Geophysical study of the crust in the central Red Sea

Geophysical study of the crust in the

central Red Sea

A thesis submitted to The University of Manchester for

the degree of Doctor of Philosophy in the Faculty of

Science & Engineering

2019

Wen Shi

School of Natural Sciences

Department of Earth and Environmental Sciences

2

Table of Contents

Table of Contents ........................................................................................................... 2

List of Figures ................................................................................................................. 6

Abstract ........................................................................................................................ 11

Declaration ................................................................................................................... 12

Copyright Statement .................................................................................................... 13

Acknowledgements ...................................................................................................... 14

The Author ................................................................................................................... 15

List of publications ....................................................................................................... 16

1. Introduction ............................................................................................................. 18

1.1 Research background and motivations .............................................................. 18

1.2 Aim of thesis ....................................................................................................... 19

1.3 Thesis content and layout .................................................................................. 19

1.4 References .......................................................................................................... 23

2. Review of literature on the Red Sea and how it prompts the present work .......... 27

2.1 Study area ........................................................................................................... 27

2.2 Geological and tectonic setting .......................................................................... 29

2.2.1 Seafloor spreading and continental rifting in the Red Sea .......................... 30

2.2.2 Seismic tomographic studies encompassing the Red Sea ........................... 34

2.3 How the previous studies prompt the present work ......................................... 36

2.4 References .......................................................................................................... 37

3. Data and methods .................................................................................................... 48

3.1 Multichannel seismic reflection data ................................................................. 48

3.2 Magnetic anomalies ........................................................................................... 51

3.2.1 Shipboard magnetic data ............................................................................ 51

3.2.2 Aeromagnetic data ...................................................................................... 54

3.2.3 2D Werner deconvolution ........................................................................... 56

3.2.4 Statistical analysis of the Werner solutions ................................................ 62

3.2.5 Reduction to the pole (RTP) ........................................................................ 62

3.3 Gravity anomalies ............................................................................................... 64

3.3.1 Free-air gravity data .................................................................................... 64

3

3.3.2 Bouguer gravity anomalies .......................................................................... 65

3.3.3 Mantle Bouguer anomalies (MBAs)............................................................. 67

3.3.4 Bouguer slab formula .................................................................................. 67

3.3.5 2D gravity forward modelling ...................................................................... 69

3.4 Isostatic loading corrections .............................................................................. 70

3.5 Bathymetry data ................................................................................................. 71

3.5.1 Smith and Sandwell (1997) global topography dataset (Version 18.1)....... 71

3.5.2 Multibeam sonar data ................................................................................. 72

3.6 References .......................................................................................................... 72

4. Paper 1: Oceanic-like axial crustal high in the central Red Sea ............................... 79

4.1 Introduction ........................................................................................................ 81

4.2 Tectonic setting .................................................................................................. 85

4.2.1 Continental rifting and seafloor spreading in the northern and southern

Red Sea ................................................................................................................. 85

4.2.2 Seismic tomographic studies encompassing the Red Sea ........................... 86

4.3 Data and methods .............................................................................................. 87

4.3.1 Seismic reflection ........................................................................................ 87

4.3.2 Magnetic anomalies .................................................................................... 88

4.3.3 Bathymetry data .......................................................................................... 92

4.3.4 Isostatic loading corrections ........................................................................ 92

4.3.5 Bouguer gravity anomalies .......................................................................... 93

4.4 Results ................................................................................................................ 94

4.4.1 Character of basement and seabed derived from seismic reflection profiles

.............................................................................................................................. 94

4.4.2 Oceanic-like axial crustal highs in isostatically corrected basement depths

.............................................................................................................................. 96

4.4.3 Correlation between Bouguer gravity anomalies and basement reflection

depths ................................................................................................................... 99

4.5 Discussion ......................................................................................................... 105

4.5.1 How does the Red Sea axial high compare with axial highs at other

spreading centres near hotspots? ...................................................................... 105

4.5.2 How thick is crust beneath the axial high and how does it relate to mantle

tomographic results? .......................................................................................... 107

4

4.5.3 What are its implications? ......................................................................... 111

4.6 Conclusions ....................................................................................................... 111

4.7 Acknowledgments ............................................................................................ 112

4.8 References ........................................................................................................ 113

5. Paper 2: Central Red Sea basement depths from Werner deconvolution of

aeromagnetic data ..................................................................................................... 128

5.1 Introduction ...................................................................................................... 130

5.2 Geological setting ............................................................................................. 132

5.3 Data and methods ............................................................................................ 135

5.3.1 Multichannel seismic reflection ................................................................ 135

5.3.2 Magnetic anomalies .................................................................................. 135

5.3.3 Bouguer gravity anomalies ........................................................................ 142

5.3.4 Bathymetry data ........................................................................................ 143

5.4 Results .............................................................................................................. 143

5.4.1 Basement depth derived from aeromagnetic data ................................... 143

5.4.2 Correlation between Bouguer gravity anomalies and magnetic basement

elevations ............................................................................................................ 145

5.4.3 Number of magnetic source solutions ...................................................... 148

5.5 Discussion ......................................................................................................... 149

5.5.1 How well do the magnetic basement topography and source solution

numbers correspond with other data? .............................................................. 150

5.5.2 The potential utility of magnetic source depth determination in the Red

Sea ...................................................................................................................... 154

5.6 Conclusions ....................................................................................................... 154


5.8 References ........................................................................................................ 156

6. Paper 3: Oceanic basement roughness in the central Red Sea ............................. 164

6.1 Introduction ...................................................................................................... 166

6.2 Tectonic setting ................................................................................................ 167

6.3 Data and methods ............................................................................................ 169

6.3.1 Multichannel seismic reflection ................................................................ 169

6.3.2 Gravity anomalies ...................................................................................... 172

6.3.3 Bathymetry data ........................................................................................ 179

5

6.4 Results .............................................................................................................. 179

6.4.1 Basement roughness along across-ridge seismic profiles ......................... 179

6.4.2 Basement roughness along the ridge-parallel gravity profiles ................. 182

6.5 Discussion ......................................................................................................... 183

6.5.1 Reduced-to-pole magnetic anomalies ....................................................... 184

6.5.2 Along-axis gradients in mantle Bouguer anomalies .................................. 185

6.6 Conclusion ........................................................................................................ 187


6.8 References ........................................................................................................ 188

7. Synthesis ................................................................................................................ 198

7.1 Crust types in the central Red Sea ................................................................... 198

7.1.1 Classifications ............................................................................................ 198

7.1.2 Character of the axial high......................................................................... 203

7.1.3 Distinct domains from north to south ....................................................... 204

7.2 Moho depth and crustal thickness ................................................................... 204

7.3 Uncertainty in the seismically derived basement depth ................................. 207


7.5 References ........................................................................................................ 208

8. Conclusion and future work ................................................................................... 213

8.1 Conclusions ....................................................................................................... 213

8.2 Future work suggestions .................................................................................. 215

8.2.1 Future work in the Red Sea ....................................................................... 215

8.2.2 Apply Werner deconvolution in other areas ............................................. 216

8.3 References ........................................................................................................ 217

Appendices ................................................................................................................. 221

Appendix 1 .............................................................................................................. 221

Appendix 2 .............................................................................................................. 227

References .............................................................................................................. 235

Word Count: 54,698 words

6

List of Figures

Figure 2.1 Map of Red Sea bathymetry showing the location of study area……………28

Figure 2.2 Simplified stratigraphic sections of the Red Sea and Gulf of Suez………….29

Figure 2.3 Seafloor spreading magnetic anomalies in the southern Red Sea……………31

Figure 2.4 Shear wave velocity map at a depth of 150 km………………………..……….35

Figure 3.1 Map of free-air gravity anomalies showing the locations of multichannel

seismic reflection profiles……………………………………………………………………………………….48

Figure 3.2 Depths derived from the seismic reflection profiles………………………………49

Figure 3.3 Confidence map showing the ability to image the basement reflection....51

Figure 3.4 (a): Tracks of shipboard magnetic surveys. (b): Residual magnetic

anomalies……………………………………………………………………………………………………………….52

Figure 3.5 (a): Extents of shipboard magnetic lines contributing to the evaluation of

seismic profiles using Werner source depths. (b): A sketch showing how the source

depths and apparent susceptibilities were projected onto the seismic profiles……53

Figure 3.6 (a): Locations of aeromagnetic survey flight lines. (b): Residual

aeromagnetic anomalies. (c): Locations of the long survey lines used in Chapter 5. (d):

Residual aeromagnetic anomalies…………………………………………………………………….55

Figure 3.7 Parameters of a vertical thin dike resolved using 2D Werner

deconvolution………………………………………………………………………………………………58

Figure 3.8 Aeromagnetic anomalies reduced to pole…………………………………………….63

Figure 3.9 (a): Free-air gravity anomalies and locations of multichannel seismic

reflection profiles. (b): Locations of gravity profiles G1-32………………………………………65

Figure 3.10 Bouguer gravity anomalies …………….………….……………………………..………..66

Figure 3.11 Bathymetry of the central Red Sea……………………………………..…………………71

7

Figure 4.1 Bathymetry of the Red Sea showing the location of study area……………..81

Figure 4.2 Free-air gravity anomalies and locations of multichannel seismic reflection

profiles……………………………………………………………………………………………………………………83

Figure 4.3 (a): Tracks of shipboard magnetic surveys. (b): Extents of magnetic lines

contributing to the seismic profiles of Werner source depths. (c): Residual magnetic

anomalies. (d): Bouguer gravity anomalies……………………………………………………………..89

Figure 4.4 Depths derived from the seismic reflection profiles and Werner

deconvolution of marine magnetic data………………………………………………………………….95

Figure 4.5 (a): Basement depths along the Red Sea seismic lines corrected for

evaporite and other sediment loading. (b): Basement depths around the Reykjanes

Ridge. (c): Locations of Reykjanes Ridge profiles. (d): Red Sea crustal deepening with

distance from the ridge-axis……………………………………………………………………………………99

Figure 4.6 Graphs showing correlation between basement reflection depths and

Bouguer gravity anomalies…………………………………………………………………………………….100

Figure 4.7 Apparent density contrasts deduced from Bouguer-basement depth

gradients…………………………………………………………………………………………………………101

Figure 4.8 Simulation using basement depth profile 21 illustrating how apparent

density contrasts inferred using the gravity slab formula are reduced by upward

continuation…………………………………………………………………………………………………………102

Figure 4.9 (a): Density structure along line PIII. (b): Free-air gravity anomaly calculated

from (a) compared with observations from the Sandwell et al. (2014) gravity field

(version 23.1). (c): Total mass anomaly per unit area along PIII……………………………..104

Figure 4.10 (a): Sodium oxide contents of axial lavas. (b): Seismically determined

estimates of crustal thickness versus average Na8.0 ……………………………………………….107

Figure 4.11 Examples of locally elevated topography at ridges located near mantle

hotspots…………………………………………………………………………………………………………110

Figure 5.1 Bathymetry of the Red Sea showing the location of study area……………131

8

Figure 5.2 Seafloor spreading magnetic anomalies in the southern Red Sea………134

Figure 5.3 (a): Locations of aeromagnetic survey flight lines and multichannel seismic

reflection profiles. (b): Residual aeromagnetic anomalies. (c): Locations of the survey

lines used for the Werner deconvolution. (d): Residual aeromagnetic anomalies along

the long survey lines plotted as profiles………………………………………………………………..136

Figure 5.4 Depths derived from the seismic reflection profiles and Werner

deconvolution of the aeromagnetic data………………………………………………………………140

Figure 5.5 Histograms and cumulative distribution functions (CDF) of differences

between the Werner source depths and the seismically derived basement depths

(ΔZ)……………………………………………………………………………………………………………………….142

Figure 5.6 Basement topography map derived from the aeromagnetic data Werner

source solutions……………………………………………………………………………………………………144

Figure 5.7 Magnetic basement depths along the spreading axis…………………………..145

Figure 5.8 (a): Bouguer gravity anomalies. (b): Map of aeromagnetic-derived

basement depths with shading from the Bouguer gravity grid………………………………146

Figure 5.9 (a): Correlation between Bouguer gravity anomalies and aeromagnetic-

derived basement depths within 60 km of the axis. (b): Differences between

aeromagnetic-derived basement depths and those predicted from Bouguer gravity

anomalies using the regression in (a).………………………………………………………………..147

Figure 5.10 The numbers of magnetic sources……………………………………………………..149

Figure 5.11 (a): Map locating the fracture zones over the basement topography map.

(b): Map locating the interpreted extent of evaporite and other sedimentary cover of

overlain on the magnetic basement depths………………………………………………………….151

Figure 5.12 (a): Map locating the fracture zones over the map of Figure 5.10. (b):

Comparison of the data in (a) with the distribution of evaporites and other

sediments……………………………………………………………………………………………………………..153

Figure 6.1 Bathymetry of the Red Sea showing the location of study area…………..167

9


reflection. (b): Locations of gravity profiles G1 -32 and seismic reflection

profiles…………………………………………………………………………………………………………………170

Figure 6.3 Depths derived from the seismic reflection profiles…………………………….171

Figure 6.4 Histograms of differences between the SSv23 and free-air anomaly data

collected on (a): RRS Shackleton and (b): RV Robert Conrad…………………………………173

Figure 6.5 Differences between the SSv23 gravity field and the shipboard gravity data

after the shipboard data were fi ltered with a 4 km along -track median

filter………………………………………………………………………………………………………………………174

Figure 6.6 Forward modelling of gravity profile G21…………………………………………… 178

Figure 6.7 (a): Basement depths along the Red Sea seismic lines corrected for the

isostatic loading of evaporites and other sediments. (b): Regional crustal subsidence

trend. (c): Residual basement reliefs. (d): The relationship between the mean

basement roughness value and filter width…………………………………………………………..180

Figure 6.8 The relationship between the basement roughness and spreading rate for

ultraslow and slow spreading ridges……………………………………………………………………181

Figure 6.9 Basement roughness values computed with a modified Bouguer slab

formula along axis-parallel gravity profiles G1-32………………………………………………….183

Figure 6.10 Aeromagnetic anomalies and reduced to pole…………………………………..185

Figure 6.11 Comparison of along-axis gradients in mantle Bouguer anomalies in the

central Red Sea with those at other mid-ocean ridges………………………………………….186

Figure 7.1 (a): Geological map of the central Red Sea interpreted in this study. (b): A

comparison of crust type classification in (a) with that of Izzeldin (1987)……………….199

Figure 7.2 Interpretations of crust types along seismic lines 17 and 21…………………200

Figure 7.3 Comparisons of crust type classification in Figure 7.1a with (a): Free-air

gravity anomalies, (b): Bouguer gravity anomalies, and (c): RTP aeromagnetic

10

anomalies (d): A comparison of Bouguer gravity anomalies in (b) with RTP

aeromagnetic anomalies in (c)……………………………………………….……………………………..202


reflection profiles, seismic refraction profile SO53-PIII, and profile A-B. (b): A

composite profile produced by projecting the seafloor and basement reflections along

profile 25 on onto profile PIII. (c): Graph showing basement reflection along profile

25 and Moho along profiles PIII and A-B………………………………….……………………….…...205

Figure 7.5 A sketch showing across-ridge crustal structures in the central Red Sea…207

Figure A1.1 Depths derived from the seismic reflection profiles and Werner

deconvolution of marine magnetic data……………………………………………………………… 226

11

Abstract

The Red Sea is an important example a continental rift proceeding to an

oceanic basin, but whether the crust in the central Red Sea is continental or oceanic

has been controversial. Contributing to this debate, the basement geometry and

roughness are assessed using seismic reflection and potential field data.

An axial crustal high with a width of 70-100 km and a height of 0.8-1.6 km is

found after correcting the seismically derived basement depths for evaporite and

other sediment isostatic loading. Basement axial highs are commonly found at mid-

ocean ridges affected by hotspots, where enhanced mantle melting results in

thickened crust. Therefore, it is suggested that the central Red Sea is underlain by

oceanic crust typical of a mid-ocean ridge near to a mantle hotspot, like the Reykjanes

Ridge. Bouguer gravity anomalies are found strongly correlated with basement

depths from seismic reflection data. The low average basement densities deduced

from Bouguer-basement depth gradients imply thickened crust and/or low mantle

densities beneath the ridge axis. Normal axial crust thickness predicted from

fractionation-corrected sodium contents (Na8.0) implies that the earliest seafloor

spreading in the central Red Sea began with thinner than average crust.

To further assess the basement geometry, the inverse method of Werner

deconvolution is improved and used to invert aeromagnetic anomalies for magnetic

basement depths. The improved Werner deconvolution effectively maps out the axial

plateau and valleys in the crustal basement. The results confirm that the basement

topography in the region away from the seismic lines also has an axial plateau within

~60 km of the axis. Magnetic basement depth near the spreading axis generally co-

varies with Bouguer gravity anomalies. Valleys in the derived depths coincide with

fracture zones interpreted previously from shipboard gravity, aeromagnetic,

bathymetric and seismic reflection data. Those valleys also correspond with areas

where the evaporites have extended into the axial valley floor, as suggested by earlier

researchers.

Basement roughness values are computed in profiles both across and parallel

to the axis. The values from axis-crossing seismic data are ~230 m, similar to those

observed at other ultraslow and slow spreading ridges. The roughness values derived

from axis-parallel profiles of the gravity field (200-550 m) are comparable with those

of the Mid-Atlantic Ridge where it has similar along-axis segmentation. Although

these basement roughness values by themselves do not exclude an extended

continental crust interpretation in the central Red Sea, they are supportive of an

oceanic crustal interpretation when considered along with other evidence.

Finally, our new produced geological map suggests that seafloor spreading and

continental rifting in the central Red Sea have been symmetric.

12

Declaration

I declare that no portion of the work referred to in the thesis has been

submitted in support of an application for another degree or qualification,

of this, or any other university or other institute of learning.

Signed: Date:

13

Copyright Statement

i. The author of this thesis (including any appendices and/or schedules to this thesis)

owns certain copyright or related rights in it (the “Copyright”) and s/he has given The

University of Manchester certain rights to use such Copyright, including for

administrative purposes.

ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic

copy, may be made only in accordance with the Copyright, Designs and Patents Act

1988 (as amended) and regulations issued under it or, where appropriate, in

accordance with licensing agreements which the University has from time to time. This

page must form part of any such copies made.

iii. The ownership of certain Copyright, patents, designs, trademarks and other

intellectual property (the “Intellectual Property”) and any reproductions of copyright

works in the thesis, for example graphs and tables (“Reproductions”), which may be

described in this thesis, may not be owned by the author and may be owned by third

parties. Such Intellectual Property and Reproductions cannot and must not be made

available for use without the prior written permission of the owner(s) of the relevant

Intellectual Property and/or Reproductions.

iv. Further information on the conditions under which disclosure, publication and

commercialisation of this thesis, the Copyright and any Intellectual Property and/or

Reproductions described in it may take place is available in the University IP Policy

(see

http://www.campus.manchester.ac.uk/medialibrary/policies/intellectualproperty.pd

f), in any relevant Thesis restriction declarations deposited in the University Library,

The University Library’s regulations:

(see http://www.manchester.ac.uk/library/aboutus/regulations) and in The

University’s policy on presentation of Theses.

14

Acknowledgements

Foremost, I would like to thank my supervisor Neil Mitchell for his guidance, support,

patience, encouragement, and great enthusiasm over the last four years.

I would like to thank my co-supervisor Lara Kalnins at the University of Edinburgh for

her helpful comments, support, and enthusiasm during my PhD research.

I would like to thank A.Y. Izzeldin in Sudan and Ian C.F. Stewart in Australia for their

constructive comments and support.

I would like to thank British Geological Survey (BGS) for providing us the software of

GRAVMAG, which is used to check some results computed using my own program.

I would like to thank the staff in School of Earth and Environmental Sciences for their

help throughout the course of this PhD.

I would like to thank my parents for their support and encouragement.

I would like to thank Lin Ma and Jianpeng Wang for their help throughout the four

years.

I would like to thank all my friends for their support and company throughout the four

years.

15

The Author

Wen Shi graduated from the China University of Mining and Technology with

a bachelor’s degree in Geophysics in 2009. He graduated from the Graduate

University of Chinese Academy of Sciences with a degree of Master of Natural Science

in Solid Earth Physics in 2012. During his master’s course, he worked with his master's

supervisor in the Institute of Geology and Geophysics, Chinese Academy of Sciences

from July 2010 to July 2012. Then he worked as a research assistant in the Shanghai

Institute of Microsystem and Information Technology, Chinese Academy of Sciences

from July 2012 to July 2014.

Wen Shi started PhD studies in Basin Studies and Petroleum Geoscience,

School of Earth and Environmental Sciences in the University of Manchester since

January 2015. His research interest is focusing on the structure of the central Red Sea

from potential field and seismic data.

16

List of publications

1. Shi, W., N. C. Mitchell, L. M. Kalnins, and A. Y. Izzeldin, 2018, Oceanic-like axial

crustal high in the central Red Sea: Tectonophysics, v. 747-748, p. 327-342.

2. Shi, W., N. C. Mitchell, L. M. Kalnins, and I. AY, 2017, Assessing the Nature of Crust

in the Central Red Sea Using Potential Fields and Seismic Reflection Data: AGU Fall

Meeting Abstracts.

3. Shi, W., N. Mitchell, and L. Kalnins, 2017, Assessing the nature of crust in the central

Red Sea using potential field methods: EGU General Assembly Conference Abstracts,

p. 542.

4. Hui Dong, Longqing Qiu, Wen Shi, Baolin Chang, Yang Qiu et al., Ultra-low field

magnetic resonance imaging detection with gradient tensor compensation in urban

unshielded environment. Appl. Phys. Lett., 2013, 102, 102602.

5. Xu P F, Shi W, Ling S Q, et al., Mapping spherically weathered “Boulders” using 2D

microtremor profiling method: a case study along subway line 7 in Shenzhen. Chinese

J. Geophysics. (in Chinese), 2012, 55(6) : 2120-2128.

17

Chapter 1.

Introduction

18

1. Introduction

1.1 Research background and motivations

Wilson (1966) noted that the Earth has been undergoing a cyclical process of

supercontinent assemblage and breakup. In these cycles, continents have repeatedly

assembled into a supercontinent and then broken apart again (e.g., Rogers and

Santosh, 2004; Santosh et al., 2009). At present, we are in a stage when continental

blocks are scattered on the surface of the Earth (Bonatti et al., 2015; Rogers and

Santosh, 2004). Geological reconstructions suggested that the most recently

assembled supercontinent (Pangea) existed roughly between ~250 Ma and ~160 Ma

(e.g., Anderson, 1982; Condie, 1998). A key step in this Wilson's cycle is ocean

formation, during which a continent rifts and proceeds to seafloor spreading.

Understanding the processes that occur during the transition from continental

rifting to seafloor spreading is important in exploring how our planet works and has

been a big challenge of Plate Tectonics for decades (e.g., Bonatti et al., 2015; Taylor et

al., 1995). On the modern Earth, there are few young ocean basins where this

transition can be observed. Those basins are in comparable ocean basin stages of

transition from continental rifting to oceanic spreading, but differ in spreading rate

and opening direction. The Woodlark Basin is small, opening relatively quickly (~60

mm yr-1) and in a complicated tectonic setting (e.g., its spreading center is highly offset

and the basin is still evolving rapidly after the Ontong Java collision with the West

Melanesian Trench) (Martinez et al., 1999; Weissel et al., 1982). The Gulf of California

rift is opening highly obliquely (e.g., Atwater and Stock, 1998; Lonsdale, 1989;

Withjack and Jamison, 1986) with a spreading rate of ~45–47 mm yr-1 (Plattner et al.,

2007). In contrast, the Red Sea is opening slowly (~10-~16 mm yr-1) and nearly

orthogonally (e.g., Chu and Gordon, 1998), so it provides an important example of

transition from nearly orthogonal slow continental rifting to seafloor spreading.

However, how far the central Red Sea is through this transition to full seafloor

spreading has been controversial (e.g., Augustin et al., 2016; Bonatti, 1985; Davies and

Tramontini, 1970; Izzeldin, 1982, 1987; Ligi et al., 2012; Mitchell and Park, 2014; Shi

19

et al., 2018; Sultan et al., 1992, 1993; Tramontini and Davies, 1969). Contributing to

this debate, the study assessed the crustal type (whether the crust is continental or

oceanic) in the central Red Sea by evaluating the basement geometry and roughness.

Moreover, the nature of transitions from oceanic to continental crusts were also

addressed in this study.

1.2 Aim of thesis

The work presented in this thesis is based primarily on the analysis and

interpretation of multichannel seismic reflection, satellite free-air gravity, shipboard

magnetic, aeromagnetic, Smith and Sandwell (1997) global topography and

bathymetry, and multibeam sonar data. Its overall objective is to address whether the

crust in the central Red Sea is continental or oceanic, and where those types extend,

with the aim of improving the understanding of how the transition from continental

rifting to seafloor spreading is evolving along the Red Sea rift. For example, Bonatti

(1985) and Ligi et al. (2011,2012) suggested that the inception of seafloor spreading

is not synchronous along the central Red Sea since the stretched continental crusts

still exist in the inter-trough zones and separate the oceanic-like ‘deeps’, while this

study suggests that this oceanic spreading inception is synchronous in the central Red

Sea along the spreading ridge because the entire axial zone including inter-trough

zones is underlain by oceanic crust. The edges of that oceanic crust and thus

transitions to extended continental crust are located at distances of ~60 km on both

sides of the ridge axis.

1.3 Thesis content and layout

This thesis is presented in Journal Format. It is divided into seven chapters.

Three research papers containing original research are presented in Chapters 4, 5, and

6. The descriptions of each chapter, author contributions, and publication status are

listed as follows.

Chapter 1

This chapter presents a general introduction to research background and

motivations, aim of thesis, and thesis content and layout.

20

Chapter 2

This chapter presents the corresponding literature review on the Red Sea and

how they prompt the present work.

Chapter 3

This chapter summarizes the data and methods applied in this study. The basic

and derived data include multichannel seismic reflection data, shipboard magnetic

data, aeromagnetic anomalies, reduction to the pole magnetic anomalies, satellite

free-air gravity anomalies, Bouguer gravity anomalies, Smith and Sandwell (1997)

global topography and bathymetry, and multibeam sonar data. The methods include

Werner deconvolution, isostatic loading corrections, Bouguer slab formula, and

gravity forward modeling.

Chapter 4

This chapter presents the first research paper: “Oceanic-like axial crustal high

in the central Red Sea”.

In this paper, we found an axial crustal high after correcting the seismically

derived basement depths for evaporite and other sediment isostatic loading, and

interpreted it as suggesting the central Red Sea is underlain by oceanic crust typical of

a mid-ocean ridge near to a mantle hotspot, like the Reykjanes Ridge. We then

discussed what our results imply about the evolution of the Red Sea rift in this area

and broader implications.

Author contributions: as first and corresponding author, I processed and

interpreted the data and produced all the figures. All the codes used in this study

were written by me. Codes were written to work in Matlab. Figures and calculations

were also performed with Generic Mapping Tools (Wessel et al., 2013). I produced

the first draft manuscript of this paper and modified it under the guidance of my main

supervisor Neil Mitchell. Co-authors Neil Mitchell, Lara Kalnins, and A.Y. Izzeldin

provided helpful comments on the draft manuscript.

Publication status: this chapter is published in the journal Tectonophysics.

21

Chapter 5

This chapter presents the second research paper: “Central Red Sea basement

depths from Werner deconvolution of aeromagnetic data”.

In this paper, we improved the inverse method of Werner deconvolution and

used it to invert aeromagnetic anomalies for magnetic basement depths in the central

Red Sea. The results confirmed that the basement topography in the region away

from the seismic lines also has an axial plateau within ~60 km of the axis. Moreover,

this exercise illustrated the potential feasibility and applicability of magnetic source

depth determination in the central Red Sea and elsewhere where there is a

magnetized basement overlain by sediments lacking magnetization.


interpreted the data and produced all the figures. All the codes used in this study

were written by me. I produced the first draft manuscript of this paper and modified

it under the guidance of my main supervisor Neil Mitchell. Co-authors Neil Mitchell,

Lara Kalnins, Ian C.F. Stewart, and A.Y. Izzeldin provided helpful comments on the draft

manuscript.

Publication status: this chapter is ready for submission to the journal Marine

Geophysical Researches.

Chapter 6

This chapter presents the third research paper: “Oceanic basement roughness

in the central Red Sea”.

In this paper, we computed basement roughness values in the central Red Sea

along lines both parallel to the axis and across it in order to assess if basement

roughness is compatible with those of other mid-ocean ridges. These roughness

values in the central Red Sea are similar to the values observed at other ultraslow and

slow spreading ridges. A change in those roughness values roughly mid-way between

the coast and the axis may mark the transition in crustal type from stretched

continental to predominantly oceanic. Although these basement roughness values by

22

themselves do not exclude an extended continental crust interpretation in the central

Red Sea, they are supportive of an oceanic crustal interpretation when considered

along with other evidence.


interpreted the data and produced all the figures. All the codes except those for the

‘reduction-to-the pole’ technique were written by me. I produced the first draft

manuscript of this paper and modified it under the guidance of my main supervisor

Neil Mitchell. Co- supervisor Lara Kalnins provided helpful comments on the draft

manuscript.

Publication status: this chapter is in preparation to be submitted to the journal

Marine Geophysical Researches.

Chapter 7

This chapter synthesizes the interpretations presented in Chapters 4, 5 and 6.

The spatial geometry of crustal type classifications and variations in crustal thickness

are discussed.

Chapter 8

This chapter summarizes the results of the earlier chapters, tries to take a

broader view in discussing them and makes suggestions for future work.

The results Chapters 4, 5, and 6 are interconnected here. To address the thesis

objective of whether the crust in the central Red Sea is continental or oceanic, the

basement geometry was assessed in Chapters 4 and 5, while the basement roughness

was assessed in Chapter 6. The results in Chapter 4 are mainly based on seismic

reflection data, whilst those in Chapter 5 are mainly based on aeromagnetic data. The

axial highs found in Chapters 4 and 5 from different data can confirm each other. The

basement roughness values computed in Chapter 6 are compatible with those of other

mid-ocean ridges, supporting the oceanic crustal interpretation proposed in Chapters

4 and 5.

23

1.4 References

Anderson, D. L., 1982, Hotspots, polar wander, Mesozoic convection and the geoid:

Nature, v. 297, p. 391.

Atwater, T., and J. Stock, 1998, Pacific-North America plate tectonics of the Neogene

southwestern United States: an update: International Geology Review, v. 40,

p. 375-402.

Augustin, N., F. M. van der Zwan, C. W. Devey, M. Ligi, T. Kwasnitschka, P. Feldens, R.

A. Bantan, and A. S. Basaham, 2016, Geomorphology of the central Red Sea

Rift: Determining spreading processes: Geomorphology, v. 274, p. 162-179.

Bonatti, E., 1985, Punctiform initiation of seafloor spreading in the Red Sea during

transition from a continental to an oceanic rift: Nature, v. 316, p. 33-37.

Bonatti, E., A. Cipriani, and L. Lupi, 2015, The Red Sea: Birth of an Ocean, The Red Sea,

Springer, p. 29-44.

Chu, D., and R. G. Gordon, 1998, Current plate motions across the Red Sea:

Geophysical Journal International, v. 135, p. 313-328.

Condie, K. C., 1998, Episodic continental growth and supercontinents: a mantle

avalanche connection?: Earth and Planetary Science Letters, v. 163, p. 97-108.

Davies, D., and C. Tramontini, 1970, The deep structure of the Red Sea: Philosophical

Transactions of the Royal Society of London A: Mathematical, Physical and

Engineering Sciences, v. 267, p. 181-189.

Izzeldin, A., 1982, On the structure and evolution of the Red Sea: PhD Diss. Univ.

Strasbourg.

Izzeldin, A., 1987, Seismic, gravity and magnetic surveys in the central part of the Red

Sea: their interpretation and implications for the structure and evolution of the

Red Sea: Tectonophysics, v. 143, p. 269-306.

Ligi, M., E. Bonatti, G. Bortoluzzi, A. Cipriani, L. Cocchi, F. Caratori Tontini, E. Carminati,

L. Ottolini, and A. Schettino, 2012, Birth of an ocean in the Red Sea: initial pangs:

Geochemistry, Geophysics, Geosystems, v. 13 (Paper Q08009,

doi:10.1029/2012GC004155).

Ligi, M., E. Bonatti, F. C. Tontini, A. Cipriani, L. Cocchi, A. Schettino, G. Bortoluzzi, V.

Ferrante, S. Khalil, and N. C. Mitchell, 2011, Initial burst of oceanic crust

24

accretion in the Red Sea due to edge-driven mantle convection: Geology, v. 39,

p. 1019-1022.

Lonsdale, P., 1989, Geology and tectonic history of the Gulf of California: The eastern

Pacific Ocean and Hawaii: Boulder, Colorado, Geological Society of America,

Geology of North America, v. N, p. 499-521.

Martinez, F., B. Taylor, and A. M. Goodliffe, 1999, Contrasting styles of seafloor

spreading in the Woodlark Basin: Indications of rift‐induced secondary mantle

convection: Journal of Geophysical Research: Solid Earth, v. 104, p. 12909-

12926.

Mitchell, N. C., and Y. Park, 2014, Nature of crust in the central Red Sea:

Tectonophysics, v. 628, p. 123-139.

Plattner, C., R. Malservisi, T. H. Dixon, P. LaFemina, G. Sella, J. Fletcher, and F. Suarez-

Vidal, 2007, New constraints on relative motion between the Pacific plate and

Baja California microplate (Mexico) from GPS measurements: Geophysical

Journal International, v. 170, p. 1373-1380.

Rogers, J. J., and M. Santosh, 2004, Continents and supercontinents, Oxford University

Press, 289 p.

Sandwell, D. T., R. D. Muller, W. H. F. Smith, E. Garcia, and R. Francis, 2014, New global

marine gravity model from CryoSat-2 and Jason-1 reveals buried tectonic

structure: Science, v. 346, p. 65-67.

Santosh, M., S. Maruyama, and S. Yamamoto, 2009, The making and breaking of

supercontinents: some speculations based on superplumes, super

downwelling and the role of tectosphere: Gondwana Research, v. 15, p. 324-

341.

Shi, W., N. C. Mitchell, L. M. Kalnins, and A. Y. Izzeldin, 2018, Oceanic-like axial crustal

high in the central Red Sea: Tectonophysics, v. 747-748, p. 327-342.

Smith, W. H. F., and Sandwell, D. T., 1997, Global sea floor topography from satellite

altimetry and ship soundings: Science, v. 277, p. 1956-1962.

Sultan, M., R. Becker, R. E. Arvidson, P. Shore, R. J. Stern, Z. Elalfy, and E. A. Guinness,

1992, Nature of the Red Sea crust: A controversy revisited: Geology, v. 20, p.

593-596.

25

Sultan, M., R. J. Stern, R. E. Arvidson, P. Shore, and R. Becker, 1993, Nature Of the Red-

Sea Crust - a Controversy Revisited - Reply: Geology, v. 21, p. 575-576.

Taylor, B., A. Goodliffe, F. Martinez, and R. Hey, 1995, Continental rifting and initial

sea-floor spreading in the Woodlark Basin: Nature, v. 374, p. 534.

Tramontini, C., and D. Davies, 1969, A Seismic Refraction Survey in The Red Sea:


Weissel, J. K., B. Taylor, and G. D. Karner, 1982, The opening of the Woodlark Basin,

subduction of the Woodlark spreading system, and the evolution of northern

Melanesia since mid-Pliocene time: Tectonophysics, v. 87, p. 253-277.

Wessel, P., W. H. Smith, R. Scharroo, J. Luis, and F. Wobbe, 2013, Generic mapping

tools: improved version released: Eos, Transactions American Geophysical

Union, v. 94, p. 409-410.

Wilson, J. T., 1966, Did the Atlantic close and then re-open?: Nature, v. 211, p. 676-

681.

Withjack, M. O., and W. R. Jamison, 1986, Deformation produced by oblique rifting:


26

Chapter 2.

Review of literature on the Red Sea and how

it prompts the present work

27

2. Review of literature on the Red Sea and how it

prompts the present work

2.1 Study area

The Red Sea is located between Africa and Arabia, with a length of ~2,000 km,

a maximum width of 355 km, and a surface area of ~458,620 km2 (Head, 1987). A

bathymetric map of the Red Sea is presented in Figure 2.1. In the north, the Red Sea

is bifurcated by the Sinai Peninsula into the Gulf of Suez and the Gulf of Aqaba (Figure

2.1). In the south, it is connected to the Indian Ocean through the Strait of Bab-el-

Mandeb and the Gulf of Aden (Figure 2.1). The Afar hotspot region (Figure 2.1) is

located near the southern end of the Red Sea (e.g., Nyblade et al., 2000; Sicilia et al.,

2008).

The study area in the central Red Sea is located between 19°N and 23°N (Figure

2.1). The central Red Sea is characterized by a complex pattern of axial deeps floored

with volcanic rocks and shallower inter-trough zones between them floored with

evaporites (e.g., Augustin et al., 2014; Bonatti, 1985; Izzeldin, 1982, 1987). The

magnetic anomalies found in this region are poorly lineated and seafloor spreading

anomalies can only be identified up to Chron 3 (e.g., Izzeldin, 1982, 1987; Ligi et al.,

2012; Phillips, 1970; Roeser, 1975; Searle and Ross, 1975). The southern Red Sea has

a comparatively straight and continuous axis with lineated magnetic anomalies (e.g.,

Augustin et al., 2014; Phillips, 1970; Roeser, 1975; Vine, 1966), whilst the northern

Red Sea has few lineated magnetic anomalies (Cochran, 2005) though a few isolated

deeps (e.g., Bonatti, 1985; Cochran and Karner, 2007; Ehrhardt and Hübscher, 2015;

Guennoc et al., 1988; Ligi et al., 2018; Martinez and Cochran, 1988; Pautot et al., 1984).

28

Figure 2.1 Map of Red Sea bathymetry (Smith and Sandwell, 1997, version 18.1) showing the

location of the central Red Sea study area. Red dots locate the prominent deeps in the central

Red Sea from Augustin et al. (2014) and Karbe (1987). From north to south, these are (1)

Nereus, (2) Thetis, (3) Hadarba, (4) Hatiba, (5) Atlantis II, (6) Erba, (7) Port Sudan, (8) Suakin,

and (9) Pelagia deeps. The relative plate motion vectors were calculated from the pole of Chu

and Gordon (1998).

Thick evaporites were deposited from Middle to Late Miocene (15–5 Ma)

(Figure 2.2) in the Red Sea basin, complicating the seismic imaging of underlying crust

structure and preventing direct sampling of basement except in isolated locations (e.g.,

Augustin et al., 2014; Bosworth et al., 2005; Girdler and Whitmarsh, 1974; Mitchell

and Park, 2014; Stoffers and Kühn, 1974; Tramontini and Davies, 1969; Whitmarsh et

al., 1974). These evaporites exceed 3 km in thickness in places (e.g., Izzeldin, 1987;

Tramontini and Davies, 1969). The seismic S-reflection from the top of the Miocene

29

evaporites is found throughout most of the Red Sea (e.g., Izzeldin, 1987; Knott et al.,

1966; Phillips and Ross, 1970; Ross and Schlee, 1973). The Plio-Pleistocene (PP)

sediments overlying the S-reflection are thin (only 0.2–0.3 km thick) and tend to be

uniform (e.g., Izzeldin, 1987; Phillips and Ross, 1970; Ross and Schlee, 1973). In deep

waters away from the coasts, they have a similar density to halite (Wheildon et al.,

1974), hence diapirism is muted, whereas diapirism occurs under the thicker and

denser PP sediments along the coasts (Mitchell et al., 2017).

Figure 2.2 Simplified stratigraphic sections of the Red Sea and Gulf of Suez (Bosworth, 2015).

2.2 Geological and tectonic setting

The Red Sea is a young ocean basin that is at various stages in the transition

from continental rifting to seafloor spreading, though the exact stage that some areas

are at have been controversial (e.g., Bonatti et al., 1981; Cochran and Martinez, 1988;

30

Hall, 1989; Rihm and Henke, 1998). How this transition is evolving along the Red Sea

rift is therefore still open to vigorous debate.

2.2.1 Seafloor spreading and continental rifting in the Red Sea

The Red Sea was formed as the African-Arabian shield rifted, allowing the

distinct Arabian and Nubian plates to separate (e.g., Ghebreab, 1998; McKenzie et al.,

1970). The extension of the Red Sea may have begun in the Eocene but developed

more greatly in the Oligocene associated with massive and rapidly erupted basalts in

Ethiopia and southern Yemen at ~30 Ma (Bosworth and McClay, 2001; Hofmann et al.,

1997; Mohr, 1983; Omar and Steckler, 1995). These rapid eruptions have been

attributed to the Afar plume penetrating the lithosphere (Furman et al., 2006; George

et al., 1998; Richards et al., 1989). The present Red Sea opening rate increases

southward from ~10 mm yr−1 at 25.5°N to ~16 mm yr−1 near 18°N away from the

Nubia/Africa spreading pole located in the Mediterranean (e.g., Chu and Gordon, 1998;

DeMets et al., 1990; DeMets et al., 2010; Reilinger et al., 2015).

2.2.1.1 Southern Red Sea

The southern Red Sea (south of 19°N; Figure 2.1) has a continuous axial zone

with extensive volcanism and seafloor spreading magnetic anomalies (e.g., Augustin

et al., 2014; Phillips, 1970; Roeser, 1975). These lineated seafloor spreading magnetic

anomalies are clearly identifiable up to anomaly Chron 3 (5 Ma, Figure 2.3) near the

axial trough between 16°N and 19°N, suggesting that recognizable seafloor spreading

began at least by 5 Ma (e.g., Cochran, 1983; Girdler and Styles, 1974; Phillips, 1970;

Roeser, 1975; Vine, 1966). Augustin et al. (2014, 2016) suggested that seafloor

spreading likely began somewhat earlier, 8-12 Ma, based on spreading rates of Chu

and Gordon (1998) and locations of volcanic ridges interpreted from multibeam sonar

and vertical gravity gradient data. In addition to the clear anomalies in the centre of

the basin, Girdler and Styles (1974) interpreted the low amplitude magnetic anomalies

over the southern Red Sea shelves as additional seafloor spreading magnetic stripes,

suggesting the Red Sea was formed by two stages of seafloor spreading. In contrast,

Cochran (1983) argued that these magnetic anomalies result from a wide region of

mafic diking and intrusions rather than a continuous oceanic crust of dykes and

31

extrusives, because the anomalies have low amplitudes (less than 200 nT) and long

wavelengths (20-50 km). However, Hall (1989) supported the interpretation of

seafloor spreading, because he found the reduced-to-pole magnetic anomalies are

linear and symmetrical about the axis. Based on magnetic and gravity modelling

constrained by the seismic refraction data of Gettings et al. (1986) and Mooney et al.

(1985), Almalki et al. (2014) recently suggested that about 75 km of oceanic crust

formed before Middle to Late Miocene (15–5 Ma) under the Farasan Bank (Figure 2.1),

which supports a two-stage spreading evolution of the Red Sea. A seismic refraction

line shot across the Yemen margin by Egloff et al. (1991) was interpreted as showing

that oceanic crust adjacent to the axis terminates southward of 16°N and that

continental-type crust lies farther south as far as 14˚N.

Figure 2.3 Seafloor spreading magnetic anomalies in the southern Red Sea (Phillips, 1970).

The bottom curve represents synthetic magnetic anomalies generated by the seafloor

spreading model beneath it using a spreading rate of 10 mm yr-1. The black blocks indicate

normal magnetization, whereas open blocks represent reversed magnetization.

Going south of 18°N, increasing influence of the Afar melting anomaly is

indicated by basalt Sr-Nd-Pb isotopes and higher 3He/4He ratios, in both major and

trace elements (Altherr et al., 1988; Moreira et al., 1996; Volker et al., 1997).

32

2.2.1.2 Central Red Sea

In the central Red Sea, the topographic structure of the axial zone becomes

less lineated, consisting of a series of ‘deeps’, which are separated by inter-trough

zones. The inter-trough zones are shallower, lack strong magnetic anomalies

compared to the ‘deeps’, and are covered by evaporites that have flowed laterally

along with overlying sediments across the axis (e.g., Augustin et al., 2014; Izzeldin,

1982, 1987).

How far the central Red Sea is through the transition from continental rifting

to full seafloor spreading has been controversial.

Bonatti (1985) suggested the central Red Sea is just at the point of transitioning

from continental rifting to oceanic spreading. He proposed that the ‘deeps’ found in

the central Red Sea are discrete seafloor spreading cells based on the presence of

normal mid-ocean ridge basalt (MORB) sampled from them and their high amplitude

magnetic anomalies. This contrasts with low amplitude anomalies outside the deeps,

which were therefore assumed to overlie stretched continental crust (Ligi et al., 2011,

2012).

However, other evidence indicates more established seafloor spreading in the

central Red Sea. An extensive aeromagnetic survey revealed that there are low

amplitude magnetic anomalies outside the ‘deeps’ that are lineated and aligned

parallel to the ridge axis (Izzeldin, 1987; Rasul et al., 2015). LaBrecque and Zitellini

(1985) showed with numerical modelling that such subdued anomalies could be

produced by widely distributed dykes, lava flows, and sills, as occur in modern-day

Afar. Low amplitudes may also have arisen from the slow spreading rate, the greater

depth of basement and alteration under the evaporites (Augustin et al., 2014; Dyment

et al., 2013; Izzeldin, 1987, 1989; Mitchell and Park, 2014). Based on seismic reflection,

gravity and magnetic data, Izzeldin (1982, 1987) suggested the inter-trough zones are

manifestations of less organized seafloor spreading and underlain by oceanic crust.

Using multibeam sonar data, Augustin et al. (2014, 2016) also interpreted these inter-

trough zones as merely areas where the off-axis evaporites have flowed into the axis,

obscuring the volcanic geomorphology. Seismic refraction data show the velocity of

33

basement under the inter-trough zone between Nereus and Thetis deeps (Figure 2.1)

is 6.86 km s-1 (Davies and Tramontini, 1970; Tramontini and Davies, 1969) which

overlaps with velocities of oceanic crust elsewhere (6.7-6.9 km s-1; Carlson, 2001,

2010). Further seismic refraction data collected by Egloff et al. (1991) along their line

PIII (near 19.5°N) also suggested the basement around the ridge axis is oceanic, which

transitions to stretched continental crustal velocities towards the coast of Sudan.

Finally, free-air gravity anomalies derived from satellite altimeter data (Sandwell et al.,

2014; Sandwell and Smith, 2009) reveal that anomalies along the spreading centre are

segmented. Mitchell and Park (2014) and Augustin et al. (2016) suggested that this

segmented pattern is similar to the segmentation observed at slow-spreading mid-

ocean ridges elsewhere (e.g., the northern Mid-Atlantic Ridge) (Schouten et al., 1987;

Sempéré et al., 1990).

Based on seismic reflection and potential field data, Izzeldin (1982, 1987)

suggested that intermediate crust separates crust that is continental near the coasts

from that which is oceanic around the axis. This area lies ~65-160 km from the ridge

axis.

2.2.1.3 Northern Red Sea

The northern Red Sea (North of 23°N; Figure 2.1), which is closer to the Euler

pole of rotation of the relative motion of the African plate with respect to the Arabian

plate (e.g., Chu and Gordon, 1998; DeMets et al., 1990; DeMets et al., 2010) and has

experienced less extension than our study area, has been thought to be underlain by

continental crust with a series of large crustal fault blocks interpreted from seismic

velocity data and from magnetic and gravity anomalies (Cochran and Karner, 2007;

Gaulier et al., 1986; Martinez and Cochran, 1988). Cochran and Karner (2007)

suggested that the strong lithosphere resulting from low mantle temperature in the

northern Red Sea is preventing the transition to oceanic spreading. If correct, this area

may still be in late stage continental rifting (e.g., Cochran, 1983; Gaulier et al., 1988;

Martinez and Cochran, 1988). In contrast, others have interpreted this region as

underlain by oceanic crust on the basis of unpublished seismic and magnetic data (e.g.,

Dyment et al., 2013; Tapponnier et al., 2013). Using remote sensing, geochemical, and

34

geochronological data, Sultan et al. (1992) carried out a plate reconstruction for the

opening of the Red Sea and found a best match of pre-existing African and Arabian

geologic features by juxtaposing present Red Sea coastlines. This has been interpreted

as indicating that the entire Red Sea basin is underlain by oceanic crust (Bosworth et

al., 1993). In addition, a few ‘deeps’ containing basalts are revealed in the northern

Red Sea (e.g., Bonatti, 1985; Ehrhardt and Hübscher, 2015; Guennoc et al., 1988; Ligi

et al., 2018; Pautot et al., 1984).

2.2.2 Seismic tomographic studies encompassing the Red Sea

Seismic tomographic studies have found that upper mantle S- and P-wave

seismic velocities of the adjacent to the Red Sea increase by up to a few percent from

south to north with increasing distance from the Afar plume. Using body wave travel

time tomography, Park et al. (2007) found a -1.5% S-wave velocity anomaly at 200 km

depth beneath the coast of the southern Red Sea, rising to ~-1% in the central Red Sea

and to 0% or more in the northern Red Sea. A similar structure was found by Park et

al. (2008) from Rayleigh wave tomography although with a more subdued northwards

increase in S-wave velocity. They suggested this structure is caused by an upwelling

of warm mantle beneath the southern Arabian shield, originating from the Afar

hotspot. They proposed that this hot plume material flows from Afar underneath the

southern and central Red Sea, and then extends northwards beneath Arabia, whereas

the northern Red Sea (north of ~23°N) is without underlying hot mantle. The upper

mantle under the Red Sea is poorly resolved in such models as they are mainly based

on teleseismic recordings along the coast of Saudi Arabia (no offshore recordings).

Nevertheless, this general pattern of upper mantle structure is corroborated by Na8.0

analyses of axial lavas (sodium oxide concentrations corrected for fractionation (Klein

and Langmuir, 1987)), which indicate the upper mantle temperature in the Red Sea

generally decreases by about 60°C from 18°N to 26°N (Haase et al., 2000).

35

Figure 2.4 Shear wave velocity map at a depth of 150 km (Chang et al., 2011). Shear wave

splitting data (blue lines) were derived from Gashawbeza et al. (2004) and Hansen et al. (2006).

More recently, Chang et al. (2011) have carried out an inversion of seismic

travel times and waveforms that provides a more complete coverage of the area

under the Red Sea (Figure 2.4). They found low velocity (hot) material is located

beneath the southern Red Sea and Gulf of Aden, where there is active seafloor

spreading. They also suggested that the hot material at a depth of ~150 km does not

extend north-westwards below the central and northern Red Sea areas, but forms a

channel extending northward beneath Arabia. The higher velocities under the central

Red Sea prompt the question of whether we should expect a non-volcanic margin in

this area and whether any seafloor spreading should be unlike at a hotspot.

36

The Afar mantle plume is located at the Afar Triangle, with hot mantle plume

material expected to spread out in a star-like pattern into the soft asthenosphere

(Courtillot et al., 1999; Ebinger and Sleep, 1998; Schilling, 1973). Recently, the

existence of another separate mantle hotspot beneath the northern Arabia and

Jordan was suggested by Chang and Van der Lee (2011), based on their tomographic

results. They suggested that this separate mantle plume could be the reason for

simultaneous northward and southward migration of the Neogene volcanism

between Afar and Jordan (Bosworth et al., 2005; Camp and Roobol, 1992). As a result

of this volcanism migration, significant upwelling mantle occurs in western Arabia

(Figure 2.4).

Cenozoic volcanism on the continental flanks of the Red Sea are asymmetric,

being more common in Arabia (e.g., Dixon et al., 1989; Makris and Rihm, 1991).

Abundant basaltic lava fields characterize the eastern flank (Saudi Arabia), while the

volcanisms are absent on the western flank (Egypt and Sudan). Besides the

explanation of double mantle plume model proposed by Chang and Van der Lee (2011),

some studies (Voggenreiter et al., 1988; Wernicke, 1985) explained this asymmetry by

suggesting that the Red Sea is opening on an east-dipping low-angle detachment fault

(the Wernicke Model). However, Camp and Roobol (1991) argued that this low-angle

detachment model does not accord with field and age data delineating the timing of

magmatism and uplift on the Arabian plate. Bosworth (2015) suggested that this

simple shear low-angle detachment geometry cannot reproduce the value or

distribution of heat flow, which is more nearly symmetric (Girdler, 1970; Girdler and

Evans, 1977; Scheuch, 1976).

2.3 How the previous studies prompt the present work

Although there are strong indications that much of the central Red Sea is

underlain by oceanic crust (Augustin et al., 2014; Izzeldin, 1982, 1987; Mitchell and

Park, 2014), there is still some doubt, given that the off-axis magnetic anomalies are

not identifiable as seafloor spreading anomalies and the seismic refraction data are

limited. Augustin et al. (2014, 2016) and Mitchell and Park (2014) have suggested that

evaporites are flowing into the axis at areas where the basement is deeper (oceanic

37

fracture zones), although this has remained to be proven given the previously

available deep-seismic data able to penetrate the evaporites. Furthermore, the more

recent mantle seismic tomography results (Chang et al., 2011) also prompt the

question of whether any underlying crust, if it is oceanic, is likely to be similar to ridges

away from hotspots rather than those proximal to hotspots.

This research has been enabled by accessing to the Izzeldin (1982, 1987)

multichannel seismic reflection data, which penetrate the evaporites, as well as the

newly released satellite gravity data (Sandwell et al., 2014) (version 23), and the

aeromagnetic anomalies (Izzeldin, 1982, 1987). This combination allows us to address

the above problems. In this study, those multichannel seismic reflection seismic,

satellite gravity, and aeromagnetic data are used to evaluate the basement geometry

and roughness in the central Red Sea. We then compare our findings with other mid-

ocean ridges and discuss what these results imply about the evolution of the Red Sea

rift in the central Red Sea.

2.4 References

Almalki, K. A., P. G. Betts, and L. Ailleres, 2014, Episodic sea-floor spreading in the

southern Red Sea: Tectonophysics, v. 617, p. 140-149.

Altherr, R., F. Henjes-Kunst, H. Puchelt, and A. Baumann, 1988, Volcanic activity in the

Red Sea axial trough—evidence for a large mantle diapir?: Tectonophysics, v.

150, p. 121-133.

Augustin, N., C. W. Devey, F. M. van der Zwan, P. Feldens, M. Tominaga, R. A. Bantan,

and T. Kwasnitschka, 2014, The rifting to spreading transition in the Red Sea:

Earth and Planetary Science Letters, v. 395, p. 217-230.






38

Bonatti, E., P. Hamlyn, and G. Ottonello, 1981, Upper mantle beneath a young oceanic

rift: peridotites from the island of Zabargad (Red Sea): Geology, v. 9, p. 474-

479.

Bosworth, W., 2015, Geological evolution of the Red Sea: historical background,

review, and synthesis, The Red Sea, Springer, p. 45-78.

Bosworth, W., P. Huchon, and K. McClay, 2005, The Red Sea and Gulf of Aden basins:

Journal of African Earth Sciences, v. 43, p. 334-378.

Bosworth, W., and K. McClay, 2001, Structural and stratigraphic evolution of the Gulf

of Suez rift, Egypt: a synthesis: Mémoires du Muséum national d'histoire

naturelle, v. 186, p. 567-606.

Bosworth, W., M. Sultan, R. Stern, R. Arvidson, P. Shore, and R. Becker, 1993, Nature

of the Red Sea crust: A controversy revisited: Comment and Reply: Geology, v.

21, p. 574-576.

Camp, V. E., and M. J. Roobol, 1991, Comment on “Topographic and volcanic

asymmetry around the Red Sea: Constraints on rift models” by TH Dixon, ER

Ivins, and JF Brenda: Tectonics, v. 10, p. 649-652.

Camp, V. E., and M. J. Roobol, 1992, Upwelling asthenosphere beneath western Arabia

and its regional implications: Journal of Geophysical Research: Solid Earth, v.

97, p. 15255-15271.

Carlson, R., 2001, The effects of temperature, pressure, and alteration on seismic

properties of diabase dike rocks from DSDP/ODP Hole 504B: Geophysical

research letters, v. 28, p. 3979-3982.

Carlson, R., 2010, How crack porosity and shape control seismic velocities in the upper

oceanic crust: Modeling downhole logs from Holes 504B and 1256D:


doi:10.1029/2009GC002955).

Chang, S.-J., and S. Van der Lee, 2011, Mantle plumes and associated flow beneath

Arabia and East Africa: Earth and Planetary Science Letters, v. 302, p. 448-454.

Chang, S. J., M. Merino, S. Van der Lee, S. Stein, and C. A. Stein, 2011, Mantle flow

beneath Arabia offset from the opening Red Sea: Geophysical Research Letters,

v. 38.

39



Cochran, J., and G. Karner, 2007, Constraints on the deformation and rupturing of

continental lithosphere of the Red Sea: the transition from rifting to drifting:

Geological Society, London, Special Publications, v. 282, p. 265-289.

Cochran, J. R., 1983, A model for development of Red Sea: AAPG Bulletin, v. 67, p. 41-

69.

Cochran, J. R., 2005, Northern Red Sea: nucleation of an oceanic spreading center

within a continental rift: Geochemistry, Geophysics, Geosystems, v. 6, Q03006,

doi:10.1029/2004GC000826.

Cochran, J. R., and F. Martinez, 1988, Evidence from the northern Red Sea on the

transition from continental to oceanic rifting: Tectonophysics, v. 153, p. 25-53.

Courtillot, V., C. Jaupart, I. Manighetti, P. Tapponnier, and J. Besse, 1999, On causal

links between flood basalts and continental breakup: Earth and Planetary

Science Letters, v. 166, p. 177-195.




DeMets, C., R. G. Gordon, D. Argus, and S. Stein, 1990, Current plate motions:


DeMets, C., R. G. Gordon, and D. F. Argus, 2010, Geologically current plate motions:


Dixon, T. H., E. R. Ivins, and B. J. Franklin, 1989, Topographic and volcanic asymmetry

around the Red Sea: Constraints on rift models: Tectonics, v. 8, p. 1193-1216.

Dyment, J., P. Tapponnier, A. Afifi, M. Zinger, D. Franken, and E. Muzaiyen, 2013, A

new seafloor spreading model of the Red Sea: Magnetic anomalies and Plate

kinematics: AGU Fall Meeting Abstracts (T21A-2512).

Ebinger, C. J., and N. Sleep, 1998, Cenozoic magmatism throughout east Africa

resulting from impact of a single plume: Nature, v. 395, p. 788.

Egloff, F., R. Rihm, J. Makris, Y. A. Izzeldin, M. Bobsien, K. Meier, P. Junge, T. Noman,

and W. Warsi, 1991, Contrasting structural styles of the eastern and western

40

margins of the southern Red Sea - the 1988 Sonne Experiment: Tectonophysics,

v. 198, p. 329-353.

Ehrhardt, A., and C. Hübscher, 2015, The northern Red Sea in transition from rifting to

drifting - lessons learned from ocean deeps. In: Rasul, N.M.A., and Stewart,

I.C.F. (Eds.), The Red Sea: The formation, morphology, oceanography and

environment of a young ocean basin, Springer Earth System Sciences, Berlin

Heidelberg, p. 99-121.

Furman, T., J. Bryce, T. Rooney, B. Hanan, G. Yirgu, and D. Ayalew, 2006, Heads and

tails: 30 million years of the Afar plume: Geological Society, London, Special

Publications, v. 259, p. 95-119.

Gashawbeza, E. M., S. L. Klemperer, A. A. Nyblade, K. T. Walker, and K. M. Keranen,

2004, Shear ‐wave splitting in Ethiopia: Precambrian mantle anisotropy

locally modified by Neogene rifting: Geophysical Research Letters, v. 31.

Gaulier, J., X. Le Pichon, N. Lyberis, F. Avedik, L. Geli, I. Moretti, A. Deschamps, and S.

Hafez, 1988, Seismic study of the crust of the northern Red Sea and Gulf of

Suez: Tectonophysics, v. 153, p. 55-88.

Gaulier, J., X. Le Pichon, N. Lyberis, F. Avedik, L. Gely, and I. Moretti, 1986, New

refraction data on the Northern Red Sea-Gulf of Suez area: EOS Trans. Am.

Geophys. Union, v. 67, p. 1208-1209.

George, R., N. Rogers, and S. Kelley, 1998, Earliest magmatism in Ethiopia: evidence

for two mantle plumes in one flood basalt province: Geology, v. 26, p. 923-926.

Gettings, M. E., H. Blank, W. Mooney, and J. Healey, 1986, Crustal structure of

southwestern Saudi Arabia: Journal of Geophysical Research: Solid Earth, v. 91,

p. 6491-6512.

Ghebreab, W., 1998, Tectonics of the Red Sea region reassessed: Earth-Science

Reviews, v. 45, p. 1-44.

Girdler, R., 1970, A review of Red Sea heat flow: Philosophical Transactions of the

Royal Society of London A: Mathematical, Physical and Engineering Sciences,

v. 267, p. 191-203.

Girdler, R., and T. Evans, 1977, Red Sea heat flow: Geophysical Journal International,

v. 51, p. 245-251.

41

Girdler, R., and P. Styles, 1974, Two stage Red Sea floor spreading: Nature, v. 247, p.

7-11.

Girdler, R. W., and R. B. Whitmarsh, 1974, Miocene evaporites in Red Sea cores, their

relevance to the problem of the width and age of oceanic crust beneath the

Red Sea: In: Whitmarsh, R.B., Weser, O.E., Ross, D.A., et al. Initial Reports of

the Deep Sea Drilling Project, v. 23, p. 913-921.

Guennoc, P., G. Pautot, and A. Coutelle, 1988, Surficial structures of the northern Red

Sea axial valley from 23°N to 28°N : time and space evolution of neo-oceanic

structures: Tectonophysics, v. 153, p. 1-23.

Haase, K. M., R. Mühe, and P. Stoffers, 2000, Magmatism during extension of the

lithosphere: geochemical constraints from lavas of the Shaban Deep, northern

Red Sea: Chemical Geology, v. 166, p. 225-239.

Hall, S. A., 1989, Magnetic evidence for the nature of the crust beneath the southern

Red Sea: Journal of Geophysical Research: Solid Earth And Planets, v. 94, p.

12267-12279.

Hansen, S., S. Schwartz, A. Al-Amri, and A. Rodgers, 2006, Combined plate motion and

density-driven flow in the asthenosphere beneath Saudi Arabia: Evidence from

shear-wave splitting and seismic anisotropy: Geology, v. 34, p. 869-872.

Head, S. M., 1987, Red Sea fisheries: In: Edwards AJ, Head SM (eds) Red Sea: key

environments: Pergamon Press, Oxford, 363-382 p.

Hofmann, C., V. Courtillot, G. Feraud, P. Rochette, G. Yirgu, E. Ketefo, and R. Pik, 1997,

Timing of the Ethiopian flood basalt event and implications for plume birth and

global change: Nature, v. 389, p. 838-841.


Strasbourg.




Izzeldin, A., 1989, Transverse structures in the central part of the Red Sea and

implications on early stages of oceanic accretion: Geophysical Journal

International, v. 96, p. 117-129.

42

Karbe, L., 1987, Hot brines and the deep sea environment. In: Edwards A.J., and Head,

S.M. (Eds.) Key Environment, Red Sea, Pergamon Press, Oxford, p. 70-89.

Klein, E. M., and C. H. Langmuir, 1987, Global correlations of ocean ridge basalt

chemistry with axial depth and crustal thickness: Journal of Geophysical

Research: Solid Earth, v. 92, p. 8089-8115.

Knott, S. T., E. T. Bunce, and R. Chase, 1966, Red Sea seismic reflection studies: The

World Rift System, Geol. Surv. Canada,Paper 66-14, 78-97.

LaBrecque, J., and N. Zitellini, 1985, Continuous sea-floor spreading in Red Sea: an

alternative interpretation of magnetic anomaly pattern: AAPG Bulletin, v. 69,

p. 513-524.




doi:10.1029/2012GC004155).

Ligi, M., E. Bonatti, W. Bosworth, Y. Cai, A. Cipriani, C. Palmiotto, S. Ronca, and M.

Seyler, 2018, Birth of an ocean in the Red Sea: Oceanic-type basaltic melt

intrusions precede continental rupture: Gondwana Research, v. 54, p. 150-160.




p. 1019-1022.

Makris, J., and R. Rihm, 1991, Shear-controlled evolution of the Red Sea: pull apart

model: Tectonophysics, v. 198, p. 441-466.

Martinez, F., and J. R. Cochran, 1988, Structure and tectonics of the northern Red Sea:

catching a continental margin between rifting and drifting: Tectonophysics, v.

150, p. 1-31.

McKenzie, D., D. Davies, and P. Molnar, 1970, Plate tectonics of the Red Sea and East

Africa: Nature, v. 226, p. 243.

Mitchell, N. C., M. Ligi, P. Feldens, and C. Hübscher, 2017, Deformation of a young salt

giant: regional topography of the Red Sea Miocene evaporites: Basin Research,

v. 29, p. 352-369.

43



Mohr, P., 1983, Ethiopian flood basalt province: Nature, v. 303, p. 577-584.

Mooney, W. D., M. E. Gettings, H. R. Blank, and J. H. Healy, 1985, Saudi Arabian

seismic-refraction profile: a traveltime interpretation of crustal and upper

mantle structure: Tectonophysics, v. 111, p. 173-246.

Moreira, M., P. Valbracht, T. Staudacher, and C. Allègre, 1996, Rare gas systematics in

Red Sea ridge basalts: Geophysical Research Letters, v. 23, p. 2453-2456.

Nyblade, A. A., R. P. Knox, and H. Gurrola, 2000, Mantle transition zone thickness

beneath Afar: implications for the origin of the Afar hotspot: Geophysical


Omar, G. I., and M. S. Steckler, 1995, Fission track evidence on the initial rifting of the

Red Sea: two pulses, no propagation: Science, v. 270, p. 1341-1344.

Park, Y., A. A. Nyblade, A. J. Rodgers, and A. Al‐Amri, 2007, Upper mantle structure

beneath the Arabian Peninsula and northern Red Sea from teleseismic body

wave tomography: Implications for the origin of Cenozoic uplift and volcanism

in the Arabian Shield: Geochemistry, Geophysics, Geosystems, v. 8 (Paper

Q06021, doi:10.1029/2006GC001566).

Park, Y., A. A. Nyblade, A. J. Rodgers, and A. Al‐Amri, 2008, S wave velocity structure

of the Arabian Shield upper mantle from Rayleigh wave tomography:


doi:10.1029/2007GC001895).

Pautot, G., P. Guennoc, A. Coutelle, and N. Lyberis, 1984, Discovery of a large brine

deep in the northern Red Sea: Nature, v. 310, p. 133.

Phillips, J. D., 1970, Magnetic anomalies in the Red Sea: Philosophical Transactions of

the Royal Society of London A: Mathematical, Physical and Engineering

Sciences, v. 267, p. 205-217.

Phillips, J. D., and D. A. Ross, 1970, Continuous seismic reflexion profiles in the Red

Sea: Philosophical Transactions of the Royal Society of London A:

Mathematical, Physical and Engineering Sciences, v. 267, p. 143-152.

44

Rasul, N. M., I. C. Stewart, and Z. A. Nawab, 2015, Introduction to the Red Sea: Its

Origin, Structure, and Environment, The Red Sea, Springer, p. 1-28.

Reilinger, R., S. McClusky, and A. ArRajehi, 2015, Geodetic constraints on the

geodynamic evolution of the Red sea, in Rasul, N., and Stewart, I. C., eds., The

Red Sea: The formation, morphology, oceanography and environment of a

young ocean basin, Springer, p. 135-149.

Richards, M. A., R. A. Duncan, and V. E. Courtillot, 1989, Flood basalts and hot-spot

tracks: plume heads and tails: Science, v. 246, p. 103-107.

Rihm, R., and C. Henke, 1998, Geophysical studies on early tectonic controls on Red

Sea rifting, opening and segmentation, Sedimentation and Tectonics in Rift

Basins Red Sea-Gulf of Aden, Springer, p. 29-49.

Roeser, H. A., 1975, A detailed magnetic survey of the southern Red Sea: Geologisches

Jahrbuch, v. 13, p. 131-153.

Ross, D. A., and J. Schlee, 1973, Shallow structure and geologic development of the

southern Red Sea: Geological Society of America Bulletin, v. 84, p. 3827-3848.




Sandwell, D. T., and W. H. Smith, 2009, Global marine gravity from retracked Geosat

and ERS‐1 altimetry: Ridge segmentation versus spreading rate: Journal of

Geophysical Research: Solid Earth (1978–2012), v. 114.

Scheuch, J., 1976, Preliminary heat flow map of the Red Sea and an attempt to provide

a geological-geophysical interpretation: Afar between Continental and

Oceanic Rifting, v. 2, p. 171-174.

Schilling, J., 1973, Afar mantle plume: rare earth evidence: Nature Physical Science, v.

242, p. 2-5.

Schouten, H., H. J. Dick, and K. D. Klitgord, 1987, Migration of mid-ocean-ridge volcanic

segments: Nature, v. 326, p. 835-839.

Searle, R. C., and D. A. Ross, 1975, A geophysical study of the Red Sea axial trough

between 20.5° and 22° N: Geophysical Journal International, v. 43, p. 555-572.

45

Sebai, A., E. Stutzmann, J.-P. Montagner, D. Sicilia, and E. Beucler, 2006, Anisotropic

structure of the African upper mantle from Rayleigh and Love wave

tomography: Physics of the Earth and Planetary Interiors, v. 155, p. 48-62.

Sempéré, J.-C., G. Purdy, and H. Schouten, 1990, Segmentation of the Mid-Atlantic

Ridge between 24 N and 30 40'N.

Sicilia, D., J.-P. Montagner, M. Cara, E. Stutzmann, E. Debayle, J.-C. Lépine, J.-J.

Lévêque, E. Beucler, A. Sebai, and G. Roult, 2008, Upper mantle structure of

shear-waves velocities and stratification of anisotropy in the Afar Hotspot

region: Tectonophysics, v. 462, p. 164-177.

Smith, W. H. F., and D. T. Sandwell, 1997, Global sea floor topography from satellite

altimetry and ship depth soundings: Science, v. 277, p. 1956-1962.

Stern, R. J., and P. Johnson, 2010, Continental lithosphere of the Arabian Plate: a

geologic, petrologic, and geophysical synthesis: Earth-Science Reviews, v. 101,

p. 29-67.

Stoffers, P., and R. Kühn, 1974, Red Sea evaporites: a petrographic and geochemical

study: Init. Rep. DSDP, v. 23, p. 821-847.



593-596.

Tapponnier, P., J. Dyment, M. Zinger, D. Franken, A. Afifi, A. Wyllie, H. Ali, and I. Hanbal,

2013, Revisiting Seafloor-Spreading in the Red Sea: Basement Nature,

Transforms and Ocean-Continent Boundary: AGU Fall Meeting Abstracts

(T12B-04).



Vine, F. J., 1966, Spreading of the ocean floor: new evidence: Science, v. 154, p. 1405-

1415.

Voggenreiter, W., H. Hötzl, and A. Jado, 1988, Red Sea related history of extension and

magmatism in the Jizan area (Southwest Saudi Arabia): indication for simple-

shear during early Red Sea rifting: Geologische Rundschau, v. 77, p. 257-274.

Volker, F., R. Altherr, K.-P. Jochum, and M. T. McCulloch, 1997, Quaternary volcanic

activity of the southern Red Sea: new data and assessment of models on

46

magma sources and Afar plume-lithosphere interaction: Tectonophysics, v.

278, p. 15-29.

Wernicke, B., 1985, Uniform-sense normal simple shear of the continental lithosphere:

Canadian Journal of Earth Sciences, v. 22, p. 108-125.

Wheildon, J., T. Evans, R. Girdler, R. Whitmarch, O. Weser, and D. Ross, 1974, Thermal

conductivity, density, and sonic velocity measurements of samples of

anhydrite and halite from Sites 225 and 227: Initial Reports of the Deep Sea

Drilling Project, v. 23, p. 909-911.

Whitmarsh, R., O. Weser, and D. Ross, 1974, Initial Reports of the Deep Sea Drilling

Project: US Government Printing Office, Washington, v. 23b.

47

Chapter 3.

Data and methods

48

3. Data and methods

3.1 Multichannel seismic reflection data

Multichannel seismic reflection data used in this study were collected in 1976

(Izzeldin, 1982, 1987), using a Vaporchoc source with a 2.4 km streamer consisting of

48 channels (50 m spacing) in deep-water survey, and with a 1.2 km streamer

consisting of 24 channels (50 m spacing) in shallow-water survey. The data were

moveout-corrected by others as described by Izzeldin (1982, 1987). Unfortunately,

we do not have access to the digital data, but we have paper copies of the seismic

reflection profiles from Izzeldin (1982, 1987). The locations of seismic reflection

profiles 7, 9, 11, 15, 17, 19, 21, 25, 27, 29, and 31 are shown in Figures 3.1.

Figure 3.1 Map of free-air gravity anomalies (Sandwell et al., 2014, version 23.1) showing

the locations of multichannel seismic reflection profiles 7, 9, 11, 15, 17, 19, 21, 25, 27, 29, 31

of Izzeldin (1982, 1987). The data were gridded using the GMT software (Wessel et al., 2013).

Two-way travel times of the basement and seabed reflections were converted

to depths below sea level (Figure 3.2) as follows. A P-wave velocity (Vp) of 1.538 km

s-1 was used for the water according to the empirical equations of Mackenzie (1981),

49

with typical Red Sea salinity of 40 ppt and temperature of 21°C. A 1.9 km s-1 Vp for

the Plio-Pleistocene sediments was chosen, based on the measurements made on

samples recovered from DSDP Leg 23 (Whitmarsh et al., 1974) and the results of

seismic reflection and refraction surveys of Egloff et al. (1991) and Gaulier et al. (1988).

A Vp of 4.21 km s-1 was used for the evaporites, based on seismic refraction data from

Tramontini and Davies (1969), Girdler and Whitmarsh (1974), and Egloff et al. (1991).

Figure 3.2 Depths derived from the seismic reflection profiles of Izzeldin (1982, 1987). Line

numbers are shown in the lower right corner of each panel. Black lines denote bathymetry

(Smith and Sandwell, 1997, version 18.1). Dark green and cyan lines are the depths of the

seabed and the S-reflection at the top of the Miocene evaporites, respectively. Red and blue

lines are the basement reflection depths computed using the P-wave velocities of 4.21 km s-1

and 3 km s-1 for the evaporites, respectively.

Depths of the basement computed using a P-wave velocity of 4.21 km s-1 for

the evaporites and seabed derived from the seismic reflection profiles are used in

Chapters 4, 5, and 6.

50

Although they were collected 43 years ago, the seismic data quality is high.

Due to the thick evaporite layer and data noise, there is no Moho reflection obtained

from those data. On the western flank of profile 15, the basement reflection is not

clear enough to be identified. About 15% of the seismic data cannot be used to

compute the basement depth, because the basement reflection is in places absent or

un-interpretable.

The main uncertainty in the seismically derived basement depth originates

from the P-wave velocity used for the evaporites. During the seismic depth conversion,

the Vp of 4.21 km s-1 for the evaporites was chosen based on laboratory sample and

field refraction measurements of the upper evaporites, which are dominated by halite

(Girdler and Whitmarsh, 1974; Tramontini and Davies, 1969). However, the seismic

data show that there are many laminations in the lower evaporites also. Those

laminations are probably due to the laminated evaporites (halite, anhydrite, limestone

and shale), which would have a different velocity. Izzeldin (1982) suggested the

average seismic velocity of the lower evaporites ranges between 3 and 3.5 km s-1. To

address how the velocity affects the basement depths found, a Vp of 3 km s-1 for the

evaporites was used to compute another set of basement depths (in Figure 3.2). We

consider the models derived using the P-wave velocities of 3 km s-1 and 4.21 km s-1 for

evaporites as the minimum and maximum salt thickness models, respectively. In

Figure 3.2, the difference between the basement depths of these two models

increases with the basement two-way time. The difference increases from ~0.2 km to

~1.8 km as the basement depth derived using 4.21 km s-1 (red lines in Figure 3.2)

increases from 2 km to 6.7 km. The average difference is 1.5 km. Due to the

uncertainty in seismically derived basement depth, the height of the found axial

plateau is potentially overestimated.

Figure 3.3 is a depth to basement confidence map. In Figure 3.3, red indicates

where the basement reflections can be imaged with a high confidence, while those in

the areas away from the seismic lines have to be estimated using other geophysical

methods (e.g., Werner deconvolution of magnetic data).

51

Figure 3.3 Confidence map showing the ability to image the basement reflection.

3.2 Magnetic anomalies

3.2.1 Shipboard magnetic data

Marine magnetic field measurements from towed magnetometers were

obtained from the National Centers for Environmental Information (NCEI)

(www.ngdc.noaa.gov/mgg). The data comprise residual magnetic anomalies after

removal of the international geomagnetic reference field (IGRF) from the total field

measurements. Figure 3.4 shows the survey locations and contoured anomalies after

further adjustments to correct IGRF errors of the individual surveys (see figure

caption).

52

Figure 3.4 (a): Tracks of shipboard magnetic surveys of RVs Jean Charcot (78005111,

83008011), Atlantis (A2093L19), Chain (CH043L01, CH043L03, CH061L02, CH100L03),

Discovery (DI103B), Glomar Challenger (DSDP23GC), Melville (INMD09MV), Robert Conrad

(RC0911A), Shackleton (SHA1079) and Vema (V1413). (b): Residual magnetic anomalies of the

surveys in (a) obtained from the National Centers for Environmental Information (NCEI)

(www.ngdc.noaa.gov/mgg) gridded and contoured every 50 nT. To reduce effects of

reference field errors, the residual anomalies of each survey were adjusted by subtracting

their mean value before gridding and contouring. (Anomalies are not reduced to the pole.)

The data were gridded and contoured using the GMT software (Wessel et al., 2013).

Major causes of magnetic anomalies are expected to be susceptibility and

remanent magnetization variations within the basement produced by intrusive or

extrusive volcanic bodies. To investigate possible magnetic sources, Werner

deconvolution was applied to calculate the magnetic source depths and apparent

susceptibilities. They were calculated along individual segments of the magnetic lines

(ungridded magnetic data) where they cross the seismic reflection profiles of Izzeldin

(1987) (Figure 3.5a), and then were projected to the seismic profiles (Figure 3.5b).

53

Figure 3.5 (a): Extents of shipboard magnetic lines (blue) contributing to the evaluation of

seismic profiles (red) using Werner source depths. (b): A sketch showing how the source

depths and apparent susceptibilities were projected onto the seismic profiles. A source

located at point M was assigned position SM along the seismic line.

In Chapter 4, the projected magnetic source solutions are used to verify the

basement depths derived from the seismic reflection data.

The data quality of shipboard residual magnetic data is generally high. Those

data can roughly reflect the fracture zones in the central Red Sea (Figure 3.4b). More

than 95% of the data have reasonable values (between -1000 nT and 1000 nT). Less

54

than 10% of the data are distorted due to the shake and course changes of the ship.

The mean value of each survey ranges from to 10-700 nT.

3.2.2 Aeromagnetic data

The aeromagnetic survey was carried out in 1976 by the Arabian Geophysical

and Surveying Company (ARGAS) as described by Saudi-Sudanese Red Sea

Commission (1976) and Izzeldin (1982, 1987). This survey covered the central Red Sea

between 18.5 °N and 23 °N with flight lines oriented N60°E (Figure 3.6a). Sixty-four

survey lines were run from coast to coast and spaced at a distance of 10 km. 464

shorter lines were added over the axial zone and some coastal zones for detailed

investigation with the lines spaced 2.5 km apart. The total length of the main survey

lines is 23,011 km, while that of additional lines is 28,210 km. The survey was flown

at an altitude of 305 m above mean sea level. Measurements of total magnetic field

were obtained using a caesium vapour magnetometer with a resolution of 0.01 nT at

a sample frequency of 1 Hz.

Diurnal variations were corrected using data from a magnetometer station in

Jeddah and the results verified by analysis of data at crossing lines. Minor line levelling

errors in the data were minimised. Residual magnetic anomalies derived from the

total field data were used in this study. Figure 3.6b shows the contoured anomalies

after further adjustments made here to correct minor remaining data offsets of the

individual lines (see figure caption). The length of survey line tends to affect the

number of magnetic sources found by the Werner deconvolution (a longer length of

data is needed to resolve deeper magnetic bodies), so we used the longer lines for our

analysis (Figure 3.6c).

In Chapter 5, aeromagnetic data are used to compute magnetic basement

depth throughout the central Red Sea.

In Chapter 6, aeromagnetic anomalies were adjusted using reduction to the

pole technique by Ian C.F. Stewart. The reduced-to-pole aeromagnetic anomalies are

symmetric and lack of correlation with basement topography, indicating they are likely

seafloor spreading anomalies, explained by Hall (1989).

55

Figure 3.6 (a): Locations of aeromagnetic survey flight lines (blue) and multichannel seismic

reflection profiles 7, 9, 11, 15, 17, 19, 21, 25, 27, 29, 31 of Izzeldin (1987) (red). Black line

locates the spreading axis. (b): Residual aeromagnetic anomalies from data collected by

Arabian Geophysical and Surveying Company (ARGAS) in 1976. The data have been gridded

and contoured every 75 nT. In order to reduce effects of remaining reference field errors, the

residual aeromagnetic anomalies of each survey line were adjusted by subtracting their mean

anomaly value before gridding and contouring. The data were gridded and contoured using

the GMT software (Wessel et al., 2013). (c): Locations of the long survey lines used to derive

magnetic basement depth in Chapter 5 (blue). (d): Residual aeromagnetic anomalies along

the long survey lines in (c).

56

The aeromagnetic data quality is high. More than 95% of the data are

continuous and smooth (Figure 3.6d). All the observed aeromagnetic values range

from -600 nT to 600 nT, which are reasonable values. The variation between the mean

values of each survey line is less than 100 nT. Those data can roughly reveal the

fracture zones in the central Red Sea (Figure 3.6b). The residual aeromagnetic

anomalies are generally consistent with the magnetic anomalies reported in other

studies, for example, Roeser (1975) and Hall et al. (1977).

3.2.3 2D Werner deconvolution

2D Werner deconvolution has been widely used in the assessment of magnetic

basement structure by estimating the magnetic source depth and susceptibility (e.g.

Karner et al., 1991; Ku and Sharp, 1983; Thakur et al., 2000).

This method was originally designed to solve the thin-dike problem (Ku and

Sharp, 1983; Werner, 1953). The edges of a thick body can also be recognized by 2D

Werner deconvolution, because the horizontal gradient of the total magnetic field

caused by the edge of a thick body is equal to the total field from a thin-dike body (Ku

and Sharp, 1983; Werner, 1953). The 2D Werner deconvolution does not require any

initial model parameter values (Subrahmanyam and Gebissa, 2017). This method has

been tested in a large number of studies with synthetic and real magnetic data. Some

typical examples are as follows.

Based on synthetic studies, Jain (1976) suggested that the 2D Werner

deconvolution can locate the tops of thin sheets and the edges of thick bodies with

high accuracy. To test the validity of the 2D Werner deconvolution, Ku and Sharp

(1983) used it to resolve real magnetic anomalies and synthetic anomalies of various

geometric models, which include thin dikes (alternatively, bodies with a large ratio of

length to thickness), thick body edges, dike-like bodies with tapering edge from thin

to thick, and high/low basement topographic relief. They suggested that the 2D

Werner deconvolution can provide reliable estimates of magnetic source bodies.

Moreover, they also suggested that iteration, statistic decision, and seven-point

Werner operator are necessary for obtaining accurate estimates. The synthetic

studies of Subrahmanyam and Gebissa (2017) confirmed that the 2D Werner

57

deconvolution can yield reliable source parameter estimates for the sources of thin,

thick, wide, and faulted bodies.

Hartman et al. (1971) extended 2D Werner deconvolution in estimating the

basement depth around Rainbow Lake and Peace River in Canada by considering the

thin-dike body as an approximation. Because each Werner solution in their study is

associated to a single magnetic anomaly, the resolution of their results is low. Stagg

et al. (1989) mapped the magnetic basement in the Great Australian Bight based on

Werner solutions. Davy and Wood (1994) simply delineated the basement top using

visual identification of Werner source clusters. Thakur et al. (2000) revealed the

basement configuration in Bay of Bengal using 2D Werner deconvolution, although

short wavelength features were absent in the magnetic derived basement. Martelet

et al. (2013) used this method to extract magnetic sources at the basement interface

in the south-western part of the Paris Basin. During the extraction, they kept the

significant clusters of solutions and rejected the disseminated solutions.

Hansen and Simmonds (1993) extended the 2D Werner deconvolution to

interpret anomalies of multiple magnetic bodies and applied it to investigate the top

and bottom of the spreading center basalts in the Cobb Offset area of the Juan de Fuca

Ridge. Hansen (2005) developed 3D multiple-source Werner deconvolution based on

the algorithm of Hansen and Simmonds (1993). Hansen (2005) suggested that 3D

multiple-source Werner deconvolution can provide useful depth estimates and

locations of the sources, but it cannot provide useful dip angle and susceptibility

information.

Although individual depth values derived using the method have large

uncertainties and the method can produce some erroneous solutions due to non-

uniqueness and calculation window sizes poorly matching those needed for the

solutions, their depths have been shown generally to cluster within basement

(Cochran and Karner, 2007; Karner et al., 1991).

58

Following Werner (1953) and Ku and Sharp (1983), the magnetic anomaly due

to a dike or other tabular body can be written as:

𝑇𝑚𝑎𝑔(𝑥, 0) =𝐴(𝑥−𝑥0)+𝐵𝐷

(𝑥−𝑥0)2+𝐷2 (3.1)

where 𝑥 is distance along a profile, 𝐴 and 𝐵 are functions of orientation and

magnetization of the dike, 𝑥0 is horizontal position of the point immediately above

the dike, and 𝐷 is depth to the top of the dike (Figure 3.7).

Figure 3.7 Parameters of a vertical thin dike resolved using 2D Werner deconvolution (Ku

and Sharp, 1983).

The interferences from neighbouring anomalies, regional trends, and

measured magnetic noise are addressed by the addition of an interference polynomial

(𝑃) to the right side of equation (3.1) (Ku and Sharp, 1983; Werner, 1953):

𝑃 = 𝐶0 + 𝐶1𝑥 + 𝐶2𝑥2 + ⋯ + 𝐶𝑛𝑥𝑛 (3.2)

where 𝑛 is the order of the interference polynomial, and 𝐶0, 𝐶1, 𝐶2, ⋯ , 𝐶𝑛 are the

coefficients.

In practice, a polynomial of order two is adequate for obtaining stable and

reliable solutions (Hartman et al., 1971; Ku and Sharp, 1983; Werner, 1953):

59

𝑇𝑚𝑎𝑔(𝑥, 0) =𝐴(𝑥−𝑥0)+𝐵𝐷

(𝑥−𝑥0)2+𝐷2+ 𝐶0 + 𝐶1𝑥 + 𝐶2𝑥2 (3.3)

Equation (3.3) can be expressed as:

𝑎0 + 𝑎1𝑥 + 𝑎2𝑥2 + 𝑎3𝑥3 + 𝑎4𝑥4 + 𝑏0𝑇𝑚𝑎𝑔 + 𝑏1𝑥𝑇𝑚𝑎𝑔 = 𝑥2𝑇𝑚𝑎𝑔 (3.4)

where

𝑎0 = −𝐴𝑥0 + 𝐵𝐷 + 𝐶0𝐷2 + 𝑥02𝐶0,

𝑎1 = 𝐴 − 2𝐶0𝑥0 + 𝐶1𝐷2 + 𝐶1𝑥02,

𝑎2 = 𝐶0 − 2𝐶1𝑥0 + 𝐶2𝐷2 + 𝐶2𝑥02,

𝑎3 = 𝐶1 − 2𝐶2𝑥0,

𝑎4 = 𝐶2,

𝑏0 = −𝑥0 − 𝐷2,

and

𝑏1 = 2𝑥0

Marquardt’s (1963) inverse modelling method was used to solve the

simultaneous equations constructed from equation (3.4) by a seven-point Werner

operator (Ku and Sharp, 1983). If the sample spacing is ∆𝑥, we have the following

results for the thin dike (Ku and Sharp, 1983):

Horizontal position:

𝑋 = 0.5𝑏1∆𝑥 + 𝑥 (3.5a)

Depth:

𝑌 = √−𝑏0 − 0.25𝑏12∆𝑥

(3.5b)

Magnetic susceptibility: 𝜒𝑚 =√𝐽𝑥

2+𝐽𝑧2

|𝑭|∆𝑥 (3.5c)

where

𝐽𝑥 =−𝐵 cos 𝐼 sin 𝛼−𝐴 sin 𝐼

2∆𝑇[(cos 𝐼 sin 𝛼)2+(sin 𝐼)2] (3.6a)

and

60

𝐽𝑧 =−𝐴 cos 𝐼 sin 𝛼+𝐵 sin 𝐼

2∆𝑇[(cos 𝐼 sin 𝛼)2+(sin 𝐼)2] (3.6b)

where 𝑱𝒔 = ( 𝐽𝑥, 𝐽𝑧 ) is the vector sum of the induced and remanent magnetization,

2∆𝑇 is thickness of the dike (∆𝑇 ≪ 𝐷), 𝐼 is magnetic inclination of the main field 𝑭,

and 𝛼 is the strike of the tabular body measured counterclockwise from magnetic

north (Figure 3.7).

Matlab codes for 2D Werner deconvolution and Marquardt’s (1963) inverse

modelling method are provided in Appendix 2.

In Chapter 4, magnetic source solutions from Werner deconvolution of

shipboard magnetic data are used to verify the seismically derived basement depths.

In Chapter 5, Werner deconvolution is used to invert aeromagnetic anomalies for

magnetic basement depths.

Although 2D Werner deconvolution can be carried out quickly and can be

easily implemented (Hassan et al., 2007; Martelet et al., 2013), it can be limited by the

following seven issues:

(1): This method is limited to the analysis of simple 2D models (Kilty, 1983).

Magnetic source parameters derived from the 2D analysis tend to be less accurate

than those from the 3D analysis, because in reality the observed magnetic field are

produced by 3D objects.

(2): This method cannot work out the shape of magnetic bodies. Due to the

thin-dike assumption, the 2D Werner deconvolution can provide just the source

clusters around the tops of thin dikes and the edges of thick bodies (Ku and Sharp,

1983; Werner, 1953). The shape of magnetic bodies instead needs to be estimated

using magnetic forward modelling.

(3): Shallow bodies mask deeper sources (Jain, 1976). The high frequency

anomalies produced by shallow bodies tend to mask the interpretation of anomalies

caused by the deeper bodies (Jain, 1976; Ku and Sharp, 1983). Nevertheless, if the

deep source is displaced from the shallow body edges horizontally by a distance

greater than the depth to the top of deep body, the deeper bodies can still be

61

recognized (Jain, 1976; Radhakrishna Murthy et al., 2000). Moreover, if the anomalies

produced by deeper bodies are much stronger than those caused by the shallow

bodies, the top of the deeper bodies can also be located (Jain, 1976).

(4): This method can produce some erroneous solutions, due to noise, non-

uniqueness, and calculation window sizes poorly matching those needed for the

solutions (Cochran and Karner, 2007; Karner et al., 1991; Kilty, 1983; Ku and Sharp,

1983).

(5): If the horizontal distance between neighbouring magnetic bodies is less

than the depth to the top of the bodies, 2D Werner deconvolution has difficulty in

separating them (Hartman et al., 1971; Jain, 1976).

(6): This method produces unacceptable errors when applied to data collected

over varied altitudes (Ostrowski et al., 1993).

(7): This method of depth estimation is easily affected by the magnetized

minerals in the materials above the target basement.

There are many other magnetic interpretation techniques based on source

depth determination that could be used in the central Red Sea, for example, Euler

deconvolution (Thompson, 1982), source parameter imaging method (Thurston and

Smith, 1997), and analytical signal method (MacLeod et al., 1993). Li (2003) assessed

those methods. He suggested that no single method is best overall and an optimum

method should be selected according to the data quality and nature of the geological

problems. Compared to traditional magnetic forward and inverse modelling methods,

those methods are quick and have strong anti-noise properties, as they can well

isolate the magnetic anomaly from the noise (Li, 2003).

There are three reasons why we prefer to apply 2D Werner deconvolution in

this study. First, this method can be easily carried out on our high-quality magnetic

data. Second, it does not require any initial source model parameter values. Third,

comparisons of the Werner solutions with the results from other studies (see 5.5.1)

show the acceptable accuracy of this method.

62

3.2.4 Statistical analysis of the Werner solutions

Although the Werner method produced stable clusters of solutions, the results

are nevertheless noisy and cannot uniquely indicate the top of basement. To obtain

reliable magnetic basement depths, the Werner deconvolution was improved by

including the seismic reflection data as constraints.

Differences between the individual Werner source depths and the seismically

derived basement depths (ΔZ) (i.e., Werner source depth minus seismically derived

basement depth) were computed. Cumulative distribution functions (CDFs) of those

differences (ΔZ) were then derived. Since the primary magnetic source around the

spreading centre is expected to lie within the shallow basement from the

magnetizations of the extrusive basalts and sheeted dykes (Tivey and Dyment, 2010;

Tivey and Johnson, 1987), we first recorded the CDF level corresponding with ΔZ=0

(i.e., magnetic sources within the seismic basement) for the axial data. We then use

this CDF level to estimate depth to basement throughout the central Red Sea,

assuming that the whole region consists of similarly magnetized basement, with none

of the magnetic field originating in the overlying sediments. CDFs were computed

from magnetic sources within rectangular cells 20 km × 10 km along the survey lines.

This statistical method needs a large sample size (greater than 200 in this study)

to compute reliable CDF levels. Hence, the calculation cell sizes are large, resulting in

the low resolution of derived basement topography maps.

In Chapter 5, statistical analysis of the Werner solutions is used to derive

magnetic basement depths from the Werner source solutions.

3.2.5 Reduction to the pole (RTP)

Due to the geomagnetic inclination, magnetic anomalies observed anywhere

other than magnetic poles are asymmetric even when the distribution of causative

bodies is symmetric (Ansari and Alamdar, 2009). To remove the skewness of the

anomalies caused by inclination, reduction to the pole was used to adjust the

aeromagnetic data. The RTP magnetic anomalies is shown in Figure 3.8.

63

Figure 3.8 Aeromagnetic anomalies of Izzeldin (1982, 1987) reduced to pole. Red lines locate

the seismic reflection profiles 7, 9, 11, 15, 17, 19, 21, 25, 27, 29, 31 from Izzeldin (1982, 1987).

The data are contoured every 80 nT. The data were gridded and contoured using the GMT

software (Wessel et al., 2013).

The magnetic anomaly due to a localized spherical source can be expressed as

(Ravat, 2007):

∆T𝑚(r) = −∂

∂β∆Vφ(r) =

∂2

∂β ∂φ

∆J

r (3.7)

where r is distance between source and observation point, ∆Vφ is anomalous

potential due to the magnetization direction φ , ∆J is the intensity of anomalous

magnetization, β is the direction of the Earth’s main field, and ∆J

r is the anomalous

source function.

The RTP magnetic anomaly ∆Tz(r) is the vertical intensity anomaly due to

vertical magnetization, which can be computed by twice differentiating the

anomalous source function ∆J

r in the vertical direction (to obtain first the anomalous

64

potential due to the vertically magnetized source, and then its anomaly in the vertical

direction) as (Ravat, 2007):

∆Tz(r) =∂2

∂z2

∆J

r =

∂2

∂z2 ∫ ∫ ∆T𝑚(r) ∂β ∂φ+∞

−∞

+∞

−∞ (3.8)

where z represents the vertical direction.

At low latitudes, the magnetic RTP becomes unstable, resulting in severe

linear artefacts along the magnetic declination direction (Arkani-Hamed, 2006;

Keating and Zerbo, 1996). Additionally, this technique is seriously affected by high-

frequency noise, which can cause distortion of the RTP result (Zhang et al., 2018).

The reduced-to-pole aeromagnetic anomaly grid used in Chapter 6 (Figure 3.8)

was provided by Ian C.F. Stewart.

3.3 Gravity anomalies

3.3.1 Free-air gravity data

Version 23.1 of the marine gravity field (referred to as “SSv23”) used in this

study were derived from satellite altimetry measurements by Sandwell et al. (2014).

The data are shown in Figures 3.1 and 3.9a.

The SSv23 data are evaluated in Chapter 6, using shipboard gravity data

collected on the RRS Shackleton (Girdler and Southren, 1987) and on RV Robert

Conrad (Cochran and Martinez, 1988). The differences between the SSv23 and the

RRS Shackleton and RV Robert Conrad data have standard deviations of 5.5 and 3.7

mGal. Sandwell (pers. comm. 2013) suggested that these biases could be due to edge

effects from when the vertical offshore altimetry deflections were converting to

gravity anomalies (Mitchell, 2015). Nevertheless, the biases are small compared to

the >100 mGal full range of the SSv23 gravity anomalies.

65

Figure 3.9 (a): Free-air gravity anomalies (Sandwell et al., 2014, version 23.1) and locations

of multichannel seismic reflection profiles 7, 9, 11, 15, 17, 19, 21, 25, 27, 29, 31 from Izzeldin

(1982, 1987). The data were gridded using the GMT software (Wessel et al., 2013). (b):

Locations of gravity profiles G1-32 and seismic reflection profiles as in (a). Cyan line

approximately locates the spreading axis. Green line locates gravity profile G21, along which

the gravity forward modelling results are shown in Figure 6.9. In Chapter 6, the study area is

divided into two areas: north of 20.25°N and south of 20.25°N, as the free-air gravity field

north of 20.25°N has cross-axis lineaments suggested to be due to oceanic spreading

segments (Mitchell, 2015).

In Chapter 6, free-air gravity anomalies are sampled from SSv23 along the

ridge-parallel gravity profiles G1-32 shown in Figure 3.9. Because the free-air gravity

field has lineations crossing the Red Sea north of 20.25°N but not south of there

(Mitchell, 2015), the profiles are each divided into two segments either side of 20.25°N.

Those segmented gravity profiles are then inverted for basement depth variations,

which are used to estimate basement roughness.

3.3.2 Bouguer gravity anomalies

Gravity anomalies arise from density variations within the crust and upper

mantle, topography of the Moho and crust-evaporite interfaces, as well as seabed

topography. Mitchell et al. (2017) computed marine Bouguer anomalies of the central

Red Sea to remove the component of the gravity field due to the seabed topography

(Figure 3.10).

66

The minor errors in Bouguer gravity anomalies are likely to arise from the

thickness variations of Plio-Pleistocene sediments (Cowie and Karner, 1990; Mitchell

et al., 2017).

Figure 3.10 Bouguer gravity anomalies from Mitchell et al. (2017) computed by removing

the component of the free-air gravity field (Sandwell et al., 2014, version 23.1) due to the

seabed topography. The data are contoured every 50 mGal. The data were gridded and

contoured using the GMT software (Wessel et al., 2013).

In Chapter 4, the correlation between the marine Bouguer anomalies and the

basement depths is examined for evidence of variations in crustal thickness or density

or in mantle density. In regions of high correlation, we solve for the apparent density

contrast that best explains the observed gravity anomaly to see if it is consistent with

the expected density contrast between the mantle and the evaporites, assuming a

constant thickness crust.

In Chapter 5, the covariation between the marine Bouguer anomalies and

magnetic basement depths is used as a test of the magnetically derived depths.

67

3.3.3 Mantle Bouguer anomalies (MBAs)

The mantle Bouguer anomaly (MBA) is a gravity anomaly, obtained by

subtracting the gravitational attraction of a reference crust and sediments from the

free-air gravity anomaly, with a correction made for cooling of the lithosphere with

age (Grindlay et al., 1998; Lin et al., 1990; Rommevaux et al., 1994).

The MBAs vary with varied crustal thickness and/or upper mantle densities,

therefore, they can be used to investigate the upwelling structure of mantle beneath

spreading ridges (Magde and Sparks, 1997).

MBAs along 11 seismic profiles were computed by removing gravity effects of

evaporite-crust and crust-mantle interfaces from the marine Bouguer gravity

anomalies of Mitchell et al. (2017) (Figure 3.10), assuming a uniform 7 km thick crust.

Densities of 2,148 kg m-3 (Wheildon et al., 1974), 2,957 kg m-3 (Hyndman and Drury,

1977), and 3,220 kg m-3 (Crough, 1983; Gvirtzman et al., 2016) were used for

evaporites, oceanic crust, and hot mantle, respectively.

Because MBAs are derived from Bouguer gravity anomalies of Mitchell et al.

(2017), MBAs can also be affected by the thickness variations of Plio-Pleistocene

sediments.

In Chapter 6, the along-axis gradients in MBAs in the central Red Sea were

compared with those at other mid-ocean ridges. As the gradients were calculated for

the spreading axis, with zero age crust, no gravity anomaly correction for lithospheric

cooling component was needed.

3.3.4 Bouguer slab formula

The gravity anomaly caused by a layer of infinite lateral extent and constant

thickness ℎ and density contrast ∆𝜌 can be computed using the Bouguer slab formula:

𝛿𝑔 = 2𝜋𝐺ℎ∆𝜌 (3.9)

Equation (3.9) can be written as:

𝜕𝛿𝑔

𝜕ℎ= 2𝜋𝐺∆𝜌 (3.10)

68

In Chapter 4, equation (3.10) is used to compute apparent density contrast

from Bouguer-basement depth gradients.

In Chapter 6, equation (3.9) is modified as:

𝑔𝑓𝑎𝑎 = 2𝜋𝐺[ℎ𝑤(𝜌𝑤 − 𝜌𝑐) + 𝑡𝑒𝑠(𝜌𝑒 − 𝜌𝑐)] + 𝑐 (3.11)

where 𝑔𝑓𝑎𝑎 is free-air anomaly along the axis-parallel profiles, ℎ𝑤 is water depth, 𝑡𝑒𝑠

is total thickness of the evaporites and other sediments, 𝜌𝑤 and 𝜌𝑐 are water and

crustal densities, 𝜌𝑒 is the mean density of the evaporite and sediment layers, and 𝑐

is a constant along each ridge-parallel gravity profile.

The physical significance of 𝑐 is to derive 𝑡𝑒𝑠 along the gravity profile from

𝑔𝑓𝑎𝑎 , when other data (e.g., seismic reflection data) can provide 𝑡𝑒𝑠 at some measure

points as constraints.

A mean density of 2,148 kg m-3 is used for the evaporite and sediment layers

based on DSDP sample measurements of Wheildon et al. (1974). The crust was

assumed to have a density typical of oceanic crust dominated by gabbro. A density of

2,957 kg m-3 was used for the oceanic crust based on DSDP sample measurements of

Hyndman and Drury (1977). A 1,020 kg m-3 density was used for the seawater.

By rearranging equation (3.11), we obtain:

𝑡𝑒𝑠 =𝑔𝑓𝑎𝑎−𝑐

2𝜋𝐺(𝜌𝑒−𝜌𝑐)− ℎ𝑤

𝜌𝑤−𝜌𝑐

𝜌𝑒−𝜌𝑐 (3.12)

Hence, the basement depth is:

ℎ𝑏 = ℎ𝑤 + 𝑡𝑒𝑠 =𝑔𝑓𝑎𝑎−𝑐

2𝜋𝐺(𝜌𝑒−𝜌𝑐)+ ℎ𝑤

𝜌𝑒−𝜌𝑤

𝜌𝑒−𝜌𝑐 (3.13)

Although absolute basement depths cannot be calculated merely from gravity

anomalies using equation (3.13) because 𝑐 is unknown, the basement depths derived

from seismic data are used to determine c for each profile.

The basement depths computed from equation (3.13) are then used to

estimate basement roughness along ridge-parallel gravity profiles.

69

Basement relief derived using this technique will be underestimated, because

the assumption of an infinite slab is not fully met. Therefore, gravity forward

modelling is needed for assessing the bias magnitude.

3.3.5 2D gravity forward modelling

In Chapter 6, we have used 2D gravity forward modelling to assess the above-

mentioned bias magnitude, which can be used to correct the roughness values.

Forward models of free-air gravity anomalies were computed along profiles G1-32 by

summing the gravitational effects of elementary mass rectangular cells.

The vertical gravitational attraction of a small rectangular cell is (Shengye and

Yuling, 2004):

∆𝑔 = 𝐺𝜌[(𝑥 + 𝑎)𝑙𝑛(𝑥+𝑎)2+𝐻2

(𝑥+𝑎)2+ℎ2 − (𝑥 − 𝑎)𝑙𝑛(𝑥−𝑎)2+𝐻2

(𝑥−𝑎)2+ℎ2 + 2𝐻 (𝑡𝑔−1 𝑥+𝑎

𝐻− 𝑡𝑔−1 𝑥−𝑎

𝐻) −

2ℎ(𝑡𝑔−1 𝑥+𝑎

ℎ− 𝑡𝑔−1 𝑥−𝑎

ℎ)] (3.14)

where 2𝑎 is plate width, ℎ is depth to upper boundary, and 𝐻 is depth to lower

boundary.

Based on the principle of superposition, the vertical gravitational attraction of

a geological body can be reproduced by the sum of the attractions of many small

individual rectangular cells constituting the body (Blakely, 1996). To forward model

the gravity anomalies, the structure derived from the along-axis profiles were

subdivided into 𝑛 small cells and equation (3.13) was applied to compute the sum of

their gravitational attractions:

𝑔𝑘 = ∑(∆𝑔)𝑖

𝑛

𝑖=1

(3.15)

where 𝑔𝑘 is the vertical gravitational attraction measured at the 𝑘𝑡ℎ measurement

point, (∆𝑔)𝑖 is the vertical gravitational attraction produced for the 𝑘𝑡ℎ measurement

point by the 𝑖𝑡ℎ small rectangular cell, (∆𝑔)𝑖 is computed from equation (3.14).

MATLAB software was used to carry out the 2D gravity forward modelling,

based on equations (3.14) and (3.15).

70

The major uncertainty of the 2D gravity forward modelling originates from its

assumption of simple 2D models. The results derived from 2D analysis is less accurate

than those from the 3D analysis, since the observed gravity field are produced by 3D

objects. Moreover, the uncertainty in the gravity anomalies derived from the 2D

forward modelling can originate from the assumptions of flat Moho. Egloff et al. (1991)

suggested that Moho is not flat in the central Red Sea. Their seismic refraction profiles

shown the Moho varies in depth from 5 km to 10 km near the ridge axis.

We reproduced the seismic profile SO53-PIII of Egloff et al. (1991) using this

2D gravity forward modelling (Figure 4.9), although we did not perform this modelling

along the seismic lines from Izzeldin (1982, 1987).

3.4 Isostatic loading corrections

When assessing whether the geometry of crustal basement is typical of

oceanic crust, it is necessary to correct the observed basement depth for the effect of

loading by the overlying evaporites and sediment.

In Chapter 4, we have used a simple 1-D Airy isostatic model (Airy, 1855; Watts,

2001) in which the isostatic depression, ∆𝑧, is:

∆𝑧 =(𝜌𝑒𝑠−𝜌𝑤)

(𝜌𝑚−𝜌𝑤)𝑡𝑒𝑠 (3.16)

where 𝜌𝑒𝑠 is the mean density of the evaporite and sediment layers, 𝜌𝑚 and 𝜌𝑤 are

the densities of mantle and seawater, and 𝑡𝑒𝑠 is the total thickness of the evaporites

and other sediments.

A mean density of 2,148 kg m-3 is used for the evaporite and sediment layers

(Wheildon et al., 1974). A density of 3,220 kg m-3 was chosen for the hot mantle

(Crough, 1983; Gvirtzman et al., 2016). A 1,020 kg m-3 density was used for the

seawater.

Because the assumption of Airy isostasy ignores lithospheric rigidity, the deep

basement will be overcorrected and the shallow basement will be undercorrected,

compared with flexural isostasy (e.g., Watts, 2001). Thus, unloaded basement relief

71

will also be underestimated. For the relatively weak Red Sea, this difference should

be moderate.

3.5 Bathymetry data

3.5.1 Smith and Sandwell (1997) global topography dataset (Version 18.1)

Version 18.1 of the Smith and Sandwell (1997) bathymetry data (referred to as

“V18.1”) used in this study is shown in Figure 3.11. These data were derived by

combining shipboard depth measurements with depths inferred from satellite

altimetry of the sea surface.

In the region away from the coasts, the differences between the seabed

reflection and the bathymetry of V18.1 are less than 200 m (Figure 3.2). In the region

near to the coasts, these differences become larger than 800 m (Figure 3.2), indicating

a decrease in the accuracy of V18.1 data.

Figure 3.11 Bathymetry of the central Red Sea (Smith and Sandwell, 1997, version 18.1). Red

lines locate the seismic reflection profiles 7, 9, 11, 15, 17, 19, 21, 25, 27, 29, 31 from Izzeldin

(1982, 1987). The data are contoured every 500 m. The data were gridded and contoured

using the GMT software (Wessel et al., 2013).

72

In Chapters 4, 5, and 6, comparisons of the bathymetry sampled along the

seismic lines with depths derived from the seabed reflection are used to verify the

positions of the seismic profiles.

3.5.2 Multibeam sonar data

Augustin et al. (2014) interpreted multibeam sonar data from the central Red

Sea to reveal the pattern of evaporite flow.

These data only cover the region around the spreading centre.

In Chapter 6, we overlay their interpreted extents of evaporite and exposed

volcanic basement on the map of magnetic source depths to help assess how well the

more extensive flow corresponds with basement valleys and indirectly assess the

deconvolution results for geological consistency.

3.6 References

Airy, G. B., 1855, On the computation of the effect of the attraction of mountain-

masses, as disturbing the apparent astronomical latitude of stations in

geodetic surveys: Philosophical Transactions of the Royal Society of London, v.

145, p. 101-104.

Ansari, A., and K. Alamdar, 2009, Reduction to the pole of magnetic anomalies using

analytic signal: World Applied Sciences Journal, v. 7, p. 405-409.

Arkani-Hamed, J., 2006, Differential reduction to the pole: revisited: Geophysics, v. 72,

p. L13-L20.




Blakely, R. J., 1996, Potential theory in gravity and magnetic applications, Cambridge

University Press.




73



Cowie, P. A., and Karner, G. D., 1990, Gravity effect of sediment compaction: examples

from the North Sea and the Rhine Graben: Earth and Planetary Science Letters,

v. 99, p. 141-153.

Crough, S. T., 1983, The correction for sediment loading on the seafloor: Journal of

Geophysical Research: Solid Earth, v. 88, p. 6449-6454.

Davy, B., and R. Wood, 1994, Gravity and magnetic modelling of the Hikurangi Plateau:

Marine geology, v. 118, p. 139-151.




v. 198, p. 329-353.




Girdler, R., and T. Southren, 1987, Structure and evolution of the northern Red Sea.





Grindlay, N. R., J. A. Madsen, C. Rommevaux-Jestin, and J. Sclater, 1998, A different

pattern of ridge segmentation and mantle Bouguer gravity anomalies along the

ultra-slow spreading Southwest Indian Ridge (15° 30′ E to 25° E): Earth

and Planetary Science Letters, v. 161, p. 243-253.

Gvirtzman, Z., C. Faccenna, and T. W. Becker, 2016, Isostasy, flexure, and dynamic

topography: Tectonophysics, v. 683, p. 255-271.



12267-12279.

Hall, S. A., G. E. Andreasen, and R. W. Girdler, 1977, Total-intensity magnetic anomaly

map of the Red Sea adjacent coastal areas, a description and preliminary

74

interpretation. In: Red Sea Research 1970-1975. Saudi Arabian Directorate

General of Mineral Resources Bulletin, 22, F1-F15.

Hansen, R., 2005, 3D multiple-source Werner deconvolution for magnetic data:

Geophysics, v. 70, p. L45-L51.

Hansen, R. O., and M. Simmonds, 1993, Multiple-source Werner deconvolution:

Geophysics, v. 58, p. 1792-1800.

Hartman, R. R., D. J. Teskey, and J. L. Friedberg, 1971, A system for rapid digital

aeromagnetic interpretation: Geophysics, v. 36, p. 891-918.

Hassan, H. H., R. A. Charters, and J. W. Peirce, 2007, Mapping Depth to Basement Using

2D Werner Inversion of High Resolution Aeromagnetic (HRAM) Data: CSPG-

CSEG, Expanded Abstracts, p. 222-225.

Hyndman, R., and M. Drury, 1977, Physical properties of basalts, gabbros, and

ultramafic rocks from DSDP Leg 37: Aumento, F., Melson, WG, et al., Init. Repts.

DSDP, v. 37, p. 395-401.


Strasbourg.




Jain, S., 1976, An automatic method of direct interpretation of magnetic profiles:

Geophysics, v. 41, p. 531-541.

Karner, G., N. Driscoll, and J. Peirce, 1991, Gravity and magnetic signature of Broken

Ridge, southeast Indian Ocean. Proceedings of the Ocean Drilling Program,

Scientific Results, p. 681-695.

Keating, P., and L. Zerbo, 1996, An improved technique for reduction to the pole at

low latitudes: Geophysics, v. 61, p. 131-137.

Kilty, K. T., 1983, Werner deconvolution of profile potential field data: Geophysics, v.

48, p. 234-237.

Ku, C. C., and J. A. Sharp, 1983, Werner deconvolution for automated magnetic

interpretation and its refinement using Marquardt's inverse modeling:


75

Li, X., 2003, On the use of different methods for estimating magnetic depth: The

Leading Edge, v. 22, p. 1090-1099.

Lin, J., G. Purdy, H. Schouten, J. Sempere, and C. Zervas, 1990, Evidence from gravity

data for focused magmatic accretion along the Mid-Atlantic Ridge: Nature, v.

344, p. 627-632.

Mackenzie, K. V., 1981, Discussion of sea water sound ‐ speed determinations:

Journal of the Acoustical Society of America, v. 70, p. 801-806.

MacLeod, I. N., K. Jones, and T. F. Dai, 1993, 3-D analytic signal in the interpretation of

total magnetic field data at low magnetic latitudes: Exploration Geophysics, v.

24, p. 679-688.

Magde, L. S., and D. W. Sparks, 1997, Three‐dimensional mantle upwelling, melt

generation, and melt migration beneath segment slow spreading ridges:

Journal of Geophysical Research: Solid Earth, v. 102, p. 20571-20583.

Marquardt, D. W., 1963, An algorithm for least-squares estimation of nonlinear

parameters: Journal of the Society for Industrial and Applied Mathematics, v.

11, p. 431-441.

Martelet, G., J. Perrin, C. Truffert, and J. Deparis, 2013, Fast mapping of magnetic

basement depth, structure and nature using aeromagnetic and gravity data:

combined methods and their application in the Paris Basin: Geophysical

prospecting, v. 61, p. 857-873.

Mitchell, N. C., 2015, Lineaments in gravity data of the Red Sea. In: Rasul, N.M.A., and

Stewart, I.C.F. (Eds.), The Red Sea: The formation, morphology, oceanography

and environment of a young ocean basin, Springer Earth System Sciences,

Berlin Heidelberg, p. 123-133.



v. 29, p. 352-369.

Ostrowski, J. S., M. Pilkington, and D. J. Teskey, 1993, Werner deconvolution for

variable altitude aeromagnetic data: Geophysics, v. 58, p. 1481-1490.

Radhakrishna Murthy, I., K. Swamy, and P. Rama Rao, 2000, Use and abuse of Werner

deconvolution technique: J. Ind. Geophys. Union, v. 4, p. 97-102.

76

Ravat, D., 2007, Reduction to Pole, in D. Gubbins, and E. Herrero-Bervera, eds.,

Encyclopedia of Geomagnetism and Paleomagnetism: Dordrecht, Springer

Netherlands, p. 856-858.


Jahrbuch, v. 13, p. 131-153.

Rommevaux, C., C. Deplus, P. Patriat, and J. C. Sempéré, 1994, Three‐dimensional

gravity study of the Mid‐Atlantic Ridge: Evolution of the segmentation

between 28° and 29 °N during the last 10 m.y.: Journal of Geophysical





Saudi-Sudanese Red Sea Commission, 1976, Aeromagnetic survey.

Shengye, Z., and P. Yuling, 2004, The principle of applied geophysics, China University

of Geosciences Press.



Stagg, H., J. Willcox, and D. Needham, 1989, Werner deconvolution of magnetic data:

Theoretical models and application to the Great Australian Bight: Journal of

the Geological Society of Australia, v. 36, p. 109-122.

Subrahmanyam, M., and F. T. Gebissa, 2017, Performance evaluation of spectral

analysis and Werner deconvolution interpretation techniques in magnetic

method: Journal of Applied Geophysics, v. 138, p. 102-113.

Thakur, N., T. G. Rao, R. Khanna, and C. Subrahmanyam, 2000, Magnetic basement in

the Bay of Bengal through Werner deconvolution: Marine Geology, v. 162, p.

599-605.

Thompson, D., 1982, EULDPH: A new technique for making computer-assisted depth

estimates from magnetic data: Geophysics, v. 47, p. 31-37.

Thurston, J. B., and R. S. Smith, 1997, Automatic conversion of magnetic data to depth,

dip, and susceptibility contrast using the SPI (TM) method: Geophysics, v. 62,

p. 807-813.

77

Tivey, M. A., and J. Dyment, 2010, The magnetic signature of hydrothermal systems in

slow spreading environments. In: Rona, P.A., Devey, C.W., Dyment, J., Murton,

B.J. (Eds.) Diversity of Hydrothermal Systems on Slow Spreading Ocean Ridges,

American Geophysical Union, Geophysical Monograph Series, v. 188, p. 43-66.

Tivey, M. A., and H. P. Johnson, 1987, The central anomaly magnetic high: Implications

for ocean crust construction and evolution: Journal of Geophysical Research:

Solid Earth, v. 92, p. 12685-12694.

Tramontini, C., and D. Davies, 1969, A seismic refraction survey in the Red Sea:


Watts, A. B., 2001, Isostasy and Flexure of the Lithosphere: UK, Cambridge University

Press, 1-284 p.

Werner, S., 1953, Interpretation of magnetic anomalies at sheet-like bodies: Sveriges

Geol: Undersok. Ser. C. Arsbok, v. 43, p. 1949.



Union, v. 94, p. 409-410.







Zhang, Q., Y. Zhang, G. Yin, and Z. Li, 2018, An improved frequency-domain algorithm

for stable reduction to the pole at low latitudes: Journal of Geophysics and

Engineering, v. 15, p. 1767-1782.

78

Chapter 4.

Paper 1: Oceanic-like axial crustal high in the

central Red Sea

79

4. Paper 1: Oceanic-like axial crustal high in the central

Red Sea

Wen Shi1*, Neil C. Mitchell1, Lara M. Kalnins2, A.Y. Izzeldin3

1School of Earth and Environmental Sciences, The University of Manchester,

Manchester M13 9PL, UK.

2School of GeoSciences, The University of Edinburgh, The King’s Buildings, Edinburgh,

EH9 3FE, UK.

3Awasconrc, Gereif W, H4, Bld 376, Khartoum, POB 410, Khartoum, Sudan.

* Corresponding author.

E-mail address: [email protected] (Wen Shi)

This paper is published in the journal Tectonophysics.



80

Abstract

The Red Sea is an important example of a rifted continental shield proceeding

to seafloor spreading. However, whether the crust in the central Red Sea is

continental or oceanic has been controversial. Contributing to this debate, we assess

the basement geometry using seismic reflection and potential field data. We find

that the basement topography from seismically derived structure corrected for

evaporite and other sediment loading has an axial high with a width of 70-100 km and

a height of 0.8-1.6 km. Basement axial highs are commonly found at mid-ocean ridges

affected by hotspots, where enhanced mantle melting results in thickened crust. We

therefore interpret this axial high as oceanic-like, potentially produced by recently

enhanced melting associated with the broader Afar mantle anomaly. We also find the

Bouguer gravity anomalies are strongly correlated with basement reflection depths.

The apparent density contrast necessary to explain the Bouguer anomaly varies from

220 kg m-3 to 580 kg m-3 with no trend with latitude. These values are too small to be

caused primarily by the density contrast between evaporites and mantle across a crust

of uniform thickness and density structure, further supporting a thickened crustal

origin for the axial high. Complicating interpretation, only a normal to modestly

thickened axial crust is predicted from fractionation-corrected sodium contents (Na8.0),

and the basement reflection is rugged, more typical of ultra-slow spreading ridges that

are not close to hotspots. We try to reconcile these observations with recent results

from seismic tomography, which show modest mantle S-wave velocity anomalies

under this part of the Red Sea.

Keywords: Red Sea, Ocean–continent transition, Oceanic crust, Seismic reflection,

Potential field, Subsalt

81

4.1 Introduction

The Red Sea is a young ocean basin transitioning from continental extension

to seafloor spreading (e.g., Bonatti et al., 1981; Cochran and Martinez, 1988; Rihm and

Henke, 1998). However, how far the central Red Sea (Figure 4.1) is through this

transition to full seafloor spreading has been debated.

Figure 4.1 Bathymetry of the Red Sea (Smith and Sandwell, 1997, version 18.1). Red dots

locate the prominent deeps in the central Red Sea from Augustin et al. (2014) and Karbe

(1987). From north to south, these are (1) Nereus, (2) Thetis, (3) Hadarba, (4) Hatiba, (5)

Atlantis II, (6) Erba, (7) Port Sudan, (8) Suakin, and (9) Pelagia deeps. Green dot marks the

Farasan Islands. The relative plate motion vectors were predicted using the Chu and Gordon

(1998) plate rotation pole.

82

Bonatti (1985) suggested the central Red Sea is just at the point of transitioning

from continental rifting to oceanic spreading. He proposed that the ‘deeps’ found in

the central Red Sea are discrete seafloor spreading cells based on the presence of

normal mid-ocean ridge basalt (MORB) sampled from them and their high amplitude

magnetic anomalies. This contrasts with low amplitude anomalies outside the deeps,

which were therefore assumed to overlie stretched continental crust (Ligi et al., 2011,

2012).

However, other evidence could support the interpretation of more established

seafloor spreading in the central Red Sea. An extensive aeromagnetic survey revealed

that there are low amplitude magnetic anomalies outside the ‘deeps’ aligned parallel

to the ridge axis (Izzeldin, 1987; Rasul et al., 2015). LaBrecque and Zitellini (1985)

showed with numerical modelling that such subdued anomalies could be produced by

widely distributed dykes, lava flows, and sills, as occur in modern-day Afar. Low

amplitudes may also have arisen from the slow spreading rate, the greater depth of

basement and alteration under the evaporites (Augustin et al., 2014; Dyment et al.,

2013; Izzeldin, 1987, 1989; Levi and Riddihough, 1986; Mitchell and Park, 2014). The

‘deeps’ are separated by inter-trough zones, which are shallower, lacking in strong

magnetic anomalies compared to the ‘deeps’, and covered by evaporites that have

flowed laterally and sediments across the axis. Based on seismic reflection, gravity

and magnetic data, Izzeldin (1982, 1987) suggested the inter-trough zones are

manifestations of less organized seafloor spreading and underlain by oceanic crust.

Using multibeam sonar data, Augustin et al. (2014, 2016) also interpreted these zones

as merely areas where the off-axis evaporites have flowed into the axis, obscuring the

volcanic geomorphology. Seismic refraction data show the velocity of basement

under the inter-trough zone between Nereus and Thetis deeps (Figure 4.1) is 6.86 km

s-1 (Davies and Tramontini, 1970; Tramontini and Davies, 1969), which overlaps with

velocities of oceanic crust elsewhere (6.7-6.9 km s-1; Carlson, 2001, 2010). Further

seismic refraction data collected by Egloff et al. (1991) along line PIII (Figure 4.2) also

suggested the basement around the ridge axis is oceanic, which transitions to

stretched continental crustal velocities towards the coast of Sudan. Finally, free-air

gravity anomalies derived from satellite altimeter data (Sandwell et al., 2014; Sandwell

83

and Smith, 2009) reveal that anomalies along the spreading centre are segmented

(Figure 4.2). Mitchell and Park (2014) and Augustin et al. (2016) suggested that this

segmented pattern is similar to the segmentation observed at slow-spreading mid-

ocean ridges elsewhere (e.g., the northern Mid-Atlantic Ridge) (Schouten et al., 1987;

Sempéré et al., 1990). The rugosity of basement computed from these anomalies is

similar to that of the similarly slow-spreading Mid-Atlantic Ridge (Shi et al., 2017).

Figure 4.2 Free-air gravity anomalies (Sandwell et al., 2014, version 23.1) and locations of

multichannel seismic reflection profiles 7, 9, 11, 15, 17, 19, 21, 25, 27, 29, 31 of Izzeldin (1987)

and seismic refraction profile SO53-PIII of Egloff et al. (1991). Purple curves show the

segmentation of gravity anomalies in the centre of the basin.

Based on seismic reflection and potential field data, Izzeldin (1982, 1987)

suggested that intermediate crust separates crust that is clearly continental near the

coasts from that which is clearly oceanic around the axis. This area lies ~65-160 km

from the ridge axis.

84

Young ocean basins such as the Red Sea offer opportunities to explore mantle

and crustal processes at the transition from rifting to seafloor spreading. In particular,

the rate and geometry of deformation may be important for how rapidly the rising

mantle cools during this phase. This in turn affects the flux of melt produced by mantle

decompression and where oceanic crust is first emplaced. Buck (1986) suggested that

lateral temperature gradients in the asthenosphere produced by rifting could lead to

more rapid upwelling, implying a greater initial melt flux. Alternatively, the surface

rifting and lithospheric mantle thinning could be laterally offset (Hopper and Roger

Buck, 1998), implying a different location of initial magma emplacement. Some of

these and other issues affecting the initial melting are illustrated in the recent

numerical geodynamic models of Harry and Bowling (1999), Corti et al. (2003),

Fletcher et al. (2009), Jeanniot et al. (2016), Ros et al. (2017), and Armitage et al.

(2018). However, the evidence needed to investigate these ideas from basins

presently transitioning to seafloor spreading is limited as examples are rare and often

complicated. For example, the Woodlark Basin is small, opening relatively fast (~60

mm yr-1) and in a complicated tectonic setting that is still evolving rapidly after the

Ontong Java collision with the West Melanesian Trench (Martinez et al., 1999; Weissel

et al., 1982). The Gulf of California rift is opening highly obliquely (e.g., Atwater and

Stock, 1998; Lonsdale, 1989; Withjack and Jamison, 1986). The Red Sea, in contrast,

is opening slowly (~10-~16 mm yr-1) and more nearly orthogonally (e.g., Chu and

Gordon, 1998), so it provides an important example of mantle and crustal dynamics

of slow orthogonal rifting.

In the present study, we use 11 lines of industrial seismic reflection data from

the central Red Sea reported in Izzeldin (1982, 1987). We verify interpreted basement

depths using Werner deconvolution applied to magnetic anomalies and then correct

those depths for isostatic loading by the evaporites and other sediments. The

basement geometry is found to reveal axial highs similar in gross morphology to,

though larger than, those of the Reykjanes Ridge, with more rapid deepening with

distance from the axis that cannot be explained by simple subsidence. This leads us

to favour an oceanic interpretation for the crust here, in which melt production has

recently increased, creating thicker crust which forms the axial high. We then discuss

85

what these results imply about the evolution of this section of the Red Sea rift and

broader implications.

4.2 Tectonic setting

4.2.1 Continental rifting and seafloor spreading in the northern and southern Red

Sea

The Red Sea opening rate increases southward from ~10 mm yr−1 at 25.5°N to

~16 mm yr−1 near 18°N with increasing distance from the Nubia/Africa spreading pole,

which lies in the Mediterranean (e.g., Chu and Gordon, 1998; DeMets et al., 1990;

DeMets et al., 2010). The extension of the Red Sea may have begun in the Eocene but

became more established in the Oligocene, associated with massive and rapidly

erupted basalts in Ethiopia and southern Yemen at approximately 30 Ma (Bosworth

and McClay, 2001; Hofmann et al., 1997; Mohr, 1983; Omar and Steckler, 1995).

These rapid eruptions have been attributed to the Afar plume penetrating the

lithosphere (Furman et al., 2006; George et al., 1998; Richards et al., 1989).

The northern Red Sea, which is closer to the pole of opening and has

experienced less extension than our study area (Figure 4.1), has been thought to be

underlain by continental crust with a series of large crustal fault blocks interpreted

from seismic velocity data and from magnetic and gravity anomalies (Cochran and

Karner, 2007; Gaulier et al., 1986; Martinez and Cochran, 1988). If correct, this area

may still be in late stage continental rifting (e.g., Cochran, 1983; Gaulier et al., 1988;

Martinez and Cochran, 1988). In contrast, others have interpreted this region as

underlain by oceanic crust on the basis of unpublished seismic and magnetic data (e.g.,

Dyment et al., 2013; Tapponnier et al., 2013). Using remote sensing, geochemical, and

geochronological data, Sultan et al. (1992) carried out a plate reconstruction for the

opening of the Red Sea and found a best match of pre-existing African and Arabian

geologic features by juxtaposing present Red Sea coastlines. This has been interpreted

as indicating that the entire Red Sea basin is underlain by oceanic crust (Bosworth et

al., 1993). In addition, a few ‘deeps’ containing basalts are revealed in the northern

Red Sea (e.g., Bonatti, 1985; Guennoc et al., 1988; Ligi et al., 2018; Pautot et al., 1984).

86

In the southern Red Sea, farther from the pole than our study area (Figure 4.1),

seafloor spreading magnetic anomalies are clearly identifiable up to Chron 3 near the

axial trough between 16°N and 19°N, suggesting that recognizable seafloor spreading

began at least by 5 Ma (e.g., Cochran, 1983; Girdler and Styles, 1974; Phillips, 1970;

Vine, 1966). Augustin et al. (2016) suggested that oceanic spreading likely began

somewhat earlier, 8-12 Ma, based on spreading rates of Chu and Gordon (1998) and

locations of volcanic ridges interpreted from multibeam sonar and vertical gravity

gradient data. These are consistent with the spreading ages (8–12 Ma) discussed by

Izzeldin (1987) and Augustin et al. (2014). In addition to the clear anomalies in the

centre of the basin, Girdler and Styles (1974) and Hall (1989) also interpreted the low

magnetic anomalies over the southern Red Sea shelves as seafloor spreading magnetic

stripes, suggesting the Red Sea was formed by two stages of seafloor spreading. Based

on magnetic and gravity modelling constrained by the seismic refraction data of

Gettings et al. (1986) and Mooney et al. (1985), Almalki et al. (2014) recently

suggested that about 75 km of oceanic crust formed before Middle to Late Miocene

(15–5 Ma) under the Farasan Bank (Figure 4.1), which supports a two-stage spreading

evolution of the Red Sea. In contrast, Cochran (1983) argued that these magnetic

anomalies result from a wide region of mafic diking and intrusions rather than a

continuous oceanic crust of dykes and extrusives, because the anomalies have low

amplitudes (less than 200 nT) and long wavelengths (20-50 km). A seismic refraction

line shot across the Yemen margin by Egloff et al. (1991) was interpreted as showing

that oceanic crust adjacent to the axis terminates southward of 16°N and that

continental-type crust lies farther south as far as 14˚N.

4.2.2 Seismic tomographic studies encompassing the Red Sea

Seismic tomographic studies have found that S- and P-wave seismic velocities

of the upper mantle adjacent to the Red Sea increase by up to a few percent from

south to north with increasing distance from the Afar plume. Using body wave travel

time tomography, Park et al. (2007) found a -1.5% S-wave velocity anomaly at 200 km

depth beneath the coast of the southern Red Sea, rising to ~-1% in the central Red Sea

and to 0% or more in the northern Red Sea. A similar structure was found by Park et

al. (2008) from Rayleigh wave tomography although with a more subdued northwards

87

increase in S-wave velocity. They suggested this structure is caused by an upwelling

of warm mantle beneath the southern Arabian shield, originating from the Afar

hotspot. They proposed that this hot plume material flows from Afar underneath the

southern and central Red Sea, and then extends northwards beneath Arabia, whereas

the northern Red Sea (north of ~23°N) is without underlying hot mantle. Shear wave

splitting directions from Hansen et al. (2006) indicate that the hot mantle flow moves

northwards rather than parallel to the Red Sea. This is supported by the azimuthal

anisotropy analysis of Sebai et al. (2006). The upper mantle under the Red Sea is

poorly resolved in such models as they are mainly based on teleseismic recordings

along the coast of Saudi Arabia (no offshore recordings). Nevertheless, this general

pattern of upper mantle structure is corroborated by Na8.0 analyses of axial lavas

(sodium oxide concentrations corrected for fractionation (Klein and Langmuir, 1987)),

which indicate the upper mantle temperature in the Red Sea generally decreases by

about 60°C from 18°N to 26°N (Haase et al., 2000).

More recently, Chang et al. (2011) have carried out an inversion of seismic

travel times and waveforms that provides a more complete coverage of the area

under the Red Sea. They found low velocity (hot) material is located beneath the

southern Red Sea and Gulf of Aden, consistent with active seafloor spreading. They

also suggested that the hot material at a depth of ~150 km does not extend north-

westwards below the central and northern Red Sea areas, but forms a channel

extending northward beneath Arabia. The comparative high velocities under the

central Red Sea coincide with our data and, as we show later, this could be important

for interpretation of our results.

4.3 Data and methods

4.3.1 Seismic reflection

The multi-channel seismic reflection survey was carried out by a geophysical

services company in 1976 (Izzeldin, 1982). The data were collected using a Vaporchoc

source with a streamer consisting of 48 channels 50 m apart for the deep-water survey

(2.4 km streamer), and of 24 channels 50 m apart for the shallow-water survey (1.2

km streamer), positioned using a local radio navigation system. The data were

88

processed (24-fold), with semblance analysis providing interval velocity every 3.6 km

along-track, and moveout corrected. The locations of seismic reflection profiles 7, 9,

11, 15, 17, 19, 21, 25, 27, 29, and 31 used in this study are shown in Figure 4.2.

Two-way travel times for the basement and seabed reflections were converted

to depths below sea level (Figure 4.4) as follows. A P-wave velocity (Vp) of 1.538 km s-

1 was used for the water according to the empirical equations of Mackenzie (1981),

with typical Red Sea salinity of 40 ppt and temperature of 21°C. A 1.9 km/s Vp for the

Plio-Pleistocene sediments was chosen based on the measurements of DSDP Leg 23

samples (Whitmarsh et al., 1974) and the results of seismic reflection and refraction

surveys by Egloff et al. (1991) and Gaulier et al. (1988). A Vp of 4.21 km/s was used for

the evaporites, based on seismic refraction data from Tramontini and Davies (1969),

Girdler and Whitmarsh (1974), and Egloff et al. (1991).

4.3.2 Magnetic anomalies

4.3.2.1 Sources of magnetic data

Marine magnetic field measurements from towed magnetometers were

obtained from the National Centers for Environmental Information (NCEI)

(www.ngdc.noaa.gov/mgg). The data comprise residual magnetic anomalies after

removal of the international geomagnetic reference field (IGRF) from the total field

measurements. Figures 4.3a and 4.3c show the survey locations and contoured

anomalies after further adjustments to correct IGRF errors of the individual surveys

(see figure caption).

Major causes of magnetic anomalies are expected to be susceptibility and

remanent magnetization variations within the basement produced by intrusive or

extrusive volcanic bodies. To investigate possible magnetic sources, Werner

deconvolution was applied to derive the magnetic source depths and apparent

susceptibilities, which were calculated along individual segments of the magnetic lines

(ungridded magnetic data) where they cross the seismic reflection profiles of Izzeldin

(1987) (Figure 4.3b), and then were projected to the seismic profiles.

89

Figure 4.3 (a): Tracks of shipboard magnetic surveys of RVs Jean Charcot (78005111,

83008011), Atlantis (A2093L19), Chain (CH043L01, CH043L03, CH061L02, CH100L03),

Discovery (DI103B), Glomar Challenger (DSDP23GC), Melville (INMD09MV), Robert Conrad

(RC0911A), Shackleton (SHA1079) and Vema (V1413). (b): Extents of magnetic lines (blue)

contributing to the seismic profiles (red) of Werner source depths. (c): Residual magnetic

anomalies of the surveys in (a) obtained from the National Centers for Environmental

Information (NCEI) (www.ngdc.noaa.gov/mgg) gridded and contoured every 50 nT. To reduce

effects of reference field errors, the residual anomalies of each survey were adjusted by

subtracting their mean value before gridding and contouring. (Anomalies are not reduced to

the pole.) (d): Bouguer gravity anomalies from Mitchell et al. (2017) computed by removing


seabed topography.

90

4.3.2.2 Werner deconvolution

Werner deconvolution is an inverse method that is used to solve for magnetic

source parameters (e.g., depth and susceptibility) from the observed magnetic field

assuming that the sources comprise thin sheet-like bodies of semi-infinite extent

(Werner, 1953). The total field from a thin sheet-like body is equal to the horizontal

gradient of the total field caused by the edge of a thick body. Werner deconvolution

exploits this idea to estimate likely parameters of dykes and other layered structures

(Ku and Sharp, 1983). Although individual depth values derived using the method

have large uncertainties and the method can produce some erroneous solutions, their

depths have been shown generally to cluster within basement (Cochran and Karner,

2007; Karner et al., 1991).

The total magnetic anomaly caused by a dike or other tabular body is given as

(Ku and Sharp, 1983):

(4.1)

where , , is

horizontal position of the top centre of the dike, is depth to the top of the dike,

is thickness of the dike ( ), is the vector sum of induced and

remanent magnetization, is magnetic inclination of the main field , and is

strike of the body measured counterclockwise from magnetic north.

Interference from neighbour anomalies or regional trends is incorporated in the form

of a polynomial (Ku and Sharp, 1983):

(4.2)

where are interference terms.

By rearranging equation (4.2), an inversion equation is obtained (Ku and

Sharp, 1983; Rao, 1984):

0

2 2

0

( )( ,0)=

( )mag

A x x BDT x

x x D

− +

− +

=-2 ( sin cos sin )x zA T J I J + =2 ( - cos sin sin )x zB T J I J + 0x

D

2 T T D =( , )x zJ JJs

I F

200 1 22 2

0

( )( ,0)=

( )mag

A x x BDT x C C x C x

x x D

− ++ + +

− +

2

0 1 2C C x C x+ +

91

(4.3)

where

and

A seven-point Werner operator was applied to construct seven simultaneous

equations for inversion equation (4.3), with a sample spacing of . Then, we obtain

the following results for the thin dike (Ku and Sharp, 1983):


(4.4a)

Depth:

(4.4b)

Magnetic susceptibility: (4.4c)

Marquardt’s (1963) non-linear least-squares best-fit method was used to solve

the simultaneous equations, producing the estimates of magnetic source depth and

susceptibility shown in Figure 4.4.

Magnetic source depths were estimated from the magnetic anomalies where

the sources lay less than 5 km from the seismic lines. If the magnetic bodies

recognized by Werner deconvolution are real, the depth estimates should define

either the upper boundaries of dykes or the edges of other causative bodies, so

magnetic source solutions tend to be tightly grouped vertically beneath the true

locations of the causative bodies (Cochran and Karner, 2007; Karner et al., 1991; Ku

2 3 4 2

0 1 2 3 4 0 1mag mag maga a x a x a x a x b T b xT x T+ + + + + + =

2 2

0 0 0 0 0

2 2

1 0 0 1 1 0

2 2

2 0 1 0 2 2 0

3 1 2 0

4 2

2 2

0 0

2 ,

2 ,

2 ,

,

,

a Ax BD C D x C

a A C x C D C x

a C C x C D C x

a C C x

a C

b x D

= − + + +

= − + +

= − + +

= −

=

= − −

1 02b x=

x

1X=0.5 b x x +

2

0 1= 0.25 Y b b x− −

2 2

=x z

m

J Jx

+

F

92

and Sharp, 1983). Therefore, the upper clusters of Werner solutions were interpreted

as the top of the magnetic basement.

4.3.3 Bathymetry data

We have used version 18.1 of the Smith and Sandwell (1997) bathymetry grid,

which combines shipboard depth measurements with depths inferred from satellite

altimetry of the sea surface. These data are shown in Figure 4.1. Comparisons of the

bathymetry sampled along the seismic lines with depths derived from the seabed

reflection were used to verify the positions of the seismic profiles.

4.3.4 Isostatic loading corrections

When assessing whether the geometry of crustal basement is typical of

oceanic crust, it is necessary to correct the observed basement depth for the effect of

loading by the overlying evaporites and sediment. We have used a simple 1-D Airy

isostatic model (Airy, 1855; Watts, 2001) in which the isostatic depression, ∆z, is:

(4.5)

where ρes is the mean density of the evaporite and sediment layers, ρm and ρw are the

densities of mantle and seawater, and tes is the total thickness of the evaporites and

other sediments. A mean density of 2,148 kg m-3 was used for the evaporite and

sediment layers based on DSDP sample measurements of Wheildon et al. (1974). A

density of 3,220 kg m-3 was chosen for the hot mantle (Crough, 1983; Gvirtzman et al.,

2016). A 1,020 kg m-3 density was used for the seawater. Reversing isostatic

depression, which was typically 1-2 km, produced the profiles shown in Figure 4.5a.

It was not possible to backstrip fully these sediments due to lack of detailed

stratigraphic data, but industry well data show the evaporites were deposited from

~15 Ma, at the start of the Middle Miocene (Hughes and Beydoun, 1992) to ~5.3 Ma,

at the end of the Miocene. This corresponds to times of active rifting, and continental

rifts are typically weak, with a low effective elastic thickness (Te) of 5-15 km (Watts

and Burov, 2003). Young, slow-spreading oceanic lithosphere is also typically weak,

with Te < 13 km and commonly Te < 5 km (Cochran, 1979; Kalnins, 2011). The

( )es wes

m w

z t

− =

−

93

assumption of Airy isostasy, ignoring lithospheric rigidity, will lead to overcorrected

deep basement and undercorrected shallow basement compared with flexural

isostasy (e.g., Watts, 2001); unloaded basement relief will thus also be

underestimated. For the relatively weak Red Sea, this difference should be moderate.

For a basin of comparable scale, Davison et al. (2012) estimated 0.5 km of isostatic

overcorrection of their deepest basement for a Te of 5 km.

To reveal the systematic trend of basement deepening with distance away

from the axis, both western and eastern sides of the unloaded basement depth

profiles were plotted together by offsetting each segment to their average axial depth

of 1.69 km (Figure 4.5d). To help assess whether the crust is oceanic, we compare the

observed subsidence with the global average oceanic crust subsidence curve (blue

solid line in Figure 4.5d) from Crosby and McKenzie (2009) using the Chron 2A to

present spreading rates of Chu and Gordon (1998). In doing so, we assume that Red

Sea opening prior to Chron 2A occurred with a similar opening pole and rate. Besides

some offsets of dated features along the Dead Sea transform fault (e.g., Barjous and

Mikbel, 1990; Garfunkel, 1981; Garfunkel et al., 1974) there are unfortunately no

independent measures of Nubia-Arabia motion to confirm this unequivocally.

However, other data from the Gulf of Aden at 14°N, 52°E show continuous spreading,

with opening rate decreasing from ~30 mm year-1 at 15-17.5 Ma to ~20 mm year-1 at

10 Ma and then remaining constant to the present (Fournier et al., 2010).


Gravity anomalies arise from density variations within the crust and upper

mantle, as well as topography on the seabed, crust-evaporite, and Moho interfaces.

Mitchell et al. (2017) computed marine Bouguer anomalies of the central Red Sea to

remove the component of the gravity field due to the seabed topography.

We examine the correlation between the marine Bouguer anomalies and the

basement depths for evidence of variations in crustal thickness or density or in mantle

density. In regions of high correlation, we solve for the apparent density contrast (Δρ)

that best explains the observed gravity anomaly (Figure 4.7) to see if it is consistent

94

with the expected density contrast between the mantle and the evaporites, assuming

a constant thickness crust.

Apparent densities were derived from Bouguer-basement depth gradients

by inverting the equation derived from the gravity slab formula:

(4.6)

where is the universal constant of gravitation. The gradients were obtained

by least-squares regression for data within 60 km of the axis (regions of high

correlation). Using equation (4.6) ignores effects of upward continuation; we explore

these potential inaccuracies in section 4.4.3.

4.4 Results

4.4.1 Character of basement and seabed derived from seismic reflection profiles

In Figure 4.4, the seabed in all the seismic profiles forms an axial trough within

~20 km of the axis. The average depth of the axial trough shallows southwards from

~1.8 km in profile 7 to ~1.4 km in profile 29. The seismically derived seabed depths

are generally consistent with Smith and Sandwell (1997, version 18.1) bathymetry,

except in the axial trough of profile 9, where within 4 km of the axis, the seismically

derived depth is 0.5 km shallower. Below the seabed, the S-reflection marking the top

of the Miocene evaporites (Ross and Schlee, 1973) is found everywhere other than

over the axial trough. The Plio-Pleistocene (PP) sediments overlying the S-reflection

are thin (0.2-0.3 km thick) and tend to be uniform, as found in shallow seismic surveys

(e.g., Phillips and Ross, 1970; Ross and Schlee, 1973).

Bg

h

2BgG

h

− =

G Bg

h

95

Figure 4.4 Depths derived from the seismic reflection profiles of Izzeldin (1987) and Werner

deconvolution of marine magnetic data. Line numbers are shown in the lower right corner of

each panel. Magnetic anomalies (purple lines) along the seismic profiles were sampled from

the EMAG2 v3 grid (Meyer et al., 2017). Black lines denote bathymetry (Smith and Sandwell,

1997, version 18.1). Dark green, cyan, and red lines are the depths of the seabed, the S-

reflection at the top of the Miocene evaporites, and the basement, respectively, derived from

the seismic reflection data. Grey circles are Werner source depth solutions, with circle size

96

proportional to 𝑙𝑜𝑔2(𝜒𝑚 + 2). Depth estimates tend to cluster vertically beneath the true

location of the causative body, with magnetic basement being interpreted around the top of

the vertical clusters of solution depths. The Werner solutions generally confirm the

seismically derived basement depths.

The basement is considerably more rugged. The basement reflection is

discontinuous, probably because of faulting, and in places completely absent or un-

interpretable, a result of varied data quality. Basement outcrops directly on the

seafloor in the axial trough and deepens progressively towards the coasts from an

average depth of 1.69 km near the axis to ~6 km depth at a distance of ~60 km on both

sides of the axial trough. Further landward, this trend changes: the basement rises

steeply towards the coasts by up to 4 km in ~60 km distance, before becoming harder

to identify in the seismic data near the coasts. On the western flank of profile 15, the

reflection basement is not clear. Across the central Red Sea, the magnetic basement

tops derived from Werner deconvolution are generally consistent with the seismic

basement reflection depths. Additionally, only a minority of magnetic sources are

found by the deconvolution within the evaporites or PP sediments.

4.4.2 Oceanic-like axial crustal highs in isostatically corrected basement depths

After correcting for evaporite and sediment loading, the data reveal axial highs

in all profiles (Figure 4.5a and 4.5d). They have plateaux 70-100 km wide with adjacent

steep slopes deepening by 0.8-1.6 km over a distance of 30-40 km (Figure 4.5a).

Within the plateaux are axial troughs, where basement typically outcrops over 14 km,

forming a valley of varied size but on average 0.43 km deep (Figure 4.5d). Three

profiles marked in green in Figure 4.5d differ from the others; these lines lie furthest

to the north and furthest from the Afar plume. The other basement depth profiles

have a broadly similar morphology. The basement deepens between ~35 km from the

axis (at the axial plateau edge) and ~60-100 km, with the average profile reaching a

minimum at ~80 km. Beyond there, the basement commonly ascends towards the

coasts. Axial crustal highs are not found in active magmatic rifts, which instead contain

basement depressions (Corti et al., 2004; Mohr, 1982; Rosendahl, 1987; Thybo and

Nielsen, 2009). However, an axial high is commonly found at spreading ridges located

near mantle hotspots where excess melting generates thicker and more elevated axial

97

crust, such as the slow-spreading Reykjanes Ridge near the Iceland hotspot (Searle

and Laughton, 1981), the ultra-slow spreading Spiess Ridge near the Bouvet hotspot

(Mitchell and Livermore, 1998), and the intermediate rate Galápagos Spreading

Centre near the Galápagos hotspot (Blacic et al., 2008).

Based on the seismic profiles of Johansen et al. (1984), the Reykjanes Ridge

axial crustal high is ~40-60 km wide and rises 0.6-1.0 km above the surrounding

topography (Figure 4.5b). Although more pronounced than the Reykjanes Ridge, the

Red Sea axial basement high may imply that the central Red Sea has similarly

experienced increased melt supply and enhanced crustal thickness in the recent

geological past. As shown in Figure 4.5a and 4.5b, the relief of the Red Sea axial high

does not vary systematically with distance from the Afar region, in contrast with the

axial relief of the Reykjanes Ridge, which increases systematically towards Iceland

(Jones et al., 2002; Vogt, 1971; White et al., 1995). Moreover, with a short-wavelength

(<10 km) relief exceeding 150 m, the basement surface around the Red Sea axis is

rougher than that near the Reykjanes Ridge.

98

Figure 4.5 to be continued on next page.

99

Figure 4.5 (a): Basement depths along the Red Sea seismic lines (Figure 4.4) corrected for

evaporite and other sediment loading. (b): Basement depths around the Reykjanes Ridge from

Johansen et al. (1984), also corrected for sediment loading. (c): Locations of Reykjanes Ridge

profiles shown over the bathymetry of Smith and Sandwell (1997, version 18.1). (d): Red Sea

crustal deepening with distance from the ridge‐axis. All profiles are shown offset to their

average axial depth of 1.69 km (depth at zero distance). The solid blue line is the global normal

oceanic lithosphere subsidence curve from Crosby and McKenzie (2009). Green lines are

profiles 7 (both western and eastern flanks) and 9 (western flank) lying farthest from the Afar

plume. Global normal oceanic lithosphere subsidence (blue line) was predicted from the

Crosby and McKenzie (2009) rate with seafloor spreading rates from Chu and Gordon (1998).

Normal oceanic subsidence curves allowing for different subsidence rates were predicted

using the axial depths and the subsidence rates of Marty and Cazenave (1989). Red subsidence

curves have been offset to common 1.69 km axial depth while blue subsidence curve is shown

without offset.

4.4.3 Correlation between Bouguer gravity anomalies and basement reflection

depths

There is a strong correlation between the Bouguer anomalies and the

basement depths, although this correlation breaks down at distances greater than ~60

km from the ridge axis, where the basement shallows while the Bouguer anomaly

stays subdued (Figure 4.6). This strong correlation suggests that the density interface

between the evaporites and basement is a prominent contributor to the Bouguer

anomaly, although other density contributions (crustal thickness, and mantle and

crustal density) may also vary coherently with the deepening of basement. The

changes near the coasts suggest a reduction in the average density of the materials

100

within and beneath basement. This may reflect a change from oceanic crust around

the axis to continental or transitional crust near the coasts.

Figure 4.6 Graphs showing correlation between basement reflection depths (red) and

Bouguer gravity anomalies (blue) (Mitchell et al., 2017) derived by correcting free-air

anomalies for seabed relief using a halite density (2,160 kg m-3). Line numbers are shown in

lower right of each panel.

If the mantle density and crustal thickness are both assumed for the sake of

argument to be uniform, the correlation would be mainly due to the density contrast

between mantle rocks and evaporites acting on the topography of the basement (a

uniform crustal thickness would contribute a uniform amount to the gravity field,

aside from upward continuation effects). The derived apparent density contrasts in

Figure 4.7 vary from 220 to 580 kg m-3, with no obvious trend with latitude. These

contrasts are rather low compared with 1,070 kg m-3 if hot mantle rocks of 3,220 kg

m-3 density (Crough, 1983; Gvirtzman et al., 2016) were contrasting with evaporite

and other sediments of 2,148 kg m-3 (Wheildon et al., 1974). The difference between

101

the 1,070 kg m-3 expected value and the 220-580 kg m-3 apparent density contrasts

could arise from a combination of upward continuation effects, thickened crust, and

hotter mantle beneath the axis.

Figure 4.7 Apparent density contrasts deduced from Bouguer‐basement depth gradients.

The red and blue symbols represent western and eastern flanks, while black symbols represent

contrasts derived from data of both flanks combined. On the western flank of profile 7, the

basement reflection was too indistinct to calculate an apparent density contrast.

The apparent density contrasts were computed based on the gravity slab

formula, so it ignores contributions to the gravity field arising from topographic

changes on the basement and Moho interfaces away from the points of observations.

We carried out a simulation in which crust with a uniform thickness of 7 km and

uniform gabbroic density of 2,900 kg m-3 (Hyndman and Drury, 1977) overlies mantle

with a uniform density of 3,220 kg m-3 (Crough, 1983; Gvirtzman et al., 2016). Figure

4.8 shows two simulations using basement relief from profile 21 (Figure 4.2). To

quantify the effect of upward continuation, theoretical Bouguer gravity anomalies

(Figure 4.8c) computed from the models with and without the interface between

basement and mantle (Figures 4.8a and 4.8b) were used to derive graphs of Bouguer

gravity anomaly versus basement reflection depth and regression lines (Figure 4.8d)

whose gradients were used to calculate apparent density contrasts. Figure 4.8d shows

that if only the topography on the evaporite-basement interface were taken into

account, the apparent density contrast between evaporite and mantle would be 1,006

102

kg m-3. It also shows that if the topography on both evaporite-basement interface and

Moho were taken into account, the apparent density contrast would be reduced by

~97 kg m-3 to 909 kg m-3. We have also run the simulation with varying basement

depths, and found that upward continuation can reduce the apparent density

contrasts we infer using the gravity slab formula by up to ~160 kg m-3.

Figure 4.8 Simulation using basement depth profile 21 illustrating how apparent density

contrasts inferred using the gravity slab formula are reduced by upward continuation. (a):

Model with evaporites (2,148 kg m‐3), 7 km thick crust (2,900 kg m‐3) and mantle (3,220 kg m‐

3). (b): Model with evaporites directly overlying mantle. (c): Theoretical Bouguer gravity

anomalies computed using 2D gravity forward modelling for the two density models. (d):

Scatterplots with regression lines of Bouguer gravity anomaly versus basement reflection

depth. The slight difference in slope translates to a ~97 kg m‐3 difference in apparent density

contrast.

Alternatively, the axial topography could reflect thickened crust. If the

basement topography is uncompensated, with a near-flat Moho, the gravity anomaly

reflects the ~730 kg m-3 density contrast between evaporite and oceanic crust, much

closer to the values observed. This further supports the view that the axial high is at

least partly due to thickened crust. For a model with 7 km of crust beneath the axis

using Airy isostasy, so the topography on the Moho compensates for the basement

103

topography, our simulations suggest that an apparent density contrast of ~575 kg m-3

would be observed. However, this is an extreme model, as it ignores lateral variations

in mantle density due to temperature variations.

Addressing those mantle temperature variations, upper mantle velocities

varying from 7.4 to 7.8 km s-1 were reported for seismic refraction profile PIII of Egloff

et al. (1991), which is located in Figure 4.2. In Figure 4.9a, we show a density structure

derived from their velocities using density‐velocity relations of Christensen and Shaw

(1970). The model in Figure 4.9a is generally isostatically balanced, though there are

small imbalances at the oceanic-continental boundary and around Suakin Deep

(Figure 4.9c). The free-air anomalies predicted using 2D gravity forward modelling

successfully reproduce the observed free-air anomalies. It implies a lateral mantle

density variation of ~300 kg m-3 (Figure 4.9a). Using the basement topography and

Bouguer anomaly from 45 km to 85 km along profile PIII (outside the axial valley and

east of the ocean-continent transition), we derived an apparent density contrast of

880 kg m-3. The difference of 190 kg m-3 between 880 kg m-3 and 1,070 kg m-3 could

be due to mantle density variation and upward continuation, since the line shows no

variation in crustal thickness, but does imply a variation in mantle density. More

generally, we conclude that a combination of crustal thickness variations, upward

continuation, and mantle density variations can potentially explain the low apparent

density contrasts in Figure 4.7.

104

Figure 4.9 (a): Density structure (kg m‐3) along line PIII located in Figure 4.2 based on the

seismic refraction velocity (Vp) model of Egloff et al. (1991, their profile SO53‐PIII) and the

density‐velocity relations of Christensen and Shaw (1970). (“Pre‐evaporites” are pre‐evaporite

sedimentary rocks.) OCT: Oceanic–continental transition. (b): Free‐air gravity anomaly

calculated from (a) compared with observations from the Sandwell et al. (2014) gravity field

105

(version 23.1). (c): Total mass anomaly per unit area along PIII, computed by integrating

density over depth to the base of the model in (a).

4.5 Discussion

As mentioned above, the axial highs with basement deepening with distance

to 60 km from the spreading axis (Figure 4.5) are more like those of oceanic crust than

continental rifts, which typically host depressions (Corti et al., 2004; Mohr, 1982;

Rosendahl, 1987; Thybo and Nielsen, 2009). Prominent axial highs are common

features of oceanic spreading ridges near mantle hotspots (Blacic et al., 2008; Cochran

and Sempéré, 1997; Hooft and Detrick, 1995; Searle and Laughton, 1981). In the

central Red Sea, the boundary of the oceanic crust to transitional or continental crust

likely occurs where the correlations between basement reflection depths and Bouguer

gravity anomalies break down, coinciding roughly with the transitions identified by

Izzeldin (1987). This boundary also coincides with a transition at ~60 km from the axis

that was interpreted by Egloff et al. (1991) from their velocity data near Suakin Deep

(Figure 4.9). We here compare the axial high to those of other spreading centres,

examine its origin in more detail and explore implications.

4.5.1 How does the Red Sea axial high compare with axial highs at other spreading

centres near hotspots?

Axial highs are usually associated with “magmatically robust” spreading

centres, where the crust is unusually thick (e.g., Blacic et al., 2008). For Reykjanes

Ridge, it has been suggested that the axial high is due to thickened crust resulting from

enhanced mantle decompression melting near to the Iceland hotspot (White et al.,

1995). Using seismic reflection and refraction data, Smallwood and White (1998)

suggested that at ~62°N the Reykjanes Ridge crust thins from 10 km on the ridge axis

to 7.8 km on 5 Ma crust ~45 km from axis.

Figure 4.11 shows a compilation of bathymetry from other spreading centres

near hotspots. It includes an area south of the Azores, where a pair of ridges

surrounding the Mid-Atlantic Ridge (MAR) form a giant V-shape in plan-view, believed

to have resulted from a pulse of magmatism from the plume that has now ended,

106

leaving the previous high rifted (Cannat et al., 1999; Escartin et al., 2001). The

Reykjanes Ridge is surrounded by more than one V-shaped ridge, suggesting multiple

pulses of magmatism (e.g., Parnell‐Turner et al., 2017; Vogt, 1971). Ridges

surrounding the Galápagos Spreading Centre have been interpreted as arising from

magmatic pulses (Kappel and Ryan, 1986). Full spreading rates in these examples vary

from ~16 mm yr−1 to ~64 mm yr−1 (Chu and Gordon, 1998; DeMets et al., 1990; DeMets

et al., 2010).

The axial relief in the central Red Sea (0.8-1.6 km) is similar to that at Spiess

Ridge, more pronounced than those at Reykjanes Ridge and Galápagos Spreading

Centre, and lower than those at the Mid-Atlantic Ridge near the Azores. The crustal

thickness beneath the Spiess Ridge was estimated to be ~11-15 km (Mitchell and

Livermore, 1998), while the Galápagos Spreading Centre axis near the Galápagos

hotspot has a crustal thickness of only ~5.6-7.5 km (Canales et al., 2002).

Unlike the Reykjanes Ridge near Iceland and the Mid-Atlantic Ridge near the

Azores, the central Red Sea axial high is not obviously surrounded by V-shaped ridges

in either the gravity field (Figure 4.2) or from the seismic data (Figure 4.5a), suggesting

that the influence of Afar hotspot on the opening of central Red Sea is not that strong.

Whether this implies a lack of fluctuations in melt supply from the plume is unclear,

as any such effect might be complicated by the fracture zones apparent from the

cross-axis trends in the gravity field (Figure 4.2). Possible V-shaped ridges appear in

the free-air gravity anomalies at 17˚-18˚N, closer to the Afar plume (Mitchell and Park,

2014).

107

Figure 4.10 (a): Sodium oxide contents of axial lavas from Haase et al. (2000) (solid circles)

and Ligi et al. (2012) (plus symbols) corrected for fractionation to 8 wt% MgO. Diamond

symbols indicate the average Na8.0 values expected at the axial locations of the eleven seismic

reflection profiles based on the dashed regression line shown. (b): Seismically determined

estimates of crustal thickness versus average Na8.0 from Klein and Langmuir (1987). The Na8.0

values at the seismic lines (orange shading) suggest that the axial crustal thickness of the

central Red Sea is ~5-10 km (orange dashed lines).

4.5.2 How thick is crust beneath the axial high and how does it relate to mantle

tomographic results?

In the central Red Sea, there is only one seismic refraction dataset capable of

revealing crustal thickness (Egloff et al., 1991), and it did not reveal thickened crust

under the spreading axis (Figure 4.9a). Alternative estimates of crustal thickness are

available from geochemistry of the axial lavas. Sodium oxide concentrations in mid-

ocean ridge basalt samples corrected for magma-chamber fractionation to 8% MgO

(Na8.0) have been interpreted by Klein and Langmuir (1987) as a measure of the depth-

extent of mantle melting and shown to correlate with the thickness of oceanic crust

derived from seismic refraction experiments. The Na8.0 values from the Red Sea

108

shown in Figure 4.10a (Haase et al., 2000; Ligi et al., 2012) increase systematically

northwards implying decreasing crustal thickness, as expected from decreasing

extents of melting and decreasing mantle temperature away from the Afar plume. We

use the regression trend in Figure 4.10a to estimate the average Na8.0 at the points

where the seismic reflection lines cross the spreading axis. From the range of Na8.0

and a regression of the Klein and Langmuir (1987) Na8.0 data on crustal thickness, the

central Red Sea axial crust thickness is estimated to be ~5-10 km. This is similar to the

mean of 7.1±0.8 km for normal oceanic crust (White et al., 1992), so the geochemical

data do not indicate particularly thick crust. Furthermore, the basement is noticeably

more rugged than the Reykjanes Ridge (Figure 4.5). This may be explained by a

combination of (1) the slower spreading rate in the Red Sea, which leads to stronger,

colder lithosphere closer to the ridge and larger abyssal hills (e.g., Malinverno, 1991;

Sauter et al., 2011; Whittaker et al., 2008) and (2) potentially thinner crust in the Red

Sea, which shows some correlation with greater roughness in slow to ultraslow

spreading systems (Sauter et al., 2018).

To reconcile these observations, we speculate that the earliest seafloor

spreading in the central Red Sea began with lower melt fluxes and thinner than

average crust. Melt production then increased, increasing the crustal thickness to near

average and creating the axial high. Based on current spreading rates of Chu and

Gordon (1998) and the basement depths of Figure 4.5d, we suggest the axial high has

developed since ~9 Ma. In Figure 4.5d, the basement is most elevated relative to the

subsidence curve from 10 to 35 km off-axis, and returns to it by ~60 km. The rate of

deepening from 10 to 35 km is too fast to be caused by normal thermal subsidence. If

we interpret these variations in basement topography as solely due to thickened crust,

the crust would be thickest 10 to 35 km from the axis and would thin to ~60 km, while

the increasing elevation with distance within 10 km of the axis is most likely due to

dynamic effects within the active rift (e.g., Buck et al., 2005; Schmalholz and

Mancktelow, 2016; Tapponnier and Francheteau, 1978). This view of near normal

crustal thickness is compatible with the recent mantle seismic velocity model of Chang

et al. (2011), who showed low S-wave velocities associated with hotter mantle from

the Afar plume extending beneath Arabia rather than beneath the central Red Sea (S-

109

wave velocity beneath the southern Red Sea and Arabia is ~0.25 km s-1 lower than that

beneath the central Red Sea). This extent of hot plume material could also explain

why the influence of Afar hotspot on the opening of central Red Sea is not that strong

(although the increased melt production could have been affected by the Afar) and

there is no relation to the distance from the Afar in the apparent density contrast

(Figure 4.7).

In the Afar region, there have been pulses of volcanism (Audin et al., 2004;

Barberi et al., 1975), so the variations in basement gradient in Figure 4.5d may have

arisen from temporal changes in composition or temperature of the upwelling mantle.

Others have remarked on the possibility of pulsating mantle plumes leaving V-shaped

ridges south of Iceland and similar V-shaped ridges have been found elsewhere (e.g.,

Parnell‐Turner et al., 2017; Vogt, 1971). However, no V-shaped ridges are observed

in the central Red Sea; the crustal thickness variations appear to be consistent along

the ridge.

Alternatively, a low initial melt supply may be a result of the mechanics of

rifting mentioned in the introduction, if early melting was suppressed at the slow

rifting rates due to conductive cooling or locally infertile mantle (Bonath, 1990; Zhou

and Dick, 2013). Such a low melt supply would not be expected if there were

enhanced mantle circulation at this stage as proposed by Buck (1986). We note that

seaward thickening of oceanic crust is not always observed in seismic refraction

datasets from other rifted margins (Peron-Pinvidic et al., 2013). However, seismic

reflection and refraction data do show the crust thickens seaward at the oceanic–

continental transition (OCT) on the Angolan margin (Contrucci et al., 2004; Moulin et

al., 2005), indicating the South Atlantic Ocean basin there may have experienced an

increase of melt production during early seafloor spreading as we suggest for the

central Red Sea.

110

Figure 4.11 Examples of locally elevated topography at ridges located near mantle hotspots.

Panels (a), (c), (e) and (g) locate profiles at the Reykjanes Ridge near the Iceland hotspot, the

Spiess Ridge near the Bouvet hotspot, the northern Mid-Atlantic Ridge near Azores hotspot,

111

and the Galápagos Spreading Centre near Galápagos hotspot, respectively. Panels (b), (d), (f),

and (h) show the profiles located in (a), (c), (e) and (g), respectively. Bathymetry data from

Smith and Sandwell (1997, version 18.1).

4.5.3 What are its implications?

If the axial high in the central Red Sea represents an increasingly thick crust

but approaching only normal crustal thickness, the earlier spreading centre would

have been deeper. The earlier evaporites may have therefore been deposited

continuously across the ridge and not only on the flanks (as might otherwise have

been the case in the south). This would in turn imply that volcanic eruption occurred

beneath or through the evaporites. Magma can heat adjacent evaporite and cause it

to flow (Schofield et al., 2014). Augustin et al. (2016) proposed that the salt craters

with raised rims found in the inter-trough zones were likely created by such eruptions,

marking locations where volcanism continued after the area was covered by

evaporites. Also, such eruptions ought to have geochemical consequences. For

example, the evaporites affected could be rich in KCl and CaCl2 but poor in MgSO4 due

to hydrothermal alteration of host basalts (e.g., Jackson et al., 2000), and sulfur

isotope compositions of marine sulphates should be negatively shifted (e.g., Mills et

al., 2017). Thus, if suitable samples could be recovered, the geochemistry of the

evaporites could help to confirm the existence of a ridge buried by evaporites and map

out its transition to exposed ridge.

4.6 Conclusions

To understand what type of crust underlies the central Red Sea, we carefully

corrected for effects of overlying evaporite and other sediments to reconstruct

basement geometry from 11 deep seismic reflection lines. The seismically derived

basement depths corrected for evaporite and other sediment loading reveal an axial

high typical of mid-ocean ridges affected by hotspots such as Reykjanes Ridge, where

enhanced mantle melting results in thickened crust. In contrast, basement axial highs

are not commonly observed at active amagmatic continental rifts. Its relief of ~1 km

relative to a background subsidence trend is within the observed range. It is similar

to that at Spiess Ridge, larger than that at Reykjanes Ridge, but smaller than that of

112

the Mid-Atlantic Ridge near the Azores. We suggest the central Red Sea is underlain

by oceanic crust and the central part of the Red Sea rift is an (ultra) slow spreading

ridge influenced by the Afar hotspot, although our data do not reveal V-shaped ridges

in this part of the Red Sea like those associated with plume pulses on the Reykjanes

Ridge near Iceland or the Mid-Atlantic Ridge near the Azores.

Bouguer gravity anomalies calculated by correcting for the seabed topography

are strongly correlated with basement reflection depths with ~60 km of the axis. The

apparent density contrast implied by the correlation (220 to 580 kg m-3) is too small

for a uniform thickness crust overlying a mantle of uniform density, which would lead

to mantle rocks contrasting with evaporites and a 1,070 kg m-3 apparent density

contrast. Around 160 kg m-3 of this difference could be caused by an upward

continuation effect (our method ignores topography of interfaces). We suggest that

the remaining discrepancy is caused by lower density mantle and/or thicker crust

towards the spreading axis, although variations in crustal density may also contribute.

Geochemical data (Na8.0) suggest that the crust has normal thickness beneath

the present axis, while the rugged basement topography is consistent with a slow to

ultra-slow spreading ridge with cold, rigid lithosphere and thin crust. To reconcile the

axial high and gravity inversion results, which suggest thickening crust towards the

present day, with these other observations, we speculate that the crust was unusually

thin earlier in the evolution of the basin and has recently thickened to a more normal

thickness for a slow-spreading ridge.

4.7 Acknowledgments

We thank David Sandwell and Walter Smith for leading the gravity and

bathymetry mapping initiatives, and for the group involved in producing the grids used

in our study. Figures were prepared using the GMT software (Wessel et al., 2013).

LMK is supported by a Royal Society of Edinburgh Personal Research Fellowship

funded by the Scottish Government. We also thank Nico Augustin and an anonymous

reviewer for helpful comments that significantly improved the article.

113

4.8 References

Airy, G. B., 1855, On the computation of the effect of the attraction of mountain-

masses, as disturbing the apparent astronomical latitude of stations in

geodetic surveys: Philosophical Transactions of the Royal Society of London, v.

145, p. 101-104.

Almalki, K. A., P. G. Betts, and L. Ailleres, 2014, Episodic sea-floor spreading in the

Southern Red Sea: Tectonophysics, v. 617, p. 140-149.

Armitage, J. J., K. D. Petersen, and M. Pérez‐Gussinyé, 2018, The Role of Crustal

Strength in Controlling Magmatism and Melt Chemistry During Rifting and

Breakup: Geochemistry, Geophysics, Geosystems, v. 19, p. 534–550.



p. 375-402.

Audin, L., X. Quidelleur, E. Coulié, V. Courtillot, S. Gilder, I. Manighetti, P. Y. Gillot, P.

Tapponnier, and T. Kidane, 2004, Palaeomagnetism and K‐Ar and 40Ar/39Ar

ages in the Ali Sabieh area (Republic of Djibouti and Ethiopia): constraints on

the mechanism of Aden ridge propagation into southeastern Afar during the

last 10 Myr: Geophysical Journal International, v. 158, p. 327-345.







Barberi, F., G. Ferrara, R. Santacroce, and J. Varet, 1975, Structural evolution of the

Afar triple junction: Afar depression of Ethiopia, v. 1, p. 38-54.

Barjous, M., and S. Mikbel, 1990, Tectonic evolution of the Gulf of Aqaba-Dead Sea

transform fault system: Tectonophysics, v. 180, p. 49-59.

Blacic, T. M., G. Ito, A. K. Shah, J. P. Canales, and J. Lin, 2008, Axial high topography

and partial melt in the crust and mantle beneath the western Galápagos

114

Spreading Center: Geochemistry, Geophysics, Geosystems, v. 9 (Paper Q12005,

doi:10.1029/2008GC002100 ).

Bonath, E., 1990, Not so hot" hot spots" in the oceanic mantle: Science, v. 250, p. 107-

111.





479.



naturelle, v. 186, p. 567-606.

Bosworth, W., M. Sultan, R. Stern, R. Arvidson, P. Shore, and R. Becker, 1993, Nature

of the Red Sea crust: A controversy revisited: Comment and Reply: Geology, v.

21, p. 574-576.

Buck, W. R., 1986, Small-scale convection induced by passive rifting: the cause for

uplift of rift shoulders: Earth and Planetary Science Letters, v. 77, p. 362-372.

Buck, W. R., L. L. Lavier, and A. N. Poliakov, 2005, Modes of faulting at mid-ocean

ridges: Nature, v. 434, p. 719.

Canales, J. P., G. Ito, R. S. Detrick, and J. Sinton, 2002, Crustal thickness along the

western Galápagos Spreading Center and the compensation of the Galápagos

hotspot swell: Earth and Planetary Science Letters, v. 203, p. 311-327.

Cannat, M., A. Briais, C. Deplus, J. Escartın, J. Georgen, J. Lin, S. Mercouriev, C. Meyzen,

M. Muller, and G. Pouliquen, 1999, Mid-Atlantic Ridge–Azores hotspot

interactions: along-axis migration of a hotspot-derived event of enhanced

magmatism 10 to 4 Ma ago: Earth and Planetary Science Letters, v. 173, p. 257-

269.

Carlson, R., 2001, The effects of temperature, pressure, and alteration on seismic

properties of diabase dike rocks from DSDP/ODP Hole 504B: Geophysical

research letters, v. 28, p. 3979-3982.

Carlson, R., 2010, How crack porosity and shape control seismic velocities in the upper

oceanic crust: Modeling downhole logs from Holes 504B and 1256D:

115


doi:10.1029/2009GC002955).



v. 38.

Christensen, N. I., and G. H. Shaw, 1970, Elasticity of mafic rocks from the Mid-Atlantic

Ridge: Geophysical Journal International, v. 20, p. 271-284.






Cochran, J. R., 1979, An analysis of isostasy in the world's oceans: 2. Midocean ridge

crests: Journal of Geophysical Research: Solid Earth, v. 84, p. 4713-4729.

Cochran, J. R., 1983, A model for development of Red Sea: Aapg Bulletin, v. 67, p. 41-

69.



Cochran, J. R., and J. C. Sempéré, 1997, The Southeast Indian Ridge between 88 E and

118 E: Gravity anomalies and crustal accretion at intermediate spreading rates:


Contrucci, I., L. Matias, M. Moulin, L. Géli, F. Klingelhofer, H. Nouzé, D. Aslanian, J.-L.

Olivet, J.-P. Réhault, and J.-C. Sibuet, 2004, Deep structure of the West African

continental margin (Congo, Zaïre, Angola), between 5 S and 8 S, from

reflection/refraction seismics and gravity data: Geophysical Journal


Corti, G., M. Bonini, D. Sokoutis, F. Innocenti, P. Manetti, S. Cloetingh, and G. Mulugeta,

2004, Continental rift architecture and patterns of magma migration: A

dynamic analysis based on centrifuge models: Tectonics, v. 23 (Paper TC2012,

doi:10.1029/2003TC001561).

Corti, G., J. Van Wijk, M. Bonini, D. Sokoutis, S. Cloetingh, F. Innocenti, and P. Manetti,

2003, Transition from continental break‐up to punctiform seafloor spreading:

116

How fast, symmetric and magmatic: Geophysical Research Letters, v. 30 (Paper

1604, doi: 10.1029/2003GL017374).

Crosby, A., and D. McKenzie, 2009, An analysis of young ocean depth, gravity and

global residual topography: Geophysical Journal International, v. 178, p. 1198-

1219.






Davison, I., L. Anderson, and P. Nuttall, 2012, Salt deposition, loading and gravity

drainage in the Campos and Santos salt basins: Geological Society, London,

Special Publications, v. 363, p. 159-174.


Geophysical journal international, v. 101, p. 425-478.




New Seafloor Spreading Model of the Red Sea: Magnetic Anomalies and Plate

Kinematics: AGU Fall Meeting Abstracts (T21A-2512).


and W. Warsi, 1991, Contrasting Structural Styles Of the Eastern And Western

Margins Of the Southern Red-Sea - the 1988 Sonne Experiment:


Escartin, J., M. Cannat, G. Pouliquen, A. Rabain, and J. Lin, 2001, Crustal thickness of

V‐shaped ridges south of the Azores: Interaction of the Mid‐Atlantic Ridge

(36–39 N) and the Azores hot spot: Journal of Geophysical Research: Solid

Earth, v. 106, p. 21719-21735.

Fletcher, R., N. Kusznir, and M. Cheadle, 2009, Melt initiation and mantle exhumation

at the Iberian rifted margin: Comparison of pure–shear and upwelling-

117

divergent flow models of continental breakup: Comptes Rendus Geoscience, v.

341, p. 394-405.

Fournier, M., N. Chamot‐Rooke, C. Petit, P. Huchon, A. Al‐Kathiri, L. Audin, M. O.

Beslier, E. d'Acremont, O. Fabbri, and J. M. Fleury, 2010, Arabia‐Somalia plate

kinematics, evolution of the Aden‐Owen‐Carlsberg triple junction, and

opening of the Gulf of Aden: Journal of Geophysical Research: Solid Earth, v.

115 (Paper B04102, doi:10.1029/2008JB006257).

Furman, T., J. Bryce, T. Rooney, B. Hanan, G. Yirgu, and D. Ayalew, 2006, Heads and

tails: 30 million years of the Afar plume: Geological Society, London, Special

Publications, v. 259, p. 95-119.

Garfunkel, Z., 1981, Internal structure of the Dead Sea leaky transform (rift) in relation

to plate kinematics: Tectonophysics, v. 80, p. 81-108.

Garfunkel, Z., J. Bartov, Y. Eyal, and G. Steinitz, 1974, Raham Conglomerate–new

evidence for Neogene tectonism in the southern part of the Dead Sea Rift:

Geological Magazine, v. 111, p. 55-64.






Geophys. Union, v. 67, p. 1208-1209.

George, R., N. Rogers, and S. Kelley, 1998, Earliest magmatism in Ethiopia: evidence

for two mantle plumes in one flood basalt province: Geology, v. 26, p. 923-926.

Gettings, M. E., H. Blank, W. Mooney, and J. Healey, 1986, Crustal structure of

southwestern Saudi Arabia: Journal of Geophysical Research: Solid Earth, v. 91,

p. 6491-6512.


7-11.



118

Red Sea: in Whitmarsh, R.B., Weser, O.E., Ross, D.A.,et al., Initial Reports of the

Deep Sea Drilling Project, v. 23, p. 913-921.









Hall, S. A., 1989, Magnetic Evidence for the Nature Of the Crust beneath the Southern

Red-Sea: Journal Of Geophysical Research-Solid Earth And Planets, v. 94, p.

12267-12279.

Hansen, S., S. Schwartz, A. Al-Amri, and A. Rodgers, 2006, Combined plate motion and

density-driven flow in the asthenosphere beneath Saudi Arabia: Evidence from

shear-wave splitting and seismic anisotropy: Geology, v. 34, p. 869-872.

Harry, D., and J. Bowling, 1999, Inhibiting magmatism on nonvolcanic rifted margins:

Geology, v. 27, p. 895-898.




Hooft, E. E., and R. S. Detrick, 1995, Relationship between axial morphology, crustal

thickness, and mantle temperature along the Juan de Fuca and Gorda Ridges:


Hopper, J. R., and W. Roger Buck, 1998, Styles of extensional decoupling: Geology, v.

26, p. 699-702.

Hughes, G., and Z. Beydoun, 1992, The Red Sea—Gulf of Aden: biostratigraphy,

lithostratigraphy and palaeoenvironments: Journal of Petroleum Geology, v.

15, p. 135-156.


ultramafic rocks from DSDP Leg 37: DSDP, v. 37, p. 395-401.

Izzeldin, A., 1982, On the structure and evolution of the Red Sea: Diss. Univ. Strasbourg.

119







Jackson, M. P., C. Cramez, and J.-M. Fonck, 2000, Role of subaerial volcanic rocks and

mantle plumes in creation of South Atlantic margins: implications for salt

tectonics and source rocks: Marine and Petroleum Geology, v. 17, p. 477-498.

Jeanniot, L., N. Kusznir, G. Mohn, G. Manatschal, and L. Cowie, 2016, Constraining

lithosphere deformation modes during continental breakup for the Iberia–

Newfoundland conjugate rifted margins: Tectonophysics, v. 680, p. 28-49.

Johansen, B., P. R. Vogt, and O. Eldholm, 1984, Reykjanes Ridge: further analysis of

crustal subsidence and time-transgressive basement topography: Earth and

planetary science letters, v. 68, p. 249-258.

Jones, S. M., N. White, and J. Maclennan, 2002, V‐shaped ridges around Iceland:

Implications for spatial and temporal patterns of mantle convection:

Geochemistry, Geophysics, Geosystems, v. 3 (Paper 1059,

doi:10.1029/2002GC000361).

Kalnins, L. M., 2011, Spatial variations in the effective elastic thickness of the

lithosphere and their tectonic implications, Oxford University, 251-325 p.

Kappel, E. S., and W. B. Ryan, 1986, Volcanic episodicity and a non‐steady state rift

valley along northeast Pacific spreading centers: Evidence from Sea MARC I:


Karbe, L., 1987, Hot brines and the deep sea environment, AJ Edwards, and SM Head

(Oxford: Pergamon Press).

Karner, G., N. Driscoll, and J. Peirce, 1991, Gravity and magnetic signature of broken

ridge, southeast Indian Ocean: Proceedings of the Ocean Drilling Program,


120









p. 513-524.

Levi, S., and R. Riddihough, 1986, Why are marine magnetic anomalies suppressed

over sedimented spreading centers?: Geology, v. 14, p. 651-654.




doi:10.1029/2012GC004155).

Ligi, M., E. Bonatti, W. Bosworth, Y. Cai, A. Cipriani, C. Palmiotto, S. Ronca, and M.

Seyler, 2018, Birth of an ocean in the Red Sea: Oceanic-type basaltic melt

intrusions precede continental rupture: Gondwana Research, v. 54, p. 150-160.




p. 1019-1022.


Pacific Ocean and Hawaii: Boulder, Colorado, Geological Society of America,

Geology of North America, v. N, p. 499-521.

Mackenzie, K. V., 1981, Discussion of sea water sound‐speed determinations: The


Malinverno, A., 1991, Inverse square-root dependence of mid-ocean-ridge flank

roughness on spreading rate: Nature, v. 352, p. 58-60.


parameters: Journal of the society for Industrial and Applied Mathematics, v.

11, p. 431-441.

121



150, p. 1-31.


spreading in the Woodlark Basin: Indications of rift‐ induced secondary

mantle convection: Journal of Geophysical Research: Solid Earth, v. 104, p.

12909-12926.

Marty, J. C., and A. Cazenave, 1989, Regional variations in subsidence rate of oceanic

plates: a global analysis: Earth and Planetary Science Letters, v. 94, p. 301-315.

McKenzie, D., D. Davies, and P. Molnar, 1970, Plate tectonics of the Red Sea and east


Meyer, B., R. Saltus, and A. Chulliat, 2017, EMAG2: Earth magnetic anomaly grid (2-

arc-minute resolution) version 3: National Centers for Environmental

Information, NOAA. Model. doi, v. 10, p. V5H70CVX.

Mills, J. V., M. L. Gomes, B. Kristall, B. B. Sageman, A. D. Jacobson, and M. T. Hurtgen,

2017, Massive volcanism, evaporite deposition, and the chemical evolution of

the Early Cretaceous ocean: Geology, v. 45, p. 475-478.



v. 29, p. 352-369.

Mitchell, N. C., and R. A. Livermore, 1998, Spiess Ridge: An axial high on the slow

spreading Southwest Indian Ridge: Journal of Geophysical Research: Solid

Earth, v. 103, p. 15457-15471.



Mohr, P., 1982, Musings on continental rifts: In: Palmason, G. (Ed.), Continental and

oceanic rifts. American Geophysical Union, Geodynamic Series, v. 8, p. 293-309.





122

Moulin, M., D. Aslanian, J.-L. Olivet, I. Contrucci, L. Matias, L. Géli, F. Klingelhoefer, H.

Nouzé, J.-P. Réhault, and P. Unternehr, 2005, Geological constraints on the

evolution of the Angolan margin based on reflection and refraction seismic

data (ZaïAngo project): Geophysical Journal International, v. 162, p. 793-810.







Q06021, doi:10.1029/2006GC001566).




doi:10.1029/2007GC001895).

Parnell‐Turner, R., N. White, T. J. Henstock, S. M. Jones, J. Maclennan, and B. J.

Murton, 2017, Causes and Consequences of Diachronous V‐Shaped Ridges in

the North Atlantic Ocean: Journal of Geophysical Research: Solid Earth, v. 122,

p. 8675–8708.

Peron-Pinvidic, G., G. Manatschal, and P. T. Osmundsen, 2013, Structural comparison

of archetypal Atlantic rifted margins: A review of observations and concepts:

Marine and Petroleum Geology, v. 43, p. 21-47.





Sciences, v. 267, p. 205-217.

Phillips, J. D., and D. A. Ross, 1970, Continuous seismic reflexion profiles in the Red

Sea: Philosophical Transactions of the Royal Society of London A:

Mathematical, Physical and Engineering Sciences, v. 267, p. 143-152.

123

Rao, V. B., 1984, On:“Werner deconvolution for automated magnetic interpretation

and its refinement using Marquardt’s inverse modeling” by CC Ku and JA Sharp

(Geophysics, v 48, p. 754–774, June, 1983): Geophysics, v. 49, p. 1119-1119.

Rasul, N. M., I. C. Stewart, and Z. A. Nawab, 2015, Introduction to the Red Sea: Its

Origin, Structure, and Environment, The Red Sea, Springer, p. 1-28.

Richards, M. A., R. A. Duncan, and V. E. Courtillot, 1989, Flood basalts and hot-spot

tracks: plume heads and tails: Science, v. 246, p. 103-107.



Basins Red Sea – Gulf of Aden, Springer, p. 29-49.Ros, E., M. Pérez‐Gussinyé,

M. Araújo, M. Thoaldo Romeiro, M. Andrés‐Martínez, and J. P. Morgan, 2017,

Lower Crustal Strength Controls on Melting and Serpentinization at Magma‐

Poor Margins: Potential Implications for the South Atlantic: Geochemistry,

Geophysics, Geosystems, v. 18, p. 4538–4557.

Rosendahl, B. R., 1987, Architecture of continental rifts with special reference to East

Africa: Annual Review of Earth and Planetary Sciences, v. 15, p. 445-503.






Sandwell, D. T., and W. H. Smith, 2009, Global marine gravity from retracked Geosat

and ERS‐1 altimetry: Ridge segmentation versus spreading rate: Journal of

Geophysical Research: Solid Earth (1978–2012), v. 114.Sauter, D., H. Sloan, M.

Cannat, J. Goff, P. Patriat, M. Schaming, and W. R. Roest, 2011, From slow to

ultra-slow: How does spreading rate affect seafloor roughness and crustal

thickness?: Geology, v. 39, p. 911-914.

Sauter, D., J. Tugend, M. Gillard, M. Nirrengarten, J. Autin, G. Manatschal, M. Cannat,

S. Leroy, and M. Schaming, 2018, Oceanic basement roughness alongside

magma-poor rifted margins: insight into initial seafloor spreading: Geophysical


124

Schmalholz, S. M., and N. S. Mancktelow, 2016, Folding and necking across the scales:

a review of theoretical and experimental results and their applications: Solid

Earth, v. 7, p. 1417-1465.

Schofield, N., I. Alsop, J. Warren, J. R. Underhill, R. Lehné, W. Beer, and V. Lukas, 2014,

Mobilizing salt: Magma-salt interactions: Geology, v. 42, p. 599-602.

Schouten, H., H. J. Dick, and K. D. Klitgord, 1987, Migration of mid-ocean-ridge volcanic

segments: Nature, v. 326, p. 835-839.

Searle, R., and A. Laughton, 1981, Fine-scale sonar study of tectonics and volcanism

on the Reykjanes Ridge: Oceanologica Acta, v. 4, p. 5-13.

Sebai, A., E. Stutzmann, J.-P. Montagner, D. Sicilia, and E. Beucler, 2006, Anisotropic

structure of the African upper mantle from Rayleigh and Love wave

tomography: Physics of the Earth and Planetary Interiors, v. 155, p. 48-62.

Sempéré, J.-C., G. Purdy, and H. Schouten, 1990, Segmentation of the Mid-Atlantic

Ridge between 24 N and 30 40'N.

Shi, W., N. Mitchell, and L. Kalnins, 2017, Assessing the nature of crust in the central

Red Sea using potential field methods: EGU General Assembly Conference

Abstracts, p. 542.

Smallwood, J. R., and R. S. White, 1998, Crustal accretion at the Reykjanes Ridge, 61–

62 N: Journal of Geophysical Research: Solid Earth (1978–2012), v. 103, p.

5185-5201.

Smith, W. H. F., and D. T. Sandwell, 1997, Global Sea Floor Topography from Satellite

Altimetry and Ship Depth Soundings: Science, v. 277, p. 1956-1962.



593-596.




(T12B-04).

Tapponnier, P., and J. Francheteau, 1978, Necking of the lithosphere and the

mechanics of slowly accreting plate boundaries: Journal of Geophysical

Research: Solid Earth (1978–2012), v. 83, p. 3955-3970.

125

Thybo, H., and C. A. Nielsen, 2009, Magma-compensated crustal thinning in

continental rift zones: Nature, v. 457, p. 873-876.




1415.

Vogt, P., 1971, Asthenosphere motion recorded by the ocean floor south of Iceland:


Watts, A., and E. Burov, 2003, Lithospheric strength and its relationship to the elastic

and seismogenic layer thickness: Earth and Planetary Science Letters, v. 213, p.

113-131.

Watts, A. B., 2001, Isostasy and Flexure of the Lithosphere: UK, Cambridge University

Press, 1-284 p.








Union, v. 94, p. 409-410.





White, R., J. Bown, and J. Smallwood, 1995, The temperature of the Iceland plume and

origin of outward-propagating V-shaped ridges: Journal of the Geological

Society, v. 152, p. 1039-1045.

White, R. S., D. McKenzie, and R. K. O'Nions, 1992, Oceanic crustal thickness from

seismic measurements and rare earth element inversions: Journal of


126

Whitmarsh, R., O. Weser, and D. Ross, 1974, Initial reports of the deep sea drilling

project: US Government Printing Office, Washington, v. 23b.

Whittaker, J. M., R. D. Müller, W. R. Roest, P. Wessel, and W. H. Smith, 2008, How

supercontinents and superoceans affect seafloor roughness: Nature, v. 456, p.

938-941.



Zhou, H., and H. J. Dick, 2013, Thin crust as evidence for depleted mantle supporting

the Marion Rise: Nature, v. 494, p. 195.

127

Chapter 5.

Paper 2: Central Red Sea basement depths

from Werner deconvolution of aeromagnetic

data

128

5. Paper 2: Central Red Sea basement depths from

Werner deconvolution of aeromagnetic data

Wen Shi1*, Neil C. Mitchell1, Lara M. Kalnins2, Ian C.F. Stewart3, A.Y. Izzeldin4




EH9 3FE, UK.

3Stewart Geophysical Consultants Pty. Ltd., Adelaide, South Australia.




This paper is ready for submission to the journal Marine Geophysical Researches.

129

Abstract

The Red Sea is an important example of a rifting continent transitioning to

oceanic basin. The geometry of its basement is central to questions concerning the

nature of the crust (whether it is oceanic or highly stretched continental crust) and to

whether the relief of the basement affects the flow of evaporites, questions that may

also apply to some more ancient rifted margins. To reconstruct basement geometry,

Werner deconvolution was used to invert a grid of aeromagnetic anomalies (with

survey lines spaced 2-10 km apart) for magnetic source depths. At any location, the

deconvolution typically yields many solutions over a wide range of depths, though

usually clustered below the depth of magnetic basement. Comparing magnetic source

depths with basement depths interpreted from eleven seismic reflection profiles, we

computed cumulative density functions (CDFs) of differences between them (ΔZ). For

the near-axis data, where we know the basement is igneous, we recorded the CDF

level corresponding with ΔZ=0 and then mapped out the depth corresponding with

that CDF level elsewhere throughout the central Red Sea. The derived magnetic

basement along the ridge axis tends to deepen by ~0.25 km from 18.5°N to 23°N,

consistent with declining influence of the Afar plume. Magnetic basement depth near

the spreading axis generally co-varies with Bouguer gravity anomalies, as expected

from the latter reflecting a strong density contrast between basement and the

overlying evaporites and other sediments. Valleys in the derived depths mostly

coincide with fracture zones interpreted previously from gravity, magnetic,

bathymetric and seismic reflection data. Those valleys also correspond with areas

where the evaporites have extended into the axial valley floor, as suggested by earlier

researchers. Nearer to the coasts, the resolved magnetic basement depth is shallower

than seismic basement, possibly because of magnetized minerals in the sedimentary

section. This suggests a second utility of the method in mapping out sedimentary

sources where seismic data are available but lithological information is poor. Overall,

our work illustrates the potential utility of magnetic source depth determination in

areas with few seismic data but where the basement has a strong magnetization

contrast with overlying sediments.

130

Keywords: Central Red Sea, Magnetic basement depth, Werner deconvolution,

Cumulative density function

5.1 Introduction

Depth of basement from magnetic anomalies can be useful in regions where

basement is not well resolved in seismic reflection or refraction data, or where seismic

data are sparse. This is the case in the Red Sea, where the geometry of the basement

is an important constraint on the type of crust (whether it is oceanic or continental),

which is important for models of the continental rift to spreading transition, a key

stage of Wilson's (1966) cycle. In the central Red Sea, which some have argued is

currently undergoing a transition to seafloor spreading (Bonatti et al., 1981; Ligi et al.,

2011, 2012), seismic reflection and refraction data able to image basement

continuously along lines and thus resolve basement geometry are sparse (Egloff et al.,

1991; Izzeldin, 1982, 1987, 1989), and otherwise only refraction data from single

hydrophones are available (Drake and Girdler, 1964; Tramontini and Davies, 1969).

The crust of the central Red Sea has also been interpreted as mainly oceanic (Izzeldin,

1987; Mitchell and Park, 2014; Shi et al., 2018). The morphology of the seabed

comprises a series of closed-contour deeps (Figure 5.1) separated by shallower areas

underlain by evaporites (Inter-Trough Zones or ITSs; Bonatti, 1985). The ITZs have

been suggested to arise from flowage of the evaporites on the flanks of the ridge

underlying the central Red Sea along fracture zones (Augustin et al., 2014, 2016;

Mitchell et al., 2017), where the basement is depressed and the evaporites likely

warmer, thus the geometry of basement is important for understanding the flowage

also. Although basement depths derived from magnetic anomalies are not usually as

reliable as those from seismic datasets, their calculation is nevertheless potentially

useful given the importance of the region and lack of publicly available seismic data.

131


locate the prominent deeps in the central Red Sea (from Augustin et al. (2014) and Karbe

(1987)). From north to south, these are (1) Nereus, (2) Thetis, (3) Hadarba, (4) Hatiba, (5)

Atlantis II, (6) Erba, (7) Port Sudan, (8) Suakin, and (9) Pelagia deeps. Yellow dot marks the

city of Jeddah. Purple dot marks the Tokar Delta. Green dot marks the Farasan Islands.

Relative plate motion vectors (blue) were predicted using the Chu and Gordon (1998) plate

rotation pole.

In regions where seismic coverage is sufficiently extensive to allow basement

structure to be mapped out continuously, magnetic methods can also be useful to

determine if magnetic sources exist within the sediments. For example, in the Red

Sea, we would not expect much induced or remanent magnetism to originate from

the evaporites if they were wholly formed by seawater evaporation. On the other

132

hand, the presence of some magnetic sources within them could suggest the presence

of terrigenous sediments containing magnetic minerals and that may have

implications for the supply of sediment derived by erosion of the adjacent land.

Terrigenous sediments within impermeable evaporite bodies may also suggest

potential petroleum reserves if suitable organic sources and thermal histories are

present.

Werner deconvolution of magnetic data has been used to investigate

basement geometry and geological structure for decades (e.g., Ku and Sharp, 1983;

Martelet et al., 2013; Stagg et al., 1989; Thakur et al., 2000; Tsokas and Hansen, 1995).

In the present study, we invert 64 lines of aeromagnetic anomalies for magnetic

source depths in the central Red Sea using Werner deconvolution. To obtain more

reliable magnetic basement depths, seismic reflection data are used as constraints.

We overlay the fracture zones interpreted previously from gravity, magnetic,

bathymetric and seismic reflection data on the map of magnetic basement depths to

assess how well these fracture zones coincide with basement valleys. We use a

correlation between basement depth and Bouguer anomalies found previously (Shi et

al., 2018), along with geological consistency, to test of the method results. We also

overlay the extents of evaporite and exposed volcanic basement suggested by earlier

researchers on the derived depth map to help assess how well the more extensive

flows correspond with those valleys. We then discuss the potential utility of magnetic

source depth determination in the central Red Sea.

5.2 Geological setting

The Red Sea has formed by the extension of the Afro-Arabian shield, creating

the separate Arabian and Nubian plates (e.g., Ghebreab, 1998; McKenzie et al., 1970).

The extension may have begun in the Eocene and developed substantially in the

Oligocene at ~30 Ma (Bosworth and McClay, 2001; Hofmann et al., 1997; Mohr, 1983;

Omar and Steckler, 1995). The 3.2-Ma-to-present Red Sea opening rate increases

southward from ~10 mm yr−1 at 25.5°N to ~16 mm yr−1 near 18°N with increasing

distance from the Nubia/ Arabia pole located in the Mediterranean (e.g., Chu and

Gordon, 1998; DeMets et al., 1990). Southward of 18°N, the Red Sea is affected by

133

the Afar plume, revealed in 3He/4He ratios and both major and trace elements of axial

lava samples (Altherr et al., 1988; Moreira et al., 1996; Volker et al., 1997). Haase et

al. (2000) provided geochemical data of axial lavas (Na8.0, sodium oxide concentrations

corrected for fractionation (Klein and Langmuir, 1987)) suggesting that the upper

mantle temperature in the Red Sea generally decreases by about 60°C from 18°N to

26°N. (The source pressure of melting also declines (Haase et al., 2000), implying that

the temperature change at a given depth may not be so large.) Chang et al. (2011)

suggested that the hot plume material from Afar at a depth of ~150 km does not

extend north-westwards below the central and northern Red Sea areas, but forms a

channel extending northward beneath Arabia.

The transition from continental extension to seafloor spreading is a key stage

in Wilson's (1966) cycle, but there are few young ocean basins where this transition

can be observed. The Woodlark Basin is small and opening quickly (~60 mm yr-1;

Martinez et al., 1999; Weissel et al., 1982) and the Gulf of California rift is opening

highly obliquely (Atwater and Stock, 1998; Lonsdale, 1989; Withjack and Jamison,

1986) with a spreading rate of ~45–47 mm yr-1 (Plattner et al., 2007). The Red Sea is

the only example of a young transitioning rift opening slowly (~10-~16 mm yr-1) and

nearly orthogonally (e.g., Chu and Gordon, 1998), so it is an important example of

transitioning from nearly orthogonal slow continental rifting to seafloor spreading.

Unfortunately, the Miocene evaporites covering much of the Red Sea, which

reach up to 4 km in places (Shi et al., 2018), make imaging of the basement with

seismic methods difficult and magnetic anomalies are affected by the greater depth

of basement, which is depressed isostatically by the evaporites, and by potential

alteration of magnetic minerals by hydrothermal circulation under them (Augustin et

al., 2014; Izzeldin, 1987; Mitchell and Park, 2014). In the northern Red Sea, low seismic

velocities (Gaulier et al., 1986) and other data have been interpreted as indicating the

presence of continental crust (Cochran, 1983), although this has recently been

disputed (Dyment et al., 2013). In the southern Red Sea, a continuous axial zone with

extensive volcanism and linear magnetic anomalies identifiable up to Chron 3 (5 Ma,

Figure 5.2) indicate that spreading has been oceanic since at least 5 Ma (Augustin et

al., 2014; Phillips, 1970; Roeser, 1975; Vine, 1966). Although not attributable to

134

individual chrons, parallel magnetic anomalies closer to the coasts may also indicate

oceanic crust, as they are symmetrical about the axis and not correlated with

basement geometry (Hall, 1989).

Figure 5.2 Seafloor spreading magnetic anomalies in the southern Red Sea (Phillips, 1970).

The bottom curve represents synthetic magnetic anomalies generated by the seafloor

spreading model beneath it using a spreading rate of 10 mm yr-1. The black blocks indicate

normal magnetization, whereas open blocks represent reversed magnetization.

In the central Red Sea (19°N to 23°N; Figure 5.1), the structure of the axial zone

is more complicated, consisting of a series of ‘deeps’ separated by inter-trough zones.

Normal mid-ocean ridge basalt (MORB) and high amplitude magnetic anomalies have

been found in these deeps (e.g., Bonatti, 1985; Pautot et al., 1984). There is still

debate concerning whether the deeps are discrete seafloor spreading cells, separated

by stretched continental lithosphere (Bonatti, 1985; Cochran, 1983; Ligi et al., 2011,

2012) or exposed parts of a continuous oceanic spreading axis, which is partly covered

by evaporites and other sediments (e.g., Augustin et al., 2014, 2016; Davies and

Tramontini, 1970; Dyment et al., 2013; Egloff et al., 1991; Izzeldin, 1982, 1987, 1989;

LaBrecque and Zitellini, 1985; Mitchell and Park, 2014; Searle and Ross, 1975; Shi et

al., 2018; Tramontini and Davies, 1969). Here, we favour the second interpretation,

135

as axial crustal highs typical of a spreading ridge are found in the entire central Red

Sea, based on seismic reflection data and Bouguer anomalies (Izzeldin, 1987; Shi et al.,

2018).


5.3.1 Multichannel seismic reflection

Multichannel seismic reflection data were collected by a geophysical services

company in 1976 as described by Izzeldin (1982). The deep-water survey used a 2.4

km, 48 channel streamer (50 m spacing), whilst a 1.2 km, 24 channel streamer (50 m

spacing) was used in shallow water. The data were moveout-corrected by others as

described by Izzeldin (1982, 1987). The locations of profiles 7, 9, 11, 15, 17, 19, 21, 25,

27, 29, and 31 used in this study are shown in Figures 5.3a and 5.3c.

Depths of basement were derived from these data as described by Shi et al.

(2018). Two-way travel times for the basement and seabed reflections were

converted to depths below sea level using P-wave velocities (Vp) of 1.538 km s-1

(Mackenzie, 1981), 1.9 km s-1 (Egloff et al., 1991; Gaulier et al., 1988; Whitmarsh et al.,

1974), and 4.21 km s-1 (Egloff et al., 1991; Girdler and Whitmarsh, 1974; Tramontini

and Davies, 1969) for water, Plio-Pleistocene sediments, and evaporites , respectively.

5.3.2 Magnetic anomalies

5.3.2.1 Sources of magnetic data

The aeromagnetic survey was also carried out in 1976 by the Arabian

Geophysical and Surveying Company (ARGAS) as described by Izzeldin (1982, 1987).

As shown in Figure 5.3a, this survey covered the central Red Sea between 18.5°N and

23°N with flight lines oriented N60°E. Sixty-four survey lines were run from coast to

coast and spaced 10 km apart. 464 shorter lines were added over the axial zone and

some coastal zones for detailed investigation with the lines spaced 2.5 km apart. The

total length of the main survey lines is 23,011 km, whereas that of the additional lines

is 28,210 km. The survey was flown at 305 m above sea level. Measurements of total

136

magnetic field were obtained using a caesium vapour magnetometer with a resolution

of 0.01 nT and sample frequency of 1Hz.

Figure 5.3 (a): Locations of aeromagnetic survey flight lines (blue) and multichannel seismic

reflection profiles 7, 9, 11, 15, 17, 19, 21, 25, 27, 29, 31 of Izzeldin (1987) (red). Black line

locates the spreading axis. (b): Residual aeromagnetic anomalies from data collected by

Arabian Geophysical and Surveying Company (ARGAS) in 1976. The data have been gridded

with a horizontal resolution of 1 arc-minute and contoured every 75 nT. In order to reduce

effects of remaining reference field errors, the residual aeromagnetic anomalies of each

survey line were adjusted by subtracting its mean anomaly value before gridding and

contouring. (c): Locations of the survey lines used for the Werner deconvolution (blue). (d):

Residual aeromagnetic anomalies along the long survey lines plotted as profiles.

137

Magnetic anomalies were corrected for Earth's main field variations. Diurnal

variations were corrected using data from a magnetometer stationed in Jeddah

(Figure 5.1) and the results verified by analysis of data at crossing flight lines. Minor

line levelling errors in the data were minimised. Figure 5.3b shows the contoured

anomalies after further adjustments made here to correct minor remaining data

offsets of the individual lines (see figure caption). The length of survey line affects the

number of magnetic sources found by Werner deconvolution (a longer length of data

is needed to resolve deeper magnetic bodies) so we used the longer lines for our

analysis (Figure 5.3c). Consequently, there are fewer deep sources revealed towards

the ends of each line.

5.3.2.2 Werner deconvolution

Werner deconvolution has been widely used to estimate depths and

susceptibilities of magnetic sources (e.g. Karner et al., 1991; Ku and Sharp, 1983;

Thakur et al., 2000). The method is valid for sources comprising thin sheet-like bodies

of semi-infinite extent (alternatively, bodies with a large ratio of length to thickness)

and the edges of thick bodies (Ku and Sharp, 1983). Although individual depth values

derived using the method have large uncertainties and the method can produce some

erroneous solutions due to non-uniqueness and calculation window sizes poorly

matching those needed for the solutions, their depths generally cluster within

basement (Cochran and Karner, 2007; Karner et al., 1991).

Following Werner (1953) and Ku and Sharp (1983), the total magnetic anomaly

due to a dike or other tabular body can be written as:

(5.1)

where is distance along a profile, and are functions of orientation and

magnetization of the dike, is horizontal position of the point immediately above the

dike, and is depth to the top of the dike.

0

2 2

0

( )( ,0)=

( )

A x x BDTmag x

x x D

− +

− +

x A B

0x

D

138

The interferences from neighbouring anomalies, regional trends, and

measured magnetic noise are addressed by the addition of an interference polynomial

( ) to the right side of equation (5.1) (Ku and Sharp, 1983; Werner, 1953):

(5.2)

where is the order of the interference polynomial, and are the

coefficients.

In practice, a polynomial of order two is adequate for obtaining stable and

reliable solutions (Hartman et al., 1971; Ku and Sharp, 1983; Werner, 1953):

(5.3)

Equation (5.3) can be expressed as:

(5.4)

where

and

Marquardt’s (1963) inverse modelling method was used to solve the

simultaneous equations constructed from equation (5.4) by a seven-point Werner

operator (Ku and Sharp, 1983). If the sample spacing is , we have the following

results for the thin dike (Ku and Sharp, 1983):


(5.5a)

P

2

0 1 2

n

nP C C x C x C x= + + + +

n 0 1 2, , , nC C C C

200 1 22 2

0

( )( ,0)=

( )

A x x BDTmag x C C x C x

x x D

− ++ + +

− +

2 3 4 2

0 1 2 3 4 0 1a a x a x a x a x b Tmag b xTmag x Tmag+ + + + + + =

2 2

0 0 0 0 0

2 2

1 0 0 1 1 0

2 2

2 0 1 0 2 2 0

3 1 2 0

4 2

2 2

0 0

,

2 ,

2 ,

2 ,

,

,

a Ax BD C D x C

a A C x C D C x

a C C x C D C x

a C C x

a C

b x D

= − + + +

= − + +

= − + +

= −

=

= − −

1 02b x=

x

1X=0.5 b x x +

139

Depth:

(5.5b)

Magnetic susceptibility: (5.5c)

where

(5.6a)

and

(5.6b)

where is the vector sum of the induced and remanent magnetization,

is thickness of the dike ( ), is magnetic inclination of the main field ,

and is the strike of the tabular body measured counterclockwise from magnetic

north.

Werner deconvolution was performed along the profiles using a moving

window with a minimum/maximum size of 0.5/60 km in 0.5 km increments and shifted

every 1 km. This returned a number of stable clusters of Werner solutions (Figure 5.4).

2

0 1= 0.25 Y b b x− −

2 2

=x z

m

J Jx

+

F

2 2

cos sin sin

2 (cos sin ) (sin )x

B A IJ

T I

− −=

+

2 2

cos sin sin

2 (cos sin ) (sin )z

A B IJ

T I

− +=

+

=( , )x zJ JJs

2 T T D I F

140

Figure 5.4 Depths derived from the seismic reflection profiles of Izzeldin (1987) and Werner

deconvolution of the aeromagnetic data. Line numbers are shown in the top left corner of

each panel. Black lines denote bathymetry (Smith and Sandwell, 1997, version 18.1). Dark

green, cyan, and red lines are the depths of the seabed, the S-reflection at the top of the

Miocene evaporites, and the basement, respectively, derived from the seismic reflection data.

Grey circles are Werner source depth solutions, with circle size proportional to 𝑙𝑜𝑔2(𝜒𝑚 + 2),

where 𝜒𝑚 is apparent susceptibility. Blue lines are the basement depths derived from

magnetic sources using the CDF level of 93.16% (corresponding with ΔZ =0 for the axial data

(Figure 5.5a)).

141

5.3.2.3 Statistical analysis of the Werner solutions

Although the Werner method produced stable clusters of solutions, the results

are nevertheless noisy and not centred on the top of basement. To obtain more

reliable magnetic basement depths, the seismic reflection data were used as

constraints. Differences between the individual Werner source depths and the

seismically derived basement depths (ΔZ) (i.e., Werner source depth minus seismically

derived basement depth) were computed and cumulative histograms (cumulative

density functions, CDFs) derived from those differences as shown in Figure 5.5. CDFs

of ΔZ were derived from the data around the spreading axis (distance from the

spreading axis ≤ 60 km) and close to the coasts (distance from the axis > 60 km)

separately (Figure 5.5a and 5.5b). Since the primary magnetic source around the

spreading centre is expected to lie within the shallow basement from the

magnetizations of extrusive basalts and sheeted dykes (Tivey and Dyment, 2010; Tivey

and Johnson, 1987), we first recorded the CDF level corresponding with ΔZ=0 (i.e.,

magnetic sources lying at the top of seismic basement) for the axial data (Figure 5.5a).

We then used this CDF level to estimate depth to basement throughout the central

Red Sea, assuming that the whole region consists of similarly magnetized basement,

with none of the magnetic field originating in the overlying sediments. CDFs were

computed from magnetic sources within rectangular cells 20 km × 10 km along the

survey lines. Figures 5.4, 5.6 and 5.7 show those magnetic basement depths.

142

Figure 5.5 Histograms and cumulative density functions (CDF) of differences between the

Werner source depths and the seismically derived basement depths (ΔZ) (i.e., Werner source

depth minus seismically derived basement depth). Left axes correspond with the histograms.

(a) and (b) are, respectively, calculated from the data over the areas around the axial trough

(distance from the ridge axis ≤ 60 km) and near the coasts (distance from the ridge axis > 60

km).


Mitchell et al. (2017) produced a map of marine Bouguer gravity anomalies of

the central Red Sea, removing the seabed effect from free-air gravity anomalies

(Figure 5.8a). Short-wavelength variations in the Bouguer anomalies are affected by

topography of the basement underlying the evaporites, due to the strong density

contrast across the basement surface. Consequently, Bouguer anomalies and

basement depths are strongly correlated within 60 km of the axis (Shi et al., 2018).

Essentially, Bouguer anomalies can be used as a proxy for basement depths. We

143

therefore examined the covariation between Bouguer anomalies and magnetic

basement depths as an independent, albeit indirect, test of the magnetically derived

depths. Figure 5.8b shows the magnetic depths in colour shaded using the Bouguer

gravity anomalies of Mitchell et al. (2017), to help the reader locate features between

the two datasets. We also used the correlation between the two data types to invert

for a single apparent basement density using the slab formula.


Augustin et al. (2014) interpreted multibeam sonar data from the central Red

Sea to reveal the pattern of evaporite flowage. We overlaid their interpreted extents

of evaporite and exposed volcanic basement on the map of magnetic source depths

to help assess how well the more extensive flow corresponds with basement valleys

and indirectly assess the deconvolution results for geological consistency.

5.4 Results

5.4.1 Basement depth derived from aeromagnetic data

In Figure 5.4, magnetic source solutions produced by Werner deconvolution

occur over a wide range of depths. Some of those solutions are clearly erroneous, for

example, the solutions that lie in water. After carrying out the statistical analysis of

the solutions (section 5.3.2.3), the derived magnetic basement depths shown in Figure

5.4 show some local discrepancies. At the axis, the magnetic basement tends to be

deeper than the seismic basement by up to 2 km (Figure 5.4). At ~60 km off-axis, the

magnetic basement is commonly shallower than seismic basement by up to a few km

in places. Near the coasts, the magnetic basement depth is shallower than seismic

derived basement, representing the effects of magnetized minerals in the

sedimentary section. Nevertheless, ignoring the local differences and the

discrepancies nearer to the coasts, the derived magnetic basement depths seem to

reveal the main geomorphological features of the basement.

In Figure 5.6, the basement has an axial plateau ~120 km wide with adjacent

steep slopes deepening by 0.8-2.5 km over a distance of ~40 km. The axial plateau

shallows southward from ~1.8 km near 23°N to ~1.2 km near 18.5°N. It is

144

discontinuous, with three main valleys located around 20.2°N, 22°N, and 23°N. Due

to the low resolution of the magnetic basement grid, axial troughs revealed in the

seismic reflection data in Figure 5.4 are not shown within the axial plateau in Figure

5.6. On both sides of the axial plateau, the trend of basement deepening toward the

coasts changes at a distance of ~60 km from the ridge axis. The apparent basement

rises steeply landward from there to the coasts by up to 3 km, partly due to the

sedimentary sources mentioned earlier and partly due to a genuine rise in basement

(Shi et al., 2018). The shallow region next to the western coast is ~10 km wider than

that next to the eastern coast. Moreover, the basement along the ridge axis tends to

deepen by ~0.25 km from 18.5°N to 23°N (Figure 5.7).

Figure 5.6 Basement topography map derived from the aeromagnetic data Werner source

solutions. The map shows the elevation within each cell (with a size of 20 km × 10 km) at

which the CDF of magnetic source elevations within that cell equals 93.16% (Figure 5.5a).

Dashed cyan line locates the spreading axis. (The statistical method used here is described

and justified in section 3.2.4 and 5.3.2.3)

145

Figure 5.7 Magnetic basement depths along the spreading axis. Red dashed line is a least-

squares regression through the data. Magnetic basement tends to deepen by ~0.25 km from

18.5°N to 23°N.

5.4.2 Correlation between Bouguer gravity anomalies and magnetic basement

elevations

The Bouguer anomalies from Mitchell et al. (2017) are partly correlated with

the magnetic basement elevations (Figure 5.8). As there is a strong correlation

between Bouguer gravity anomalies (Mitchell et al., 2017) and basement seismic

reflection depths around the spreading axis (Shi et al., 2018), this indirectly suggests

that the magnetic basement elevations are reasonable. Both Bouguer anomalies and

the magnetic basement are elevated around the spreading centre (i.e., in the area

within ~ 60 km of the axis) and depressed in the region flanking the axial zone.

The apparent density contrast computed from the Bouguer-basement depth

correlation line (white dashed line in Figure 5.9a) based on the slab formula is 427 kg

m-3. This is a strong density contrast compared with values of 220-580 kg m-3 deduced

from similar gradients using seismic reflection depth by Shi et al. (2018). If a mean

density of 2,148 kg m-3 is used for the evaporite and sediment layers (Wheildon et al.,

1974), the apparent basement density is 2,575 kg m-3.

146

Figure 5.8 (a): Bouguer gravity anomalies from Mitchell et al. (2017) computed by removing


seabed topography. The data are contoured every 50 mGal. Black lines mark fracture zones

interpreted here from the gravity data. (b): Map of aeromagnetic-derived basement depths

(Figure 5.6) with shading from the Bouguer gravity grid of (a) with a light direction of N30°W

to highlight the fracture zones.

In Figure 5.9a, the diversity of magnetic source depths increases where the

Bouguer anomalies are small, particularly those below 30 mGal. These occur outside

the axial zone (Figure 5.8a). Although they could be erroneous, we speculate that this

could be caused by magnetic bodies (e.g., basaltic sills) in the evaporite and sediment

layers. Alternatively, they may be caused by crustal density or thickness variations, so

that, for example, basement of depths shallower than the regression line in Figure 5.

9a may overlie lower density or thicker crust.

Basement depths were predicted from the Bouguer gravity anomalies of

Mitchell et al. (2017) (Figure 5.8a) using the regression in Figure 5.9a. Figure 5.9b

shows the aeromagnetic-derived basement depths minus those predicted depths in

order to study the origins of deviations from the correlation in Figure 5.9a. Those

differences can be as large as 0.8 km and occur in coherent patches of a few cells

across outside the axial zone. They do not obviously relate to the fracture zones

147

(Figure 5.8a) suggesting that these large differences likely do not arise from varied

crustal thickness or density structure affecting the Bouguer anomalies.

Figure 5.9 (a): Correlation between Bouguer gravity anomalies from Mitchell et al. (2017)

(Figure 5.8a) and aeromagnetic-derived basement depths (in Figure 5.6) within 60 km of the

axis. Contours represent the number of points falling within each 0.02 km by 2 mGal block

148

(contour interval 150 counts). White dashed line is regression of the data. (b): Differences

between basement depths from Figure 5.6 and those predicted from Bouguer gravity

anomalies from Figure 5.8a using the regression in (a).

In areas that lie ~70–160 km from the ridge axis, the correlation between

Bouguer gravity anomalies and magnetic basement depths breaks down. The Bouguer

gravity anomaly stays subdued while the estimated magnetic basement shallows. We

suggest that this likely arises from magnetic minerals supplied from erosion of the

adjacent land areas. For example, magnetic basement is shallow offshore the Tokar

Delta, where the Plio-Pleistocene sediments in exploration wells reach 2 km in

thickness (Hughes and Beydoun, 1992). The westerly 10 km of seismic refraction

profile PIII of Egloff et al. (1991) closest to the delta revealed a 3-km thick lower

velocity body interpreted as terrigenous sediment. Arabian-African shield rocks on

the adjacent land contain ophiolites and, on the Arabian side, are overlain by volcanic

rocks (Stern and Johnson, 2010). Shallow magnetic basement around other parts of

the coasts therefore likely also represent the effects of terrigenous sediments

containing minerals with greater remanent and induced magnetization.

5.4.3 Number of magnetic source solutions

The derived number of magnetic source solutions should ideally reflect the

number of thin sheet-like bodies and edges of thick bodies, e.g., basaltic dykes and

edges between oceanic crust and continental or transitional crust. In Figure 5.10, the

number of source solutions per kilometre ranges from 9 to 15 near the spreading

centre and at a distance of ~60 km from the axis on both sides of the axial zone. In

other regions, the numbers are lower: 5 counts km-1 is common. The along-rift

continuity of high numbers around the axis and mid-way between axis and the coasts

ceases at ~21°N. This may be due to the fracture zone. Additionally, the numbers of

source solutions near the western coast are ~2 counts km-1 higher than numbers near

the eastern coast.

149

Figure 5.10 The numbers of magnetic sources revealed by the Werner deconvolution

normalized by the cumulative length of survey lines in each cell (cells as Figure 5.6).

5.5 Discussion

Shi et al. (2018) found an axial crustal high after correcting the seismically

derived basement depths for evaporite and other sediment isostatic loading, and

interpreted it as suggesting the central Red Sea is underlain by oceanic crust typical of

a mid-ocean ridge near to a mantle hotspot, like the Reykjanes Ridge. In this study,

we confirm that the basement topography in the region away from the seismic lines

also has an axial plateau within ~60 km of the axis. Shi et al. (2018) have discussed

how this axial high could have formed.

Along the ridge axis, the magnetic basement tends to be elevated from 23°N

to 18.5°N by ~0.25 km. Moreover, the plateau shallows on average southward from

~1.8 km around 23°N to ~1.2 km around 18.5°N (Figure 5.7). These variations are

consistent with declining influence of the Afar plume to the north.

150

5.5.1 How well do the magnetic basement topography and source solution numbers

correspond with other data?

Figure 5.11a shows the fracture zones interpreted from the Bouguer gravity

anomalies of Mitchell et al. (2017) and from magnetic, bathymetric, and seismic

reflection data by Izzeldin (1982, 1989) superimposed on the map of magnetic

basement topography. Fracture zones are usually characterised by valleys and

escarpments due to faulting and by thinner crust (e.g., Menard and Atwater, 1969;

Searle, 2013). In Figure 5.11a, although some discrepancies occur at 18.8°N and

20.5°N, the valleys found in the derived magnetic depths coincide with fracture zones

near the axis around 20.2°N, 22°N, and 23°N.


151

Figure 5.11 (a): Map of the basement topography of Figure 5.6 overlain with fracture zones

interpreted by Izzeldin (1982) (blue and white lines from magnetic and bathymetric data,

respectively). Black lines locate the fracture zones identified here from the gravity data in

Figure 5.8a. Cyan line locates the spreading axis. Black box locates the area of (b). (b): Map

locating the interpreted extent of evaporite and other sedimentary cover of Augustin et al.

(2014) overlain on the magnetic basement depths. White lines mark the boundary of

multibeam bathymetry survey area. Heavy black lines indicate the front of evaporite flows.

The prominent deeps in this region are labeled.

Evaporites including halite were widely deposited during the Miocene in the

Red Sea (Girdler and Whitmarsh, 1974; Stoffers and Kühn, 1974). Because of the weak

rheology of halite, the evaporites tend to flow downslope toward the axial valleys, due

152

to subsidence and as their lateral constraints are lost (Feldens and Mitchell, 2015;

Mitchell et al., 2017). From multibeam sonar data, Augustin et al. (2014) interpreted

the extents of evaporite and exposed volcanic basement. Figure 5.11b shows those

extents overlain on the map of magnetic basement depths. It shows that the

evaporites have extended into the axial valleys around 21.7°N and 23°N, where the

derived magnetic basement depth has cross-axis valleys with depths of ~1.2 km and

~1 km, respectively. Those two valleys (fracture zones) also correspond with the

inter-trough zones separating the deeps. This is consistent with the basement in the

inter-trough zones being deeper than that in the deeps and with the inter-trough

zones being merely areas where evaporites blanket the axial valley and obscure the

volcanic geomorphology (Augustin et al., 2014, 2016). The volcanic basement is

generally exposed where it is elevated. There is nevertheless some disagreement

between these data, e.g., the evaporite flow around 20.5°N does not correspond with

an axial valley, although there are valleys nearby.


153

Figure 5.12 The number of Werner sources from Figure 5.10 overlain with (a) the fracture

zones from Figure 5.11a and (b) the distribution of evaporites from Figure 5.11b.

We interpret the high numbers of magnetic source solutions near the axis

(Figures 5.10 and 5.12) as due to the dikes and lava flows, while those mid-way

between axis and the coasts are due to the edges between oceanic crust and

continental or transitional crust. The higher count in the axial zone could also be due

to the shallower and more exposed crust, hence a stronger signal than the other

regions. Seafloor spreading ridges are often offset by transform faults and fracture

zones (e.g., Schouten et al., 1985; Searle, 2013). In Figure 5.12a, the ends of short

axial zone segments generally coincide with the fracture zones interpreted previously

154

from other data. In Figure 5.12b, the presence of underlying cross-axis valleys

suggested by the evaporite flows at 20.5°N, 21.7°N, and 23°N coincide with low

numbers of magnetic sources. This may be because the Werner deconvolution

method missed some magnetic causative bodies in depressions, since high frequency

magnetic anomalies tend to be suppressed due to upward continuation effects.

5.5.2 The potential utility of magnetic source depth determination in the Red Sea

Although containing some noise of up to a few km in depth, as the Werner

deconvolution here reveals the main basement features seen in seismic, Bouguer

gravity, and bathymetry data, we suggest the derived magnetic basement has a

reasonable morphology about the axis.

Due to non-uniqueness effects and poorly matching of calculation window

sizes, some erroneous solutions are produced during the calculation.

There are many other magnetic interpretation techniques based on source

depth determination that could be used in the central Red Sea, e.g., Euler

deconvolution (Thompson, 1982), source parameter imaging method (Thurston and

Smith, 1997), and analytical signal method (MacLeod et al., 1993). Compared to

traditional magnetic forward and inverse modelling methods, those methods are

quick and have strong anti-noise properties, as they can well isolate the magnetic

anomaly from the noise.

5.6 Conclusions

To assess the basement geometry in the central Red Sea, we have inverted

aeromagnetic anomalies for source depths using Werner deconvolution. After

applying a statistical method to resolve the depths most likely to correspond with

seismic basement, the method effectively maps out the major topographic features

of the crustal basement near the axis.

The results confirm that the basement topography in the region has an axial

plateau within ~60 km of the axis, suggesting the entire axial zone including inter-

155

trough-zones in the central Red Sea is underlain by oceanic crust. The basement axial

plateau shallows southward, consistent with increasing influence of the Afar plume.

Magnetic basement elevation near the spreading axis generally co-varies with

Bouguer gravity anomalies. Their correlation implies a strong apparent density

contrast of 427 kg m-3 between basement and the overlying evaporites. The axial

valleys revealed in the basement coincide with fracture zones previously interpreted

from Bouguer gravity data of Mitchell et al. (2017) and from magnetic, bathymetry,

and seismic reflection data by Izzeldin (1982). Those valleys also correspond with

areas where the evaporites have invaded the axial valley, likely along fracture zones

(Augustin et al., 2014). Nearer to the coasts, the resolved magnetic basement is

shallower than seismic basement. We suggest this is due to the effects of magnetized

minerals in the sedimentary section. Additionally, we suggest the high numbers of

magnetic source solution near the axis are due to dikes and shallow and exposed crust,

whereas those located at a distance of ~60 km from the axis are due to the contact

between oceanic crust and continental or transitional crust. Overall, the exercise

illustrates the potential feasibility and applicability of magnetic source depth

determination in the central Red Sea and elsewhere where magnetized basement is

overlain by sediments lacking magnetization.

5.7 Acknowledgments

We thank the Arabian Geophysical and Surveying Company for the

aeromagnetic survey. We thank David Sandwell and Walter Smith for leading the

gravity and bathymetry mapping initiatives, and for the group involved in producing

the grids used in our study. Figures were prepared using GMT software (Wessel et al.,

2013). LMK was supported by a Royal Society of Edinburgh Personal Research

Fellowship funded by the Scottish Government. A Royal Society (International

Exchanges Scheme) grant to NCM enabled discussions with colleagues Nico Augustin

and Froukje van der Zwan that contributed to our conclusions.

156

5.8 References

Altherr, R., F. Henjes-Kunst, H. Puchelt, and A. Baumann, 1988, Volcanic activity in the

Red Sea axial trough—evidence for a large mantle diapir?: Tectonophysics, v.

150, p. 121-133.



p. 375-402.











479.



naturelle, v. 186, p. 567-606.



v. 38.







69.

157






Drake, C. L., and R. Girdler, 1964, A geophysical study of the Red Sea: Geophysical








v. 198, p. 329-353.

Feldens, P., and N. C. Mitchell, 2015, Salt Flows in the central Red Sea. In: Rasul, N.M.A.,

and Stewart, I.C.F. (Eds.), The Red Sea: The formation, morphology,

oceanography and environment of a young ocean basin, Springer Earth System

Sciences, Berlin Heidelberg, p. 205-218.






Geophys. Union, v. 67, p. 1208-1209.


Reviews, v. 45, p. 1-44.





158






12267-12279.

Hartman, R. R., D. J. Teskey, and J. L. Friedberg, 1971, A system for rapid digital

aeromagnetic interpretation: Geophysics, v. 36, p. 891-918.




Hughes, G., and Z. Beydoun, 1992, The Red Sea—Gulf of Aden: biostratigraphy,

lithostratigraphy and palaeoenvironments: Journal of Petroleum Geology, v.

15, p. 135-156.


Strasbourg.









Karner, G., N. Driscoll, and J. Peirce, 1991, Gravity and magnetic signature of Broken

Ridge, southeast Indian Ocean. Proceedings of the Ocean Drilling Program,





159






p. 513-524.




doi:10.1029/2012GC004155).




p. 1019-1022.


Pacific Ocean and Hawaii: Geology of North America, Geological Society of

America, Boulder, Colorado, v. N, p. 499-521.



MacLeod, I. N., K. Jones, and T. F. Dai, 1993, 3-D analytic signal in the interpretation of

total magnetic field data at low magnetic latitudes: Exploration Geophysics, v.

24, p. 679-688.



11, p. 431-441.

Martelet, G., J. Perrin, C. Truffert, and J. Deparis, 2013, Fast mapping of magnetic

basement depth, structure and nature using aeromagnetic and gravity data:

combined methods and their application in the Paris Basin: Geophysical

prospecting, v. 61, p. 857-873.


spreading in the Woodlark Basin: Indications of rift ‐ induced secondary

160


12909-12926.



Menard, H., and T. Atwater, 1969, Origin of fracture zone topography: Nature, v. 222,

p. 1037.



v. 29, p. 352-369.




Moreira, M., P. Valbracht, T. Staudacher, and C. Allègre, 1996, Rare gas systematics in

Red Sea ridge basalts: Geophysical Research Letters, v. 23, p. 2453-2456.







Sciences, v. 267, p. 205-217.

Plattner, C., R. Malservisi, T. H. Dixon, P. LaFemina, G. Sella, J. Fletcher, and F. Suarez-

Vidal, 2007, New constraints on relative motion between the Pacific plate and

Baja California microplate (Mexico) from GPS measurements: Geophysical



Jahrbuch, v. 13, p. 131-153.




Schouten, H., K. D. Klitgord, and J. A. Whitehead, 1985, Segmentation of mid-ocean

ridges: Nature, v. 317, p. 225-229.

161

Searle, R., 2013, Mid-ocean ridges, Cambridge University Press.







Stagg, H., J. Willcox, and D. Needham, 1989, Werner deconvolution of magnetic data:

Theoretical models and application to the Great Australian Bight: Journal of

the Geological Society of Australia, v. 36, p. 109-122.

Stern, R. J., and P. Johnson, 2010, Continental lithosphere of the Arabian Plate: a

geologic, petrologic, and geophysical synthesis: Earth-Science Reviews, v. 101,

p. 29-67.



Thakur, N., T. G. Rao, R. Khanna, and C. Subrahmanyam, 2000, Magnetic basement in

the Bay of Bengal through Werner deconvolution: Marine Geology, v. 162, p.

599-605.

Thompson, D., 1982, EULDPH: A new technique for making computer-assisted depth

estimates from magnetic data: Geophysics, v. 47, p. 31-37.

Thurston, J. B., and R. S. Smith, 1997, Automatic conversion of magnetic data to depth,

dip, and susceptibility contrast using the SPI (TM) method: Geophysics, v. 62,

p. 807-813.

Tivey, M. A., and J. Dyment, 2010, The magnetic signature of hydrothermal systems in

slow spreading environments. In: Rona, P.A., Devey, C.W., Dyment, J., Murton,

B.J. (Eds.) Diversity of Hydrothermal Systems on Slow Spreading Ocean Ridges,

American Geophysical Union, Geophysical Monograph Series, v. 188, p. 43-66.

Tivey, M. A., and H. P. Johnson, 1987, The central anomaly magnetic high: Implications

for ocean crust construction and evolution: Journal of Geophysical Research:

Solid Earth, v. 92, p. 12685-12694.



162

Tsokas, G., and R. Hansen, 1995, A comparison of inverse filtering and multiple source

Werner deconvolution for model archaeological problems: Archaeometry, v.

37, p. 185-193.


1415.

Volker, F., R. Altherr, K.-P. Jochum, and M. T. McCulloch, 1997, Quaternary volcanic

activity of the southern Red Sea: new data and assessment of models on

magma sources and Afar plume-lithosphere interaction: Tectonophysics, v.

278, p. 15-29.








Union, v. 94, p. 409-410.







Wilson, J. T., 1966, Did the Atlantic close and then re-open?: Nature, v. 211, p. 676-

681.



163

Chapter 6.

Paper 3: Oceanic basement roughness in the

central Red Sea

164

6. Paper 3: Oceanic basement roughness in the central

Red Sea

Wen Shi1*, Neil C. Mitchell1, Lara M. Kalnins2, Ian C.F. Stewart3, A.Y. Izzeldin4




EH9 3FE, UK.

3Stewart Geophysical Consultants Pty. Ltd., Adelaide, South Australia.




This paper is in preparation to be submitted to the journal Marine Geophysical

Researches.

165

Abstract

The Red Sea is a rare example of a continental rift proceeding to an oceanic

basin, but the nature of the crust underlying it (whether oceanic or highly extended

continental) has been controversial. We continue our investigation of crustal type

here by assessing the basement roughness (the root-mean-square variation of

basement relief) along profiles across and parallel to the spreading axis in the central

Red Sea. If the crust were oceanic, the across-axis roughness would be typical of

abyssal hill topography formed by faulting and volcanism, whereas the axis-parallel

roughness would mimic typical variations due to the ridge segmentation including

fracture zones. We estimated roughness values from depths of basement interpreted

from across-ridge seismic reflection profiles. The mean across-ridge roughness value

of 230 m is consistent with those observed over ultraslow and slow spreading ridges.

Basement roughness values along ridge-parallel profiles were computed from the

free-air gravity field using densities appropriate for oceanic crust and a modified

Bouguer slab formula, since suitable ridge-parallel seismic profiles are not available.

The derived roughness values around the axial trough are comparable with those of

slow spreading ridges such as the Mid-Atlantic Ridge. Systematic bias in the measured

roughness arising from the slab approximation was assessed using forward 2D

modelling and found to be ~30%. Correcting for this bias still leaves roughness values

within the range of values for oceanic crust. The axis-parallel roughness values change

mid-way between the coast and the rift axis, where a transition in crustal type from

stretched continental to predominantly oceanic has been suggested previously based

on the seismic reflection data. Although the basement roughness values by

themselves do not rule out extremely extended continental crust, combined with

other evidence they support an oceanic crustal interpretation.

Keywords: Central Red Sea, Basement roughness, Oceanic crust, Seismic reflection,

Potential field

166

6.1 Introduction

The Red Sea is currently transitioning from continental rifting to seafloor

spreading (Cochran and Martinez, 1988; Rihm and Henke, 1998). Identifiable ocean

floor magnetic anomalies suggest that seafloor spreading has been occurring in the

southern Red Sea since at least 5 Ma (Cochran, 1983; Girdler and Styles, 1974; Phillips,

1970; Vine, 1966). However, whether the crust in the central Red Sea (Figure 6.1) is

continental or oceanic has been debated (Bonatti, 1985; Ligi et al., 2012; Mitchell and

Park, 2014; Shi et al., 2018). Shi et al. (2018) found an oceanic-like axial crustal high

in the central Red Sea after correcting seismically derived basement depths for

isostatic loading by evaporites and other sediments. They suggested that the axis of

the central Red Sea rift is an ultra-slow spreading mid-ocean ridge affected by the Afar

hotspot, somewhat like the Reykjanes Ridge, an axial high affected by the Iceland

hotspot. If so, the entire central Red Sea is likely underlain by oceanic crust.

Basement roughness is defined as the root-mean-square deviation of residual

basement relief along a profile after removal of its systematic trend (Malinverno,

1991). The roughness can provide observational constraints on the nature of changes

in crustal thickness and tectonics (Ma and Cochran, 1997). Basement roughness has

been used in investigations of crustal structure, spreading rate, faulting models, and

ridge morphology at mid-ocean ridges (e.g., Bird and Pockalny, 1994; Malinverno and

Gilbert, 1989; Minshull, 1999; Sauter et al., 2018; Small, 1994). Sauter et al. (2011)

and Sauter et al. (2018) suggested that spreading rate, mantle temperature, and

lithosphere composition could affect the lithospheric strength and thus the basement

roughness.

In this study, we computed basement roughness values in the central Red Sea

along lines both parallel to the axis and across it in order to assess if those basement

roughness values compatible with those of other mid-ocean ridges. We then reassess

the other evidence supporting the oceanic crustal interpretation.

167


locate prominent deeps in the central Red Sea from Augustin et al. (2014) and Karbe (1987).

From north to south, these are (1) Nereus, (2) Thetis, (3) Hadarba, (4) Hatiba, (5) Atlantis II, (6)

Erba, (7) Port Sudan, (8) Suakin, and (9) Pelagia deeps. The relative plate motion vectors were

predicted based on the poles of Chu and Gordon (1998).

6.2 Tectonic setting

The Red Sea is a young ocean basin formed by the Arabian plate being split

from Nubian plate (Ghebreab, 1998; McKenzie et al., 1970). It is one of the few places

on Earth where a geologically recent or active transition from continental extension

to seafloor spreading can be observed (Cochran and Martinez, 1988; Rihm and Henke,

1998). The extension of the Red Sea may have first started in the Eocene and

developed substantially in the Oligocene at ~30 Ma (Bosworth and McClay, 2001;

168

Hofmann et al., 1997; Mohr, 1983; Omar and Steckler, 1995). The 0-3 Ma Red Sea

spreading rate increases southward from ~10 mm yr−1 at 25.5°N to ~16 mm yr−1 near

18°N (Chu and Gordon, 1998; DeMets et al., 1990; DeMets et al., 2010; Reilinger et al.,

2015).

The southern Red Sea (south of 19°N; Figure 6.1) has a well-developed

continuous axial zone typical of seafloor spreading, with volcanic geomorphology in

multibeam data, large amplitude magnetic anomalies and basaltic lavas recovered in

dredges as evidence of extensive volcanism (Augustin et al., 2014; Haase et al., 2000;

Phillips, 1970; Roeser, 1975). Ocean floor magnetic anomalies there are clearly

identifiable up to Chron 3, suggesting that full seafloor spreading started at least by 5

Ma (Cochran, 1983; Girdler and Styles, 1974; Phillips, 1970; Roeser, 1975; Vine, 1966).

Based on spreading rates of Chu and Gordon (1998) and locations of volcanic ridges,

Augustin et al. (2014, 2016) suggested that seafloor spreading began at 8–12 Ma, with

the oceanic crust created older than 5 Ma buried under the thick Miocene evaporites

(Girdler, 1984; Girdler and Whitmarsh, 1974).

A series of depressions described as ‘deeps’ occur in the central Red Sea

(between 19°N and 23°N; Figure 6.1) (Bonatti, 1985; Pautot et al., 1984). High

amplitude magnetic anomalies, volcanic geomorphology, and normal mid-ocean ridge

basalt (MORB) suggest that they are oceanic spreading centres (Augustin et al., 2014,

2016; Bonatti, 1985; Izzeldin, 1987; Pautot et al., 1984). However, the nature of crust

underlying the shallower inter-trough zones and crust off-axis covered by evaporites

and hemipelagic sediments has been more controversial. Some authors have

interpreted the low amplitude magnetic anomalies found in these areas as indicating

a highly extended continental crust (Bonatti, 1985; Cochran and Karner, 2007; Ligi et

al., 2011, 2012), whereas others have interpreted seismic velocity structures obtained

from limited seismic refraction data and segmented structure of the gravity field as

indicating oceanic crust while explaining the low amplitude magnetic anomalies by

various mechanisms (e.g., Augustin et al., 2014, 2016; Dyment et al., 2013; Egloff et

al., 1991; Girdler, 1985; LaBrecque and Zitellini, 1985; Mitchell and Park, 2014; Searle

and Ross, 1975; Shi et al., 2018; Tramontini and Davies, 1969).

169

In the northern Red Sea (north of 23°N; Figure 6.1), the ‘deeps’ are less

pronounced and become more widely spaced, although the basaltic intrusions lavas

have been recovered from them (Bonatti, 1985; Cochran, 2005; Guennoc et al., 1988;

Pautot et al., 1984). The presence of large fault blocks of continental crust has been

inferred from the gravity anomalies combined with seismic refraction data (Cochran

and Karner, 2007; Martinez and Cochran, 1988). In contrast, others have suggested

this region is also underlain by oceanic crust based on unpublished seismic reflection

and magnetic data (Dyment et al., 2013; Tapponnier et al., 2013). From

reconstructions of geological features across the Red Sea, Sultan et al. (1992, 1993)

suggested that the entire Red Sea basin is underlain by oceanic crust.


6.3.1 Multichannel seismic reflection

Multichannel seismic reflection data used in this study were collected in 1976

(Izzeldin, 1982, 1987), using a Vaporchoc source with a 2.4 km streamer consisting of

48 channels during deep-water surveying and with a 1.2 km streamer during shallow-

water surveying. The data were processed using a 24-fold stack, with stacking

velocities computed every 3.6 km along track. The locations of seismic reflection

profiles 7, 9, 11, 15, 17, 19, 21, 25, 27, 29, and 31 are shown in Figures 6.2a and 6.2b.

Basement depths were derived from these data as described by Shi et al.

(2018). They converted two-way travel times of the basement and seabed reflections

to depths below sea level using P-wave velocities (Vp) of 1.538 km s-1, 1.9 km s-1, and

4.21 km s-1 for water, Plio-Pleistocene sediments, and evaporites, respectively. These

velocities were chosen based on the earlier studies. The Vp of water was chosen

according to the empirical equations of Mackenzie (1981), with typical Red Sea salinity

of 40 ppt and temperature of 21 °C. The Vp of Plio-Pleistocene sediments was chosen

based on the measurements of DSDP Leg 23 samples (Whitmarsh et al., 1974) and the

results of seismic reflection and refraction surveys by Egloff et al. (1991) and Gaulier

et al. (1988). The Vp of evaporites was chosen based on seismic refraction data from

Tramontini and Davies (1969), Girdler and Whitmarsh (1974), and Egloff et al. (1991).

Figure 6.3 shows those depths derived from the seismic reflection data.

170


of multichannel seismic reflection profiles 7, 9, 11, 15, 17, 19, 21, 25, 27, 29, 31 from Izzeldin

(1982, 1987). (b): Locations of gravity profiles G1-32 and seismic reflection profiles as in (a).

Cyan line approximately locates the spreading axis. Green line locates gravity profile G21,

along which the gravity forward modelling results are shown in Figure 6.6. The study area was

divided into two areas: north of 20.25°N and south of 20.25°N, as the free-air gravity field

north of 20.25°N has cross-axis lineaments suggested to be due to oceanic spreading

segments (Mitchell, 2015).

171

Figure 6.3 Depths derived from the seismic reflection profiles of Izzeldin (1987). Line

numbers are shown in the lower right corner of each panel. Magnetic anomalies (purple lines)

along the seismic profiles are reduced-to-pole magnetic anomalies (Figure 6.10). Black lines

denote bathymetry (Smith and Sandwell, 1997, version 18.1). Dark green, cyan, and red lines

are the depths of the seabed, the S-reflection at the top of the Miocene evaporites, and the

basement, respectively. RTP: Reduced-to-pole.

172

The basement depths corrected for isostatic loading of the evaporites and

other sediments by Shi et al. (2018) were used here to estimate basement roughness

along the seismic profiles. To compute residual basement reliefs, we needed to

remove the systematic trend with basement age, but, in the central Red Sea, the crust

deepens too fast to be caused by normal thermal subsidence, likely an effect of varied

crustal thickness (Shi et al., 2018). Therefore, rather than use a theoretical subsidence

curve, we have removed a running median average. For example, the residual

basement reliefs in Figure 6.7c were obtained by removing the regional variation of

Figure 6.7b from the basement depths in Figure 6.7a.

The results are unfortunately non-unique; Figure 6.7d shows how the

basement roughness varies with filter width. However, after experimenting with

various filter widths, 60 km appears to do a good job of attenuating most of the abyssal

hill topography while still recording the regional variation (Figure 6.7b). Furthermore,

the roughness in Figure 6.7d varies gradually with filter width. In Figure 6.8, the value

obtained with the 60 km filter width is shown with the red star, whereas the error bar

represents the effect of filter width varying from 120 to 460 km.

6.3.2 Gravity anomalies

6.3.2.1 Free-air gravity data

We have used version 23 of the marine gravity field (referred to below as

“SSv23”) derived from satellite altimetry measurements by Sandwell et al. (2014). The

data are shown in Figure 6.2a.

Shipboard gravity data obtained from the National Centers for Environmental

Information (NCEI) (www.ngdc.noaa.gov/mgg) were used to evaluate the SSv23 data.

These data were collected on the RRS Shackleton in 1987 with a LaCoste and Romberg

gravity meter (Girdler and Southren, 1987) and on RV Robert Conrad during cruise

2507 in 1984 with a Bell BGM-3 gravity meter (Cochran and Martinez, 1988).

Some spikes occur in shipboard datasets, e.g., due to errors in Eötvös

correction and centripetal accelerations occurring with course changes (Mitchell, 2015;

Wessel and Watts, 1988). To reduce their effects, a 4-km running median average

173

filter was applied to the shipboard values before comparing the two datasets and

calculating the distribution of their differences shown in Figures 6.4 and 6.5.

Figure 6.4 Histograms of differences between the SSv23 and free-air anomaly data collected

on (a): RRS Shackleton and (b): RV Robert Conrad (SSv23 minus shipboard value). Vertical lines

show the means and standard deviations of the differences.

The differences between the SSv23 and the RRS Shackleton and RV Robert

Conrad data have standard deviations of 5.5 and 3.7 mGal (Figure 6.4). The RV Robert

Conrad data more closely follow SSv23 than the RRS Shackleton data, most likely

because of the superior Bell BGM-3 gravity meter used on the RV Conrad (Mitchell,

2015). Figure 6.5 shows a map of the differences. Blue areas indicate that SSv23 is -

10 mGal smaller than the shipboard values in the northernmost Red Sea, east side at

25-26 °N, centre at 24 °N, and centre at 20 °N. Their origins are difficult to identify.

Sandwell (pers. comm. 2013) suggested that these biases could be due to edge effects

from when the vertical offshore altimetry deflections were converting to gravity

anomalies (Mitchell, 2015). Nevertheless, the biases are small compared to the >100

mGal full range of the SSv23 gravity anomalies.

(a) (b)

174

Figure 6.5 Differences between the SSv23 gravity field and the shipboard (RRS Shackleton

and RV Conrad) gravity data (SSv23 minus shipboard) after the shipboard data were filtered

with a 4 km along-track median filter.

Free-air gravity anomalies were sampled from SSv23 along the ridge-parallel

gravity profiles G1-32 shown in Figure 6.2b. Because the free-air gravity field has

lineations crossing the Red Sea north of 20.25°N but not south of there (Mitchell,

2015), the profiles were each divided into two segments either side of 20.25°N. Those

segmented gravity profiles were then inverted for basement depth variations as

follows.

175

6.3.2.2 Bouguer slab formula

The gravity anomaly caused by a layer of infinite lateral extent and constant

thickness ℎ and density contrast ∆𝜌 can be computed using the Bouguer slab formula:

𝛿𝑔 = 2𝜋𝐺ℎ∆𝜌 (6.1)

where, 𝐺 is the gravitational constant.

To use equation (6.1) to invert the gravity data for basement relief, we

assumed a simplified structure of water underlain by evaporite and other sediments

of uniform density, in turn underlain by crust and mantle also each of uniform density.

As the lines are parallel to the axis, temperature in the upper mantle is expected to

vary only modestly along each line, and hence we can ignore its effect on the gravity

variations. In the central Red Sea, a 200-300 m thick layer of Plio-Pleistocene sediment

overlies the Miocene evaporites (Egloff et al., 1991; Izzeldin, 1987; Whitmarsh et al.,

1974). However, it is almost uniform in thickness (Mitchell et al., 2017; Ross and

Schlee, 1973) and has a similar density to the underlying evaporites (Mitchell et al.,

2010; Wheildon et al., 1974), so little error is introduced by using a single density for

both units. If the crust is oceanic, its upper layer (seismic layer 2) comprises low-

density lavas and dykes (Searle, 2013). We have chosen to ignore this layer in the

calculations and instead use a uniform gabbro density in the inversion because a

global study of seismic refraction data revealed that variations in crustal thickness

arise mainly from variations in the gabbro layer, while the lavas and dykes are more

uniform (Mutter and Mutter, 1993). Hence, its effect on the gravity variations should

be small. Similarly, the crust was also considered to be a layer of uniform density,

because seismic refraction data typically show little variation in seismic velocity

(Grevemeyer and Weigel, 1996), although this may be less true of some fracture zones

(White and Williams, 1986). The effect of topography of the Moho is ignored due to

its small density contrast and deep location. We return to some of these assumptions

later.

Using equation (6.1), the free-air anomaly variation along the axis-parallel

profiles can then be expressed in the following form:

176

𝑔𝑓𝑎𝑎 = 2𝜋𝐺[ℎ𝑤(𝜌𝑤 − 𝜌𝑐) + 𝑡𝑒𝑠(𝜌𝑒 − 𝜌𝑐)] + 𝑐 (6.2)

where 𝑔𝑓𝑎𝑎 is free-air anomaly, ℎ𝑤 is water depth, 𝑡𝑒𝑠 is total thickness of the

evaporites and other sediments, 𝜌𝑤 and 𝜌𝑐 are water and crustal densities, 𝜌𝑒 is the

mean density of the evaporite and sediment layers, and 𝑐 is a constant along each

ridge-parallel gravity profile.

A mean density of 2,148 kg m-3 was used for the evaporite and sediment layers

(Wheildon et al., 1974). The crust was assumed to have a density typical of oceanic

crust dominated by gabbro. A density of 2,957 kg m-3 was used for the oceanic crust

based on DSDP sample measurements of Hyndman and Drury (1977). A 1,020 kg m-3

density was used for the seawater.

By rearranging equation (6.2), we obtain:

𝑡𝑒𝑠 =𝑔𝑓𝑎𝑎−𝑐

2𝜋𝐺(𝜌𝑒−𝜌𝑐)− ℎ𝑤

𝜌𝑤−𝜌𝑐

𝜌𝑒−𝜌𝑐 (6.3)

Hence, the basement depth is:

ℎ𝑏 = ℎ𝑤 + 𝑡𝑒𝑠 =𝑔𝑓𝑎𝑎−𝑐

2𝜋𝐺(𝜌𝑒−𝜌𝑐)+ ℎ𝑤

𝜌𝑒−𝜌𝑤

𝜌𝑒−𝜌𝑐 (6.4)

Although absolute basement depths cannot be calculated merely from gravity

anomalies using equation (6.4) because 𝑐 is unknown, the basement depths in Figure

6.3 derived from seismic data were used to determine c for each profile.

The basement depths computed from equation (6.4) were then used to

estimate basement roughness along ridge-parallel gravity profiles, which are shown in

Figure 6.9a and 6.9b.

6.3.2.3 2D gravity forward modelling

Basement relief derived using the Bouguer slab formula will be

underestimated, because the assumption of an infinite slab is not fully met. We have

used gravity forward modelling to assess the bias magnitude, which can be used to

correct the roughness values. Forward models of free-air gravity anomalies were

177

computed along profiles G1-32 (Figure 6.2b) by summing the gravitational effects of

elementary mass rectangular cells.

The vertical gravitational attraction of a small rectangular cell is (Shengye and

Yuling, 2004):

∆𝑔 = 𝐺𝜌[(𝑥 + 𝑎)𝑙𝑛(𝑥+𝑎)2+𝐻2

(𝑥+𝑎)2+ℎ2 − (𝑥 − 𝑎)𝑙𝑛(𝑥−𝑎)2+𝐻2

(𝑥−𝑎)2+ℎ2 + 2𝐻 (𝑡𝑔−1 𝑥+𝑎

𝐻− 𝑡𝑔−1 𝑥−𝑎

𝐻) −

2ℎ(𝑡𝑔−1 𝑥+𝑎

ℎ− 𝑡𝑔−1 𝑥−𝑎

ℎ)] (6.5)

where 2𝑎 is plate width, ℎ is depth to upper boundary, and 𝐻 is depth to lower

boundary.

Based on the principle of superposition, the vertical gravitational attraction of

a geological body can be reproduced by the sum of the attractions of many small

individual rectangular cells constituting the body (Blakely, 1996). To forward model

the gravity anomalies, density structure derived from the along-axis profiles were

subdivided into 𝑛 small cells and equation (6.6) was applied to compute the sum of

their gravitational attractions:

𝑔𝑘 = ∑(∆𝑔)𝑖

𝑛

𝑖=1

(6.6)

where 𝑔𝑘 is the vertical gravitational attraction measured at the 𝑘𝑡ℎ measurement

point, (∆𝑔)𝑖 is the vertical gravitational attraction produced for the 𝑘𝑡ℎ measurement

point by the 𝑖𝑡ℎ small rectangular cell, (∆𝑔)𝑖 is computed from equation (6.5).

The density model in Figure 6.6a for gravity profile G21 was obtained using the

modified Bouguer slab formula (equation (6.4) first to estimate the

evaporite/basement depth, with bathymetry data used to constrain the top two

interfaces assuming constant thickness of Plio-Pleistocene (PP) sediments (water-PP

sediments and PP sediments-evaporites)). The same densities of seawater, evaporites,

and oceanic crust were used as in section 6.3.2.2. The forward model in Figure 6.6b

derived from the density structure in Figure 6.6a using equations (6.5) and (6.6) tends

to underestimate the extreme gravity anomalies (peaks and troughs). After scaling

the basement relief by a factor of 1.3 (i.e., maintaining the same mean depth but

increasing the rugosity about that mean depth), the calculated forward model does a

178

better job of reproducing the observed free-air anomalies. We concluded that the

systematic calculation bias arising from the use of the gravity slab formula is ~30%,

which is uniform along the central Red Sea gravity profiles as density interfaces show

no systematic variation with latitude. In order to get more accurate basement

roughness values from the slab formula, the roughness values were therefore

multiplied by a factor of 1.3, as shown by the red symbols in Figure 6.9.

Figure 6.6 Forward modelling of gravity profile G21. (a): Density model obtained from the

slab-formula-based results. The structure was subdivided into n=3600 small cells, and the

number of measurement points m is 720. (b): Forward 2D model predictions of free-air

gravity anomalies based on the slab formula results (red) underestimate the gravity anomaly

variation (blue line is free air anomaly from satellite altimetry). After multiplying the

basement relief variation by a factor of 1.3, the forward model (black line) better predicts the

observations. Based on this result, the systematic bias in basement roughness derived from

axis-parallel gravity profiles is 30%.

(a)

(b)

179


Version 18.1 of the Smith and Sandwell (1997) bathymetry data used in this

study is shown in Figure 6.1. These data were derived by combining shipboard depth

measurements with, in areas lacking shipboard depth measurements, variations

inferred from the effect of the seabed density contrast on free air anomalies, in turn

derived from satellite altimetry measurements. As the density of ship tracks along the

centre of the Red Sea is high, the high relief areas around the rift axis are mainly

constrained by survey data rather than gravity-based interpolation. Seabed depths

sampled along the gravity profiles were used in inverting basement depths with

equation (6.4).

6.4 Results

6.4.1 Basement roughness along across-ridge seismic profiles

The relationship between basement roughness and filter width is shown in

Figure 6.7d. The mean roughness value derived from data both north and south of

20.25°N increases from 120 m at the filter width of 10 km to 460 m at the filter width

of 240 km. For filter widths of less than 70 km, the mean roughness values in the

northern and southern regions are very close to each other. When the filter width is

greater than 70 km, the roughness value of the northern region is slightly higher than

that of the southern region. As the basement deepens to its deepest value at a 60 km

distance either side of the axis (Figure 6.7a), we suggest that 60 km is the most

appropriate filter width to use here and so the roughness value of 230 m is our best

estimate for the central Red Sea.

180

Figure 6.7 (a): Basement depths along the Red Sea seismic lines (Figures 6.2a and 6.2b)

corrected for the isostatic loading of evaporites and other sediments. (b): Regional systematic

trend estimated from (a) using a 60 km running median average filter. (c): Residual basement

reliefs obtained by removing the systematic trend in (b) from the isostatically adjusted

basement depths in (a). Grey shadings (within ~60 km of the axis) in (a), (b), and (c) show the

area Shi et al. (2018) interpreted as underlain by oceanic crust. (d): The relationship between

181

the mean basement roughness value and filter width. Blue and black lines represent

roughness values in northern (north of 20.25°N) and southern (south of 20.25°N) regions,

while red line represents the values derived from data of both regions combined. The

roughness value of 230 m at the filter width of 60 km is considered to be the most appropriate

value for the central Red Sea (see main text).

Figure 6.8 The relationship between the basement roughness and spreading rate for

ultraslow and slow spreading ridges, modified from Sauter et al. (2018). The error bar for the

Red Sea value indicates the range of roughness value derived from various filter widths in

Figure 6.7d. The other error bars represent the standard deviations of the corresponding

roughness values. SWIR: Southwest Indian Ridge, MCSC: Mid-Cayman Spreading Center, MAR:

Mid-Atlantic Ridge, SPR: South Pandora Ridge. Data for the Red Sea are determined in this

study. Data for Arctic ridge are from Weigelt and Jokat (2001) and Ehlers and Jokat (2009).

Data for SWIR are from Sauter et al. (2011) and Sloan et al. (2012). Data for MCSC are from

Sauter et al. (2018). Data for MAR are from Goff (1991), Goff et al. (1995), Neumann and

Forsyth (1995), Minshull (1999), and Lizarralde et al. (2004). Data for Sheba ridge are from

d'Acremont et al. (2010). Data for SPR are from Lagabrielle et al. (1996). The continuous black

line is the power-law of Malinverno (1991). The dashed blue and red lines are power-laws

derived by Ehlers and Jokat (2009) from basement topography adjacent to ridges with axial

valleys and with axial highs, respectively.

Figure 6.8 shows the relationship between the basement roughness and

spreading rate for the central Red Sea together with other ultraslow and slow

spreading ridges from Sauter et al. (2018). The roughness values at ultraslow

182

spreading ridges range from 100 to >500 m, while those at slow spreading ridges are

200-240 m. The Red Sea basement roughness value (230 m) is generally consistent

with these data. It lies near the curve of Malinverno’s (1991) global study of abyssal

hill roughness and near to the curve of Ehlers and Jokat (2009) of abyssal hills of near

hotspots. The 60 km filter width leaves some abyssal hill topography (Figure 6.7b),

suggesting it may be too small. Increasing the filter width would produce a higher

roughness value as shown by the error bar in Figure 6.8. This would still leave the

roughness within the range of values typical for an ultra-slow spreading ridge.

6.4.2 Basement roughness along the ridge-parallel gravity profiles

In Figure 6.9a and 6.9b, the basement roughness values in both areas range

from 200 m to 550 m and are comparable to those of the Mid-Atlantic Ridge

(horizontal pink lines in Figure 6.9a and 6.9b). The roughness values around the

spreading axis are higher than those near the coasts. Those values change about mid-

way between the coast and the spreading axis (40-80 km off-axis), coinciding with a

transition in crustal type from stretched continental to predominantly oceanic

interpreted by Izzeldin (1982, 1987) from varying character of the basement reflection

in regional seismic profiles and by Shi et al. (2018) from a change in Bouguer anomaly-

basement depth correlation.

183

Figure 6.9 Basement roughness values (green symbols) computed with a modified Bouguer

slab formula along axis-parallel gravity profiles G1-32 (Figure 6.2b) using a crustal density

(2,957 kg m-3) appropriate for oceanic crust dominated by gabbro. (a): North of 20.25°N. (b):

South of 20.25°N. Red symbols show the values corrected for bias arising from the slab

formula based on the results of gravity forward modelling (see section 3.2.3). Horizontal pink

lines represent basement roughness values calculated along lines parallel to the Mid-Atlantic

Ridge axis from bathymetry where the basement is only weakly sedimented (bathymetry data

derived from the Global Multi-Resolution Topography Synthesis (Ryan et al., 2009) over

latitudes 22°N to 32°N and 80 km off-axis, where the bathymetry has a similar segmentation

structure to the central Red Sea).

6.5 Discussion

In this study, we have shown that basement roughness values in the central

Red Sea are similar to those of the similarly ultraslow and slow spreading ridges,

although they by themselves do not rule out an extremely extended continental crust

184

interpretation. This is, however, consistent with the oceanic-like seismic velocity

structures in the central Red Sea (Davies and Tramontini, 1970; Egloff et al., 1991;

Tramontini and Davies, 1969) and an axial high similar to hotspot-affected ridges (Shi

et al., 2018) and supports the interpretation that rifting has proceeded to seafloor

spreading in the central Red Sea, as well as in the south.

After correcting seismically derived basement topography for isostatic loading

of evaporites and other sediments, Shi et al. (2018) revealed that the basement has

an axial high, as shown in Figure 6.7a. Such a high is not typically found at continental

rifts but is common for mid-ocean ridges near to mantle hotspots, such as the

Reykjanes Ridge near Iceland or Spiess Ridge on the similarly ultra-slow spreading

Southwest Indian Ridge near to the Bouvet hotspot.

Some other clues for assessing crust type in the central Red Sea are reassessed

as follows.

6.5.1 Reduced-to-pole magnetic anomalies

Figure 6.10 shows the aeromagnetic data originally presented by Izzeldin (1982,

1987, 1989) after reduction to pole. Izzeldin (1987) interpreted spreading anomalies

up to Chron 2A (3 Ma) north of 20˚N and up to Chron 3A (6 Ma) south of 20˚N from

these data. However, although less easily identified, linear anomalies also exist

further from the axis, and they lie parallel with and are symmetrical about the axis

(e.g., at 60 km from the axis in line 9). That symmetry and their lack of correlation

with basement topography (Figure 6.3) is a strong indication that they are seafloor-

spreading anomalies, as argued for similar anomalies in the southern Red Sea (Hall,

1989). The low magnetic amplitudes here could be due to the deeper basement,

alteration under the evaporites, overlapping flows and effects of crustal segmentation

(Augustin et al., 2014; Dyment et al., 2013; Izzeldin, 1987, 1989; LaBrecque and

Zitellini, 1985; Levi and Riddihough, 1986; Mitchell and Park, 2014; Searle and Ross,

1975).

185

Figure 6.10 Aeromagnetic anomalies of Izzeldin (1982, 1987) reduced to pole. Red lines

locate the seismic reflection profiles (Figure 6.2a). The data are contoured every 80 nT.

6.5.2 Along-axis gradients in mantle Bouguer anomalies

Mantle Bouguer anomalies (MBAs) are marine Bouguer anomalies in which the

gravitational effect of the cooling lithospheric plate is removed from free-air

anomalies as well as the effect of the seabed density contrast (Grindlay et al., 1998;

Lin and Morgan, 1992). As MBAs vary with varied crustal thickness and/or upper

mantle densities, they are commonly interpreted in terms of the three-dimensional

upwelling structure of mantle beneath ridges (Magde and Sparks, 1997). (As regions

of upwelling mantle are hot, and hence less dense, they reduce gravity anomalies. In

addition, crust is generally thicker above such regions because of greater melting,

which also reduces the MBA.) Wang and Cochran (1995) measured along-axis

gradients in MBA by dividing the peak to trough amplitude by distance and found them

to be ~0.1 mGal km-1 at mid-ocean ridges with axial highs including the slow-spreading

Reykjanes Ridge, while finding MBAs at slow spreading ridges that are not affected by

mantle hotspots to be higher (e.g., gradients along parts of the Mid-Atlantic Ridge

186

away from hotspots are 0.3-1.2 mGal km-1 and those along Southwest Indian Ridge

are 0.4– 0.7 mGal km-1) (Figure 6.11).

Figure 6.11 Comparison of along-axis gradients in mantle Bouguer anomalies (MBA) in the

central Red Sea with those at other mid-ocean ridges, modified from Wang and Cochran

(1993). RR: Reykjanes Ridge, CIR: Central Indian Ridge, SEIR: Southeast Indian Ridge, GSC:

Galápagos spreading center, EPR: East Pacific Rise.

We have computed differences in MBAs between the seismic reflection

profiles where they cross the spreading axis, using the depth of basement in those

profiles as a constraint, along with the seabed reflection where the axis is covered

with evaporites and other sediments. Mantle Bouguer anomalies along 11 seismic

profiles (Figure 6.3b) in the central Red Sea were computed by subtracting gravity

effects of evaporite-crust and crust-mantle interfaces from the marine Bouguer

gravity anomalies of Mitchell et al. (2017), assuming a uniform 7 km thick crust.

Densities of 2,148 kg m-3 (Wheildon et al., 1974), 2,957 kg m-3 (Hyndman and Drury,

1977), and 3,220 kg m-3 (Crough, 1983; Gvirtzman et al., 2016) were used for

evaporites, oceanic crust, and mantle, respectively. As the differences were

calculated for the spreading axis, with zero age crust, no lithospheric cooling

component was needed. Figure 6.11 shows those differences converted to along-axis

gradients in MBA, which range from 0.02 to 0.3 mGal km-1. Given that the seismic

lines are not optimally located for sampling the largest along-axis MBA gradients (the

187

seismic lines are spaced typically ~50 km (Figure 6.2), whereas peaks and troughs in

MBA profiles can be 20 km or less (Magde and Sparks, 1997), the 0.3 mGal km-1 value

may be the most representative of the steepest values.

This 0.3 mGal km-1 value is larger than the 0.1 mGal km-1 value of Wang and

Cochran (1995) for the Reykjanes Ridge, suggesting that crustal thickness is less

uniform along-axis in the central Red Sea (Bell and Buck, 1992). This may suggest the

axial crust is not as hot and/or thick as at the Reykjanes Ridge. The geochemistry of

basaltic lavas (Na8.0) also suggest that the present axis has a normal or only modestly

thick crust (Haase et al., 2000), and hence is less likely to have redistributed crustal

material by the mechanism of Bell and Buck (1992). Although these various lines of

evidence for a somewhat cold axial lithosphere would seem to be incompatible with

an axial high, Shi et al. (2018) suggested this could be explained by crustal thickening

from an early rifting stage which produced less magma and thinner than average

oceanic crust. Furthermore, seismic tomography results indicate that the hot material

from the Afar plume has mainly propagated under Arabia rather than the central Red

Sea (Chang et al., 2011).

6.6 Conclusion

To further assess the nature of crust in the central Red Sea, we computed

basement roughness values in profiles both across and parallel to the axis and

compared them with those observed at ultraslow and slow spreading ridges.

Basement roughness values from axis-crossing seismic data are ~230 m, similar to the

values observed at other ultraslow and slow spreading ridges (e.g., the Mid-Atlantic

Ridge, the Sheba Ridge, and the Southwest Indian Ridge). The roughness values

derived from axis-parallel profiles of the gravity field (200-550 m) are comparable with

those of the Mid-Atlantic Ridge where it has similar along-axis segmentation. A

change in those roughness values roughly mid-way between the coast and the axis

may mark the transition in crustal type from stretched continental to predominantly

oceanic. Although these basement roughness values by themselves do not exclude an

extended continental crust interpretation in the central Red Sea, they are supportive

of an oceanic crustal interpretation when considered along with other evidences,

188

including oceanic-like seismic velocity structures, crustal axial high, reduced-to-pole

magnetic anomalies, and along-axis gradients in mantle Bouguer anomalies.

6.7 Acknowledgments



in our study. Figures were prepared using GMT software (Wessel et al., 2013). LMK

was supported by a Royal Society of Edinburgh Personal Research Fellowship funded

by the Scottish Government. A Royal Society (International Exchanges Scheme) grant

to NCM enabled discussions with colleagues Nico Augustin and Froukje van der Zwan

that contributed to our conclusions.

6.8 References







Bell, R., and W. Buck, 1992, Crustal control of ridge segmentation inferred from

observations of the Reykjanes Ridge: Nature, v. 357, p. 583-586.

Bird, R. T., and R. A. Pockalny, 1994, Late Cretaceous and Cenozoic seafloor and

oceanic basement roughness: Spreading rate, crustal age and sediment

thickness correlations: Earth and planetary science letters, v. 123, p. 239-254.

Blakely, R. J., 1996, Potential theory in gravity and magnetic applications, Cambridge

University Press.





naturelle, v. 186, p. 567-606.

189







69.

Cochran, J. R., 2005, Northern Red Sea: nucleation of an oceanic spreading center

within a continental rift: Geochemistry, Geophysics, Geosystems, v. 6, Q03006,

doi:10.1029/2004GC000826.





d'Acremont, E., S. Leroy, M. Maia, P. Gente, and J. Autin, 2010, Volcanism, jump and

propagation on the Sheba ridge, eastern Gulf of Aden: segmentation evolution

and implications for oceanic accretion processes: Geophysical Journal















v. 198, p. 329-353.

190

Ehlers, B.-M., and W. Jokat, 2009, Subsidence and crustal roughness of ultra-slow

spreading ridges in the northern North Atlantic and the Arctic Ocean:






Reviews, v. 45, p. 1-44.

Girdler, R., 1984, The evolution of the Gulf of Aden and Red Sea in space and time:

Deep Sea Research Part A. Oceanographic Research Papers, v. 31, p. 747-762.

Girdler, R., 1985, Problems concerning the evolution of oceanic lithosphere in the

northern Red Sea: Tectonophysics, v. 116, p. 109-122.

Girdler, R., and T. Southren, 1987, Structure and evolution of the northern Red Sea.


7-11.





Goff, J. A., 1991, A global and regional stochastic analysis of near‐ridge abyssal hill

morphology: Journal of Geophysical Research: Solid Earth, v. 96, p. 21713-

21737.

Goff, J. A., B. E. Tucholke, J. Lin, G. E. Jaroslow, and M. C. Kleinrock, 1995, Quantitative

analysis of abyssal hills in the Atlantic Ocean: A correlation between inferred

crustal thickness and extensional faulting: Journal of Geophysical Research:

Solid Earth, v. 100, p. 22509-22522.

Grevemeyer, I., and W. Weigel, 1996, Seismic velocities of the uppermost igneous

crust versus age: Geophysical Journal International, v. 124, p. 631-635.

Grindlay, N. R., J. A. Madsen, C. Rommevaux-Jestin, and J. Sclater, 1998, A different

pattern of ridge segmentation and mantle Bouguer gravity anomalies along the

191

ultra-slow spreading Southwest Indian Ridge (15° 30′ E to 25° E): Earth

and Planetary Science Letters, v. 161, p. 243-253.











12267-12279.





ultramafic rocks from DSDP Leg 37: Aumento, F., Melson, WG, et al., Init. Repts.

DSDP, v. 37, p. 395-401.


Strasbourg.











p. 513-524.

192

Lagabrielle, Y., E. Ruellan, M. Tanahashi, J. Bourgois, G. Buffet, G. de Alteriis, J. Dyment,

J. Goslin, E. Gràcia-Mont, and Y. Iwabushi, 1996, Active oceanic spreading in

the northern North Fiji Basin: results of the NOFI cruise of R/V L'Atalante

(Newstarmer Project): Marine Geophysical Researches, v. 18, p. 225-247.

Levi, S., and R. Riddihough, 1986, Why are marine magnetic anomalies suppressed

over sedimented spreading centers?: Geology, v. 14, p. 651-654.




doi:10.1029/2012GC004155).




p. 1019-1022.

Lin, J., and J. P. Morgan, 1992, The spreading rate dependence of three‐dimensional

mid‐ocean ridge gravity structure: Geophysical Research Letters, v. 19, p. 13-

16.

Lizarralde, D., J. B. Gaherty, J. A. Collins, G. Hirth, and S. D. Kim, 2004, Spreading-rate

dependence of melt extraction at mid-ocean ridges from mantle seismic

refraction data: Nature, v. 432, p. 744.

Ma, L. Y., and J. R. Cochran, 1997, Bathymetric roughness of the Southeast Indian

Ridge: Implications for crustal accretion at intermediate spreading rate mid‐

ocean ridges: Journal of Geophysical Research: Solid Earth, v. 102, p. 17697-

17711.



Magde, L. S., and D. W. Sparks, 1997, Three‐dimensional mantle upwelling, melt

generation, and melt migration beneath segment slow spreading ridges:


Malinverno, A., 1991, Inverse square-root dependence of mid-ocean-ridge flank

roughness on spreading rate: Nature, v. 352, p. 58-60.

193

Malinverno, A., and L. E. Gilbert, 1989, A stochastic model for the creation of abyssal

hill topography at a slow spreading center: Journal of Geophysical Research:

Solid Earth, v. 94, p. 1665-1675.



150, p. 1-31.



Minshull, T., 1999, On the roughness of Mesozoic oceanic crust in the western North

Atlantic: Geophysical Journal International, v. 136, p. 286-290.







v. 29, p. 352-369.

Mitchell, N. C., M. Ligi, V. Ferrante, E. Bonatti, and E. Rutter, 2010, Submarine salt

flows in the central Red Sea: Geological Society of America Bulletin, v. 122, p.

701-713.




Mutter, C. Z., and J. C. Mutter, 1993, Variations in thickness of layer 3 dominate

oceanic crustal structure: Earth and Planetary Science Letters, v. 117, p. 295-

317.

Neumann, G. A., and D. W. Forsyth, 1995, High resolution statistical estimation of

seafloor morphology: Oblique and orthogonal fabric on the flanks of the Mid-

Atlantic Ridge, 34–35.5 S: Marine Geophysical Research, v. 17, p. 221-250.



194





Sciences, v. 267, p. 205-217.

Reilinger, R., S. McClusky, and A. ArRajehi, 2015, Geodetic constraints on the

geodynamic evolution of the Red sea, in Rasul, N., and Stewart, I. C., eds., The

Red Sea: The formation, morphology, oceanography and environment of a

young ocean basin, Springer, p. 135-149.



Basins Red Sea-Gulf of Aden, Springer, p. 29-49.


Jahrbuch, v. 13, p. 131-153.



Ryan, W. B., S. M. Carbotte, J. O. Coplan, S. O'Hara, A. Melkonian, R. Arko, R. A. Weissel,

V. Ferrini, A. Goodwillie, and F. Nitsche, 2009, Global multi ‐ resolution

topography synthesis: Geochemistry, Geophysics, Geosystems, v. 10.

Sandwell, D., E. Garcia, K. Soofi, P. Wessel, M. Chandler, and W. H. Smith, 2013,

Toward 1-mGal accuracy in global marine gravity from CryoSat-2, Envisat, and

Jason-1: The Leading Edge, v. 32, p. 892-899.




Sauter, D., H. Sloan, M. Cannat, J. Goff, P. Patriat, M. Schaming, and W. R. Roest, 2011,

From slow to ultra-slow: How does spreading rate affect seafloor roughness

and crustal thickness?: Geology, v. 39, p. 911-914.

Sauter, D., J. Tugend, M. Gillard, M. Nirrengarten, J. Autin, G. Manatschal, M. Cannat,

S. Leroy, and M. Schaming, 2018, Oceanic basement roughness alongside

magma-poor rifted margins: insight into initial seafloor spreading: Geophysical


195

Searle, R., 2013, Mid-ocean ridges, Cambridge University Press.



Shengye, Z., and P. Yuling, 2004, The principle of applied geophysics, China University

of Geosciences Press.



Sloan, H., D. Sauter, J. A. Goff, and M. Cannat, 2012, Abyssal hill characterization at

the ultraslow spreading Southwest Indian Ridge: Geochemistry, Geophysics,

Geosystems, v. 13.

Small, C., 1994, A global analysis of mid-ocean ridge axial topography: Geophysical




Sultan, M., R. Becker, R. Arvidson, P. Shore, R. Stern, Z. El Alfy, and R. Attia, 1993, New

constraints on Red Sea rifting from correlations of Arabian and Nubian

Neoproterozoic outcrops: Tectonics, v. 12, p. 1303-1319.



593-596.




(T12B-04).




1415.

Wang, X., and J. R. Cochran, 1995, Along-axis gravity gradients at mid-ocean ridges:

Implications for mantle flow and axial morphology: Geology, v. 23, p. 29-32.

196

Weigelt, E., and W. Jokat, 2001, Peculiarities of roughness and thickness of oceanic

crust in the Eurasian Basin, Arctic Ocean: Geophysical Journal International, v.

145, p. 505-516.

Wessel, P., and A. B. Watts, 1988, On the accuracy of marine gravity measurements:

Journal of Geophysical Research, v. 93, p. 393-413.





White, R. S., and C. A. Williams, 1986, Oceanic fracture zones: Journal of the Geological

Society, v. 143, p. 737-741.



197

Chapter 7.

Synthesis

198

7. Synthesis

7.1 Crust types in the central Red Sea

7.1.1 Classifications

In this study, we assessed crust types and their basement geometries in the

central Red Sea using multichannel seismic reflection data, ship-board magnetic data,

aeromagnetic data, and satellite gravity data.

Figure 7.1a shows our interpreted crust types. The presence of oceanic crust

was interpreted based on axial crustal high revealed in seismically derived basement

topography corrected for evaporite and other sediment loading in Chapter 4 (Figure

4.5), an axial high similarly in basement topography from Werner deconvolution of

aeromagnetic data in Chapter 5 (Figure 5.6), and basement roughness in Chapter 6

(Figure 6.9), which is comparable with roughness of ultra-slow spreading mid-ocean

ridges. Transitional crust lies between that oceanic crust and the extended

continental crust nearer to the two coasts of the Red Sea. Locating such transitional

crust is difficult and commonly is located differently by different researchers working

on the same margins (e.g., Eagles et al., 2015). We here locate uncertain transitional

crust where the Bouguer-basement depth correlations break (Chapter 4; Figure 4.6).

The loss of this correlation towards the coasts implies a lower net density underlying

the basement, hence more likely continental crust. Furthermore, we have found a

change in basement roughness implied by along-axis roughness of gravity anomalies

(Chapter 6; Figure 6.9), suggesting a decrease in relief due to oceanic fracture zones.

The narrow zones around the basement troughs where the trend changes from

decreasing to increasing towards the coast are interpreted as oceanic–continental

transitions (OCTs) (see also Figure 7.2). In Figure 7.1a, the lineaments interpreted by

Izzeldin (1987) from gravity, magnetic, and bathymetric data of Izzeldin (1987) are

probably fracture zones.

199

Figure 7.1 (a): Geological map of the central Red Sea based on the interpretation principles

given in main text. Cross-axis lineaments are from Izzeldin (1987). (b): A comparison of crust

type classification in (a) with that of Izzeldin (1987).

200

In Figure 7.1b, we compare our crust type classification with that of Izzeldin

(1987), who interpreted continental crust primarily from a change in the topographic

character of basement seismic reflections. In Figure 7.1b, the locations of oceanic and

uncertain transitional crust interpreted in our study are generally consistent with his.

Thus, the transitions we have identified also coincide with the reflection character

changes Izzeldin (1987) identified.

Figure 7.2 Interpretations of crust types along seismic lines 17 and 21. OCT: oceanic–

continental transition.

In Figures 7.1a and 7.2, the edges of oceanic crust tend to be symmetrical

about the axis, suggesting a symmetrical spreading in the central Red Sea.

In Figures 7.3a, 7.3b and 7.3c, Free-air gravity anomalies (Figures 3.1 and 3.9a),

Bouguer gravity anomalies (Figure 3.10) and RTP aeromagnetic anomalies (Figure 3.8)

are overlain on our crust type classification (Figure 7.1a). The high gravity anomalies

(Figures 7.3a and 7.3b) and high magnetic anomaly amplitudes (Figure 7.3c) near the

axis are in line with the basement axial high. Over our interpreted extended

continental crust towards the coasts, the gravity and magnetic anomies are generally

subdued (Figures 7.3a, 7.3b and 7.3c). Over the OCT areas (Figures 7.3a and 7.3b), the

Free-air and Bouguer gravity anomalies are generally also subdued. Within these OCT

areas, they can be observed some linear magnetic anomalies that are aligned with the

axis and occurring on both sides of the axis (Figure 7.3c). We interpret these as caused

201

by volcanic dyke swarms intruded parallel to the axis or extrusives associated with

them, possibly during the final breakup stage.

202

Figure 7.3 Comparisons of crust type classification in Figure 7.1a with (a): Free-air gravity

anomalies (Sandwell et al., 2014, version 23.1) contoured every 20 mGal, (b): Bouguer gravity

anomalies from Mitchell et al. (2017) contoured every 50 mGal, and (c): RTP aeromagnetic

anomalies contoured every 80 nT. (d): A comparison of Bouguer gravity anomalies in (b) with

RTP aeromagnetic anomalies in (c).

203

In Figure 7.3d, RTP aeromagnetic anomalies (Figure 3.8) are overlain on the

Bouguer gravity anomalies (Figure 3.10). Figure 7.3d shows that the cross-axis

magnetic highs generally overlie the Bouguer gravity highs. At 21.5°N and 23°N, both

the cross-axis magnetic and Bouguer gravity highs are subdued (see also Figures 7.3b

and 7.3c), due to the fracture zones. There are some linear high magnetic anomalies

in the OCT areas, over which the Bouguer gravity anomalies are low. The low Bouguer

gravity anomalies reflect deep basements. The magnetic anomalies here lack of

correlation with basement topography. This could indicate that they are seafloor-

spreading magnetic anomalies, as argued for similar anomalies in the southern Red

Sea (Hall, 1989). Here, we suspect that these high magnetic anomalies may be

produced either by dykes intruding the last stretched continental crust of the OCT or

youngest oceanic crust.

7.1.2 Character of the axial high

We suggested that the axial high was created by the increased melt supply and

enhanced crustal thickness (Shi et al., 2018). The axial relief in the central Red Sea

(0.8-1.6 km) is similar to that at the ultra-slow spreading Spiess Ridge (full rate, ~12.8

mm yr−1 (DeMets et al., 1994)) near the Bouvet hotspot (Mitchell and Livermore, 1998),

but larger than those at the slow-spreading Reykjanes Ridge near Iceland (0.6–1.0 km)

(Figure 4.5) and the intermediate rate Galápagos Spreading Centre near the Galápagos

hotspot (Blacic et al.,2008). This could be explained by the colder, stronger lithosphere

close to the ridge in the Red Sea due to the greater distance from the hotspot and the

slower spreading rate. The distances between the spreading ridges and their

associated hotspots are ~1,000 km for the central Red Sea ridge, ~300 km for the

Reykjanes Ridge, and ~100 km for the Galápagos Spreading Centre. The full spreading

rates of the central Red Sea, the Reykjanes Ridge, and the Galápagos Spreading Centre

are ~14 mm yr−1 (Chu and Gordon, 1998; DeMets et al., 2010), ~20 mm yr−1 (DeMets

et al., 2010; Talwani et al., 1971), and 55 mm yr−1 (DeMets et al., 1990), respectively.

As shown in Figure 7.1a, the width of Red Sea oceanic crust decreases

northwards from ~140 km at 19°N to ~100 km at 21°N, and then it increases to ~120

km at 23°N. Although oceanic crust is a bit narrower in the north than in the south,

its width does not vary systematically with distance from the Afar region.

204

The axial high is found along the whole central Red Sea, a distance of ~725 km.

South of our study area, the axial high probably extends southwards to around 16°N

in the southern Red Sea, because the Bouguer anomalies are similarly elevated and

this region is also undergoing ultra-slow seafloor spreading (e.g., Cochran, 1983;

Girdler and Styles, 1974; Phillips, 1970; Vine, 1966) with underlying hot mantle (e.g.,

Chang et al., 2011; Park et al., 2007, 2008).

7.1.3 Distinct domains from north to south

Cross-axis gravity lineations crossing the Red Sea occur in the region north of

20.25°N but not south of there (Mitchell, 2015). Thus, the central Red Sea can be

divided into two distinct domains as shown in Figure 6.2. The basement roughness

measured along strike is larger north of 20.25°N (with a mean value of ~350 m) than

it is south of 20.25°N (with a mean value of ~ 290 m) (Figure 6.8). This implies that

fracture zones have only developed north of 20.25˚N.

7.2 Moho depth and crustal thickness

None of the data in the 11 seismic reflection profiles in this study reveal the

Moho. To investigate the crustal thickness, geometry and type across the Red Sea,

Moho depths along seismic refraction profile SO53-PIII of Egloff et al. (1991) and

profile A-B of Salem et al. (2013) were plotted together with the basement reflections

along seismic profile 25 (Figure 7.4).

205


of multichannel seismic reflection profiles 7, 9, 11, 15, 17, 19, 21, 25, 27, 29, 31 of Izzeldin

(1987), seismic refraction profile SO53-PIII of Egloff et al. (1991), and profile A-B of Salem et

al. (2013). (b): A composite profile produced by projecting the seafloor and basement

206

reflections along profile 25 of Izzeldin (1987) on onto profile PIII of Egloff et al. (1991). (“Pre-

evaporites” are pre-evaporite sedimentary rocks.) (c): Graph showing basement reflection

along profile 25 and Moho along profiles PIII and A-B. OBS: Ocean bottom seismographs.

Profile 25 is the only seismic profile near profiles PIII and A-B (Figure 7.4a). In

Figure 7.4b, the basement top from Egloff et al. (1991) (blue) is elevated above that

from seismic profile 25 (red) at distances -60– -45 km. This may be explained by the

interface between the two layers marking a velocity change within igneous basement

rather than the basement-evaporite interface exactly (Mitchell et al., 2010), since the

seismic velocity of evaporites (~4.2 km s-1) overlaps with seismic velocities of volcanic

extrusives (Hammer et al., 1994; Mitchell, 2001). The difference in the basement

depth within the distance of -20–20 km is probably due to the different locations of

profiles 25 and PIII.

In Figure 7.4c, necking zones are probably located around the distance of ~53

km on either side of the axis, where the crust is thinnest. This is consistent with the

seismic interpretations of Egloff et al. (1991), which also suggested that the

continental crust is thinned from the coast up until ~53 km from the axis (their OBS 9

position) (Figure 7.4b). As demonstrated by profiles 25 and A-B, the trend of crustal

thickening occurs on both sides of the axis, suggesting a symmetric crustal structure

in the central Red Sea. The symmetry of structures in the gravity and magnetic data

(Figure 7.3) also suggest that the rift was symmetric not asymmetric in this part of the

Red Sea.

Based on the above studies, across-ridge crustal structures in the central Red

Sea are sketched in Figure 7.5.

207

Figure 7.5 A sketch showing across-ridge crustal structures in the central Red Sea.

7.3 Uncertainty in the seismically derived basement depth

As discussed in Chapter 3 (3.1), the main uncertainty in the seismically derived

basement depth and the thickness of evaporates originates from the P-wave velocity

(Vp) used for the evaporites. In Chapter 3, P-wave velocities of 3 km s-1 and 4.21 km

s-1 for evaporites were used to compute the minimum and maximum salt thickness

models (Figure 3.2). If evaporite Vp is 3 km s-1 rather than 4.21 km s-1, the axial relief

reduces by about 40% (Figure 3.2), the axial relief (corrected for the isostatic loading)

would be 0.5-1.0 km rather than 0.8-1.6 km. However, even for the smaller axial relief

estimates, the axial high and the rising in basement towards the coasts are not

eliminated, so this uncertainty does not affect the interpretations of axial high and

ocean-continent boundaries. Lateral variations in evaporite Vp could distort the

seismically derived basement depth. However, we suggest this variation is minor,

since layering within the evaporites, which is an indicator of the presence of non-halite

components with lower velocities, does not vary along-axis (Izzeldin, 1987).

7.4 Acknowledgments



in our study. Figures were prepared using GMT software (Wessel et al., 2013). We

208

thank Ian C.F. Stewart for providing reduced-to-pole aeromagnetic anomaly grid. A

Royal Society (International Exchanges Scheme) grant to NCM enabled discussions

with colleagues Nico Augustin and Froukje van der Zwan that contributed to our

conclusions.

7.5 References




Blacic, T. M., G. Ito, A. K. Shah, J. P. Canales, and J. Lin, 2008, Axial high topography

and partial melt in the crust and mantle beneath the western Galápagos

Spreading Center: Geochemistry, Geophysics, Geosystems, v. 9 (Paper Q12005,

doi:10.1029/2008GC002100 ).

Bosworth, W., P. Huchon, and K. McClay, 2005, The Red Sea and Gulf of Aden basins:

Journal of African Earth Sciences, v. 43, p. 334-378.



v. 38.







69.

DeMets, C., R. G. Gordon, D. F. Argus, and S. Stein, 1990, Current plate motions:


DeMets, C., R. G. Gordon, D. F. Argus, and S. Stein, 1994, Effect of recent revisions to

the geomagnetic reversal time scale on estimates of current plate motions:

Geophysical research letters, v. 21, p. 2191-2194.



209







v. 198, p. 329-353.






Geophys. Union, v. 67, p. 1208-1209.


7-11.



12267-12279.

Hammer, P. T., L. M. Dorman, J. A. Hildebrand, and B. D. Cornuelle, 1994, Jasper

Seamount structure: Seafloor seismic refraction tomography: Journal of







150, p. 1-31.

Mitchell, N. C., 2001, Transition from circular to stellate forms of submarine volcanoes:




210





v. 29, p. 352-369.

Mitchell, N. C., M. Ligi, V. Ferrante, E. Bonatti, and E. Rutter, 2010, Submarine salt

flows in the central Red Sea: Geological Society of America Bulletin, v. 122, p.

701-713.

Mitchell, N. C., and R. A. Livermore, 1998, Spiess Ridge: An axial high on the slow

spreading Southwest Indian Ridge: Journal of Geophysical Research: Solid

Earth, v. 103, p. 15457-15471.





Q06021, doi:10.1029/2006GC001566).




doi:10.1029/2007GC001895).



Sciences, v. 267, p. 205-217.

Salem, A., C. Green, S. Campbell, J. D. Fairhead, L. Cascone, and L. Moorhead, 2013,

Moho depth and sediment thickness estimation beneath the Red Sea derived

from satellite and terrestrial gravity data: Geophysics, v. 78, p. G89-G101.






211





593-596.

Talwani, M., C. C. Windisch, and M. G. Langseth, 1971, Reykjanes ridge crest: A

detailed geophysical study: Journal of Geophysical Research, v. 76, p. 473-517.




(T12B-04).




1415.



Union, v. 94, p. 409-410.

212

Chapter 8.

Conclusion and future work

213

8. Conclusion and future work

8.1 Conclusions

Understanding the processes that occur during the transition from continental

rifting to seafloor spreading is important in exploring how the Earth works and has

been a major challenge of Plate Tectonics. The Red Sea provides an important

example of transition from nearly orthogonal slow continental rifting to seafloor

spreading. Seafloor spreading in the southern Red Sea began at least by 5 Ma (e.g.,

Cochran, 1983; Girdler and Styles, 1974; Phillips, 1970; Roeser, 1975; Vine, 1966). But

how far the central Red Sea is through the transition to full seafloor spreading has

been controversial and is still open to debate. Contributing to this debate, this study

assessed the crustal type (whether the crust is continental or oceanic) in the central

Red Sea by evaluating the basement geometry and roughness.

We corrected for effects of overlying evaporite and other sediments to

reconstruct basement geometry from 11 seismic reflection lines. The seismically

derived basement depths corrected for evaporite and other sediment loading reveal

an axial high. Such a high is not typically found at continental rifts but is common for

mid-ocean ridges near to mantle hotspots, such as the Reykjanes Ridge near Iceland

or Spiess Ridge on the similarly ultra-slow spreading Southwest Indian Ridge near to

the Bouvet hotspot. Its relief of ~1 km relative to a background subsidence trend is

within the observed range. It is similar to that at Spiess Ridge, larger than that at

Reykjanes Ridge, but smaller than that of the Mid-Atlantic Ridge near the Azores.

To further assess the basement geometry, the inverse method of Werner

deconvolution is improved and used to invert aeromagnetic anomalies for magnetic

basement depths. The improved Werner deconvolution effectively maps out the axial

plateau and valleys in the crustal basement. The results confirm that the basement

topography in the region away from the seismic lines also has an axial plateau within

~60 km of the axis. The basement axial plateau shallows southward, consistent with

increasing influence of the Afar plume.

214

Basement roughness values are computed in profiles both across and parallel

to the axis. The values from axis-crossing seismic data are ~230 m, similar to those

observed at other ultraslow and slow spreading ridges. The roughness values derived

from axis-parallel profiles of the gravity field (200-550 m) are comparable with those

of the Mid-Atlantic Ridge where it has similar along-axis segmentations. Although

these basement roughness values by themselves do not exclude an extended

continental crust interpretation in the central Red Sea, they are supportive of an

oceanic crustal interpretation when considered along with other evidences including

basement axial high, reduced-to-pole magnetic anomalies, and along-axis gradients in

mantle Bouguer anomalies.

Therefore, this study suggests that the entire axial zone in the central Red Sea

is underlain by oceanic crust and the central part of the Red Sea rift is an ultra-slow

spreading ridge influenced by the Afar hotspot, further supporting the idea that the

oceanic spreading inception is synchronous in the central Red Sea along the spreading

ridge.

Bouguer gravity anomalies are found strongly correlated with basement

depths from seismic reflection data within ~60 km of the axis. The apparent density

contrast implied by the correlation (220 to 580 kg m-3) is too small to be caused

primarily by the density contrast between evaporites and mantle across a crust of

uniform thickness and density structure, implying thickened crust and/or low mantle

densities beneath the ridge axis.

Geochemical data (Na8.0) suggest that the crust has normal thickness beneath

the present axis, while the rugged basement topography is consistent with a slow to

ultra-slow spreading ridge with cold, rigid lithosphere and thin crust. To reconcile the

axial high and gravity inversion results, which suggest thickening crust towards the

present day, with these other observations, this study suggests that the crust was

unusually thin earlier in the evolution of the basin and has recently thickened to a

more normal thickness for a slow-spreading ridge.

Finally, a geological map of the central Red Sea was produced based on the

studies in previous chapters. It shows that the transitions between oceanic and

215

extended continental crust tend to be symmetrical about the axis, suggesting that

seafloor spreading has been symmetric. The extended continental crust on either side

of the oceanic crust also appears to be symmetric implying that continental extension

was symmetric here too.

8.2 Future work suggestions

8.2.1 Future work in the Red Sea

To further study the transition from continental rifting to seafloor spreading

along the Red Sea rift, four suggestions are proposed:

(1) Basement geometry in the southern Red Sea

This study has suggested that the central Red Sea rift is an ultra-slow spreading

ridge influenced by the Afar hotspot, since there is an axial crustal high found along

the ridge. To some extent, this interpretation could be assessed by checking whether

there is a similar axial crustal high in the southern Red Sea or not, because the

southern Red Sea rift is an ultra-slow spreading ridge (e.g., Cochran, 1983; Girdler and

Styles, 1974; Phillips, 1970; Vine, 1966). Moreover, the comparison of axial reliefs in

the central and southern Red Sea could be used to assess indirectly effects of

differences in enhanced mantle melting volumes in these two regions.

Therefore, the basement geometry (corrected for isostatic loading) in the

southern Red Sea is suggested to be evaluated using seismic reflection, seismic

refraction, gravity, and magnetic data.

(2) Magnetic source distributions and susceptibilities in the southern and

northern Red Sea

Werner deconvolution method could be used to find the source of magnetic

anomalies in the southern Red Sea (where we are certain that seafloor spreading has

begun) and in the northern Red Sea, and compare the source depths, susceptibilities,

and numbers along the entire Red Sea. The trends with latitude in these parameters

could help to improve the understanding of the transition. Besides, there is a

possibility that these magnetic source properties instead reveal the influence of the

216

sedimentary magnetism, i.e., from magnetic minerals carried into the Red Sea by wind

at present and rivers during wetter periods of the past.

(3) Numerical modelling of structural development in the central Red Sea

To clarify the 28-0 Ma structural development of southeastern Red Sea margin,

Dwivedi and Hayashi (2009) carried out a numerical modelling using a 2D elastic finite

element (FE) package (Hayashi, 2008) in which the geological structure was

asymmetrical as suggested by Voggenreiter et al. (1988). This numerical modelling

was performed along a seismic refraction profile (Mooney et al., 1985), which

provided some constraints of crustal geometry. This seismic section is close to the

Farasan Islands. We suggest that similar numerical modelling could be performed

along the seismic refraction profile SO53-PIII of Egloff et al. (1991) in the central Red

Sea.

(4) Analogue modelling of the Red Sea rift

Analogue experiments carried out by Molnar et al. (2017) were used to

investigate how propagating rifts interact with preexisting structures during the

transition from continental rifting to seafloor spreading. However, these analogue

experiments only simulated the structural evolution around the Danakil microplate in

the southern Red Sea. We suggest that similar analogue experiments could be

performed along the whole Red Sea rift to help to assess the structural evolution in its

geological past.

8.2.2 Apply Werner deconvolution in other areas

This study has shown the feasibility and applicability of Werner deconvolution

of magnetic data in the central Red Sea. The Werner method could be useful in studies

of oceanic crust in other areas that lack of other information from seismic or other

methods. For example, lineaments in the central Pacific gravity field were originally

thought to originate from rolls in Earth’s mantle (Haxby and Weissel, 1986). The more

recent maps of the gravity field show them to be more extensive but with orientations

that may suggest other origins (Gans et al., 2003; Mitchell and Davies, 2018). The

Werner method could provide an alternative measure of basement elevations to help

217

to address the gravity lineament origins. In the Gulf of Mexico, the locations of ocean-

continent boundaries have not been precisely determined and the kinematics of

Jurassic opening still remain poorly known, because seismic imaging of underlying

crust structure is complicated by the thick evaporite and sediment layers, the

combination of which exceeds 12 km in thickness (Galloway et al., 2000; Ibrahim et al.,

1981; Nguyen and Mann, 2016). In this area, if not too badly affected by sedimentary

magnetic sources, the Werner method could be used to derive the basement

topography, which might be useful for addressing those unsolved questions.

The Woodlark Basin is another young ocean basin where the transition from

continental extension to seafloor spreading can be observed (Martinez et al., 1999;

Weissel et al., 1982). The structure of the crust and upper mantle in the Woodlark

Basin is important for understanding the processes of this transition (Ferris et al.,

2006). The correlation between the Bouguer anomalies and the basement depths

could be used to assess the variations in crustal thickness or density or in mantle

density in this region.

8.3 References





69.

Dwivedi, S. K., and D. Hayashi, 2009, Numerical modeling of the development of

southeastern Red Sea continental margin: Earthquake Science, v. 22, p. 239-

249.




v. 198, p. 329-353.

218

Ferris, A., G. A. Abers, B. Zelt, B. Taylor, and S. Roecker, 2006, Crustal structure across

the transition from rifting to spreading: the Woodlark rift system of Papua New

Guinea: Geophysical Journal International, v. 166, p. 622-634.

Galloway, W. E., P. E. Ganey-Curry, X. Li, and R. T. Buffler, 2000, Cenozoic depositional

history of the Gulf of Mexico basin: AAPG bulletin, v. 84, p. 1743-1774.

Gans, K. D., D. S. Wilson, and K. C. Macdonald, 2003, Pacific Plate gravity lineaments:

Diffuse extension or thermal contraction?: Geochemistry, Geophysics,

Geosystems, v. 4, doi:10.1029/2002GC000465, 9.


7-11.

Haxby, W. F., and J. K. Weissel, 1986, Evidence for small‐scale mantle convection

from Seasat altimeter data: Journal of Geophysical Research: Solid Earth, v. 91,

p. 3507-3520.

Hayashi, D., 2008, Theoretical basis of FE simulation software package: Bull. Fac. Sci.

Univ. Ryukyus, v. 85, p. 81-95.

Ibrahim, A., J. Carye, G. Latham, and R. Buffler, 1981, Crustal structure in Gulf of

Mexico from OBS refraction and multichannel reflection data: AAPG Bulletin,

v. 65, p. 1207-1229.


spreading in the Woodlark Basin: Indications of rift ‐ induced secondary


12909-12926.

Mitchell, N. C., and H. Davies, 2018, Equatorial Pacific gravity lineaments:

interpretations with basement topography along seismic reflection lines:

Marine Geophysical Research, p. 1-15.

Molnar, N., A. Cruden, and P. Betts, 2017, Interactions between propagating rotational

rifts and linear rheological heterogeneities: Insights from three‐dimensional

laboratory experiments: Tectonics, v. 36, p. 420-443.




219

Nguyen, L. C., and P. Mann, 2016, Gravity and magnetic constraints on the Jurassic

opening of the oceanic Gulf of Mexico and the location and tectonic history of

the Western Main transform fault along the eastern continental margin of

Mexico: Interpretation, v. 4, p. SC23-SC33.



Sciences, v. 267, p. 205-217.


Jahrbuch, v. 13, p. 131-153.


1415.

Voggenreiter, W., H. Hötzl, and A. Jado, 1988, Red Sea related history of extension and

magmatism in the Jizan area (Southwest Saudi Arabia): indication for simple-

shear during early Red Sea rifting: Geologische Rundschau, v. 77, p. 257-274.




220

Appendices

221

Appendices

Appendix 1

Depths derived from the seismic reflection profiles of Izzeldin (1982, 1987) and

Werner deconvolution of marine magnetic data

222

223

224

225

226

Figure A1.1 Depths derived from the seismic reflection profiles of Izzeldin (1982, 1987) and

Werner deconvolution of shipboard marine magnetic data. Line numbers are shown in the

lower right corner of each panel. Dark green, cyan, and red lines are the depths of the seabed,

the S-reflection at the top of the Miocene evaporites, and the basement, respectively, derived

from the seismic reflection data. Grey circles are Werner source depth solutions, with circle

size proportional to 𝑙𝑜𝑔2(𝜒𝑚 + 2). (For more details, please refer to Chapter 4 (Figure 4.4).)

227

Appendix 2

Matlab codes for 2D Werner deconvolution and Marquardt’s (1963) inverse

modelling method

(1) Codes for 2D Werner deconvolution

%%%%%%% clear; clc; close all; pathname='D:\Magneticdepth\forthesis\'; filename='78005111cut377_524distance9.txt'; fpath1=[pathname filename]; fid1=fopen(fpath1,'r'); data=fscanf(fid1,'%f %f %f %f',[4 inf]); %% Load data%% data=data'; fclose('all'); [mm_dat,nn_dat]=size(data); Lo1=data(:,1); %% Load longitude%% La1=data(:,2); %% Load latitude%% x1=data(:,3); %% Load distance%% y1=data(:,4); %% Load magnetic anomalies%% inter=1; inx=0:inter:max(x1); iny=interp1(x1,y1,inx,'PCHIP'); %%Interpolation, 'inx' has a resolution of 1 m%% inx=inx'; iny=iny'; n_iny=length(iny); result=zeros(n_iny,2); %% Array 'result' is used to save the distance and depth of the magnetic source%% result2=zeros(n_iny,8); %% Array 'result2' is used to save distance, a0, a1, a2, a3, a4, b0, and b1 (Chapter3; Equation 3.4)%% m_result=0; m_result2=0;

228

delt_x=100; %% Sample spacing is 100 m%% while delt_x<(n_iny/6) %% Sample spacing should be shorter than 1/6 length of the survey line%% disp(delt_x); start_i=3*delt_x+1; %% Starting point for the calculation %% for i_window=start_i:1000:n_iny-start_i %% Calculation window%% syms a0_w a1_w a2_w a3_w a4_w b0_w b1_w f_w=sym('A', [7 1]); TMAG=zeros(7,1); %% Magnetic anomalies at 7 sampling points%% TMAG(1)=iny(i_window-3*delt_x); TMAG(2)=iny(i_window-2*delt_x); TMAG(3)=iny(i_window-delt_x); TMAG(4)=iny(i_window); TMAG(5)=iny(i_window+delt_x); TMAG(6)=iny(i_window+2*delt_x); TMAG(7)=iny(i_window+3*delt_x); for i_werner=1:7 %% Conduct simultaneous equations (there are 7 equations) %% n_werner=i_werner-4; f_w(i_werner)= a0_w+a1_w*n_werner+a2_w*(n_werner)^2+... a3_w*(n_werner)^3+a4_w*(n_werner)^4+b0_w*TMAG(i_werner)... +b1_w*n_werner*TMAG(i_werner)-(n_werner)^2*TMAG(i_werner); end A_m=zeros(7,7); %% Initial coefficients (left side) %% b_m=zeros(7,1); %% Initial coefficients (right side) %% a_m=zeros(7,1); %% Initial solution for Marquardt's method%% for p_A=1:7 %% Get initial coefficients (both left and right sides) %% nn_A=p_A-4; A_m(p_A,1)=1; A_m(p_A,2)=nn_A; A_m(p_A,3)=nn_A^2; A_m(p_A,4)=nn_A^3; A_m(p_A,5)=nn_A^4; A_m(p_A,6)=TMAG(p_A); A_m(p_A,7)=nn_A*TMAG(p_A); b_m(p_A)=(nn_A^2)*TMAG(p_A); end

229

a_m=A_m\b_m; %% Compute initial solution for Marquardt's method%% j_illcodition=rcond(A_m); f_illcodition=j_illcodition<1e-12; fnan=any(isnan(a_m)); finf=any(isinf(a_m)); if fnan||finf||f_illcodition %% If the initial solution is derived from ill-conditioned equations or has some extraneous roots, use pseudoinverse to compute the initial solution again%% a_m=pinv(A_m)*b_m; x0_w=a_m'; else x0_w=a_m'; end [r,mf]=Marquardt(f_w,x0_w,0.4,1.2,1.5,[a0_w;a1_w;a2_w;a3_w;a4_w;b0_w;b1_w]); %% Solve the simultaneous equations using Marquardt's method%% solution_w=double(r); %% Solution of the simultaneous equations%% sa0_w=solution_w(1)*(delt_x^2); %% a0 %% sa1_w=solution_w(2)*delt_x; %% a1 %% sa2_w=solution_w(3); %% a2 %% sa3_w=solution_w(4)/delt_x; %% a3 %% sa4_w=solution_w(5)/(delt_x^2); %% a4 %% sb0_w=solution_w(6)*delt_x^2; %% b0 %% sb1_w=solution_w(7)*delt_x; %% b1 %% accurecy_w=double(mf); sx0=0.5*sb1_w; %% Distance of the magnetic source%% sD=sqrt(-sb0_w-(sx0^2)); %% Depth of the magnetic source%% if sD>300 && sD<16000 %% Depth between 300 m and 16000 m%% m_result=m_result+1; result(m_result,1)=inx(i_window)+sx0; %%% Save distance%% result(m_result,2)=sD; %%% Save depth%% m_result2=m_result2+1; result2(m_result2,1)=inx(i_window)+sx0; %%% Save distance%% result2(m_result2,2)=sa0_w; %% Save a0%%

230

result2(m_result2,3)=sa1_w; %% Save a1%% result2(m_result2,4)=sa2_w; %% Save a2%% result2(m_result2,5)=sa3_w; %% Save a3%% result2(m_result2,6)=sa4_w; %% Save a4%% result2(m_result2,7)=sb0_w; %% Save b0%% result2(m_result2,8)=sb1_w; %% Save b1%% end end delt_x=delt_x+500; end result=result(1:m_result,:); result2=result2(1:m_result2,:); M_print=result'; fidout1=fopen([fpath1(1:length(fpath1)-4),'_Distance_sourcedepth.dat'],'wt'); fprintf(fidout1, '%12.6f %12.6f\n',M_print); %% Write distances and depths of the magnetic sources to text file%% fclose('all'); M_print2=result2'; fidout2=fopen([fpath1(1:length(fpath1)-4),'_aANDb.dat'],'wt'); fprintf(fidout2, '%12.12f %12.12f %12.12f %12.12f %12.12f %12.12f %12.12f %12.12f\n',M_print2); %% Write distances, a0, a1, a2, a3, a4, b0, and b1 to text file%% fclose('all'); fpath2=[fpath1(1:length(fpath1)-4) '_Distance_sourcedepth.dat']; fid2=fopen(fpath2,'r'); if fid2>0 %% Confirm the text file is available%%

dat2=fscanf(fid2,'%f %f',[2 inf]); %% Load distances and depths of the magnetic sources%%

dat2=dat2'; fclose('all'); [mm_dat2,nn_dat2]=size(dat2);

data3=zeros(mm_dat2,4); %% Array 'mm_data3' is used to save distance, depth, and susceptibility of magnetic source%%

fpath4=[fpath1(1:length(fpath1)-4) '_aANDb.dat'];

231

fid4=fopen(fpath4,'r'); dat4=fscanf(fid4,'%f %f %f %f %f %f %f %f',[8 inf]); %% Load distance, a0, a1, a2, a3, a4, b0, and b1%%

dat4=dat4'; fclose('all'); mm_data3=0; %% Pointer of 'mm_data3'%% for kk=1:mm_dat2 flag1=dat2(kk,1)>0 && dat2(kk,1)<max(x1); %% Confirm the solutions in text file are available%% flag2=dat2(kk,2)>300 && dat2(kk,2)<16000; %% Depth between 300 m and 16000 m%% if flag1 && flag2 mm_data3=mm_data3+1; data3(mm_data3,1)=dat2(kk,1); data3(mm_data3,2)=dat2(kk,2); sa0_w=dat4(kk,2); %% a0%% sa1_w=dat4(kk,3); %% a1%% sa2_w=dat4(kk,4); %% a2%% sa3_w=dat4(kk,5); %% a3%% sa4_w=dat4(kk,6); %% a4%% sb0_w=dat4(kk,7); %% b0%% sb1_w=dat4(kk,8); %% b1%% sx0=0.5*sb1_w; sD=sqrt(-sb0_w-(sx0^2)); sC2=sa4_w; %% Compute C2 (Chapter3; Equation 3.4)%% sC1=sa3_w+2*sx0*sC2; %% Compute C1 (Chapter3; Equation 3.4)%% sC0=sa2_w+2*sC1*sx0-sC2*(sx0^2)-sC2*(sD^2); %% Compute C0 (Chapter3; Equation 3.4)%% sA=sa1_w+2*sC0*sx0-sC1*(sx0^2)-sC1*(sD^2); %% Compute A (Chapter3; Equation 3.4)%% sB=(1/sD)*(sa0_w+sA*sx0-sC0*(sD^2)-sC0*(sx0^2)); %% Compute B (Chapter3; Equation 3.4)%% sI=30; %% Magnetic inclination%% salp=45; %% Strike of the tabular body%% sVx=cosd(sI)*sind(salp);

232

sVz=sind(sI); b_width=100; %% Half of the dike thickness is assumed as 100 m%% sJx=((-1)*(sB*sVx+sA*sVz))/(2*b_width*((sVx)^2+(sVz)^2)); %% Compute magnetization in x direction%% sJz=(-sA*sVx+sB*sVz)/(2*b_width*((sVx)^2+(sVz)^2)); %% Compute magnetization in z direction%% H_earth=40000; %% Earth main field is assumed as 40000 nT%% M_total=sqrt((sJx)^2+(sJz)^2); sms=M_total/(H_earth); %% Susceptibility %% data3(mm_data3,3)=sms; data3(mm_data3,4)=M_total; end end data3=data3(1:mm_data3,:); Ds2=data3(:,1); depth=data3(:,2); Msms=data3(:,3); Mtotal=data3(:,4); M_print=[Ds2,depth,Msms]'; fidout1=fopen([fpath1(1:length(fpath1)-4),'_WernerSolutions.dat'],'wt');

fprintf(fidout1, '%12.6f %12.6f %12.6f\n',M_print); %% Write distance, depth, and susceptibility of magnetic source to text file%%

fclose('all'); figure subplot(2,1,1); %% Plot out magnetic anomalies%% plot(x1./1000,y1); xlim([min(x1)./1000 max(x1)./1000]); xlabel('Distance (km)'); ylabel('Magnetic anomaly (nT)'); ss=char(filename(1:length(filename)-4)); bb=strrep(ss,'_','\_'); title(bb,'fontsize',12); subplot(2,1,2); %% Plot out Werner Solutions%% plot(Ds2./1000,depth./1000,'* ','Linewidth',3); for j_s=1:length(Ds2)

233

ssm=num2str(Msms(j_s),6); text(Ds2(j_s)./1000,depth(j_s)./1000,ssm(1:4)); end xlim([min(x1)./1000 max(x1)./1000]); set(gca,'YDir','reverse'); xlabel('Distance (km)'); ylabel('Depth (km)'); set(gcf, 'position', [20 50 1500 700]); saveas(gcf,[pathname,'WernerSolutions.emf']); saveas(gcf,[pathname,'WernerSolutions.fig']); close all; end %%%%%%%

(2) Codes for Marquardt’s (1963) inverse modelling method

%%%%%%% function [x,Marf] = Marquardt(f,x0,Pa,u,v,aANDb) %% f: simultaneous equations; x0: initial solution; Pa: coordination factor; u: damping factor; v: amplification factor; aANDb: Independent variables; %% format long; jin=1.0e-8; %% Required smallest value of dx (dx is increment of independent variables) is 1.0e-8%% FF=transpose(f)*f; %% Sum of the squares of f %% m=length(f); x0=transpose(x0); %% Transpose%% n=length(x0); A=jacobian(f,aANDb); %% Compute Jacobian matrix%% sw=1; %% sw is used to record the norm (size) of dx (dx is increment of independent variables) later%% nn=0; %% nn is the number of iterations%% while sw>jin&&nn<20000 nn=nn+1; Fx=zeros(m,1); for i=1:m Fx(i,1)=Funval(f(i),aANDb,x0); %% Values of f at x0 %% end

234

FFx=Funval(FF,aANDb,x0); %% Value of FF at x0 %% Ax=Funval(A,aANDb,x0); %% Value of A at x0 %% gFFx=transpose(Ax)*Fx; %% Gradient of FF %% AW=transpose(Ax)*Ax; while 1 dx=-(AW+u*eye(size(AW)))\gFFx; %% Increment of independent variables%% x1=x0+dx; for i=1:m Fx1(i,1)=Funval(f(i),aANDb,x1); %% Values of f at x1 %% end FFx1=Funval(FF,aANDb,x1); %% Values of FF at x1 %% sw=norm(dx); %% The size of dx %% if sw<=jin %% If sw is small enough (smaller than the required smallest value), break the loop%% break; end if FFx1>=FFx+Pa*transpose(gFFx)*dx %% If FFx1 is not small enough (still >=FFx+Pa*transpose(gFFx)*dx), update u %% u=u*v; continue; %% Increase u and continue testing u in the present loop %% else u=u/v; break; %% Decrease u and go to the outer iteration, which is searching for the final solution %% end

end x0=x1; %% x1 becomes a new initial solution %%

end x=x0; %% The final solution %% Marf=Funval(FF,aANDb,x); %% The minimum value of FF%% format short; %%%%%%%

235

References


Strasbourg.






11, p. 431-441.

Documents

Geophysical study of the crust in the central Red Sea