Signal test for acoustic fibre optics for the purpose of monitoring varying salinity

(1)

Signal test for acoustic fiber

optics for the purpose of

monitoring varying salinity

(2)

Signal test for acoustic fiber optics

for the purpose of monitoring

varying salinity

Pauline Kruiver

Edwin Obando Hernandez Manos Pefkos

(3)

Title

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity Client Rijkswaterstaat Centrale Informatievoorziening, DELFT Project 11203677-003 Attribute 11203677-003-ZKS-0003 Pages 74

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity

Keywords

Acoustic measurements, signal test, salinity, speed of sound, fibre optics, distributed acoustic sensing

Summary

See management summary (English) & management samenvatting (Dutch) References

DI01 2019 – subproject 3

Version Date Author Initials Review Initials Approval Initials 0.8 19 March 2020 Pauline Kruiver Edwin Obando Hernandez Manos Pefkos

Victor Hopman Toon Segeren

Status final

(4)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity i

1 Management summary

Salt intrusion of surface waters in the Netherlands poses a problem for fresh water resources, for example at intake points of drinking water. Currently, the tools and instruments to monitor and understand salt water intrusions are insufficient. The point sensors of the monitoring network provide valuable information, but only at fixed point locations. RWS is looking for new technology to gather 2D or 3D information about salinity variations in fresh, brackish and salt waterbodies. Two promising techniques are Distributed Acoustic Sensing (DAS) with fibre optic cables and Electrical Resistivity Tomography (ERT). This report describes the results of a laboratory test using DAS. The results of a laboratory test using ERT are described in a separate report.

DAS using fibre optic cables is an emerging technique which might provide a means to monitor salinity in water via the link between the speed of sound in water and salinity. A signal test was performed in the laboratory to investigate the response of this new type of sensor and to determine the speed of sound. Three poles of closely wound fibre optic cable were situated in a large tank of water and different acoustic sources were tested.

The suitability of the source depends on the frequency content of the signal. The DAS system used in the tests had an upper limit of 60 kHz for sampling frequency. This means that sources with frequencies of ~30 kHz and higher were not registered properly, as expected from the Nyquist criterion. The lower frequency sources, such as a 3.5 kHz pinger and the 4-24 kHz sweep of an X-star are suitable sources in this setting. The acoustic waves were recorded by the DAS cable with a good signal to noise ratio.

The speed of sound through water depends on the temperature and salinity. Measurements were performed in a tank with fresh water of ~14°C. The quality of the data was good and allowed for determination of first arrivals of the acoustic signal. These travel times were converted to speed of sound, which resulted in an average of 1484 m/s. Next, an increase in water temperature was used as an analogy of increasing the salinity. Repetition of the measurements in the hot water tank (~35°C) resulted in an average speed of sound of 1520 m/s.

The signal test showed that the DAS system can faithfully record the acoustic signals. In addition, the signal test showed that the DAS system is capable of determining differences in speed of sound in different situations.

The differences in speed of sound due to salt water intrusions in the Amsterdam Rijnkanaal near Diemen (one of the potential test locations) are in the order of 5 to 10 m/s. The canal is much wider than the laboratory setup. We therefore expect that the DAS system will be suitable to detect salt water intrusions in real situations. We have included recommendations for a field test, to be combined with an ERT field test.

(7)

11203677-003-ZKS-0003, March 19, 2020, final

2 Management samenvatting

Indringing van zout water in zoet water vormt een probleem voor de zoetwatervoorziening, bijvoorbeeld bij inlaten van drinkwater. Op dit moment zijn de tools en instrumenten onvoldoende om de zoutindringing te monitoren en te begrijpen. De puntmetingen van met meetnet geven waardevolle informatie, maar slechts op een beperkt aantal vaste locaties. RWS is op zoek naar nieuwe technologieën om twee of driedimensionale informatie te verzamelen over variaties in zoutgehaltes in zoet, brak en zout water. Twee veelbelovende technieken zijn Distributed Acoustic Sensing (DAS) met glasvezel en Electrical Resistivity Tomography (ERT). Dit rapport beschrijft de resultaten van een laboratoriumtest met DAS. Een apart rapport beschrijft de resultaten van een laboratoriumtest met ERT.

DAS met glasvezel is een nieuwe techniek die mogelijk zoutgehaltes kan meten en monitoren via de relatie tussen de geluidssnelheid door water en het zoutgehalte. We hebben een signaaltest uitgevoerd om te testen welke respons deze nieuwe sensor geeft. De tests zijn uitgevoerd in het laboratorium in een grote watertank. Drie palen waar glasvezel omheen gewikkeld is zijn in de tank geplaatst en verschillende akoestische bronnen zijn gebruikt om de respons te meten.

Of een akoestische bron geschikt is hangt af van de frequentie van het signaal. Het DAS systeem dat in de tests gebruikt is heeft een maximale registratiefrequentie van 60 kHz. Dit betekent dat het signaal van akoestische bronnen met frequenties van ~30 kHz en hoger niet goed geregistreerd kan worden, zoals verwacht wordt uit de Nyquist theorie. Het signaal van de akoestische bronnen met lagere frequenties, bijvoorbeeld de 3,5 kHz pinger en de X-star met een sweep van 4 tot 24 kHz, konden wel goed opgepikt worden door het DAS systeem. De signaal-ruis verhouding was goed.

De geluidssnelheid door water is afhankelijk van de druk, de temperatuur en het zoutgehalte. We hebben metingen gedaan van de geluidssnelheid in een tank gevuld met zoet water met een temperatuur van ~14°C. De datakwaliteit was goed en de eerste aankomsten van het akoestische signaal konden bepaald worden. De looptijden zijn omgezet naar geluidssnelheden, waarbij de gemiddelde waarde 1484 m/s was. Vervolgens is het water in de tank verwarmd als substituut voor het verhogen van het zoutgehalte. De metingen zijn herhaald toen het water ~35°C was. De gemiddelde geluidssnelheid in deze situatie was 1520 m/s. De signaaltest heeft aangetoond dat het DAS system de akoestische signalen betrouwbaar kan meten. Bovendien kan met het DAS system het verschil in geluidssnelheid in verschillende situaties bepaald worden.

In het Amsterdam Rijnkanaal bij Diemen (één van de mogelijke testlocaties) zijn variaties in geluidssnelheid van ongeveer 5 tot 10 m/s te verwachten als gevolg van variaties in zoutgehalte. Het kanaal is bovendien veel breder dan de laboratorium opstelling. Daarom verwachten wij dat het DAS systeem geschikt zal zijn om in de werkelijke situatie zoutpluimen te registreren. We hebben aanbevelingen opgenomen voor een veldtest, waarbij de DAS en ERT techniek gecombineerd wordt.

(8)

11203677-003-ZKS-0003, March 19, 2020, final

3 Introduction

3.1 Background

Salt intrusion of surface waters in the Netherlands poses a problem for fresh water resources. Mitigating measures are designed and implemented during large-scale reconstruction works, such as the new locks at IJmuiden and Terneuzen. At existing locks with salt intrusion problems (e.g. Krammer, IJmuiden, Den Oever and Rhine delta) solutions such as bubble screens, pumps and narrowing the waterway are investigated. The current hydrodynamic flow models and system knowledge of salt intrusion, however, are insufficient for studying and monitoring possible solutions effectively (Schroevers, 2014). The measurement techniques and strategies currently employed by RWS are not suitable for the determination of the salt load or the shape of the salt plume along the water column. In 2017, various techniques were considered to measure and monitor salt intrusions (Schroevers et al., 2017). The most promising innovative techniques were acoustic measurements with fibre optics (Distributed Acoustic Sensing, DAS) and Electrical Resistivity Tomography (ERT).

DAS is an emerging technique which has been developed over recent years. The basic idea is that acoustic waves deform the fibre optic cable and this deformation is measured using the backscatter of a light pulse. In the oil industry, DAS has been applied for several years (e.g. Li et al., 2015). Geothermal applications are emerging, e.g. Mondanos and Coleman (2019). Applications of the technique in the shallow subsurface are being investigated (e.g. Jousset et al., 2018).

For both techniques, DAS and ERT, laboratory tests were performed in 2018 and 2019 in two different experimental setups. This report describes the signal test for DAS. The ERT laboratory experiment is reported in a separate document (Hopman et al., 2020). The experiments in the laboratory were combined with another KPP project consisting of a signal test for the detection of holes in the clay cover in a water bottom (Kruiver et al., 2019a).

3.2 Scope of the signal test

The overall goal of the KPP project “Proof of concept new measurement techniques for salt intrusion” is to obtain a future-proof strategy for acquiring data and monitoring salt intrusion in Dutch surface waters. The signal test using DAS is the first step in reaching this goal. The aim of the signal test is to determine whether the differences in acoustic travel times required for the determination of the salinity can be measured with sufficient resolution using DAS.

3.3 Acknowledgements

The application of DAS to shallow subsurface applications is new. We collaborate with the English company Silixa, who provided the DAS system (iDAS). Athena Chalari and Stoyan Nikolov provided a workshop on the iDAS technique. Francesco Ciocca and Simon Dathan assisted in operating the system in the laboratory. Rufat Aghayev and Simon Dathan are acknowledged for their help in interpreting the data.

(9)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 4 The experimental set-up and the performance of the experiments were facilitated by our colleagues Marcel Busink, Marcel Grootenboer, Edvard Ahlrichs, Mike van der Werf, Marco de Kleine and independent contractor Maarten Malingré. We express our thanks for their help.

3.4 Outlook for the future

When the signal test indicates that the travel times can be determined, a field test (in 2020) is the next step. The field test can include both DAS and ERT measurements. A possible test location is a site in the Amsterdam-Rijn Kanaal, near Diemen. In Schroevers et al. (2017) a field test is described. The recommendations for a field test are updated using the results from the laboratory experiments and presented in chapter 10. This includes several options for test locations. The selection of the most suitable test location is the first step in preparing for a field test.

3.5 Reading guide

Chapter 4 describes the theory of DAS and the method of measuring salt using fibre optics. The experimental set-up is summarised in chapter 5 and the full laboratory report is included in appendix B. Chapter 6 provides a description of the processing steps. In chapter 7, the results of the signal test are given. The discussion and conclusions are included in chapters 8 and 9. The report ends with recommendations for a field test in chapter 10.

(10)

11203677-003-ZKS-0003, March 19, 2020, final

4 Method of measuring salt with fibre optics

4.1 Influences on speed of sound in water

The speed of soundwaves (compressional waves or p-waves) in water depends on the temperature, pressure and salinity. Del Grosso’s equation (1974) is based on a polynomial of 20 components, which has been simplified to 9 components for regular use:

C = 1402,39 + 0,156 P + 5,011 T − 0,05509 T2+ 0,2215 ∙ 10−3T3+ 1,330 S + 0,13 · 010−3S2− 0,0128 T S + 0,097 · 10−3_T2_S Equation 4.1 in which C = speed of sound (m/s) T = temperature (o_C) ₀o_{≤ T ≤ 40}o_C S = salinity (o_/ oo) 0 ≤ S ≤ 35 o/oo P = static pressure (kg/cm2₎ _{P ≤ 1000 kg/cm}2

The interactive version of the equation reformulated by Wong and Zhu (1995) on

http://resource.npl.co.uk/acoustics/techguides/soundseawater/content.html requires input of the pressure in the SI units of kPa. The simplified equation is accurate to 0.5 promille within the specified range. When the temperature and pressure are known, the salinity can be derived from the speed of sound. Other speed of sound equations are given by Unesco (1981a, b, c). Chloride concentrations in fresh, brackish and salt water (according to waterinfo.rws.nl) and their corresponding salinity values are included in Table 4.1.

Table 4.1 Definition of fresh, brackish and salt water (source: waterinfo.rws.nl).

Chloride concentration (mg/L) Salinity (o_/ oo) Fresh (zoet) < 150 < 0.3

Medium fresh (matig zoet) > 150 > 0.3

Light brackish (licht brak) > 300 > 0.54

Weak brackish (zwak brak) > 1.000 > 1.8

Brackish (brak) > 3.000 > 5.4

Strong brackish to salt (sterk brak - zout) > 10.000 > 18 Salinity and chloride can be converted using:

𝑆𝑎𝑙𝑖𝑛𝑖𝑡𝑦 (𝑝𝑝𝑡) = 0.00180665 𝐶𝑙−_{(𝑚𝑔/𝐿)}

Equation 4.2

Various influences on the speed of sound and the uncertainty in the speed of sound estimation are described in Kruiver et al. (2019b, in Dutch). The influences can be summarised as follows:

• Uncertainty in travel times:

o Picking of arrival time of the direct wave.

o Recognition of direct, reflected and refracted waves. • Uncertainty in path length:

o Uncertainty in the position of the source and receivers. The signal is recorded over a gauge length of 10 m length of the fibre optic cable.

(11)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 6 o Temperature gradients, affecting the ray paths.

• Influences on the speed of sound in water:

o Uncertainty and variations in temperature. The effect of a temperature variation of 1°C is 2-3 times larger than the effect of 1 o_/

oo variation in salinity.

o Flow of water: the transit time current measurement is based on this effect. In our case it is an unwanted effect.

o Presence of air bubbles: the acoustic signal is distorted or attenuated by the presence of air bubbles in the water column (Fox at al., 1955 and Brennen, 1995)

o Density variations by suspended matter: the effects of dilute suspensions of solid particles on the sound speed may generally be neglected for practical sonar applications (Richards and Leighton, 2000).

The expected ability of the iDAS to record differences in travel times related to variations in salinities is described in the next section.

4.2 DAS as receiver

The DAS system that we used is the iDAS (intelligent Distributed Acoustic Sensor) by Silixa. Figure 4.1 shows the principle of iDAS operation where the acoustic field interacts with the backscattered light generated along a continuous fibre optic cable. By analysing the backscattered light and measuring the time between the laser pulse being launched and the signal being received, the iDAS can measure the acoustic signal at all points along the fibre. The physical parameter that iDAS measures is the rate of change of strain in the fibre or strain rate. This strain rate is averaged over a length of fibre equal to the gauge length. This gauge length is a fixed value, specific for the interrogation unit. The iDAS unit used in the experiments has a gauge length of 10 m. This means that the light that is backscattered over a cable length of 10 m is averaged into one data point. However, the spatial resolution of the fibre optic cable is much finer, because the cable can be interrogated at intervals varying from 0.25 to 2.0 m. This is a setting in the acquisition software. Setting the spatial resolution to 0.25 m results in one data point for every 0.25 m along the cable. This is the finest sampling for the system used in the tests. The gauge length is one of the aspects that are being improved in DAS systems in newer units by technological innovations. The newer versions of Silixa’s DAS system (Carina) has an adjustable gauge length of 2, 10 or 30 m. The gauge length of 2 m is a technical improvement over the system that was used in the tests. In addition, noise levels in the Carina system are reduced using a specially manufactured fibre option cable.

The length of the cable, the duration of the pulse, the speed of light and data-transfer rate pose a limit to the sampling frequency of the recorded signal. For the experimental set-up with 3 poles in November 2018, the maximum achievable frequency was 40 kHz. For the experimental set-up with 2 poles in February and March 2019 the limit of maximum sampling frequency was reached at 60 kHz.

(12)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 4.1 Visualisation of the iDAS technique (source: Silixa).

Application of iDAS for salinity measurements

The ability of the system to record signals with certain frequencies is linked to the sampling frequency. According to Nyquist theory, the sampling frequency needs to be at least twice as high as the source frequency. Ideally, the sampling frequency should be higher. Because of the limit in sampling frequency of the iDAS system of 40-60 kHz, the choice of usable frequencies of the acoustic sources is limited as well.

The variations in salinity that RWS requires to be measured depend on fresh or salt water conditions. For salt water conditions, the current salinity meters need to be able to detect differences of 1 o_/

oo and for fresh water conditions of 0.025o/oo. To illustrate the effect of

measurement frequency and frequency of the acoustic source in relation to these salinity variations, a calculation example is presented in Kruiver et al. (2019b) and summarised here. This calculation example is inspired by the Amsterdam Rijnkanaal near Diemen, which is approximately 100 m wide and has a water temperature of ~4 to 5 °C in winter. For salt water conditions at 4.7°C, an arbitrary salinity of 17o_/

oo, distance of 100 m and a difference in salinity

of 1o_/

oo , the travel time difference corresponds to 0.06 ms. For fresh water conditions at 4.5°C,

an arbitrary salinity of 0.088o_/

oo, distance of 100 m and a difference in salinity of 0.025o/oo, the

travel time difference corresponds to0.0015 ms (see example in Kruiver et al., 2019b). The interplay between these travel time differences and the source signal is illustrated in Figure 4.2 for a perfectly sampled signal and in Figure 4.3 for a down-sampled signal. If we extrapolate the findings of Figure 4.3 to the maximum iDAS sampling rate of 40-60 kHz, it is clear that the required travel time differences cannot be measured using a standard 210 kHz echosounder source combined with the iDAS sampling rate. A lower frequency source is required.

(13)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 4.2 Interplay between the measured travel time differences between two situations with a difference in salinity equal to the RWS requirements for electrical conductivity (red and black) and the frequency of the source for a perfectly sampled signal. The difference between the red and the black line is Δt (left for fresh water Δt~0.0015 ms and right for salt water Δt~0.06 ms for an example distance of 100 m). For a source with a frequency of 210 kHz (top panels) and 3.5 kHz (bottom panels).

(14)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 4.3 Same as Figure 4.2, but with a lower sampling frequency. The source of 210 kHz is sampled with 600 kHz (top panels) and the source of 3.5 kHz sampled with 40 kHz (bottom panels).

(15)

11203677-003-ZKS-0003, March 19, 2020, final

5 Summary of experimental set-up

The full laboratory report is included in appendix B (in Dutch) and summarised in this section. The setup consists of a tank (18 m x 5.5 m x 2.5 m) with three poles with one densely wrapped fibre optic cable around each pole (Figure 5.1). Pole C is situated horizontally on the bottom in the sand bed in the length of the tank and is therefore not visible on the pictures. Poles A and B are vertical with the bottom of the pole resting on the sand bed and the top in the air.

Figure 5.1 Set-up during the cold and warm water test on 22 February and 8 March 2019. Left panel is a schematic top view.

The measurements were performed in three rounds. The very first measurements were conducted on 21-23 November 2018. The purpose of these measurements was testing different sources and investigating the response of the poles wound with fibre optic cable. In addition, the settings were optimised. Because this was the first time of using the iDAS in this setting, all available sources were used, even though not all of them were suitable from a transmitter and receiver frequency perspective. Based on the tests, a source with a suitable frequency was

(16)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 11 selected for the cold and warm water experiments in February and March 2019 (Table 5.1). After the first test days, the position of the poles was adjusted as well. The speed of sound can be determined more reliably for a simple set-up with the source and two poles in one line. The final set-up is shown in Figure 5.1, where the distance between poles A and B was 9.09 m. The sources used in the different experiments are listed in Table 5.1.

To test the effect of varying salinity on the speed of sound in the laboratory setting, we chose to use an analogy of temperature. Instead adding many kilograms of salt and having to dispose of the salt water and sediments on the bottom of the tank after the experiment, we heated the water in the tank to increase the speed of sound. On 22 February 2019, we heated ~1 m3_of

water to ~65°C using two portable boilers. After finishing the cold water experiment, we added this hot water between poles A and B while the X-star source was active and the iDAS system was recording the response. After this, the X-star was disconnected and removed from the tank during heating in order to avoid wear. The basin water was pumped through the boilers heating the water. After one week of heating, which was the maximum available time for this part of the experiment, the temperature had risen from ~14°C to ~35°C. It was expected that this rise in temperature would be sufficient for a detectable effect on the speed of sound (Equation 4.1). The X-star measurements were repeated in the warm water tank on 8 March 2019.

Table 5.1 Summary of sources and sample frequencies during the various test days.

Date Source Sampling

frequency

Purpose

21-23 November 2018

Knudsen 3.5 kHz Pinger 40 kHz Testing different sources and response from the fibre optic poles Knudsen dual frequency

echosounder (33 & 210 kHz) 1)

40 kHz

Hammer 30 kHz

“Seismic tube” (experimental source) 1)

40 kHz

22 February 2019 Edgetech X-star SB-424 60 kHz Measuring in cold water

8 March 2019 Edgetech X-star SB-424 60 kHz Measuring in warm water

(17)

11203677-003-ZKS-0003, March 19, 2020, final

6 Processing

6.1 Processing flowchart

The data from the iDAS measurements were processed using different custom-built routines. The flowchart is shown in Figure 6.1. After reading the files, the data was first plotted (§ 6.2). Based in visual quality control, several files were selected for further analysis. The X-star source emits a sweep of frequencies. During normal use of the X-star, the instrument also records the signal. While doing the recording by the instrument itself, the signal is automatically deconvoluted. In our tests, the X-star was used as a source only and the signal was recorded by the iDAS. In this case, the signal needs to be deconvolved before further processing (§ 6.3). This is an extra step relative to the other acoustic sources. The data is recorded continuously. Therefore, the individual shots (i.e. regulated bursts of acoustic energy) need to be recovered from the files containing several shots (§ 6.4). The calculation of the sound od speed is based on first arrival picking (§ 6.5).

Figure 6.1 Processing flowchart

6.2 Raw data visualisation and Quality Control

Initially, the raw data from each experiment was normalised in order to remove some coherent noise and plotted. Quality control (QC) was then performed visually on the produced images with the aim of selecting the datasets with optimal data quality for further processing. The overall noise levels and the ability to distinguish the signals coming from different poles and from each individual shot were the main criteria for this selection. Noise comes from the system and from ambient noise. An example of a dataset selected for processing is shown in Figure 6.2. All channels along the fibre optic cable are plotted through time. The trace of each channel is the amplitude record of vibration versus time. The plot shows the traces in the x-axis versus time on the y-axis. The amplitude of the trace is visualised by the colour, with blue colours denoting low amplitudes and red colours denoting high amplitudes, via the colours of the rainbow.

(18)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 6.2 Normalised raw data plot for the water tank experiment on 22/11/2018 using a Hammer source. The colour bar indicates the normalised amplitude of the signal. The distance on the x-axis denotes the distance along the fibre optic cable. Each observation point along the cable is referred to as a channel, analogous to a classical geophone streamer with e.g. 24 channels where each channel is a physical geophone.

The selected datasets for processing come from the experiments described in Appendix B, where the experimental configuration (sources, geometry of poles, introduction of hot water in the tank, etc.) used for obtaining each one is discussed.

The raw data can be converted to strain rate (measured in nanostrain/second) with the application of Equation 6.1:

𝑠𝑡𝑟𝑎𝑖𝑛 𝑟𝑎𝑡𝑒 = 𝑠𝑐𝑎𝑙𝑒 𝑓𝑎𝑐𝑡𝑜𝑟 (𝑛𝑚) ∗ 𝑖𝐷𝐴𝑆 𝑟𝑎𝑤 𝑑𝑎𝑡𝑎 ∗ 𝑠𝑎𝑚𝑝𝑙𝑖𝑛𝑔 𝑓𝑟𝑒𝑞. (𝐻𝑧) 𝑔𝑎𝑢𝑔𝑒 𝑙𝑒𝑛𝑔𝑡ℎ (𝑚)

Equation 6.1

This conversion was not applied to the processed datasets, as it is essentially the application of a numerical factor which doesn’t alter the physical meaning of the measurements. Further interpretation of the data only used the timing of the first strong amplitude, not the amplitude value itself.

6.3 X-Star source signal deconvolution

The data from the experiments in which the Edgetech X-star Sb-424 was used as a source required some additional processing. This instrument deconvolves the signal it emits after it is returned to the instrument and acts simultaneously as a receiver. When the source is used with a different receiver, such as the iDAS, then the source signal needs to be deconvolved by processing. The source signal’s precise signature had to be identified and extracted from the

(19)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 14 data, using the existing knowledge on the sampling frequency (60 kHz) and the frequency range of the source sweep (4-24 kHz).

For each dataset, the Fourier amplitude spectrum of individual traces was initially computed using the Fast Fourier Transform (FFT) in order to identify the most prominent frequencies in each trace, while also verifying that the source’s frequency range is present. The traces were split into 0.02-second-long windows, using a Tukey window (employs cosine tapering with a shape parameter of 0.05 – this determines the fraction of the window inside the cosine tapered region). An RMS analysis of the calculated FFTs was next performed on the windowed traces to separate those containing good signal.

Following this, a Butterworth bandpass filter (Figure 6.3) was applied on the average trace of all selected traces containing good signal (i.e. excluding records with low quality shots, or test shots for optimisation of the acquisition settings), leaving only the frequencies contained in the source signal (4-24 kHz).

Figure 6.3 The bandpass Butterworth filter applied on the signals.

The resulting Fourier amplitude spectrum of the average original signals (Figure 6.4) and filtered signals (Figure 6.5) suggests that the source signal increases quadratically in the frequency domain between 4-24kHz. In this way, the spectral properties of the source signal were identified, which enabled the reconstruction of a new source signal.

It was found that a chirp signal with a quadratically increasing frequency from 4kHz to 24kHz, shown in Figure 6.6, best mimics the spectral behaviour of the source signal we investigated. Analysis of different shots showed that the source sweep is consistent and repeatable between shots. Comparison with instrument specifications was not possible, because the sweep characteristics and the deconvolution algorithm of the X-star are not disclosed by the manufacturer.

(20)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 6.4 The Fourier amplitude Spectrum of the original signal.

Figure 6.5 The Fourier amplitude spectrum of the filtered signal, after a bandpass filter has been applied.

The reconstructed source signal then had to be extracted from the raw data. Two methods were employed to accomplish this. Initially, cross-correlation with the source signal was performed using each of the selected trace’s FFT. An example of cross-correlation being applied to the traces gathered after a certain shot during the X-star source experiment (28/02/2019) is shown in Figure 6.7. Poles A and B can be distinguished. However, it is unclear where the first arrival of the source signal occurs on each one. Several spurious features are present, such as a seemingly simultaneous event between traces 30 and 80 around 0.724

(21)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 16 seconds. These features hinder the interpretation of the first arrival times of the source signal on the poles.

This led to the application of deconvolution as a means of extracting the source signal from the acquired datasets. An example of its application is shown in Figure 6.8. The aforementioned features have now been removed, and the incidence of the wavefront on each pole is clearly visible. Following the deconvolution of the reconstructed source signal, the data were prepared for further analysis.

(22)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 6.7 The X-star dataset for a specific shot after cross-correlation with the reconstructed source signal.

Figure 6.8 Application of deconvolution on the same dataset as that of Figure 4.6.

6.4 Automatized Selection for X-star data

In this step, the aim was to split each signal into separate windows containing the energy of each individual shot. This was specifically done for the X-star data, because the data from this source were mainly used in the analysis of the cold and hot water experiments. The records were registered in continuous mode, resulting in many shots per record. For the analysis, a method had to be found to identify the signature of each shot and to capture all of its energy evenly. The shot rate of the X-star was two shots per second. Initially, each record containing multiple shots was split into 0.5-second windows by identifying the peak amplitude contained within each window. Each shot was thus identified and the time sample number that corresponds to that peak was recorded. The number of shots in each experiment can be determined by visual inspection of the raw data.

(23)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 18 Every shot was isolated into a 40-milliseconds window in a way that the highest peak in that window was positioned in the middle (even number of samples on either side). In this way it was ensured that the selected data corresponds to a single shot wherein the highest energy is concentrated.

Figure 6.9 An example of a windowed shot, where the energy from a single shot (Poles A and B) has been isolated from other shots.

6.5 Automatic Picking of First Arrivals

The automatic picking was designed for the windowed shots of the X-star data. In principle, the routine could also be applied to other sources, if these data would have been separated shots. However, the windowing per shot was not performed for the other sources, because the focus of the analysis was on the X-star data. This is related to the suitable and well-defined frequency content of the X-star source.

To compute the speed of sound it is necessary to estimate the arrival time of sound waves at both pole A and B that are located at known positions. Each record contains 60 shots recorded at both poles (30 second records with 2 shots a second). In total 46 records contained 2,760 shots recorded at both poles. As previously described, each individual shot of 40-ms was deconvolved to extract the actual response on the propagation medium. An example of the automatic picking in the collected shot gather for pole A is displayed in Figure 6.10.

(24)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 6.10 Example of automatically picked first arrivals of the deconvolved signals for cold water.

To cope with the large number of shots/records the first arrival picking is performed in automatized manner. The extracted data should provide a big enough data base for statistical analysis.

The automatic picking was performed along all wave arrivals for poles A and B. Prior to the first arrival picking, for each signal the analytical signal is computed using the Hilbert Transform. The envelope of the analytical signal enhances the prominent amplitudes and minimizes the non-coherent amplitudes.

An example of the automatized picking is displayed in Figure 6.11 for a shot in the cold-water case. The algorithm worked well in that case over the entire length of the fibre optic cables on poles A and B. The algorithm performed less well in the warm water case (Figure 6.12), where there seem to be mispicks of one cycle at certain sections of the poles. Apparently, the data are noisier and the amplitudes of the wavefront arriving at different sections of the pole are less constant. Due to time limitations we have excluded the obvious mispicks from the analysis, rather than improving our automatic picking procedure. Instabilities in the acoustic signal and mispicks might be related to lack of coupling of the fibre cable onto the pole.

(25)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 6.11 Example of automatically picked first arrivals of the deconvolved signals for cold water.

(26)

11203677-003-ZKS-0003, March 19, 2020, final

7 Results

7.1 Arrival times of various sources

During the various test days, several sources were tested in order to assess their suitability in combination with the iDAS. This was part of the signal test. Based on the frequencies of the sources and the sampling frequency of the iDAS as explained in section 4.2, we expected beforehand that certain sources might be unsuitable. The sampling frequency varied among the different tests and was optimised to a frequency as high as possible in later stages during the experiments.

The response of the iDAS to various sources and the analysed arrival times of the following sources are presented and discussed in this sub-section:

• Hammer of unknown and uncontrollable frequency content (sampling frequency: 30 kHz) • Knudsen 3.5 kHz Sonar Ping (sampling frequency: 40 kHz)

• Knudsen dual frequency (33 & 210 kHz) echosounder (sampling frequency: 40 kHz) • Seismic tube with various frequency settings (sampling frequency: 40 kHz)

7.1.1 Hammer Source

During the several test days, a hammer was hit against the side of the tank in various positions. During the first tests, the hammer blows were too loud, which resulted in clipping of the signal. Next, more gentle blows were applied, which resulted in a better defined recording by the iDAS. The 6 hammer shots used in the experimental configuration of the 22nd_{November 2018 were}

analysed. The seismic traces of one shot on selected traces on a short section of pole C are shown in Figure 7.1. The position of the hammer shot relative to the pole was unknown. Therefore, we chose the portion of the traces that appeared to be closest to the source and therefore representing simultaneous arrival of the wavefront.

First arrivals were hand-picked and presented in Table 7.1. Note that the mean arrival times presented for each shot are not absolute travel times but are all relative to the start time of the window in which each shot was captured, as determined by the algorithm developed in Section 6.4. All the shots contained in a file are located and separated, with time then starting from 0 seconds for each shot. The low standard deviation of these results indicates that this source signal’s arrival times on Pole C were uniform along this portion of the pole. In addition, repeatability is clearly seen between shots. These factors facilitated the picking of the first arrival times of the source signal due to its distinguishable and consistent pattern, as seen in Figure 7.1. The iDAS system is therefore able to discern and faithfully reproduce a hammer source signal under these experimental conditions.

(27)

11203677-003-ZKS-0003, March 19, 2020, final

Table 7.1: Shot statistics for hammer source

Figure 7.1 First arrival time picking results for the 1st_{shot performed with a hammer source.}

7.1.2 Knudsen 3.5 kHz Sonar Ping on pole C

Eight recordings using the Knudsen sonar ping source with a 3.5 kHz central frequency were analysed for the experimental setup of 22nd_{November 2018. The arrival times across 12}

channels along the fibre optic cable were picked. The channels of choice were no. 150 followed by no. 400 to 625 in steps of 25 and no. 950. A calculation of velocity can be performed from these measurements, since they describe the movement of a wavefront along Pole C, over which distance is known. The difference in channel number between the channels farthest away from each other (950 and 150) can be converted into absolute distance by accounting for the optic fibre cable’s wrapping factor. This distance amounts to 8.168 meters. It is important to note that the calculated velocity is in fact an apparent velocity, since the exact distance from the source to each section of the pole is unknown due to the frequent movement of various sources on the first test day. The pattern of waveform propagation, however, suggests that the source was positioned approximately in line with pole C. Table 7.2 summarises the information used and the resulting apparent velocities.

Shot Number Mean arrival time (s)

(within window of current shot)

Standard deviation of first arrivals within one window (s)

1 0.00986 2.18E-06 2 0.00990 5.40E-06 3 0.00640 3.13E-06 4 0.00843 2.97E-06 5 0.01071 2.69E-05 6 0.00953 2.55E-06

(28)

11203677-003-ZKS-0003, March 19, 2020, final

Table 7.2: Arrival times at channels 150 and 950 which were used for deriving the apparent velocity along Pole C for the Pinger source. The mean apparent velocity and its standard deviation are also shown in the final two columns.

The standard deviation of the apparent velocity, shown in Table 7.2 is quite high. A possible cause might be related to interference of different waves, e.g. the direct wave through water and the wave through the PVC tube. Based on the difference in speeds in water (~1500 m/s) and PVC (~1060 m/s, https://www.rshydro.co.uk/sound-speeds/) it is expected that these waves do not interfere in the first arrival picking procedure. We have not found an explanation for the large standard deviation. The high standard deviation might suggest that the sonar ping is not an optimal source for our purposes when used for measurements with an iDAS system. A standard deviation of 14.7 m/s is too high compared to the minimum velocity differences we are interested in measuring. In addition, the lack of consistency and reproducibility would further complicate the use of this source in such a context. Picking the first arrival times, however, was easily done as the first breaks were clearly distinguishable, as evidenced by Figure 7.2. In the next section we will show that for a different configuration, the combination of the Pinger source and iDAS cable performs better.

Shot Number Ch. 150 first arrival time [t] Ch. 950 first arrival time [t] Apparent Velocity [m/s] Mean App. Velocity [m/s] Standard Deviation of App. Velocity [m/s] 1 0.01324 0.00731 1377.4030 1391.6 14.7 2 0.01320 0.00728 1379.7297 3 0.01307 0.00727 1408.2151 4 0.01310 0.00732 1413.1487 5 0.01319 0.00733 1393.8566 6 0.01309 0.00726 1401.0291 7 0.01317 0.00720 1368.1742 8 0.01314 0.00727 1391.4821

(29)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 7.2 First arrival time picking for shot 1 of the Knudsen 3.5 kHz Sonar Ping source from Channel 400 to Channel 625 in steps of 25 channels. The wavefront travels along the pole, reaching the high trace numbers first and the lower trace numbers later.

7.1.3 Knudsen 3.5 kHz Sonar Ping on poles A and B (Far End)

The same source was also recorded on the other two poles, A and B, which have an orientation different from pole C. Ten recordings on poles A and B using the Knudsen sonar ping source with a 3.5 kHz central frequency were analysed for the experimental setup of 21st_November

2018. In this case, the arrival of the acoustic signal on the two poles are clearly visible Figure 7.3.

Table 7.3 Mean picked first arrival times and their standard deviation for Poles A and B for the Knudsen 3.5 kHz Sonar Pinger source.

The picked arrival times at each pole are displayed in Table 7.3. An example of picked first arrivals on Pole B for the first recording are shown in Figure 7.4.

Shot Number

Pole A: Mean first arrival time [s]

Pole A: Mean first arr. time Standard Deviation [s]

Pole B: Mean first arrival time [s]

Pole B: Mean first arr. time Standard Deviation [s] 1 0.0155 2.6e-04 0.01267 3.5e-05 2 0.0151 2.6e-04 0.01241 3.0e-05 3 0.01547 2.8e-04 0.01268 2.6-05 4 0.01518 3.0e-04 0.01241 3.7e-05 5 0.01546 2.8e-04 0.01267 3.8-05 6 0.0152 2.6e-04 0.0127 3.3e-05 7 0.01553 2.6e-04 0.0124 4.8e-05 8 0.01517 2.6e-04 0.0127 4.4e-05 9 0.01548 2.9e-04 0.01238 3.5e-05 10 0.0155 3.0e-04 0.01269 4.5e-05

(30)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 25 In the results presented in the previous section (7.1.2), the same source was used with a different experimental configuration: the wavefront travelled along pole C in that case. In the configuration described in this section, the wavefront hits the entire pole (either A or B) approximately simultaneously along the entire pole. We observe that the standard deviation of the first arrival times for both poles is now significantly lower compared to the previous results, indicating that the arrival times are consistent and clear. This is further supported by the fact that there were no complications with picking the first breaks, as seen in Figure 7.4.

The incoming wavefront was therefore more accurately sampled with the two vertical poles A and B, rather than with the horizontal pole C. This shows that directionality of the source relative to the fibre optic cable is important.

The standard deviations of the mean first arrivals on pole B are around an order of magnitude smaller (order 10-5_{) compared to the arrivals on pole A (order 10}-4_{, Table 7.3). While this is}

supported by Figure 7.3, where the arrivals on Pole A are more inconsistent compared to pole B, an explanation for this effect has not yet been found.

Figure 7.3 The first recording’s signal using the Knudsen 3.5 kHz Sonar Ping source, placed at the far end of the water tank (21 November 2018). The feature on the left is Pole A, while the feature on the right is Pole B.

(31)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 7.4 Picked first arrival times on Pole B for the Knudsen 3.5 kHz Sonar Pinger source. The wavefront hits the pole simultaneously on all traces.

Figure 7.5 Picked first arrival times on Pole A for the Knudsen 3.5 kHz Sonar Pinger source. The wavefront hits the pole more irregularly compared to Pole B. Note that, to generate this Figure, the processing was repeated and the time axis has been slightly shifted upwards.

7.1.4 Knudsen dual frequency echosounder (33 and 210 kHz)

Several measurements on the 22nd_{November 2018 used a dual frequency echosounder with}

frequencies of 33 and 210 kHz. Raw data visualisation of these measurements, such as that seen in Figure 7.6, showed that they contained excessive amounts of noise, up to a level where no processing could lead to the recovery of a clear signal.

(32)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 27 The sampling rate for these experiments was 40 kHz. According to the Nyquist-Shannon sampling theorem, since the source signal frequency is higher than half the sampling frequency, the source signal cannot be faithfully recorded. The iDAS system used in our experiments cannot thus be used with this specific source. This would only be feasible with an iDAS system capable of sampling at frequencies of minimum 66 kHz to detect the signal of the lower frequency of the two frequencies of this source.

Figure 7.6 An example of a dataset acquired with the Knudsen 33&210 kHz echosounder as a source showing noise only due to the low sampling rate, forbidding the recovery of a clear signal

7.1.5 Seismic Tube: Experimental Source

The seismic tube is an experimental source, which can be controlled to emit a pulse with a frequency between 0.1 and 100 kHz or a sweep e.g. from 0.1 to 10 kHz. The purpose of using this source was for the co-located experiment of KPP project “Direct measurements using fibre optic cable” (RWS KPP project 2019 BO03). A raw plot of normalised data acquired by the iDAS system with the experimental seismic tube as source (Figure 7.7) on the 22nd_November

2018 shows two discernible shots occurring between 2-3 seconds and 5-6 seconds. After processing this dataset and displaying it in “wiggle” form, such as in Figure 7.2, the recorded source signal was not clear due to the low sampling rate. The signal was still obscured by high noise levels. This made data quality insufficient for first arrival picking. The seismic tube is not suitable as a source for this particular salt water intrusion setup.

(33)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 7.7 Excessive noise levels obscuring seismic tube’s raw signal., due to a combination of source frequency and sample rate of the iDAS system.

7.2 Speed of sound in cold and hot water

The data collected on 28 February 2019 (cold water) and 8 March 2019 (hot water) were used to analyse the speed of sound. On both days, the source was the Edgetech X-star Sb-424. As explained in chapter 6, the data collected using sweep source along each record the 60 shots were separated using common window length. All individual records were deconvolved using averaged band-pass filtered record portion used to recreate the sweep source. In addition to the cold water and the warm water situation, a blob of hot water was added to the cold-water tank on 28 February 2019.

7.2.1 Cold water case

An example of a deconvolved shot gather for cold water case is displayed in Figure 7.8. For both poles a clear and consistent clear first arrivals occur at 0.009 and 0.016 seconds approximately. The first arrivals were automatically picked for Poles A and B using the algorithms described in section 6.5. For the arrival time analysis in cold-water up to 100 traces were selected along both poles.

[s

]

(34)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 7.8 Example of a deconvolved shot gather for first arrival determination for hot-water case.

The picked first arrivals are displayed in Figure 7.9. In general, the average picked arrival-times appears to be rather consistent along all collected traces. The computed standard deviation as a function of trace is shown in Figure 7.10. It appears that pole B (closer to the source) has both a stronger signal and a lower standard deviation (0.033 ms) compared to Pole A (0.047 ms). In some segments, the standard deviation is very similar in both poles, e.g. for traces 15 to 30 and 60 to 80.

Figure 7.9 Stability of the first arrivals on two poles in the cold water case: the red line shows the average over all shots (realstart_4kHz-24kHz_190228131311.tdms) as a function of trace. The grey band shows the standard deviation. The top line refers to pole B (closer to the source) and the bottom line to pole A (further away from the source).

(35)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 7.10 Variability of standard deviation (std) for all records as a function of trace (average std of first arrivals for pole B in green = 0.033 ms and pole A in blue = 0.047 ms).

Using the distance between the poles, the difference in arrival times was used to calculate the speed of sound. In an approximation, we assume that the wave front hits the pole simultaneously on the entire pile, as a plane wave. Each trace on one of the poles was combined with all traces of the other pole. This combination of first arrivals on poles A and B resulted in 14,400 pairs. The computed speed of sound is shown as a histogram in Figure 7.11a. The average speed of sound is 1,487.1 m/s with a standard deviation of 4.8 m/s.

a) all traces b) centre traces of pole

Figure 7.11 Distribution of the computed speeds of sound for the cold-water case. a) Using all traces, resulting in an average of 1487.1 m/s with a standard deviation of 4.8 m/s. b) Using only the centre 10 traces of each pole resulting in an average of 1483.6 m/s with a standard deviation of 4.4 m/s.

(36)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 31 However, considering the distributed nature of the collected traces an important aspect to take into consideration is the validity of plane wave assumption. In Figure 7.12 the zoomed view of first wave arrivals shows a marked hyperbolic shape caused by the proximity of the source, while for pole A there is almost no hyperbolic feature that is attenuated by the longer distance to the source. Based on the geometry (position of source and length and positions of poles) and an indicative speed of sound of 1500 m/s, we expect a difference in first arrivals of 0.07 ms between the centre and the end of pole B and 0.02 ms for pole A. The observed curvature in Figure 7.12 is somewhat smaller. The variation between traces seems to be larger than the geometric effect. In the hot water, was, we therefore used a different criterion to select the traces used to determine the most representative speed of sound.

Figure 7.12 Zoomed in on first arrivals for Poles A and B for cold-water case.

To analyse the implications of the hyperbolic or non-plane wave assumption, the velocity for cold-water condition (that is more stable than hot-water condition) is computed using a selection of traces from the middle portion (e.g. traces 280 – 290 of pole B in Figure 7.12). For these traces at the top of the hyperbola, the simultaneous arrival at the pole is fulfilled. The average speed of sound in this case is 1483.6 m/s with a standard deviation of 4.4 m/s (Figure 7.11b). This is slightly lower than using all traces. The changes in the speed of sound that are relevant

(37)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 32 for the salinity are small. Therefore, the geometry effects need to be taken into account in a field test.

7.2.2 Hot water case

The measurements were repeated using the same source (Edgetech X-star Sb-424), placed at the same position when the water in the tank had been heated uniformly to ~35 °C. The data was processed using the same algorithms as the cold-water case. An example of a shot in the hot-water case is shown in Figure 7.13.The deconvolved signals depict clear and coherent wave arrivals for both poles. The hyperbolic shape of first arrivals, however, is not so clear in this case (Figure 7.14). We therefore used a different criterion to select the best traces to include in the calculation of the speed of sound in the hot water case.

(38)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 7.14 Zoomed in on first arrivals for Poles A and B for hot-water case.

The picked average arrival times on poles A and B are shown in Figure 7.13 (zoom without pick in Figure 7.15). In the hot-water case, the arrival times show more variability compared to the cold-water case (Figure 7.9). The average arrival times show marked variation between traces and poles (Figure 7.15). For pole A, it appears that waves arrive earlier in the middle of the pole, while at both ends wave appears a bit later. If this was due to geometrical effects, this would effect would be larger at pole B than at pole A. However, we observe the opposite. An alternative explanation might be that it is caused by a non-uniform temperature distribution or by a deviation from a plane wave. For Pole B, which is closer to the source, the variation seems to even higher compared to Pole A.

The variability can be better observed when plotting together the standard deviation respect to the traces for both poles (Figure 7.16). The average standard deviation for pole B (0.033 ms) is almost 1/3 of the average standard deviation obtained at Pole A (0.10 ms). The major difference is observed in those traces located at both ends of both the poles.

(39)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 7.15 Stability of the first arrivals on two poles in the hot water case: the red line shows the average over all shots (for the record AB_UTC_20190308_144402.412.tdms) as a function of traces. The grey band shows the standard deviation. The top line refers to pole B (closer to the source) and the bottom line to pole A (further away from the source).

Figure 7.16 Variability of standard deviation (std) for all records as a function of trace (average std of first arrivals for pole B in green = 0.100 ms and pole A in blue = 0.033 ms).

(40)

11203677-003-ZKS-0003, March 19, 2020, final

a) all traces b) data with std <0.05 ms

c) data with std < 0.02 ms

Figure 7. 7.17 Distribution of the retrieved speed of sound in the warm water case for various ranges of standard deviation for all traces displayed in Figure 7.16, a) all data, b) data with std of arrival times <0.05 ms, and c) data with std of arrival times < 0.02 ms. The average values and standard deviations are included in Table 7.4.

The resulting speeds of sound are shown in the histograms in Figure 7. 7.17. For a temperature of 35 °C and salinity of 0, a speed of sound of 1,520 m/s is expected. When all data are included, the average speed is 1524 m/s. This is close to what is expected. However, this includes the obvious mispicks of the traces that seem to be unstable over the shots. If we want to improve the speed estimate, we can reject the combinations that show a standard deviation of the picked arrival times of either above 0.05 ms or even above 0.02 ms (even stricter criterion). The statistics of the remaining data are included in Table 7.4. As expected, rejection of a larger portion of the data results in a smaller standard deviation for the calculated speed of sound. With the strictest rejection criterion, for which we believe all mispicks are removed, the average speed of sound is 1,520 m/s. This is significantly faster than in the cold water case.

Table 7.4 Speed of sound for all possible combinations and for rejecting combinations with a standard deviation of the picked arrival times above a threshold.

Number of Combinations Travel Time Std (ms) Combinations used (%) Speed of Sound (m/s) Std Speed of sound (m/s) 10,000 All 100 1524.1 21.6 5,760 >0.05 57.6 1520.7 17.4 1,230 >0.02 12.3 1520.3 15.9

(41)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 36 7.2.3 Cold to hot water

During the cold-water experiment on 28 February 2019, 1 m3_{of hot water (~65 °C) was added}

to the tank via a tube in between poles A and B (Figure 7.18). The X-star source was turned on and continuous measurements were collected for about 30 minutes. Figure 7.19 shows the deconvolved signal and frequency content of the raw data along the selected traces for one shot collected at pole B. The waveform shows different types of vibrations, such as the direct wave (our target) and low frequency energy which is probably noise. The shot-gather shows a prominent energy in the range of 4 kHz to 16 kHz that occurs along all traces. There is also energy present of low frequency, which is probably noise.

Figure 7.18 Schematic representation of the addition of the hot water blob on 28 February 2019, between pole A and B.

Figure 7.19 Top: deconvolved no-filtered signal of shot 60, shown as time traces. Bottom: Power spectral density as a function of trace.

The same record is band-pass filtered using a frequency limit of 4 kHz – 16 kHz to isolate the energy of the main signal as observed in the spectrogram (Figure 7.20). After band-pass filtering, shot gather depicts a clean and coherent waveform with a consistent energy pattern around 7 kHz. The highest energy seems to be concentrated in the middle of the shot gather and with less energy amplitude towards both ends of the pole (Figure 7.20).

noise signal

(42)

11203677-003-ZKS-0003, March 19, 2020, final

Figure 7.20 Top: deconvolved band-pass filtered signal of shot 60, shown as time traces. Bottom: Power spectral density as a function of trace.

Figure 7.21 Spectrograms of deconvolved signals over time during addition of 1 m3_{of hot water in the cold-water} tank. Top: pole B (closer to the source), bottom: pole A.

The computed spectrum along selected traces for each record are averaged and plotted as a function of time (Figure 7.21). The spectrogram is computed using 140 traces from 1,800 shots (260,400 traces) at each pole, consecutively collected over ~16 minutes (from 13:36:58 to 13:52:47). Both spectrograms represent the time of the experiment of adding hot water. The variation of spectral energy is not so clear from this figure. The variation is better visible in

(43)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 38 Figure 7.22. In this figure, the peak amplitude extracted from both spectrograms is compared with the velocity computed in time domain from consecutive 1800 shots. In the top panel, there is a change in character of the peak spectral energy: it is much noisier from ~13:44 on, probably due to the water flow and bubbles. In the bottom panel, there is a similar sustained increment in speed of sound during the first 7 to 8 minutes of the experiment from 1481 m/sec up to 1484 m/sec remaining close to that value until the end of the experiment. According to the log, the hot water was added at ~13:43 and lasted about 2 minutes. Synchronisation between different measuring systems was not accomplished (lesson-learned for next experiment). The marked change close to 13:44 coincides probably with the moment that the hot water was added to the tank.

Figure 7.22 Top: Extracted peak energy from the computed spectrogram (Figure 7.21). Bottom: Computed speed of sound from first arrival differences between poles A and B derived from 1,800 shots.

(44)

11203677-003-ZKS-0003, March 19, 2020, final

Signal test for acoustic fiber optics for the purpose of monitoring varying salinity 39 In order to investigate if this change in character in the peak energy is related to the addition of the hot water, the same analysis was carried for the cold-water case (Figure 7.23). Here, almost no variation was observed in the energy of the spectrogram, indicating that the pervious case indeed seems to provide information of the modification of the water properties due to adding hot-water during the measurements.

Figure 7.23 Cold water case, prior to addition of hot water. Top: Power spectral density over time for poles B and A. Bottom: energy of the peak in PSD over time, orange for pole B and blue for pole A

Signal test for acoustic fibre optics for the purpose of monitoring varying salinity

Signal test for acoustic fiber

optics for the purpose of

monitoring varying salinity

Signal test for acoustic fiber optics

for the purpose of monitoring

varying salinity

Contents

1

Management summary

2 Management samenvatting

3 Introduction

4 Method of measuring salt with fibre optics

5 Summary of experimental set-up

6 Processing

7 Results

a) all traces b) data with std <0.05 ms

c) data with std < 0.02 ms