper, Mahmood Pirooznia et al. Fenoglio-Marc et al. [2004] calculated the rate of VCM in Mediterranean Sea, combining T/P data and a set of TG records based on sea level height changes. Another research was performed by Kuo [2004] in which he introduced a new method, namely Gauss-Markov stochastic adjustment model, and used it to estimate the vertical crustal motions (VCMs) in areas near to open seas, semi-enclosed seas and lakes. Kuo et al. [2008] combined geocentric sea level measurements from high-precision T/P satellite altimetry observations and TG records at 25 stations distributed in coastal regions of the Baltic Sea and around the Great Lakes. Ostanciaux et al. [2012] computed VCM rates by combining TG records and local SA and then compared this database to previous studies that use geodetic techniques and TG records in order to evaluate the consistency of their results and previous ones.
It should be pointed out that the quality of SA observations in coastal regions depends on the range between satellite orbit and a water surface, which is derived by measuring the time elapses between sending and receiving radar pulses from the satellite board and the surface of the Earth (the altimeter waveforms) [Fu and Cazenave, 2001].
However, near the coasts, altimeter waveforms may be corrupted due to variety of reasons, and the range observations are infected by some biases [Mathers et al., 2004]. To improve the precision of those altimeter observations with complex returning waveforms, a waveform re-tracking method is implemented to reprocess the original returning pulse. "Re-tracking" refers to the analyses of the waveforms, received in the satellite antenna as pulses to improve the range [Davis, 1995;Gomez-Enri et al., 2009]. Three common re-tracking algorithms such as Offset Center Of Gravity (OCOG), -Parameter and threshold can be used for this purpose [see Rapley et al., 1987;Martin et al., 1983;Davis, 1997;Deng, 2003;Lee, 2008;Zhang, 2009]. Based on the author's knowledge, the re-tracking correction has not been investigated in the studies of VCM using TG and SA observations and most of the papers ignored this correction. However, this effect may be very significant as it is later shown in this paper.
In this study, missions of T/P, Jason-1 and Jason-2 altimeters are used for VCM's determination at coastal parts of Caspian Sea (see Figure 1). These missions have 10-day repeat cycles. Since VCM is determined by difference between SA and TG observations, those passes of altimeter which cross near the TG station (adjacent to the coast) should be used. Due to the problems of SA in the shallow water of coastal regions, choosing a suitable SA points near the TG stations are very important. So in this study, two strategies are implemented (i) selecting the closest SA point to the TG stations on the condition that the SA data is not corrupted. (ii) Extrapolating SA points by Spatio-Temporal Kriging method to evaluate the altimeter observation at TG stations. In both cases, however, the re-tracking algorithms are used to improve the range observations near the coast. Tide effect should also be removed from SA and TG observations. In this study, to eliminate tide phenomena, it is modeled by considering significant tidal frequencies (empirical tide modeling), and we do not use global tide model such as FES2004 for tide correction.

Mahmood Pirooznia et al.
4 Due to sample rate of SA observations (approximately 10 days for T/P, Jason-1 and Jason-2), we encounter with aliasing phenomenon for tide modeling (see section 2.4 for more explanation about aliasing phenomenon). So To deal with it, aliasing frequencies are used for tide modeling. Eventually the rate of tide free sea level height time series of SA and TG are determined which indicate the absolute and relative changes in sea levels, respectively. Difference of the above mentioned two rates gives the VCMs. These VCMs in turn, have been compared with VCMs determined by continuous GPS stations.

Data source and pre-analysis of observations
In this section we concisely describe our data sources and some preprocessing strategies, including re-tracking of SA observations, selection of suitable SA point and tidal analysis, being used to prepare the two data sources for determination of VCM at Caspian Sea.

Data sources
Tables 1 and 2 show the features of SA and TG data which are used in this study. Sensor Geophysical Data Record (SGDR) of SA is used in this study due to containing along-track waveform information that may be used for applying re-tracking corrections for the range observations.
Tide gauge data in this study is provided from Ports and Maritime Organization (PMO) of Iran which were recorded by floating pressure tide gauges. This data is not available in international databases and no processing for quality control of which were applied.

Re-tracking of SA observations
The observable quantity in satellite altimetry is the range observation which is known as satellite range or satellite height above sea level. The range observation is contaminated by some source of errors which must be corrected via global or local modeling and can be summarized as follows: wet troposphere correction, dry troposphere correction, ionosphere correction, electromagnetic bias correction, ocean tide correction, inverse barometer correction, pole tide correction and center of gravity movement. They are better known as "geophysical corrections" [Benada, 1997;Pirooznia et al., 2016a;Soltanpour et al. 2017]. Also the re-tracking correction is used to correct range observation. The three common re-tracking methods, namely OCOG, -Parameter, and Threshold are implemented. These corrections should be applied on satellite range for each repeat cycle to create the time series of sea height [Khajeh et al., 2014;Khaki et al., 2014]. Thus, corrected SSHs are derived from the following equations. (1) where C is the speed of light (299,792,458 ), the pulse duration (3.125 ns for T/P), G 0 the nominal tracking gate (for T/P, Jason-1 and jason-2, it is 32.5, 31 and 31, respectively ), G the computed tracking gate (middle point of the leading edge which is computed from re-tracking algorithms), the re-tracking correction, h the altitude of the satellite from the reference ellipsoid, R the observed range, the geophysical corrections, and SSH is corrected sea surface height. As can be seen in computational analysis, the mean computed biases may be greater than 35 cm.

Selection of suitable SA points
Due to the well-known problems of SA technique near the coastal regions and shallow waters, choosing a suitable SA point to compare with TG station is of crucial importance. For this purpose, two strategies are considered in this study. For the first strategy, we select those SA ground tracks that pass very close to the TG station provided that the range observation is available on them (see Table 3). It should be mentioned that the closer the satellite ground tracks get to the TG stations, the more accurate would be the results. In 2002, Nerem and Mitchum [2002] showed that as the satellite tracks get closer to the coastal TG station, their time series get more similar.
Furthermore, increasing the number of cycles will result in the reduction of total error of modeling. Thus if there are more than one pass near the TG station, we consider all of them. For example, according to Table 3, for Anzali TG, three points are suitable points for comparison. This enables more reliable estimation of VCM.
Choosing the pass of SA close to the TG station as it is done in the first approach, means that the selected SA points are still some distances apart from TG stations. Moreover it means that no rigorous criteria is applied to determine the true distance. How much the two passes can be distance apart? Moreover, there is no SA pass that exactly crosses the location of TG station. To deal with this issue, an extrapolation method is used to predict the value of SA observation at the location of TG station in order to further improve the accuracy of VCM. In this approach three point-wise SA sea level time series are created. Then, the Spatio-Temporal ordinary Kriging is used for predicting point-wise sea level time series at TG stations. Ordinary kriging is a geo-statistical interpolation method, which was originally developed for spatial data by Krige (1952). In this study, SA measurements along the Caspian Sea, which are scattered in space and time, are used to predict the sea level at any time and location along the Caspian Sea. Kriging creates a statistically unbiased estimator that is optimal with a respect to the mean squared prediction error. The predictor is a weighted average of the observed values, in which the weights depend on spatial or Spatio-Temporal dependence of observed and predicted locations. The dependency is expressed by the covariance C ((x 1 ; t 1 ); (x 2 ; t 2 )) between the sea levels at two space-time points (x 1 ; t 1 ) and (x 2 ; t 2 ). The required covariance models reflect that the sea level, are estimated based on empirical covariance values between SA observations at various locations. It is obvious that the construction of an appropriate covariance function C ((x 1 ; t 1 ); (x 2 ; t 2 )) to predict the sea level is of particular importance for SA data along the Caspian Sea. For more details about Spatio-Temporal kriging and empirical covariance function see Boergens, et al. [2016]. Figure 2 shows a typical example of an empirical and modeled covariance functions for cycle 10 of T/P satellite pass 16. The modeled covariance function is used for prediction.  In table 4 the extrapolated SA points near each tide gauge station is shown. Figure 3 shows the position of TG stations, SA points, extrapolated SA points and available GPS stations in the southern coast of Caspian Sea. The GPS stations will be later used to determine the absolute vertical deformation at this region and in order to make a comparison with the deformation obtained by combination of SA and TG observations.

Tidal analysis
Finally, the tidal effects should be removed from both data sources. Since, global tide models are limited in the number of tidal components (frequencies), and for comparison of these two data sources, it is necessary that remove the similar tidal components from both observational sources; instead of using global tide model, local tide model 7 Vertical crustal motion at Caspian Sea  should be conducted in order to calculate the tidal effects precisely. In this case, tide modeling is done by the following equation [Hou and Vaníček, 1994;Okwuashi and Ndehedehe, 2017]: To determine the optimum number of tidal frequencies, a spectral analysis is conducted which reveals that tide gauge observations in our case include 40 significant frequencies.
It should be mentioned that the analysis of SA time series, suffers from aliasing phenomena. According to Shannon sampling theorem (William, 1977;Shannon, 1984), if is the highest present frequency in a signal, a sampling rate of at least = 2 (sample-second) would be adequate to reconstruct the whole signal in which, is the well-known Nyquist frequency. In other words, a signal can be fully reconstructed or recovered by sampling it at series of points which are 1/2 seconds apart. If there is any frequency content in the signal that is higher than this sampling frequency, it will be aliased to the lower frequency. In the case of sea tide, periods of SA observations (10 days) are longer than period of tidal components and thus the determination of tidal effects using this data is subjected to aliasing problem. To deal with this, aliasing period is defined by the following equation [Lindsley and Long, 2011]. (4) in which, is the frequencies of tidal constituent, is sampling period in days (orbit repeat period for satellite Jason2 is = 9.9156 ) and [•] is used for a Fix function 1 which rounds decimal numbers toward zero [Lindsley and Long, 2011]. the confidence intervals for the model parameters, and thus, they may strongly influence modeling and predictions Liu 1993a, 1993b]. Offset in the time series of tide gauge may occur during data acquisition process, due to some phenomena such as, boat hitting, tide gauge damage, earthquake motions and wrong recalibration of the instrument. Occurring of offsets in the observations can change the statistical parameter of signal and may affect on the trend computation [Gazeaux et al. 2013;Montillet et al. 2015;Lopez Bernal et al. 2016]. The first step in a dealing with these types of errors is modeling sea level time series. In this study, the functional model of the tide (Eq.) is used 1 The difference between Int and Fix is in negative numbers; for a number like -9.2 Int returns -10, whereas Fix returns -9.
2 2 ( -[ + 0.5]) = to model sea level. Before that modeling, the low pass filtering (Savitzky-Golay filter) is used to reduce the noises in sea level time series [Shekhar, 2016]. Then, the time series is constructed (modeled) using significant frequencies and the 3-sigma threshold method is used to eliminate the possible outliers in the observations. Subsequently, for offset detection, noise assessment algorithm (least square variance component estimation) is used to find the effective noise component in the time series, and using the cofactor matrix of the noise as the weight matrix of observations, the observations are adjusted and the signal is processed to find the possible times of offsets in order to remove their corresponding observations from the time series. For more details see Pirooznia, et al. [2019].
For modeling sea level time series, the significant frequencies should be found. For this purpose, spectral analysis is performed on the sea level time series. To this end, angular frequencies are created using the period of observations ( = 2 / ). Assume that ₁, ₂,..., is the set of real and positive angular frequencies of the observation vector that may be equal to tidal angular frequencies, or they can be constructed by period of observations. In this study, in order to find effective frequencies, the angular frequencies are determined based on the period of observations. To check the effective frequencies, the significant test is employed [Amiri-Simkooei and Asgari, 2011] and the effective frequencies which pass this test are selected. The result of spectral analysis on SA and TG of Caspian Sea shows that the annual and semiannual components have the highest effect on time series (see Figures 4 and 5). As mentioned previously, spectral analysis of SA time series due to sampling rate of observations faces with the aliasing phenomenon which creates the aliasing frequency. As an example, for the component of half daily S2, the main frequency is 2 cycles per day, whose aliased frequency is 0.01702 cycles per day corresponds to period of 58.74 days (see Figure 4). Figure 4 shows a typical power spectrum for SA time series of Anzali-crossover point, alias frequency of S2 component, annual  The tables which include amplitudes and phases of tidal components (both SA and TG data stations) are added in the appendix A. from SA and TG time series ordinates (see Figures 6-8), it is evident that the Caspian Sea level has been decreasing during the last decades. According to Arpe et al. [2012], the Volga River as the main source of Caspian Sea water supply, has brought less water to the sea.

Numerical results and discussion of VCM
In this section, we present the numerical results of VCM at the coastal area of Caspian Sea. To do so, the SA and TG observations that are processed based on the steps of previous section are subtracted from each other.
As mentioned previously, to improve the accuracy of SA time series, the re-tracking correction is implemented to range observation. Table 5 shows the correlation coefficient between the sea level that is resulted by Anzali, Noshahr and Neka tide gauge observations and SA observations (Anzali-16 point, Noshahr-16 point and Neka-107 point), without applying any re-tracking and while the re-tracking is performed. In this region the threshold re-tracking method, by 35% as threshold value, presents the highest correlation. That means, by using this re-tracking method the most accurate measurement of sea level will be achieved. It should be emphasized that the vertical datum of SA and TG observations are different; thus, to compute the correlation between SA and TG time series, after normalizing the time series of observations, the correlations are computed at the times when SA and TG observations overlap. Also due to the well-known problems of SA technique near the coastal regions and shallow waters, choosing a suitable SA point to compare with TG station is of crucial importance. For this purpose, two strategies are considered in this study.

11
Vertical crustal motion at Caspian Sea   Figure 9 shows Mean value of VCM at TG stations.
Analyzing GPS time series at Geodynamic stations is an alternative approach to obtain VCM's. This analysis supplies the auxiliary source to be compared with the method of this study. In this study, four geodynamic stations are used, namely, MABD, NKAD, RSHT and TKBN. Table 7 shows the approximate positions of GPS stations, time span of their observations and their distances to closest TG stations.
These GPS time series were provided by national cartography center (NCC) of Iran. All processing steps were done in NCC. The processing of GPS raw data was done by GAMIT software, also it was integrated using GLOBK software [Herring et al., 2015]. According to recent report in NCC (www.ncc.org.ir), the first step in processing this data, is controlling the quality of rinex data. This operation is done with TEQC software. Then data processing on a daily basis is done using double differences method. Finally, the precise orbital data (SP3 files from International GNSS service (IGS)) with a delay of two weeks, mapping function equations for troposphere modeling, ocean and solid tide modeling and solar radiation model are applied. Moreover, the IGS stations are referenced with a respect to ITRF 2014 reference frame to minimize the calculation errors and increase the accuracy. Figure 10 shows vertical GPS time series at geodynamic stations.
12 Figure 9. Mean VCM estimated from SA-TG. In order to compare VCMs from GPS with those from combination of SA and TG, it is necessary that GPS stations are located near to TG stations. Even though these GPS stations are not located sufficiently close to our TG stations (see table 7), both of them are positioned on the same faults (see Figure 11).

13
Vertical crustal motion at Caspian Sea   The listed locations for TG and geodynamic stations shows that there is a notable distance between them. Due to long distances (few Kilometers) between geodynamic (GPS) and TG stations, comparing their results may look unsound. However, from tectonic point of view, these stations are situated on a common fault and their comparison would be meaningful.
The resulted VCM's from previous section are compared with the crustal motions deduced from GPS time series at geodynamic stations. The comparison shows an acceptable consistency (Table.8). Table 8 indicates that there is a good consistency between (SA-TG) and GPS vertical rates. The mean value of the trend differences between the closest GPS and the SA-TG vertical rates is 0.82, 0.21 and 2.71 mm/year for Anzali, Noshahr and Neka TG stations, respectively. At all stations, the rates of VCM are negative which indicates the subsidence phenomenon. Various factors may be responsible to this subsidence: tectonic changes due to existing faults, irregular groundwater extraction and etc. [Pacheco-Martínez et al., 2015;Carreón-Freyre et al., 2016]. According to the studies conducted by the Iranian ministry of energy (http://www.moe.gov.ir) using the data of piezometric wells, the rate of groundwater extraction has significantly increased in the recent years.
This conclusion can be used as a confirmation to the negative rates of VCM in these areas. For instance, the highest rate of vertical change at MABD station might have originated from the extravagant amount of groundwater extraction which caused the crust to subside. To compute the correlations between piezometric wells data and GPS time series, after normalizing the time series, the correlations are computed at the times when piezometric wells data and GPS observations overlap. Table 9 shows the computed correlations.

Mahmood Pirooznia et al.
14  Figure 12 shows the groundwater level time series near Anzali, Noshahr and Neka TG stations. The rates of groundwater level time series are negative which may be the sign of groundwater extraction

Conclusion and recommendation
In this paper, the pattern of vertical deformation in coastal parts of Caspian Sea is determined using a combination of satellite altimetry observations and tide gauge measurements. Satellite altimetry measures the absolute sea level, while tide gauge data computes the relative sea level. Thus the difference between these two observations can be used to determine the vertical motion at ground station. The result show that the sea level of Caspian Sea from 1992 to 2014 has had a decreasing trend. Moreover, the VCMs at the desired stations are negative that may indicate land subsidence. Typically, the numerical results illustrate that by considering SA crossover points, the vertical crustal motion at Anzali, Noshahr and Neka are -1.56, -1.28 and -5.01 mm/year, respectively. The innovations of this study are as follows: using re-tracking correction for SA data neat coastal TG stations that increases the correlation between SA and TG sea level time series, application of Spatio-Temporal kriging method by creating the empirical covariance functions to interpolate SA signal at the TG stations and implementing local and more accurate tide modeling instead of using global tide model such as FES2004. The future issues that may significantly improve the quality of the results are reported below The precision of orbit determination for satellite altimetry is 1-2cm. This leads to a precision of about 4-5cm in sea surface height. An increase in precision of orbit determination will result in a more accurate sea surface height data.

15
Vertical crustal motion at Caspian Sea  Table 9. The correlation coefficient between ground water level data of three wells near TG stations and GPS. TG stations should gather data in a unified vertical datum. This unification is of fundamental importance and simplifies the calculation process. In some researches the Mean Sea Level (MSL) which is determined from SA and TG data is used as an alternative for estimation of VCM, in this case there is no need to calculate the rate of change and VCM is calculated by subtraction of MSL from SA and TG data. This needs all TG stations to be geo-referenced.
TG stations should be calibrated precisely. Furthermore, TG observations at nearby stations, should be given appropriate weightings and combined with each other in order to increase the accuracy. To integrate the elevations of different TG's, stations must be fixed to a local benchmark; this benchmark then must be transferred to a national and finally an international datum. This procedure requires a modern geodetic technique.
Various tectonic factors involved in vertical deformation and land subsidence due to groundwater extraction may be considered as causes of VCM at coastal regions of Caspian Sea.
One of the drawback of this study is the lack of TG data for a long time. With long time data, it is possible to measure trends more accurately, and provide robust estimation of the trend.