Microseismicity Cloud Can Be Substantially Larger Than the Associated Stimulated Fracture Volume: The Case of the Paralana Enhanced Geothermal System

The problem

Temperatures high enough for economic geothermal energy are found almost everywhere, if wells are drilled deep enough (Tester et al., 2007). An enhanced geothermal system (EGS) extracts this heat by circulating water through the rock, from an injection well to a production well (Majer et al., 2007). At those depths, permeability is usually low (Majer et al., 2007). Hydraulic stimulation increases it by injecting cold water at high pressure. The stimulated volume sets the size of the heat resource. However, it cannot be observed directly.

Stimulation nearly always triggers small earthquakes (Majer et al., 2007; Zang et al., 2014). These are usually attributed to hydroshearing, in which raised fluid pressure lets existing fractures slip (Pearson, 1981) and remain more permeable (Chen et al., 2000). Several studies have linked the growing earthquake volume to the growing stimulated volume (Baisch et al., 2004; Bendall et al., 2014; Rothert and Baisch, 2011). Our goal was to test whether the earthquakes and the permeability increase occupied the same volume during the 2011 stimulation at Paralana.

What we did

We used earthquake catalogues from two stimulations about 4 km deep in South Australia. At Habanero in 2003, 20,000 m³ of water, a cube about 27 m across, was injected over 30 days and 10,436 earthquakes were located. At Paralana in 2011, 2,200 m³ of water, a cube about 13 m across, and 110 m³ of acid were injected over 3.4 days, and 3,809 earthquakes were located. Injectivity, the injection rate divided by the wellhead pressure, remained almost constant at Habanero and increased at Paralana.

Whether a single fracture slips depends on its unknown strength and stress. Counting many earthquakes averages out these variations. We therefore counted hypocentres per unit area of the plane of injection, the gently dipping fracture zone outlined by the earthquakes, in rings 50 m wide around the well. We assumed that a fracture slips once fluid pressure rises by a critical amount, then slips again with each further rise. Hypocentre density should then increase in proportion to the pressure rise beyond that critical amount.

For Paralana, we simulated flow from the well with the FEHM simulator (Zyvoloski, 2007). Permeability in the simulation grew with time and fell with distance from the well, following a six-parameter formula that assumes no particular physical cause. We inverted for these and three other parameters. A Markov chain Monte Carlo method ran 300,000 simulations to find values that reproduce both the injectivity and the hypocentre density.

What we found

At Habanero, hypocentre density was roughly uniform within 300 m of the well. It rose linearly with injection pressure. The best fit gives about 280 earthquakes per square kilometre for each megapascal of pressure above a critical pressure rise of 4.6 MPa. Beyond 300 m, density fell with distance, quickly at first and then more slowly. This pattern is expected for pressure spreading from the well. We suggest that hypocentre density is a reasonable proxy for fluid pressure. However, it will be less accurate where other processes, such as thermal contraction, trigger most earthquakes.

Injectivity alone did not constrain permeability at Paralana. Two scenarios, with permeability increases extending 400 m and 50 m from the well, both reproduced the rise in injectivity. Only the change in hypocentre density over time distinguished them.

In our inversion, only a narrow range of scenarios fitted both data sets. In all of them, permeability increased no further than 40 m from the well, and in most no further than 25 m. Earthquakes occurred up to 400 m away. For the best-fitting scenario, the volume where permeability rose by more than 1% was 0.4% of the seismically active volume, roughly one part in 250. In that scenario, permeability at the well rose by a factor of about 30. This seems reasonable in comparison with factors of 50 at Desert Peak (Dempsey et al., 2015) and 400 at Basel (Häring et al., 2008). The inversion assumes flow spreads equally in all directions, although the Paralana earthquakes did not. In two synthetic tests that break this symmetry, the inversion appeared to recover the maximum extent of the permeability increase.

If hydroshearing had increased permeability substantially, the stimulated and seismically active volumes should have agreed more closely. We therefore suggest that hydroshearing was not the main cause of the injectivity increase at Paralana. At Habanero, more than 10,000 earthquakes occurred without an increase in injectivity. Hydroshearing may not have increased permeability there either. At Paralana, we consider acid, cooling of the rock, and hydraulic fracturing, alone or together, more likely causes. Acid can dissolve minerals in fractures near the well and widen them. No acid was injected at Habanero, and injectivity there did not rise. Cooling can make rock contract and widen fractures. In our model, cold water reached no further than 60 m from the well, similar to the extent of the permeability increase. Fluid pressure may also have exceeded the minimum stress in the rock and opened a hydraulic fracture.

Limitations

Our inversion gives a radially averaged permeability and cannot represent flow channels. The six-parameter formula also excludes permeability increases that reverse, for example when cooled rock warms again.

Our results come from one stimulation. We therefore conclude only that permeability increase and induced earthquakes can be decoupled in some cases. Where they are, the volume of earthquakes is a poor indicator of the stimulated volume. Measuring stimulation success by the extent of the earthquakes alone can then be misleading.

Glossary

An enhanced geothermal system (EGS) produces heat from hot rock whose permeability may have been increased by injecting fluid.

Permeability measures how easily fluid flows through rock.

Hydraulic stimulation is the injection of cold water at high pressure to increase the permeability of rock around a well.

Injectivity is the injection rate divided by the wellhead pressure. A rising injectivity indicates that water enters the rock more easily.

Hydroshearing is slip of existing fractures caused by raised fluid pressure. It can leave the fractures more permeable.

A hypocentre is the point where an earthquake starts.

Hypocentre density is the number of hypocentres per unit area of the plane of injection, counted up to a given time.

A megapascal (MPa) is a unit of pressure, about ten times atmospheric pressure.

The critical pressure rise is the increase in fluid pressure needed before a fracture first slips.

The seismically active volume is the volume of rock in which induced earthquakes occurred.

The stimulated volume is the volume of rock whose permeability was increased by stimulation.

Inversion estimates the inputs of a model, here the permeability, from observations of its outputs.

Markov chain Monte Carlo is a method that runs a model many times to find the range of parameter values consistent with the data.

A hydraulic fracture is a new tensile crack opened when fluid pressure exceeds the smallest compressive stress in the rock.