Professional Documents
Culture Documents
FA S T T R A C K PA P E R
Accepted 2009 June 30. Received 2009 June 29; in original form 2008 March 18
SUMMARY
Hydraulic fracture stimulation is frequently performed in hydrocarbon reservoirs and geother-
mal systems to increase the permeability of the rock formation. These hydraulic fractures are
often mapped by hypocentres of induced microearthquakes. In some cases microseismicity
exhibits asymmetry relative to the injection well, which can be interpreted by unequal con-
ditions for fracture growth at opposite sides of the well or by observation effects. Here we
GJI Volcanology, geothermics, fluids and rocks
investigate the role of the lateral change of the minimum compressive stress. We use a simple
model to describe the relation among the lateral stress gradient, the mean viscous pressure
gradients in the fracture wings, the fracture geometry, and the net pressure in the fracture.
Our model predicts a faster fracture growth in the direction of decreasing stress and a limited
growth in the opposite direction. We derive a simple relationship to estimate the lateral stress
gradient from the injection pressure and the shape of the seismic hypocentre cloud. The model
is tested by microseismic data obtained during stimulation of a Canyon Sands gas field in
West Texas. Using a maximum likelihood method we fit the parameters of the asymmetric
fracture model to the space–time pattern of hypocentres. The estimated stress gradients per
metre are in the range from 0.008 to 0.010 times the bottom-hole injection overpressure
(8–10 kPa m−1 assuming the net pressure of 1 MPa). Such large horizontal gradients in the
order of the hydrostatic gradient could be caused by the inhomogeneous extraction of gas
resulting in a lateral change of the effective normal stress acting normal to the fracture wall.
Key words: Downhole methods; Geomechanics; Fracture and flow; Earthquake source
observations.
Journal compilation
C 2009 RAS
Asymmetric hydraulic fractures as a result of driving stress gradients 635
2 A S Y M M E T R I C H Y D RO F R AC T U R E
I N C A N YO N S A N D S
The asymmetry observed by Fischer et al. (2008, see Fig. 1) was
manifested both during the injection phase in the different length
and speed of growth of the fracture wings, and also after shut-in by Figure 1. Distribution of microseismicity accompanying hydraulic stimula-
continued growth in one direction only. This is illustrated in Fig. 2 tion of Canyon Sands formation in Western Texas. Different colours indicate
that shows the time dependence of the position of microearthquake events belonging to different injection stages 1, 2, 3 and 4 stimulated from
hypocentres along the extension of the hydrofracture. These plots injection points indicated by black rectangles. Note the highly asymmetric
clearly show that during injection period the fracture extended faster fracture, in particular in the depth layers A and B.
in the direction of positive coordinates (NE direction in the map view
of Fig. 1) and after shut-in continued growing only to NE. Because
ever, the maximum SNR, which is proportional to the maximum
the asymmetry, namely of the after-shut-in growth, occurred in all
observed amplitude, should decrease with increasing hypocentral
stimulated depth layers in the whole depth range of 250 m, it cannot
distance and converge to the minimum SNR. As we do not observe
be explained by a local scale heterogeneity of the rock stiffness or
this convergence at the shorter (left) fracture tip in any depth layer,
permeability in one sediment layer. Also an unfavourable orientation
we can rule out the influence of the different distances between the
of the perforations shots appears rather unlikely to explain the same
fracture tips and the monitoring system on the observed fracture
asymmetry in the different depth levels, because the orientation
asymmetry.
of the perforation guns was not controlled during the treatment
As an alternative explanation we suggest that the fracture asym-
and thus uncorrelated in the different layers. Furthermore, the guns
metry was caused by a gradient of the stress driving the fracture
with shots covering the whole circle in 60◦ intervals were used
growth. In the next section, we describe the fracture model includ-
(J. Le Calvez, 2008, personal communication) leading to an almost
ing the effect of stress gradient; the possible sources of the gradient
preferable orientation in each layer.
are discussed in Section 4.
To check whether the asymmetry is not an artefact of the mon-
itoring conditions, we determined the signal-to-noise-ratio (SNR)
of events along the fracture strike (Fig. 3). Because the minimum
3 MODEL
SNR is proportional to the minimum detectable signal at the mon-
itoring array, it is independent of hypocentral distance and should Our model is limited to the period of fluid injection when the frac-
be approximately constant. This is in agreement with the measured ture continues growing. It is based on the following simplifying
minimum SNR in Fig. 3 ranging mostly between 10 and 15 dB. How- assumptions.
C 2009 The Authors, GJI, 179, 634–639
Journal compilation
C 2009 RAS
636 T. Fischer, S. Hainzl and T. Dahm
200
along fracture from injection [m]
150
100
50
−50
−100
−150
0 10 20 30 40 50 60 70 80
time since perforation [min]
C 2009 The Authors, GJI, 179, 634–639
Journal compilation
C 2009 RAS
Asymmetric hydraulic fractures as a result of driving stress gradients 637
t t
a1 = −v1 (τ ) d(τ ) and a2 = v2 (τ ) d(τ ), and obtain a time-
0 0
dependent solution by means of integration of eq. (4) as
t t t t
g g
net v1 (τ )a dτ − v1 (τ ) dτ = net v2 (τ )a dτ + v2 (τ ) dτ
p0 0 0 p0 0 0
g a1 (t)2 g a2 (t)2
− net
+ a1 (t) = net + a2 (t).
p0 2 p0 2
(5)
150 150
a1, m
a1, m
100 100
50 50
0 0
0 50 100 150 200 0 50 100 150 200
a ,m a ,m
2 2
Figure 5. Relation between the lengths of hydrofracture wings a1 and a2 for different ratios g/ p net 0 of the stress gradient g and injection net pressure
p net net 0.25 . Assuming p net = 1 MPa the different values of g/ p net
0 . While in (a) the injection net pressure p 0 is kept constant, in (b) it is increasing with time as t 0 0
correspond to the stress gradient g of 1, 3 and 10 kPa m−1 .
√
hydraulic fracture (Economides & Nolte 2003, pp. 6–4). In this case a = 2hw/q and b = 4 2hC/q (Economides & Nolte 2003; Shapiro
the growth of the smaller wing is faster. et al. 2006). Here h and w are the average height and width of the
fracture, C the leakoff coefficient, and t0 the initiation time of the
hydrofracture.
4 A P P L I C AT I O N T O T H E C A N Y O N
We use this parametrization to fit the observed hydrofracture
S A N D S D ATA S E T A N D D I S C U S S I O N
growth a2 (t). According to our model, the expected temporal growth
We test our model predictions of the asymmetry by means of the of the shorter fracture wing a1 (t) is then directly determined from
microseismic data set acquired during a hydraulic fracture treatment a2 (t) and the parameter g/ p net
0 using eq. (7). Thus our model consists
of a tight gas reservoir in Canyon Sands, West Texas (see Figs 1 of three parameters (a, b and g/ p net
0 ) and an unknown hydrofracture
and 2). initiation point x0 and time t0 . As input data we use the positions of
For the following analysis, we assume that the seismicity clouds microearthquake hypocentres along the strike of the hydrofracture
match approximately the underlying hydrofracture geometry and as a function of the earthquake origin times. We simultaneously
that pnet
0 was constant during injection. To determine the time- optimize all five model parameters by means of a grid-search using
dependence of the lengths of the hydrofracture wings a1 (t) and a maximum likelihood method.
a2 (t), we apply the methodology described in our previous paper The results for the depth layers A and B stimulated by the in-
(Fischer et al. 2008). Here, we give only a brief description. Based on jections during the stages 2, 3 and 4 in which the asymmetry is
the mass balance equation for fluids, the wing-length L of a straight most pronounced are given in Table 1; the error intervals of g refer
planar height-fixed fracture (PKN model) is given as a function of to a 86 per cent decrease of the likelihood function. Fig. 6 shows
√
time t according to L(t) = (t − t0 )/(a + b t − t0 ) with constants the fits of the seismicity evolution. It appears that our model fits the
Figure 6. The time dependence of the positions of microearthquake hypocentres along the strike direction of the hydrofracture (dots) for different depth
intervals and injection stages: stage 2 injection in layer B (a), stage 3 injection in layer B (b), and stage 4 injection in layer A. The horizontal grey bars indicate
the injection period for which the fit of the asymmetric rupture growth was determined. The parameters of the different curves are given in Table 1; thin lines
refer to the lower and upper bound of the estimated g-value. In (a), the dashed lines indicate the trend after shut-in.
C 2009 The Authors, GJI, 179, 634–639
Journal compilation
C 2009 RAS
Asymmetric hydraulic fractures as a result of driving stress gradients 639
Table 1. Resulting parameters of the maximum likelihood method. model to the depth intervals with the most asymmetric behaviour
Layer Stage x0 (m) t0 (min) a (min m−1 ) b g/ p net (1 m−1 ) showed that the ratio between the lateral stress gradient and the
0
overpressure ranges from 0.008 to 0.010 1 m−1 . Provided the net
A 4 −6.0 0.0 0.17 0.01 0.008 [0.007, 0.009] pressure is in the range of 1 MPa, the lateral stress gradients would
B 3 10.0 1.2 0.20 0.0 0.008 [0.007, 0.009]
span the range from 8 to 10 kPa m−1 . We assume that such large
B 2 12.0 11.8 0.22 0.0 0.010 [0.006, 0.013]
lateral gradients, which are in the order of the hydrostatic gradient,
could be explained by unequal depletion of a tight gas field in both
observational data quite well. For the deeper layer B, the larger wing horizontal and vertical directions.
is found to grow linearly with time in agreement with our previous
results (Fischer et al. 2008) showing a zero fluid leakoff (b = 0). AC K N OW L E D G M E N T S
As the net pressure is not known for this data set, we scale the es-
We are grateful for critical comments and inspiring inputs from
timated stress gradient in Table 1 to the same units by the assumed
J. Mls, J. Le Calvez, E. Brown, L. Eisner and P. Bulant. The opinion
net pressure p net
0 = 1 MPa. The resulting lateral stress gradients
of two anonymous reviewers and the detailed review report of the
g = 10, 8 and 8 kPa m−1 for the injection stages 2, 3 and 4 approach
editor, J. Renner helped us to improve the manuscript. The work of
the hydrostatic gradient of 10 kPa m−1 , which appears as a rather
the first two authors was supported by the European Union Transfer
large value and could indicate that the net pressure was probably
of Knowledge project IMAGES (MTKI-CT-2004-517242) and of
smaller.
the first one by the Czech research projects MSM0021620855 and
In general, g could be interpreted by a lateral variation of the rock
AV0Z30120515.
strength, normal stress on the fracture wall and also by inclinations
of the reservoir layers. However, inclined layers would result in only
a very small gradient in fractions of the lithostatic stress. Further- REFERENCES
more, the well logs and locations of events show almost horizontal
Baisch, S. & Harjes, H.-P., 2003. A model for fluid-induced seismicity at
layers, which rules out this interpretation. We also expect that the the KTB, Germany, Geophys. J. Int., 152, 160–170.
strength variations in horizontal layers are negligible compared to Economides, M.J. & Nolte, K.G., 2003. Reservoir Stimulation, John Wiley,
the possible pore pressure changes influencing the effective stress. Hoboken, NJ, pp. 5.1–5.14.
Accordingly, the pore pressure gradient due to inhomogeneous de- Fischer, T., Hainzl, S., Eisner, L., Shapiro, S. & Le Calvez, J., 2008. Mi-
pletion of the gas field appears the most likely interpretation. In par- croseismic signatures of hydraulic fracture growth in sediment forma-
ticular, a local depletion of a low-permeability reservoir can result tions: observations and modeling J. geophys. Res., 113, B02307, doi:
in suprahydrostatic lateral pore-pressure gradients if the reservoir 10.1029/2007JB005070.
is exerted to lithostatic conditions. Geertsma, J. & de Klerk, F., 1969. A rapid method of predicting width and
extent of hydraulic induced fractures, J. Petrol. Technol., 21, 1571–1581,
paper SPE 2458.
5 C O N C LU S I O N S House, L., 1987. Locating microearthquakes induced by hydraulic fracturing
in crystalline rocks, Geophys. Res. Lett., 14(9), 919–921.
We referred to the asymmetric shape of hydraulic fractures cre- Kochnev, V.A., Goz, I.V., Polyakov, V.S., Murtayev, I.S., Savin, V.G., Zom-
ated by an injection of fluids in the rock formations and suggested mer, B.K. & Bryksin, I.V., 2007. Imaging hydraulic fracture zones from
that, besides other possible interpretations, the asymmetry could be surface passive microseismic data, First Break, 25, 77–80.
easily explained by the horizontal gradient of the confining stress, Khristianovich, S.A. & Zheltov, Y.P., 1955. Formation of Vertical fractures
which results in a different closure pressure on opposite sides of by means of highly viscous liquid, in Proceedings of the Fourth World
injection. We propose a fracture model and derive simple analytical Pet. Congress, Rome, Vol. 2, pp. 579–586.
Nordgren, R.P., 1972. Propagation of a Vertical Hydraulic Fracture, SPE J.
equations to relate the lateral stress gradient to the parameters of the
12(8), 306–314, paper SPE 7834.
hydrofracture. This model indicates that the lateral stress gradient Perkins, T.K. & Kern, L.R., 1961. Widths of Hydraulic Fractures, J. Petrol.
can be determined from the net pressure and the length difference Technol., 13(9), 937–949, paper SPE 89.
between the fracture wings. We tested the model on the asymmet- Shapiro, S.A., Dinske, C. & Rothert, E., 2006. Hydraulic-fracturing con-
ric microseismicity accompanying the growth of hydraulic fracture trolled dynamics of microseismic clouds, Geophys. Res. Lett., 33, L14312,
in the Canyon Sands formation in West Texas. Application of our doi: 10.1029/2006GL026365.
C 2009 The Authors, GJI, 179, 634–639
Journal compilation
C 2009 RAS