SPECIAL g e o p h y H y d r oSECTION: s iH y d r o g e o p h y s i c s cs

MRS: A new geophysical technique for groundwater work

JEAN ROY, IGP

uring the last 10 years, applied geophysics techniques have made signicant progress in the exploration, quantication, and management of groundwater (GW) through wider applications of classic techniques and their integration. These include resistivity, including resistivity imaging and vertical electric sounding (VES); induced polarization (IP); spontaneous polarization (SP); time- and frequency-domain electromagnetics (TDEM, FDEM); ground-penetrating radar (GPR); very low-frequency EM (VLF); seismic; magnetic; gravity; and gamma-ray spectrometry. During that ten-year interval, however, one technique that stands out as truly new and highly relevant for GW is magnetic resonance sounding (MRS), a eld application of nuclear magnetic resonance (NMR). Functionally, MRS ts between two established techniques: atomic absorption spectrometry (AAS) and time-domain electromagnetics (TDEM). AAS is used in laboratories on carefully prepared samples and has no in-situ depth of penetration, but it performs well in element discrimination and determination of their concentration. TDEM has good depth of penetration (i.e., in suitable cases, it can measure in-situ ground conductivity as a function of depth down to several hundred meters) but it can not discriminate between elements. MRS shares some of these characteristics: it has excellent element selectivity but for one element onlyhydrogen, a major component of the water molecule. Also, MRS has moderate depth of penetration, in particular over resistive terrain (i.e., up to 150 m), while quantifying water content and pore size as a function of depth.
Figure 1. MRS setup. (bottom) Typical MRS eld layout using a square loop. (top) NUMISPLUS system of IRIS Instruments: (1) DC/ DC converter, (2) main unit, (3) wire loop, (4) tuning box, (5) rechargeable battery, and (6) control and data acquisition PC.

NMR in a nutshell NMR is one of the numerous processes of interaction between EM elds and matter. Most familiar ones occur at the level of electrons; NMR works at the nuclei level by exploiting two nuclear properties: (1) a net angular momentum and (2) a net magnetic moment . Only ~42 isotopes (~30 elements involved) are abundant enough and have both of these properties in exploitable magnitude. The gyromagnetic ratio = / is an atomic constant that uniquely characterizes each of these isotopes. Here, we are only concerned with hydrogen nuclei (1H+) with = 2.675 x 108 rads-1T-1. At equilibrium, the net magnetic moment of the volume investigated for a given isotope is aligned with the ambient (static) magnetic eld Bs. We can put it out of this alignment by (1) momentarily changing Bs or (2) exciting the volume at the resonance, Larmor frequency L = Bs/2. After excitation, because of their angular momentum, the excited nuclei will not immediately return to equilibrium but will precess around their ambient magnetic eld at frequency L during a relaxation time characterized by decay time constant Td . There is a far-ranging analogy between the precession of nuclei in a static magnetic eld and the precession of a top in
1226 The Leading Edge October 2009

a gravity eld. Through discrete dierences in energy levels, a quantum perspective is also useful for several aspects. Two types of rotations are relevant: precession of the nuclei around Bs and nutation around the excitation eld Bl . The various NMR decay time constants (T1, T2, and T2*) are signicant to petrophysical studies. In ground geophysics, we exploit the NMR concept with magnetometers and for MRS. In borehole geophysics, NMR logging tools provide highly relevant diagnostic information for petroleum exploration; due to cost factors, NMR logging is not yet generalized for GW projects. MRS implementation For MRS work, we designate the Earths magnetic eld, Be, as static (i.e., Bs = Be). The practical implementation uses a large loop laid on the ground generally in a manner quite similar to a single loop time-domain EM setup (Figure 1,

Hydrogeophysics

Figure 2. Two popular loop congurations for MRS: A = square and B = square-8. The latter conguration allows attenuation of noise interference at the cost of shallower depth of investigation.

bottom part), but additional loop shapes are also used (Figure 2). The MRS instrument (Figure 1, upper part) energizes this loop during the excitation step and uses the same loop during the detection step. A laptop PC provides control, monitoring, data recording, processing, and inversion. In this implementation, each module is < 20 kg so the entire system can be backpack-transported. Data acquisition. MRS data acquisition starts with a magnetic survey to check eld homogeneity and determine the local value of L. In conductive areas, we add a VES or an

EM sounding to get the subsurface geoelectrical characteristics of the site. The MRS system is tuned to the local Larmor frequency, L , and a sounding is implemented by varying the strength or pulse moment of the excitation. The pulse moment (Q in Ams) is the product of loop current times pulse duration. Due to signal-to-noise ratio (SNR) consideration, each measurement is repeated a number of times for signal stacking purposes and associated improvement in SNR. A sounding acquired in a park in the Netherlands at the margin of a dunes area nicely illustrates the approach (Figure 3). In this data summary (left panel), three quantities are displayed for each Q value used: the initial value (E0 - *) of the NMR signal, the average noise level ()both in nV using the left y-axisand the signal decay time constant (T2* - ) in msusing the right y-axis. The sounding parameter, Q for MRS, is the variable that allows depth discrimination. For example, in a Schlumberger VES, the sounding parameter is the operator-controlled interelectrode distance. Data inversion. Prior to data inversion, an initial model of the subsurface MRS response is constructed using the value of Be and its dip, the geoelectrical section, and some acquisition parameters such as loop size and shape. Using that initial model, the inversion attempts a least-squares t of the data to the model, using free-water content ( MRS) and signal decay rates (e.g., T2*) as inverted parameters over discrete depth intervals (Figure 3, right panels). Below the water table, MRS is an estimate ( MRS) of the eective porosity, while the signal decay rate is related to the water-bearing pore size. A more

Figure 3. MRS data set (at Waalwijk-1 in the Netherlands) and inversion. (left) MRS data summary of eld and model response of inversion. (middle) MRS inversion results: water content and decay time constant versus depth, (right) Lithological log inferred from three nearest boreholes; in this model, the WVP1A unit has a higher permeability than the SDL1B and Kd31k units. The symbols for the formations correspond to the standard used in the Netherlands. (Log from TNO, 1998 data)
October 2009 The Leading Edge 1227

Hydrogeophysics

Figure 4. MRS investigation on a paleo-channel near Maun, Botswana. (a) Summary of data set. (b) MRS inversion results with multidecay time analysis. (c) TDEM resistivity. (d) Lithology of site bh 8351 (after Mangisi, 2004; Roy and Lubczynski, 2005).

complex, two-pulses excitation scheme is also used (Legtchenko et al., 2003), from which an estimate of T1 (e.g., T1*) is made. The left panel of Figure 3 shows one peak without major inexions so that only one main aquifer is resolved under these conditions which t with the known hydrogeological information over the depth range considered. Often, because of mixed grain size or presence of ne sediment, the transition near the water table is gradual rather than abrupt. At the Waalwijk-1 site (Figure 3), the estimated depth to the water table is ~8 m. The inversion strategy and parameters also contribute to how smooth the transition is between vadoze zone and saturated formations. MRS and GW investigation Raw or inverted MRS data typically provide water content and the NMR decay time constant as a function of depth. Information acquired through MRS surveys allows, under suitable conditions, not only detection and positive identication of water-bearing layers but also the determination of their vertical geometry (depth and thickness), their freewater content (the amount of water free to move under realistic hydraulic gradients) and an estimate of key parameters such as hydraulic conductivity (K) and transmissivity (T). For a given lithology/mineralogy, the longer the NMR decay rate, the coarser the water-bearing pore size below the water table. This important observation was rst explained in 1962 by Korringa et al. in their KST model and later conrmed
1228 The Leading Edge October 2009

through empirical observations (Kenyon et al., 1989). In fact, through spectra analysis of the decay rate, the relationship between NMR decay rate and pore size allows the determination of pore-size distribution. In-situ pore-size estimation by NMR is made possible by the fact that the smaller the pore, the faster the relaxation of the precessing 1H+ nuclei through repeated contacts with the solid grain surface. Because of the close link between pore size, throat size, hydraulic permeability, and hydraulic conductivity, NMR logs can reliably supply ow properties. MRS, which is less advanced than its borehole-logging counterpart, is less reliable in environments where magnetic minerals are present. Also, in most cases, MRS supplies an average decay rate instead of a decay rate spectrum. Above the water table, in particular at depths below the reach of GPR, MRS can supply information that is dicult to acquire noninvasively, such as water content and water lm thickness or water drop size (Roy and Lubczynski, 2005). However, exploitation of MRS in the vadoze zone still needs calibration. GW resources assessment needs data about four dominant characteristics: recharge, aquifer storage, ow property estimation, and GW quality. These characteristics are usually quantied using a combination of techniques including pump and recovery tests, other hydrogeological methods, and numerical model methods including 1D and distributed models. MRS is likely to play an increasingly signicant role in such resource mapping and quantication strategies.

Hydrogeophysics

Figure 5. MRS forward models (Bs = 48000 nT, dip = 65, 150-m square loop). (left) Aquifer depth versus Q relationship with 20% porosity, 10m thick saturated aquifer at mean depth 10-165 m. (right) MRS equivalence under indicated conditions when thickness < 10 m. With porosity*thickness of 2 m of water at a mean depth of 45 m, the water content and thickness can no longer be discriminated and only their product determined.

MRS capabilities and limitations. A decade of tests and evaluations leads to the conclusion that MRS is a highly appropriate method for many GW applications due to (1) its inherent selectivity to 1H+ and therefore near-surface GW; (2) its performance as a noninvasive sounding tool (i.e., information as a function of depth); and (3) the relevance of its inverted parameters to aquifer and aquitard characterizations, MRS and Td. MRS is most commonly used in a sounding mode; i.e., 1D, water quantity ( MRS) and signal decay time (Td) are estimated as functions of depth for both the vadoze and saturated zones. The hydrogeological signicance of MRS needs careful consideration. K and T calibrations exploiting relaxation time Td have progressed signicantly and lithology-dependent factors have already been evaluated (Vouillamoz, 2003). Signal decay spectral analysis, currently limited to MRS data sets with high SNR, allows the water content to be resolved into components of pore size such as ne, medium, and coarse (Figure 4). The excitation moment Q displayed along the y-axis to stress the relationship (sounding parameter) between Q and depth is an alternate way of displaying a summary of the MRS data set. The MRS technique is sensitive to ambient noise: MRS data cannot be acquired near power lines, industrial installations nor during magnetic storms. Current implementations of the technique are not yet compatible with all geological settings: magnetic materials and some stratigraphic combination of aquifers and conductive layers may generate masking eects (Roy and Lubczynski, 2003). Some geological structures (e.g., a conductive shear zone in an otherwise re1230 The Leading Edge October 2009

sistive environment) can focus natural and cultural current/ noise thereby lowering the eective SNR of the data. Also, magnetic gradients are problematic for MRS surveys. Magnetic gradients can be diverse in origin and scale, ranging from dykes to magnetite grains in sands to magnetic lms on the surface of mineral grains. Magnetic gradients, if not accounted for, can shorten the measured NMR decay rate beyond the aperture window of MRS instrumentation and thus MRS may become insensitive to water in some magnetic rocks. Field verication testing in Canada and Asia appears to have found a solution to the magnetic gradient problem using a spin-echo technique (Legchenko, personal communication, 2008 and 2009). Spin-echo techniques have the potential to widen the range of applicable environments and to improve the reliability of estimates of MRS and Td. Two aspects of aquifer storage quantication through MRS are summarized in Figure 5. Aquifer depth estimation using Q as sounding parameter is shown on the left. The model responses of an aquifer at progressively larger mean depth clearly show that the peak response is reached at larger Q as aquifer depth increases; aquifer thickness exploits a similar relationship. In MRS, an equivalence limitation exists which is analogous to the one in resistivity work where only a layers conductance can be estimated instead of both its resistivity and thickness for some specic structures. A series of MRS responses for aquifers of the same porosity*thickness product (2 m of water) is shown on the right of Figure 5. MRS soundings easily resolve an 80-m aquifer but will only determine the porosity*thickness product when the thickness is reduced

Hydrogeophysics

Figure 6. MRS water content MRS and GW storage parameters; in addition to the bound water/free water boundary, the unsaturated zone, including the dewatered cone of pump GW extraction, has an extra boundary, specic retention versus specic yield. Except for well-sorted coarsegrain aquifers, eective porosity (saturated part) is larger than specic yield (dewatered cone). Also, except for carbonates, MRS detects essentially only free water; this is a characteristic of current instrumentation. Capillary water on this gure is from the hydrogeology context: contrary to NMR petrophysics, it is not limited to micropore-size (after Lubczynski and Roy, 2005).

to less than 10 m under the modeled conditions. Contrary to most noninvasive geophysical techniques, MRS can discriminate to some extent the type of water detected mostly through decay rate analysis: i.e., bound water versus free water and eventually retained water versus water released by gravity. Currently, with some exceptions for carbonates, instrumental characteristics limit MRS response mostly to free water (i.e., clay-bound water is not detected with MRS contrary to the case with NMR logs). In this MRS context, Figure 6 summarizes water types both in saturated and unsaturated zones. One of the most dicult properties to discriminate, from a geophysical perspective, is the separation between the specic yield and the specic retention of the vadoze zone or the dewatered cone during the GW pumping process. Specically, determining the fraction of groundwater which can ow by gravity in contrast to the fraction retained on the pores walls. This ratio is highly dependent on the distribution of the rocks grain size and surface properties. MRS has the potential to contribute to the determination of these two critical characteristics. Ongoing R&D Research and development directions involve SNR improve1232 The Leading Edge October 2009

ment, 2D and 3D capability, and a widening of the NMR signal aperture-window. Important progress is being made with respect to the hydrogeological control and calibration of the technique (Figure 7). After suitable development along several R&D directions, one can expect better ground penetration, higher GW selectivity, and higher relevance of inverted parameters than GPR, possibly with less spatial resolution. However, it is most likely that the optimal use of the MRS technique will be tightly integrated with other geophysical techniques to supply the most relevant information in a rapid and cost-eective way. The technology is currently available from companies in France, Russia, Germany, and the United States. Active working groups in MRS are located in many countries, including Australia, China, France, Germany, India, the Netherlands, Russia, Spain, and the USA. Three international workshops have allowed users and designers to share their experience and knowledge on the techniqueBerlin (1999), Orleans (2003), and Madrid (2006). Each was documented in special issues of geoscientic journals. A fourth workshop is in preparation for 2009.

Hydrogeophysics

Figure 7. Calibration of MRS results in terms of hydraulic parameters. (left) T from pump tests versus MRS T-estimator: MRST1*2Z showing lithology dependence (broadly classied as granites, sands, and chalk). (right) MRS-T calibration after integration of lithology-dependent factor (Vouillamoz, 2003).

Suggested reading. Applications of the integrated NMRTDEM method in groundwater exploration in Israel by Goldman et al. (Journal of Applied Geophysics, 1994). Pore-size distribution and NMR in microporous cherty sandstones by Kenyon et al. (SPWLA 13th Annual Logging Symposium, 1989). Theory of spin pumping and relaxation in systems with a low concentration of electron spin resonance centers by Korringa et al. (The Physical Review, 1962). A review of the basic principles for proton magnetic resonance sounding measurements by Legchenko and Valla (Journal of Applied Geophysics, 2002). A complex geophysical approach to the problem of groundwater investigation by Legchenko et al. (2003 SAGEEP Proceedings). Magnetic resonance sounding applied to aquifer characterization by Legchenko et al. (Ground Water, 2004). MRS contribution to hydrogeological system parametrization by Lubczynski and Roy (Near Surface Geophysics, 2005). Hydrological Verication of Magnetic Resonance Soundings, Maun Area, Botswana by Mangisi (masters thesis, ITC, 2004). Magnetic Resonance SoundingA Reality in Applied Hydrogeophysics edited by Plata et al. (Boletin Geologico y Minero, Special Issue, 2007). The case of an MRS-elusive second aquifer by Roy and Lubczynski (Proceedings of the 2nd MRS International Workshop, 2005). MRS multi-exponential decay analysis: aquifer pore-size distribution and vadose zone characterization by Roy and Lubczynski (Near Surface Geophysics, 2005). Principles of Magnetic Resonance, 3rd edition, by Slichter (Springer-Verlag, 1996). La charactrization des acquifres pau une mthod non-invasive: les sondages par rsonance magntique protonique by Vouillamoz (Thse de lUniversitde Paris, 2003). Resonance Soundinga Reality in Applied Hydrogeophysics edited by Yaramanci et al. (in issues 34 of Journal of Applied Geophysics, 2008).

Acknowledgments: The support of ITC and the collaboration of BRGM, CSIR-Envirotek, DWA-B, DWA-N, Ecole Polytechnique, GSD, GSN, IGM, IRIS Instruments, MBG, UQAC, WCS, and WRC is gratefully acknowledged. In particular, I want to acknowledge 10+ years of collaboration with my former hydrogeologist colleague at ITC, M. W. Lubczynski. My thanks to Rick Miller and an anonymous reviewer for help in improving the clarity of the text. Corresponding author: jeanroy_igp@videotron.ca

October 2009

The Leading Edge

1233