www.nature.

com/scientificreports

OPEN Well‑scale demonstration

of distributed pressure sensing
using fiber‑optic DAS and DTS
Gerald K. Ekechukwu & Jyotsna Sharma*

In this study, we used data from optical fiber-based Distributed Acoustic Sensor (DAS) and Distributed
Temperature Sensor (DTS) to estimate pressure along the fiber. A machine learning workflow was
developed and demonstrated using experimental datasets from gas–water flow tests conducted in a
5163-ft deep well instrumented with DAS, DTS, and four downhole pressure gauges. The workflow
is successfully demonstrated on two experimental datasets, corresponding to different gas injection
volumes, backpressure, injection methods, and water circulation rates. The workflow utilizes the
random forest algorithm and involves a two-step process for distributed pressure prediction. In the
first step, single-depth predictive modeling is performed to explore the underlying relationship
between the DAS (in seven different frequency bands), DTS, and the gauge pressures at the four
downhole locations. The single-depth analysis showed that the low-frequency components (< 2 Hz)
of the DAS data, when combined with DTS, consistently demonstrate a superior capability in
predicting pressure as compared to the higher frequency bands for both the datasets achieving an
average coefficient of determination (or ­R2) of 0.96. This can be explained by the unique characteristic
of low-frequency DAS which is sensitive to both the strain and temperature perturbations. In the
second step, the DTS and the low-frequency DAS data from two gauge locations were used to predict
pressures at different depths. The distributed pressure modeling achieved an average ­R2 of 0.95 and
an average root mean squared error (RMSE) of 24 psi for the two datasets across the depths analyzed,
demonstrating the distributed pressure measurement capability using the proposed workflow.
A majority of the current DAS applications rely on the higher frequency components. This study
presents a novel application of the low-frequency DAS combined with DTS for distributed pressure
measurement.

Prediction of downhole pressures plays a vital role in a variety of applications including the management and
evaluation of petroleum, geothermal, and groundwater resources. Traditionally, pressure is measured using
gauges. While offering a cost-effective measurement solution, pressure gauges suffer from many limitations, such
as frequent calibration needs, low tolerance in harsh environments (such as high-temperature, high-pressure,
corrosive conditions, characteristic of petroleum and geothermal ­reservoirs1), hysteresis errors, and ability to
only provide pressure at the discrete gauge location measuring at pre-determined points (in other words, single-
point sensing).
Distributed fiber optics sensing (DFOS) is an advanced non-invasive, real-time sensing technology that can
overcome many limitations of traditional ­gauges2. Among other things, fiber optic sensors are insensitive to
electromagnetic noise, resistant to corrosion and high-pressure and high-temperature conditions, and do not
require electronics along the optical path, making them suitable for many downhole sensing applications. The
optical fiber functions both as the sensor and the channel to transmit the data, providing a truly distributed
measurement in real-time along the cable. These sensors are capable of measuring physical properties such as
temperature (via Distributed Temperature Sensing or DTS), vibration (via Distributed Acoustic Sensing or DAS),
and strain (via Distributed Strain Sensing or DSS), simultaneously along the entire fiber.
Although DAS and DTS have been used for a variety of applications, ranging from flow ­profiling2–5, fracture
­monitoring6–10, seismic ­measurement11–14, leak ­detection15,16 and others, the use of a combination of DAS and
DTS has not been explored for the prediction of downhole pressures. In this study, we develop and demonstrate
a machine learning and signal processing-assisted workflow for continuous real-time distributed measurement
of pressure using DAS and DTS data.

Department of Petroleum Engineering, Patrick F. Taylor Hall, LSU, Baton Rouge, LA 70803, USA. *email: JSharma@
LSU.edu

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 1

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 1.  The backscattered light spectrum generated from typical DFOS.

DFOS overview and low‑frequency DAS.  DFOS utilizes the optical time-domain reflectometry
(OTDR) principle to measure the spatial distribution of an intended m ­ easurand17. In the OTDR method, suc-
cessive short laser pulses are sent out of the transmitter into the fiber, causing interactions with the heteroge-
neous crystalline core of the fiber optic cable which acts as scattering sites sending the backscattered light or
signals back to the interrogation unit located at the surface. These impurities or molecular level heterogeneities
are inherently and unintentionally produced during the optical fiber manufacturing. The interactions between
the light and the scattering sites along the core of the fiber sense the temperature, vibration, and strain variation
originating from events in the vicinity of the fiber-optic cable. The spectrum of the backscattered light consists
of the Rayleigh, Brillouin, and Raman components (Fig. 1). Typically, the DTS uses the backscattered Raman
component to measure temperature, DAS uses the backscattered Raleigh component to measure vibrations or
dynamic strain, while DSS uses the Brillouin component to measure ­strain18–20.
Acoustic disturbance on a fiber generates microscopic elongation or compression of the fiber (micro-strain),
which causes a change in the phase relation and/or amplitude. The raw DAS data are usually delivered in the form
of delays in the optical phase [− π to + π] between two points along the fiber cable. The phase delay varies linearly
with a small length change (the axial strain) between two locations separated by the gauge length. Differences
in the strain obtained during successive pulses or the time differential of the measured optical phase (i.e., strain
rate) may also be provided as a signal response by some sensing unit providers. 21.
A majority of the DAS applications discussed in the previous section have focused on the high-frequency
bands (> 2 Hz) of the DAS data. The use of low-frequency DAS in the oil and gas industry is fairly recent and
only a few researchers have explored it for practical applications. For instance, Jin and ­Roy21 used the low-
frequency DAS data (< 0.05 Hz) to constrain the length, density, and width of reservoir fractures, while Becker
et al. 1 employed a low-frequency DAS system to detect strain at mHz frequencies. Shragge et al.22 used low-
frequency DAS for the acquisition of S-wave profiles. Technical improvements in optical interrogators have
enabled improved SNR close to DC frequencies, so it is expected that the application space of low-frequency
DAS will continue to grow. A unique characteristic of low-frequency DAS is that it is more sensitive to both the
temperature and the strain effects, and thus the dynamic strain rates are clearly o ­ bserved23. Thus, in principle,
the low-frequency DAS phase data has a more direct relationship with pressure, which is further discussed in
the next section and also demonstrated in the workflow presented in this study.

Theoretical background.  When light travels through a fiber of length L and refractive index n, the optical
phase ∅ is related to the wavenumber k by the following expression:
φ = nkL, where k = 2π/ (1)
Direct pressure exposure induces changes in a phase differential d∅/∅ which changes the properties of the
optical fiber. The changes in the optical phase induce strain ( dL/Lorεzg ) , modifies the index of refraction ( dn/n )
of the material (the photo-elasticity effect), and causes waveguide dispersion (dk/k) as shown ­below23,24:
d∅/∅ = dL/L + dn/n + dk/k (2)
24
Hocker showed that the third term representing waveguide mode dispersion effects is negligible. The phase
delay induces strain as shown below with dn/n now represented by the second term of the left side of the equa-
tion below 25,26.

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 2

Vol:.(1234567890)
www.nature.com/scientificreports/

dφ n2  
= εzg − (P11 + P12 )εrg + P12 εzg (3)
φ 2
where Pij is the strain optic (elastooptic or Pockels) coefficients and εzg and εrg are the axial and the radial com-
ponents of the induced strain in the fiber, respectively. Accounting for double transit, substituting Eq. 1 into
Eq. 3 and rearranging the equation, the strain sensitivity is given a­ s27:
εzg = dφ/2πnLξ (4)
where,
n2  
ξ =1− P12 − vg (P11 + P12 ) (5)
2
Equation (4) above shows the relationship between the pressure-induced strain and the phase differential.
­Budiansky23 then further provided an expression for pressure sensitivity due to the induced strain as follows:
 
1 − 2 1 − f vp − 2fvg
εzg /p = − (6)
fEg + (1 − f )Ep

The expression has been deemed accurate for f = (a/b)2 ≪ 1 and Ep ≪ Eg  . Where, a is the radius of the
fiber, b is the radius of the coating, Eg and vg are the Young’s modulus and the Poisson ratio of the glass, respec-
tively, while Ep and vp are the Young’s modulus and the Poisson ratio of the cladding, respectively. Hughes and
­Jarzynski28 showed that the sensitivity is generally governed by both the bulk and the Young’s moduli of the
coating materials, which are also temperature dependent. Furthermore, Giallorenzi et al26 observed that the
pressure sensitivity of coated fibers can be frequency-dependent and this dependency is affected by the combined
or synergistic effects of all the coatings and the fiber cable.
DAS measurements based on Rayleigh backscattering are temperature and strain dependent. However, the
way it affects the measurement is different for both strain and temperature. The strain affects the measurement by
directly changing the actual fiber length, but also through changes of the refractive index (photo-elastic effect).
The temperature affects the measurement, again through changes of the fiber length (thermal expansion), but
also through changes of the refractive index (thermo-optic effect). So at all times, the DAS measurement is
affected by both strain and temperature. However thermal changes typically have a response time much slower
than strain changes, and hence will have a much lower frequency ­content18. Jin and ­Roy29 also observed that
the low-frequency component of DAS signals are affected by thermal perturbations and proposed the following
relationship of low-frequency (LF) DAS with temperature and strain variations:
LFDAS = C1 �ε + C2 �T (7)
where �ε and T are the strain and the temperature variations, respectively. C1 and C2 are coefficients depend-
ent on the fiber structure and material ­properties30. C2 can be significantly dependent on the thermal expansion
coefficients of the entire multilayer structure of the fiber which could vary from fiber to fiber.
The above discussion establishes the physical dependencies between the pressure perturbations and the DAS-
based measurements. However, the above equations also highlight the complex and often non-linear relations that
depend on the fiber and the coating material properties (such as, thickness, elasticity, strain-optic coefficients etc.)
and the dynamic environmental conditions (such as temperature and frequency effects), which may not be fully
known analytically without assumptions or limitations. Machine learning algorithms have been demonstrated
to effectively “learn” the complex non-linear relationships between a given set of target prediction output and
input features. Thus, a machine learning approach was adopted in this study to directly learn the relationship
between pressure and the DAS and DTS measurements using the observed data.

Data acquisition
Experimental set‑up.  The data analyzed in this study was obtained from two-phase (nitrogen gas and
water) flow experiments conducted in a 5163-ft deep test-well located in the Petroleum Engineering Research
and Technology Transfer (PERTT) lab facility at LSU (Fig. 2a). The wellbore consists of a 9.625 in diameter cas-
ing that is cemented in place, with a 2.875 in diameter concentric tubing to 5025 ft. depth (Fig. 2b). DAS and
DTS fiber cables, along with four downhole pressure and temperature (P/T) gauges are attached to the outside
of the tubing as shown in Fig. 2b. The data acquisition parameters and sensor specifications are summarized in
Table 1.
The DAS was acquired at a frequency of 10 kHz, hence, the maximum frequency we can measure is 5 kHz
based on the Shannon-Nyquist criterion 31. To obtain the different frequency components of the signal, spectral
decomposition is performed on the raw DAS time-domain data by applying the Fast Fourier Transforms. The
frequencies are then split up into pre-specified bands consisting of different frequency ranges and then called
the frequency band energy (FBE) data. It is preferred to analyze the DAS data in the FBE domain as it provides
a simplified snapshot of the acoustic energy over a fixed duration and over different frequency ranges at any
given ­time23. FBE data is also much smaller in size as compared to the original time-domain DAS data, making it
easier to identify signals of importance and interpret vibration data only on those particular signals, leading to a
significant reduction in turnaround time for data i­ nterpretation23. For this study, we would analyze seven differ-
ent frequency bands as follows: Band-LF [0–2 Hz], Band-0 [2–5000 Hz], Band-1 [2–10 Hz], Band-2 [10–50 Hz],

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 3

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 2.  (a) DAS and DTS installation on the test-well at PERTT. (b) Test-well schematic showing the location
of the four downhole gauges and fiber-optic DAS and DTS 32.

DTS DAS Pressure and temperature gauge

Optical Mode Multi-mode Optical Mode Single-mode Pressure Accuracy < ± 1.0 psi
Spatial Resolution 3.28 ft Spatial Resolution 4.92 ft Pressure Resolution 0.035 psi
Sampling Interval 1.64 ft Sampling Interval 2.53 ft Temperature Accuracy  ± 0.27 ○F
Temporal Resolution 12 s Temporal Resolution 10 s Temporal Resolution 1 s
Range 15 km Range 16 km Maximum Pressure 10,000 psi
Accuracy  ± 1.8 °F Frequency 10 kHz Maximum Temperature 302 ○F

Table 1.  Data acquisition and sensor ­parameters32–34.

Dataset Water circulation rate [GPM] Gas injection volume [BBL] Injection method Backpressure [PSI]
1 0 2 Injection Line Choke closed
2 100 5 Tubing 300

Table 2.  Experimental parameters used for the two datasets.

Band-3 [50–200 Hz], Band-4 [200–500 Hz], and Band-5 [500–1000 Hz]. The acoustic energy contained in
frequency bands above 1000 Hz is insignificant in our specific dataset and they were therefore not a­ nalyzed22.

Experimental procedure and datasets analyzed.  Two-phase flow experiments using water and nitro-
gen gas were conducted in the test-well to understand gas–water flow dynamics at well-scale conditions. The
wellbore was initially filled with water in both the tubing and the casing, and a fixed volume of nitrogen gas
(measured in barrels or bbl) was injected either down the tubing or the 0.5 in diameter gas injection line strapped
to the tubing (as shown in Fig. 2b). The objective of the experiments was to observe and characterize the gas rise
in water using the fiber-optic sensors and downhole gauges as described in detail in the r­ eferences32,33.
Two different experimental datasets were used in this study to demonstrate the proposed distributed pressure
measurement workflow using DTS and DAS. The datasets correspond to different gas injection volumes, water
circulation rates, backpressure, and injection method, as summarized in Table 2. The injection sequence is as fol-
lows: preconditioning stage—which involves the injection of water down through the tubing and up the annulus
back to the surface with the objective of removing any leftover gas from the wellbore, gas injection stage—which
is the injection of a fixed volume of nitrogen gas slug either down the injection line or down the tubing, and the
post-injection or simply water circulation stage—in which water is once more injected to displace the gas in the
well. The objective was to demonstrate that the methodology works for different operational conditions. In the
first experiment (Dataset-1) nitrogen was injected through the gas line and allowed to rise to the surface through

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 4

Vol:.(1234567890)
www.nature.com/scientificreports/

Figure 3.  Waterfall plots showing the DAS values in the different frequency bands for Dataset-1.

the annulus without any water circulation and with the choke closed at the surface, while in the second experi-
ment (Dataset-2) the gas injection down the tubing is immediately followed by water injection to push the gas
down the tubing and eventually up through the annulus and back to the surface, while a constant backpressure
is maintained on the casing at the surface. Both two-phase flow scenarios will create some pressure disturbance
that will be recorded by the pressure gauges. The main objective of this work is to model the relationship between
the DAS and DTS values to the pressure gauge readings and then use the developed model to predict pressure
at different depths along the wellbore for distributed pressure measurement. The temperature data read by the
downhole gauges was only used for the DTS depth calibration.
Figures 3 and 4 show the DAS waterfall plots for Bands-LF, 0, 1, and 2 for Dataset-1 and Dataset-2, respec-
tively which indicate the gas flow signature in water which is observed more clearly in Dataset-1 (with no water
circulation) as compared to Dataset-2 (with water circulation). Plots for DAS Bands 3 to 5 are included in the
supplementary material (Figs. S1 and S2). A detailed interpretation of the gas signature can be found in the
­references33. Since the LFDAS data are sensitive to both dynamic strain and temperature changes, Band-LF (0 to
2 Hz) has both positive and negative numbers depending on whether the fiber section is experiencing compres-
­ henomenon32.
sive or tensile strain or heating or cooling p
Figures 5 and 6 show the pressure and temperature profiles for Dataset-1 and Dataset-2, respectively, at the
four depths where the gauges are installed (487 ft., 2023 ft., 3502 ft., and 5025 ft.). The pressure and temperature
data in Dataset-1 spanned a period of about 12 h while that of Dataset-2 was about 5 h. For Dataset-1, the effect
of gas was observed at about 2.5 h elapsed time, once the gas is in the annulus. The elapsed time is the time
difference between any given time and the reference time, where the reference time corresponds to the start of
the preconditioning stage described earlier. As expected, the pressures at the different gauges are lower for the
top gauge and increases as we go deeper into the well. This is a result of hydrostatic pressure which increases as
the depth increases. In addition, as the gas migrates out of the annulus to the surface we see that the pressure
increases since a lighter fluid (nitrogen) is being replaced by a heavier one (water) in the annulus. The gas rise
signature and arrival at surface are clearly observed in the DAS plots in Figs. 3 and 4. For Dataset-1, the decrease
in pressure at about 11.5 h of elapsed time was a result of opening of the choke valve at the surface which marks
the end of the experiment. For Dataset-2, which involves continuous water circulation post gas-slug injection,
additional pressure effect arises due to turbulence flow, which are also evident in the high vibration energy in
the DAS plots in Fig. 4. The maximum temperatures at the different gauges also showed an increasing trend
down the wellbore as expected from the geothermal gradient. The temperature readings in Dataset-2 are a few
degrees lower than Dataset-1 due to the cooling effect from the water circulation (at 100 GPM, see Table 2). In
Dataset-2 the pressures were more erratic than those for Dataset-1 due to the dynamic flow effects resulting
from the water circulation.

Data preparation.  One of the key steps in the data preparation was to align the downhole sensor data
spatially and temporally. As summarized in Table 1, DAS, DTS, and the pressure gauges had sampling times of
10 s, 12 s, and 1 s, respectively. While the DTS and DAS produced distributed measurements every 1.64 ft and

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 5

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 4.  Waterfall plots showing the DAS values in the different frequency bands for Dataset-2.

Figure 5.  Pressures from downhole gauges and temperature from DTS at the four gauge locations for
Dataset-1.

Figure 6.  Pressures from downhole gauges and temperature from DTS at the four gauge locations for
Dataset-2.

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 6

Vol:.(1234567890)
www.nature.com/scientificreports/

Figure 7.  Workflow for pressure prediction using DTS and DAS.

2.53 ft, respectively, along the fiber, the pressure gauges measured pressure at only four discrete locations (487
ft., 2023 ft., 3502 ft., and 5025 ft.). The downhole temperature gauge data in our case was only used for depth
calibration of the DTS. The first data preparation step was that the three different datasets had to be resampled to
ensure that they had the same sampling interval and corresponding timestamps. Therefore, in order to prepare
the data points to use in the machine learning model, the DAS and DTS were time-matched with a criterion that
the DTS is matched with the DAS if their timestamps are within ± 3 s apart. This is a reasonable criterion since
the temperature is not changing rapidly (Figs. 5 and 6). For some machine learning algorithms, the features or
input variables in the dataset need to be transformed via normalization. Normalization ensures fast convergence
of the gradient-based learning process, such as neural network models. Min–max scaling was performed on one
feature at a time to scale the data (yi) to [-1, 1] using the following equation:
′ yi − ymin
yi = 2
ymax − ymin
−1 (8)

The chosen machine learning method used in the main body of our work – the random forest—is robust and
its accuracy remains the same with or without normalization.

Methodology
Analysis workflow.  The analysis workflow developed in this study is illustrated in Fig. 7 and described as
follows:
(1) Data preparation DTS, DAS, and pressure gauge data is time and depth matched and normalized.

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 7

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 8.  Schematic illustration of the random forest algorithm.

(2) Single-depth analysis The machine learning model is implemented independently at the four gauge
depth (487 ft., 2023 ft., 3502 ft., and 5025 ft.). At each depth, the input features for the model are the DAS (one
frequency band at a time) and DTS data, while the target output variable is the change in pressure relative to the
initial pressure at the first time-step (∆P). 70% of the data were randomly selected for model training and the
remaining 30% is used for blind testing. The performance is evaluated for each frequency band individually to
select the one with the best performance for pressure prediction. This analysis is repeated at all four gauge depths
and all seven frequency bands, for the two experimental datasets.
(3) DAS frequency band selection The best performing frequency band is selected based on the single-depth
analysis at all four gauge locations, for both datasets. This frequency band is used for the distributed pressure
analysis.
(4) Distributed pressure analysis Here the objective is to predict pressure at different depths using the DAS
and DTS data. The machine learning model is trained using data at any two gauge depths and then blind-tested
for predicting the pressures at the other gauge depths different from the ones used for training. The input features
here are DTS, DAS (only the frequency band selected in (3), and elapsed time, and the target predicted is the
change in pressure (∆P).

Random forest algorithm.  Five different machine learning algorithms were considered for our workflow
including random ­forest35,36, gradient boosting machine (GBM)37,38, extreme gradient boosting (XGBoost)39,
support vector regression (SVR)40,41, and different architectures of shallow artificial neural network (ANN). Of
these, the random forest algorithm was selected as the model of choice in the main body of this work based on
the consistently high performance (high ­R2 and low RMSE) and low computational time when compared with
the other algorithms. The results of the comparison are presented in the supplementary material (Table S1 and
Fig. S3).
Random forest is an ensemble machine learning technique based on several decision trees 35,36 (Fig. 8). First,
the dataset needs to be split into training and testing or evaluation datasets. The training set is then sampled
randomly based on the number of decision trees to be trained. Each subset of the training set is further split into
training and validation datasets (otherwise known as out-of-bag or OOB samples). Each decision tree builds
its own model and uses the validation samples for evaluation. The decision tree model is a sequence of rules
based on the features (nodes) and splitting criteria. All input variables and possible split points are evaluated
and the split points that minimizes the cost function (mean squared error or MSE) across all training samples
and validation samples are selected. The cost function is calculated as:
n
1  2
Costfunction = yi − 
yi (9)
n
i=1

where yi and  yi are the actual target and predicted target values, respectively, and n is the number of samples.
Decision trees have several advantages in that they implicitly perform feature selection, they are not affected by
the non-linearity of the predictors and they are relatively easy to interpret. However, they suffer from high vari-
ance, that is if we split the data set into two parts at random and then try to train on them, the results could be
very different. Hence, in order to build a model with low variance and better accuracy, the ensemble approach
is used to combine several decision tree models to obtain a stronger model. The ensemble methods usually
involve creating multiple different subsets from the training data, building multiple predictive models, and then

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 8

Vol:.(1234567890)
www.nature.com/scientificreports/

Predictor Target Number of datapoints

Random Forest Model (Input variables) (Output variable) (Training/Testing data)
2668/1144 (Dataset-1)
Single-depth analysis DAS (all frequency bands), DTS ∆P
553/238 (Dataset-2)
7624/7624 (Dataset-1)
Distributed depth analysis DAS (Band-LF), DTS, Time ∆P
1582/1582 (Dataset-2)

Table 3.  Machine learning model features for the single-depth and distributed pressure modeling.

Figure 9.  Cross-plots of input features DTS and different DAS frequency bands for Dataset-1.

combining the predictions. The random forest employed in this work is based on the bootstrap aggregation or
bagging for ­short35. Bagging involves bootstrapping the training data to get subsets, learn one model for each
set which is usually run in parallel, and then average the model prediction.
The most important hyperparameters that need to be considered in the random forest modeling procedure
are as follows:

Number of trees: This is the number of trees that are used in the algorithm. The number of decision trees
used in this study was 100 based on a parametric study that showed no appreciable improvement in the
performance scores beyond this value.
Splitting criteria: The mean square error or MSE was used as the splitting criteria.
Stopping criteria: This can be specified by either the maximum depth of each tree or the minimum samples
required to split an internal node. If the maximum depth is specified, then the splitting stops after the speci-
fied value is reached otherwise if the nodes are expanded until all leaves are pure or until all leaves contain
less than minimum samples for a split. For this study, there was no performance improvement beyond a
maximum depth of 10.
Minimum Sample Split: The minimum samples required to split an internal node is 2 and the minimum
number of samples to be in a leaf node is 1.

Performance metrics.  We employed the coefficient of determination or R-squared ­(R2) and the root mean
squared errors (RMSE) to quantify the performance of our models. These performance metrics are robust
enough to give the relative performance across the different scenarios and have been widely used in machine
learning model performance assessment. They are calculated as:

n  2  n
SSR i=1 y
i − y 1  2
2 and RMSE = 
2
R = = n  yi − 
yi (10)
SST n
i=1 yi − y i=1

yi is the predicted value, yi is the actual target value, y is the mean of the target values, and n is the number
where 
of samples. SSR is the "regression sum of squares" and quantifies how far the estimation is from the target feature
mean prediction (based on no relationship with predictors). SST is the "total sum of squares" and quantifies how
much the data point vary around their mean.

Results
Table 3 summarizes the input and output features used for the machine learning models for the single-depth
and distributed pressure modeling steps. We first present the results of the exploratory data analysis of the input
and output variables to identify patterns and data distributions. This is followed by the discussion of results
from the random forest models for the single-depth analysis at each gauge depth and the distributed pressure
prediction scenarios.

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 9

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 10.  Cross-plots of input features DTS and different DAS frequency bands for Dataset-2.

Figure 11.  Cross-plots of pressure and different DAS frequency bands for Dataset-1.

Figure 12.  Cross-plots of pressure and different DAS frequency bands for Dataset-2.

Descriptive data exploration.  Figures 9 and 10 show the cross-plots between DAS and DTS for the DAS
frequency Bands-LF, 0, and 1 for Dataset-1 and Dataset-2, respectively. The high-frequency Bands 0 and 1 show
a similar DAS-DTS relationship which is distinct from the low-frequency DAS data in Band-LF. For example,
Fig. 9a shows a linear DAS-DTS relationship which is not seen for Figs. 9b and c. For the higher frequency bands
in both datasets, the relationship between DTS and DAS cannot be clearly explained. Figures 11 and 12 show the
cross-plots between DAS and pressure at various depths for Datasets-1 and Dataset-2, respectively. Again, the
higher frequency DAS Bands 0 and 1 show a similar relationship with pressure, while the DAS Band-LF shows
a more distinct trend. The exploratory data analysis demonstrates the unique properties of low-frequency DAS
which can be attributed to the sensitivity of DAS to temperature and strain variations at low f­ requency1, 23.

Single‑depth predictive modeling.  In this section, the results of the single-depth analysis are discussed
for both datasets at the four pressure gauge locations. The DAS and DTS were used as the input features while
the change in pressure with respect to initial pressure at the first time-step (∆P) was used as the output variable
(Table 3). Figure 13 shows the RMSE and R ­ 2 values for the pressure predictions of the testing sets of Dataset-1
at the four gauge locations for the seven DAS frequency bands. Figure 14 shows the predicted and the actual
pressure profiles for the testing subsample for Dataset-1 for Bands-LF, 0, and 1. The predicted pressure profiles
for Bands 2 to 5 for both the datasets are included in the supplementary material (Figs. S4, S5). The ­R2 of the
Band-LF ranged from 0.90 to 0.99 with an average performance across all depths of 0.97, while the average ­R2
value across all depths for the higher frequency bands ranged from 0.81 to 0.83. Similarly, the RMSE values

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 10

Vol:.(1234567890)
www.nature.com/scientificreports/

Figure 13.  The RMSE and ­R2 values for the testing subsample at each gauge depth for Dataset-1.

varied between 0.8 to 11.4 psi for the Band-LF and between 8 to 23 psi for the higher frequency bands. The
average RMSE across all four depths was 4.7 psi for Band-LF compared to 14.2 psi average RMSE for the higher
frequency bands. The results clearly demonstrate that the low-frequency DAS data gives a more accurate predic-
tion for pressure for Dataset-1.
Figure 15 shows the results from single-depth pressure prediction for Dataset-2 across the four gauge depths
for the testing subsample. The ­R2 values for the Band-LF ranged from 0.90 to 0.96 with an average of 0.94 across
all depths. While the average ­R2 values for the higher frequency DAS bands ranged from 0.64 to 0.75. Similarly,
the RMSE values varied between 6.7 to 11.3 psi for the Band-LF and between 12.3 to 39.5 psi for the higher
frequency bands. Figure 16 compares the predicted and the actual pressure trends for the testing subsample for
Dataset-2. Similar to Dataset-1, Figs. 15 and 16 clearly demonstrate that the random forest model using low-
frequency DAS as input gives a more accurate prediction of pressure.

Distributed pressure predictive modeling.  The results from the single-depth pressure modeling clearly
established that the low-frequency DAS (or Band-LF) gave a consistently better performance compared to the
higher frequency DAS bands (> 2 Hz). Therefore, for the distributed pressure modeling, we used the input vari-
ables of DAS Band-LF, DTS, and elapsed time as the input features, and the change in pressure from the original
(ΔP) as the output for the random forest model. Training of the model was performed with datasets from any
two gauge depths while the resulting model was used to predict pressures at the other two depths. Figure 17a–h
show the predicted versus the actual pressure plots for eight different scenarios, for Dataset-1 and Datset-2,
respectively. For example, in Fig. 17a the DTS, DAS and pressure data at 487 ft and 2023 ft were used for training
the random forest model and the trained model was used to predict pressures at 3502 ft using the DAS Band-LF
and DTS at that depth. The ­R2 values for Dataset-1 for all eight scenarios were higher than 0.99 with the RMSE
ranging between 2.5 to 4.2 psi. Similarly, Dataset-2 results in Fig. 18a–h show strong model performance with
­R2 greater than 0.95 in all cases. Although for Dataset-2 the RMSE was higher (44.7 psi) compared to Dataset-1,
which is likely due to the dynamic effects resulting from water circulation.

Discussion
Prediction of downhole pressure is crucial for wide-ranging potential applications including the management
and evaluation of petroleum, geothermal, and groundwater resources. For oil operators, downhole pressure
monitoring supports the determination of well productivity, estimation of flow rates, and sizing of surface and
downhole equipment. The industry primarily relies on downhole and surface gauges to meet its pressure data
needs, however, this often results in a deficiency of crucial data, due to the low spatial and temporal resolution
achieved from gauges which only provide measurement at a handful of locations. Distributed pressure meas-
urement simultaneously along the entire wellbore in real-time will give the operators and drillers never-before-
seen visibility of the dynamics of fluid flow along the well and may reduce exposure to incidents and improve
reservoir management.

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 11

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 14.  Pressure prediction using DAS Bands LF, 0, and 1 for the testing set for Dataset-1 (RMSE in psi).

Although the adoption of DAS and DTS is increasing rapidly, well-scale or field-scale distributed pressure
sensing has not been reported using these measurements. This study presents the first well-scale application of
fiber optics data for pressure prediction. To model the pattern in the data, we have trained a machine learning
algorithm and then used the developed model to predict the pressure data at different depths. In a typical oilfield
scenario, surface and downhole pressure gauges are commonly available which can be used for the model train-
ing, in conjunction with the DTS and DAS (which includes low-frequency data). The trained model can then
be used for the distributed pressure prediction at various locations where DTS and DAS data is available. While
data from two downhole gauge locations was used for training the model to demonstrate the workflow, even a
single pressure gauge data may also be used for model training if that is the only pressure measurement available
with co-located DAS and DTS information. 
The single-depth analysis showed that the low-frequency DAS (combined with DTS) consistently demon-
strated superior capability to predict pressure compared to the higher frequency DAS (> 2 Hz). A plausible expla-
nation for the better performance shown by the Band-LF (0–2 Hz) is the higher DAS sensitivity to temperature
and strain in the low-frequency range. The pressure response to the fluid compression in turn is related to the
longitudinal strain experienced by the fiber through the mechanical properties of the fiber (as discussed in Sec.

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 12

Vol:.(1234567890)
www.nature.com/scientificreports/

Figure 15.  The RMSE and ­R2 values for the testing subsample at each gauge depth for Dataset-2.

1.2). The results are consistent with those from some recent studies that have also shown that low-frequency DAS
gives a better correlation with pressure. For instance, Becker et al.1 in their lab-scale experiment showed that the
low-frequency DAS band (100 mHz) showed higher sensitivity to fluid pressure. In our well-scale experiments,
the pressures investigated were up to 3200 psi whereas the maximum pressure in the Becker study was less than
1 psi. This study demonstrates that low-frequency DAS combined with DTS can be used for distributed pressure
measurement at well-scale.

Conclusions
This study presents the first well-scale application of distributed fiber-optic data for pressure prediction. The
complex relationship between DAS, DTS, and pressures were modeled by training a machine-learning algorithm
and the developed model was used to predict the pressure at different depths in a 5163 ft. deep weelbore. In a
typical oilfield scenario, surface and downhole pressure gauges are commonly available, which can be used for
model training, and subsequently the trained model can be used to predict pressure at different spatial loca-
tions where DAS and DTS information is available. The results demonstrate the frequency dependence of the
pressure measured by the optical fiber. The low-frequency DAS components (< 2 Hz), together with DTS gave

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 13

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 16.  Pressure prediction using DAS Bands LF, 0, and 1 for the testing set for Dataset-2 (RMSE in psi).

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 14

Vol:.(1234567890)
www.nature.com/scientificreports/

Figure 17.  Comparison of the predicted and the actual pressure profiles obtained for the distributed pressure
analysis for different combinations of training and testing depths for Dataset-1 (RMSE in psi).

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 15

Vol.:(0123456789)
 www.nature.com/scientificreports/

Figure 18.  Comparison of the predicted and the actual pressure profiles obtained for the distributed pressure
analysis for different combinations of training and testing depths for Dataset-2 (RMSE in psi).

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 16

Vol:.(1234567890)
www.nature.com/scientificreports/

more accurate pressure predictions, as compared to the high-frequency DAS components. In the single-depth
­ 2) was about 0.96 and the aver-
analysis with low-frequency DAS, the average coefficient of determination (or R
age RMSE was about 7 psi, for the two datasets analyzed. For the distributed pressure analysis the average ­R2
was over 0.95 and average RMSE was 24 psi for the two datasets for all depths analyzed, demonstrating strong
model performance. This study presents a novel application of the low-frequency DAS combined with DTS for
distributed pressure measurement at well-scale.

Received: 29 December 2020; Accepted: 31 May 2021

References
1. Becker, M., Coleman, T., Ciervo, C., Cole, M. & Mondanos, M. Fluid pressure sensing with fiber-optic distributed acoustic sensing.
Lead. Edge 36, 1018–1023 (2017).
2. Johannessen, K., Drakeley, B. K. & Farhadiroushan, M. Distributed Acoustic Sensing - A New Way of Listening to Your Well/
Reservoir. in SPE Intelligent Energy International (Society of Petroleum Engineers, 2012). https://​doi.​org/​10.​2118/​149602-​MS
3. Chaudhuri, A., Osterhoudt, C. F. & Sinha, D. N. An Algorithm for Determining Volume Fractions in Two-Phase Liquid Flows by
Measuring Sound Speed. J. Fluids Eng. 134, (2012).
4. Paleja, R. et al. Velocity Tracking for Flow Monitoring and Production Profiling Using Distributed Acoustic Sensing. in SPE Annual
Technical Conference and Exhibition (Society of Petroleum Engineers, 2015). https://​doi.​org/​10.​2118/​174823-​MS
5. Xiao, J. et al. Dynamic Water Injection Profiling in Intelligent Wells Using Distributed Acoustic Sensor with Multimode Opti-
cal Fibers. in SPE Annual Technical Conference and Exhibition (Society of Petroleum Engineers, 2015). https://​doi.​org/​10.​2118/​
174865-​MS
6. Bakku, S. K., Fehler, M. & Wills, P. Monitoring hydraulic fracturing using distributed acoustic sensing in a treatment well. in SEG
Technical Program Expanded Abstracts 2014 5003–5008 (Society of Exploration Geophysicists, 2014). https://​doi.​org/​10.​1190/​
segam​2014-​1280.1
7. Denney, D. Distributed acoustic sensing for hydraulic- fracturing monitoring and diagnostics. J. Pet. Technol. 64, 68–74 (2012).
8. MacPhail, W. F. P., Lisoway, B. & Banks, K. Fiber Optic Distributed Acoustic Sensing of Multiple Fractures in a Horizontal Well.
in SPE Hydraulic Fracturing Technology Conference (Society of Petroleum Engineers, 2012). https://​doi.​org/​10.​2118/​152422-​MS
9. Martinez, R., Hill, A. D. & Zhu, D. Diagnosis of Fracture Flow Conditions With Acoustic Sensing. in SPE Hydraulic Fracturing
Technology Conference (Society of Petroleum Engineers, 2014). https://​doi.​org/​10.​2118/​168601-​MS
10. Molenaar, M. M. & Cox, B. E. Field Cases of Hydraulic Fracture Stimulation Diagnostics Using Fiber Optic Distributed Acoustic
Sensing (DAS) Measurements and Analyses. in SPE Unconventional Gas Conference and Exhibition (Society of Petroleum Engineers,
2013). https://​doi.​org/​10.​2118/​164030-​MS
11. Nizkous, I., Gerritsen, S., Kiyashchenko, D. & Joinson, D. Distributed Acoustic Sensing (DAS) VSP for Imaging and Velocity Model
Building. in International Petroleum Technology Conference (International Petroleum Technology Conference, 2015). https://​doi.​
org/​10.​2523/​IPTC-​18483-​MS
12. Ugueto, G. A. et al. Application of Integrated Advanced Diagnostics and Modeling To Improve Hydraulic Fracture Stimulation
Analysis and Optimization. in SPE Hydraulic Fracturing Technology Conference (Society of Petroleum Engineers, 2014). https://​
doi.​org/​10.​2118/​168603-​MS
13. Webster, P., Wall, J., Perkins, C. & Molenaar, M. Micro-seismic detection using distributed acoustic sensing. in SEG Technical Pro-
gram Expanded Abstracts 2013 2459–2463 (Society of Exploration Geophysicists, 2013). https://d ​ oi.o
​ rg/1​ 0.1​ 190/s​ egam2​ 013-0​ 182.1
14. Karrenbach, M. et al. DAS Microseismic Monitoring and Integration With Strain Measurements in Hydraulic Fracture Profiling.
in Proceedings of the 5th Unconventional Resources Technology Conference (American Association of Petroleum Geologists, 2017).
https://​doi.​org/​10.​15530/​urtec-​2017-​26707​16
15. Boone, K., Ridge, A., Crickmore, R. & Onen, D. Detecting Leaks in Abandoned Gas Wells with Fibre-Optic Distributed Acoustic
Sensing. in International Petroleum Technology Conference (International Petroleum Technology Conference, 2014). https://​doi.​
org/​10.​2523/​IPTC-​17530-​MS
16. Thodi, P., Paulin, M., Forster, L., Burke, J. & Lanan, G. Arctic Pipeline Leak Detection using Fiber Optic Cable Distributed Sensing
Systems. in OTC Arctic Technology Conference (Offshore Technology Conference, 2014). https://​doi.​org/​10.​4043/​24589-​MS
17. Hill, D. Distributed Acoustic Sensing (DAS): Theory and Applications. in Frontiers in Optics 2015 FTh4E.1 (OSA, 2015). https://​
doi.​org/​10.​1364/​FIO.​2015.​FTh4E.1
18. Hartog, A. An Introduction to Distributed Optical Fibre Sensors. (CRC Press Taylor & Francis Group, 2017).
19. Miah, K. & Potter, D. A review of hybrid fiber-optic distributed simultaneous vibration and temperature sensing technology and
its geophysical applications. Sensors 17, 2511 (2017).
20. Latal, J. et al. Rock massif temperature changes measurement with regard to thermal responses generated by a thermal response
test device. in (eds. Baldini, F., Homola, J. & Lieberman, R. A.) 877416 (2013). https://​doi.​org/​10.​1117/​12.​20172​41
21. Jin, G. & Roy, B. Hydraulic-fracture geometry characterization using low-frequency DAS signal. Lead. Edge 36, 975–980 (2017).
22. Shragge, J. et al. Low-frequency ambient Distributed Acoustic Sensing (DAS): Useful for subsurface investigation? in SEG Technical
Program Expanded Abstracts 2019 963–967 (Society of Exploration Geophysicists, 2019). https://d ​ oi.o
​ rg/1​ 0.1​ 190/s​ egam2​ 019-3​ 2164​
79.1
23. Budiansky, B., Drucker, D. C., Kino, G. S. & Rice, J. R. Pressure sensitivity of a clad optical fiber. Appl. Opt. 18, 4085 (1979).
24. Hocker, G. B. Fiber-optic sensing of pressure and temperature. Appl. Opt. 18, 1445 (1979).
25. Bucaro, J. A., Dardy, H. D. & Carome, E. F. Fiber-optic hydrophone. J. Acoust. Soc. Am. 62, 1302–1304 (1977).
26. Giallorenzi, T. et al. Optical fiber sensor technology. IEEE J. Quantum Electron. 18, 626–665 (1982).
27. SEAFOM. Measuring Sensor Performance Document – 02 (SEAFOM MSP-02). (2018).
28. Hughes, R. & Jarzynski, J. Static pressure sensitivity amplification in interferometric fiber-optic hydrophones. Appl. Opt. 19, 98
(1980).
29. Jin, G., Krueger, K. & Roy, B. Low Frequency DAS Well Interference Evaluation. (2019).
30. Jin, G. et al. Novel Near-Wellbore Fracture Diagnosis for Unconventional Wells Using High-Resolution Distributed Strain Sensing
during Production. SPE J. (2021). https://​doi.​org/​10.​2118/​205394-​PA
31. Oshana, R. Overview of Digital Signal Processing Algorithms. in DSP Software Development Techniques for Embedded and Real-
Time Systems 59–121 (Elsevier, 2006). https://​doi.​org/​10.​1016/​B978-​07506​7759-2/​50006-5
32. Sharma, J., Cuny, T., Ogunsanwo, O. & Santos, O. Low-Frequency Distributed Acoustic Sensing for Early Gas Detection in a
Wellbore. IEEE Sens. J. 21, 6158. https://​doi.​org/​10.​1109/​JSEN.​2020.​30387​38 (2020).
33. Sharma, J., Santos, O. L. A., Feo, G., Ogunsanwo, O. & Williams, W. Well-scale multiphase flow characterization and validation
using distributed fiber-optic sensors for gas kick monitoring. Opt. Express 28, 38773 (2020).

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 17

Vol.:(0123456789)
 www.nature.com/scientificreports/

34. Feo, G., Sharma, J., Kortukov, D., Williams, W. & Ogunsanwo, T. Distributed fiber optic sensing for real-time monitoring of gas
in riser during offshore drilling. Sensors 20, 267 (2020).
35. Breiman, L. Random forests. Mach. Learn. 45, 5–32 (2001).
36. Breiman, L., Friedman, J. H., Olshen, R. A. & Stone, C. J. Classification And Regression Trees. (Routledge, 2017). https://​doi.​org/​
10.​1201/​97813​15139​470
37. Friedman, J. H. Machine. Ann. Stat. 29, 1189–1232 (2001).
38. Friedman, J. H. Stochastic gradient boosting. Comput. Stat. Data Anal. 38, 367–378 (2002).
39. Chen, T. & Guestrin, C. XGBoost. in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and
Data Mining 785–794 (ACM, 2016). doi:https://​doi.​org/​10.​1145/​29396​72.​29397​85
40. Cortes, C. & Vapnik, V. Support-vector networks. Mach. Learn. 20, 273–297 (1995).
41. Vapnik, V. The Nature of Statistical Learning Theory. (Springer, 1995).

Author contributions
J.S conceptualized and supervised this study; G.E. conducted the machine learning modeling; J.S. and G.E.
analyzed the data. Both authors wrote the manuscript and the supplementary material.

Funding
National Academy of Sciences’ Gulf Research Program, and the DOE Israel-U.S. Binational Industrial Research
and Development (BIRD) Foundation.

Competing interests 
The authors declare no competing interests.

Additional information
Supplementary Information The online version contains supplementary material available at https://​doi.​org/​
10.​1038/​s41598-​021-​91916-7.
Correspondence and requests for materials should be addressed to J.S.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note  Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access  This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.

© The Author(s) 2021

Scientific Reports | (2021) 11:12505 | https://doi.org/10.1038/s41598-021-91916-7 18

Vol:.(1234567890)