Measurement: Ruijie Xi, Xiaohui Zhou, Weiping Jiang, Qusen Chen

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Measurement 122 (2018) 247–256

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

Simultaneous estimation of dam displacements and reservoir level variation T


from GPS measurements

Ruijie Xia,b, Xiaohui Zhoua, , Weiping Jianga,c, Qusen Chena,b
a
School of Geodesy and Geomatics, Wuhan University, 129 Luoyu Road, Wuhan 430079, Hubei, China
b
Nottingham Geospatial Institute, The University of Nottingham, Nottingham NG7 2TU, United Kingdom
c
GNSS Research Center, Wuhan University, 129 Luoyu Road, Wuhan 430079, Hubei, China

A R T I C LE I N FO A B S T R A C T

Keywords: GPS has been widely applied to monitor displacements using direct measurements. Even conventional ways
Dam safety monitoring using direct measurement can achieve millimeter level (better than 1 mm in horizontal and 2 mm in vertical
GPS component) accuracy displacement, the causative loads such as hydraulic thermal loads always have to be
Multipath effect obtained by other sensors or methods. This paper proposes a novel method to derive reservoir level variation
SNR
data with displacement estimation simultaneously for the mechanism explanation of deformations. This method
Reservoir water level variation
is based on multipath theory for specular reflections and the use of Signal-to-Noise Ratio (SNR) data, rather than
direct measurements in GPS. We took data from an existing GPS monitoring system which has been set up on the
Xilongchi Dam in Shanxi Province, China. A whole year of data from the sites located beside the reservoir was
used to conduct the experiments. The result indicates that the water level variation daily time series estimated
with SNR measurements can be achieved to centimeter level. The correlation between the estimated value and
the daily records can even reach up to 0.98. It can be concluded that the GPS-based system designed by this
paper can monitor displacements and reservoir level synchronously, and GPS is also confirmed to be a viable tool
for both the monitoring of dam displacements and a reservoir water level indicator.

1. Introduction GPS [5,6]. Conventional SHM procedures are usually labor-cost heavy
and inaccessible to critical parts of the structure under surveillance [6],
Concrete dams play a key role in supporting economies. Health resulting in insufficient frequency inspection; the accuracy of the total
monitoring of concrete dams, therefore, has become a topic of great station and leveling significantly depends on the weather condition and
importance and involves monitoring the dynamic behavior of large baseline length; TSL limits its application in large scale dam deforma-
dams. The performance of these structures under environmental and tion monitoring; the photogrammetry method suffers from asynchrony
operational loads may decrease over time mainly due to seasonal of monitoring points; Ground-based InSAR aggravates the financial
changes and fluctuations of the reservoir level. However, the response burden to the establishment of monitoring system.
under these loads is still poorly understood at present. Normally the Up to date, the displacements and reservoir level data have always
temperature data is easy to obtain from thermometers, therefore the been provided separately by structural health monitoring (SHM) sen-
acquisition of water level data and deformations occurring in dams will sors and specific water level indictors such as a graduated pole and
be important for interpreting the deformation mechanism of dams. pneumatic water gauge [1]. Even still, the mechanism for interpreting
Health monitoring of concrete dams involves monitoring the static can be carried out with the displacements and reservoir level data from
behavior of large dams through both displacement sites and reservoir different sensors [7]. The identical data sources could be self-consistent,
level variations. This is implemented through the measurements from synchronous and cost-effective. To the best of the authors’ knowledge,
various sensors, from conventional structural health monitoring (SHM) it is still rarely researched.
platforms [1–5], such as traditional visual inspections [1], the Ground- In this study, a GPS-based dam deformation monitoring system was
based Interferometric synthetic aperture radar (InSAR) method [2], designed to provide displacements and reservoir level data synchro-
terrestrial laser scanning (TLS) technology [3,4], photogrammetry nously. Compared to the above-mentioned methods, continuous GPS
methods and so on, to satellite-based geodetic measurements including measurements can overcome almost all the shortcomings of other


Corresponding author.
E-mail address: xhzhou@whu.edu.cn (X. Zhou).

https://doi.org/10.1016/j.measurement.2018.03.036
Received 10 August 2017; Received in revised form 3 March 2018; Accepted 14 March 2018
Available online 15 March 2018
0263-2241/ © 2018 Elsevier Ltd. All rights reserved.
R. Xi et al. Measurement 122 (2018) 247–256

technologies due to their merits of all-climate working, high automa-


tion, high accuracy, real-time processing and low-cost effectiveness [6].
Later studies have proved its feasibility and performance for effective
and timely monitoring of engineered structures [8–12]. However,
compared to the previous GPS dam deformation monitoring experi-
ments [8–10], the system with GPS Relative Positioning method has
been operating for more than 6 years automatically to provide long-
term precious displacements data and the accuracy is much higher than
the Precise Point Positioning (PPP) method [11,12]. More importantly,
a novel method for reservoir level estimation is within this dam mon-
itoring system to realize the displacements and reservoir level estima-
tion with only one sensor. This method is based on multipath theory for
specular reflections and the use of Signal-to-Noise Ratio (SNR) data,
rather than direct measurements in GPS. In the past decades, the re-
flected GPS signals caused by multipath effect were widely used for
altimetry, ocean winds, and soil moisture etc. [13–21]. Among these
studies, Larson et al. [14] confirmed the validity of reflected signals of
GPS to be a tide gauge in the sea level variation determination with a
geodetic-quality GPS receiver. However, the stations are normally set at
places which are free from obstructions. For engineering application
purposes, the observation environment of the site locations that are
selected is not always ideal compared to that of the coastal sea level
retrieval. The obstructions around the antenna could make it difficult to
separate the signal reflected by the water surface and to obtain the
water level accordingly. Thus, more consideration should be taken into
the reservoir level estimation, such as the satellite and the signal se-
lection.
In this study, a near real-time GPS-based deformation monitoring
system which served for the Xilongchi Dam is introduced first. During
the data processing schemes, the accuracy assessment of the system and
the displacements time series observed will be given. After that, the
reservoir level retrial method and the detailed technical GPS issues will Fig. 1. Google Earth image for the upper reservoir of Xilongchi pumped-storage power
be presented. With a whole year SNR data, the effectiveness of this station. The layout of monitoring and reference stations, and Dam Coordinate System
method is assessed and the relationship between the dam displacement (DCS) are shown in (a). (b) and (c) show the photograph of S071 and TN01.
and the water level is also primarily analyzed.
site are recorded at 5-s interval and transferred through optical fiber to
2. GPS deformation monitoring system of Xilongchi dam the server 24 h a day continuously. Then, in the server end, the data
processing software computes the position every 4 h automatically and
2.1. Data description and GPS deformation monitoring system saves the results into the database simultaneously.

Xilongchi Dam is a pumped storage power station which provides


power during peak times of electricity usage. It is a medium size 2.2. Data processing strategy
(401.16 m crest length and 50 m high) dam located about 90 km north
of Taiyuan, Shanxi province, China. The reservoir was formed by ex- In the monitoring system, the high precision GPS data processing
cavation on an up-hill land with a height of 1460 m and was put into method plays a key role in acquiring reliable displacements. Therefore,
service in 2011. The capacity of the dam is about 4.94 million m3 and specific data processing models and strategies shown in Table 1 were
the reservoir working depth is 25.5 m. The aim of this reservoir is to selected for short baselines to satisfy the high-accuracy and near real-
pump water and draw off water to provide electricity. In the period of time requirement of the monitoring system. A home-made dam
the water level rising and falling, the water pressure on the dam could
be the dominant factor of deformations. Therefore, the displacements Table 1
The data processing models and strategies for GPS deformation monitoring.
and reservoir level variation are both important for the deformation
mechanisms explanation. Item Models & Constraints
To ensure the structural safety, a GPS-based deformation mon-
itoring system was built and completed in June 2009. The distribution Observations Original L1 carrier phase observations
Elevation angle cutoff 10°
of reference and monitoring points of the geodetic monitoring scheme is
Sampling rate 5s
shown in Fig. 1. There are five monitoring stations around the reservoir, Satellite obit Broadcast ephemeris
two of them are located on the dams (L022 and L132). The others are Satellite clock biases Double-difference elimination
set directly beside the reservoir (S171, S071 and S191). Two reference Receiver clock biases Double-difference elimination
stations were constructed on the north-east corner (TN02) and south- Satellite Antenna Phase Centre Offset Default values from manufacturer
(PCO)
east corner (TN01) of the reservoir at stable points. All the baseline Satellite Antenna Phase Centre GPS Model
lengths between the arbitrary monitoring station and the reference variation (PCV)
station are below 800 m. At each station, a TRIMBLE NETRS receiver Receiver Antenna PCO and PCV Not applied
and a CHOKE RING antenna (TRM29659.00) are installed similarly to Troposphere Delay Double-difference mitigation
Ionosphere Delay Double-difference mitigation for the first
the stations served for the International GPS Service (IGS). The pho-
order effect
tographs of the monitoring station S071 and the reference station TN01 Ambiguity Resolution Bootstrap and Decision-function
can be seen in Fig. 1(b and c). The real-time GPS observations on each

248
R. Xi et al. Measurement 122 (2018) 247–256

monitoring specific software named DDMS (Dam Deformation Mon-


itoring System) [22] was also developed.
In GPS data processing, double-difference [23,24] observation is
ideal for the short baselines. Thus, the satellite-dependent terms (sa-
tellite clock offsets and carrier phase fractional biases) and the propa-
gation medium-dependent terms (tropospheric and ionospheric delays)
could be neglected. According to Jiang et al. [25], satellite orbit will not
degrade those baselines less than 1 km. The accuracy of the broadcast
ephemeris is good enough for deformation monitoring. During the data
processing, raw L1 carrier phase observations are utilized directly to
estimate all parameters due to its low noise and can provide a result as
good as dual-frequency observations.
Ambiguity resolution is another key issue in the high-precision GPS
data processing. In this software, ambiguity resolution follows a se-
quential strategy, which is also called “Bootstrap and decision-func-
tion”. The detailed method of “Bootstrap and decision-function” can be
found from Dong et al. [23]. Where possible, ambiguities were fixed to
integers (routinely close to 100%).

2.3. Monitoring precision assessment

Jiang et al. [22] has tested the precision of this deformation mon- Fig. 2. Displacement time series of 4-h session after removal of mean value from Jan. 1st
itoring system whose precision can be achieved as 0.8 mm, 0.5 mm and 2014 to Jan. 20th 2014 (totally 120 sessions, with 6 solutions a day and for 20 days). The
1.4 mm for north-south, east-west, and vertical component, respec- error bar is the standard error for every session, and the repeatability is also shown in the
tively, for a 4-h session. In this section, we will also give an assessment figure.

for the GPS positioning precision. Normally, in GPS data processing


precision assessment, the repeatability of the session solution is esti- time series, the displacement time series (N, E, U) of each of the
mated following Eq. (1) in each case in terms of north-south, east-west baselines considered are shown in Fig. 3.
and vertical components [24]. Broadly speaking, apparent signals can be easily noticed in Fig. 3.
1 First, steps are apparent at the beginning in the time series of sites S191,
n n (Xi − X )2 2 n Xi
⎡ n − 1 ∑i = 1 ⎤ ∑i = 1 L132 and S071 which coincide with events in the site logs, that is, the
σi2 σi2
R=⎢ n 1
⎥ , X =
n 1 changing of the antennas. Unusual dilatations in north or vertical
⎢ ∑i = 1 ⎥ ∑i = 1
⎢ σi2 ⎥ σi2
(1) components of all stations appear in late 2009. From the weather data,
⎣ ⎦
we knew that it was the result of the heavy snow covering the antenna.
where n is the number of sessions processed; Xi is the solution estimated In S171, the displacement amplitude can reach to 8–10 mm. All the
for session i ; σi is the standard error of solution estimated coordinate gaps in the time series are due to the equipment failure. Secondly,
component for session i . X is the weighted average value of Xi , as periodic signals can be clearly observed in all the time series. The east
shown in Eq. (1). component of site S071 shows a ∼4 mm annual stepped variation, ra-
TN01 and TN02 are the reference stations, whose observation en- ther than seasonal motion. King and Williams [26] also found that a
vironment is ideal. The baseline they formed was selected to do the short baseline has 10 mm downward deviations during the local winter
precision assessment. In this study, the 4-h session coordinated solu- months. This phenomenon needs more analysis. Thirdly, obvious trend
tions from Jan. 1st 2014 to Jan. 20th 2014 was applied to calculate the signals can be found in L022 and L132, which are located directly on
repeatability. Fig. 2 shows the displacement time series and the re- the dams. Site L022 moves gradually to the north-east direction, and
peatability values. As shown, the repeatability for north–south, east–- both of them show a subsidence in the vertical component. Finally, time
west and vertical components are 0.6 mm, 0.5 mm and 1.5 mm re- series of TN01 show less scatters than other stations since it is built on a
spectively, which is the same as Jiang et al. [22]. Therefore, we can stable location.
conclude that the precision of this monitoring system is better than Evidently, the signals in the deformation time series are now rou-
1 mm in horizontal and better than 2 mm in vertical for a 4-h data tinely used to infer changes in the dam’s shape at secular and seasonal
processing session. timescales, such as temperature and hydrological loading. We have not
yet begun to analyze the source of these signals, but the temperature
2.4. Coordinate time series and water level data are important for the mechanism of the de-
formation. Section 3 will demonstrate the way to obtain water level
In the data processing, TN02 was specially chosen as the reference variation data from GPS SNR observations.
site and its coordinates held fixed. Another reference station, TN01, is
processed as a monitoring station. Applying the data processing stra-
tegies mentioned in Section 2.2, we processed the GPS data for 6 and a 3. Water level data acquisition from SNR observations
half years every 4 h. Generally, the information of the dam’s deforma-
tion is expressed in an independent coordinate system, such as the Dam 3.1. Methodology
Coordinate System (DCS). In reality, the orientation of this reservoir
was aligned to the World Geodetic System 84 (WGS84) coordinate In the previous research, Larson et al. [14] confirmed the capability
system, which is consistent with the coordinates of monitoring stations of the GPS tide gauge by using a geodetic-quality receiver to obtain the
obtained directly from data processing procedure. Therefore, it is rea- coastal sea level. Similarly, the method behind the GPS tide gauge can
sonable for us to take coordinates in WGS84 to illuminate the dam be used in the water level reflection of a reservoir. The basic theory is
status. Therefore, the Dam Coordinate System (DCS) shown in Fig. 1 is that the interference between the direct and reflected signals produces a
used in the following analysis. After eliminating the gross errors by QIR characteristic pattern in the SNR data. The pattern depends on the
criterion (3σ rejection criterion) and removing the mean value from height of the GPS antenna above the reflecting surface [14,20]. Fig. 4

249
R. Xi et al. Measurement 122 (2018) 247–256

Fig. 3. Displacement time series of monitoring stations (from 2009 to 2016).

gives a geometry diagram of the water level reflection. designed to suppress reflections (the reflected wave in Fig. 4) and the
The height of the water level Hw can be calculated by choke-ring GPS antenna often equipped, the reflections cannot be
completely suppressed, especially in terms of snow, water and soil en-
Hw = HG−h (2)
vironments [27–30]. These direct and indirect signals are mixed in the
where HG is the height of the GPS antenna. h denotes the vertical dis- antenna, producing constructive and destructive interference oscilla-
tance between the GPS antenna phase center and the water surface. In tions that are observable in the signal-to-noise ratio (SNR) measure-
the previous GPS positioning stage, HG can be determined by GPS data ments (Fig. 4). SNR is the carrier-to-noise power-density ratio to express
processing with millimeter level accuracy and the value will be stable in the GPS signal power [31,32]. It is an observable commonly recorded
a short time. Then, the next mission is to obtain h from GPS SNR by GPS receivers and reported in the RINEX format [33,34] which is
measurements. sensitive to carrier phase multipath. By using SNR, we can better un-
Although the gain pattern of the geodetic-quality GPS receiver is derstand the probable location of reflecting objects and severity of

250
R. Xi et al. Measurement 122 (2018) 247–256

SNR Data

50

40

30 observed direct signal

10 15 20 25 30 35 40 45
SNR Data with Direct Signal Removed
10

-5

-10
10 15 20 25 30 35 40 45
Fig. 4. Direct and indirect signals mixed in the antenna and the geometry of the water Satellite Elevation Angle (degrees)
level reflection. h is the vertical distance between the GPS antenna and the water surface.
HG and Hw denote the height of the GPS antenna and the water level to a same horizontal Fig. 6. L1 SNR data for one GPS satellite is shown in black. The direct signal is fitted by a
plane. first-order polynomial in blue (top); SNR data with the direct signal removed was shown
in bottom.

cosine function [14,16]:

SNR = A·cos(2π·f ·t + ϕ0)


Receiver and Antenna type

SNR observation index


= A (θ)·cos 2π· ( 2h
λ
·sin(θ) + ϕ0 )
in the file Header
≅ A·cos ( 2h
2π· λ ·sin(θ) + ϕ0 ) (3)

where λ is the GPS carrier wavelength, A is the amplitude, θ is the


Satellite list
satellites elevation angles and ϕ0 is the phase offset. Then, in case of
horizontal planar reflectors, the multipath signal frequency is constant
T:\scanning\Elsevier\Journal\MEASUR\5355\ for sine of the satellite elevation angle and changing with different h , as
in Eq. (4). The water level variations will drive the apparent height h
SNR for L1 and L2 of G15 increasing or decreasing to H, and the dominant frequency of SNR time
SNR for L1 and L2 of G21 series will change accordingly. The exact relationship between h and
SNR for L1 and L2 of G14 the frequency is shown in Eq. (5), if the L1 (the wavelength is
∼0.1902 m) SNR observation is used. In the following terms, all the
SNR for L1 and L2 of G22
spectral analysis for the frequency term can be transferred to the re-
SNR for L1 and L2 of G18 flector height meters as shown in Eq. (5). However, it should be no-
SNR for L1 and L2 of G24 ticed, the absolute vertical distance h between the antenna phase center
and the water surface may have a stable topographic bias [19]. Thus,
SNR for L1 and L2 of G12
the relative variance of the water level can be obtained with certain
accuracy.
Fig. 5. SNR observations in the RINEX file. 2h
⎧λ =f
⎨ sin(θ) = t (4)
multipath errors resulting from these reflections [31,32]. Fig. 5 shows a ⎩
standard GPS RINEX file and the S1/S2 observables indicated for the
L1/L2 phase observations. In the following term, we predominantly λ 0.1902
h= ·f = ·f = 0.0951·f
discuss how to use the SNR time series extracted from RINEX file to 2 2 (5)
reflect the water level variations.
Fig. 6(top) shows a pattern of SNR time series with the elevation The dominant frequency can be estimated by spectral analysis to
variations. Signal (shown in Fig. 6) rises in strength by 10 dB-Hz as the determine the effective reflector height. This is the same method of
elevation angle of the satellite rises from 10° to 45°. This rise is pri- deriving the snow depth and sea surface variations by SNR measure-
marily due to a geodetic antenna’s gain pattern. Generally, the rise ments [13,14,16,19]. Fig. 7 shows an example of the variation of water
trend is recognized as a direct signal which can be fitted by a first-order surface changing the frequency of the reflecting SNR signal. We chose
polynomial and removed. The residual SNR signal is regarded as the the SNR data from PRN 3 in days of year 10, 27 and 32 in 2013. From
reflected signal (Fig. 6(bottom)). A quasi-sinusoidal pattern is illu- the left panel of Fig. 7, these three series show obvious and distinctive
strated clearly in Fig. 6(bottom) especially when the elevation angles frequencies. Then the Lomb Scargle Periodogram (LSP) [35,36] was
are low (between 10° and 20°). used to estimate the frequency f and then convert it to h as Eq. (5). As is
In general, the dominant frequency is proportional to the vertical shown, there is only one dominant frequency and this frequency is
distance between the GPS antenna phase center and the underlying much higher than other frequencies in all three days. Therefore, the
reflected surface h. For detrended SNR, signals can be expressed as a vertical distances between the GPS antenna and water surface in the
three days were approximately 10.2 m, 12.9 m and 17.5 m respectively.

251
R. Xi et al. Measurement 122 (2018) 247–256

Fig. 7. Left panel represents SNR data for representative tracks at site s191 in days of year 10, 27 and 32 (2013). Lomb-Scargle periodograms computed for the SNR data are shown in the
right panel. The corresponding reflector heights are also shown.

3.2. Technical GPS issues value. However, fortunately, the water level in a reservoir has its
own upper and lower bounds, which means the extremely high and
Severe variation of water level occurs when the electricity demands low peaks out of the designed scale of the water level may not be the
are rising, leading to the displacements of the dam. Therefore, it is results of the reflected signal from the water surface. According to
critical to monitor the exact water level with the monitoring system. the designed range of water level variation and a number of tests,
Previous studies have demonstrated that the quality of L1 and L2 we concluded that the LSP peaks (Reflector height h ) shown lower
SNR data is sufficient for GPS reflection. However, considering the poor than 9 m and higher than 25 m can be ignored. Meanwhile, in the
observation environment in deformation monitoring and the demand of height between 9 and 25 m area, if the dominant one does not ex-
obtaining a reliable water level, we must assess the GPS SNR data ceed 2 times the rest of peaks, the results will also be discarded.
quality and choose the appropriate satellites. This section mainly de- Fig. 10 displays the LSP of two SNR time series. It can be seen that
monstrates the issues we should consider regarding the GPS retrieval. both of them show many peaks under 9 m, which can be recognized
as the background noise. However, the results from G03 only have
1. Generally, the L2 code is encrypted with a much lower power level one dominant peak and G16 has two dominant peaks over 9 m.
than the equivalent L1 [14,17]. Therefore, our research is based Multiple peaks will make it difficult to identify the correct retrieval
entirely on L1 SNR data. value. Therefore, the result from G16 will be discarded for the
2. As displayed on the bottom panel in Fig. 6, the data whose elevation purpose of high accuracy.
angles are between 10° and 20° shows an apparent distinctive fre-
quency. Meanwhile, as shown in Fig. 8, the detrended L1 SNR data 3.3. Water level retrieval and results
of seven satellites show a significant consistency. Therefore, only
data whose elevation angles are between 10° and 20° will be applied To verify the feasibility of the method, we compared the retrieved
to estimate the multipath frequency. This elevation angle generally water level with the readings from a graduated scale. The retrieved
spans 30 min for each satellite. water level was obtained through L1 SNR data of the whole of 2013 at
3. Choosing the best SNR data from appropriate satellites is important three sites (S191, S171 and S071), and the scale readings are obtained
for GPS retrieval. In this study, only the SNR data reflected from the at 8o’clock in the morning each day. Since the retrieved values are
reservoir water surface can accurately provide the height of the session-wised, the average values are obtained through the division of
water level. Before the GPS retrieval, the “sky plot” can be used to the estimated value by 8 for comparison.
primarily analyze the condition of GPS satellites above the parti- As previously mentioned, the displacement data is resolved at a 4-h
cular site. Fig. 9 identifies the tracks of available satellites that may session. However, under some circumstances, 4 h may not be enough to
provide reflected SNR signals in site S191 and site S071. The two red capture a satellite whose elevation is between 10° and 20°. Therefore,
cycles represent an elevation angle of 10° and 20° respectively. 24 h were used as a session to acquire the daily average water level
Seemingly only the satellites on the east side of S191 and on the data. Moreover, more redundant satellites are available to improve the
west side of S071 could have the footprint located on the surface of precision at the water’s quiet period with a 24-h session.
the reservoir according to the location of the two sites (in Fig. 1). It Fig. 11 shows the estimated value from GPS and the daily records
can be seen that, from Figs. 9 and 11 satellites are available to measured. We can see that the water level is different from month to
provide reflection signal for S191. Only 7 satellites can be used in month due to the different requirements of power. In January, March,
S071 due to the site location. However, it is sufficient for GPS re- April, September and October, most of the time, the water level is
trieval in this site. During the experiments, tracks without 300 stable. However, during summer and winter, the water level varies
points will be discarded. dramatically.
4. One thing which should be mentioned is that, normally, a large Site S191 is located in the west side of the reservoir. No obvious
background noise can exist in the SNR time series because of the obstacles exist around the site and most of the multipath effect would
obstacles around the site. In this case, the peaks of the LSP will be come from the water surface. Therefore, the GPS results clearly follow
aliased by other frequencies, which will cause a wrong retrieval the general signature of the water level variations over the whole year.
During the stable period, the estimated results agree well with the daily
records. In summer and winter, although the GPS results cannot com-
pletely follow the daily records, they can follow the fluctuation of the
water level. We calculated the correlation between the water level re-
cords and the estimated values, with the correlation coefficient 0.96.
Fig. 11 also shows the GPS retrieval results of sites S171 and S071.
Since both sites are located in the corner of the reservoir, the ob-
servation environments may not be as good as that of S191. However,
both sites can still reflect the water level variations and most of values
Fig. 8. Observed L1 SNR data for seven satellites, whose elevation angles are between 10° in the stable stage show a good agreement with the daily records, with a
and 20°.
few values showing a larger discrepancy. The apparent discrepancy

252
R. Xi et al. Measurement 122 (2018) 247–256

Fig. 9. Sky plot of S191 and S071. The figure only shows traces of the satellites that may provide available data for GPS reflection. As to in situ environment, water surface is located on
the east side of S191 and on the west side of S071. Site S071 is in the east-south corner of the reservoir, which may lead to fewer satellites available for GPS retrieval. This can be
confirmed by Fig. 1.

that day.
It should be noted that even the daily records might not accurately
represent the variation of the water level during the summer months
since there is only one reading for each day. Hence, we can consider
that, the uncertainty in the water level could be equally as big in the
value obtained from the GPS data. However, the GPS SNR measure-
ments could have a higher temporal resolution.
Quasi-static monitoring of dams involves measurement of static
Fig. 10. LSP of data of G03 and G16. factors such as ambient temperatures, reservoir level, opening and
closing of joints, displacements and strains which are measured accu-
may be a result of the reflected signals from obstructions around the rately by instruments [1,7,37]. Therefore, the water level data provided
sites. The correlation coefficients in both sites are 0.98 and 0.93 re- by GPS SNR observations can be used to explain the interpretation of
spectively. The reason for better correlation coefficients of S171 is that the causative relationship between the hydraulic load and the dam
the water retrieval during summer agrees with the exacted water deformation in structural health monitoring (SHM). In this section, we
fluctuation much better than sites S191 and S071. briefly analysed whether the reservoir level fluctuations are responsible
To give the accuracy assessment of this retrieval, we selected the for the dam deformations.
data for three days when the water was in a quiet condition. All the Fig. 14 shows the whole-year deformation series of site S191 and
satellites LSP results are demonstrated in Fig. 12, and corresponding S071 for the 4-h session from 8 to 12o’clock at the local time. The water
dominant peaks are listed in Table 2. We can see that all the satellites level records measured are also shown in the figure. As mentioned
show a similar dominant peak around 14 m and other peaks are much above, the gross errors have been rejected from the deformation series.
less than the dominant one. From the mean values, and comparing them Thus, gaps may exist in the time series.
with the measured records of the three days, we can notice that the In Fig. 14, a linear correlation between the dam displacements and
accuracy of the daily mean value can be achieved to centimeter level. the reservoir level could be found, especially in the vertical component.
For site S071, the displacement shows a dramatic dynamic response in
the east-west direction during summer when the water level variations
4. Discussion were significant. As site S071 is on the east side of reservoir, the water
level variation could therefore cause the deformation in the east-west
From Fig. 11, we figure out that the estimated water level shows a direction. In other directions, there are similar signals that could be
good agreement with the exact record during the stable stage. However, noticed among sites’ displacements and the reservoir level. However,
only a similar trend can be observed during summer and winter. This is more analysis is needed for the comprehensive deformation mechanism
because the daily records only provide one value in a single day (gen- revealing.
erally at 8o’clock in the morning of the local time) and the estimated
value is computed during the time when the sites can track satellites 5. Conclusions
whose elevation angels are between 10° and 20°. In reality, water level
variation changes within a day. Hence, the estimated value from every In this contribution, we designed and developed a GPS-based dam
satellite will show a different value with the daily records and other deformation monitoring system, which can provide safety displace-
satellites. Fig. 13 delivers an example. The blue line represents the ments and reservoir level variation synchronously. Based on the system,
trend of h in Fig. 4 on June 14, 2013, and the five points with detrended the displacements for 6.5 years were observed and the reservoir level
SNR series are from the satellites with PRN number 1, 3, 10, 11 and 26. data for a whole year was estimated. From the research presented, our
The LSPs for the data are also shown in Fig. 13. As shown, the five conclusions are summarized as follows:
satellites illustrate an increasing h in the reservoir water level with the
time, which indicates the height of the water level was decreasing on – The displacement monitoring accuracy of this system can be

253
R. Xi et al. Measurement 122 (2018) 247–256

Fig. 11. GPS reservoir water level retrievals from site S191 (top), S171 (middle) and S071 (bottom) and the measured records of the reservoir water level. The mean values of the measured
and estimated water level height were removed from the time series.

achieved to better than 1 mm and 2 mm in horizontal and vertical Table 2


components respectively. Except for the equipment replacement and The retrieval results in three days (unit: meter).
heavy snow event, the near real-time displacements can be con-
PRN number DOY = 094 DOY = 095 DOY = 096
tinuously observed. In the displacement time series, all the stations
exhibit obvious periodic signals and stable trend signals can only be G01 14.562 14.497 14.588
shown in the stations on dams. G03 14.544 14.551 14.624
G05 14.469 14.424 14.515
– Based on the three monitoring stations directly beside the reservoir,
G06 14.545 14.551 14.569
the accuracy of the water level data estimated with GPS SNR mea- G10 14.442 14.551 14.497
surements can be achieved to centimeter level. From a whole year G14 14.662 14.551 14.805
SNR data in 2013, the correlation coefficients between the GPS es- G19 14.632 14.714 14.642
timations and the real records of reservoir level variations can reach G26 14.430 14.461 14.479
G31 14.461 14.497 14.533
to 0.96, 0.98 and 0.93 respectively.
Mean 14.527 14.533 14.583
– From the analysis of the preliminary causative relationship between
the hydraulic load and dam displacements, the water level

Fig. 12. The LSP for the three days (Day of year = 94, 95 and 96) when the water level did not have variations.

254
R. Xi et al. Measurement 122 (2018) 247–256

Fig. 13. (a) The water level fluctuation trend and five detrended SNR series for GPS satellites with PRN number 1, 3, 10, 11 and 26 for elevation angles of 10–20°. (b) LSP for the data
shown in a.

variations could be one of the factors that results in the deformation Acknowledgment
of this dam.
We are grateful to the anonymous reviewers and editors for their
Finally, the GPS-based system designed by this paper can monitor helpful constructive suggestions and comments that helped to sig-
displacements and reservoir level synchronously, and GPS is also con- nificantly improve the paper quality. This work is supported by “the
firmed to be a viable tool for both the dam displacements monitoring Open Foundation of Key Laboratory of Precise Engineering and Industry
and a reservoir water level indicator. Surveying of National Administration of Surveying, Mapping and
Geoinformation” (Grant No. PF2017-8). This research was supported by
the National Natural Science Foundation of China (Grant Nos.

Fig. 14. Reservoir water level fluctuation and deformation series in three directions in site s191 and s071 during 2013.

255
R. Xi et al. Measurement 122 (2018) 247–256

41525014, 41304007 and 41210006), the Program for Changjiang [17] M.D. Jacobson, Snow water equivalent estimation for a snow-covered prairie grass
Scholars of the Ministry of Education of China. The Chinese Scholarship field by GPS interferometric reflectometry, Positioning 2014 (1) (2014).
[18] K.M. Larson, E.E. Small, E. Gutmann, et al., Using GPS multipath to measure soil
Council (CSC) has provided the first author a scholarship which allows moisture fluctuations: initial results, GPS Solut. 12 (3) (2008) 173–177.
him to visit the University of Nottingham in the United Kingdom for [19] K.M. Larson, F.G. Nievinski, GPS snow sensing: results from the EarthScope plate
two years to research and study in the UK from November 2016. Miss boundary observatory, GPS Solut. 17 (1) (2013) 41–52.
[20] M. Ozeki, K. Heki, GPS snow depth meter with geometry-free linear combinations of
Roxanne Parnham at the Sino-UK Geospatial Engineering Centre of the carrier phases, J. Geod. 86 (3) (2011) 209–219.
University of Nottingham is acknowledged for the proofreading. [21] C. Chew, E.E. Small, K.M. Larson, An algorithm for soil moisture estimation using
GPS-interferometric reflectometry for bare and vegetated soil, GPS Solut. (2015)
1–13.
References [22] Weiping Jiang, Hongfei Liu, Wanke Liu, et al., CORS development for Xilongchi
Dam deformation monitoring, Geomat. Inf. Sci. Wuhan Univ. 37 (8) (2012).
[1] S.I. Pytharouli, S.C. Stiros, Ladon dam (Greece) deformation and reservoir level [23] D.N. Dong, Y. Bock, Global Positioning System network analysis with phase am-
fluctuations: evidence for a causative relationship from the spectral analysis of a biguity resolution applied to crustal deformation studies in California, J. Geophys.
geodetic monitoring record, Eng. Struct. 27 (3) (2005) 361–370. Res. Solid Earth 94 (B4) (1989) 3949–3966.
[2] M. Talich, M. Glöckner, O. Böhm, et al. The application of the Ground-based InSAR [24] G. Blewitt, Carrier phase ambiguity resolution for the Global Positioning System
technique for the deformation monitoring of concrete hydropower dam Orlík on applied to geodetic baselines up to 2000 km, J. Geophys. Res. Solid Earth 94 (B8)
Vltava River, vol. 3, in: Proceedings of the 6th International Conference on (1989) 10187–10203.
Engineering Surveying, Prague, Czech Republic, 2014. [25] Weiping Jiang, Jingnan Liu, Shirong Ye, The systematical error analysis of baseline
[3] R. Vezočnik, T. Ambrožič, O. Sterle, et al., Use of terrestrial laser scanning tech- processing in GPS network, Geom. Inform. Sci. Wuhan Univ. 26 (3) (2001)
nology for long term high precision deformation monitoring, Sensors 9 (12) (2009) 196–199.
9873–9895. [26] M.A. King, S.D.P. Williams, Apparent stability of GPS monumentation from short-
[4] D. González-Aguilera, J. Gómez-Lahoz, J. Sánchez, A new approach for structural baseline time series, J. Geophys. Res.: Solid Earth 114 (B10) (2009).
monitoring of large dams with a three-dimensional laser scanner, Sensors 8 (9) [27] E.D. Gutmann, K.M. Larson, M.W. Williams, et al., Snow measurement by GPS in-
(2008) 5866–5883. terferometric reflectometry: an evaluation at Niwot Ridge, Colorado, Hydrol.
[5] H. Zhu, J. Yin, L. Zhang, et al., Deformation monitoring of dam model test by op- Process. 26 (19) (2012) 2951–2961.
tical fiber sensors, Chin. J. Rock Mech. Eng. 6 (2008) 015. [28] K. Yu, C. Rizos, D. Burrage, et al., An overview of GNSS remote sensing, Eurasip J
[6] W. Li, C. Wang, GPS in the tailings dam deformation monitoring, Procedia Eng. 26 Adv Signal Process 134 (2014-08-27) (2014) 1–14.
(2011) 1648–1657. [29] K. Yu, C. Rizos, D. Burrage, et al., GNSS remote sensing, Eurasip J. Adv. Signal
[7] P. Bukenya, P. Moyo, H. Beushausen, et al., Health monitoring of concrete dams: a Process. 2014 (1) (2014) 1–3.
literature review, J. Civ. Struct. Health Monit. 4 (4) (2014) 235–244. [30] K. Boniface, J.J. Braun, J.L. McCreight, et al., Comparison of snow data assimilation
[8] Zhenghang Li, Zhizhao Liu, Zemin Wang, Study on monitoring dam deformation system with GPS reflectometry snow depth in the Western United States, Hydrol.
with GPS positioning, J. Wuhan Univ. Hydrol. Electr. Eng. 29 (6) (1996). Process. 29 (10) (2015) 2425–2437.
[9] K.W. Hudnut, J.A. Behr, Continuous GPS monitoring of structural deformation at [31] A. Bilich, K.M. Larson, P. Axelrad, Modeling GPS phase multipath with SNR: case
Pacoima Dam, California, Seismol. Res. Lett. 69 (4) (1998) 299–308. study from the Salar de Uyuni, Boliva, J. Geophys. Res.: Solid Earth 113 (B4)
[10] E. Gökalp, L. Taşçı, Deformation monitoring by GPS at embankment dams and (2008).
deformation analysis, Surv. Rev. 41 (311) (2009) 86–102. [32] A. Bilich, K.M. Larson, Mapping the GPS multipath environment using the signal-to-
[11] C.O. Yigit, M.Z. Coskun, H. Yavasoglu, et al., The potential of GPS Precise Point noise ratio (SNR), Radio Sci. 42 (2007), http://dx.doi.org/10.1029/
Positioning method for point displacement monitoring: a case study, Measurement 2007RS003652.
91 (2016) 398–404. [33] W. Gurtner, RINEX: The Receiver Independent Exchange Format Version 2.10, ftp.
[12] A. Martín, A.B. Anquela, A. Dimas-Pagés, et al., Validation of performance of real- igscb.jpl.nasa.gov/igscb/data/format, 2002.
time kinematic PPP. A possible tool for deformation monitoring, Measurement 69 [34] W. Gurtner, L. Estey, RINEX – The Receiver Independent Exchange Format-version
(2015) 95–108. 3.00, Astronomical Institute, University of Bern and UNAVCO, Bolulder, Colorado,
[13] K.M. Larson, E.D. Gutmann, V.U. Zavorotny, et al., Can we measure snow depth 2007.
with GPS receivers? Geophys. Res. Lett. 36 (17) (2009) 153–159. [35] J.D. Scargle, Studies in astronomical time series analysis. II-Statistical aspects of
[14] K.M. Larson, J.S. Löfgren, R. Haas, Coastal sea level measurements using a single spectral analysis of unevenly spaced data, Astrophys. J. 263 (1982) 835–853.
geodetic GPS receiver, Adv. Space Res. 51 (8) (2013) 1301–1310. [36] N.R. Lomb, Least-squares frequency analysis of unequally spaced data, Astrophys.
[15] F.G. Nievinski, K.M. Larson, Forward modeling of GPS multipath for near-surface Space Sci. 39 (2) (1976) 447–462.
reflectometry and positioning applications, GPS Solut. 18 (2) (2014) 309–322. [37] T. Bayrak, Modelling the relationship between water level and vertical displace-
[16] W. Wan, K.M. Larson, E.E. Small, et al., Using geodetic GPS receivers to measure ments on the Yamula Dam, Turkey, Nat. Hazards Earth Syst. Sci. 7 (2) (2007)
vegetation water content, GPS Solut. 19 (2) (2014) 1–12. 289–297.

256

You might also like