theadvantages$of$complementingmt$profilesin$36d...

63
1 The advantages of complementing MT profiles in 3D environments with 2 geomagnetic transfer function and interstation horizontal magnetic transfer 3 function data: Results from a synthetic case study 4 5 6 Joan Campanyà 1 , Xènia Ogaya 1 , Alan G. Jones 1,2 , Volker Rath 1 , Jan Vozar 1 & Naser 7 Meqbel 3 8 9 1 Dublin Institute for Advanced Studies (DIAS), Ireland 10 2 Now at: Complete MT Solutions, Ottawa, Canada 11 3 GFZ German Research Centre For Geosciences, Potsdam, Germany 12 13 14 15 16 E-mail: [email protected] 17 18 19 20 21 22 23 24 25 26 Page 1 of 63 Geophysical Journal International 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Upload: duongtu

Post on 01-Feb-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

1  

The  advantages  of  complementing  MT  profiles  in  3-­‐D  environments  with  2  

geomagnetic  transfer  function  and  inter-­‐station  horizontal  magnetic  transfer  3  

function  data:  Results  from  a  synthetic  case  study 4  

5  

6  

Joan Campanyà1, Xènia Ogaya1, Alan G. Jones1,2, Volker Rath1, Jan Vozar1 & Naser 7  

Meqbel3 8  

9  

1Dublin Institute for Advanced Studies (DIAS), Ireland 10  

2Now at: Complete MT Solutions, Ottawa, Canada 11  

3GFZ German Research Centre For Geosciences, Potsdam, Germany 12  

13  

14  

15  

16  

E-mail: [email protected] 17  

18  

19  

20  

21  

22  

23  

24  

25  

26  

Page 1 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 2: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

SUMMARY 27  

As a consequence of measuring time variations of the electric and the magnetic field, 28  

which are related to current flow and charge distribution, magnetotelluric (MT) data in 29  

two dimensional (2-D) and three dimensional (3-D) environments are not only sensitive 30  

to the geoelectrical structures below the measuring points but also to any lateral 31  

anomalies surrounding the acquisition site. This behaviour complicates the 32  

characterisation of the electrical resistivity distribution of the subsurface, particularly in 33  

complex areas. In this manuscript we assess the main advantages of complementing the 34  

standard MT impedance tensor (Z) data with inter-station horizontal magnetic tensor (H) 35  

and geomagnetic transfer function (T) data in constraining the subsurface in a 3-D 36  

environment beneath a MT profile. Our analysis was performed using synthetic 37  

responses with added normally distributed and scattered random noise. The sensitivity of 38  

each type of data to different resistivity anomalies is evaluated, showing that the degree 39  

to which each site and each period is affected by the same anomaly depends on the type 40  

of data. A dimensionality analysis, using Z, H and T data, identifies the presence of the 41  

3-D anomalies close to the profile, suggesting a 3-D approach for recovering the 42  

electrical resistivity values of the subsurface. Finally, the capacity for recovering the 43  

geoelectrical structures of the subsurface is evaluated by performing joint inversion 44  

using different data combinations, quantifying the differences between the true synthetic 45  

model and the models from inversion process. Four main improvements are observed 46  

when performing joint inversion of Z, H and T data: (1) superior precision and accuracy 47  

at characterising the electrical resistivity values of the anomalies below and outside the 48  

profile; (2) the potential to recover high electrical resistivity anomalies that are poorly 49  

recovered using Z data alone; (3) improvement in the characterization of the bottom and 50  

lateral boundaries of the anomalies with low electrical resistivity; and (4) superior 51  

Page 2 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 3: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

imaging of the horizontal continuity of structures with low electrical resistivity. These 52  

advantages offer new opportunities for the MT method by making the results from an 53  

MT profile in a 3-D environment more convincing, supporting the possibility of high-54  

resolution studies in 3-D areas without expending a large amount of economical and 55  

computational resources, and also offering better resolution of targets with high 56  

electrical resistivity. 57  

58  

1. INTRODUCTION 59  

Over the last thirty years theoretical and technological developments have justified and 60  

established the use of the magnetotelluric (MT) method in many regions with widely 61  

differing objectives, by adjusting it to the specific requirements of each area of study 62  

(e.g., Chave & Jones, 2012). Thanks to this relatively rapid progress, MT studies have 63  

been undertaken from the deepest seas (e.g. Heinson et al., 2000; Baba et al., 2006; 64  

Seama et al., 2007) to the highest mountain ranges on Earth (e.g., Le Pape et al., 2012; 65  

Kühn et al., 2014), characterizing the geoelectrical subsurface in a large variety of 66  

regions and for both academic and commercial applications (e.g., Heinson et al., 2006; 67  

Tuncer et al., 2006; Falgàs et al., 2009; Jones et al., 2009; Campanyà et al., 2011; La 68  

Terra & Menezes, 2012; Ogaya et al., 2014; Piña-Varas et al., 2014). However, 69  

limitations of the MT method still exist and there is ample room for improvement. 70  

One of the main issues hindering high resolution of the subsurface using MT is that in 3-71  

D environments the data can be strongly affected by geoelectrical structures located 72  

some distance away from the sites where the data are acquired, affecting most data sets 73  

to at least some degree (e.g. Brasse et al., 2002), and potentially leading to inaccurate 74  

characterization of the subsurface below the study area (e.g. Jones & Garcia, 2003). In 75  

recent years, the availability of 3-D forward and inversion codes (e.g. Farquharson et al. 76  

Page 3 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 4: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

2002; Siripunvaraporn et al. 2005a; Avdeev & Avdeeva, 2009; Egbert & Kelbert 2012; 77  

Kelbert et al., 2014; Grayver, 2015; Usui, 2015) has improved the characterization of the 78  

subsurface in these situations, and 3-D inversion has become a more routine procedure 79  

for the EM community, although more experience is needed before it becomes as firmly 80  

established as is 2-D inversion (e.g., Miensopust et al., 2013; Tietze & Ritter, 2013; 81  

Kiyan et al., 2014). 82  

However, difficulties associated with logistics, acquisition costs and instrumentation 83  

availability still often require MT field specialists to acquire data predominantly along 2-84  

D profiles, even when the study is performed in a 3-D environment. Jones (1983), 85  

Wanamaker et al. (1984), Berdichevsky et al. (1998), Park & Mackie (2000), Ledo et al. 86  

(2002) and Ledo (2005) studied the limitations of 2-D interpretation of 3-D MT data and 87  

made various recommendations, showing some situations where 2-D inversion of 3-D 88  

data can properly reproduce the geoelectrical subsurface when the content in the 2-D 89  

modes is appropriately chosen for inversion. Siripunvaraporn et al. (2005b) suggest that 90  

if the data are 3-D in nature but collected along a profile rather than on a grid, the results 91  

from the 3-D inversion of the MT profile are superior to the results obtained by 92  

performing 2-D inversion of the same data, even with the far coarser mesh employed in 93  

3-D as compared to 2-D inversion. Following this idea, several publications recommend 94  

inverting all four components of the MT impedance tensor (Z), e.g. Siripunvaraporn et 95  

al. (2005b), Tietze & Ritter (2013) and Kiyan et al. (2014), instead of the approach of 96  

inverting only the off-diagonal components (Zo) (e.g. Sasaki, 2004; Sasaki & Meju, 97  

2006). However, the presence of anthropogenic noise can strongly reduce the quality of 98  

the Z data, in particular the quality of the diagonal components, making it sometimes 99  

worthless to use all components of the Z tensor for the inversion process because of 100  

scatter and/or extremely large errors in the diagonal components of Z. 101  

Page 4 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 5: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Inter-station horizontal magnetic transfer function (H) and the geomagnetic transfer 102  

function (T) data, which are not affected by electric field effects caused by galvanic 103  

distortion from charges on conductivity boundaries or gradients on small-scale 104  

structures, have been used to improve MT results and constrain the subsurface using 2-D 105  

or 3-D inversion (e.g. Soyer & Brasse, 2001, 2-D H only of field data; Gabàs & 106  

Marcuello, 2003, 2-D Z + T theoretical study; Siripunvaraporn & Egbert, 2009, 3-D Z, T 107  

and Z+T of synthetic and field data; Berdichevsky et al., 2010a, 2-D Z + T of field data; 108  

Habibian et al., 2010, 2-D Z + T of field data; Tietze & Ritter, 2013), 3-D Z, T and Z+T 109  

of synthetic and field data, Habibian & Oskooi, 2014, 2-D H cf. T theoretical study, 110  

Meqbel et al., 2014, Z, T and Z+T of field data, Rao et al., 2014, Z and Z+T of field 111  

data; Yang et al., 2015, Z and Z+T of field data; Varentsov, 2015a,b, performing 2D, 112  

2D+ and quasi 3-D inversion using Z, T and H with theoretical and field data), but no-113  

one has yet undertaken an examination of the advantages of including H and T , 114  

simultaneously, when performing 3-D inversion of Z data (Z+H+T in 3-D). It should be 115  

noted that the magnetic effects caused by galvanic distortion charges do affect Z, H and 116  

T (e.g., Chave & Smith, 1994; Chave & Jones, 1997), however, these effects rapidly 117  

decrease with increasing period due to the square root dependence of their sensitivity on 118  

frequency, and can effectively be ignored over broad frequency bands. This rapid 119  

decrease is true for pure galvanic distortion but not in other situations as for example in 120  

the case of presence of current channelling caused by thin-long structures (e.g. Lezaeta 121  

and Haak, 2003; Campanyà et al., 2012). 122  

Another area for improvement for the MT geophysical technique is its issue with 123  

recovering high electrical resistivity anomalies in areas with contrasting low electrical 124  

resistivity values, or areas beneath strong low resistivity anomalies (e.g. Vozoff, 1991; 125  

Berdichevsky et al., 1998; Bedrosian, 2007; Berdichevsky & Dmitriev, 2008; Chave & 126  

Page 5 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 6: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Jones, 2012; Miensopust et al., 2013). The capacity at recovering low electrical 127  

resistivity structures is a particular strength of the MT method as a geophysical 128  

technique when characterizing geoelectrical structures from the first few metres to the 129  

asthenosphere (e.g. Jones, 1999; Ledo et al., 2011; Rosell et al., 2011; Campanyà et al., 130  

2012; Ogaya et al., 2013; Dong et al., 2014; Muñoz, 2014). However, issues in 131  

recovering high electrical resistivity anomalies are somewhat of a limitation for the MT 132  

method in certain resource exploration applications, for example where resistive 133  

hydrocarbons are present (e.g. Simpson & Bahr, 2005; Unsworth et al., 2005) and very 134  

high quality data are required. 135  

136  

The aim of this paper is to evaluate and highlight, from numerical experiments, the 137  

advantages in combining Z (full tensor) or Zo (off-diagonal components only) responses 138  

with H and T responses when characterizing and modelling the subsurface below a MT 139  

profile in a 3-D environment, and to evaluate if the sensitivity and resolution issues 140  

discussed above can be addressed. 141  

142  

2. EXPERIMENTAL DESIGN 143  

The numerical experiment consisted of calculating the forward responses of a defined 144  

synthetic model with embedded 3-D anomalies. These responses were used for a 145  

sensitivity test, evaluating the effects of each anomaly on the different types of data, Z or 146  

Zo, H and T. Dimensionality analyses and inversions were also performed characterising 147  

the 3-D behaviour of the responses and recovering the electrical resistivity structures of 148  

the subsurface by performing 3-D joint inversion with different combinations of Z, H 149  

and T. The fit of the data and the accuracy for recovering the electrical resistivity 150  

Page 6 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 7: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

distribution of the subsurface was quantified with the aim of providing a robust support 151  

to the conclusions extracted from the experiment. 152  

153  

2.1 Synthetic Model 154  

The background structure of the synthetic model (Fig. 1) is a 1-D layered Earth 155  

comprised of four layers: (1) a resistivity of 100 Ωm between 0 m (the surface) and 90 m 156  

depth; (2) a resistivity of 400 Ωm between 90 m and 1500 m depth; (3) a low-resistive 157  

layer of 10 Ωm between 1500 m and 2000 m depth; and (4) a layer from 2000 m to 90 158  

km depth with a resistivity of 200 Ωm. A half space of 20 Ωm at 90 km depth was added 159  

at the bottom of the model to ensure that the EM fields at the adopted periods do not 160  

penetrate away from the utilized mesh, also creating a more common response for the 161  

longest periods simulating the existence of a hypothetical asthenosphere. This 1-D 162  

layered background was used to simulate a more realistic and complex situation than a 163  

homogeneous background, and at the same time to highlight the inherent deficiencies of 164  

H and T data for resolving laterally uniform structures. 165  

Within this layered background structure there are three 3-D resistivity bodies; two low 166  

resistivity structures (L1 and L2) and a high resistivity body (H1) (Fig. 1). Both low and 167  

high resistivity anomalies have been added to highlight the resolution differences 168  

between these two types of anomalies when characterising the subsurface. L1 and L2 are 169  

located between 500 m and 1000 m depth and have a resistivity value of 3 Ωm. H1 is 170  

located between 1500 m and 2000 m depth, with a resistivity value of 400 Ωm. L1 and 171  

H1 cross the single north-south MT profile whereas the L2 structure lies completely 172  

outside the simulated linear acquisition profile to its west; this geometry was chosen 173  

with the aim of evaluating the capacity of the various inversions in determining 174  

structures outside the profile and to appraise if unexpected artefacts associated with L2 175  

Page 7 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 8: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

may be artificially created below the profile. Due to the importance of the role of H and 176  

T responses in characterizing the conducting 10 Ωm layer between 1500 m and 2000 m 177  

depth, this layer has been labelled as target body L3. 178  

179  

The forward response of the synthetic model at the 28 sites located along the north-south 180  

profile, covering a range of periods between 0.0001 and 1000 s (frequencies of 10000 – 181  

0.001 Hz), was calculated using the ModEM code (Egbert & Kelbert 2012, Kelbert et 182  

al., 2014), modified for inclusion on H responses (see supplementary material, S1, for 183  

more details), with a large 3-D mesh that comprised 222 x 182 x 132 cells in the north-184  

south, east-west and vertical directions, respectively. The core of the mesh comprised 185  

162 x 112 x 132 cells with a horizontal cell size of 120 m x 120 m, while the distance 186  

between sites was 600 m. The lateral size of the padding cells increased by a factor of 187  

1.3 from the edges of the core outward to the boundaries. The thickness of the top layer 188  

was 10 m, with an increasing factor of 1.048 for the subsequent layers in the z-direction, 189  

ensuring that the thickness of each cell within the depth of interest, namely the first 3 190  

km, is smaller than the horizontal cell size. 191  

192  

2.2 Response types 193  

As mentioned above, three different EM tensor relationships were studied in our 194  

numerical experiment: (1) the standard MT impedance tensor (Z) (Eq. 1); (2) the 195  

geomagnetic transfer function (T) (Eq. 2); and (3) the inter-station horizontal magnetic 196  

transfer function (H) (Eq. 3). 197  

𝒆! = 𝒁𝒉!                      (1)

ℎ!! = 𝑻𝒉!                      (2)

𝒉! = 𝑯!"𝒉!                  (3)

Page 8 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 9: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

198  

where  𝒉! and 𝒉! are two component vectors comprising the horizontal magnetic 199  

components at the local site l ℎ!! , ℎ!! and the neighbouring reference site n   ℎ!!, ℎ!! , 200  

respectively. The neighbouring reference site is the site used to relate the horizontal 201  

magnetic fields of all the studied sites (local sites) with the horizontal magnetic fields of 202  

this site (neighbouring reference site). 𝒆! is a two component vector comprising the 203  

horizontal electric components at the local site l   𝑒!! , 𝑒!! and ℎ!! is the vertical component 204  

of the magnetic field recorded at the local site l. 𝒁  and 𝑯!" are 2x2 complex matrices 205  

and 𝑻 is a 1x2 complex vector. In all cases dependence on frequency is assumed. 206  

The H transfer function relates the horizontal magnetic field between two different sites 207  

(one of the local sites and the neighbouring reference site), characterizing the differences 208  

associated with the electrical resistivity distribution below the two sites. This requires 209  

that the structures below the neighbouring reference site should also be characterised 210  

(e.g. Soyer, 2002), which can be achieved by having the neighbouring reference site 211  

within the area of study. 212  

213  

Note that, in case the neighbouring reference site used to acquire and process the data is 214  

outside the study area, for the inversion process the H data can be rearranged using as a 215  

neighbouring reference site any of the sites in the study area, even if the data at this site 216  

were not acquired during the whole survey. This can be achieved following steps 217  

analogous to the ones used in the ELICIT methodology to process MT data (Campanyà 218  

et al., 2014), in which inputs and outputs related by the transfer functions do not need to 219  

be acquired simultaneously. 220  

221  

Page 9 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 10: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

For the experiment presented in this paper, site 26 (identified with a green arrow in Fig. 222  

1) has been chosen as a neighbouring reference site for all of the H responses as there 223  

are no prescribed anomalies below it and thereby facilitating simple evaluation of the 224  

properties of the H responses. 225  

226  

2.3. Sensitivity Analysis 227  

A non-linear sensitivity test was performed to identify the sites and periods responsive to 228  

the geoelectrical anomalies and to characterize and quantify the effects of those 229  

anomalies on the Z, H and T responses, individually and collectively. The sensitivity test 230  

consisted of calculating the differences between the forward responses of the 1-D 231  

layered model with the studied anomalies included (L1, L2 and H1) and the forward 232  

responses of the background 1-D layered model without any anomalies. The differences 233  

between the model responses were divided by the adopted error, thus highlighting how 234  

sensitive the responses are to the anomalies with respect to the errors in the data. 235  

Impedance error bounds were set to 5 % of |𝑍𝑖𝑗| for diagonal and off-diagonal 236  

impedance tensor elements, in combination with an error floor of 5 % of   𝑍!" ∗ 𝑍!"!! 237  

for the diagonal elements. Error floor was added to avoid the extremely small error 238  

values associated with the periods of the Zxx and Zyy components not affected by the 3-239  

D anomalies. For the T and the H responses we used constant error bounds of 0.02. 240  

These errors were chosen consistent with the errors assumed in other publications with 241  

both real and synthetic data (e.g. Soyer, 2002; Habibian et al., 2010; Egbert & Kelbert, 242  

2012; Tietze & Ritter, 2013). Figures 2, 3, 4 and 5 show the effects of the L1, L2, H1 243  

and L1+L2+H1 anomalies, respectively, individually and collectively as contoured 244  

pseudo-sections, with distance along the profile on the abscissa and period, on a decadic 245  

logarithm scale, along the ordinate. Differences within the assumed errors, values 246  

Page 10 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 11: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

between -1 and 1, are white in colour and show the periods for the various components 247  

at the sites that are insensitive to the anomalies. Therefore, for these periods the 248  

anomalies cannot be resolved within the assumed precision. Dark blue and dark red 249  

areas show periods at the sites that are highly sensitive to the anomaly, with anomalous 250  

responses of four or more standard deviations of the assumed errors. Note the somewhat 251  

simplicity of anomalous response for single features (Figs. 2-4), but the complexity of 252  

the response when all features are included (Fig. 5). The total summed effect of each 253  

anomaly on the responses is quantified in Fig. 6. In this figure, the sensitivities of Zo 254  

(off-diagonal components of Z) are also shown, as the off-diagonal components have 255  

been often used as a data type, although most practitioners now invert the full impedance 256  

tensor Z. These pseudo-sections show the sensitivity of each period at each site, with 257  

respect to the adopted error, between the responses of the 1-D layered model with the 258  

studied anomaly and the responses of the 1-D layered model without any anomaly. The 259  

sensitivity values for each period and site were calculated following equation 5: 260  

261  

𝑆 = 1+ (𝑎𝑏𝑠 !"#$%  !"#$%&#"!!"#"!""#"

− 1)                        ∀  !"#$%  !"#$%&#"!!"#"!""#"

>  1      (5) 262  

263  

where S is the sensitivity per each period and site. Note that this equation has the 264  

condition that each component of the analysed tensor relationship, for each period and 265  

site, is included in the summand only if difference between model response and data is 266  

greater than the assumed error, which means is sensitive to the analysed anomaly. 267  

268  

Four main conclusions can be drawn from the sensitivity tests: 269  

(1) The influence of each anomaly at a given site and at a given period depends upon the 270  

types of responses used. Whereas the Z and H responses are more sensitive at the sites 271  

Page 11 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 12: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

directly above the anomalies, the T responses are more sensitive at the sites located at 272  

the edge of the anomalies. This can be understood as a consequence of the physics of 273  

induction. For a uniform source field there is no vertical field above a laterally uniform 274  

Earth, the greatest effects in T are off to the side of the anomaly (see, e.g., Jones, 1986). 275  

(2) The anomalies affect a narrower range of periods for H and T responses than for Z 276  

responses. For the Z responses, the anomalous effect extends to longer periods without 277  

seeing the end of the effect of the anomalies for the analysed periods. This is because the 278  

galvanic charges on the boundaries of conductivity contrasts, once imposed, remain in 279  

place at all longer periods. There are both magnetic and electric effects of the charges 280  

(see, e.g. Chave & Smith, 1994; Chave and Jones, 1997; Jones 2012) but the magnetic 281  

effects drop off very quickly with increasing period for isolated anomalies. 282  

(3) For the H and T responses the influence of anomaly L1 is stronger than for 283  

anomalies L2 and H1, though the difference is not as large as for Z and Zo responses 284  

(Fig. 6). This is because MT is primarily sensitive to elongated conductors, and in our 285  

numerical experiment L1 is the longest and most conducting of the three anomalies. This 286  

could result in the inversion modelling of the Zo and Z responses naturally focussing 287  

more on fitting data responding to the L1 structure than to the other anomalies, as this 288  

will lead to the greatest overall misfit reduction, whereas for the H and T responses the 289  

L1, L2 and H1 anomalies will have similar weight during the inversion process. 290  

(4) With the 3-D geoelectrical structures (L1, L2 and H1) orientated perpendicular to the 291  

MT profile, the presence of sensitive periods for the Zxx, Zyy, Hxy, Hyx, Hyy and Ty 292  

components implies that the analysed responses are sensitive to the 3-D nature of the 293  

anomalies (Berdichevsky & Dmitriev, 2008). However, in this numerical experiment, 294  

the effect of the L1 anomaly on the y component of the T responses is smaller than the 295  

Page 12 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 13: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

assumed error (Fig. 2c), which means that in this situation the T responses “see” the L1 296  

anomaly as a 2-D structure within the precision of the responses. 297  

298  

The different influences of each anomaly in relation to the examined types of responses 299  

suggest that the use of H and T responses as an accompaniment to the commonly used 300  

Zo and Z responses will introduce complementary information that enhances resolution 301  

and restricts acceptable model space, whilst simultaneously increasing the weight to the 302  

L2 and H1 anomalies during the inversion process. 303  

304  

2.4. Dimensionality analysis 305  

The dimensionality of the synthetic model was analysed using the Z, H and T responses 306  

separately (Fig. 7). Normally distributed scattered random noise was added to the 307  

responses in a manner consistent with the errors assumed for the sensitivity test. In 308  

particular, the added noise at each site and frequency was derived by generating random 309  

numbers from the normal distribution with mean parameter (0) and standard deviation 310  

(1) and multiplying this number by the data error of the responses. The same errors as 311  

those assumed for the sensitivity test were assumed for the distorted responses. 312  

For the Z responses, dimensionality analysis was performed using the phase tensor 313  

(Caldwell et al., 2004), as this method is routinely used for evaluating the dimensionality 314  

of the study area without being affected by galvanic distortion. To assess the presence of 315  

3-D structures, a variant of the β criterion was used. Typically β is used to justify 2-D/3-316  

D interpretation based on a threshold value (see e.g. Caldwell et al., 2004; Thiel et al. 317  

2009; Booker, 2014). In this case we took into account the errors of the data, as 318  

suggested by Booker (2014), interpreting all deviations from zero that are above the 319  

error in determining β as a 3-D. The ellipses are each filled with a colour associated with 320  

Page 13 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 14: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

the magnitude of β divided by this error. Values greater than one require the presence of 321  

3-D structures (Fig. 7a). The error in β was calculated by computing a thousand β values 322  

using different values of the Z components (Zxx, Zxy, Zyx and Zyy, real and imaginary) 323  

randomly distributed within the total range of the assumed error. 324  

For the H responses dimensionality analysis was performed following similar criteria to 325  

that used by Berdichevsky et al. (2010b), plotting the variations of the H responses: H-I, 326  

where I is the identity matrix. In this case 1-D structures are represented by dots 327  

(Phi_max = Phi_min = 0), 2-D structures by lines (Phi_max > Phi_min = 0) and 3-D 328  

structures by ellipses (Phi_max ≥ Phi_min ≠ 0). The colour represents the magnitude of 329  

Phi_min divided by the error of Phi_min. Only values greater than one require the 330  

presence of 3-D structures to explain the observed responses (Fig. 7b). Error in Phi_min 331  

values were derived following equivalent steps to that the calculation of the error in β. 332  

333  

Finally, induction arrows (Schmucker, 1970), following the Parkinson criteria (i.e., the 334  

real arrows generally point towards current concentrations in conducting anomalies, 335  

Jones, 1986), were used to study the dimensionality of the area using the T responses 336  

(Fig. 7c). For the T responses, the presence of real and imaginary arrows pointing in 337  

non-parallel directions at the same period and site is associated with a 3-D structure, as 338  

the imaginary induction arrow is approximately parallel to the difference between two 339  

real induction arrows at consecutive periods at the same site (Marcuello et al., 2005). 340  

The coloured circles show the differences between the angle of the real and the 341  

imaginary arrows, for each period and site, divided by the error at constraining the angle 342  

difference. Note that to define the angle difference between real and imaginary arrows 343  

we computed the differences using the same and opposite directions of the real and 344  

imaginary arrows, and selected the minimum angle difference value. Values greater than 345  

Page 14 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 15: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

one at dividing the angle difference by the error at determining the angle difference 346  

suggest the presence of 3-D structures (Fig. 7c). Errors at defining the angle difference 347  

between real and imaginary arrows were derived following equivalent steps to that the 348  

calculation of the error in β. 349  

350  

Comparing the results in Fig. 7 between all sites and periods, differences between the 351  

angles of the phase tensor ellipses (removing the 90º ambiguity), the directions of the 352  

arrows (T responses) and the angles of Phi_max (H responses), all of them are also 353  

related to a 3-D environment. Note that for structures crossing the MT profile (L1 and 354  

H1), the phase tensor and the induction arrows appear to characterize the 3-D behaviour 355  

at the sites located at the edges of the anomalies, whereas looking at the variations of H, 356  

sites located above of the anomalies seems to be more affected by the 3-D effect of the 357  

anomalies. 358  

359  

2.5. 3-D Inversion of the MT Profile 360  

Nine inversions were performed using different combinations of Z, Zo, H and T 361  

responses employing the ModEM algorithm (Egbert & Kelbert 2012, Kelbert et al., 362  

2014) modified for inclusion on H responses, (see supplementary material, S1, for more 363  

details). In all cases, the same inversion parameters were used to avoid differences 364  

associated with inversion set-up parameters and thus focusing only on the differences 365  

related to the investigated data types. 366  

367  

2.5.1 Inversion parameters 368  

The 3-D solution mesh, purposely chosen with a lower resolution than the forward mesh 369  

described above, comprised 137 x 120 x 99 cells in the north-south, east-west and 370  

Page 15 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 16: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

vertical directions, respectively. The centre of the mesh comprised 99 x 62 x 99 cells 371  

with a horizontal cell size of 200 m x 200 m, while the distance between sites was 600 372  

m, thus ensuring the presence of free cells between the sites. The lateral extent of the 373  

padding cells increased by a factor of 1.3. The thickness of the top layer was 10 m, 374  

increasing by a factor of 1.069 for the subsequent layers in the z-direction. The same 375  

responses studied in the dimensionality analysis, with added noise, were used for the 376  

inversion process, but only for periods between 0.0001 s (frequency of 10000 Hz) and 377  

31.62 s (frequency of 0.0316 Hz), focusing on the periods affected by the anomalies and 378  

avoiding the periods affected by the assumed hypothetical asthenosphere at 90 km depth. 379  

380  

Several tests were executed to choose the initial electrical resistivity model and the 381  

smoothing parameters. The starting model was chosen after inverting the Z responses, 382  

which are sensitive to lateral-vertical variations and to the actual resistivity of the 383  

subsurface, with seven different half-space starting models of 80, 100, 120, 150, 200 and 384  

400 Ωm, respectively. Inversion results starting with a half-space model of 120 Ωm 385  

recovered the most similar electrical resistivity values to the synthetic model used to 386  

generate the synthetic data, having a lower average distance between the final inversion 387  

result and the original synthetic model. For all the inversions a similar nRMS was 388  

obtained regardless of the different starting models. The smoothing values used for this 389  

test were 0.3 in all directions. As shown below, these smoothing values do not differ 390  

significantly from the smoothing values finally used during the inversion problem. 391  

392  

It is well known that regularization, necessary in any ill-posed inverse problem, has a 393  

large influence on the results of the inversion and particularly on resolution. In the case 394  

of the ModEM code we have to choose horizontal and vertical smoothing parameters. 395  

Page 16 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 17: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

To take care of that, we performed additional tests to determine the appropriate 396  

regularization parameters, being aware that different data types may react differently to 397  

the same smoothing parameters. Inverting Z and Z+T+H several combinations of 398  

horizontal and vertical smoothing parameters were executed using smoothing values 399  

between 0.1 and 0.6 (always assuming the same smoothing values for north-south and 400  

east-west directions). Searching for the simplest model that can fit the data, the largest 401  

smoothing parameters were chosen that adequately recover the synthetic model while 402  

still allowing the inversion processes to fit the data within the error bars. Based on these 403  

tests, the chosen starting model for all the inversions presented in this manuscript was a 404  

homogeneous half-space of 120 Ωm, and the smoothing values for all the inversions 405  

were 0.4 for north-south and east-west directions, and 0.3 for the vertical direction. Note 406  

that despite a smoothing of 0.5 for the north-south and east-west direction was also 407  

appropriate to fit Z data, it was not appropriate when inverting the combined Z+T+H 408  

responses, suggesting that the joint inversion process including all possible data requires 409  

more detail in the model to fit the data. No outstanding differences were observed when 410  

using 0.5 or 0.4 smoothing parameter for north-south and east-west directions when 411  

performing the inversion process with Z data. In the presented inversions, the a priori 412  

model was the same as the initial model (both are homogeneous), assuming that there is 413  

no prior information about the electrical resistivity values of the subsurface.” 414  

415  

An additional test was performed to evaluate if the choose neighbouring reference site 416  

for the H responses in the sensitivity test was adequate for the inversion process. Four 417  

inversions with Z and H responses were performed using four different reference sites 418  

for the H responses: (1) using a site close to the L2 anomaly; (2) on top of the L1 419  

anomaly; (3) on top of the H1 anomaly; (4) as in the sensitivity test, on top of a non-420  

Page 17 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 18: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

complex environment (almost 1-D). Results from the inversion process (Fig. 8a) show 421  

that locating the reference site on top of a non-complex environment, such as the one 422  

used for the sensitivity test, facilitate convergence of the inversion process (lower 423  

nRMS) also obtaining the model that better recovers the original synthetic model. Figure 424  

8b shows the nRMS of each inversion for each site and data type. From the results of 425  

this test, it is strongly recommended not to use a reference site on top of anomalies (L1 426  

and H1 in our case), in particular on top of a strong conductive anomaly such as L1. In 427  

this experiment, the inversion process using a reference site on top of L1 was not able to 428  

fit the data and the obtained model was the one that recovered the original synthetic 429  

model worst of all (Fig. 8). The inversion process using a neighbouring reference site 430  

close to L2 also does not converge to an nRMS lower than 1.2 but the model from the 431  

inversion process is superior to the two cases where the neighbouring reference site was 432  

on top of L1 and H1 anomalies. Note that the nRMS for the H response associated with 433  

the neighbouring reference site is always close to 1 in all cases. This is because the H 434  

responses compare the signal between two sites, and the signal between the 435  

neighbouring site with itself is the same. The fact that it is one and not zero is because of 436  

the noise we added in the responses. For the following inversions presented in this 437  

manuscript, the same site as the one used in the sensitivity test was used as a reference 438  

site for the H responses. 439  

440  

For the numerical experiment, four inversions were performed to examine and highlight 441  

the improvements that can be made by complementing the Zo and Z responses with the T 442  

and H responses, whilst also evaluating the necessity of using diagonal components of Z 443  

in these situations. The final inversions were (1) Zo, (2) Z, (3) Zo+H+T, and (4) Z+H+T. 444  

Four additional inversions, (5) Zo+H, (6) Zo+T, (7) Z+H and (8) Z+T were performed to 445  

Page 18 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 19: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

differentiate the contributions of the H and T responses independently. Finally, the 446  

capacity of the H and T responses to characterize the subsurface by themselves was 447  

considered in inversion (9) H+T. This latter one is interesting as no electric fields are 448  

involved, thus there should be primary sensitivity to lateral conductivity variations rather 449  

than to the absolute values of the conductivities themselves. 450  

451  

2.5.2. Results 452  

The results are displayed showing a north-south section of the obtained electrical 453  

resistivity models beneath the MT profile and with horizontal slices of the models at 650 454  

and 1550 m depths. Figure 9 shows the results of inversion (1) Zo, (2) Z, (3) Zo+H+T 455  

and (4) Z+H+T, highlighting the advantages of complementing Z or Zo with H and T. 456  

Figure 10 displays the results of inversions (5) Zo+T, (6) Zo+H, (7) Z+T, (8) Z+H and 457  

(9) H+T demonstrating the consequences of different contributions of H and T 458  

responses. The same colour scale was used in all figures. White dashed lines indicate the 459  

positions of the main resistivity structures, L1, L2, L3 and H1, to facilitate the 460  

comparison of the obtained inversion results with the original synthetic model. 461  

2.5.2.1 Fit of the data 462  

The fit of the data of each inversion process is shown in Fig. 11 plotting the summed 463  

normalized root mean square (nRMS) as calculated in the ModEM code (Egbert & 464  

Kelbert 2012, Kelbert et al., 2014) for each site, type of response and for each type of 465  

inversion, providing valuable information on how data misfit distributes along the 466  

profile for different inversions and different response types. Different colours were used 467  

for each type of response: red for Zo responses, black for Z responses, blue for T 468  

responses, and green for H responses. No systematic problems related to over-fitting 469  

were observed with a particular response type or site locations, although some of the 470  

Page 19 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 20: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

sites positioned at the edges of the profile have slightly larger nRMS than the ones 471  

located within the profile. Note that the nRMSes associated with Z and Zo responses are 472  

similar when only Z or Zo were used in the inversion process or when they were 473  

combined with T and H responses during the inversion process. This result shows that 474  

all datasets are fit to a similar level and that no remarkable decrease in resolution of a 475  

particular type of response is observed when other responses are used during the 476  

inversion process. 477  

478  

More detailed information about the fit of the data can be found in Fig. S2 479  

(Supplementary Material), where differences between data and model responses divided 480  

by the error were plotted for each type of inversion, component of the data, site and 481  

period in a pseudo-section format. Grey areas are periods with differences lower than the 482  

error; red and blue areas are periods with differences larger than the errors. 483  

484  

2.5.2.2 Electrical resistivity model beneath the MT profile 485  

Models resulting from inversion of the Zo or Z responses (Figs. 9a1, 9a2) show the 486  

existence and presence of the main anomalies but do not resolve or characterize them 487  

well; in particular the resistivity values directly below L1 are poorly resolved due to the 488  

shielding effects of L1. These two models appear to be unable to differentiate between 489  

anomaly L1 and the conducting anomalous layer L3 that underlies L1, implying a 490  

continuous horst-like structure with the L3 getting shallow to the depths of L1 in the 491  

middle of the profile. Additionally, the high electrical resistivity anomaly H1 is inferred 492  

but not well recovered, showing resistivity values around 80 Ωm, which are almost an 493  

order of magnitude lower than the correct value (400 Ωm). Note that these values (∼80 494  

Ωm) are also lower than the initial half-space model used for the inversion process (120 495  

Page 20 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 21: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Ωm). The use of the H and T responses as complements to the Zo or Z responses (Figs. 496  

9a3, 9a4) significantly improves resolution of all targets, conductive and resistive, 497  

showing L1 and L3 structures as disconnected (Z+H+T) or almost disconnected 498  

(Zo+H+T), thus better constraining the bottom of anomaly L1 and clearly recovering 499  

anomaly H1 in the northern part of the profile. The use of the H and T responses also 500  

better defines the north and south boundaries of anomalies L1 and H1. By 501  

complementing the Zo or Z responses with the H responses, the continuity of the low 502  

resistivity layer (L3) is better defined, although in none of the models is the continuity of 503  

L3 correctly recovered below L1 due to the electromagnetic shielding effects of L1 to 504  

any structure immediately below it. 505  

506  

2.5.2.3 Horizontal slices at 650 m and 1550 m depth 507  

The horizontal slices through the various 3-D resistivity models at 650 m depth are 508  

shown in Fig. 9b. If only the Zo responses are inverted, the model is symmetrical on both 509  

sides of the profile, thus characterizing the anomalies to extend outside of the profile but 510  

without indicating on which side the anomalies are located. This illustrates and 511  

emphasizes that the off-diagonal components contain no directional information for off-512  

profile structures. The use of diagonal components in inversion of the full Z responses 513  

characterizes the dimensionality of the structures whilst locating L2 to be west of the 514  

MT profile and inferring the westerly extension of L1 (compare Fig. 9b1 with Fig. 9b2). 515  

From Figs. 9b3 and 9b4 it is clear that by complementing the Zo or Z responses with H 516  

and T responses the correct dimensionality of the structures is recovered in both cases, 517  

showing in particular a remarkable improvement when Zo responses are complemented 518  

with H and T responses. Fairly equivalent models are obtained from the Zo+H+T and 519  

Z+H+T responses, but with slightly improved resolution, closer to the original synthetic 520  

Page 21 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 22: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

model, when using Z+H+T. The use of the H and T responses also improves the 521  

characterization of L2 by showing more reliable low resistivity values than when only 522  

using Zo or Z responses, even if the resistivity values are still higher than in the original 523  

synthetic model. Additionally, the north and south boundaries and the east and west 524  

propagation of L1 are better defined when complementing the Zo or Z responses with the 525  

H and T ones. 526  

527  

Figure 9c shows the horizontal slices of the final 3-D models at 1550 m depth. At this 528  

depth the synthetic model is a uniform 10 Ωm conducting layer (defined as anomaly L3) 529  

within which a 400 Ωm resistive anomaly (H1) is embedded. Models from the 530  

conventional MT, Zo and Z, responses (Figs. 9c1, 9c2), exhibit a more complex situation 531  

than might be expected from the original synthetic model, somewhat suggesting the 532  

presence of the H1 anomaly with the highest electrical resistivity values (∼ 100 Ωm) 533  

either directly below or outside of the profile, which is still lower than the 120 Ωm of 534  

the starting model used for the inversion process. By carrying out joint inversion 535  

including the H and T responses, the models better define the L3 layer and more 536  

accurately recover H1 below and outside the MT profile, with electrical resistivity 537  

values of around 250 Ωm, which is greater than the 120 Ωm of the starting model used 538  

for the inversion process (Figs. 9c3, 9c4) and closer to the 400 Ωm of the original 539  

synthetic model (Fig. 9c5). 540  

541  

2.5.2.4 Contribution of H and T 542  

Figure 10 shows the results of the inversion process combining Zo and Z responses with 543  

either T or H responses separately. From Fig. 10a, results below the profile, it can be 544  

seen that the inclusion of the H responses is responsible for superior differentiation 545  

Page 22 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 23: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

between L1 and L3. This differentiation also occurs when inverting Z+T but not when 546  

Zo is complemented with T responses (Zo+T). i.e., the use of diagonal terms of the 547  

impedance tensor is key to resolving the lack of interconnection between L1 and L3. In 548  

addition, the H responses contribute to better defining the continuity of the low electrical 549  

resistivity L3 layer. The main improvement gained by using the T responses is superior 550  

definition of the north and south boundaries of L1. 551  

Figure 10b shows the results at 650 m depth when inverting the H or T responses as a 552  

complement to the Zo and Z responses. In all cases, comparing with the results in Figs. 553  

9b1 and 9b2, the resolution and characterization of the anomalies are improved, 554  

especially in defining the L2 anomaly. The main differences between complementing 555  

with either the H or T responses are observed when characterizing the propagation of 556  

anomaly L1 to the west, which is constrained by the H and Z responses but not by the T 557  

responses, as expected from our sensitivity tests above. Figure 10c displays the model 558  

slices at 1550 m depth, showing the contribution of including the H and T responses 559  

separately from each other (cf. Figs. 9c). In all cases, the use of either or both the H and 560  

T responses lead to superior recovery of the H1 anomaly and show its propagation to the 561  

east. Figure 10c also shows that by using the H responses as a complement to the Z and 562  

Zo responses, the models better reproduce the electrical resistivity values of the original 563  

synthetic model off-profile. 564  

565  

Figures 10a5, 10b5 and 10c5 show that the magnetic-field only H and T responses alone 566  

are insufficient to characterize the subsurface by themselves. The associated models 567  

locate the anomalies along the north-south and east-west directions, but have no 568  

resolution in the vertical direction and do not recover the electrical resistivity values. 569  

This is because the magnetic field is sensitive to lateral resistivity contrasts (both 570  

Page 23 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 24: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

amplitude and position), but is insensitive to 1-D layering and to the actual resistivity 571  

values of the synthetic model 572  

 573  

2.5.2.5 Accuracy of the models 574  

One way to quantify how well the models obtained from the inversion process reproduce 575  

the synthetic model is to calculate the differences between them in model space. We 576  

have calculated these differences with a focus on the structures below the profile, where 577  

we expect to obtain higher resolution. Figure 12 shows the differences between the 578  

synthetic model and the models obtained from the inversion process, using decadic 579  

logarithm electrical resistivity values. Blue colours indicate results from the inversion 580  

process that are more resistive than the synthetic model, red colours indicate results 581  

more conductive than the synthetic model, and white colours indicate an almost-perfect 582  

fit between the inversion process model and the synthetic model. Note that the colour 583  

scale indicates differences between the two models that are greater than ±0.023; this 584  

value is equivalent to a difference of approx. ±0.5 Ωm for electrical resistivity values of 585  

10 Ωm, and a difference of approx. ±50 Ωm for electrical resistivity values of 1000 Ωm. 586  

The average differences for each type of inversion are shown in Table 1. 587  

588  

Table 1. Average differences between the final inversion models and the original 589  synthetic model. Differences calculated using log10 electrical resistivity values. 590  Inverted

Data Zo Z Zo+H+T Z+H+T Zo+T Zo+H Z+T Z+H H+T

Difference 0.151 0.149 0.133 0.134 0.144 0.141 0.134 0.138 0.214

591  

These results indicate the superior performance when reproducing the electrical 592  

resistivity values of the synthetic model when the Z or Zo responses are combined with T 593  

Page 24 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 25: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

and H responses. The results also corroborate that the electrical resistivity values 594  

obtained when using only the T and H responses are not correctly reproduced. 595  

596  

By conducting a more detailed analysis of Fig. 12, the following important points 597  

associated with the recovering of each anomaly become apparent: 598  

1) Structure L1: The edges and bottom are better recovered when T and H responses are 599  

each or both involved in the inversion process. Despite the electrical resistivity values 600  

always remaining more conductive than in the synthetic true model, they are better 601  

recovered when the H responses are included in the inversion process; 602  

2) Structure H1: In all the inversion processes the electrical resistivity values associated 603  

with H1 are less resistive than in the synthetic model, however, recovery of these 604  

resistivity values, and the geometry of the anomaly, is always improved when the H and 605  

T responses are each or both involved in the inversion process. 606  

3) Structure L3: The electrical resistivity values are improved when the H responses are 607  

included in the inversion process. Note that when the T responses are used in the 608  

inversion process, the top of the anomaly is also better characterized, particularly in the 609  

southern part of the profile. 610  

4) Background 1-D model: Results from the inversion processes complementing Z or Zo 611  

with the H responses are the ones that best recover the electrical resistivity values of the 612  

background 1-D resistivity model; not only do the results show smaller differences with 613  

respect to the original synthetic model, but they also show a more homogeneous (less 614  

blobby) resistivity model than the other inversion results. 615  

616  

3. DISCUSSION 617  

Focusing on the recovery of electrical resistivity structures in 3-D areas, the use of the 618  

full impedance tensor responses (Z), instead of using only the off-diagonal elements 619  

Page 25 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 26: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

(Zo), is highly recommended for all 3-D inversions because of the far superior capacity 620  

to reproduce the dimensionality of the geoelectrical structures in the subsurface. It 621  

reaffirms and further validates the conclusions drawn and emphasized previously by 622  

Siripunvaraporn et al. (2005b), Tietze & Ritter (2013), and Kiyan et al. (2014). 623  

However, it is not always possible to use all of the components of the MT impedance 624  

tensor because diagonal components, being of lower amplitude, are far more affected by 625  

the presence of noise. This requires the necessity of including robust complementary 626  

information in these situations. Siripunvaraporn & Egbert (2009) highlighted the 627  

advantages of using Z+T data during the 3-D inversion process, suggesting that 628  

resistivity values, geometry of the anomalies, and depths are all better characterized. 629  

However, not all of the studies show major advantages when performing joint inversion 630  

of Z+T responses in a 3-D environment. Tietze & Ritter (2013) present a situation where 631  

results from Z responses alone are more reliable than the results from Z+T responses. In 632  

previous synthetic studies working with T and H responses, Habibian and Oskooi (2014) 633  

show that inversion processes using either or both T and H responses have vertical 634  

resolution and are able to recover the correct electrical resistivity values of the 635  

subsurface, which does not happen in the experiment shown in this manuscript (Fig. 10). 636  

However, in the previous study the background structure was homogeneous with 637  

uniform resistivity and the same as the one used for the starting model for the inversion. 638  

This configuration masks the incapacity of the H and T responses on their own to 639  

recover vertical anomalies or the correct electrical resistivity values. The numerical 640  

experiment design in this manuscript, using a 1-D layered model as a background, 641  

highlight this deficiency of the H and T responses, showing that vertical resolution or 642  

the obtained electrical resistivity values from an inversion using either or both T and H 643  

responses are only reliable if additional information defining the background structures 644  

Page 26 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 27: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

of the study area is added. However, complementing Z and Zo responses, T and H aids 645  

superior determination of the electrical resistivity values of the subsurface, including the 646  

characterisation of the absolute electrical resistivity values (see Table 1 and Fig. 12). 647  

Results from Varentsov (2015a,b) combining Z, H and T responses using 2-D, 2-D+ and 648  

a quasi 3-D approaches encourage the use of H responses for the inversion process, 649  

improving the accuracy of the results, and better separating close spaced conductors in 650  

complex geoelectrical situations. 651  

652  

In this work, the advantages of using T and H responses as a complement to Z or Zo 653  

responses in a 3-D environment are evaluated from the results of a numerical 654  

experiment. Starting with the sensitivity of each response type to the studied anomalies, 655  

the results of our numerical experiment exhibit that both the H and T responses are also 656  

sensitive to the dimensionality of the structures and can be used to successfully 657  

reproduce the dimensionality of the anomalies when complementing Zo or Z responses 658  

(Figs. 9 and 10). The sensitivity tests performed (Fig. 6) demonstrate that the Zo and Z 659  

responses are intrinsically more sensitive to the existence of the anomalies than are the 660  

H and T responses. However, the dimensionality analysis (Fig. 7) and the capacity for 661  

recovering the detail of the geoelectrical structures (Figs. 9, 10 and 12) are remarkably 662  

improved when the Zo or Z responses are complemented with either, and especially both, 663  

the H and T responses. Each response type has a specific behaviour when characterizing 664  

the dimensionality of the structures and consistency between all of them will support the 665  

ultimately adopted dimensionality of the study area. The use of the T and H responses 666  

can also improve the dimensionality analysis in situations where the phase tensor is not a 667  

strong tool; see for example the results close to the L2 and H1 anomalies (Fig. 7). In that 668  

case the phase tensors do not correctly define the presence of 3-D structures having no 669  

Page 27 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 28: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

periods showing absolute values of the β angle divided by the associated error greater 670  

than 1. However, by using the H and T responses the characterisation of 3-D structures 671  

of these areas is significantly improved. Comparing the dimensionality analysis from the 672  

H and T responses close to resistive anomaly H1, induction arrows clearly point in 673  

different directions than the direction of Phi_max from the H responses, which is also 674  

suggestive of a 3-D situation. 675  

676  

Focusing now on the ability of recovering the correct electrical resistivity distribution of 677  

the subsurface, the observed improvements are associated with the use of different types 678  

of responses that better restrict the model space than if any are used individually. More 679  

specifically, this phenomenon occurs because the Z responses are far more sensitive to 680  

the L1 anomaly than to the L2 and H1 anomalies, whereas for our model the T and H 681  

responses have similar sensitivity to all of the anomalies (Fig. 6). Such sensitivity means 682  

that the inversion of the Zo or Z responses focuses mainly on L1, with less importance 683  

placed upon the L2 and H1 anomalies. Results from the sensitivity tests (Figs. 2, 3, 4 684  

and 5) can also be used to understand how inclusion of the H and T responses improves 685  

the characterization of the subsurface. 686  

687  

Inspecting the resolved structures beneath the MT profile (Figs. 9, 10 and 12), four main 688  

improvements are apparent when performing joint inversion of the Zo or Z responses 689  

with H and T ones: 690  

(1) Characterization of the bottom of the L1 anomaly with better differentiation between 691  

the L1 and L3 structures; 692  

(2) The possibility of recovering the high resistivity anomaly, H1, surrounded by low 693  

resistivity values, L3, that are poorly recovered when using only Zo or Z responses; 694  

Page 28 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 29: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

(3) Better resolution in determining the resistivity values of the subsurface, obtaining 695  

results more similar to the original synthetic model; 696  

(4) Superior definition of the shape of the anomalies in the north-south direction, i.e. 697  

along profile, with a better fit inside the white dashed lines in the figures. 698  

699  

Despite all these advantages, we would like to point out that in all models the location of 700  

the L3 layer at the edges of the profile is somewhat shallower than what it should be. 701  

This problem is not properly corrected when complementing Z or Zo with T and H 702  

responses, although results with T responses seem to slightly improve the depth of the 703  

conductive layer L3 (Figs. 9, 10 and 12). 704  

705  

Looking outside the line of the profile (Figs. 9b, 9c and 10b, 10c), L2 and H1 are better 706  

recovered when the inversion is complemented with H and T responses. However, 707  

although the characterization of the off-profile anomalies is improved when 708  

complementing the Z responses with the H and T ones, the shape of the anomalies and 709  

the electrical resistivity values are still not properly constrained, probably as a 710  

consequence of a combination of (i) the non-uniqueness of the MT method, and (ii) 711  

scatter and error. 712  

713  

Inspecting the results obtained from the T responses, the main signal associated with the 714  

anomalies is located at the edge of the structures (Figs. 2, 3, 4, 5, 6), making the T 715  

responses a good type of data to define boundaries of the structures along the MT 716  

profile. Note that from Fig. 6d this edge-detection is more precise for low resistivity 717  

anomalies (L1) than for high resistivity anomalies (H1). This can also be observed when 718  

recovering the anomalies from the inversion process (Figs. 9 and 10) where the north 719  

Page 29 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 30: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

and south boundaries of L1 and H1 are improved when the T responses are included. 720  

This result, highlighting the capacity of the T responses characterising the edges of the 721  

anomalies, is consistent with previous results, such as Siripunvaraporn & Egbert (2009) 722  

and Habibian and Oskooi (2014). In addition, the top of the low electrical resistivity 723  

layer L3 is improved when the inversion of Z or Zo data is complemented with T data 724  

(Figs. 10a1, 10a3, 12c and 12d). This makes T data a good complement to improve the 725  

definition in recovering the edges and shape of the anomalies below a profile located in 726  

a 3-D environment, in agreement with results suggested by Siripunvaraporn & Egbert 727  

(2009). Note that, apart from the H1 anomaly, no major improvement associated with 728  

the use of T responses has been observed in defining the electrical resistivity values of 729  

the anomalies or the background 1-D model (Figs. 12c, 12d). Results from Fig. 10 730  

shows that, although the use of T data as a complement of Zo provides an improvement 731  

in characterizing the subsurface, the results are remarkably better when Z and T 732  

responses are both used in the inversion process. 733  

734  

Examining the results from the H responses, the major advantage is that their use makes 735  

the resistivity model more consistent as a whole as all the sites are related to the 736  

neighbouring reference site and the intensity of the anomalies tend to be consistent along 737  

the whole profile. This effect is also observed in the sensitivity test (Figs. 4b and 6c). In 738  

this case the neighbouring reference site is located close to H1 and is affected by its 739  

presence, which is why the effect of H1 can be seen at sites south of the profile. This 740  

property of H responses increases the consistency of the model as a whole for shallow 741  

and deep structures independently of the extension of the area of the study. This level of 742  

consistency is not required when using the Zo, Z or T responses, where each site is 743  

mostly focused on the anomalies nearby but without any explicit link to anomalies that 744  

Page 30 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 31: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

are located far away along the study area, where the periods of the site are insensitive. 745  

The consequences of this property in the inversion process can be seen in Fig. 12, where 746  

the models including the H responses obtain electrical resistivity values of the anomalies 747  

and the background 1-D resistivity model closer to the original synthetic model. Note 748  

that from Fig. 10, although results are better quality when Z+H are used in the inversion 749  

process, instead of Zo+H, differences are not as remarkable as the differences observed 750  

between Z+T and Zo+T, having similar results from both types of inversion. Note also 751  

that, from the performed test shown in Fig. 8, to facilitate and improve the results from 752  

the inversion process using H responses it is recommended to use a neighbouring 753  

reference site located on top of a non-complex geoelectrical area. Results from the 754  

dimensionality analysis or from prior inversions using Z alone or Z+T can be used to 755  

choose the optimum neighbouring reference site for Z+H or Z+T+H inversion. 756  

757  

4. CONCLUSIONS 758  

759  

The use of the H and T responses is demonstrated to be a crucial complement to the Zo 760  

or Z ones in order to better resolve and characterize the structures beneath MT profiles 761  

in 3-D environments. For the synthetic model considered here, large differences were 762  

seen whether the diagonal components of Z were used or not, having far better results 763  

when the diagonal components were included. These differences were reduced when 764  

including the H and T responses in the inversion process, although results using the 765  

diagonal components of Z still recover the subsurface better than when the diagonal 766  

components are not used. From the results of this experiment, four main improvements 767  

are observed when carrying out joint inversion of the Zo or Z responses with the H and T 768  

ones: 769  

Page 31 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 32: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

(1) More accurate characterization of the electrical resistivity values of the studied 770  

anomalies and also of the background resistivity model; 771  

(2) Superior capability for recovering high electrical resistivity anomalies than by using 772  

the Zo or Z responses alone; 773  

(3) Improved definition of the bottom of low electrical resistivity anomalies and the 774  

north, south, east and west boundaries of the anomalies beneath and outside the MT 775  

profile; and 776  

(4) Superior imaging of the horizontal continuity of low electrical resistivity anomalies. 777  

Additionally, note that the use of the H responses increases the reliability of the 778  

resistivity model as a whole, reducing the possibility for the inverse problem to generate 779  

spurious anomalies if the anomaly is not consistent with the data acquired at the 780  

neighbouring reference site. Although any site could be used as a neighbouring reference 781  

site, as far as the structures below the neighbouring reference site are properly 782  

constrained, it is recommended to use a site located on top of a simple geoelectrical area 783  

and it is particularly strongly recommended not to use a site on top of a strong anomaly. 784  

785  

All of these advantages incentivise the use of H and T responses as complements Z or Zo 786  

responses, in particular when characterizing the subsurface beneath a MT profile in a 3-787  

D environment. This supports the possibility of high-resolution studies in 3-D 788  

environments without expending large amounts of economical and computational 789  

resources, and also opening the door for the achievement of targets with high electrical 790  

resistivity values. 791  

792  

5. ACKNOWLEDGMENTS 793  

Page 32 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 33: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

We would like to acknowledge the financial support to JC and XO from the IRECCSEM 794  

project (www.ireccsem.ie) funded by a Science Foundation of Ireland Investigator 795  

Project Award (SFI: 12/IP/1313) to AGJ. Gary Egbert, Anna Kelbert and Naser Meqbel 796  

are very gratefully thanked for making their ModEM code available to the community. 797  

The authors wish to acknowledge Irish Centre for High-End Computing (ICHEC) for the 798  

computing facilities provided. Finally, the authors would like to thank Stephan Thiel, the 799  

two anonymous reviewers, and the Associate Editors Mark Everett and Ute Weckmann 800  

for their very helpful suggestions and comments on our submitted versions that 801  

improved our final paper. 802  

803  

804  

6. REFERENCES 805  

1. Avdeev, D.B. & Avdeeva, A., 2009. 3D magnetotelluric inversion using a 806  

limited-memory quasi-Newton optimization, Geophysics, 74(3), F45– F57. 807  

2. Baba, K., Chave, A.D., Evans, R.L., Hirth, G. & Mackie, R.L., 2006. Mantle 808  

dynamics beneath the East Pacific Rise at 17S: insights from the Mantle 809  

Electomagnetic and Tomography (MELT) experiment, J. Geophys. Res., 111, 810  

B02101, doi:10.1029/2004JB003598. 811  

3. Bedrosian, P.A., 2007. MT+, Integrating Magnetotellurics to Determine Earth 812  

Structure, Physical State, and Processes. Surv. Geophys, 28: 121-167. doi: 813  

10.1007/s10712-007-9019-6 814  

4. Berdichevsky, M.N., Dmitriev, V. & Pozdnjakova, E.E., 1998. On two-815  

dimensional interpretation of magnetotelluric soundings, Geophys. J. Int., 133, 816  

585–606. 817  

5. Berdichevsky M.N. & Dmitriev V.I., 2008. Models and Methods of 818  

Magnetotellurics.Springer-Verlag Berlin Heidelberg. 819  

6. Berdichevsky, M.N., Golubtsova, N.S., Varentsov, Iv. M., Pushkarev, P.Yu., 820  

Rybin, A.K. & Sokolova, E. Yu., 2010a. Geoelectric section of the Central Tien 821  

Shan: Sequential inversion of the magnetovariational and magnetotelluric data 822  

Page 33 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 34: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

along the Naryn Line, Physics of the Solid Earth, 46(8), 698-706. 823  

doi:10.1134/S1069351310080069. 824  

7. Berdichevsky, M.N., Sokolova, E.Yu., Varentsov, Iv.,M., Rybin, A.K., 825  

Baglaenko, N.V., Batalev, V.Yu., Golubtsova, N.S., Matayukov, V.E. & 826  

Pushkarev, P.Yu., 2010b. Geoelectric section of the Central Tien Shan: Analysis 827  

of magnetotelluric and magnetovariational Responses along the Naryn 828  

Geotransverse, Physics of the Solid Earth, 46(8), 679-697. 829  

doi:10.1134/S1069351310080057. 830  

8. Booker, J.R., 2014. The Magnetotelluric Phase Tensor: A Critical Review. Surv. 831  

Geophys. 35,7-40. doi:10.1007/s10712-013-9234-2 832  

9. Brasse, H., P. Lezaeta, V. Rath, K. Schwalenberg, W. Soyer, & V. Haak, 2002. 833  

The Bolivian Altiplano conductivity anomaly, J. Geophys. Res., 107(B5), 2096, 834  

doi:10.1029/2001JB000391 835  

10. Caldwell, G.T., Bibby, H.M., Brown, C., 2004. The Magnetotelluric Phase 836  

Tensor, Geophys. J. Int., 158, 457–469 837  

11. Campanyà, J., Ledo, J., Queralt, P., Marcuello, A., Liesa, M. & Muñoz, J.A. 838  

2011. Lithospheric characterization of the Central Pyrenees based on new 839  

magnetotelluric data. Terra Nova. 23, 213 - 219. doi: 10.1111/j.1365-840  

3121.2011.01001.x . 841  

12. Campanyà, J., Ledo, J., Queralt, P., Marcuello, A., Liesa, M. & Muñoz, J.A. 842  

2012. New geoelectrical characterization of a continental collision zone in the 843  

West- Central Pyrenees: Constraints from long period and broadband 844  

magnetotellurics. Earth Planet. Sci. Lett., 333-334, 112-845  

121.doi:10.1016/j.epsl.2012.018 846  

13. Campanyà, J., J. Ledo, P. Queralt, A. Marcuello, & A.G. Jones, 2014. A new 847  

methodology to estimate magnetotelluric (MT) tensor relationships: Estimation 848  

of Local transfer-functIons by Combining Inter-station Transfer-functions 849  

(ELICIT), Geophy. J. Int., 198, 484-494, doi: 10.1093/gji/ggu147. 850  

14. Chave, A.D. & Jones, A.G., 1997. Electric and magnetic field galvanic distortion 851  

decomposition of BC87 data, Journal of Geomagnetism and Geoelectricity, 49, 852  

767-789. 853  

15. Chave, A.D. & Jones, A.G., 2012. The Magnetotelluric Method. Cambridge 854  

University press. United Kingdom. 855  

Page 34 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 35: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

16. Chave, A.D. & Smith, J.T., 1994. On electric and magnetic galvanic distortion 856  

tensor decompositions, Journal of Geophysical Research-Solid Earth, 99, 4669-857  

4682. 858  

17. Dong, H., Wei, W., Gaofeng, Y., Jin, S., Jones, A.G., Jing, J., Zhang, L., Xie, C., 859  

Zhang, F. & Wang, H., 2014. Three-dimensional electrical structure of the crust 860  

and upper mantle in Ordos Block and adjacent area: Evidence of regional 861  

lithospheric modifications, Geochem. Geophys. Geosyst., 15, 862  

doi:10.1002/2014GC005270. 863  

18. Egbert, G. D., & Kelbert, A., 2012. Computational recipes for electromagnetic 864  

inverse problems, Geophys. J. Intern., 189(1), 251-267. 865  

19. Falgàs, E., Ledo, J., Marcuello, A. & Queralt., P., 2009. Monitoring freshwater-866  

seawater interface dynamics with audiomagnetotellurics data, Near Surface 867  

Geophysics,7 (5-6), 391-399. 868  

20. Farquharson, C., Oldenburg, D., Haber, E. & Shekhtman, R., 2002. An al- 869  

gorithm for the three-dimensional inversion of magnetotelluric data, SEG 870  

Technical Program Expanded Abstracts, Society of Exploration Geo- physicists. 871  

21. Gabàs, A. & Marcuello, A., 2003. The relative influence of different types of 872  

magnetotelluric data on joint inversions. Earth, Planets and Space. 55 (5), 243- 873  

248. 874  

22. Grayver, A.V., 2015. Parallel three-dimensional magnetotelluric inversion using 875  

adaptive finite-element method. Part I: theory and synthetic study, Geophysical 876  

J. Int., 202, 584-603. 877  

23. Habibian, B.D.; Brasse, H.; Oskooi, B.; Ernst, T.; Sokolova, E.; Varentsov, Iv. & 878  

EMTESZ Working Group, 2010. The conductivity structures across the Trans-879  

European Suture Zone from magnetotelluric and magnetovariational data 880  

modeling. Phys. Earth Planet.Int., 183, 377-386. 881  

24. Habibian, B.D. & Oskooi, B., 2014. A resolution comparison of horizontal and 882  

vertical magnetic transfer functions. Journal of the Earth and Space Physics, 40 883  

(3), 47-53. 884  

25. Heinson, G.S.; Constable, S.C. & White, A., 2000. Episodic melt transport at 885  

mid-ocean ridges inferred from magnetotelluric sounding. Geophy. Res. Lett., 27, 886  

2317- 2320. 887  

Page 35 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 36: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

26. Heinson, G.S., Direen, N.G., Gill, R.M., 2006. Magnetotelluric evidence for a 888  

deep crustal mineralizing system beneath the Olympic Dam iron oxide copper-889  

gold deposit, southern Australia. Geology, 34, 573-576, doi:10.1130/G22222.1. 890  

27. Jones, A.G., 1983. The problem of current channelling: A critical review, 891  

Surveys in Geophysics, 6, 79-122. 892  

28. Jones, A.G., 1986. Parkinson's pointers' potential perfidy!, Geophysical Journal 893  

of the Royal Astronomical Society, 87, 1215-1224. 894  

29. Jones, A.G., 1999. Imaging the continental upper mantle using electromagnetic 895  

methods, Lithos, 48, 57-80. 896  

30. Jones, A. G. 2012, Distortion of magnetotelluric data: its identification and 897  

removal, in A. D. Chave and A. G. Jones, eds., The Magnetotelluric Method: 898  

Theory and Practice: Cambridge University Press. 899  

31. Jones, A.G., Evans, R.L., Muller, M.R., Hamilton, M.P., Miensopust, M.P., 900  

Garcia, X., Cole, P., Ngwisanyi, T., Hutchins, D., Fourie, C.J.S., Jelsma, H., 901  

Evans, S., Aravanis, T., Pettit, W., Webb, S., Wasborg, J. & Team, S., 2009. 902  

Area selection for diamonds using magnetotellurics: Examples from southern 903  

Africa. in Lithos, pp. 83-92. 904  

32. Jones, A.G. & Garcia, X., 2003. Okak Bay AMT data-set case study: Lessons in 905  

dimensionality and scale, Geophysics, 68, 70-91. 906  

33. Kelbert, A., Meqbel, N., Egbert, G. D., & Tandon, K., 2014. ModEM: A 907  

Modular System for Inversion of Electromagnetic Geophysical Data. Computers 908  

& Geosciences. 66,40-53. http://dx.doi.org/10.1016/j.cageo.2014.01.010. 909  

34. Kiyan, D., A.G. Jones, & J. Vozar, 2014. The inability of magnetotelluric off-910  

diagonal impedance tensor elements to sense oblique conductors in 3-D 911  

inversion. Geophys. J. Int., 196, 1351-1364, doi: 10.1093/gji/ggt470. 912  

 913  35. Kühn, C., Küster, J. & Brasse, H., 2014. Three-dimensional inversion of 914  

magnetotelluric data from Central Andean continental margin, Earth, Planets 915  

and Space, 66:112. 916  

36. La Terra, E.F. & Menezes, P.T.L., 2012. Audio-magnetotelluric 3D imaging of 917  

the Regis kimberlite pipe, Minas Gerais, Brazil. J. of Appl. Geophys., 77, 30-38. 918  

37. Le Pape, F.; Jones, A.G.; Vozar, J. & Wenbo, W.; 2012. Penetration of crustal 919  

melt beyond the Kunlun Fault into northern Tibet. Nature Geoscience, 5, 330-920  

335. 921  

Page 36 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 37: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

38. Ledo, J., Queralt, P., Marti, A. & Jones, A.G., 2002. Two-dimensional inter-922  

pretation of three-dimensional magnetotelluric data: an example of limitations 923  

and resolution, Geophys. J. Int., 150, 127–139. 924  

39. Ledo, J., 2005. 2-D versus 3-D magnetotelluric data interpretation, Surv. 925  

Geophys., 26, 511–543. 926  

40. Ledo, J.; Jones, A.G.; Siniscalchi, A.; Campanyà, J.; Kiyan, D.; Romano, G.; 927  

Rouai, M.; TopoMed MT Team, 2011. Electrical signature of modern and 928  

ancient tectonic processes in the crust of the Atlas mountains of 929  

Morocco. Physics of the Earth and Planetary Interiors, 185, 82-930  

88,doi:10.1016/j.pepi.2011.01.008. 931  

41. Lezaeta, P., Haak, V. ,2003. Beyond magnetotelluric decomposition: induction, 932  

current channelling, and magnetotelluric phases over 90. J.Geophys.Res. 108 933  

(B6), 2305 ,http://dx.doi.org/10.1029/2001JB000990. 934  

42. Marcuello, A., Queralt, P., Ledo, J., 2005. Applications of dispersion relations to 935  

the geomagnetic transfer function. Physics of the Earth and Planetary Interiors, 936  

150, 85-91. 937  

43. Miensopust, M.P., P. Queralt, A.G. Jones, & the 3D MT modellers, 2013. 938  

Magnetotelluric 3D inversion -a review of two successful workshops on forward 939  

and inversion code testing and comparison. Geophys. J. Int., 193, 1216-1238, 940  

doi: 10.1093/gji/ggt066. 941  

44. Meqbel, N.M; Egbert, G.D., Wannamaker, P.E., Kelbert, A., Schultz, A., 2014. 942  

Deep electrical resistivity structure of the northwestern U.S. derived from 3-D 943  

inversion of USArray magnetotelluric data. Earth and Planetary Science Letters, 944  

402, 290-304. doi:10.1016/j.epsl.2013.12.026. 945  

45. Muñoz, G., 2014. Exploring for geothermal Resources with Electromagnetic 946  

Methods, Surveys in Geophysics, 35 (1), 101-122. 947  

46. Ogaya, X., Ledo, J., Queralt, P., Marcuello, A. & Quintà, A., 2013.First 948  

geoelectrical image of the subsurface of the Hontomín site (Spain) for CO 2 949  

geological storage: a magnetotelluric 2D characterization. International Journal 950  

of Greehouse Gas Control, 13, 168-179. 951  

47. Ogaya, X., Queralt, P., Ledo, J., Marcuello, A. & Jones, A.G., 2014. 952  

Geoelectrical baseline model of the subsurface of the Hontomin site (Spain) for 953  

CO 2 geological storage in a deep saline aquifer: A 3D magnetotelluric 954  

characterization. International Journal of Greenhouse Gas Control, 27, 120-138. 955  

Page 37 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 38: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

48. Park, S.K. & Mackie, R.L., 2000. Resistive (dry?) lower crust in an active 956  

orogen, Nanga Parbat, northern Pakistan, Tectonophysics, 316, 359-380. 957  

49. Piña-Varas, P., Ledo, J., Queralt, P., Marcuello, A., Bellmunt, F., Hidalgo, R. & 958  

Messeiller., 2014. 3-D Magnetotelluric Exploration of Tenerife Geothermal 959  

System (Canary Islands, Spain). Surveys in Geophysics, 35 (4), 1045-1064. 960  

50. Rao, C.K., Jones, A.G., Moorkamp, M. & Weckmann, U. 2014, Implications for 961  

the lithospheric geometry of the Iapetus suture beneath Ireland based on 962  

electrical resistivity models from deep- probing magnetotellurics. Geophysical 963  

Journal International, 198, 737-759, doi: 10.1093/gji/ggu136. 964  

51. Rosell, O., Martí, A., Marcuello, À., Ledo, J., Queralt, P., Roca, E. & Campanyà, 965  

J., 2011. Deep electrical resistivity structure of the northern Gibraltar Arc 966  

(western Mediterranean): evidence of lithospheric slab break-off. Terra 967  

Nova. 23, 179 - 186. doi: 10.1111/j.1365-3121.2011.00996.x. 968  

52. Sasaki, Y., 2004. Three-dimensional inversion of static-shifted magnetotelluric 969  

data, Earth Planets and Space, 56, 239-248. 970  

53. Sasaki, Y. & Meju, M.A., 2006. Three-dimensional joint inversion for 971  

magnetotelluric resistivity and static shift distributions in complex media, 972  

Journal of Geophysical Research-Solid Earth, 111, 11. 973  

54. Schmucker, U., 1970. Anomalies of geomagnetic variations in the Southwestern 974  

United States. J. Geomagn. Oceanogr., Univ. of California Press, Berkeley. 975  

55. Seama, N., Baba, K., Utada, H., Toh, H., Tada., N., Ichiki, M. & Matsuno, T., 976  

2007. 1-D electrical conductivity structure beneath the Philippine Sea: Results 977  

from an ocean bottom magnetotelluric survey. Physics of the Earth and 978  

planetary interiors, 162, 2-12. 979  

56. Siripunvaraporn, W., Egbert, G., Lenbury, Y. & Uyeshima, M., 2005a. Three- 980  

dimensional magnetotelluric inversion: data-space method, Phys. Earth planet. 981  

Inter., 150, 3–14. 982  

57. Siripunvaraporn, W. and Egbert, G., 2009. WSINV3DMT: vertical magnetic 983  

field transfer function inversion and parallel implementation, Phys. Earth Planet. 984  

Inter., 173, 317—329. 985  

58. Siripunvaraporn, W., Egbert, G. & Uyeshima, M., 2005b. Interpretation of two-986  

dimensional magnetotelluric profile data with three-dimensional inversion: 987  

synthetic examples, Geophys. J. Int., 160, 804–814. 988  

Page 38 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 39: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

59. Simpson, F. & Bahr, K., 2005. Practical Magnetotellurics. Cambridge University 989  

press. United Kingdom. 990  

60. Soyer , W & Brasse, H., 2001. A magneto-variational array study in the Central 991  

Andes of N Chile and SW Bolivia, Geophysical Research Letters, 28, 3023-992  

3026. 993  

61. Soyer, W., 2002. Analysis of geomagnetic variations in the Central and Southern 994  

Andes. PhD Thesis, Freie Universität Berlin, Fachbereich Geowissenschaften, 995  

Berlin. 996  

62. Thiel, S., Heinson, G., Gray D.R., and Gregory. R.T., 2009. Ophiolite 997  

emplacement in NE Oman: constraints from magnetotelluric sounding. 998  

Geophysical Journal International, 176(3):753-766. 999  

63. Tietze, K. & Ritter, O., 2013. Three-dimensional magnetotelluric inversion in 1000  

practice-the electrical conductivity structure of the San Andreas Fault in Central 1001  

California, Geophys. J. Int., 195, 130-147. 1002  

64. Tuncer, V.M., Unsworth, E.M. & Pana, D., 2009. Deep electrical structure of 1003  

northern Alberta (Canada): implications for diamond exploration, Can. J. Earth 1004  

Sci., 46, 139-154. 1005  

65. Unsworth, M.J., 2005. New developments in conventional hydrocarbon 1006  

exploration with EM methods. Canadian Society of Exploration Geophysicists 1007  

Recorder, pp. 34–38. [April.] 1008  

66. Usui, Y., 2015. 3-D inversion of magnetotelluric data using unstructured 1009  

tetrahedral elements: applicability to data affected by topography, Geophys. J. 1010  

Int., 202, 828-849.  1011  

67. Varentsov Iv.M., 2015a. Methods of joint robust inversion in MT and MV 1012  

studies with application to synthetic datasets // Electromagnetic Sounding of the 1013  

Earth's Interior: Theory, Modeling, Practice (Ed. V.V. Spichak). Amsterdam: 1014  

Elsevier. P. 191-229.  1015  

68. Varentsov Iv.M., 2015b. Arrays of simultaneous EM soundings: design, data 1016  

processing, analysis, and inversion // Electromagnetic Sounding of the Earth's 1017  

Interior: Theory, Modeling, Practice (Ed. V.V. Spichak). 1018  

Amsterdam: Elsevier. P. 271-299. 1019  

69. Vozoff, K., 1991. The magnetotelluric method, Electromagneitc methods in 1020  

applied geophysics –Aplications, Chapter 8, 1991. 1021  

Page 39 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 40: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

70. Wanamaker, P.E., Hohmann, G.W. & Ward, S.H., 1984. Magnetotelluric 1022  

responses of three-dimensional bodies in layered earths, Geophysics, 49, 1517–1023  

1533. 1024  

71. Yang, B., Egbert, G.D., Kelbert, A., Meqbel, N., 2015. Three-dimensional 1025  

electrical resistivity of the north-central USA from EarthScope long period 1026  

magnetotelluric data. Earth and Planetary Science Letters, 422, 87–93. 1027  

1028  

1029  

FIGURE CAPTIONS: 1030  

 1031  

Figure 1: Synthetic geoelectrical model composed of four layers: (1) Between 0 m and 1032  

90 m depth with an electrical resistivity of 100 Ωm; (2) Between 90 m and 1500 m depth 1033  

with an electrical resistivity of 400 Ωm; (3) Between 1500 m and 2000 m depth with an 1034  

electrical resistivity of 10 Ωm; (4) Layer from 2000 m to 90 km depth with an electrical 1035  

resistivity of 200 Ωm. Two low electrical resistivity structures (L1 and L2) and a high 1036  

electrical resistivity body (H1) are also present. L1 and L2 are located between 500 m 1037  

and 1000 m depth and have an electrical resistivity value of 3 Ωm. H1 is located 1038  

between 1500 m and 2000 m depth with electrical resistivity values of 400 Ωm. The 1039  

green arrows points to site 26, which is used as a neighbouring reference site for H 1040  

responses. 1041  

1042  

Figure 2: Results of the sensitivity test for the L1 anomaly. Values obtained by 1043  

calculating the differences between the forward responses of the 1-D layered model 1044  

containing L1 anomaly and the forward responses of the background 1-D layered model 1045  

without anomalies. The differences are divided by the assumed error, thus showing how 1046  

sensitive the responses are to the anomaly with respect to the errors. White areas are 1047  

Page 40 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 41: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

periods that are non-sensitive to L1 (differences smaller than the assumed error). Red 1048  

and blue areas are periods that are sensitive to L1. 1049  

1050  

Figure 3: Results of the sensitivity test for the L2 anomaly. Values obtained by 1051  

calculating the differences between the forward responses of the 1-D layered model 1052  

containing L2 anomaly and the forward responses of the background 1-D layered model 1053  

without anomalies. The differences are divided by the assumed error, thus showing how 1054  

sensitive the responses are to the anomaly with respect to the errors. White areas are 1055  

periods that are non-sensitive to L2 (differences smaller than the assumed error). Red 1056  

and blue areas are periods that are sensitive to L2. 1057  

1058  

Figure 4: Results of the sensitivity test for the H1 anomaly. Values obtained by 1059  

calculating the differences between the forward responses of the 1-D layered model 1060  

containing H1 anomaly and the forward responses of the background 1-D layered model 1061  

without anomalies. The differences are divided by the assumed error, thus showing how 1062  

sensitive the responses are to the anomaly with respect to the errors. White areas are 1063  

periods that are non-sensitive to H1 (differences smaller than the assumed error). Red 1064  

and blue areas are periods that are sensitive to H1. 1065  

1066  

Figure 5: Results of the sensitivity test for the combined L1+L2+H1 anomalies. Values 1067  

obtained by calculating the differences between the forward responses of the 1-D 1068  

layered model containing L1+L2+H1 anomalies and the forward responses of the 1069  

background 1-D layered model without anomalies. The differences are divided by the 1070  

assumed error, thus showing how sensitive the responses are to the anomalies with 1071  

respect to the errors. White areas are periods that are non-sensitive to L1+L2+H1 1072  

Page 41 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 42: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

anomalies (differences smaller than the assumed error). Red and blue areas are periods 1073  

that are sensitive to L1+L2+H1. 1074  

1075  

Figure 6: Results from the sensitivity test to quantify the effect of each anomaly on the 1076  

Z, Zo, H and T responses. White areas are periods that are not sensitive to the anomalies. 1077  

Note that all plots have different colour scales. Grey thick lines at the top of each 1078  

pseudo-section show the location of the studied anomalies along the profile. 1079  

1080  

Figure 7: a) Phase tensor dimensionality analysis using Z responses. White-grey colours 1081  

are periods affected by signal associated with 1-D or 2-D structures. Rainbow colours 1082  

are periods affected by signal associated with 3-D anomalies. b) Results from the 1083  

dimensionality analysis using variations of the H responses (H-I), where I is the 1084  

identity. Dots are related to periods affected by signal associated with 1-D structures. 1085  

Lines are related to periods affected by signal associated with 2-D structures. Ellipses 1086  

are related to periods affected by signal associated with 3-D anomalies. White-grey 1087  

colours are periods that, because of the assumed errors, do not require the presence of 3-1088  

D anomalies. Rainbow colours are periods affected by signal that, even with the 1089  

assumed errors, requires the presence of 3-D anomalies. c) Induction arrows from T 1090  

responses, following the Parkinson criteria, used to characterize the dimensionality of 1091  

the area. White-grey colours are periods that, because of the assumed errors, do not 1092  

require the presence of 3-D anomalies. Rainbow colours are periods affected by signal 1093  

that, even with the assumed errors, requires the presence of 3-D anomalies. 1094  

1095  

Figure 8: a) Results of the different inversion process below the MT profile using 1096  

neighbouring reference sites located at different places along the MT profile. White 1097  

Page 42 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 43: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

dashed lines are plotted to better compare the results with the original synthetic model. 1098  

b) nRMS of each inversion processes for each site and for each type of data. Circles are 1099  

the nRMS of the corresponding site. 1100  

1101  

1102  

Figure 9: a) Results of the different inversion process below the MT profile. b) 1103  

Horizontal slices showing the results of the inversion processes at 650 m depth. c) 1104  

Horizontal slices showing the results of the inversion processes at 1550 m depth. a5), 1105  

b5) and c5) are the original synthetic model. White dashed lines are plotted to better 1106  

compare the results with the original synthetic model. 1107  

1108  

Figure 10: a) Results of the different inversion processes beneath the MT profile. b) 1109  

Horizontal slices showing the results of the inversion processes at 650 m depth. c) 1110  

Horizontal slices showing the results of the inversion processes at 1550 m depth. White 1111  

dashed lines are plotted to better compare the results with the original synthetic model. 1112  

1113  

Figure 11: nRMS of each inversion processes for each site and for each type of data. 1114  

Circles are the nRMS of the corresponding site. Excellent nRMS misfits of between 1.0 1115  

and 1.05 were achieved for all inversions using different types of responses. 1116  

1117  

Figure 12: Differences, using decadic logarithm electrical resistivity values, between 1118  

models obtained from the inversion process and the synthetic model. 1119  

Page 43 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 44: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

NS

De

pth

(m

)

1000

2000

3000

0

0

4000

-4000

0-6000 6000

No

rth

ing

(m

)

Easting (m)

0

4000

-4000

0-6000 6000

No

rth

ing

(m

)

Easting (m)

8000

-8000 -8000

8000

b) Horizontal slice (700 m depth) c) Horizontal slice (1500 m depth)

0 8000-8000 -4000 4000

Distance Profile (m)

Figure 1 (Campanyà et al. 2016)

400 ohm·m

200 ohm·m

(L3) 10 ohm·m

3 ohm·m

400 ohm·m

(L1)3 ohm·m

(L2)3 ohm·m

(H1)400 ohm·m

(L3)10 ohm·m

(H1)

(L1)

1 14 217 28

a) MT profile

Electrical resistivity [log10(ohm·m)]

0 0.5 1.5 2.51 2

Synthetic model

Page 44 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 45: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Figure 2 (Campanyà et al. 2016)

L1 anomaly

Difference / Errorfloor

a)

Hxx real Hxy real Hyx real Hyy real

Hxx imag Hxy imag Hyx imag Hyy imag

Zxx real Zxy real Zyx real Zyy real

Zxx imag Zxy imag Zyx imag Zyy imag

Pe

rio

ds

(lo

g10(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g10(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g10(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g10(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g10(s

))

-4-3

-2-10123

Sites1 26Sites1 26 Sites1 26 Sites1 26

b)

c) Tx real Tx imag Ty real Ty imag

3-3-4 -2 0 2 41-1

Page 45 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 46: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Figure 3 (Campanyà et al. 2016)

a) Zxx real Zxy real Zyx real Zyy real

Zxx imag Zxy imag Zyx imag Zyy imag

Pe

rio

ds

(lo

g1

0(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g1

0(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g1

0(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g1

0(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g1

0(s

))

-4-3

-2-10123

Sites1 26Sites1 26 Sites1 26 Sites1 26

b)

c)

Difference / Errorfloor

L2 anomaly

3-3-4 -2 0 2 41-1

Hxx real Hxy real Hyx real Hyy real

Hxx imag Hxy imag Hyx imag Hyy imag

Tx real Tx imag Ty real Ty imag

Page 46 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 47: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Figure 4 (Campanyà et al. 2016)

Difference / Errorfloor

a) Zxx real Zxy real Zyx real Zyy real

Zxx imag Zxy imag Zyx imag Zyy imag

Pe

rio

ds

(lo

g10(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g10(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g10(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g10(s

))

-4-3

-2-10123

Pe

rio

ds

(lo

g10(s

))

-4-3

-2-10123

Sites1 26Sites1 26 Sites1 26 Sites1 26

b)

c)

H1 anomaly

3-3-4 -2 0 2 41-1

Hxx real Hxy real Hyx real Hyy real

Hxx imag Hxy imag Hyx imag Hyy imag

Tx real Tx imag Ty real Ty imag

Page 47 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 48: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Figure 5 (Campanyà et al. 2016)

Difference / Errorfloor

3-3

Hxx real Hxy real Hyx real Hyy real

Hxx imag Hxy imag Hyx imag Hyy imag

Tx real Tx imag Ty real Ty imag

a) Zxx real Zxy real Zyx real Zyy real

Zxx imag Zxy imag Zyx imag Zyy imag

Periods

(log

10(s

))

-4-3

-2-10123

Periods

(log

10(s

))

-4-3

-2-10123

Periods

(log

10(s

))

-4-3

-2-10123

Periods

(log

10(s

))

-4-3

-2-10123

Periods

(log

10(s

))

-4-3

-2-10123

Sites1 26Sites1 26 Sites1 26 Sites1 26

b)

c)

-4 -2 0 2 41-1

L1, L2 and H1 anomalies

Page 48 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 49: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

42.51 63.51

71 4

Periods

(log

10(s

))

-4-3

-2-10123

Periods

(log

10(s

))

-4-3

-2-10123

HT

Periods

(log

10(s

))-4-3

-2-10123

Periods

(log

10(s

))

-4-3

-2-10123

Z_all

Zo

Sites1 26Sites1 26 Sites1 26 Sites1 26

L1 L2 H1 L1, L2, H1

Sites1 26Sites1 26 Sites1 26 Sites1 26

Sites1 26Sites1 26 Sites1 26 Sites1 26

Sites1 26Sites1 26 Sites1 26 Sites1 26

1891 1261

96481 1681 126 10011 50

126.51

1051 51 101 5.53

a)

b)

c)

d)

Figure 6 (Campanyà et al. 2016)

1001 50 1001 50

1161

Page 49 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 50: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Periods

(log

10(s

))-4

-3

-2

-1

0

1

2

3

a)

Figure 7 (Campanyà et al. 2016)

41 6 11 16 21 26

Phase Tensor (Z)

abs(Skew angle beta /error Skew angle beta)sites

Pe

rio

ds

(lo

g10(s

))

-4

-3

-2

-1

0

1

2

3

c)

41 6 11 16 21 26

Induction Arrows (T)

sites

Pe

rio

ds

(lo

g10(s

))

-4

-3

-2

-1

0

1

2

3

b)

41 6 11 16 21 26

variation of H

0 1 2 3

abs(Phi_min /error Phi_min)sites

0 1 2 3

0 1 2 3

abs( angle difference / error angle difference)

L2 L1 H1

L2 L1 H1

L2 L1 H1

real imag0.3 0.3

Page 50 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 51: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

0 0.5 1.5 2.51 2

Electrical resistivity [log10(ohm·m)]

RMS:1.09

RMS:1.98RMS:1.25

Depth

(km

)

1

2

3

0

0-4 4Distance Profile (km)

8-8

2 431

b1 b2 b3 b4

De

pth

(km

)

1

2

3

0

a1) Z+H, close L2 a2) Z+H, top L1S N S N

0-4 4Distance Profile (m)

RMS:1.35De

pth

(km

)

1

2

3

0

0-4 4Distance Profile (m)

Figure 8 (Campanyà et al. 2016)

a3) Z+H, top H1 a4) Z+H

1.5

1

0.5

RM

S

2

2.5

1.5

1

0.5

2

2.5

Z H

Sites1 28 Sites1 28Sites1 28 Sites1 28

b)

a)

Page 51 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 52: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Depth

(km

)

1

2

3

0

0-4 4Distance Profile (km)

0 0.5 1.5 2.51 2

Electrical resistivity [log10(ohm·m)]

b)

c)

0

4

-4Nort

hin

g (

km)

0-4 4Easting (km)

0-4 4Easting (km)

0-4 4Easting (km)

0-4 4Easting (km)

0-4 4Easting (km)

c1) Zo c3) Zo+H+T c2) Z c4) Z+H+T c5) S. Model

0

4

-4Nort

hin

g (

km)

0-4 4Easting (km)

0-4 4Easting (km)

0-4 4Easting (km)

0-4 4Easting (km)

0Easting (km)

-4 4

b1) Zo b3) Zo+H+T b2) Z b4) Z+H+T b5) S. Model

a)

Depth

(km

)

1

2

3

0

Depth

(km

)

1

2

3

0

0-4 4

Distance Profile (km)

a1) Zo a2) Z

a3) Zo+H+T a4) Z+H+T

a5) Synthetic Model

8-8

-8 8

S N S N

Figure 9 (Campanyà et al. 2016)

Page 52 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 53: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

b)

c)

a)

0 0.5 1.5 2.51 2

Electrical resistivity [log10(ohm·m)]

De

pth

(km

)

1

2

3

0

a1) Zo+T a2) Zo+H S N S N

0-4 4Distance Profile (m)

0

4

-4

0-4 4

No

rth

ing

(km

)

Easting (km)0-4 4

Easting (km)

b1) Zo+T b2) Zo+H

0-4 4Easting (km)

0-4 4Easting (km)

c1) Zo+T c2) Zo+H

De

pth

(km

)

1

2

3

0

De

pth

(km

)

1

2

3

0

0-4 4Distance Profile (m)

Figure 10 (Campanyà et al. 2016)

a3) Z+T a4) Z+H

a5) T+H

0-4 4Easting (km)

0-4 4Easting (km)

0-4 4Easting (km)

b3) Z+T b5) H+T b4) Z+H

0

4

-4

0-4 4

No

rth

ing

(km

)

Easting (km)0-4 4

Easting (km)0-4 4

Easting (km)

c3) Z+T c5) H+T c4) Z+H

Page 53 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 54: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Figure 11 (Campanyà et al. 2016)

1.5

1

0.5

RM

S

1.5

1

0.5

RM

S

1.5

1

0.5

RM

S

Sites1 28 Sites1 28 Sites1 28

Zo inversion Zo+T inversion Zo+H inversion

Z inversion Z+T inversion Z+H inversion

Zo+T+H inversion Z+T+H inversion T+H inversion

Zo Z T H

1.5

1

0.5

1.5

1

0.5

1.5

1

0.5

Page 54 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 55: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Figure 12 (Campanyà et al. 2016)

De

pth

(km

)

1

2

3

0

e) Zo+H f) Z+H

0-4 4Distance Profile (m)

De

pth

(km

)

1

2

3

0

De

pth

(km

)

1

2

3

0

0-4 4Distance Profile (m)

g) Zo+T+H h) Z+T+H

i) T+H

b) Z S N

De

pth

(km

)

1

2

3

0

c) Zo+T d) Z+T

a) Zo S N

De

pth

(km

)

1

2

3

0

0 0.5-0.5-0.75 0.75-0.25 0.25

Difference electrical resistivity [ohm·m]log10(Synth. mod) - log10(Inv. mod)

Page 55 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 56: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

Supplementary  material  S1  Campanyà  et  al.,  2016  GJI    Supplementary 1: Implementation of the H responses

Following the mathematical recipes (Egbert and Kelbert, 2012) and the practical

implementation of ModEM (Kelbert et al., 2014) we modified the code accordingly to allow

inverting for horizontal magnetic inter-station transfer functions. The modifications are

implemented mainly in the data functional module within the interface layer (Kelbert et al.,

2014, Fig. 1). In the data functional module (data responses routine), all kinds of transfer

functions are computed from the electric field solution (primary field) obtained from solving

the induction problem in 3-D. For the horizontal inter-station transfer functions we use the

magnetic field (dual field) derived from the electric field using Faraday's law. To simulate the

process of measuring EM components (e.g., inter-station horizontal magnetic transfers

functions) the magnetic field is interpolated at specific locations (local and neighbouring

reference sites). Using the interpolated horizontal magnetic field components (ℎ! and ℎ!) at

the local (𝒉!) and neighbouring reference site (𝒉!), we construct the inter-station horizontal

magnetic transfer functions (𝑯) equation S1.

𝒉! = 𝐇  !"𝒉!      (𝑆1)

For the inversion process in ModEM we use the None Linear Conjugate Gradients algorithm

(NLCG), which is based on a direct minimization of the penalty functional (Egbert and

Kelbert, 2012). All gradient-based optimization algorithm (including NLCG) require the

sensitivity of the data at each iteration. The sensitivity matrix in ModEM is computed by

multiplying several sub matrices (Egbert and Kelbert, 2012; Eqs. 14 and 15). The matrix that

depends on the inverted data type (e.g., H) is the so-called L matrix that represents the

linearized data functional. Mimicking the computation of the L matrix for the impedances

transfer functions (Z; Egbert and Kelbert, 2012; Eqs. 42 and 43) we computed the sensitivity

matrix for the inter-station horizontal magnetic transfer functions (𝑯). Hence, most of the

modifications required to compute the sensitivity matrix in case of inverting for H are

implemented in the none linearized data functional routine within the data functional module.

Page 56 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 57: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

a) Zo inversionlo

g10 f

req. (H

z)Zxy real Zxy imag Zyx real Zyx imag

Sites1 28

Sites1 28

Sites1 28

Sites1 28

b) Z inversion

log1

0 fr

eq. (

Hz)

Zxx real Zxy real Zyx real Zyy real

log1

0 fr

eq. (

Hz)

Zxx imag Zxy imag Zyx imag Zyy imag

Sites1 28

Sites1 28

Sites1 28

Sites1 28

Figure S2a,b (Campanyà et al. 2016)

(data - model_response) / error

Page 57 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 58: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

c) Zo+T inversionlo

g10 f

req. (H

z)Zxy real Zxy imag Zyx real Zyx imag

d) Z+T inversion

log1

0 fr

eq. (

Hz)

Zxx real Zxy real Zyx real Zyy real

log1

0 fr

eq. (

Hz)

Zxx imag Zxy imag Zyx imag Zyy imag

Figure S2c,d (Campanyà et al. 2016)

log1

0 fr

eq. (H

z)

Tx real Tx imag Ty real Ty imag

Sites1 28

Sites1 28

Sites1 28

Sites1 28

log1

0 fr

eq. (

Hz)

Tx real Tx imag Ty real Ty imag

Sites1 28

Sites1 28

Sites1 28

Sites1 28

(data - model_response) / error

Page 58 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 59: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

e) Zo+H inversionlo

g10 f

req. (H

z)

Zxy real Zxy imag Zyx real Zyx imag

log1

0 fr

eq. (

Hz)

Hxx real Hxy real Hyx real Hyy real

Figure S2e (Campanyà et al. 2016)

log1

0 fr

eq. (

Hz)

Hxx imag Hxy imag Hyx imag Hyy imag

Sites1 28

Sites1 28

Sites1 28

Sites1 28

(data - model_response) / error

Page 59 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 60: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

f) Z+H inversionlo

g10 f

req. (H

z)

Zxx real Zxy real Zyx real Zyy real

log1

0 fr

eq. (

Hz)

Zxx imag Zxy imag Zyx imag Zyy imag

Figure S2f (Campanyà et al. 2016)

log1

0 fr

eq. (

Hz)

Hxx real Hxy real Hyx real Hyy real

log1

0 fr

eq. (

Hz)

Hxx imag Hxy imag Hyx imag Hyy imag

Sites1 28

Sites1 28

Sites1 28

Sites1 28

(data - model_response) / error

Page 60 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 61: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

g) Zo+T+H inversionlo

g10 f

req. (H

z)

Zxy real Zxy imag Zyx real Zyx imag

log1

0 fr

eq. (

Hz)

Tx real Tx imag Ty real Ty imag

Figure S2g (Campanyà et al. 2016)

log1

0 fr

eq. (

Hz)

Hxx real Hxy real Hyx real Hyy real

log1

0 fr

eq. (

Hz)

Hxx imag Hxy imag Hyx imag Hyy imag

Sites1 28

Sites1 28

Sites1 28

Sites1 28

(data - model_response) / error

Page 61 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 62: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

h) Z+T+H inversionlo

g10 f

req. (H

z)

Zxx real Zxy real Zyx real Zyy real

log1

0 fr

eq. (

Hz)

Tx real Tx imag Ty real Ty imag

Figure S2h (Campanyà et al. 2016)

log1

0 fr

eq. (

Hz)

Hxx real Hxy real Hyx real Hyy real

log1

0 fr

eq. (

Hz)

Hxx imag Hxy imag Hyx imag Hyy imag

Sites1 28

Sites1 28

Sites1 28

Sites1 28

log1

0 fr

eq. (

Hz)

Zxx imag Zxy imag Zyx imag Zyy imag

(data - model_response) / error

Page 62 of 63Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Page 63: Theadvantages$of$complementingMT$profilesin$36D ...complete-mt-solutions.com/users/ajones/publications/185pp.pdf · 121" the case of presence of current channelling caused by thin-long

i) T+H inversionlo

g10 f

req. (H

z)

Tx real Tx imag Ty real Ty imag

Figure S2i (Campanyà et al. 2016)

log1

0 fr

eq. (

Hz)

Hxx real Hxy real Hyx real Hyy real

log1

0 fr

eq. (

Hz)

Hxx imag Hxy imag Hyx imag Hyy imag

Sites1 28

Sites1 28

Sites1 28

Sites1 28

(data - model_response) / error

Page 63 of 63 Geophysical Journal International

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960