Electrical Resistivity Monitoring Primer
Data acquisition
Collection of electrical resistivity data in the field (or in a laboratory) requires a dedicated instrument known as a resistivity meter. While multiple research groups have built their own resistivity systems, in general users will use a commercially purchased instrument (see here for a list of different vendors). These instruments can be used to collect different kinds of surveys.
Instrument overview
Within resistivity systems there are different types and generations of instruments. These are
- Four electrode systems. These were the first generation of instruments. They can be used to take a standard electrical resistivity measurement.
- Single channel, Multi electrode systems. These systems were developed in the 1990s, and add a multiplexer to four electrode systems so that a system can be hooked up to multiple electrodes at the same time and collects data on different electrode combinations without the need to manually move electrodes.
- Multi channel, multi electrode systems. These systems (which were introduced by different manufacturers around 2010) allow the collection of multiple potential measurements at the same time. They can speed up acquisition speed substantially.
Multi electrode systems require cables which connect the system to multiple electrodes. The most common design for these cables are multiconductor cables in which each cable is connected to an electrode. However, different cable designs exist.
Electrical resistivity systems are typically time domain systems which use a square wave, however there are some specialty frequency domain systems, as well as systems which can do both time domain and frequency domain measurements. In virtually all commercial systems we use a single current injection source, but (similar to medical imaging applications), multi source systems have been proposed.
Measurement geometries
We can use different electrode geometries. Different common electrode geometries are typically used with different advantages in terms of signal/noise and vertical and horizontal resolution. Each measurement contains information about the subsurface. As is intuitively clear (and described in detail in a 1993 article by Xu and Noel ) using a certain number of electrodes we can take a finite number of independent measurements. Xu and Noel introduce the concept of complete datasets – datasets that have all possible independent measurements. Different equivalent complete datasets exist for all geometries. There are multiple papers and studies which discuss the sensitivity of different array types and the respective value of different arrays. This includes e.g. a paper in Vadose Zone Journal by Furman et al (note that a pdf of this paper can be found on the internet through google scholar) and course notes from Loke on resistivity imaging.
Types of surveys
The term electrical resistivity survey is used both to refer to the process of data collection as well as the resulting measurements. Historically (until about the mid 1980s) data were collected using basic four electrode systems in modes known as profiling and sounding in which electrodes were moved manually. Since the early 1990s (enabled by improvements in instrumentation, and specifically the introduction of multi electrode systems) numerous other configurations have been developed. The main ones are
- Collection of 2D profiles using a linear array of electrodes located on the surface.
- Collection of data using 3D grids of electrodes
- Collection of data using electrodes placed along boreholes (either in a single borehole or in multiple boreholes)
- Collection of data with electrodes towed behind a boat (marine resistivity) or behind a vehicle using a landstreamer. This is a special case of the 2D profiles: basically at every position of the tow vehicle we collect a small 2D profile
- Collection of time-lapse data (also known as 4D data), which can be done using all the configurations listed above.
Sounding and profiling
Sounding (also known as Vertical Electrical Sounding or VES) is a survey type which has as its objective to obtain a 1D distribution of resistivity with depth. Sounding is often contrasted with profiling, which has as its objective to obtain a distribution of the lateral distribution of resistivities. Of course, modern hardware typically can collect data which has both sounding and profiling information, but there is often still interest in VES surveys, specifically for water investigation studies.
In a VES study a symmetric array configuration is used – either a wenner or a schlumberger array. The midpoint of the array remains constant, but the size of the array increases. As the size increases, the depth to which the measurement is sensitive increases. The exact sensitivity depends on the distribution of electrical conductivities. Each measurement results in a value which can be plotted on a graph. Historically this graph was interpreted in terms of distributions of electrical properties by manual curve fitting, but nowadays numerous numerical codes exist which can do this.
2D (Linear) profiles
One popular survey type which is made possible by multi electrode instruments is the 2D profile. For a 2D profile we collect data using an array of generally equidistantly spaced electrodes laid out along a straight line on the surface. This configuration remains one of the most popular ones used to date. The reasons for this popularity are threefold:
- This kind of configuration can be easily collected (just set up a straight line with equally spaced electrodes) using current hardware;
- The data resulting from this kind of survey can be relatively easily processed on a standard PC using both commercial and open source resistivity processing software;
- The processed data looks like a fence diagram of electrical properties – something which is intuitively clear to interpreters. Note that the operative words here are “looks like”: unless the electrical properties are laterally invariant there will be 3D effects which need to be taken into consideration in the acquisition and processing, and ignoring this (as some processing codes do) can lead to artefacts. However, in many cases the resistivity profile is “sufficiently close” to a fence diagram that the advantage conferred by the 2D profile outweighs the inaccuracies.
3D surveys
While the majority of surveys uses 2D profiles there is an increase in 3D surveys where electrodes are placed in 3D grids along the surface and possibly in boreholes. While collecting and processing such surveys is substantially more involved than 2D surveys, 3D surveys provide substantial value as they can allow for better resolution of subsurface electrical property distributions than 2D surveys.
Single channel vs multi channel systems
Virtually all modern electrical resistivity systems are multi-electrode systems. This means that they can address multiple electrodes connected to them. These systems are either single channel or multi channel. Below a detailed discussion of the multi channel systems is provided.
Multi channel system basics and acquisition time speedup
Multi channel resistivity instruments measure several potential differences during one current injection. The gain in acquisition speed is not the same as the number of channels but depends on how many of the readings in the measurement sequence share the same pair of current electrodes. That is set by the array type, and it changes when reciprocal measurements are added.
Where the time goes
For resistivity instruments, most of the acquisition time is spent injecting current: a square wave with a given on-time, repeated over several cycles (stacks) to average out noise. Switching the multiplexer and reading the voltages takes comparatively little time. A multi channel instrument does not shorten an injection; it lets you record up to C potential dipoles while that injection is on, where C is the number of receiver channels.
So the number of injections, not the number of readings, sets the survey time. If a sequence contains kp readings that use current pair p, those readings need ⌈kp/C⌉ injections, and the total time is approximately
T ≈ tinj × Σp ⌈ kp / C ⌉
where tinj is the time for one injection including stacking. On a single-channel system every reading needs its own injection. The multi channel speedup can therefore never exceed the smaller of two numbers: the number of channels, and the average number of readings per current pair. The second number is a property of the array and the measurement sequence, not of the instrument.
The figure below shows for several standard arrays which measurements share an injection.
Dipole–dipole.
For a current dipole A–B, the readings at separations n = 1, 2, 3, … all use the same two current electrodes. With dipole length a and n up to 8, each injection serves up to eight readings, so an instrument with eight or more channels collects a full “row” of the pseudosection in one injection. Beyond that, extra channels do not help unless the sequence has more readings per current pair.
Wenner (α).
The current electrodes sit at the outside of the array, 3a apart, and the only Wenner reading for that pair is the one with M and N at the third points. A different spacing a means a different current pair. A conventional Wenner sequence therefore has exactly one reading per injection, and a multi channel instrument gives no speedup. The same holds for Wenner-β and Wenner-γ.
Schlumberger (Wenner–Schlumberger).
In the usual multi-electrode form (A, M, N, B with MN = a and AM = NB = na), each reading also has its own current pair. A few readings coincide when sequences with different MN lengths produce the same A–B spacing, which gives a small gain (about 7% in the example below). The potential electrodes that are not centred between A and B can also be read, which is exactly the idea behind the multiple gradient array.
Multiple gradient.
Dahlin and Zhou (2006) designed this array for multi channel instruments: one widely spaced current pair, and every adjacent potential dipole between A and B is recorded in the same injection. Its sensitivity is close to that of Schlumberger, with dipole–dipole-like efficiency.
Adding reciprocal measurements
A reciprocal reading swaps the roles of the electrode pairs: current goes through the former potential electrodes M–N and the voltage is measured on the former current electrodes A–B. By the reciprocity theorem the transfer resistance should be identical, so the difference between a normal and a reciprocal reading is a direct, data-driven estimate of measurement error. In time-lapse work these reciprocal errors are the standard basis for the error model used in the inversion, and for removing bad readings before differencing.
On a single-channel system a full reciprocal set doubles the survey time. On a multi channel system the cost depends on how the reciprocal readings group by current pair, and this differs sharply by array:
- Dipole–dipole: the reciprocal of a dipole–dipole reading is again a dipole–dipole reading, with the potential dipole on the other side of the current dipole. One injection can therefore record normal readings to its right and reciprocal readings to its left (Figure above, second row). With enough channels, reciprocals come close to free.
- Schlumberger: the reciprocal readings inject through the small central dipole M–N and measure on the wide outer pair A–B. All depth levels at one midpoint share that small current dipole, so the reciprocal set groups very well (Figure above, fifth row): 4.6× faster than single-channel in the example below, even though the normal set barely speeds up. Note that the reciprocal configuration measures over a long potential dipole, which picks up more ambient noise, and that a short current dipole may limit the current you can drive into high-resistance ground.
- Wenner: the reciprocal set has the same one-reading-per-pair structure. The only sharing comes from a geometric coincidence: the current pair of a Wenner reading with spacing a is also the potential pair of the reciprocal of a Wenner reading with spacing 3a. That gives about 15% saving, regardless of the number of channels.
- Multiple gradient: the reciprocal readings inject through small adjacent dipoles. They group about as well as the normal set, so with enough channels each set is fast, but adding reciprocals still roughly doubles the time.
Example: a 48-electrode line
To make this discussion more concrete, we enumerate the injections needed for typical sequences on a 2D line of 48 electrodes with unit spacing a. Injections were grouped optimally by current pair (a reading and its polarity-reversed version count as the same pair). The sequences are:
- Dipole–dipole: dipole lengths a–4a, n = 1–8 (1,016 readings).
- Wenner α: a = 1–15 electrode spacings (360 readings).
- Schlumberger: MN = a–3a, n = 1–6 (576 readings).
- Multiple gradient: MN = a–3a, A–B spanning 8, 12 and 16 potential dipoles (1,968 readings).
| Array | Data set | Readings | Injections, 1 channel | Injections, 10 channels | Speedup | Time, 1 channel (min) | Time, 10 channels (min) |
|---|---|---|---|---|---|---|---|
| Dipole–dipole | Normal only | 1,016 | 1,016 | 162 | 6.3× | 33.9 | 5.4 |
| Dipole–dipole | Normal + reciprocal | 2,032 | 2,032 | 272 | 7.5× | 67.7 | 9.1 |
| Wenner | Normal only | 360 | 360 | 360 | 1.0× | 12.0 | 12.0 |
| Wenner | Normal + reciprocal | 720 | 720 | 615 | 1.2× | 24.0 | 20.5 |
| Schlumberger | Normal only | 576 | 576 | 537 | 1.1× | 19.2 | 17.9 |
| Schlumberger | Normal + reciprocal | 1,152 | 1,152 | 624 | 1.8× | 38.4 | 20.8 |
| Multiple gradient | Normal only | 1,968 | 1,968 | 232 | 8.5× | 65.6 | 7.7 |
| Multiple gradient | Normal + reciprocal | 3,936 | 3,936 | 472 | 8.3× | 131.2 | 15.7 |
These times assume 2 s per injection (for example 0.5 s on-time, one full ± cycle). Actual acquisition time scales with on-time and number of stacks; the ratios do not. Contact-resistance checks and multiplexer switching add a fixed overhead that is not included.
Limitations in actual speedup and practical considerations
The description above gives the theoretical speedup. However, there are a number of additional aspects to consider.
Sequence order matters.
An electrode that has just carried current stays polarized for a while, and using it shortly afterwards as a potential electrode biases the reading (Dahlin, 2000). Multi channel sequences reuse electrodes intensively, so the order of injections should be optimized to avoid this and we may need to include wait times.
Multi channel instruments may have limits in which potential electrodes can be used
Multi channel instruments differ in how potential dipoles can be configured. Some require the potential dipoles associated with one current injection to be adjacent or to form a chain of consecutive electrodes or to have one common electrode; other systems allow any combination.
Speed is not the only criterion.
Wenner and Schlumberger have higher signal-to-noise ratios than dipole–dipole at large separations, which can matter in noisy or highly conductive settings. A slower array with clean data can be a better choice than a fast noisy array.
Multi channel systems may require compromises in settings
In a multi channel measurement we have one setting for the voltage we apply (and thus the current we inject). All channels in an injection share the same current. The near channels see large voltages and the far channels small ones, so the current and input range cannot be optimal for every channel at once. The readings with the smallest signals, such as dipole–dipole at large n, remain the noisiest.
Total time in the field includes deployment time
The numbers above compare acquisition time only. Whether that difference matters depends on field logistics. These are the same for single and multi channel systems. Specifically, the comparison does not include the time to lay out and pick up cables, install electrodes, check contact resistances, or move the spread to the next line or roll it along. That work takes the same time with a single-channel or a multi channel instrument, and on a 2D survey it often takes longer than the measurements themselves. If setting up a line takes one hour, saving 30 minutes of acquisition shortens the field day by much less than the speedup factor suggests.
Value of multi channel systems in time lapse monitoring is dependent on process speed
In time-lapse monitoring we study processes. Many processes change slowly: seasonal water table movement, most remediation treatments, slow plume migration, freeze–thaw, long-term moisture trends. Measuring once a day, or even less often, samples such processes adequately, and a survey that takes an hour instead of ten minutes makes no practical difference. A single-channel system on a permanent installation is often entirely sufficient. On the other hand, fast processes need fast surveys: if the subsurface electrical property distribution changes appreciably during a survey, the data is hard to invert. Thus, for cases where we need to image e.g. rapid movement of infiltration fronts after rain or remediation amendment emplacement one could benefit from the speed provided by a multi channel system.