IEEE Signal Processing - March 2018 - 101

origin time of the event is unknown, and,
azimuth may be determined directly from the
Signal processing of
therefore, one must carry out migrationP-wave because the particle polarization is
data from receiver arrays
based imaging in time, as well as space,
parallel to the raypath direction for P-waves
plays a key role in the two i.e., four-dimensional (4-D) imaging. A
in an isotropic medium. The accuracy of this
microseismic monitoring
historical overview of the development of
measurement as well as determination of
techniques in use today:
this method is given in [11] and originated
back azimuth from higher SNR S-waves is
from studies of natural seismicity. In some
discussed in [14]. Monitoring from multiple
downhole and surface
cases, the migration-based approach is
boreholes is becoming more common, but it
monitoring.
viable for downhole monitoring, e.g., [18],
rarely solves the issue-only a small fraction
while picking and triangulation can be used
of the events can be observed on two or more
(for stronger microseismic events) in the surface monitorwells. In fact, the most challenging limitation is the availability
ing scenario.
of additional downhole monitoring wells, since drilling a dediThe most significant challenge in surface microseismic
cated monitoring well is expensive and wells previously used for
monitoring is the difficulty of directly observing P- and S-wave
production or exploration may not be suitable. Figure 2 illustrates
signals. Surface monitoring relies on thresholding the stacked
an example of a microseismic processing result from a downhole
traces to detect microseismic events. Adapting the threshold
monitoring array. The microseismic events are mapped as points
value to SNR according to binary hypothesis testing allows us
representing the source locations and mapping the injected
to balance false positives versus missed detections. Assumpfluid penetration into the rock formation.
tions about the correct noise model for surface monitoring cast
The example is a microseismic data set acquired during
some doubt on this approach, which requires an estimate of
hydraulic fracturing of the Lower Silurian and Ordovician Forthe noise. In reality, some sedimentary basins are favorable
mations in the Baltic Basin margin, Lubocino, North Poland
for surface monitoring, as they do not severely attenuate and
[35]. Two wells were drilled at this site: one vertical and one
their near-surface conditions enlarge signal due to the impedhorizontal. The horizontal section of the stimulated well was
ance contrast and free surface boundary condition [9]. Meandivided into multiple sections called stages, and each was
while, some basins severely attenuate the signal resulting in
hydraulically stimulated separately. The microseismic events
unreliable detections [35].
show that stages remained mostly separated and did not interfere with one another.
To increase the SNR by constructive superposition, stackFailure mode, source-mechanisms, and magnitudes
ing the traces from a large number of receivers is a well-known
Most signal processing for microseismicity is targeted mainmethod in active seismic, where it is used for migration-based
ly at obtaining locations for the stress-induced events along
imaging of subsurface structures. In active seismic, the source
with a geometrical interpretation of these "dots in the box."
of a seismic signal generated by explosions, mechanical vibraHowever, the amplitude and polarity of the recorded waves
tors, or airguns (for marine seismic) is controlled during the
contain additional information about the source mechanisms
data collection. However, in passive seismic monitoring, the
of the events. These source mechanisms can represent

-200
-250
-2,700

-350
-400
-500
-550

-2,800
-2,850

-600
-650
-700
-300

Stages 1-6
Located Microseismic Events

-2,750

-450

Depth (m)

Northing (m)

-300

-200 -100
Easting (m)
(a)

-2,900
-200

-250 -300 -350 -400 -450 -500 -550 -600 -650
-700
Northing (m)
(b)

0
-
-2 10
-3 00 0 )
00
(m
ng
sti
a
E

Figure 2. (a) A map view and (b) an oblique view of microseismic events located in the calibrated anisotropic velocity model. Events are colored according to stages. The true perforation intervals are shown as corresponding diamonds (Figure used with permission from [35]).

IEEE Signal Processing Magazine

March 2018

101

Table of Contents for the Digital Edition of IEEE Signal Processing - March 2018