Non-penetrated large-/mid-scale fractures: blind spot or 'nebula' for artificial tracers in inter-well tests

© Authors 2021.
                                                           This work is distributed
                                                                 under the Creative
                                                             Commons Attribution
                                                                         4.0 License

                                                                                         DOI.org/10.5194/egusphere-egu21-13435

Non-penetrated large-/mid-scale fractures: blind spot or ‘nebula’ for
artificial tracers in inter-well tests
Horst Behrens, Julia Ghergut, Martin Sauter
Applied Geology Dept., University of Göttingen, 37077, Germany
Correspondence to: Julia Ghergut (iulia.ghergut@geo.uni-goettingen.de)

Abstract. This brief note updates a study series (2015, 2017, 2019, 2020, 2021) concerning the ability of conservative
artificial tracers, in inter-well forced-gradient flow, to ‘detect’ and quantify the properties of ‘nearby’ fractures or faults
(non-penetrated by any of the well-screens involved in the tracer test). In an earlier short note (2015), we reproved the
malpractice of proclaiming ‘tracer test failure’ from the objective datum of a very long fluid residence time (a feature
unrelated to the tracer but intrinsic to fluid turnover in the georeservoir subject to the given forced-gradient flow). We now
revisit a particular reservoir setting in the Upper Rhine Rift Valley, to find that a long residence time for circulating fluids
may indeed be associated with ‘tracer-test failure’, but in a different way than previously recriminated. Again, this is not a
test design issue, but a more general problem posed by a georeservoir class we may deem ‘very large’ in terms of fluid
turnover volume under the typical conditions of a geothermal heat extraction project. With ever increasing georeservoir size,
the usefulness of artificial tracers in inter-well circulation may indeed become questionable, especially in the presence of
large-scale faults. Similar issues may occur, however, with smaller-scale fractures, if non-penetrated by any of the wells
involved in forced-gradient circulation; ambiguity between longitudinal and transversal contributions to the reservoir’s
hydraulic and fluid transport properties (transmissivity, and fluid residence times) persists, and remains difficult to resolve
even by a so-called ‘joint inversion’ of artificial-tracer and pressure signals, the more so as ‘ultra-rapid’ diffusivity of
poroelastically-induced stress changes at early stages of reservoir operation (at the very onset of fluid injection / depletion)
cannot be monitored in practice.

Keywords: large-scale fault, permeability anisotropy, reservoir boundary, drain, preferential flow, tracer test, forced-gradient, fluid
residence time, Upper Rhine Rift Valley, Oberrheingraben

2                                       Behrens et al.: Large- and mid-scale fractures, ‘blind spot’ or ‘nebula’ for artificial tracers

1 Long residence times – bad tracer tests?

In an earlier short note (Ghergut et al. 2015), we criticized the common practice of mistaking the objective datum of a long
residence time (of fluid in a georeservoir subject to forced-gradient inter-well flow) for the ‘failure’ of a particular tracer test;
we argued that fluid residence times are intrinsic to the flow field (whether natural or forced-gradient), not steerable by the
design and sizing of a tracer test; what can be influenced by the latter is just the ability to detect and meter (and accurately
quantify) a tracer signal at a particular (passive or active) monitoring well, sooner or later (or maybe ‘never’) after having
added a well-defined tracer quantity (in a well-defined manner) into the given flow field.
We now revisit the georeservoir setting described by Meixner (2009), to find that a ‘very long’ residence time may indeed
curb tracer-test effectiveness, but in different ways than had been revealed so far. Again, this is not a test design issue, but a
more general problem posed by a georeservoir class we may deem ‘very large’ in terms of fluid turnover volume under the
typical conditions of a geothermal heat extraction project. The operational ‘size’ of such projects is limited by the maximum
pressure buildup and drawdown that can be sustained at injection and production wells, respectively. In practice, for
reservoirs in the Upper Rhine Rift Valley on its Eastern side (Meixner 2009; Herzberger et al. 2009, 2010; Eggeling et al.
2013; Meixner et al 2016), this amounts to inter-well distances in the range of 1 – 3 kilometres. Lithostratigraphy of target
formations (in particular, ‘aquifer’ layer thickness and porosity values) is something we can roughly ‘choose’, but not
influence (not beyond the near-well scale, if at all; definitely not at reservoir scale). Swept reservoir volumes ( V) will then
depend on chosen flow regimes – with flow rates (Q) more or less dictated by economical considerations or constraints,
again beyond the reach of tracer-test design considerations. Fluid residence times will thus result as the quotient n  V / Q
(as a roughly defined average, assuming some ‘representative’ value n for the transport-effective porosity), but this
calculation gets complicated by the presence of large-scale fault zones, which may drain or divert a unknown fraction q* of
the total discharge Q, whose effect on the overall residence time may be strong even though the void-space volume within
fault zones is usually negligible, compared to the total reservoir void-space volume ( V*

Behrens et al.: Large- and mid-scale fractures, ‘blind spot’ or ‘nebula’ for artificial tracers                               3

fluid transport properties of neighboring (non-penetrated) fractures or faults, in the typical forced-gradient circulation setup
of a geothermal well doublet. A complementary use of natural tracers is not addressed here.

2 Tracer signals in the ‘presence’ of large-scale faults

Relying on the structural-hydrogeological model proposed by Meixner (2009) for a geothermal reservoir in South-West
Germany (on the East side of the Upper Rhine Rift, the particular site serving as a ‘pilot’ to demonstrate various ways of
‘optimizing’ electricity production by means of a geothermal well doublet), we set up a distributed-parameter model of the
more or less contributing ‘reservoir’ layers, with two ‘remote’ fault zones in guise of reservoir boundaries, and a large-scale
fault-zone between the injection and production well (fig. 1, with further specifications in the lower section of fig. 2). The
existence and approximate location of these fault zones had been inferred by Meixner (2009) from dedicated (‘local’)
seismic exploration data, corroborated with larger-scale tectonic models.

Figure 1. Outline of reservoir compartments (red: most-permeable, yellow: fairly-permeable, blue: least-permeable layers).

https://doi.org/10.5194/egusphere-egu21-13435

4                                       Behrens et al.: Large- and mid-scale fractures, ‘blind spot’ or ‘nebula’ for artificial tracers

We used FEFLOW v.5.4 (Diersch 2014) to numerically simulate fluid ‘age’ (residence time) redistribution under forced-
gradient circulation, tracer test signals (fig. 2), and heat transport (fig. 3) for a number of contrasting assumptions as to
    • the nature of boundary conditions and hydrogeological characteristics of the two ‘remote’ fault zones,
    • the degree of permeability contrast between adjacent compartments,
    • anisotropy of permeability for selected reservoir layers,
    • hydrogeological characteristics of the naturally-occurring fault, located between the injection and production wells.

Figure 2. Conservative-tracer signal predictions over three decades of reservoir operation, for selected fault-zone scenarios.

                                                                                     https://doi.org/10.5194/egusphere-egu21-13435

Behrens et al.: Large- and mid-scale fractures, ‘blind spot’ or ‘nebula’ for artificial tracers                                   5

It appears that a tracer slug sizing allowing for tracer signals to become detectable during the first three years after tracer
injection in all scenarios considered here (as specified in the lower section of fig. 2) is not feasible in practice. In some of the
scenarios considered, the system will act like a ‘very large’ reservoir, with fluid residence times in the order of decades, and
extreme dilution of injected tracers. Even using preparative-scale cleaning of samples, brine separation, sample enrichment
by solid phase extraction, evaporative up-concentrating etc., followed by advanced chromatography techniques to enhance
the separation between tracer and natural background signals, it will not be possible to lower tracer detection limits beyond a
certain threshold (the latter being controlled, primarily, by the amount of certain naturally-occurring aromatics , even at
trace level, in the reservoir brine). On practical reasons, the tracer slug sizing will be limited to some hundred kilogram of
one or two organic tracers. This means that part of the scenarios illustrated in fig. 2 will remain indistinguishable during the
first ~three years after tracer injection.

3 Long residence times: loss of tracer ‘sensitivity’?

For such reservoir structures, there is no clear-cut correspondence between early-vs.-late ‘arrival’ of detectable tracer at the
production well, and small-vs.-large reservoir (unlike the generic correlations outlined in Ghergut et al. 2013). How much
information will actually be lost, and what degree of uncertainty will affect temperature predictions, as a consequence of the
chosen practical ceiling on tracer slug sizes for injection? This amounts to giving up all signal ‘information’ contained
below a certain threshold on the tracer flux-concentration axis of fig. 2, whose ordinate value is determined by the slug size
(Minj) and the quantification (QL) limit for the respective tracer species (assuming state-of-the-art laboratory-instrumental
facilities were available).
On the other hand, the thermal lifetime of this reservoir is going to exceed three decades in any of the scenarios considered
(with a fluid turnover rate of 30 L/s or roughly one million m³ per year), and it is controlled primarily by the degree of
permeability anisotropy within the most conductive layer (cooling plume growth exemplified comparatively in fig. 3).
As long as the middle fault (M) acts like a vertical drain, irrespective of lateral fault (L) behavior, tracer signals recorded
(provided they were detectable and meterable) within one decade from tracer injection would allow to roughly tell the
existence and large-scale properties of these fault zones, the signals being separated by at least one year in terms of the
(improperly so called) ‘arrival time’, and by ~two orders of magnitude per year, in terms of increase rate (for the scenarios
illustrated in fig. 2, the l.-h.s. range of BTCs). In other words, tracer signals would be worth being metered (‘whatever it
takes’ in terms of laboratory-instrumental power) at their ascending stage during this ‘early’ first decade.
However, after peak passage, they soon become insensitive to large-scale fault characteristics, with signal differences
spanning less then 1/3 magnitude order, for the entire second decade.                During the third decade, sensitivity to fault

https://doi.org/10.5194/egusphere-egu21-13435

6                                      Behrens et al.: Large- and mid-scale fractures, ‘blind spot’ or ‘nebula’ for artificial tracers

characteristics would be regained to a certain extent, spanning approximately one magnitude order – but, after having
‘waited’ for two decades, the cumulated effects of further uncertainty sources (like flow-rate variability, etc.) are likely to
override this scarcely ‘one’ magnitude order of fault-zone sensitivity.

Figure 3. Cold-water plume growth around the injection well-screen after three decades of fluid turnover at ~1 million m³/y,
with a moderate vs. strong permeability anisotropy (scenarios plotted in black and orange in figure 2).

If the middle fault (M) does not act like a vertical drain, fluid residence times become so large that tracer signals would stay
undetectable for at least two decades. – The accelerating effect of a vertically-draining middle fault seems counter-intuitive
at first sight, but can be explained by the flow cross section ‘shrinking’ effect of vertical drainage (cf. small inserts at the
upper r.-h.s. corner of fig. 2, showing tracer ‘plume’ shapes after two and three decades), independently of the transport-
effective porosity of adjacent rock.

4 Discussion

Such large residence times, and the associated limitations to fault-zone sensitivity of artificial tracers are not a test design
issue, but a more general problem posed by a georeservoir class we may deem ‘very large’ in terms of fluid turnover volume
under the typical conditions of a geothermal heat extraction project. Beyond certain georeservoir sizes, the usefulness of
artificial tracers in inter-well circulation may indeed become questionable (or redeemable only at very high costs), especially
in the presence of large-scale fractures or faults, even when not intersected by any of the well-screens involved in forced-

                                                                                    https://doi.org/10.5194/egusphere-egu21-13435

Behrens et al.: Large- and mid-scale fractures, ‘blind spot’ or ‘nebula’ for artificial tracers                                   7

gradient flow. The transport-effective size of a georeservoir is determined not only by lithostratigraphy, but also by barrier
and drainage characteristics of fault zones, especially when these are found ‘inside’ the reservoir.
Similar effects have been found to occur, however, with smaller-scale fractures located in some distance from the wells, and
not penetrated by any of the well screens – as seen, for instance, with a geothermal well doublet in a porous-fissured, more
or less karstified, maybe faulted and fractured Jurassic ‘Malm’ formation in 2 – 2.5 miles depth in Southern Germany, for
which a conceptual-hydrogeologic model was outlined by Behrens et al. (2020, 2021), along with some competing
hypotheses as to what role fractures might play for heat and solute transport at reservoir scale (fig. 4).

Figure 4. Effect of mid-scale, non-penetrated fractures on heat and tracer transport across the heterogeneous reservoir.

Ambiguity between longitudinal and transversal contributions (relative to forced-gradient flow direction) to the reservoir’s
hydraulic and fluid transport properties (hydraulic transmissivity, and fluid residence times) persists, and remains difficult to
resolve even by a so-called ‘joint inversion’ of artificial-tracer and pressure signals, the more so as ‘ultra-rapid’ diffusivity of
poroelastically-induced stress changes (cf. figures 8 – 10 of Behrens et al. 2020) at early stages of reservoir operation, i. e., at
the very onset of fluid injection and/or depletion cannot be monitored in practice.

We gratefully acknowledge long-term support from Germany’s Federal Ministry for Economic Affairs and Energy
(BMWi) within the applied-research projects “LOGRO”, “TRENDS”, and “UnLimiteD”.

https://doi.org/10.5194/egusphere-egu21-13435

8                                          Behrens et al.: Large- and mid-scale fractures, ‘blind spot’ or ‘nebula’ for artificial tracers

References

Behrens, H., Ghergut, J., Sauter, M., Wagner, B., Wiegand, B.: Premature decline of production temperature – can tracer test tell why?
    Procs. 45th Workshop on Geothermal Reservoir Engineering, Stanford University, CA, SGP-TR-216, 2020.
Behrens, H., Ghergut, J., Sauter, M.: First-order discontinuity in cumulative tracer recovery: need for endo-tracer push-pull. Procs. 46th
    Workshop on Geothermal Reservoir Engineering, Stanford University, CA, SGP-TR-218, 2021.
Birkigt, J., Stumpp, C., Maloszewski, P., Richnow, H. H., Nijenhuis, I.: Evaluation of hydraulic characteristics in a pilot-scale constructed
     wetland using a multi-tracer experiment. Geophys. Res. Abs., 15, 10065, 2013.
Diersch, H.-J. G.: FEFLOW – Finite-element modeling of flow, mass and heat transport in fractured and porous media . Springer, Berlin
     Heidelberg, 2014.
Eggeling, L., Genter, A., Kölbel, T., Münch, W.: Impact of natural radionuclides on geothermal exploitation in the Upper Rhine Graben.
    Geothermics, 47, 80–88, 2013.
Fox, G. A., Heeren, D. M., Miller, R. B., Mittelstet, A. R., Storm, D. E.: Flow and transport experiments for a streambank seep originating
     from a preferential flow pathway. J. Hydrol., 403, 360–366, 2011.
Ghergut, J., Bedoya-González, D. A., Bensabat, J., McDermott, C., Wagner, B., Wiegand, B., Sauter, M.: Endo-tracer aided visibility of
    ‘nearby’ fault reactivation: scoping simulations, II. Procs. 44th Workshop on Geothermal Reservoir Engineering, Stanford University,
    CA, SGP-TR-214, 217–227, 2019.
Ghergut, J., Behrens, H., Licha, T., Sauter, M.: Tracer-based prediction of thermal reservoir lifetime: scope, limitations, and what reactive
    tracers can tell. Procs. 38th Workshop on Geothermal Reservoir Engineering, Stanford University, CA, SGP-TR-198, 309–315, 2013.
Ghergut, J., Behrens, H., Sauter, M.: Long residence times – bad tracer tests? Geophys. Res. Abs., 17, 13288, 2015.
Ghergut, J., Behrens, H., Sauter, M., Schmidt, S., Wagner, B.: On the visibility of non-intersected ‘nearby’ fractures in inter-well tracer
    tests: scoping simulations, I. Procs. 42nd Workshop on Geothermal Reservoir Engineering, Stanford University, CA, SGP-TR-212,
    320–325, 2017.
Herzberger, P., Kölbel, T., Münch, W.: Geothermal Resources in the German Basins. GRC Transact., 33, 395–397, 2009.
Herzberger alias Orywall, P., Münch, W., Kölbel, T., Bruchmann, U., Schlagermann, P., Hötzl, H., Wolf, L., Rettenmaier, D., Steger, H.,
    Zorn, R., Seibt, P., Möllmann, G.-U., Sauter, M., Ghergut, J., Ptak, T.: The Geothermal Power Plant Bruchsal. Procs. World
    Geothermal Congress, paper 0619, 2010.
Kordilla, J., Sauter, M., Reimann, T., Geyer, T.: Simulation of saturated and unsaturated flow in karst systems at catchment scale using a
    double-continuum approach. Hydrol. Earth Syst. Sci., 16, 3909–3923, 2012.
Maloszewski, P., Stichler, W., Zuber, A.: Interpretation of environmental tracers in groundwater systems with stagnant water zones.
    Isotopes in Environmental and Health Studies, 40, 21–33, 2004.
Maloszewski, P., Stichler, W., Zuber, A., Rank, D.: Identifying the flow systems in a karstic-fissured-porous aquifer, the Schneealpe,
    Austria, by modelling of environmental 18O and 3H isotopes. J. Hydrol., 256, 48–59, 2002.
Meixner, J.: Konzeptionelle hydrogeologische Modellansätze als Vorstudie für ein integriertes Standortmodell. Master’s Thesis,
    Karlsruher Inst. of Technology (Germany), KIT Diploma Theses / Archives, Ch. 5, 2009.
Meixner, J., Schill, E., Grimmer, J. C., Gaucher, E., Kohl, T., Klingler, P.: Structural control of geothermal reservoirs in extensional
    tectonic settings: An example from the Upper Rhine Graben. J. Struct. Geol., 82, 1–15, 2016.
Stumpp, C., Maloszewski, P.: Quantification of preferential flow and flow heterogeneities in an unsaturated soil planted with different
    crops using the environmental isotope δ18O. J. Hydrol., 394, 407–415, 2010.

                                                                                          https://doi.org/10.5194/egusphere-egu21-13435