Late Cenozoic Geology of La Paz, Bolivia, and Its Relation to Landslide Activity

Late Cenozoic geology of La Paz, Bolivia, and its relation to landslide activity

by Nicholas J. Roberts B.Sc. (Physical Geography), Simon Fraser University, 2004 M.Sc. (Earth and Environmental Sciences), University of Waterloo, 2007

Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

in the Department of Earth Sciences Faculty of Science

 Nicholas J. Roberts 2016 SIMON FRASER UNIVERSITY Spring 2016

All rights reserved. However, in accordance with the Copyright Act of Canada, this work may be reproduced, without authorization, under the conditions for Fair Dealing. Therefore, limited reproduction of this work for the purposes of private study, research, education, satire, parody, criticism, review, and news reporting is likely to be in accordance with the law, particularly if cited appropriately. Approval

Name: Nicholas Jason Roberts Degree: Doctor of Philosophy Title: Late Cenozoic geology of La Paz, Bolivia, and its relation to landslide activity Examining Committee: Chair: Doug Stead Professor

John Clague Senior Supervisor Professor

Reginald Hermanns Supervisor Geological Survey of Norway

Bernhard Rabus Supervisor Professor

Brent Ward Internal Examiner Professor

Scott Burns External Examiner Professor Emeritus Department of Geology Portland State University

Date Defended/Approved: April 20, 2016

Abstract

La Paz lies in a deeply incised valley on the Bolivian Altiplano. It has experienced frequent damaging historic landslides and numerous, much larger, prehistoric landslides. I documented the Neogene and Pleistocene lithostratigrahic and magnetostratigraphic framework of La Paz, produced an inventory of recent (1995-2014) landslides, and characterized ongoing (2008-2011) slow ground motion using radar interferometry (InSAR). The upper part of the sediment sequence beneath the Altiplano is glacial in origin and fines distally away from the Cordillera Real. It records at least 15 late Pliocene and Early Pleistocene glaciations, most of which predate the oldest known North American continental glaciation. The plateau surface formed by ca. 1.0 Ma, but most likely before ca. 1.8 Ma. After that the headwaters of the Amazon River extending westward through the Cordillera Real incised the underlying sediments. The poorly lithified fill sequence is exposed in steep slopes, promoting instability. Between 1995 and 2014, La Paz experienced 43 discrete landslides and slow ongoing landslides at 13 additional locations. Landslides were most frequent late in the rainy season and generally happened after particularly wet periods weeks in length, indicating a strong hydro-meteorological control. The margins of several landslides coincide with buried culverted streams, indicating that this engineering practice reduced slope stability. InSAR results show that about one-third of slopes in La Paz are moving at rates up to ~20 cm/a. They also identify previously unknown landslides, detect hectare-scale movements of as little as ~0.5 cm/a, and indicate several distinct failure mechanisms. Many recent landslides correspond with large, creeping paleolandslide deposits, indicating that the reduced residual strength and modern activity of these deposits influences the localization of recent failures. My findings highlight aspects of slope instability in La Paz that can be used to reduce risk. Future failures are most likely to happen in previously displaced fine-grained sediments, particularly the slowly moving paleolandslides south and east of the city centre. Several key areas require detailed ground-based monitoring, particularly during the rainy season when cumulative precipitation thresholds are exceeded. The practice of burying river channels should be re-assessed, and a survey of existing culverted channels conducted.

iii

Keywords: Andean landscape evolution; paleomagnetism; Pliocene tropical glaciation; landslide inventory; urban landslide hazard and risk; Homogeneous Distributed Scatterer Synthetic Aperature interferometry (HDS-InSAR)

Acknowledgements

I thank Dr. John Clague for his mentorship and support throughout my degree. Working with John has greatly advanced my abilities as a scientist and writer. His passion for geoscience and its communication has inspired me.

My supervisory committee provided invaluable expertise and guidance. Dr. Reginald Hermanns was instrumental in initiating this project and establishing my collaborations with Bolivian scientists. He also reminded me to always consider how my research could benefit the residents of La Paz. Dr. Bernhard Rabus’ expertise in synthetic aperture RADAR and the training he provided me was critical to my application of RADAR interferometry. His keen interest in the other aspects of my research also lead to fruitful discussions on landscape processes. In addition, Dr. Stephanie Chang (University of British Columbia) helped me to better understand the interconnections between society and natural hazards, and to consider their ramifications on land-use.

I could not have completed this work without the involvement and local knowledge of my collaborators in La Paz. Mr. Marco-Antonio Guzmán (Universidad Mayor de San Andrés [UMSA]) provided crucial details on recent landslides in La Paz, particularly those predating my first field visit in 2009, and kept me apprised of new landslides as they occurred. Ms. Estela Minaya (Director of Obseravatrio San Calixto [OSC]) was instrumental in my research through discussions on various aspects of La Paz’s geology, provision of logistical support, and engagement with local decision makers. Undergraduate students from the Faculty of Geology at UMSA and the staff at OSC were valuable and eager field assistants during each of my field visits.

Although not part of the initial plan for my thesis, paleomagnetism became a core part of my research because of the generous support of Dr. René Barendregt (University of Lethbridge). In additional to training in paleomagnetic theory and field sampling, René provided me with extensive use of his laboratory equipment over the last five years. Dr. Randy Enkin (Geological Survey of Canada) contributed additional expertise that was crucial to interpreting paleomagnetic results. Ms. Corinne Griffing (SFU) assisted with paleomagnetic field sampling, laboratory processing, and magnetization interpretation.

The support of Macdonald, Dettwiler and Associates Ltd. facilitated the RADAR remote sensing component of my research. Dr. Harold Zwick and, subsequently, Mr. Christian Nadeau, granted me access to processing facilitates and in-house software. Mr. Jayson Eppler, Dr. Jayanti Sharma, Mr. Mike Kubanski and Mr. Parwant Ghuman contributed processing expertise and countless discussions on RADAR theory and applications.

A variety of sources provided funding that enabled this research: NSERC (Discovery Grant to Dr. Clague and Post-graduate Scholarship to me); SFU (Special Graduate Entrance Scholarship, Steel Memorial Graduate Scholarship, Emergency Preparedness Scholarship, and International Research Travel Award); American Society for Photogrammetry and Remote Sensing (Robert Colwell Memorial Fellowship and Ta Liang Memorial Award); Geological Remote Sensing Group, Geological Society of London (Fieldwork Award); and the Mackenzie King Trust (Mackenzie King Open Scholarship).

Finally, I thank my friends and family. My parents, Duncan and Lynda Roberts, provided encouragement and support throughout my long time as a student. Hazel Choi kept me grounded and motivated, particularly during the tough times. Hazel – thank you for all the things you have done, big and small, to help me through this.

Table of Contents

Approval ...... ii Abstract ...... iii Acknowledgements ...... v Table of Contents ...... vii List of Tables ...... xi List of Figures...... xii

Chapter 1. Introduction ...... 1 1.1. Urbanization of the La Paz area ...... 1 1.2. Existing knowledge ...... 7 1.3. Opportunities afforded by modern techniques ...... 9 1.3.1. Paleomagnetism of sediment ...... 9 1.3.2. RADAR interferometry ...... 10 1.4. Research objectives ...... 10 1.4.1. Thesis outline ...... 11

Chapter 2. Late Pliocene and early Pleistocene history of the northeast Altiplano and adjacent Cordillera Real from magnetostratigraphy of the La Paz basin ...... 12 2.1. Introduction ...... 12 2.2. Study area ...... 14 2.2.1. Physiography ...... 14 2.2.2. Stratigraphy ...... 16 Late Neogene sediments ...... 16 Quaternary sediments and their interpretation by previous researchers ...... 19 2.3. Methods ...... 21 2.3.1. Stratigraphy ...... 21 Long-distance correlations and mapping ...... 21 2.3.2. Paleomagnetism ...... 22 Sampling ...... 22 Sample storage and analysis ...... 23 Statistical analysis ...... 24 2.4. Paleomagnetic characteristics of sediments ...... 24 2.4.1. Magnetic susceptibility and magnetization intensity ...... 24 2.4.2. Demagnetization characteristics and magnetic stability ...... 25 2.5. Lithostratigraphy and magnetostratigraphy ...... 30 2.5.1. Patapatani West ...... 34 Stratigraphy ...... 34 Facies interpretation...... 38 2.5.2. Patapatani East ...... 39 Stratigraphy ...... 40 Facies interpretation...... 41 2.5.3. Tangani ...... 42 Stratigraphy ...... 42

vii

Facies interpretation...... 44 2.5.4. Minasa ...... 45 Stratigraphy ...... 45 Facies interpretation...... 47 2.5.5. Purapura...... 47 Stratigraphy ...... 49 Facies interpretation...... 50 2.5.6. Jacha Kkota ...... 50 Stratigraphy ...... 52 Facies interpretation...... 52 2.6. Correlations and chronology ...... 53 2.6.1. Section correlations ...... 53 2.6.2. Composite polarity sequences ...... 56 2.6.3. Correlation with the Geomagnetic Polarity Time Scale ...... 59 Possible interpretations ...... 63 2.6.4. Polarity signatures and ages of previously defined geologic units ...... 63 2.7. Evolution of the Andean landscape...... 65 2.7.1. Lateral facies and landscape variability ...... 66 Lower glacial sequence ...... 66 Gravel sequence ...... 67 Upper glacial sequence and Altiplano gravels ...... 68 2.7.2. Plio-Pleistocene glacial record ...... 69 Chronology ...... 70 Earliest glaciation (g1 to g3) ...... 70 Mid-Piacenzian warm period (g4 to g6) and end of the Pliocene (g7 to g11) ...... 71 Early Pleistocene (g12 to g17) ...... 72 Subsequent Pleistocene glaciations ...... 74 Extent of glaciation ...... 75 Comparison with glaciations in other regions ...... 76 Terrestrial records...... 76 Marine records ...... 78 Drivers of glaciation in the Cordillera Real ...... 79 2.7.3. Tectonic evolution ...... 81 Paleo-elevation ...... 81 Faulting ...... 82 Incision-driven flexural uplift ...... 82 2.7.4. Basin aggradation ...... 83 2.7.5. Breaching of the Cordillera Real and Incision of the Altiplano ...... 84 2.7.6. Slope instability in the upper Río La Paz watershed ...... 87 2.7.7. Paleontological implications ...... 87 Great American biotic interchange ...... 88 2.8. Conclusions ...... 89

Chapter 3. Urban landslides in La Paz, Bolivia, from 1995 to 2014 ...... 92 3.1. Introduction ...... 92 3.2. Setting ...... 94 3.2.1. Physiography ...... 94 3.2.2. Geology ...... 94 3.2.3. Weather systems ...... 98

viii

3.3. Methods ...... 99 3.4. Results ...... 101 3.4.1. Landslide inventory ...... 101 Spatial distribution and geology ...... 101 Discrete events ...... 101 Ongoing slow failures ...... 106 Mechanism and rate...... 107 Discrete events ...... 107 Ongoing slow failures ...... 109 Impacts ...... 110 Discrete events ...... 110 Ongoing slow failures ...... 113 Temporal distribution of discrete events ...... 113 3.4.2. Relation to precipitation ...... 116 3.5. Discussion ...... 119 3.5.1. Landslide types...... 119 3.5.2. Controlling factors ...... 121 Slope properties and processes ...... 121 Precipitation triggers ...... 123 3.5.3. Hazard and risk hot spots and recommendations ...... 125 Landslide-prone areas ...... 125 Secondary hazards ...... 126 Landslide-dammed lake impoundment and outburst floods ...... 126 Debris fluidization ...... 127 Water shortages ...... 127 3.6. Conclusions ...... 128

Chapter 4. Recent slope creep at La Paz, Bolivia, from advanced spaceborne RADAR interferometry ...... 131 4.1. Introduction ...... 131 4.2. Background ...... 134 4.2.1. Application of InSAR in landslide investigation ...... 134 4.2.2. La Paz and surrounding area ...... 137 Physiography and geology relevant to slope stability ...... 137 Mass movements ...... 140 4.2.3. San Antonino case study area ...... 142 4.3. InSAR processing ...... 147 4.3.1. Scene details ...... 147 4.3.2. Processing chain ...... 148 4.4. Results ...... 149 4.4.1. Homogeneous distributed scatterer density and quality ...... 149 4.4.2. Approximate detection thresholds ...... 150 4.4.3. Spatial extent of motion within and south of La Paz ...... 153 4.4.4. Detection of previously unknown landslides ...... 155 4.4.5. San Antonino case study ...... 156 4.5. Discussion ...... 160 4.5.1. Spatial density of HDS ...... 160 4.5.2. Relation of slope deformation to the geology and physiography of La Paz ...... 161 4.5.3. Relation to historic landslides ...... 162

4.5.4. San Antonino case study ...... 163 Spatial patterns of ground motion ...... 163 Temporal patterns of ground motion ...... 165 Similar instability on opposite slopes of the Pampahasi plateau ...... 167 Resettlement after recent landslides ...... 168 4.5.5. Discrimination of kinematics and failure mode ...... 168 4.5.6. Relationships between ancient and historic instability ...... 171 4.6. Conclusions ...... 173

Chapter 5. General conclusions ...... 176 5.1. Recommendations for reducing landslide risk in La Paz ...... 178 5.2. Future research ...... 180 5.2.1. Characterization of the Neogene and Pleistocene sediment sequence ...... 180 5.2.2. Landslide investigations ...... 181 Appendix A Paleomagnetic directions by unit and section ...... 183 Appendix B Clast fabrics ...... 194 Appendix C Tuff comparisons ...... 195 Magnetic intensity ...... 195 Radiometric ages ...... 196 Magnetic remanence ...... 197 Appendix D Correlations of paleosols using paleomagnetic directions ...... 199 Correlation D ...... 200 Correlation E ...... 200 Correlations G and H ...... 200 Correlation K ...... 201 Correlation N ...... 201 Appendix E Polarities and ages of previously defined geologic units ...... 202 Patapatani Drift ...... 202 Chijini Tuff ...... 202 Calvario Drift ...... 203 Purapurani Gravel ...... 203 Kaluyo Gravel ...... 204 Sorata Drift ...... 204 Altiplano surface gravel ...... 205 La Paz Formation ...... 205 Appendix F Chuquiaguillo graben ...... 206 Appendix G Fill-sequence aggradation and incision rates ...... 208 Aggradation rates ...... 208 Incision rates ...... 208 Appendix H InSAR processing methodology ...... 212 Pre-processing ...... 212 Differential InSAR ...... 212 Long-range atmospheric phase screen generation ...... 213 Homogeneous Distributed Scatterer InSAR ...... 213

References ...... 219

List of Tables

Table 2.1. Stratigraphic section details...... 22

Table 2.2. Remanence directions by section and polarity...... 29

Table 2.3. Radiometric ages of tuff exposures in the upper Río La Paz valley system...... 54

Table 2.4. Polarities and consequent age approximations of geologic units of the Altiplano fill sequence...... 64

Table 2.5. Calculated fill-sequence aggradation and incision rates from sections at La Paz...... 84

Table 2.6. Calculated aggradation rates of other late Cenozoic continental fill sequences in the Central Andes...... 86

Table 3.1. Discrete landslide events in La Paz between 1995 and 2014...... 102

Table 3.2. Ongoing slow landslides in La Paz showing motion between 1995 and 2014...... 104

Table 3.3. Documented losses from discrete landslide events and ongoing failures between 1995 and 2014...... 111

Table 4.1. RADARSAT-2 scene details...... 147

List of Figures

Figure 1.1. Aerial view of the La Paz area looking north...... 2

Figure 1.2. Settlement of the La Paz area over time...... 3

Figure 1.3. Population growth in the La Paz metroplitan area since the mid- sixteenth century...... 4

Figure 2.1. Physiographic setting of the Altiplano...... 15

Figure 2.2. Simplified geology of the La Paz area...... 17

Figure 2.3. Equal-area stereographic projections of paleomagnetic directions...... 26

Figure 2.4. Examples of demagnetization characteristics showing the range of magnetization types...... 27

Figure 2.5. View south down the Río Kaluyo valley showing the general appearance glacial deposits below the Chijini Tuff...... 31

Figure 2.6. Examples of sediments and paleosols at the Patapatani West section...... 32

Figure 2.7. Lithostratigraphy and magnetostratigraphy of the Patapatani West section...... 35

Figure 2.8. Lithostratigraphy and magnetostratigraphy of the Patapatani East section...... 40

Figure 2.9. Lithostratigraphy and magnetostratigraphy of the Tangani section...... 43

Figure 2.10. Lithostratigraphy and magnetostratigraphy of the Minasa section...... 46

Figure 2.11. Lithostratigraphy and magnetostratigraphy of the Purapura section...... 48

Figure 2.12. Lithostratigraphy and magnetostratigraphy of the Jacha Kkota section...... 51

Figure 2.13. Comparison of radiometric ages and directional means of tuff exposures of La Paz and Achocalla basins...... 55

Figure 2.14. Magnetostratigraphic correlations of sections through the Altiplano fill sequence...... 57

xii

Figure 2.15. Comparisons of mean remanence directions of paleosols from select stratigraphic position at separate sections...... 58

Figure 2.16. Correlation of the measured polarity sequence with the geomagnetic polarity time scale and global glacial-interglacial record...... 61

Figure 3.1. Overview of the February 2011 Pampahasi landslide...... 93

Figure 3.2. Locations of discrete failure events and ongoing slow failures in La Paz between 1995 and 2014 and their spatial relationship with past large landslides...... 96

Figure 3.3. Buried culverts along the lateral margins of the 2009 Retamani landslide...... 98

Figure 3.4. Examples of landslide events in La Paz between 1995 and 2014 illustrating the range of mechanisms...... 108

Figure 3.5. Time series of landslide events in La Paz from 1995 to 2014...... 114

Figure 3.6. Distribution of landslide events and average precipitation by month...... 115

Figure 3.7. Comparison of size and initial failure mechanism of landslides and preceding cumulative 14-day precipitation...... 118

Figure 4.1. Setting of La Paz and the surrounding area showing the extent of urban development and the distribution of landslides...... 139

Figure 4.2. Evidence of recent instability in La Paz...... 141

Figure 4.3. Overview of the case-study area, southern San Antonio district...... 144

Figure 4.4. Features of prehistoric and recent instability in the western San Antonio district...... 145

Figure 4.5. Typical density and spacing of HDS-InSAR results exemplified by a part of the densely urbanized Santa Barbara area in Distrito Centro...... 150

Figure 4.6. InSAR-measured line-of-sight ground motion in La Paz and the surrounding area...... 151

Figure 4.7. InSAR-measured line-of-sight ground motion in San Antonio study area...... 157

Figure 4.8. Comparison of displacement time-series from the San Antonio district on slopes a) west and b) east of the Pampahasi plateau the western slopes with c) precipitation records...... 159

xiii

Figure 4.9. InSAR-measured line-of-sight ground motion of in the eastern half of the San Antonio study area a) before and b) after the 2011 Pampahasi landslide...... 160

Figure 4.11. Examples of landslide failure mechanisms inferred from HDS- InSAR results...... 169

xiv

Chapter 1. Introduction

Late Cenozoic landscape evolution at La Paz, Bolivia, has produced a striking terrain comprising steep valleys incised into the east margin of the Altiplano plateau and high peaks of the Cordillera Real (Fig. 1.1). Deep incision by tributaries of the upper Río La Paz has exposed up to 1 km of the Neogene and Pleistocene sedimentary fill of the La Paz and Achocalla basins. Laterally and vertically extensive exposures along valley slopes provide perhaps the best exposures of the sediments underlying the Altiplano in Bolivia. The steepness of many of these slopes and the weak geologic materials in which they are developed are responsible for frequent and diverse mass movements that, owing to the large population in the La Paz valley, result in extremely high landslide risk. Geologic investigation in and around La Paz affords three important opportunities: 1) to decipher the history of late Pliocene and Pleistocene glaciation in the equatorial latitudes of South America; 2) to investigate the causes and dynamics of landslides at La Paz; and 3) to reconstruct the evolution of the landscape of the eastern Altiplano and Cordillera Real.

1.1. Urbanization of the La Paz area

The history of development of the metropolitan area centred at La Paz (Figs. 1.2 and 1.3) – which now contains nearly 20% of Bolivia’s population – provides context for understanding the relation between the city’s residents and infrastructure and the landslides that threaten them. Although the parts of this thesis dealing with landslides (Chapters 3, 4 and 5) focus on the physical processes and controls, human elements are crucial to understanding the pressures driving development of particularly landslide- prone areas in and around the city. Future work, including the vulnerability of the

population to landslides, will require a detailed consideration of La Paz’s historical, social, economic and political landscape. The brief history of La Paz provided here is limited to its growth, with only brief consideration of social and economic vulnerably and the inequities among residents. O’Hare and Rivas (2005) and Nathan (2008) provide more detailed summaries of social vulnerably and inequity in La Paz.

Figure 1.1. Aerial view of the La Paz area looking north. The steep area in the foreground is the northernmost Achocalla basin, containing satellite communities of La Paz. The city of La Paz occupies the steep valley in the right middleground. The flat extensive surface is the Altiplano plateau on which the city of El Alto is located. Peaks in the backgroud are part of the Cordillera Real northwest of La Paz (photo Corinne Griffing, 2011; used with permission).

The valley system in which La Paz is located has a long history of settlement. Rock carvings, burial towers, and agricultural terraces in the Achocalla basin span ca. 3000 years from the Formative period to the Inca period (Stercker, 2011), that is from approximately 1500 BC until the influx of Europeans in 1535 (Silverman and Isbell, 2008). Pre-Hispanic sites throughout the La Paz basin – including Miraflores, Llojeta,

Figure 1.2. Settlement of the La Paz area over time. Areas undergoing the greatest development in each period are labelled. a) Sixteenth century settlements, both before and after the arrival of European migrants in 1535, were concentrated along Río Choqueyapu (after Saignes, 1985). b) By 1955 urbanization had spread across flat valley bottoms at Miraflores and Obrajes, and up the western slopes of the Río Choqueyapuy valley (1955 air photo from Instituto Geográfico Militar). Settlement had also begun in the northern part of the then satellite city of El Alto. c) By 2016 urban development has nearly filled the upper Río La Paz valley system and spread over 10 km west across the Altiplano(March 2016 satellite image from GoogleEarth™). Most new settlement in recent decades has been in areas of known slope instability including Villa Armonía, Villa San Antonio, Llojeta, Pampahasi and Cotahuma (details in Chapters 3 and 4). See Fig. 1.3 for population-growth context.

Figure 1.3. Population growth in the La Paz metroplitan area since the mid- sixteenth century. Population estimates until 2001 are from Leonard (1948), Prada (2000), Arbona and Kohl (2004) and references therein. Estimates for 2012 are from Bolivia’s eleventh census (Instituto Nacional de Estadística, 2012).Population growth after 2012 is approximate and based on extrapolations of growth trend between 1985 and 2012.

and Pampahasi – also date to as early as the Formative period (Aranda, 2008; Lémuz and Aranda, 2008). According to Crespo (1906, p. 7) one of the longest established Aymara groups inhabited the upper Río La Paz valley under the name Chuquiabo, from which the modern Aymara name for La Paz – Chuquiago Marka – is derived. As part of their conquest of the Andes, the Inca subjugated the Aymara population and established the town of Chuquiapu (or Chuqueyapu) along the same stretch of Río Choqueyapu where the Aymara lived near the end of the twelfth century (Crespo, 1906, p. 13) (Fig. 1.2a). These pre-Hispanic settlements likely favoured the valley floor because of its mild

climate relative to the higher exposed slopes of the Cordillera Real and Altiplano plateau. Both the Aymara and Inca populations, however, were also attracted to the local gold deposits within the valley (Berthelet, 1986).

The first European residents arrived by 1535 (Crespo, 1906), after which more Spaniards occasionally visited and settled in the valley (Leonard, 1948). The Spanish settlement – Nuestra Señora de La Paz – was founded in 1548 along Río Choqueyapu (Fig. 1.2a) as a stopping point and customs centre between mines at Potosí and La Plata to the south and the commercial centres of Lima and Arequipa to the north (Crespo, 1906; Santa Cruz, 1942). The specific site was chosen due to its pleasant weather (Klein, 1992) and, likely, its placer gold deposits (Loza, 1949). The plan for the city centre was modelled on Spanish cities, but residences outside of the core were haphazardly constructed following topography (Loza, 1949). By 1560 the diverse indigenous population had grown to ~450, and in 1573 centralized in a second settlement – San Pedro y Santiago de Chuquiabo – across the river from La Paz (Saignes, 1985) (Fig. 1.2a). The European population was a minority group, comprising only 200 Spanish immigrants in 1586 (Crespo, 1906). By this time, the La Paz basin had suffered its first historically recorded landslide, which in April 1582 killed nearly all of the ~2000 residents of the indigenous village at Hanco Hanco (variously referred to as Ancoanco, Ancu-Ancu, or Hanko Hanko) (Santa Cruz, 1941; Sanjinés, 1948) a short distance south of the city centre at the approximate location of present-day Lllojeta.

La Paz’s importance grew as the focus of mining switched from gold to the famously rich silver mine of Cerro Rico mine at Potosí, and as export through Lima expanded. The population of the La Paz area increased slowly but steadily during the seventeenth, eighteenth, and nineteenth centuries (Fig. 1.3), particularly following Bolivia’s independence from Spain in 1825 (Leonard,1948), which allowed local investment of wealth that would previously have gone to the Spanish crown (Mesa et al., 2012). At some point, the community of San Pedro y Santiago de Chuquiabo was absorbed into the city of La Paz. The population surpassed 30,000 in 1831, and by 1845 had reached ~42,000. Most growth was accommodated by increased densification of already developed areas, with little change to the city limits; by the end of the nineteenth century La Paz covered only 250 ha, similar to its size shortly after founding in 1548

(Leonard, 1948), despite its population nearing 60,000 (Leonard, 1948; Prada, 2000). When Sir Martin Conway visited La Paz in 1899, the 2-km stretch between the city and the neighbourhood of Obrajes was agricultural land, and Obrajes itself comprised the villas of La Paz’s regular residents (Conway, 1901, p. 92).

The physiography of the upper Río La Paz valley system prevented expansion of La Paz in the way typical of Spanish-planned cities elsewhere in Latin America. Racial classification established in 1902 (Prada, 2000) increased segregation within the basin, with indigenous and mixed-race populations relegated to slopes surrounding the city (Arbona and Kohl, 2004). Growth accelerated during the early twentieth century, and by the census in 1942 its population had increased by more than fivefold to ~301,000, occupying former agricultural land as far south as Obrajes and beginning to creep up the slopes of the Río Choqueyapu valley (Leonard, 1948). La Paz’s affluent population expanded down the Río La Paz valley to more spacious land at favourable elevations and with a better climate, which due to the expansion of public transport and the wider availability of private vehicles was now within commuting distance of the city centre (Leonard, 1948). In contrast, marginalized groups expanded up the steep slopes of the basin where land was cheaper and city taxation did not apply (Leonard, 1948) (Fig. 1.2b). By 1950, urban expansion spilled onto the Altiplano directly west of La Paz, initiating the satellite city of El Alto (Arbona and Kohl, 2004) (Fig. 1.2b), with early growth fueled by the arrival of workers recently freed from haciendas (Arbona and Kohl, 2004). The population increase in both El Alto and La Paz has since been dominated by migration from rural communities to large cities. Accelerated growth of El Alto in the 1980s was driven by farm failures due to El Niño-related droughts (Arbona and Kohl, 2004) and state-run mine closures driven by political changes (McFarren, 1992). In 1988 El Alto became an autonomous municipality, but many of its residents continue their daily commutes into the valley below for employment and commerce.

By the start of the twenty-first century, the combined population of La Paz and El Alto had reached nearly 1.4 million (Arbona and Kohl, 2004). In 2012, the date of the most recent census, the population of El Alto (842,378) exceeded that of La Paz (757,184) (Instituto Nacional de Estadística, 2012). Extreme physiographic barriers are slowing expansion in the upper Río La Paz valley, which is near capacity and long ago

surpassed its sustainable capacity. Replacement of adobe structures with taller buildings results in localized densification (Arbona and Kohl, 2004). However, most growth is by informal development on slopes of marginal stability (Fig. 1.2c) as many residents strive to be closer to opportunities in La Paz (O’Hare and Rivas, 2005). By contrast, rapid growth in El Alto continues westward and southward due a complete lack of physiographic barriers. If growth continues at the same rate as it has over the past two decades, the population of the metropolitan area of La Paz and El Alto will approach 2.5 million by 2050, resulting in further exposure of the population to landslides (Fig. 1.3). Although the majority of this growth can be expected in El Alto, many new residents are likely to commute to La Paz on a daily basis. In addition, the pressure will become greater than ever to develop highly unstable slopes in La Paz that have thus far been avoided.

1.2. Existing knowledge

The deeply incised upper Río La Paz valley system and the Neogene and Pleistocene sedimentary sequence it exposes have drawn the attention of geographers and geologists for over 150 years. The first scientific explorations of the eastern Bolivian Altiplano, including La Paz, were made by d’Orbigny (1842) during his regional survey of South America. Conway (1901) first recognized the gigantic Achocalla earthflow, which underlies the entire Achocalla basin (~ 50 km2), directly south of the city of La Paz. Gregory (1913) provided the most detailed early account of the nature of the sedimentary sequence exposed at La Paz, noting its high degree of lateral variability. Troll and Finsterwalder (1935) described some aspects of the local geology, including extensive glacial deposits exposed along many valley slopes, and produced the first topographic maps and a geologic map of the area1. Several other early workers provided wide-ranging, sometimes fantastical, opinions about the origin and downcutting of the sedimentary sequence, but these have had limited influence on our understanding of La

1 Dobrovolny (1955) notes that the Troll and Finsterwalder’s (1935) geology map was accidently destroyed before publication.

Paz’s geologic history. By 1945, details of the geology and processes of the La Paz area were still very limited.

Most of the current knowledge of the geology and geomorphic activity of La Paz stems from three large research projects in the mid-twentieth century. Each of these projects addressed the aforementioned themes of the formation of the basin fills beneath the Altiplano and their more recent incision and failure. Friedrich Ahlfeld undertook the first national geologic reconnaissance of Bolivia (Ahlfeld, 1946), paying particular attention to La Paz (Ahlfeld 1945a,b). Ernst Dobrovolny of the U.S. Geological Survey performed detailed geologic mapping of the La Paz area during the 1950s to help guide urban zoning (Dobrovolny, 1955). His research included the first characterization of several large prehistoric landslides, including the Achocalla earthflow (Dobrovolny, 1962). His final report (Dobrovolny, 1962) remains the primary source of information used by many geologists and engineers in La Paz. A joint Bolivian-French research group comprising scientists from Bureau de Recherches Géologiques et Minières (BRGM – the French Geologic Survey) and Office de la Recherche Scientifique et Technique d'Outre-Mer (ORSTOM – the French Office of Scientific and Technical Research Overseas) expanded on Dobrovolny’s work, again with a focus of guiding future urban development. The BRGM-ORSTOM project produced a 23-volume series of reports on a variety of aspects of the physical environment of the La Paz area, including stratigraphy (Bles et al., 1977), geomorphology (Malatrait, 1977; Vargas, 1977), and geotechnique (Anzoleaga et al., 1977). In subsequent decades, project members published papers on the geochronology of the sedimentary sequence (Lavenu et al., 1989; Thouveny and Servant, 1989) and, to a lesser extent, on mass movements of the upper Río La Paz watershed.

Other notable contributions include radiometric dating campaigns in the Central Andes (Evernden et al., 1977; Marshall et al., 1992) and Clapperton’s (1979) documentation of a Pliocene glacier that reached La Paz.

1.3. Opportunities afforded by modern techniques

Scientific developments in the decades since the last major geologic investigations in the La Paz area have provided new opportunities to advance understanding of its geology and geohazards. My PhD research employs two techniques from separate scientific fields to provide new insights into landscape evolution at La Paz on timescales ranging from months to millions of years.

1.3.1. Paleomagnetism of sediment

By the time of the first detailed geologic investigations in La Paz (Ahlfeld, 1945a,b, 1946; Dobrovolny, 1955, 1962), a chronology of reversals of Earth’s magnetic field was emerging with potential application as a geochronologic tool (Opdyke and Channell, 1996). However, block sampling procedures, the practice of blanket magnetic cleaning, and the low sensitivity of early magnetometers at that time often made accurate magnetic measurement of sediments a challenge (Tarling, 1983; Opdyke and Channell, 1996). Understanding of Geomagnetic field behaviour, and sampling and measuring techniques had improved greatly by the time of the BRGM-ORSTOM study, but paleomagnetic characterization of poorly lithified rocks and non-lithified sediments in terrestrial sections, such as those at La Paz, required tedious collection and transport of oriented sediment blocks (Johnson et al., 1975). Using such an approach, Thouveny and Servant (1989) provided some important constraints on the chronology of the La Paz sedimentary sequence. However, their work was limited to sampling fine-grained beds. The use of small plastic cubes, and later cylinders (in the early 1990’s), for collecting oriented samples of sediments for magnetic characterization, made sampling easier and increased the opportunities for magnetostratigraphic studies of sediment sequences. I used this collection technique to considerably refine the chronostratigraphy of the La Paz sedimentary sequence (Chapter 2). This sampling approach allowed me to determine the paleomagnetic record of thin beds and lenses within the predominantly coarse- grained units (diamict and gravel), a sampling strategy which had not been attempted in the earlier studies.

1.3.2. RADAR interferometry

Synthetic aperture radar interferometry (InSAR) enables the measurement of sub-centimetre ground movements using repeat-pass satellite images (Massonnet and Feigl, 1998). Developed in the late twentieth century, this remote sensing technique is now commonly applied to investigations of slow-moving landslides. Advances in radar technology (Sansosti et al., 2014) and the pioneering of new InSAR techniques (Ferretti et al., 2001, 2011; Berardino et al., 2002; Eppler and Rabus, 2011) in the past 15 years have expanded the capabilities of InSAR in landslide investigations by increasing spatial resolution, reducing or eliminating sources of error, and facilitating the production of displacement time-series. I applied a state-of-the-art InSAR technique to address questions about the extent and character of slow slope movements in La Paz (Chapter 4). This new technology provides insights into slope movements that could not have been adequately investigated at the time of earlier research programs.

1.4. Research objectives

The primary objective of my PhD research has been to better understand late Cenozoic landscape evolution at La Paz, including the formation, incision, and subsequent modification of the Neogene and Pleistocene sedimentary sequence underlying the eastern Altiplano. I focus on: (1) paleoenvironments of the eastern Bolivian Altiplano and adjacent Cordillera Real during the Pliocene and Quaternary, including the unique archive of late Pliocene and early Pleistocene tropical glaciation; and (2) landslide activity of the upper Río La Paz valley system.

Within this overall objective, I have three complementary sub-objectives:

1. to reconstruct paleoenvironments of the eastern Bolivian Altiplano and adjacent Cordillera Real during the Pliocene and Pleistocene, 2. to characterize the occurrence and properties of landslides in La Paz in recent decades, and

3. to characterize the extent and rate of slow creep-like ground motion in La Paz over several years.

1.4.1. Thesis outline

This thesis comprises three body chapters that contribute to an improved understanding of the late Cenozoic history and related geohazards of the La Paz valley. Each chapter considers a progressively shorter time scale and builds on the forgoing chapters to evaluate relations among the geology of the La Paz valley, discrete landslide events, and slow slope deformation.

Chapter 2 provides a geological framework of the La Paz area. I characterize the nature and ages of late Pliocene and early Pleistocene glaciations, when glaciers flowing from the Cordillera Real reached to near La Paz. Methods employed in this study include lithostratigraphic and magnetostratigraphic investigation of sections within the upper Río La Paz valley system along a transect oblique to the Cordillera Real.

Chapter 3 provides an inventory of landslides in the La Paz basin between 1995 and 2014. I describe landslide events and locations of slow ongoing failures during this 21-year period and identify causes, triggers, spatial patterns, and temporal trends.

Chapter 4 identifies and quantifies slow displacements of slopes in and around La Paz using InSAR. Data were acquired roughly monthly between 2008 and 2011. I compare the extent and nature of InSAR-measured slope deformation with the geology of the La Paz area (Chapter 2), the historic landslides (Chapter 3), and much larger ancient landslides underlying most of the basin’s slopes.

Chapter 5 summarizes the contributions of my research and identifies considerations important to mitigating landslide risk in La Paz. It closes with suggestions for future work that might address outstanding questions about the late Cenozoic landscape evolution of the Bolivian Andes and further improve understanding of landslide processes and hazard at La Paz.

Chapter 2. Late Pliocene and early Pleistocene history of the northeast Altiplano and adjacent Cordillera Real from magnetostratigraphy of the La Paz basin

2.1. Introduction

Late Cenozoic continental clastic sequences of the Bolivian Altiplano are important archives of landscape evolution, climate fluctuation, and land-mammal successions of the Central Andes. Locally, they record multiple glaciations (Dobrovolny, 1962; Clapperton, 1993) for latitudes at which relatively little is known about the cryosphere prior to the last glacial cycle (Ehlers and Gibbard, 2007; De Schepper et al., 2014). Deformation of the clastic sequences indicates post-Miocene tectonic dislocations (Lavenu, 1977; Lavenu, et al., 1989, 2000) with the potential to provide insight into plateau formation and tectonic history. The northern Altiplano contains a rich fossil record of Andean mammals (Hoffstetter, 1986; Marshall and Sempere, 1991; Marshall et al., 1992) that reveals the late Tertiary exchange of land mammals between North and South America (Simpson, 1980; Woodburne, 2010; Cione et al., 2015). The aggradation and incision history of the sequence may also help explain the high incidence of slope instability in the vicinity of the city of La Paz.

The upper Río La Paz, which is part of the headwaters of the Amazon River system, has incised 300 to 800 m into Plio-Pleistocene terrestrial sediments of the Bolivian Altiplano (Ahlfeld, 1946; Dobrovolny, 1962). These sediments fill a large paleobasin along the eastern limit of the Altiplano and include the earliest known evidence of tropical glaciation of the Cenozoic Period (Ehlers and Gibbard, 2007; De

Schepper et al., 2014). The sequence was noted by numerous geographic explorers in the mid- and late nineteenth century and early twentieth century (d’Orbigny, 1842; Conway, 1901; Gregory, 1913), but was not described in detail until the geology of the region was mapped by Ahlfeld (1945a, 1945b, 1946), Dobrovolny (1955, 1956, 1962), and scientists of a joint Bolivian-French research team (Anzoleaga et al., 1977; Bles et al., 1977; Malatrait et al., 1977; Servant, 1977; Vargas, 1977; Ballivián et al. 1978; Thouveny and Servant, 1989). Radiometric dating of volcanic units (Clapperton, 1979; Lavenu et al., 1989; Marshall et al., 1992) and magnetostratigraphy of four, largely non- overlapping sections (Thouveny and Servant, 1989) constrain ages of parts of the sedimentary sequence. However, many differences in interpretations have arisen, due largely to limited chronologic control and uncertain correlations between sections, particularly of glacial units. Consequently, the timing and number of glaciations (Dobrovolny, 1962; Servant, 1977; Clapperton, 1979), ages of some faunal assemblages (Hoffstetter, 1986, p. 224), and initiation of incision of the La Paz valley remain uncertain.

In this chapter, I infer the late Pliocene and early Pleistocene history of the eastern Altiplano based on lithostratigraphic and magnetostratigraphic analysis of six sections of the Altiplano fill sequence along an 18-km transect oblique to the trend of the Andes. The sections are located along the western part of the upper Río La Paz drainage system; four of the sections reach the Altiplano surface. Based on paleomagnetic remanence measurements and a volcanic marker bed, I correlate these sections, as well as the magnetostratigraphic sections described by Thouveny and Servant (1989) from now-developed parts of La Paz. The correlations highlight facies changes away from the crest of the Andes and tie the upper part of the fill sequence to the geomagnetic polarity timescale. Based on the sedimentology of the fill sequence, I identify the onset and minimum number of late Cenozoic glaciations in the Central Andes, constrain the uplift history of the Cordillera Real, and determine the age of the plateau surface and thus the lower limiting age of drainage extension into the Altiplano by the headwaters of the Amazon River.

2.2. Study area

2.2.1. Physiography

The Central Andes comprise two parallel, northwest-trending ranges – the Cordillera Occidental on the southwest and Cordillera Oriental on the northeast – separated by the high Altiplano plateau (Fig. 2.1a), which is second only to the Tibetan Plateau in both height and area. The northwest portion of the basin drains into Lake Titicaca, which itself drains via Río Desaguadero to Lago Poopó, an endorheic salt lake on the central Altiplano. To the south, drainage is into Salar de Uyuni, a closed basin on the Altiplano. Large expanses of the northern Altiplano have been little modified by erosion during the late Cenozoic, with late Pliocene volcanic deposits well preserved at the surface (Lavenu et al., 1989; Marshall et al., 1992). In some areas, however, the plateau surface was uplifted during the late Cenozoic, and young folded and faulted sediments were planated, for example in the Corque basin (Roperch et al., 1999). The plateau rises gradually along its eastern margin as sediments thicken toward their sources in the Cordillera Oriental, forming a thick clastic apron.

The Cordillera Real is the highest part of the Cordillera Oriental and extends nearly 200 km in a northwest direction from Cerro Gigante to Ancohuma. A tributary of Río Bení, which is part of the Amazon River watershed, penetrates the Cordillera Real just south of the Illimani massif, and drains ~2500 km2 of the Altiplano (Fig. 2.1a). It is the larger of two streams that drain to the Atlantic and have penetrated the Altiplano; the other, farther to the north, reaches to near the southeast shore of Lake Titicaca. At La Paz, the Río La Paz, the northern tributary of Río Beni, has incised five parallel, northeast-trending valleys as much as 500 m into the Altiplano surface (Fig. 2.1b). The upper parts of the two westernmost of these tributaries (Ríos Choqueyapu and Orkojahuira) are the Río Kalyuo and Chuquiaguillo, respectively (Fig. 2.1b).

Figure 2.1. Physiographic setting of the Altiplano.

a) Extent of the Altiplano Plateau. Extent (b) and profile (c) of constituent parts of the Central Andes: Cordillera Oriental and Cordillera Occidental enclosing the Altiplano Plateau. La Paz, Bolivia, (box ‘d’) is located southeast of Lake Titicaca at the edge of drainage incised into the eastern Altiplano by headwaters of the Amazon basin. The Cordillera Real comprises the high Cordillera Oriental between Cerro Gigante (100 km south east of La Paz) and Ancohuma (80 km northwest of La Paz), and includes the Illimani massif just southeast of La Paz. d) La Paz and surrounding area. Locations of stratigraphic sections represented with black lines: PTW, Patapatani West section; PTE, Patapatani East section; TNG, Tangani section; MIN, Minasa section; PUR, Purapura section; JKK, Jacha Kkota section (includes ‘upper’ part located 3 km southeast of main section); VIS, Viscachani section of Thouveny and Servant (1989). Terrain is from ASTER DEM.

2.2.2. Stratigraphy

During the Neogene, the Bolivian Altiplano consisted of internally drained basins. Consequently, it is underlain by a thick sequence of sediments shed from the elevated cordilleras along its margins (Ahlfeld, 1946; Newell, 1949; Evernden et al., 1977). Where exposed along the northeast margin of the Altiplano (Fig. 2.2), these sediments unconformably overlie tectonized, Silurian and Devonian sedimentary and metasedimentary rocks of the Cordillera Real piedmont (Dobrovolny, 1962). This basement consists predominantly of heavily folded and fractured quartzite and shale (Ahlfeld, 1945a, 1946; Dobrovolny, 1962), which Bles et al. (1977) named the Sica-Sica Formation. The tectonized metasediments are locally overlain by up to 1000 m of less deformed, Cretaceous red conglomerate with lenses of siltstone and sandstone of the Aranjuez Formation (Ahlfeld, 1945a, 1946; Dobrovolny, 1962; Bles et al. (1977). Several large granitic plutons of Oligocene to Early Miocene age intruded the basement rocks and are now exposed along the crest of the Cordillera Real (Evernden et al., 1977), forming its highest peaks.

Late Neogene sediments

Long, continuous exposures of the late Neogene and Pleistocene sedimentary fill are limited to the Río La Paz drainage system along the flank of the Cordillera Real. The generally fine-grained La Paz Formation (Ahlfeld, 1946, Dobrovolny, 1962; Bles et al., 1977) forms the lower half of the sequence (Fig. 2.2a) and comprises weakly consolidated, interlensing beds of silt, sand, and gravel of fluvial and lacustrine origin

Figure 2.2. Simplified geology of the La Paz area.

a) Geologic cross-section extending perpendicular away from the Cordillera Real (after Bles et al., 1977). b) Surficial geology based on mapping by Dobrovolny (1962) and Anzoleaga et al. (1977) and my new observations and interpretations (see text). The geologic units underlying the Altiplano are generalized as five groups: Paleozoic basement rock (metasedimentary units including the Sica Sica and Aranjuez formations); the La Paz Formation (Miocene to Pliocene); the Chijini Tuff, which most previous researchers included in the La Paz Formation (Anzoleaga et al., 1977; Bles et al., 1997; Lavenu et al., 1989; Thouveny and Servant, 1989; Marshall et al., 1992), but Dobrovolny (1962) described as occurring between glacial units near the Cordillera Real; glacial and glaciofluvial units (including the Patapatani drift, Calvario drift, Purapurani gravels, Kaluyo gravels, and Sorata drift); and Altiplano gravels. Younger units nested within the valleys (paleolandslide depositsts and fluvial and glaciofluvial underlying terraces) are from Anzoleaga et al. (1977). Locations of stratigraphic sections represented with black lines: PTW, Patapatani West section; PTE, Patapatani East section; TNG, Tangani section; MIN, Minasa section; PUR, Purapura section; JKK, Jacha Kkota section (includes ‘upper’ part located 3 km southeast of main section); VIS, Viscachani section of Thouveny and Servant (1989).

(Ahlfeld, 1945a,b; Dobrovolny, 1962; Bles et al., 1977). Limited lateral (<1 km) and vertical (<10 m) continuity of these beds (Ahlfeld. 1946; Dobrovolny, 1962; Bles et al., 1977) suggests an alluvial system of variable energy. A ca. 5.5-Ma (Lavenu et al., 1989; Servant et al., 1989), ~2-m-thick, welded dacitic (based on composition in Vatin- Perignon et al., 1996) tuff (Cota Cota tuff) is exposed in the lower part of the La Paz Formation. A younger, rhyolitic (Lavenu et al., 1989; based on composition described in Vatin-Perignon et al., 1996) tuff (Chijini Tuff) is present throughout the La Paz basin and crops out extensively in the city of La Paz (Fig. 2.2b), with measured thicknesses up to 14 m. Dobrovolny (1962) places the top of the La Paz Formation at the base of this second tuff. In contrast, Ahlfeld (1946) and Bles et al. (1977) place the Chijini Tuff below the top of the La Paz Formation.

Limited, heavily fragmented fossils recovered from the La Paz Formation suggest a Pliocene to very early Pleistocene age (Ahlfeld, 1946, Ahlfeld and Branisa, 1960, Dobrovolny, 1962; Hoffstetter, 1986; Mones and Mehl, 1990). Radiometric ages on the Cota Cota tuff (ca. 5.5 Ma: Lavenu et al., 1989; Servant et al., 1989) and the Chijini tuff (ca. 3.2-2.6 Ma: Clapperton, 1979; Lavenu et al., 1989; Marshall et al., 1994) and magnetostratigraphy (Thouveny and Servant, 1989) confirm that the La Paz Formation spans most of the Pliocene and may include the late Miocene and earliest Pleistocene, based on the timescale of Gradstein et al. (2012).

Quaternary sediments and their interpretation by previous researchers

The upper, generally coarse-grained part of the fill sequence consists of thick glacial diamictons interbedded with gravels (Fig. 2.2a). The latter are interpreted in previous studies as either glaciofluvial (Ahlfeld, 1945a [p. 13], 1946; Dobrovolny, 1962 [p. 43 – Kaluyo Gravel or lower Milluni]) or interglacial (Dobrovolny, 1962 [p. 43 – Purapurani Gravel]; Bles et al., 1977 [p. 7]; Thouveny and Servant, 1989 [p. 341]) origin. Dobrovolny (1962) identifies three glacial units in the Altiplano sequence: the Patapatani Drift directly below the Chijini Tuff, and the Calvario and Milluni drifts above it. The latter two units are found throughout most of the valley system and are separated by the Purapurani Gravel, which comprises up to several hundred metres of granitic, well sorted, sub-rounded, pebble-cobble gravel. Dobrovolny (1962) includes fine-grained sediments overlying the Chijini Tuff (Ahlfeld’s [1945a, 1946] upper La Paz Formation) within the Calvario Drift. He argues that these sediments are the fine-grained facies equivalent of the Calvario till nearer the Cordillera. He reports the Patapatani Drift from only a single, 7-m-high exposure on the east slope of Río Kaluyo valley (Dobrovolny, 1962; see Dobrovolny, 1956, p. 63 for site description), and suggests it is of Pleistocene age based on glyptodon remains in the upper La Paz Formation to the south, which he considers to be glaciofluvial deposits associated with the drift (Dobrovolny, 1962, p. 26).

Bolivian and French researchers (Bles et al., 1977; Servant, 1977; Ballivián et al. 1978; Thouveny and Servant, 1989) interpret the sequence differently. They propose an additional tuff within the Purapurani Gravel (the Sopari tuff [Servant, 1977; Thouveny and Servant, 1989] or Purapurani Tuff [Lavenu et al., 1989]) and suggest Dobrovolny (1962) mistook it as the Chijini tuff at his sole Patapatani Drift exposure. In their ‘two-tuff’ scheme, the Patapatani Drift is equivalent to the Calvario Drift, meaning that the first glaciation of the Cordillera Real postdates eruption of the Chijini Tuff. These later workers also divide Dobrovolny’s (1962) Milluni Drift into the Kaluyo and Sorata glaciations separated by a well developed paleosol (Bles et al., 1977; Servant, 1977; Ballivián et al., 1978). Regardless of the stratigraphic interpretation, most previous workers (Dobrovolny, 1962; Bles et al., 1977; Lavenu et al., 1989; Thouveny and Servant, 1989; Marshall et al., 1992) consider glaciation to have begun in the early Pleistocene.

Clapperton (1979), however, suggests that the earliest glaciation is Pliocene in age based on a biotite K-Ar age of 3.27 ± 0.14 Ma for tuff (almost certainly the Chijini tuff) directly overlying 2 m of till in the Río Kaluyo/Choqueyapu valley a short distance downstream of Dobrovolny’s Patapatani type section (Fig. 2.1b). More accurate dating of the Chijini Tuff using potassium feldspar indicates ages of 2.650 ± 0.012 (40Ar/39Ar, Marshall et al., 1992) to 2.8 ± 0.1 Ma (K-Ar, Lavenu et al., 1989). However, neither of these studies nor a contemporary study applying magnetostratigraphy (Thouveny and Servant, 1989) document till below the tuff. To explain the disparate ages, Clapperton (1993) proposes a pre-Chijini Tuff, apparently exposed nowhere else, which he calls the Chacaltaya Tuff.

Late Pleistocene glacial deposits and Holocene colluvium and alluvium are inset into or lie upon the Plio-Pleistocene sequence described above and cover much of the valley floor and valley walls (Ahlfeld, 1954a, Dobrovolny, 1962, Bles et al., 1977). Dobrovolny (1962) mapped end moraines of two glacial events on the valley bottoms north of the city. However, the more southerly of his two end moraines in the Río Choqueyapu valley is the debris of a large slump on the east side of Río Kaluyo (Limanpata landslide; Anzoleaga et al., 1977; Vargas, 1977; cf. Lavenu et al., 1989, p. 44) and possibly a large down-thrust block on the west side of the river (Vargas, 1977). Heim (1951) noted a diamicton of unknown age with striated clasts in a foundation excavation farther down valley (~3500 m asl) in the city centre. Northwest of La Paz on the plateau surface, a series of end moraines clearly visible in aerial photographs and satellite images extend ~20 km southwest across the plateau surface from the crest of the Cordillera Real. One of the moraines, along Río Seco just northwest of Río Kaluyo, yielded beryllium-10 terrestrial cosmogenic nuclide (TCN) exposure ages of 34-22 ka (Smith et al. 2005).

2.3. Methods

2.3.1. Stratigraphy

I described and measured ~ 1100 vertical metres of sediment at six sections (Figs. 2.1b and 2.2b) during five field visits between 2010 and 2014. Field characterization included descriptions of texture, structure, lithology, colour, clast size and shape, sorting, weathering features, and the nature of contacts. I defined unit boundaries by major changes in material properties and by paleosols or erosional contacts indicating major depositional hiatuses. I measured unit thickness and locations of features with a handheld laser rangefinder, and measured stratal thickness and dimensions of coarse clasts with a graduated metric scale. The sections include exposures in steep valley slopes, gulley systems, and road cuts, along a roughly north- south trend, within the Río Choquyapu/Kalyuo and Orkojauira/Chuquiaguillo valleys and the Achocalla basin, directly south of La Paz (Fig. 2.1b).

Long-distance correlations and mapping

I determined the general location and extent of units mapped by previous workers using geologic maps compiled by Dobrovolny (1962), Anzoleaga et al. (1977) and Bles et al. (1977). I mapped section locations using high-resolution (2-m pixel resolution or finer) ortho-rectified satellite imagery available in GoogleEarth, assisted by precise positions measured in the field using a handheld GPS. I determined the approximate (± 10 m) elevations of sections using the terrain layer in GoogleEarth and GPS measurements.

2.3.2. Paleomagnetism

Sampling

I collected 808 samples at 124 levels at the six sections (Table 2.1). Commonly six samples were taken at each site, but as few as three and as many as 16 were collected at some sites. Samples were collected in oriented plastic cylinders (6.58 cm3 volume; standard rock specimen dimensions) largely from well sorted, horizontally bedded zones of silt to fine sand. Prior to sampling, I cleared exposure surfaces to remove colluvial and slopewash deposits.

Table 2.1. Stratigraphic section details.

Section Section height (m) Sample Average sample Samples Remanence calculation sites spacing (m) Total a Sampledb Total Useful Used PCAc GCd

Patapatani West 232 229 47 4.9 328 284 251 251 0 Patapatani East 50 36.5 10 3.7 66 48 41 41 0 Tangani 250 190.5 0 6.4 193 181 160 160 0 Minasa 410 410 14 29.3 84 71 61 61 0 Purapura 250 91 14 6.5 83 70 66 60 6 Jacha Kkota 175 59 9 6.6 54 53 45 41 4

Overall 1367 1016 94 10.8 808 707 624 614 10

a Including covered slopes and undescribed exposures. b Including only described, sampled exposures. c Primary remanence directions determined by Principle Component Analysis (PCA of samples included in overall statistics [Fig. 2.3, Table 2.2] and for group mean directions calculation). d Primary remanence directions determined by intersection of Great Circles (PCA of samples included in overall statistics [Fig. 2.3, Table 2.2]; GC used for group mean directions calculation).

As the only previous paleomagnetic study (Thouveny and Servant, 1989) includes very limited sampling of coarse-grained deposits, I paid special attention to gravel and diamicton units to improve the completeness of the magnetostratigraphic record. I sampled fine-grained sequences at regular intervals, whereas samples from diamicton and gravel sequences were limited to available lenses of silt to sand. I also collected some samples of low-stone content diamicton, avoiding granules and pebbles. Where possible, I collected samples both above and below unit boundaries, particularly where pedogenesis indicated lengthy hiatuses. Samples were spaced, on average,

every 5 m except at the Minasa section, the longest and coarsest of the sections, where the average sample spacing was 30 m (Table 2.1).

Sites that produced indeterminate polarity or incoherent demagnetization characteristics were re-sampled during subsequent field visits, in many cases with improved results. Due to limited exposure, steepness, or very coarse texture, some units could not be sampled. Where possible, I attempted to capture these gaps by sampling likely correlative units at nearby sections.

Sample storage and analysis

I stored samples in magnetic shields in the Paleomagnetism Laboratory at the University of Lethbridge immediately after each field season, following unshielded transport from La Paz. For each sample, I measured bulk magnetic susceptibility (MS) with a Sapphire Instruments SI-2B magnetic susceptibility and anisotropy meter. Magnetic susceptibility provides a reliable measure of bulk magnetite or maghemite content, and provides a proxy of the degree to which a sediment or rock has become magnetized (Butler, 1992; Opdyke and Channell, 1996). I then measured natural remanent magnetization (NRM) of each sample with an AGICO JR-6A spinner magnetometer. I re-measured remanence following stepwise demagnetization in fields up to 200 militesla (mT) in an ASC Scientific D-2000 alternating field demagnetizer.

For each sample group, I demagnetized a pilot specimen with high magnetic susceptibility relative to other samples of the group at closely spaced (2.5-10 mT) demagnetization steps to determine a characteristic magnetization for each sample group and identify the remanence carrier(s). I demagnetized the remaining specimens from each group at 4 to 10 steps (5-30 mT spacing) on the basis of their pilot response in order to cover at least 50% of the natural remanent magnetization (NRM) intensity and until 20% or less of the NRM intensity remained.

Statistical analysis

I determined directions of remanence components by principle component analysis (PCA; Kirschvink, 1980) using AGICO’s Remasoft v. 3.0 (paleomagnetic data analysis software). The component of magnetization that trends to the origin of the orthogonal projection after the highest levels of AF demagnetization, is assumed to represent the primary component of magnetization, and in these semi-arid terrestrial sedimentary environments, is assumed to have been acquired shortly after sediment deposition (Opdyke and Channell, 1996). Using the statistical module in Remasoft v. 3.0, I then determined mean remanence directions of samples by group, stratigraphic unit, and polarity type.

I used the intersection of great circles (McFadden and McElhinny, 1988) to identify the primary component of magnetization for a small number of samples (Table 2.1) for which final cleaning fields did not produce a clear end point. In some instances the higher levels of demagnetization yielded incoherent magnetization, and in a few other cases a cleaned direction was not obtained by 200 mT demagnetization (the limit of the equipment used). For these groups, I fitted a Great Circle for each sample to the final portion of the demagnetization record, which comprises systematic shifts in remanence direction, using Remasoft v. 3.0. I then calculated the intersection of sample- specific great circle fits within each sample group using the statistical module in Remasoft v. 3.0.

2.4. Paleomagnetic characteristics of sediments

2.4.1. Magnetic susceptibility and magnetization intensity

Mean magnetic susceptibility of sample groups ranges from 0.20 to 56.0 x 10-4 SI units (Appendix A, Tables A1-A6); the average of all samples is 5.0 x 10-4 SI units). Sediments are magnetic, as demonstrated by the reasonably high susceptibility values and reliable remanence measurements obtained for most of the sample collection.

Susceptibility is highest for paleosols (1.0-56.0 x 10-4 SI units, average 13.0 x 10-4 SI units) and the Chijini Tuff (1.0-25.0 x 10-4 SI units, average 8.0 x 10-4 SI units). Susceptibility of the indurated tuff (cliff exposure) is five or more times higher the loose basal ash. Unweathered clastic sediments have lower values, although the range is large (0.20-13.0 x 10-4 SI units). In general, samples with magnetic susceptibility less than ~0.50 x 10-4 SI units yielded unstable magnetizations and were not useful for polarity assignment. Natural remanent magnetization (intensity) ranges from 0.1 to 1350 mA/m and generally, only samples from volcanic units or paleosols have intensities larger than 100 mA/m. Most samples with magnetizations less than ~0.5 mA/m provide incoherent data.

2.4.2. Demagnetization characteristics and magnetic stability

Stepwise alternating field demagnetization (magnetic cleaning) improves clustering of directions about the expected north or south Geocentric Axial Dipole field direction (Fig. 2.3a-b), indicating that secondary remanence components were removed. Most samples show magnetizations characteristic of magnetite, although some include harder magnetization components that were not removed by 200 mT AF demagnetization, suggesting the presence of hematite. Often these samples were also slightly redder in color. Many non-volcanic units include multiple magnetization components (e.g. Fig. 2.4). A low-coercivity component, presumably viscous remnant magnetization (VRM), was removed with 5-20 mT AF demagnetization. For many of the normally magnetized samples and some of the reversely magnetized samples, this viscous component is aligned with or close to Earth’s present field (PEF: D = 352.5°, I = -10.1) for the study area. Samples from the Chijini Tuff are very stably magnetized; the primary component is isolated after 5 mT AF demagnetization, and in many cases is reliably characterized by the NRM. Regardless of material, samples typically contain a high-coercivity (primary) component which decays in linear fashion toward the origin, when plotted on an orthogonal projection. This behaviour is often similar for all samples within a group and is assumed to record detrital remanent magnetization (DRM). An additional secondary component, intermediate between the presumed VRM and primary

DRM, was present in a small number of samples. This secondary component may be an artefact of weathering.

Figure 2.3. Equal-area stereographic projections of paleomagnetic directions. a) Natural remnant magnetization (NRM) of all samples (n = 624). b) Primary remnant magnetization determined after magnetic cleaning (n = 624). c) Mean primary remanence direction for all normally magnetized samples (open circle; n = 425), all reversely magnetized samples (closed circle; n = 199), antipode of all reversely magnetized samples (split circle), and all samples regardless of polarity (normal samples and antipode of reversed samples; cross). d) Means by stratigraphic section for normal and reversed samples. Open and closed circles are upper and lower hemisphere projection, respectively. Dashed and solid circles are angular errors (α95). PEF (star) and GAD (triangle) are Earth’s present magnetic field direction and the geocentric axial dipole location, respectively, for the sampling location.

Figure 2.4. Examples of demagnetization characteristics showing the range of magnetization types.

Orthogonal plots and equal-area stereographic projections of remanence with step-wise AF treatment. Samples show normal magnetization (a-d), reversed magnetization (e-h), normal magnetization with reversed overprint (i-l), and reversed magnetization with normal overprint (m- p). For each, magnetization examples comprise two high-quality records and two lower quality records of sufficient quality to approximate primary remanence. Open and closed circles on orthogonal plots represent vertical and horizontal planes, respectively. Open and closed circles on stereo plots are upper and lower hemisphere projection, respectively. Examples include all six magnetostratigraphic sections and all material types (silt/sand lenses, till matrix, gravel matrix, paleosols, and tuff). Section abbreviations: PTW, Patapatani West; PTE, Patapatani East; TNG, Tangani; MIN, Minasa, PUR, Purapura; JKT, Jacha Kkota. Although of high quality, one sample (i) was not included in the statistics as its primary remanence direction is an outlier from the other five samples in this group.

Primary remanence inclinations (Fig. 2.3b) show a large degree of scatter that likely represents recording of secular variation of Earth’s magnetic field by the sample collection. Outliers (about 10% for normal magnetizations and 15% for reversed magnetizations) likely reflect lower fidelity of sediment recorders where deposits are coarse grained (predominantly tills) or the incomplete removal of overprints (unconsolidated sediments preclude thermal demagnetization). Nevertheless, polarity can be confidently determined from most of the outlier samples. Given the low sampling latitude of the study area (16.5°S), declination values are the main determinant in assigning polarity to individual units. A few outlier samples with steep inclinations mismatched to declinations (i.e. steep positive inclination and northerly declination, or steep negative inclination and southerly declination) are not used for polarity assignment or statistical analysis. However, samples with shallow inclinations mismatched with declinations are considered for polarity assignment and, if in agreement with the other samples in the group, for statistical analysis. Shallower inclinations than the geocentric axial dipole (GAD) may reflect secular variation, or be the result of loading (‘inclination shallowing’) of fine sediment deposited in water (Granar, 1958; Blow and Hamilton, 1978).

All units sampled at the Purapurani section are normally magnetized. The other five sections include both normal and reversely magnetized samples. Mean remanence directions by polarity at all sections fall within 10° of the GAD direction (Fig. 2.3c; Table 2.2). Angular error of the section-specific polarity means typically decreases as the number of samples (n) increases and the coarseness of deposits decreases. Error limits

are also slightly higher for reversely magnetized sample populations, where incomplete removal of normal overprints may prevent the primary remanence direction from being completely (accurately) resolved.

Table 2.2. Remanence directions by section and polarity.

Site Grouping n D I k α 95

Patapatani West Normal-polarity samples 179 4.7 -31.8 17.77 2.6 Reverse-polarity samples 72 178.9 38.3 8.02 6.3 All specimens* 251 3.1 -33.6 12.93 2.6

Patapatani East Normal-polarity samples 26 358.2 -32.7 26.14 5.7 Reverse-polarity samples 15 178.5 25.1 9.15 13.4 All specimens* 41 358.3 -30.0 15.60 5.8

Tangani Normal-polarity samples 75 353.7 -22.1 20.48 3.7 Reverse-polarity samples 85 176.7 28.4 22.90 3.3 All specimens* 160 355.2 -25.4 21.06 2.5

Minasa Normal-polarity samples 38 357.1 -27.1 40.67 3.7 Reverse-polarity samples 23 171.7 36.1 25.99 6.1 All specimens* 61 355.2 -30.5 30.44 3.4

Purapura Normal-polarity samples 66 359.1 -33.1 16.78 4.4 Reverse-polarity samples ― ― ― ― ― All specimens* 66 359.1 -33.1 16.78 4.4

Jacha Kkota Normal-polarity samples 41 353.1 -27.3 29.69 4.2 Reverse-polarity samples 4 187.9 38.2 95.20 9.5 All specimens* 45 354.3 -28.4 28.80 4.0

All sections Normal-polarity samples 425 359.5 -29.5 18.67 1.6 Reverse-polarity samples 199 177.2 32.7 12.55 2.9 All specimens* 624 358.8 -30.5 16.08 1.4

* Irrespective of sign (upper hemisphere).

Mean cleaned remanence directions for all reversely magnetized samples and all normally magnetized samples are separated by roughly 180° of declination at sections where both polarities are present (Fig. 2.3c; Table 2.2) and by almost exactly 180° (182.3°) for the overall collection (Fig. 2.3d; Table 2.2). Passage of the ‘reversal test’ for paleomagnetic stability indicates that AF demagnetization satisfactorily removed secondary remanence components, yielding a primary DRM, and that the sample size is sufficient to average out the effects of secular variation (Butler, 1992, p. 127). The mean

of all normal polarities and antipodal reversed polarities has an inclination identical to that of the GAD for the sampling sites (-30.5°) and a mean declination 2.2° degrees west of the GAD direction (Fig. 2.3d).

Of the total sample collection (n = 808), 88% (n = 707) are of sufficient quality to assign polarity. Of these 707 samples, 83 are statistical outliers compared to the rest of their sampling group. Calculated remanence directions are, therefore, based on 77% (n = 624) of the sample collection (Fig. 2.3). In paleomagnetic analysis of glacial units, 50% or more of samples often have weak or unstable magnetizations (Barendregt et al., 2010), probably due largely to the presence of randomly oriented sand or pebbles. The sample collection from the La Paz area is thus of good quality given the coarse nature of the sediments at most sampling locations.

2.5. Lithostratigraphy and magnetostratigraphy

Starting with the Patapatani West section nearest the Cordillera Real (Figs. 2.5- 2.6), the lithostratigraphy and magnetostratigraphy of the six sections (Figs. 2.7-2.12; Tables A1-A6) are described below for a transect extending south into the generally level portion of the plateau. Clast lithologies in gravels and diamictons are almost exclusively of two types – granite and metasedimentary rocks (argillite and phyllite) (Fig. 2.6) – the relative proportions of which are represented vertically through each section (Figs. 2.7-2.12). The lithology of fine-grained units (silt and sand) and of the matrix material of coarse-grained units (gravel and diamicton) was not systematically characterized, but is broadly of four types (Figs. 2.7-2.12): largely granitic, largely argillitic/phyllitic, similar amounts of granite and argillite/phyllite, and uncertain. All diamictons are matrix-supported and include common striated and faceted clasts (Fig. 2.6b,c,d,e,f,h,k). Gravels are typically matrix-supported and in nearly all cases lack striated and faceted clasts (Fig. 2.6j). Contacts between most units are sharp to slightly gradational and planar to slightly undulating.

section of age reported - Ar 39 Ar/ 40 1962 ) type

ion (PTW),ion including the Chijini Tuff and sediments middle ground; the sample of the Chijini Tuff that yielded the

. View south down the Río Kaluyo valley showing the general appearance glacial deposits below the Chijini the deposits glacial below general appearance showing the valley Kaluyo Río the down Viewsouth Tuff

. 5 . 2

the Patapatani drift, is located across the valley in the left Figure The exposure in the right foreground is the middle portion of the PatapataniWest sect below and just above (45it to 110 m Fig. in 2.7 ). The Patapatani East section (PTE, Fig. 2.8), which is Dobrovolny’s ( here collectedwas at this site. La Paz and El Alto are visible in the background (~5 kmsouth). 31

Figure 2.6. Examples of sediments and paleosols at the Patapatani West section.

(see Fig. 2.7 for stratigraphic position of each photograph). a) Poorly sorted, weakly stratified gravel at the base of the section (unit 1). b) Matrix-supported diamicton (unit 3). c) Glacially striated phyllite clast within diamicton (unit 5). d) Weakly stratified, matrix-supported diamicton with faceted granite and phyllite clasts (unit 6). e) Contact between two weakly stratified diamictons (units 7 and 8) with a well developed, laterally extensive paleosol. f) Matrix-supported diamicton dominated by granitic clasts (unit 9). g) Cemented portion of the Chijini Tuff (unit 10) cut by a high-angle fault. h) Striated phyllite boulder in matrix-supported diamicton (unit 11). i) Well developed paleosol at the top of unit 14 with pedons supporting clay skins. j) Erosional contact between granitic gravel (unit 16) and overlying mainly phyllitic gravel (unit 17). k) Weakly stratified diamicton (unit 19). l) Soil developed in gravel directly below the Altiplano surface (unit 20).

Oxidized zones with sharp upper contacts at all six sections show varying combinations of columnar to blocky structure forming pedons, clay skins on clasts and pedons, and high magnetic susceptibility relative to underlying material. These are common pedogenic features (Brady and Weil, 2002) and the contacts are thus interpreted as paleosols and record long periods of subaerial exposure and weathering. I attribute the magnetic enrichment in these horizons (Tables A1-A6) to authigenic production of ferromagnetic minerals during pedogenesis, which is common in interglacial soils (Opdyke and Channell, 1996, p. 46).

I classify the paleosols as either poorly developed or well developed (Figs. 2.7- 2.12) depending on the degree of development of soil horizons and pedogenic structures (Retallack, 1988; Catt, 1990). The soil developed on the Altiplano surface (Fig. 2.6i) has an intermediate degree of soil development between the two end members. Poorly developed paleosols, which are more weakly developed than the Altiplano soil, have B horizons thinner than 30 cm, minor reddening, and little or no clay translocation and pedon formation. Well developed paleosols are more strongly formed than the modern Altiplano soil and have B horizons thicker than 50 cm, with pedons coated with clay. They are evidence of strong physical and chemical weathering. Two soils at the Patapatani West section (Fig. 2.6 are examples of well developed soils; they display reddening that decreases downward (Fig. 2.6e) and pedons with clay skins (Fig. 2.6i).

2.5.1. Patapatani West

The Patapatani West section (232 m in height; Figs. 2.5 and 2.7, Table A1) is located on the west bank of Río Kaluyo, where it curves east above the Limanpata landslide. It is the farthest upstream exposure in the Río Kaluyo/Choqueyapu valley (Fig. 2.1b) and consists of two exposures that are easily correlated using a 10-m-thick tuff at the same elevation. The upper exposure (137 m) extends from the base of the tuff to the Altiplano surface and is exposed in a gulley entering the west side of Quebrada Aquatiña. The lower portion (105 m) was exposed during recent (ca. 2007) roadwork just downvalley of Quebrada Aquatiña. The top of this lower section aligns with Dobrovolny’s (1956, 1962) 7-m type section of the Patapatani Drift on the opposite side of the valley and thus greatly extends the exposure below it (Fig. 2.5). Nineteen metres of the lower section is covered by spoil dumped downslope during road construction. A 3-m exposure on the west side of Quebrada Aquatiña provides the only details on stratigraphy and paleomagnetism in this largely covered zone (Fig. 2.7). The incised Altiplano sequence is buried beneath late Pleistocene and Holocene glacial and colluvial deposits between the base of the exposure (4170 m asl) and river level (4125 m asl).

Stratigraphy

The section consists almost entirely of diamicton and gravel, with a 10-m-thick tuff ~100 m above the base (Fig. 2.7). It records at least six polarity reversals. The sequence below the tuff comprises normally magnetized gravel (unit 1; Fig. 2.6a) and stratified dipping diamicton (unit 2), overlain by 87 m of massive to weakly stratified diamicton (units 3-9; Fig. 2.6c-f). I differentiate eight diamictons on the basis of an angular unconformity (top of unit 2), polarity reversals (units 3-5), and four oxidized, laterally continuous zones characterized by downward decreasing hue, clay skins on clasts, vertical columnar structures and sharp upper contacts (top of units 5-8; e.g. Fig. 2.6e). Polarity changes from reversed to normal across the second of these zones. The upper three oxidized zones have higher values of magnetic susceptibility that the sediments below them (Table A1). The proportion of granitic clasts increases upward from <10% in the gravel and dipping stratified diamicton (units 1 and 2), to ~50% in the middle diamictons (units 3-7), to ~90% in the units immediately below the tuff (units 8 and 9). All

Figure 2.7. Lithostratigraphy and magnetostratigraphy of the Patapatani West section. See Figs. 2.1 and 2.2 for location. See Tables A1 (Appendix A) for magnetostratigraphic details.

nine units below the tuff include striated and faceted clasts (e.g. Fig. 2.6c). Elongate clasts in the seven diamictons for whici I measured fabrics are preferentially oriented from SW-NE to SE-NW (Appendix B).The tuff (unit 10) is normally magnetized and consists of 20-45 cm of loose friable basal ash and ~9.5 m of cliff-forming, weakly cemented ash. It includes rare granitic and phyllite pebbles. Several high-angle faults cut the tuff in the the road cut exposure (Fig. 2.6g), but the sense and magnitude of offset are uncertain due to the homogeneity of the tuff. These structures likely continute into the underlying diamicton sequence, but covered by colluvium. At the southern margin of the road cut and in another road cut ~200 m downstream, the tuff has slumped several metres or more. No faulting is evident in the tuff at the natural exposure in Quebrada Aquatiña 120 m farther north.

Above a sharp, slightly wavy contact, the tuff is overlain by about 72 m of diamicton (units 11- 15) separated by three oxidized zones with abrupt planar upper contacts, columnar structure, and higher magnetic susceptibility (Table A1). A ~6-cm- thick layer of white homogenous silt-size ash rests on the top of the oxidized, columnar horizon at the top of unit 11 across a diffuse, slightly wavy contact. The silt is chalky and contains occasional (<5%) mineral crystals, and itself has a weakly developed columnar fabric and further magnetic enrichment (Table A1). Glass shards from the ash are optically isotropic, contain vesicle-wall fragments and have worn edges. Polarity changes from normal to reversed across the paleosol developed on unit 13. The oxidized zone at the top of unit 14 (Fig. 2.6i) is nearly devoid of clasts and has strongly developed pedogenic columnar structure (main columns are ~20 cm wide and up to 60 cm long, although finer columns occur within them). Excavation of this zone revealed ~25 cm of reddish, slightly waxy, well sorted clayey sand overlying ~15 cm of reddish brown, blocky clayey sand with waxy coatings on peds. All five diamicton units contain subangular to subrounded stones, and many of them are striated and faceted (e.g. Fig. 2.6h). About 90% of the stones are granitic. A sample of 50 elongate stones at the base of unit 11 shows a strong mode trending NE-SW (Appendix B).

About ~175 m above the base of the section, the sequence grades over a vertical distance of ~25 cm into poorly stratified, cobble-boulder gravel (unit 16), comprising largely of subrounded to rounded granitic clasts up to 2 m across. The gravel ranges

from clast supported to matrix-supported, with a matrix of coarse sand. Localized, ~1- cm-thick infillings of clayey silt between a few boulders were the only material suitable for paleomagnetic sampling. All six samples collected from these infillings are reversely magnetized, and at least one of these six samples (Fig. 2.4p) has a normally magnetized secondary remanence component. Above the granitic gravel is a more poorly sorted gravel (unit 17 and 18) with sub-rounded to subangular clasts consisting largely of phyllite and argillite. The contact between the two gravel units is erosional, with a highly irregular contact (Fig. 2.6j). Locally, a boulder lag marks the contact. Only two sets of samples were obtained from the gravel; one set just above the top of the granitic gravel is reversely magnetized (unit 17), and the other at the top of the non-granitic gravel (unit 18) is normally magnetized.

A thin laminated silt bed overlies the phyllitic/argillitic gravel and separates it from ~16.5 m of normally magnetized diamiction (unit 19), which in turn is capped by up to 1 m of gravel (unit 20) that directly underlie the Altiplano surface. Clasts in units 19 and 20 are almost entirely sub-rounded to sub-angular argillite and phyllite (Fig. 2.6k-i). The upper contact of the diamicton is oxidized and to a depth of ~0.2 m and has magnetic susceptibility 2-4 times greater than the rest of the unit and underlying gravel units (Table 1).

Samples from three stratigraphic levels within the normally magnetized upper 20 m of the sequence (Fig. 2.7) – top of the non-granitic gravel (unit 17, Fig. 2.4k), middle of the uppermost diamicton (unit 18, Fig. 2.4l), and the silt bed at the contact between the two (Fig. 2.4i) – show signs of a reversely magnetized overprint removed during stepwise demagnetization. Overprints in sediments are commonly of lower quality and weaker than the primary remnant magnetization, and thus are typically less coherent. Only one sample (Fig. 2.4i; LPZ1323) of the 18 from the three uppermost stratigraphic levels demonstrates an overprint fully antipodal to the high-coercivity remanence component, for both inclination and declination, but this sample’s primary remanence direction is an outlier compared to the other five samples from the same unit. Numerous other samples from these three groups show an overprint with reversed inclination, or reversed declination only. At low-latitude sites, inclinations of overprints can be variable and non-diagnostic.

Facies interpretation

The abundance of striated and faceted clasts in all 14 diamicton units indicates that each was deposited from ice (Eyles and Eyles, 2010), either beneath ice or, more likely by glacigenic mass flows off a glacier front. Oxidized, magnetically enriched zones occur at the tops of eight diamictons. I interpret these to be paleosols recording deep weathering during subaerial exposure between glacial periods. The strong unimodal clast fabrics of the seven diamicton units for which measurements were made suggests emplacement under conditions of high shear stress, consistent with subglacial deposition (Eyles and Eyles, 2010) by ice flowing initially to the south and later to the south-southeast (Appendix B). However, I would have expected grounded glacier ice to have removed the bounding paleosols (Fig. 2.5). Additionally, the strata are very crudely layered. I thus consider it more likely that the diamictons were deposited at the margins of glaciers by glacigenic mass flows on aprons at the edge of the subsiding La Paz basin. In this environment, clasts might align transverse to the main ice-flow direction (Clague, 1974); if so, the fabrics suggest ice fronts approximately parallel the Cordillera Real. Although there is uncertainty in the details of the depositional environment, either interpretation requires glaciers to have reached at least 14 km from the high Cordillera Real prior to and shortly after deposition of the Chijini Tuff.

The Patapatani West sediment sequence is interpreted as follows. The poorly sorted basal gravel (unit 1) records the approach of glacier ice to the site. Preservation of striations on clasts in the gravel indicates deposition from a meltwater stream very near the glacier front. Subsequently, the glacier deposited diamicton (unit 2) on top of the gravel, either subglacially or as glaciogenic mass flow. Given the lack of a paleosol at the contact between the two units, they may record the same glacial advance, although this is not certain. The massive diamictions below the tuff (units 3-9) are either basal tills or ice-marginal glaciogenic mass flow deposits. They record at least seven distinct glacial events separated by periods of at least 103 years, sufficient for reversal of Earth’s geomagnetic field and soil formation to depths of ~0.2 m to over 1 m below former land surfaces. Tuff was subsequently emplaced on a weathered till surface through a combination of airfall and one or more pyroclastic flows. Tills above the tuff (units 11-15) record at least four additional glacial phases separated ice-free periods

lasting probably on the order of 103 years. Well after the first post-Chijini glacier withdrew, ash of a distant volcanic eruption blanketed the soil and was in turn weathered during a lengthy period prior to re-advance of ice.

Pebbles, cobbles, and boulders in both the granitic and non-granitic gravels are rounded, suggesting either fluvial or glaciofluvial deposition and transport over a distance sufficiently far from the ice front to remove striations and faceting. Given the great thickness of some of the gravel units, I interpret them to be glaciofluvial or paraglacial in origin. The major difference in the clast composition of units 16 and units 17 and 18 indicates a change in provenance, specifically a change in the location of the source area or the rock exposed at the same location, and hints at a substantial hiatus. However, the polarity reversal within the non-granitic gravel (units 17 & 18) suggests either prolonged aggradation that spans a geomagnetic field reversal or an unrecognized unconformity within the unit.

The uppermost till (unit 19), below the Altiplano surface records at least the last glaciation prior to abandonment of the present upland surface. The paleosol at its upper contact, however, indicates lengthy subaerial exposure of the till prior to deposition of the final cap of non-granitic gravel on the Altiplano surface.

2.5.2. Patapatani East

The Patapatani East section (50 m in height; Fig. 2.8, Table A2) is located low on the eastslope of the Río Kaluyo valley directly east of the Patapatani West section. Dobrovolny (1956, 1962) describes the upper ~25 m of this section and assigns the diamicton units directly below the tuff to the Patapatani Drift. A road cut created in 2003 or 2004, based on GoogleEarth imagery, forms the lower 10 m of the Patapatani East section, 15 m below and 40 m southwest of the base of Dobrovolny’s (1956, 1962) natural exposure of the Patapatani Drift. Colluvial deposits cover the steep slope between the natural exposure and road cut.

Figure 2.8. Lithostratigraphy and magnetostratigraphy of the Patapatani East section. See Figs. 2.1 and 2.2 for location. See Tables A2 (Appendix A) for magnetostratigraphic details.

Stratigraphy

The road cut exposes 3.5 m of matrix-supported diamicton (unit 1), overlain across a sharp undulatory contact by 7.5 m of poorly sorted, nearly massive gravel (unit 2). Thick, well sorted, sandy silt lenses in the diamicton occur to within 80 cm of its upper contact and contain near-vertical, millimetre-scale clastic dykes. Both the diamicton and gravel are reversely magnetized.

The natural exposure above the road cut comprises three matrix-supported diamicton units (3-5). Clast composition ranges from 60% granitics in the lower diamicton (unit 3: at least 1.5 m thick) to 80% in the upper diamicton (unit 5: 3 m thick). Elongate clasts in the middle, normally magnetized diamicton (unit 4: 1.5 m thick) have a preferred NW-SE orientation (Appendix B). The sharp planar contact between the lower two diamicton coincides with the top of a 30-cm-thick, strongly oxidized zone

characterized by low clast content, a columnar structure, and ferro-manganese oxide coatings on ped surfaces. The contact between the upper two diamictons is a slightly recessive zone with protruding subangular granitic boulders. The diamicton sequence is abruptly overlain by 5 m of normally magnetized tuff (unit 6). Its thin (10-15 cm) basal zone of loose friable ash transitions upward over a few centimetres into hard tuff that forms a pronounced overhang.

The tuff is capped by 1 m of poorly sorted gravel (unit 7) containing granitic and argillite/phyllite pebbles. The gravel, in turn, is overlain across a sharp undulating contact by diamicton (unit 8) that is similar to unit 5, but contains larger boulders. This unit is discontinuously exposed for ~20 m upslope. What may be the same diamicton crops out much higher on the otherwise grassy surface of the slope and is characterized by the presence of large (0.75-2 m) sub-rounded to sub-angular granitic boulders projecting from the slope. Unit 7 gravel and base of the unit 8 diamicton are normally magnetized.

Facies interpretation

The presence of striated, faceted clasts in all five diamicton units at the Patapatani East section and the strong clast fabric in unit 4 suggest deposition as till beneath or at the margins of glaciers. A glacier reached the section and deposited the lowest till (unit 1) during a period of reversed field directions. When this glacier retreated, or possibly during a subsequent glaciation, a high-energy stream deposited coarse gravel (unit 2), likely during the same period of reversed polarity. After a passage of time sufficient for the geomagnetic field to reverse, tills (units 3-5) were deposited beneath or at the margin of a glacier. At least two distinct glaciations were separated by lengthy period during which a well developed paleosol developed (unit 3-4 contact). Subsequently, the Chijini Tuff was emplaced on a subaerial surface. Gravel (unit 7) may be outwash deposited near an ice margin, given that it includes striated clasts. Unit 8 records yet another glaciation during which ice reached and probably overrode the site.

2.5.3. Tangani

The Tangani section (250 m in height; Fig. 2.9, Table A3) is located in a deeply incised network of gullies and hoodos on the east bank of Río Choqueyapu, 0.5 km upstream from the hairpin curve on the autopista. The Chijini Tuff is not present at this section and it is not exposed in the gorge incised into the valley floor below. However, Bles et al. (1977) and Ballivián et al. (1978) report a thin exposure of Chijini Tuff just downstream at an elevation (3920 m asl), which is the approximate elevation of the base of the Tangani section. The uppermost 80 m of the main section are exposed only in vertical inaccessible headwalls, however exposures in Quebrada Tangani, ~300 m to the southeast, provide access to units 10-12, which I correlate to the upper part of the Tangani section by elevation and by a thick, laterally persistent silt bed 185 m above the base of the main section.

Stratigraphy

At the base of the Tangani section, two normally magnetized, predominantly granitic, weakly stratified, pebble-cobble gravels comprising units 1 (8 m) and 3 (3.5 m) are separated by 3 m of massive, multi-lithic, matrix-supported diamicton with clasts up to boulder size (unit 2). Unit 3 gravel is overlain by two multi-lithic, matrix-supported diamicton units (4 and 5). The lower 80-m-thick diamicton is massive and includes numerous granite boulders, whereas the upper 7-m-thick diamicton is weakly stratified and contains only pebbles and cobbles. The contact between the two diamictons is marked by a weakly oxidized zone up to 20 cm thick with fewer clasts and higher magnetic susceptibility (Table A3). Polarity changes from normal to reversed across this contact. Unit 5 becomes more stratified in its upper 2 m, grading from massive diamicton into weakly stratified, poorly sorted, pebble-cobble diamicton with laminated fine sand lens at the top of the unit that is deeply oxidized and has weak columnar structure.

The diamicton is overlain by 14 m of weakly stratified, predominantly granitic pebble-cobble gravel (unit 6) ranging from clast- to matrix-supported. It contains multiple decimetre-thick, subhorizontal, weakly laminated silty fine sand beds. The gravel is overlain by two diamictons, units 7 (5.5 m) and 8 (7 m), similar to unit 5 below it. The

Figure 2.9. Lithostratigraphy and magnetostratigraphy of the Tangani section. See Figs. 2.1 and 2.2 for location. See Tables A3 (Appendix A) for magnetostratigraphic details.

contact between the two diamictons and between the lower diamicton and the gravel below are sharp. The uppermost part of unit 8 fines upward, is oxidized, and has a weakly columnar structure. Its magnetic susceptibility is four times higher than non- oxidized diamicton directly below it (Table A3). The lower of the two diamictons (unit 7) has indeterminate polarity, whereas the upper diamicton (unit 8) has reversed polarity.

The upper half of the section comprises over 100 m of weakly stratified, pebble- cobble-boulder gravel (units 9-12). Clasts are dominantly granitic and rounded to sub- rounded. This gravel sequence is interrupted by a single, slightly finer oxidized zone (unit 9-10 contact) with enhanced magnetic susceptibility (Table A3) 13 m above the diamicton (unit 8) below. The gravel units of this sequence contain several silt beds, one of which is ~1.5 m thick and has a lateral extent of several hundred metres. The polarity of the gravel sequence is reversed except for a single, coherent, normally magnetized group (unit 11: e.g. Fig 2.4d) ~40 m above the top of unit 9; the gravel sequence above unit 9 is divided into units 10, 11 and 12 based on these polarity reversals. A contorted, poorly sorted non-granitic gravel (unit 13) cuts across the entire sequence at a steep angle parallel to the modern Río Kaluyo valley slope along an angular unconformity. The matrix of this gravel layer records normal, highly noisy remanence.

Facies interpretation

Striated and faceted clasts in all five diamictons at the Tangani section indicate deposition by glaciers. The oldest diamictons at the locality (units 2 and 4) are separated by coarse gravel units (1 and 3) deposited by high-energy streams. Unit 4 records glacial deposition in a setting that allowed for deposition of thick till, likely at an ice margin. The surface of this till was exposed subaerially for a period sufficiently long for a soil to form and the geomagnetic field to reverse. This nonglacial period was followed by renewed glaciation with deposition of the unit 5 diamicton. An even longer period of subaerial exposure followed, producing the well developed paleosol at the top of unit 5. Unit 6 likely records deposition by a braided gravel bed stream with silt lenses representing slack-water deposits.

Two additional glacial events (units 7 and 8) were each followed by a lengthy depositional hiatus and associated pedogensis. The thick gravel sequence in the upper part of the section (units 9-12) indicates prolonged valley-wide deposition by a high- energy stream, probably in a glacial setting. The occurrence of two, closely spaced polarity reversals within the gravel sequence suggests that gravel deposition spanned a substantial period of time, perhaps during more than one glaciation.

2.5.4. Minasa

The Minasa section (410 m in height; Fig. 2.10, Table A4) is located along Río Minasa, a west-bank tributary of Río Orkojahuira in barrio Villa El Carmen. It extends from Puente Colonial (~3910 m asl) to Huari Pampa (4320 m asl), an undeveloped section of the Altiplano surface between Ríos Kaluyo/Choqueyapu and Chuquiaguillo/Orkojahuira. Low natural exposures and road cuts along the south bank of Río Minasa form the lower half of the section. High natural exposures along the steep westernmost gulley provide continuous exposure of the upper half of the section. Río Minasa follows the trace of a high-angle fault. Strata on the south side of the fault have been displaced upward relative to the north side from ~10 m at the base of the section to <50 cm just below the plateau surface. There is no obvious displacement of the Huari Pampa surface along this structure.

Stratigraphy

The normally magnetized, 10-m-thick Chijini Tuff (unit 1) at the base of the section is overlain by over 307 m of weakly stratified, pebble-cobble-boulder gravel (units 2 [134 m] and 3 [173]) containing occasional silt beds. The lower third of the gravel sequence is multi-lithic, the middle is dominantly granitic, and the upper third is mainly non-granitic. The lithologic changes happen over vertical distances of tens of metres between these zones. A 20-cm oxidized zone with greater matrix content, a sharp planar upper contact, and elevated magnetic susceptibility (Table A4) marks the top of unit 2, just above the transition from multi-lithic to granitic gravels. The thick gravel sequence is

Figure 2.10. Lithostratigraphy and magnetostratigraphy of the Minasa section. See Figs. 2.1 and 2.2 for location. See Tables A4 (Appendix A) for magnetostratigraphic details.

reversely magnetized at all sampling heights, although there are large vertical sections of gravel with no samples.

The overlying normally magnetized sequence comprises alternating weakly stratified diamicton and gravel units (4-15) varying from 3 to 16 m thick. All units contain predominantly non-granitic clasts (<20%, and in most cases <10%). Several units (9, 10, 12, 14, and 15) are gravelly at the base and grade upward over several metres into diamicton; the lower contacts of the gravels are sharp. The gravel-diamicton sequence contains three decimetre-scale, slightly finer, horizontal zones of oxidation with weak columnar structure (upper contacts of units 11, 14, and 15). The highest two have elevated magnetic susceptibilities (Table A4), and the uppermost forms the modern Altiplano surface.

Facies interpretation

Following the eruption of the Chijini Tuff (unit 1), a very thick gravel sequence (units 2 and 3) accumulated at the Minasa site. These two units record major aggradation of a trunk river system, likely by meltwater streams flowing from one or more glaciers to the north. The presence of just one paleosol suggests few breaks in deposition, and the small number of fine beds within the two gravel units indicates deposition on a braidplain by high energy river flow. The diamicton units are tills, but the glacial environment included subglacial or proglacial deposition of gravels. The interfingering of the till-gravel units in the upper part of the section and the general lack of paleosols complicate identification and discrimination of separate glacial events. The presence of two paleosols (capping units 11 and 14), however, suggest at least three separate glacial periods. Following the last of these, the modern soil at the plateau surface began for form within till of the final glacial period (unit 15).

2.5.5. Purapura

The Purapura section (250 m in height; Fig. 2.11, Table A5) mostly follows the old railway ascending from La Paz to El Alto and includes natural exposures as well as

Figure 2.11. Lithostratigraphy and magnetostratigraphy of the Purapura section. See Figs. 2.1 and 2.2 for location. See Tables A5 (Appendix A) for magnetostratigraphic details.

railway and road cuts. Nearly 130 m of discontinuous exposure in the middle part of the section has not been previously described in detail or sampled paleomagnetically. The section roughly coincides with the ‘Pura Pura’ (Aqueducto) section of Bles et al., (1977) and Ballivián et al. (1978), particularly the lowest 80 m. The Purapura section is ~2 km up valley of the approximate location Thouveny and Servant (1989) give for their Purapura magnetostratigraphic section.

Stratigraphy

The base of the section (unit 1) consists of 1.5 m of weakly stratified, well rounded, granitic pebble-cobble gravel. A 1.5 m zone of tuffaceous, silty very fine to medium sand with scattered pebbles (unit 2) overlies the gravel and, in turn, is overlain across a sharp contact by 5 m of Chijini Tuff (unit 3). The tuff has a slight recessive zone at the base, corresponding to a thin layer of non-lithified ash. Both the tuff and underlying tuffaceous sand are normally magnetized. The tuff is sharply overlain by 4 m of laminated sand and silt (unit 4).

The fine sediments of unit 4 are overlain by weakly stratified diamicton capped by a thin, laterally continuous silt bed (unit 5; 11 m), and by three massive diamicton units: 6 (31 m), 7 (3.5 m) and 8 (0.5 m). The diamicton units are separated by two ~10-cm- thick paleosols characterized by clay enrichment, oxidization, and elevated (3-15 times) magnetic susceptibility (Table A5). All three diamictons are multi-lithic and normally magnetized. An overlying horizontally stratified, rounded gravel (unit 9) is almost completely granitic and of indeterminate polarity. The gravel is at least 50 m thick, but was inaccessible and has not been described in detail or sampled.

The upper 60 m of the section comprises normally magnetized, non-granitic diamicton and gravel (units 10-15), but over half of this part of the section is covered by colluvium and urban development. The diamicton is massive to weakly stratified and contains typical sub-rounded to angular striated clasts. Ten metres below the plateau surface, a 15-cm-thick zone of oxidized, massive silty sand with elevated magnetic susceptibility (3-10 times that of the diamictons above and below; Table A5) divides the diamicton into two units (13 and 14). The lower contact of this oxidized zone is

gradational, whereas the upper contact is sharp and planar. The uppermost 2 m of the sequence is gravel (unit 15); no polarity data are available for this unit.

Facies interpretation

The Chijini Tuff (unit 3) overlies a fluvial channel deposit (unit 1). Silt and sand (unit 4) accumulated in a pond or lake following emplacement of the tuff. Subsequently, glaciers reached the site of the Purapura section (units 6-8) at least three times separated by nonglacial intervals of sufficient length to form paleosols. The last of the three glaciations was followed by deposition of a thick unit of paraglacial or outwash gravel (unit 9) probably in a large river valley draining the adjacent Andes.

At least two later glaciations (units 10 and 12-14) were separated soil formation (unit 13-14 contact). The predominance of non-granitic clasts and matrix in these deposits, compared to the mixed and predominantly granitic lithologies lower in the sequence, suggest a change in glacier provenance. The final of the two glaciations ended with or was followed by deposition of a thin gravel cap on the plateau surface.

2.5.6. Jacha Kkota

The Jacha Kkota section (175 m in height; Fig. 2.12, Table A6) follows a heavily gullied ridge that descends ~200 m from the Altiplano surface to the floor of the Achocalla basin just west of Laguna Jacha Kkota. The Achocalla basin formed during a gigantic, early Holocene earthflow (Dobrovolny, 1968) dated to 11,485–10,965 14C cal yr (Hermanns et al., 2012), which left this ridge as an intact remnant of the source area. The section sampled includes the Chijini Tuff and 42 m of fine-grained sediments below the tuff. The Achocalla magnetostratigraphic section of Thouveny and Servant (1989) extends from the base of the tuff 80 m upslope along the same ridge, but does not reach the Altiplano surface.

I include the magnetostratigraphy of the uppermost 10 m of the sedimentary sequence at a nearby section (~3.2 km to the southeast) to extend the Jacha Kkota

section to the Altiplano surface. The exact stratigraphic alignment of the two sections is uncertain, but in view of the similar elevations of the Altiplano surface and the Chijini Tuff at both sites, the sequence probably starts 40 to 45 m above the top of Thouveny and Servant’s (1989) Achocalla section, in agreement with the stratigraphy of the Achocalla basin margins reported by Bles et al. (1977) and Ballivián et al. (1978).

Figure 2.12. Lithostratigraphy and magnetostratigraphy of the Jacha Kkota section. See Figs. 2.1 and 2.2 for location. See Tables A6 (Appendix A) for magnetostratigraphic details.

Stratigraphy

The predominantly fine-grained sequence of the Jacha Kkota section begins at the bottom with 9.5 m of reversely magnetized laminated silt (unit 1). The silt is overlain across a sharp undulating contact by 8 m of interbedded pebble gravel, silt, and thin sand beds (unit 2). Unit 2, in turn, is overlain across a sharp contact by 5 m of cross- bedded sand and silt (unit 3) that is capped by a thin oxidized zone with a sharp planar upper contact, and magnetic susceptibility more than double that of underlying sediments (Table A6). The next unit in the sequence comprises 19.5 m of interbedded silt and sand (unit 4). It is sharply overlain by 5.5 m of Chijini Tuff with its characteristic thin loose ashy base (unit 5). Units 2 through 5 are normally magnetized. Above the tuff are at least 10 m of interbedded silt, sand, and pebble gravel (unit 6). These sediments are coarser and lithologically more variable than the correlative reversely magnetized, supra-tuff sediments (mainly clay and silt beds) described by Thouveny and Servant (1989, p. 340). Deposits directly below the plateau surface comprise weakly laminated, largely non-granitic diamicton (unit 7) with fewer and smaller clasts than the uppermost diamicton nearer the Andes, and a capping poorly sorted gravel (unit 8) of similar composition. Both units are normally magnetized. A silty oxidized zone characterized by weak columnar structure, and a sharp planar upper contact separates the capping gravel from the diamicton that underlies it. This paleosol extends for at least several kilometres around the rim of the basin.

Facies interpretation

The Jacha Kkota section records low-energy fluvial and fluvio-lacustrine environments during most of its history. Silts dominate the stratigraphy, but there were periodic increases in energy marked by sand and pebble gravel beds. Several non- depositional and erosional hiatuses interrupted deposition. At least two hiatuses represent periods sufficiently long for pedogenesis (top of unit 3) and reversal of the magnetic field (unit 1-2 contact). The final stages of aggradation, following emplacement of the Chijini Tuff, involved higher energy transport and deposition (units 7-8) interrupted by a period of soil formation (unit 7-8 contact). Given its finer and slightly more stratified

nature, relative to other sections, the diamicton (unit 7) likely records ice-marginal deposition (possibly as mass flows).

2.6. Correlations and chronology

2.6.1. Section correlations

Multiple lines of evidence suggest the same tuff (Chijini Tuff) is exposed throughout the western part of the La Paz and Achocalla basins, and specifically at sections that I investigated (Appendix C) except for the Tangani section, which lacks tuff. Variability in magnetic intensity and magnetic susceptibility at individual exposures indicate that these properties alone are insufficient to differentiate tuffs in the La Paz area, despite Thouveny and Servant’s (1989) suggestion that different tuffs in the La Paz basin have distinct magnetic intensity. The most reliable published radiometric ages suggest all dated tuff exposures above ~3550 m asl (elevation of the Cota Cota tuff) were emplaced around 2.74 Ma (Table 2.3; Fig. 2.13a). The normal polarity of all tuffs for which remanence is known (Tables A1-A6; Thouveny and Servant, 1989) is in agreement with this late Gauss age (Fig. 2.13a). Similarity of remanence directions of non-faulted tuff exposures (Fig. 2.13b) suggests that the tuff was emplaced within a sufficiently short period (<10-100 a) so that secular variation is undetectable (cf. Bogue and Coe, 1981). The remanence of tuff at the Patapatani West section appears to be affected by faulting (Fig. 2.13c).

Lateral continuity of the Chijini Tuff, Altiplano surface and, to a lesser degree, lithologically conspicuous gravel units, enable reliable alignment of the magnetostratigraphic sequences of the six sections, as well as the upper three sections measured by Thouveny and Servant (1989) (Fig. 2.14). The sections of Thouveny and Servant (1989) are their Viscachani section on the east side of the Río Choqueyapu valley, ~3 km downstream of the Tangani section; their Purapura section, a short but unknown distance south of my Purapura section; and their Achocalla section in the same

ample " " " Clapperton (1979) Clapperton (1979) Source (1992) al. et Mashall (unpublished) 1 Appendix Lavenu et al. (1989) Lavenu et al. (1989) Lavenu et al. (1989) (1992) al. et Mashall Lavenu et al. (1989) Lavenu et al. (1989) Lavenu et al. (1989) a a " " " by source) Chacaltaya Chacaltaya Chijini Chijini Chijini Patapatani Patapatani Chijini Chijini Chijini Chijini Sopari Tuff (suggested

* * * * * d 7 of Lavenu et al., 1989 ) indicates s 0.4 0.053 0.2 0.04 0.13 0.1 0.1 5 0.037 0.072 0.14 0.1 1.3 1.1 0.1 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± (Ma) ure 7.7 2.8 2.7 2.8 1.6 182 2.65 2.38 3.28 11.6 12.1 2.74 3.27 Age and error 2.757 3.082 Fig

sanidine biotite sanidine Mineral dated feldsparK feldsparK feldsparK biotite biotite sanidine biotite sanidine biotite feldsparK biotite plagioclase (number 15). Those younger than 4.4 Ma are plotted Fig.in

Ar Ar Ar Ar

39 39 39 39 K-Ar K-Ar K-Ar K-Ar K-Ar K-Ar K-Ar K-Ar K-Ar K-Ar K-Ar Ar/ Ar/ Ar/ Ar/ Method 40 40 40 40 e 4160 4160 3900 3900 " " 3880 " 3950 4190 3900 3900 4100 4020 ― (masl) Elevation c c " " " 1992 ), not standard mean error. ― ― ― a new unpublished age 68° 7.83’ W 68° 9.33’ W 68° 9.33’ W Longitude 68° 7.83’ W 68° 7.25’ W 68° 8.12’ W 68° 8.04’ W 68° 5.85’ W 68° 5.88’ W

14) and - " " " ― ― ―

Latitude 16° 25.80’ S 16° 25.50’ S 16° 27.67’ S 16° 28.10’ S 16° 38.22’ S 16° 25.87’ S 16° 27.67’ S 16° 27.08’ S 16° 27.05’ S 1993 ). c n given by Marshall et al. ( " " " Radiometric ages of tuff exposures in the upper Río La Paz valley system. valley La Paz Río upper the tuff in exposures Radiometric ages of o Kaluyo, left bank left Kaluyo, o Location í a for comparison with the Geomagnetic Polarity Time Scale. Río Kaluyo, right bank right Kaluyo, Río Río Choqueyapu left bank left Choqueyapu Río bank left Kaluyo, Río bank right Choqueyapu Río bank right Choqueyapu Río bankRio Chuquiaguillo, right basin Achocalla Amachuma, R Río Choqueyapu left bank left Choqueyapu Río Río Choqueyapu right bank right Choqueyapu Río bank Chuquiaguillo, right Río Rio Chuquiaguillo, right bankRio Chuquiaguillo, right

. Chuquiaguillo valley is ~4030 masl. 2.13 position on left bank, a short distance upstream of the Limanpata landslide. 3 .

― ― ― 2 " " " MB153 MB155 MB160 LGM02 LGM03 MB154 MG159 PH53a LGM01 Sample ID 2 3 5 7 8 1 4 6 9 Suggested late by Clapperton ( Coordinate is from lower slope on right bank of Río Kaluyo, but map of sample locations ( Error is standard deviatio Elevation underestimates true topographic position of tuff as local valley bottom is 3925 masl and lowest tuff observed thin is part of Typographic error in location in original source, which gives the location as 1 degree of longitude farther west (southwest o f Lake Titicaca).

11 12 13 14 15 10 Ages are from previous studies (numbers 1 Most* reliable published ages (mean 2.74 = Ma). a b c d e No. Table

gullied ridge as my Jacha Kkota section, but extending from the base of the tuff to within ~40 m of the Altiplano surface, and thus filling in much of the missing post-tuff magnetostratigraphy at Jacha Kkota.

Figure 2.13. Comparison of radiometric ages and directional means of tuff exposures of La Paz and Achocalla basins. a) Radiometric ages (and errors) of tuffs plotted with the Geomagnetic Polarity Time Scale for the late Pliocene (Gauss Chron) and Early Pleistocene (Matuyama Chron). Polarity boundaries based on the LR04 time scale (Lisiecki and Raymo, 2005), except for Réunion subchron (ca. 2.15 Ma: on Ogg et al., 2012). Ages are from previous studies (Clapperton, 1979; Lavenu et al., 1989; Marshall et al., 1992) and a new unpublished sample, using potassium (black) and biotite (grey). Most reliable ages overlap at ca. 2.74 Ma (dotted line) with a range of ca. 0.2 Ma (grey band). Table 2.2 provides details of all fifteen radiometric ages. Three biotite ages (no. 7, 8 and 13; 7.70 to 12.1 Ma) and a single sanidine age (no. 5; 182 Ma), which are considered to be erroneous ages, are not shown as they predate the 4.4-Ma upper age limit of the plot. b) Mean primary remanence directions and angular error (α95) of tuff units by section (PWT, Patapatani West; PTE, Patapatani East; MIN, Minasa; PUR, Purapura; JKT, Jacha Kkota). c) Primary remanence directions and angular error (α95) of six samples from either side of a fault cutting the tuff at the Patapatani West section showing systematic shift in remanence directions between the footwall and hanging wall.

Polarity reversals both below and above the tuff align in number and spacing between my sections (Fig. 2.14). These polarity reversals refine chronostratigraphic correlations across the 18-km transect of the Altiplano fill sequence, allowing comparison of the thickness and composition of roughly time-synchronous parts of the

sequence. The broad agreement of magnetozones in both number and relative thickness at different sections above the tuff, and repetition of the same polarity sequence at multiple sections up to 18 km apart below the tuff further support the argument that the tuffs at all sections are the same.

Similarity of primary remanence directions for paleosols at similar stratigraphic positions (Fig. 2.15) suggests coeval deposition of their parent material (Appendix D). Such agreement refines temporal correlation among some sections (correlations D, E, G, H, K, and N in Fig. 2.14) and suggests that long periods of subaerial exposure were coincident at sections up to 18 km apart. Groups with remanence directions clustering near the GAD position (Fig. 2.15b,c,f) leave open the possibility that they record similar positions of the local geomagnetic field at different times of parent material deposition, but this is unlikely considering the similar stratigraphic position of the groups. Agreement of exotic remanence directions (Fig. 2.15e and to a lesser degree Fig. 2.15d), which are much less common that those along the GAD, or remanence agreement between sections directly below as well as directly above a polarity reversal (Fig. 2.15a) make coincident positions of the geomagnetic field at different times highly unlikely; it is more probable that such agreement results from remanence acquisition during the same short time period.

2.6.2. Composite polarity sequences

The Patapatani West section spans the full chronostratigraphic range documented here and can thus be considered a master section for the upper Altiplano fill sequence in the La Paz area. With only two exceptions, magnetostratigraphy from this section records all of the polarity zones downvalley at other sections reported here and by Thouveny and Servant (1989) (Fig. 2.14). The lowermost reversely magnetized magnetozone, recorded at two stratigraphic positions at the Viscachani section of Thouveny and Servant (1989) (Fig. 2.14e), is buried beneath valley-bottom deposits in Río Kaluyo, but is almost certainly present in the La Paz Formation throughout the Achocalla basin and the southern La Paz basin, where that unit has thicknesses of >500 m.

Figure 2.14. Magnetostratigraphic correlations of sections through the Altiplano fill sequence. Sections are: a) Patapatani West; b) Patapatani East; c) Tangani; d) Minasa; e) Viscachani (Thouveny and Servant, 1989); f) Purapura (Thouveny and Servant, 1989); g) Purapura; and h) Jacha Kkota. See Figs. 2.7-2.12 and in Thouveny and Servant (1989) for details. Suggested inter-section correlation (A to N) are based on magnetozone and lithologic boundaries (likely within ca. 103-104 a) and on similarity of remanence directions (likely within ca. 10-102 a). See Fig. 2.15 and Tables A1-A6 (Appendix A) for comparison of remanence directions

Figure 2.15. Comparisons of mean remanence directions of paleosols from select stratigraphic position at separate sections. Close directional agreement supports temporal coincidence of paleosols at specific stratigraphic levels, strengthening proposed section correlations (Fig. 2.14) and suggesting occurrence of spatially extensive long-lived subaerial exposure. Correlations levels are: a) Correlation D; b) Correlation E (mean excludes directional data from lower of two paleosols at one section [grey]); c) Correlation G (mean excludes data from outlier section [TNG]); d) Correlation H; e) Correlation K; and f) Correlation N (mean excludes data from outlier section [JKT]). Open and closed circles are upper and lower hemisphere projection, respectively, of mean remanence direction of each section. Dashed and solid ellipses are angular errors (α95) for individual sections (PWT, Patapatani West; PTE, Patapatani East; MIN, Minasa; PUR, Purapura; JKT, Jacha Kkota). Crosses and heavier circles are mean directions and angular errors, respectively, of all likely correlative groups at a given stratigraphic position. Remanence directions in light grey are excluded from calculation of overall means. Star and triangle symbols are Earth’s present magnetic field direction and the geocentric axial dipole location, respectively, for the sampling location.

A short, normally magnetized section recorded in the Purapurani Gravel at the Tangani section and at Thouveny and Servant’s (1989) Purapura section is also absent at the Patapatani West section. A strong erosional unconformity at the base of the overlying non-granitic gravel and the limited thickness of the Purapurani Gravel (10 m) at Patapatani West compared to exposures to the south and southeast where it attains thickness of 100-250 m suggests the short normal magnetozone has probably been completely removed in this part of the Río Kaluyo valley. At the Tangani section only the upper limit of this magnetozone is well constrained (Fig. 2.9). Detailed sampling of the middle part of the Patapatani Gravel at Tangani, Minasa, and Jacha Kkota would better constrain this magnetozone.

At Thouveny and Servant’s (1989) Purapura section this magnetozone either spans two sampling sites (<10 m) as stated in their text (Thouveny and Servant, 1989, p. 341) and shown in their polarity sequence interpretation (their Figure 6d), or three sampling sites (~20 m) as suggested by reported remanence directions (their Figure 6d). They do not provide the position of the magnetozone within the Purapurani Gravel other than to suggest in a generalized stratigraphic interpretation (Thouveny and Servant, 1989, their Figure 2) that this ~35-m section spans the upper part of the Calvario Drift and most of the Purapurani Gravel. Given the great thickness of granitic gravel, including the new Purapura section (Fig. 2.14), it appears they have only captured part of the gravel sequence.

In conclusion, from bottom up, the composite polarity sequence (Fig. 2.14: R-N- R-N-R-N-R-N-R-N) comprises ten alternating magnetozones starting with reversed polarity (measured only at the Viscachani section of Thouveny and Servant, 1989) and ending with normal polarity of the Altiplano surface, measured at all four sections that extend to the Altiplano surface.

2.6.3. Correlation with the Geomagnetic Polarity Time Scale

The long, and likely complete, polarity record of the La Paz sedimentary sequence can be compared with the astronomically tuned Geomagnetic Polarity Time

Scale (GMPT) of Lisiecki and Raymo (2005) (Fig. 2.16). The age of the Chijini Tuff ties the middle of the composite polarity sequence to the late Gauss, shortly before the Plio- Pleistocene boundary.

Below the tuff, the relative thicknesses of fully recorded magnetozones at the Patapatani West and Viscachani sections roughly match the durations of polarity intervals of the Gauss Chron (Fig. 2.16), as would be expected for long-term sedimentation at a relatively constant rate. The Gauss Chron is preceded by a long (ca. 0.6 Ma) reversely magnetized period (4.184-3.588 Ma; Fig. 2.16b) at the end of the Gilbert Chron, likely represented by the lowest magnetozone at the Viscachani section (Fig. 2.14). It is unlikely that the latest Gilbert Chron is instead represented by part of the N-R-N-R-N polarity sequence below the tuff at the Patapatani West section, because this interpretation would require a very long late Pliocene hiatus. Additionally, the occurrence and progressive expansion of Andean ice caps recorded by glacial deposits (Fig. 2.14) of the fill sequence coincides well with the timing of global cool periods during the Gauss Chron (discussion below), but is unlikely during the Gilbert Chron (Zanclean stage), which was globally warm (Lisiecki and Raymo, 2005; Fig. 2.16c).

The normally magnetized Altiplano surface must predate the Brunhes Chron (0.780 Ma) as overprints on samples from several stratigraphic levels at the Patapatani West section directly below the plateau (Fig. 2.4i,k,l) record a reversed geomagnetic field some time subsequent to their emplacement. The two uppermost normal magnetozones of the composite sequence (N4 and N5) thus represent subchrons of the Matuyama Chron. It is unlikely, although not impossible, that either the Gilsa event (8 ka long; Channell et al., 2002) or other short-lived geomagnetic excursions of the Matuyama (Channell et al., 2002) would be captured in the stratigraphy. Capturing these short polarity events in the sporadic depositional record of the fill sequence would require their chance coincidence with a major landscape-altering depositional event. Even in marine basins characterized by continuous deposition, these events are commonly not detected (Channell et al., 2002; Laj and Channell, 2015).

Figure 2.16. Correlation of the measured polarity sequence with the geomagnetic polarity time scale and global glacial-interglacial record.

a) The Altiplano polarity sequence derived from Fig. 2.14 (periods of normal and reversed polarity in black and white, respectively) with the 40Ar/39Ar age of the Chijini Tuff in red. Stratigraphic position below the local Altiplano surface (y-axis) for magnetozones R4 to N5 is only approximate and based on relative stratigraphic positions farther south (Fig. 2.14c,d,g,h). Glacial units from the composite stratigraphic sequence are labeled as g1 to g16. b) The astronomically tuned LR04 geomagnetic polarity time scale of Lisiecki and Raymo (2005), with addition of the normal Réunion subcrhon (of the reversely magnetized Matuyama Chron) based on Ogg et al. (2012). c) The LR04 benthic ‰18 paleo-temperature profile of Lisiecki and Raymo (2005) derived form 57 globally distributed records (conditions warmer than the Holocene mean in yellow, and those cooler than the Holocene mean in blue). Marine Isotope Stages (MIS) labeled in grey (glacials as even numbers, interglacials as odd numbers). MIS scheme follows Lisiecki and Raymo (2005) from present (MIS 1) the start of the Pleistocene (MIS 104) and Shackelton et al. (1995) for the Pliocene. d) Epochs and ages of the late Cenozoic (Gradstein et al., 2012). e. Various records of the onset and expansion of glaciation from around the world (see text for details). f) Biostratigraphic sections from elsewhere in the Central Andes (Tarija basin, sub-Andean fold- thrust belt, southeast Bolivia [MacFadden et al., 1983]; Uquía Formation, Cordillera Oriental, northwest Argentina [Marshall et al., 1982; Reguero et al., 2007;]; Inchasi section, Cordillera Oriental, south-central Bolivia [MacFadden et al., 1993]), including timing of main pulses of the Great American Biotic Interchange (GABI).

Thicknesses of magnetozones above the tuff approximate the relative durations of polarity intervals of the latest Gauss and early Matuyama, implying a relatively uniform long-term accumulation rate similar to that apparent below the tuff. The two normal periods are thus most likely the Réunion event (2.121-2.153 Ma, Springer et al., 2014; 2.115-2.153 Ma, Channell et al., 2003; 2.128-2.148 Ma, Ogg, 2012) and the Olduvai subchron (1.968-1.781 Ma, Lisiecki and Raymo, 2005).

Alternatively, the uppermost magnetozone could represent the Jaramillo subchron (1.075-0.991 Ma; Lisiecki and Raymo, 2005). However, that subchron’s age, so long after the end of the Gauss, and its short duration relative to the Olduvai would require a large decrease in long-term aggradation during the early Matuyama followed by rapid deposition of the complex stratigraphy within ~30-100 m of the Altiplano surface just before 1 Ma. In this scenario, the missing normally magnetized period (either Réunion or Olduvai) requires either a substantial hiatus or a non-sampled section. By analogy, the short-lived normal magnetozone in the granitic gravels appears to have been removed by erosion at the Patapatani West section and missed during sampling at several other sections (Fig. 2.14). Absence of either of the Réunion or Olduvai subchron requires either a lengthy (ca. 30 ka or more) depositional hiatus or a major basin-wide erosional event in what otherwise appears to be a ‘regular’ aggradational record over at

least ca. 1 Ma. The non-sampling of a subchron within the Matuyama is a more likely explanation.

Some versions of the geomagnetic polarity timescale divide the Réunion event into two parts, separated by a short-duration reversely magnetized period (e.g. Mankinen and Dalrymple, 1979). However, oceanic magnetic anomalies (Cande and Kent, 1992), deep-sea sediment cores (Channell et al., 2002, 2003), and terrestrial basalt sequences (Singer et al., 2014), including a reinvestigation of the lavas from which the event was first identified on Réunion Island (Baksi and Hoffman, 2000), all suggest only a single short normally magnetized period in the 2.1-2.0-Ma interval. Given the absence of strong evidence for two separate Réunion events, the Altiplano surface can be no older than the Olduvai subchron.

Possible interpretations

My preferred interpretation is that the polarity sequence includes all polarity reversals from just below the Gauss-Gilbert boundary to shortly before the end of the Olduvai. In this scenario the fill sequence described here spans nearly 2 Ma of the latest Zanclean stage, the Piacenzian stage, and the majority of the Gelasian stage, and the modern Altiplano surface formed not long before ca. 1.8 Ma.

An alternative interpretation is that the sequence extends upward as far as the Jaramillo subchron near the end of the Early Pleistocene, and either the Réunion event or the Olduvai subchron is not represented. In this scenario, the modern Altiplano surface formed by ca. 1.0 Ma. However, features of the preserved glacial sequence make this unlikely (as discussed below).

2.6.4. Polarity signatures and ages of previously defined geologic units

Geologic units of the Altiplano fill show consistent polarities or polarity sequences between sections (Appendix E; Table 2.4), which confirm many of the lithostratigraphic

and Kkota, and

Comment Correlative to composite sequence of all seven units all seven of sequence composite to Correlative Remanence less suggests of duration a than decades centuries few or -4 ), with the aggradation terminating in the Olduvai subchron. b b - <10 ― 0.34 0.58 0.10 0.10 0.10 ≥ ≥ >1.54 Duration 2.4 Fig. 2.13a <2.1 2.74 2.74 1.95 Upper ( >1.78

>1.781 >1.781 ― >2.2 2.74 2.74 ≥2.0 1.95 Lower Approximate age range (Ma) age range Approximate <3.32 >3.32 c Polarity sequence R-N-R-N-R-N-R-N-R-N N R-N-R N-R-N-R-N N N-R R-N N

1962 ) d

. Description sequence glacial Lower sequence Gravel gravels Altiplano Upper glacialUpper sequence Polarities and consequent age approximations of geologic units of the Altiplano fill sequence. fill the of Altiplano geologic units of approximations Polarities age and consequent

. presumed presence of the penultimate normalmagnetozone from sections farther up valley (Tangani; Purapura of Thouveny Servant, 1989 ) 4 .

2 a Nomenclature after Dobrovolny ( Based on magnetostratigraphic correlation to the GMPT (LR04) and new 40Ar/39Ar age. Polarity sequence based on that measured below the tuff at Viscachani and Jacha Kkota, that measured above the tuff at Jacha Unit

Age Age estimates are based on the preferred correlation to the GMPT a b c Purapurani Gravel Purapurani Patapatani Drift Patapatani Chijini Tuff Drift Calvario Gravel Kaluyo Drift Sorata gravel surface Altiplano Formation Paz La upper Table

correlations within the La Paz basin suggested by previous researchers. Additionally, these polarities confirm longer distance correlations proposed by Dobrovolny (1962) between coarse-grained glacial sediments nearer the Cordillera Real and fine-grained sequences south of La Paz, both above and below the Chijini Tuff. The obviously different polarity records of the glacial sequences of similar thickness (~100 m) below (N-R-N-R-N) and above (N-R) the Chijini Tuff argues against the suggestion (Servant, 1977; Bles et al., 1977; Ballivián et al. 1978; Thouveny and Servant, 1989) that Patapatani and the Calvario drifts are coeval.

My new polarity sequence also greatly improves age constraints for the Altiplano sedimentary sequence previously provided by radiometric ages on the Chijini Tuff (Evernden et al., 1977; Servant, 1978; Clapperton, 1979; Lavenu et al., 1989; Marshall et al., 1992) and a magnetostraigraphy spanning only a portion of the sequence (Thouveny and Servant, 1989). Consequently, ages of many of the units can now be refined (Appendix E; Table 2.4), and in some cases tied to specific short-lived global climate events (see discussion below).

2.7. Evolution of the Andean landscape

The Altiplano sedimentary sequence exposed in the La Paz and Achocalla basins provides insight into the late Pliocene and Early Pleistocene evolution of the eastern Altiplano and Cordillera Real, which may be more generally applicable to larger parts of the Altiplano and Cordillera Oriental. Some similar interpretations have been suggested by previous authors, and the addition of detailed magnetostratigraphy not only confirms some of those interpretations, but provides new insights as well.

The Piacenzian stage (late Pliocene) is a key period in Earth’s climate, as it is characterized by initial warmth and subsequent climatic deterioration leading into the Pleistocene and its paleoenvironments thus provide opportunities to better understand the sensitivity and variability of Earth’s climate system (Lunt et al., 2010). The mid- Piacenzian warm period (3.265-3.025 Ma; Dowsett et al., 2010) is of special interest

because it is the most recent time in Earth’s history when climate was globally warm (~2- 3°C warmer than present; Haywood and Valdes, 2004; Ravelo et al., 2004; Lawrence et

al., 2006; Lunt et al., 2010), with atmospheric CO2 concentrations similar to those of today (Pagani et al., 2010; Bartoli et al., 2011). Consequently, this period is an important target for paleoclimate reconstruction (e.g. Dowsett and Cronin, 1990; Kleiven et al., 2002; Ravelo et al., 2004; Lawrence et al., 2006; Dwyer and Chandler, 2009; Dowsett et al., 2010; De Schepper et al., 2014) used to improve climate models (e.g. Haywood and Valdes, 2004; Haywood et al., 2007, 2011, 2013; Dolan et al., 2015). However, mid- Piacenzian terrestrial paleoclimates of the South American tropics remain poorly constrained (e.g. Salzmann et al., 2011).

2.7.1. Lateral facies and landscape variability

My magnetostratigraphic results allow for a correlation of sediments throughout the western part of the upper Río La Paz valley system (Fig. 2.14). I document spatial differences in facies in time-correlative units as a function of distance from the Cordilleran Real. I divide the fills into three parts on the basis of vertical facies variation: a lower glacial sequence (equivalent to units 1-9 and 11-15 at the Patapatani West section) separated by a gravel sequence (Patapatani West units 16-18) from an upper glacial sequence and the Altiplano surface gravels (Patapatani West units 19-20).

Lower glacial sequence

Below the extensive granitic gravel the sediment sequence near the Cordillera Real is dominated by till and coarse gravel, whereas farther south it consists mainly of interbedded silt and sand. The proximal and distal sediment sequences record the same polarity sequence (from bottom up N-R-N-R-N-R) and records glacial deposition and correlative distally fining proglacial environments (e.g. Eyles and Eyles, 2010), extending from near the start of the Gauss Chron (ca. 3.6 Ma) to at least the start of the Matuyama (ca. 2.6 Ma). It corresponds to Dobrovolny’s (1962) Patapatani Drift and the uppermost La Paz Formation below the Chijini Tuff and to his Calvario Drift, including its distal fine facies, above the tuff. In the Río Choqueyapu valley, distal outwash facies of the oldest

documented glacial events (lowest units in the Río Kaluyo valley; Fig. 2.14a) are underlain by lithologically identical, but reversely magnetized, fine-grained deposits (Fig. 2.14e). It is therefore possible that evidence of even earlier Cordilleran ice sheets is buried beneath valley-bottom deposits at the Patapatani West section.

Paleosols within the glacial sequence (Figs. 2.14 and 2.15 correlations D, E, G, and H) indicate sheet-like aggradation of the glacial/proglacial sequence, interrupted by long periods of landscape stability and weathering. The Chijini Tuff was emplaced during a nonglacial period probably similar to those evidenced by paleosols below and above it. Pulses of sedimentation separated by long periods of land surface stability and weathering suggest repeated glaciations, each accompanied by transfers of large volumes of sediment from the Cordillera Real, alternating with interglaciations, during which there was much reduced and more localized sediment delivery to the basin. This style of aggradation reflects the infilling of an endorheic subsiding basin.

Gravel sequence

The granitic (R-N-R) and non-granitic (bottom up R-N) gravel units are markedly different from the underlying, dominantly diamictic sequence. No tills occur within these thick gravels, even near the Cordillera Real (northeast of the Patapatani West section) and they are of far greater thickness and lateral continuity than the gravel units within the glacial sequence below. However, as in the sequence below, the gravel units fine distally into the Achocalla area. A portion of the reversely magnetized fine-grained sequence at the Jacha Kkota section measured by Thouveny and Servant (1989) probably correlates with the granitic gravel, and the polarity sequence of overlying fines likely correlates with the non-granitic gravel unit, but has not yet been measured.

The gravels occur as two thick, lithologically distinct units, in contrast with the localized thin gravels alternating with glacial diamictons and interglacial paleosols in the underlying and overlying sequences. Some previous workers (Bles et al., 1977; Clapperton, 1979; Thouveny and Servant; 1989) suggest these thick gravels record interglacial periods. However, polarity reversals within each gravel indicate they record long depositional periods (likely 100 ka or more), suggesting persistence on longer

timescales than the typical ca. 41-ka periodicity of Early Pleistocene interglacials. The long-lived periods of large-scale aggradation recorded by these gravels is in marked contrast to typical nonglacial intervals of the sequence below, which are represented by soil formation, indicative of long periods of landscape stability.

The gravels are probably outwash deposits associated with the expansion of glaciers in the Cordillera Real, having been transported sufficiently far enough to erode striations from clast surfaces. The correlative ice margin and associated tills are probably located in a part of the fill sequence, possibly northwest of La Paz, not yet exposed by incision. An analogous facies relationship appears to exist between the sub- tuff tills at the Patapatani West section and a thick sequence of gravels exposed below tuff in Río Chuquiaguillo, but their temporal relationship has not been confirmed paleomagnetically.

The major lithologic difference between the two gravel units suggests a long passage of time, sufficient for a major change in the source area, either in location or degree of incision. The erosional contact between the two gravels at the Patapatani West section is unique among the study sites. It is most easily explained by local topographic control (see Tectonics section below). These two gravel units likely record two periods of tectonic uplift, each beginning prior to one of the three subchrons of the Matuyama Chron (Fig. 2.16b).

Upper glacial sequence and Altiplano gravels

The upper part of the sedimentary sequence comprises one or more tills ranging from ~5 to over 80 m thick, capped by a thin (generally <5 m) gravel unit. Both the till and the capping gravel are composed almost entirely of argillite. What appears to be a single till is present at most sections, although it is possible that soils that separate tills lower in the sequence may have been removed by erosion. Notably, in the upper reaches of Quebrada Minasa (Fig. 2.14d), a paleosol separates two tills high in the sequence, and the capping gravel is absent. The glacier that deposited at least one of these tills was extensive enough to reach the present-day Achocalla basin. Where

present, the capping gravel overlies a well developed paleosol, suggesting that it was deposited well after the glaciation that deposited the underlying till.

The upper glacial sequence records a return to a time of extensive glaciation, similar to that recorded by the lower glacial sequence below the granitic gravel. The upper tills and gravel are normally magnetized. Their position within the polarity sequence indicates they were deposited during either the Olduvai or Jaramillo subchron (Fig. 2.16b).

2.7.2. Plio-Pleistocene glacial record

I have identified 17 separate glacial units (designated g1 to g17) in the Altiplano sedimentary sequence (Fig. 2.14a,c,d and Fig. 2.16a). Given the length of hiatuses necessary for polarity reversals and pedogenesis, these glacial events are probably separated by periods of subaerial exposure on the order of 103 to 104 years. The seemingly regular alternation between glacial and non-glacial states, together with the temporal magnitude of ice-free conditions and the number of distinct glacial deposits in each of the sub-tuff magnetozones, suggest that the glacial events are separate glaciations. The Altiplano fill sequence at La Paz thus provides a record of multiple glacial-interglacial cycles of the late Pliocene and Early Pleistocene.

Of the 17 glacial units defined by polarity reversals and unconformities, eight unambiguously correspond to seven specific marine isotope stages of the late Pliocene: MG2, M2, KM2, G10, G6-G4, and G2 (solid diamonds in Fig. 2.16c). The other three late Pliocene glacial units are each constrained to a small number of cold peaks (open diamonds in Fig. 2.16c). Pleistocene glacial units can, in most cases, only be constrained to multiple possible cold peaks (range bars in Fig. 2.16c; dark and light indicate preferred and alternative correlation to the GMPT, respectively), although in some cases the most likely cold peaks can be identified (open ovals in Fig. 2.16c; dark and light indicate preferred and alternative correlation to the GMPT, respectively). Given the near-absence of erosional evidence in the fill sequence, it is reasonable to assume that any ice advances as far south as the Patapatani West section have been recorded.

Chronology

Association of terrestrial glacial deposits with specific marine isotope stages (MIS), especially those prior to the Late Pleistocene (ca. 0.126 Ma; Gradstein et al., 2012), is rarely possible (Balco et al., 2005a). For the Altiplano fill, such correlations are assisted by the general lack of erosional hiatuses, the regular nature of glacial/nonglacial depositional patterns, and the limited number of, or dominance of only a few, cold peaks during short polarity periods (i.e. Mammoth subchron, mid Gauss, and Kaena subchron). Comparison of the number of glacial events in each magnetozone of the Altiplano fill sequence (Fig. 2.16a) to specific cold periods of the global, astronomically tuned, benthic oxygen isotope (δ18O) record (Lisiecki and Raymo, 2005) (Fig. 2.16c) suggests, and in some cases confirms, which globally cool periods are recorded by ancient glacial deposits in the tropical Andes, as well as which warm periods are recorded by paleosols. The strongest, longest cold peaks are most likely to generate far- reaching Andean glaciers. Long hiatuses in the sequence can be similarly linked to the warmest, longest interglaciations. Some caution is required in such correlations, however, because the marine oxygen isotope record, and particularly the LR04 record, are spatially averaged and thus could lead to underestimation of the local severity of cold or warm peaks due to out-of-phase ice expansion in different regions (cf. Dolan et al., 2011). Additionally, precipitation variability can result in ice extent fluctuation, particularly if glaciers are precipitation limited (Ward et al., 2007).

Earliest glaciation (g1 to g3)

The ice-proximal (unit g1) and ice-marginal (unit g2) deposits at the base of the Patapatani West section are the first conclusive evidence of an icecap in the Cordillera Real. Because there is no paleosol or polarity reversal at the contact between units g1 and g2 (Fig. 2.7), it is not known if there is a long hiatus between them. The units correspond either with MIS MG4 and MG2, respectively, or with MIS MG2 entirely; these are the only periods of the early Gauss when global temperatures were cooler than present (Fig. 2.16c). Lisiecki and Raymo (2005) suggest the pronounced MG2 peak in the oxygen isotope record of the eastern tropical south Pacific (Schakelton, 1995; ODP site 846), which is of similar magnitude to the well documented M2 cool peak (Lisiecki

and Raymo, 2005, their Figure 9), may be an artefact from coring or splicing error. Alternatively, the MIS MG2 glacial period might have been stronger in the tropical south than elsewhere, in agreement with an ice margin sufficiently far south of the Cordillera Real to deposit the striated, stratified diamicton (unit g2). A locally strong MIS MG2 cold peak could also account for the underlying striated gravel (unit g1).

The lowest massive till (unit g3) of the Patapatani Drift sequence (Patapatani West section) records the first ice advance to the northern limit of the city of La Paz. It was almost certainly emplaced during MIS M2 (ca. 3.3 Ma), which is by far the most severe cold peak of the Mammoth subchron (Fig. 2.16c) and a globally recognizable cooling event (De Shepper et al., 2009, 2014; Dolan et al., 2015).

Mid-Piacenzian warm period (g4 to g6) and end of the Pliocene (g7 to g11)

The fill sequence records three glaciations during the globally mild climate of the mid-Piacenzian warm period: one in the mid Gauss and two in the Kaena subchron. Marine isotope stage KM, which spans the polarity reversal at the base of the Kaena subchron, is the first strong colder interval after MIS M2 (Fig. 2.16c); it is particularly pronounced at ODP site 846 (Shackelton, 1995; Lisiecki and Raymo, 2005). It is the only logical period during the mid-Gauss when glaciers could expand to deposit till (unit g4) of the N2 magnetozone. Overlying reversely magnetized till (unit g5) may record the later part of the same glaciation. The weakly developed paleosol capping this till most likely formed during the relatively long warm period between MIS KM2 and G22, and the overlying till (unit g6) was deposited during the minor cold peak of MIS G22. In this scenario, the paleosol at the boundary between magnetozones R3 and N3 formed during the G21 warm interval, which lasted ca. 15,000 years.

Normally magnetized tills underlying and overlying the Chijini Tuff record five latest Pliocene glaciations between the close of the mid-Piacenzian warm period (3.025 Ma) and the Gauss-Matuyama boundary (2.608 Ma). The glacial units below the tuff represent three separate glaciations during the late Gauss, prior to ca. 2.74 Ma. The paleosols between them are much better developed than the single one below (i.e. at the unit g5-g6 contact), suggesting they formed during longer periods (>15,000 years) or

under warmer/wetter climatic conditions. The first of these tills (unit g7) most likely records MIS G20 or G18, because one of the two overlying paleosols most likely represents MIS G17, the most pronounced late Gauss interglaciation; MIS G20 is more likely as the marine oxygen-isotope record from the equatorial east Pacific (Shackleton, 1995; ODP site 846) indicates that the G18 excursion is relatively minor. The next glacial unit (g8) can be constrained only to one of the four cold peaks prior to the MIS G11 interglaciation. The last pre-eruption glacial unit (g9) was probably deposited during MIS G10, the most severe late Gauss cold period before 2.74 Ma.

The two glacial units overlying the tuff match cool peaks of the post-2.74 Ma portion of the Gauss, based on the LR04 polarity timescale, in number and approximate duration. The thick till (unit g10 – >45 m) directly overlying the tuff likely records glaciation during the ca. 75-ka-long MIS G6-G4 cold period. The capping paleosol and overlying <10-m-thick till (unit g11) likely represent MIS G3 and MIS G2, respectively. The cold peak at the Gauss-Matuyama boundary (MIS 104) might fall in a period of transitional polarity (cf. Harland et al., 1982), although the revised polarity timescale of Ogg (2012) places the Gauss-Matuyama polarity boundary at 2.581 Ma, leaving MIS 104 completely within the Gauss. It is thus possible that the second glacial unit above the tuff (g11) records MIS 104, not MIS G2.

Early Pleistocene (g12 to g17)

The Early Pleistocene glacial record is more difficult to constrain due to the length of reversed polarity intervals of the Matuyama and the large number of cold peaks within them (Fig. 2.16c). Additionally, correlation of the polarity sequence above the tuff to the GMPT is less certain. Assuming the preferred GMPT correlation in which magnetozones N4 and N5 record the Réunion and Olduvai subcrons, the three glacial units of magnetozone R4 (g12, g13, and g14) would be associated with glaciations of the earliest Gauss Chron (2.608-2.148 Ma). These were the strongest cold peaks (MIS 100, 98, and 96) and occurred at the start of the Matuyama. However, the three units might have been deposited during any of the early Matuyama cold peaks prior to the MIS 83 interglacial just below the start of the Réunion subchron. Till deposited during MIS 82 would not have sufficient time for paleosol formation at the top of the Calvario

Drift sequence (unit g14) and subsequent deposition of the overlying lowermost granitic gravels (end of magnetozone R4) before the start of the Réunion subchron (magnetozone N4). The thick glacial units of magnetozone N5 (g15 and g16) most likely record MIS 74 and 70, the strongest cold intervals during the Olduvai subchron. However, given the complexity of the drift sequence below the Altiplano surface at Río Minasa (Fig. 2.10), it is possible that some of the lesser cold peaks during the Olduvai (particularly MIS 72) also generated ice caps that reached to at least the Minasa section. At all but the Minasa section, the terminal cold peak of the Olduvai subchron (MIS 64) is not recorded in the glacial sequence, as the glacial unit (either g 15 or g16) is capped by a well developed paleosol (Fig. 2.14), indicating a long period of ice-free conditions prior to the deposition of the surface gravel on the Altiplano surface before the end of this polarity interval. The youngest glacial unit (g17), present only at the Minasa section, could record ice advance as late as MIS 64.

If, alternatively, the less favoured paleomagnetic scenario is correct, the uppermost magnetozone (N5) corresponds to the Jaramillo subchron and the three glaciations recorded in magnetozone R4 (units g12, g13, and g14) might correspond to any of the cold peaks of the Matuyama before the base of the Jaramillo (1.075 Ma), excluding the Réunion and Olduvai subchrons. With this alternative GMPT correlation, the penultimate till unit of the sequence (g15) would date to MIS 30, which is the first, as well as the larger and longer, of two cold intervals during the Jaramillo. However, the presence of two paleosols formed within the ~80-m-thick N5 till sequence (Sorata Drift) at the Minasa section (Fig. 2.10) differentiates three till units (g15, g16 and g17) and suggests that the Sorata Drift represents at least three glaciations. Theoretically, it is possible that the youngest glacial unit (g17) records the final cold interval of the Jaramillo subchron (the first half of the split peak of MIS 28; Fig. 2.16c). However, I consider this interpretation unlikely as it requires short-lived, relatively mild MIS 28 to have produced a far-reaching Andean ice cap when many other more substantial cool intervals between the end of the Réunion subchron and start of the Jaramillo apparently did not. Additionally, the short period between this cold interval and the end of the Jaramillo (ca. 10 ka) is probably insufficient to produce the paleosol at the top of the Sorata Drift and then later deposit the Altiplano surface gravel. Even if the youngest glacial unit (g17) records MIS28, it is difficult to explain occurrence of two separate

glaciations (g15 and g16) prior to it during the Jaramillo subchron. Thus the record of multiple glaciations in magnetozone N5 of the sedimentary sequence adds further credence to the preferred correlation – the Altiplano surface in the La Paz area dates to the Olduvai subchron, not the Jaramillo.

Records of more recent glaciations in the Cordillera Real are derived from Late Pleistocene (MIS 2 glaciation) end moraines in the Río La Paz valley system (Dobrovolny, 1962) and on the plateau surface northwest of La Paz (Smith et al. 2005), and inferred from magnetic susceptibility of a sediment core from Lake Titicaca spanning the period from ca. 0.37 Ma to present (MIS 10-2 glaciations) (Fritz et al., 2007). Thus, a substantial gap remains in the glacial history of the Cordillera Real, of at least 0.62 Ma, if the Altiplano surface formed in the Jaramillo subchron, but likely more than 1.4 Ma, assuming the Altiplano fill sequence terminates in the Olduvai subchron.

Subsequent Pleistocene glaciations

The Cordillera Real experienced ice cap growth during Middle and Late Pleistocene glaciations that are younger than the events described here. Moraine complexes extending from the Cordillera Real onto the Altiplano margin northwest of La Paz record glacier expansion during the Late Pleistocene, including MIS2 and the Younger Dryas/Antarctic Cold Reversal (Smith et al., 2005, 2008; Zech, 2008). More extensive moraine complexes may date to older glaciations. Sediment cores from Lake Titicaca provide evidence of four glacial periods (likely MIS 10, 8, 6, and 2-4) (Fritz et al., 2007, their Figure 5).

Glaciations younger than the formation of the Altiplano surface at La Paz are absent from the stratigraphy that I report due to topographic inversion. At La Paz, these younger advances were confined to the deepening, upper Río La Paz valley system. However, the stratigraphic record has been almost completely destroyed due to episodic incision of the valley system during interglacial periods. Preferential preservation of older glacial sequences along the eastern Andean front in Patagonia (Mercer, 1976, 1983) similarly results from incision-driven topographic inversion (Lagabrielle et al., 2010).

Younger glaciations in the La Paz basin are recorded only by undated deposits, likely representing the most recent advances. An end moraine in the Río Chuquiaguillo valley bottom (~3950 m asl) about 3 km up-valley from Río Minasa (Dobrovolny, 1962) is the clearest evidence. In the Río Kaluyo valley a short distance down-valley of the Patapatani West and Patapatani East sectionsDobrovolny (1962, pp. 49-50) interprets two valley-traverse ridges, each associated with a chain of small lakes, as recording two distinct glacial advances. However, the lower ridge and lakes at about 4100 m asl are related to the Limanpata landslide, sourced from the valley’s eastern wall (eastern) valley wall (Anzoleaga et al., 1977; Vargas, 1977). The upper ridge and lakes at ~4250 m asl could be the result of block subsidence (Vargas, 1977), approximately where the Chuquiaguillo graben (Appendix F) intersects the west valley slope. Terraces of Miraflores gravel along Río Orkojauira/Chuquiaguillo, downstream of the end moraine mapped by Dobrovolny (1962) and along Río Choqueyapu downstream of the Limanpata landslide, are likely of glaciofluvial origin (Dobrovolny, 1962; Anzoleaga et al., 1977). The dominantly gravelly sediments that underlies these surfaces could be composite deposits recording two or more glaciations. The till reported by Heim (1951) south of the city centre (~3500 m asl) may also record an advance.

Extent of glaciation

Ice-margin positions indicated by the transition from diamicton facies to gravel facies generally reach farther from the Cordillera Real over time. The first three glacial units (g1-g3) record proglacial, ice-proximal, and full glacial depositional environments, indicating that glaciers approached and eventually reached the most northerly stratigraphic section by MIS M2 (ca. 3.3 Ma). How far south of the site of this section that ice reached during the mid-Piacenzian warm period (3.265-3.025 Ma; units g4-g6) and early late Gauss (until ca. 2.85 Ma; units g7 and g8) is unclear because these units are exposed at only a single locality. The till underlying the tuff (unit g9) reaches ~5 km south of the Patapatani West section (Clapperton, 1979), but not ~1.5 km beyond that point to the Viscachani section (Ballivián et al., 1978; Thouveny and Servant, 1989). Tills directly overlying the tuff are present at least as far south as the Purapura section, and higher tills, underlying the granitic gravels, reach to almost the northern limit of the Achocalla basin (Dobrovolny, 1962). Tills above the thick gravel sequence (units g15-

g17) reach well into the Achocalla basin, but it is unclear whether they represent a subglacial or ice-marginal position.

Comparison with glaciations in other regions

Terrestrial records

The fragmentary nature of Pliocene and Early Pleistocene terrestrial glacial deposits (De Schepper et al., 2014) and the limited geochronological constraints on their ages (Balco et al., 2005a) allow only approximate comparison of glaciations of the glacial record at La Paz with those in other regions (Fig. 2.16e). Ice caps formed in the Patagonian Andes in the latest Miocene (>4.6 Ma) (Mercer, 1983), well over 1 Ma before the first evidence of tropical Andean glaciation. Subsequent evidence for Patagonian glaciations includes a till tightly constrained to the latest Gilbert Chron (3.71-3.588 Ma) based on magnetic polarity and K/Ar geochronology (Mercer, 1976), a till overlying a 3.46-Ma basalt flow (Mercer, 1976), and outwash gravel overlain by a late Gauss (2.73 ± 0.06 Ma) basalt flow (Mercer, 1983). The latter two may fall within the pre-Chijini Tuff part of the La Paz sequence (i.e. between units g1 and g9). Singer et al. (2014) constrain the ages of six younger Patagonian tills using paleomagnetism and high- resolution 40Ar/39Ar chronologies of basalts with which they are interlayered. The first till of the sequence is older than the Réunion subchron and may correspond to one or more of the reversely magnetized glacial deposits above the Chijini Tuff (units g12-14). Subsequent Patagonian tills were deposited during the Réunion subchron, the Olduvai subchron, between these subchrons (two events), and between the Olduvai and Jaramillo subchrons. Of these five events, only the one during the Olduvai is possibly represented in the La Paz sedimentary sequence (unit g15, g16 or g17), assuming that the preferred correlation of the Altiplano sequence to the GMPT is correct. If the alternative correlation is correct, one of the final glaciations of the La Paz sequence (unit g15, g16 or g17) probably coincides with the Great Patagonian Glaciation, which occurred within the Jaramillo subchron (Griffing, 2012) before 1.016 ± 0.005 Ma (Ton- That et al., 1999) and likely during MIS 34.

The earliest, reliably dated terrestrial evidence of an ice sheet in North America is end-Pliocene till (Barendregt et al., 2010) and outwash (Froese et al., 2000; Barendregt et al., 2010; Hidy et al., 2013) representing the most extensive glaciation in central Yukon Territory, Canada. These deposits are similar in age to glacial units directly above and below the Chijini Tuff (g9-g11). Subsequent montane and ice sheet glaciations in Yukon Territory, documented in each of the main (>50-ka-duration) polarity intervals of the Early Pleistocene (Barendregt et al., 2010) may coincide with Matuyama-age glacial units near La Paz (g12-g17). The Laurentide Ice Sheet first reached to near its southern limit by 2.41 ± 0.14 Ma, possibly coincident with deposition of unit g12, g13, or g14 near La Paz, but not again until sometime between the end of the Olduvai subchron and ca. 1.6 Ma (Balco et al., 2005a). Gao et al. (2012) identified what they interpret to be till in James Bay Lowland, Canada (~53°N) and assign it to a pre-Laurentide, Pliocene (3.6- 3.4 Ma) glaciation. The age is based on the magnetic polarity sequence (bottom up: N- R-N-R) of overlying lacustrine sediments. If this interpretation is correct, the till correlates with or is older than the earliest glaciations recorded in the La Paz area. However, the fossil pollen used to correlate the polarity sequence to the GMPT may be reworked and the reversely magnetized intervals could, alternatively, record portions of the Matuyama Chron. Additional research is required before the conclusions of Gao et al. (2012) can be accepted.

No direct evidence for Pliocene or earliest Pleistocene glaciation has yet been found in Europe, Scandinavia, or Greenland (De Schepper et al., 2014). Early Pliocene (ca. 4 Ma) terrestrial glacial deposits of at least three localized glaciations have been found in Iceland and dated to the early Gauss, the Kaena subchron (possibly correlative with units g1/g2 and g5 or g6, respectively), and the late Gauss after ca. 2.9 Ma (possibly correlative to unit[s] g8, g9, g10 or g11) (Geirsdóttir, 2011). Large ice caps formed in Iceland five to 13 times during the Matuyama Chron (Geirsdóttir, 2011), a period when six glacial events (units g12-g17) are recorded at La Paz. The earliest known evidence of Cenozoic glaciation in the tropics outside the Cordillera Real is glacial diamictons on Mount Kenya from the Olduvai subchron (possibly coincident with unit g15, g16 or g17) and the preceding part of the Matuyama Chron (Mahaney et al., 2013), and glaciofluvial sediments from the earliest Matuyama (ca. 2.6 Ma) in the Colombian Andes (Helmens et al., 1997) (possibly coincident with unit g12, g13, or g14).

Marine records

The onset and intensification of high-latitude glaciation recorded in continuous marine sediments are similar in timing to the glacial record of the Cordillera Real (Fig. 2.16e). Glaciomarine sediments on the continental shelf of southeast Alaska and ice- rafted detritus in the abyssal North Pacific dating to 3.5-3.0 Ma (Lagoe et al., 1993) record expansion of glaciers in the St. Elias Mountains from the earliest Gauss Chron to the end of the Kaena subchron (units g1-g6 and possibly g7). Glacial expansion in ranges adjacent to the Bering Sea is signalled by possible ice-rafted debris prior to 3.8 Ma, by first occurrence of sea-ice dinoflagellates at ~3.4 Ma, shortly before or coincident with the earliest glaciation recorded at La Paz (unit g1), and later by the occurrence of sea-ice diatoms and common dropstones at ~2.7 Ma, likely coincident with glaciation in the Cordillera Real at the end of the Gauss Chron (units g7-g9) (Takahashi et al., 2011). Records of ice-rafted debris indicate incremental late Pliocene expansion of the East Antarctic ice sheet (MIS M2) (McKay et al., 2012), Greenland ice sheet (MIS G22) (Kleiven et al., 2002), and other parts of the circum-North Atlantic (MIS G6) (Kleiven et al., 2002), corresponding to, respectively, glacial units g3, g6, and g10.

Additionally, La Paz glacial units appear to coincide with specific glaciations recoded by peaks in ice-rafted detritus (Fig. 2.16e). Greenland (Kleiven et al., 2002) and East Antarctica (Passchier, 2011) experienced slight increases in ice calving during MIS MG2, a time when ice advanced to just north of La Paz (unit g2 and possibly g1), followed by substantial increases during M2 when till of unit g3 was deposited. Both high-latitude ice sheets again expanded at MIS KM2 and G22 (Kleiven et al., 2002; Passchier, 2011) during the mid-Piacenzian warm period, coincident with deposition of units g4/g5 and g6, respectively. Pronounced increases in ice rafting in the North Atlantic suggest a particularly large expansion of the Greenland ice sheet during MIS G6, G4, 104, 100, 98, and 96 (Kleiven et al., 2002), possibly corresponding to deposition of units g10, g11, g12, g13, and g14.

Drivers of glaciation in the Cordillera Real

The causes of late Pliocene and Early Pleistocene glaciation in the Central Andes remain unclear. In light of the region’s climate sensitivity (Markgraf et al., 2010), there may be several interacting drivers. Expanding ice extent through the late Pliocene and Early Pleistocene suggests decreasing temperature, increasing precipitation, or a combination of the two. This expansion appears to parallel global climatic deterioration during the latest Pliocene and Early Pleistocene (Fig. 2.16c). However, long-term global climatic trends cannot alone explain the pattern of cryosphere variability in the tropical Bolivian Andes, because glaciations during the warm mid-Piacenzian were more extensive than during the globally cooler Last Glacial Maximum.

Elevation of the Central Andes during the Miocene and early Pliocene (Gregory- Wodzick, 2000; Barnes and Ehlers, 2009) likely had a strong influence on glacier extent. The central Andes were probably higher in the Pliocene than today (section 2.7.3). Subsequent denudation of the Cordillera Real, evidenced by the huge fan of thick Plio- Pleistocene flanking the mountain range at and northwest of La Paz, suggests an overall reduction of mean elevation of the range and thus the source area of glaciers since the late Pliocene. Regardless of the nature of adiabatic temperature change in cordilleran ice-source areas since the start of the late Pliocene, continued uplift of the Altiplano probably continued the previous non-adiabatic cooling trend preceding the late Pliocene (Ehlers and Poulsen, 2009).

Altiplano uplift also increased austral summer precipitation leading up to the onset of late Pliocene glaciation (Ehlers and Poulsen, 2009). The net effect on mass balance would have depended on regional temperature and ice cap hypsometry; melt and accumulation would be enhanced, respectively, below and above the snowline. The seasonality of precipitation has a particularly strong influence on tropical glaciers (Kaser and Osmaston, 2002) including the modern Zongo glacier on Huayna Potosí (Sicart et al., 2003). However, precipitation changes in the eastern Central Andes during glacial- interglacial cycles are unclear, even for the late Quaternary. In the Cordillera Oriental at Cochabamba, ~230 km southeast of La Paz, precipitation during the maximum Late Pleistocene glaciation was possibly similar to the modern regime (Kull et al., 2008). In

contrast, some proxy records suggest variable humidity since the Late (Bräuning, 2009) and Middle Pleistocene (Fritz et al., 2007), with ice advance generally coinciding with periods of greater humidity (Fritz et al., 2007).

Due to their high elevation, the Central Andes disrupt atmospheric circulation (Gregory-Wodzick, 2000). Late Miocene uplift changed the prevailing wind direction, and thus water vapor source, from the south Pacific to the equatorial Atlantic (Ehlers and Poulsen, 2009). The Cordillera Occidental thus posed an important barrier to moisture reaching the Cordillera Oriental in the latest Miocene. However, owing to the volcanic growth and destruction of its peaks, paleo-physiography of the Cordillera Occidental is not well constrained. The modern spatial patterns of precipitation in the Central Andes were established by the start of the late Pliocene (Bershaw et al., 2010), which seems to generally coincide with the start and intensification of glaciations in the Cordillera Real. However, late Pliocene circulation in the region is so far only broadly understood (Dowsett et al., 2010; Haywood et al., 2011).

Geographic and atmospheric mechanisms already postulated to explain intensification of Northern Hemisphere glaciation might also play a role in the late Pliocene initiation and subsequent expansion of ice sheets in the Cordillera Real. Changes in oceanic circulation ca. 4-3 Ma due to closure of the Central American (Bartoli et al., 2005; Lunt et al., 2008) and Indonesian Sea Ways (Cane and Molnar, 2001) altered southern hemisphere heat and moisture availability, possibly favouring growth of ice sheets in the tropical eastern Andes. Although the Antarctica Circumpolar Current played an important role in the thermal isolation and onset of early Cenozoic glaciation in Antarctica (Mercer, 1983 and references therein; Zachos et al., 2001), its establishment greatly predates the onset of glacial condition in the Central Andes. A complete circum-Antarctic seaway had already formed by the late Oligocene upon the opening of Drake Passage (Lawver and Gahagan, 2003; Livermore et al., 2005; Pfuhl and McCave, 2005).

Bartoli et al. (2011) report an overall trend of reduction in global atmospheric CO2 concentrations during the progressive glacial expansion recorded at La Paz. Gradual

reduction of atmospheric CO2 concentrations between the end of the early Pliocene (ca. 3.8 Ma) and the first glacial evidence at La Paz (3.4 Ma) (Bartoli et al., 2011, their Figure 3) may have contributed to formation of the first ice caps in the Cordillera Real. However, the expansion of cordilleran ice nearly to La Paz during the mid-Piacenzian warm period (Fig. 2.16), a time of elevated atmospheric CO2 (Bartoli et al., 2011, their Figure 3), suggests that other factors played important roles.

Paleo-climatic and paleo-glacier modeling might provide additional insight into the causes of late Neogene and Quaternary ice-cap variability in the Cordillera Real. Such simulations should integrate the ice extents at La Paz that I have documented for the late Pliocene and Early Pleistocene and that other studies (Dobrovolny, 1962; Smith et al., 2005) have documented for the Late Pleistocene. However, precipitation reduction during the Late Pleistocene is the simplest explanation of reduced ice cap extent relative to the late Pliocene and Early Pleistocene. Seasonal (Zhou and Lau, 1998; Vuille, 1999) and inter-annual (Garreaud et al., 2003; Vuille et al., 2003) precipitation variation in the region demonstrate the region’s sensitivity to changes in prevailing wind direction that could drive moisture changes over multiple glacial cycles.

2.7.3. Tectonic evolution

Paleo-elevation

Uncertainty in the height of the Central Andes during the late Pliocene stems from differing interpretations of the timing of Cenozoic uplift (Gregory-Wodzick, 2000). Constraining their paleo-elevation is important for modeling Pliocene climate (Dowsett et al., 2010; Haywood et al., 2011), as the Andes influence Southern Hemisphere atmospheric circulation (Gregory-Wodzick, 2000). Disagreement centres on whether uplift progressed steadily and slowly since ca. 40 Ma or accelerated (~2.5 km) ca. 10-6 Ma (Barnes and Ehlers, 2009). Isotopic paleo-elevation proxies from carbonate deposits (Garzione et al. 2006; Ghosh et al., 2006) and mammal teeth (Bershaw et al., 2010) suggest that the Central Andes achieved most of their present altitude by, respectively, ca. 6.8 Ma and ca. 3.6 Ma. However, some portion of the late Cenozoic oxygen isotope

depletion in these proxies may result from tectonically induced changes in paleoclimate (Ehlers and Poulsen, 2009) or moisture source (Bershaw et al., 2010).

Glacier expansion north of La Paz suggests that by the late Pliocene the Cordillera Real was at least as high as today. The average elevation of the Cordillera Real was perhaps even greater than today, given that glacial limits during relatively weak, short cold peaks of the mid-Piacenzian and early late Gauss (before ca. 2.85 Ma) had similar positions near La Paz as Late Pleistocene end moraines (Dobrovolny, 1962; Smith et al., 2005). Additionally, lithologic trends in the Altiplano sedimentary sequence suggest substantial erosion of the Cordillera Real through the late Pliocene and Early Pleistocene, involving unroofing and incision of one or more granitic plutons followed by erosion of underlying host rocks. However, this possibility does not consider late Pliocene differences in moisture availability (Haywood and Valdes, 2004; Ravelo et al., 2004) and temperature on glacier mass balance in the Cordillera Real or to what degree rock uplift counteracted exhumation (England and Molnar, 1990).

Faulting

High-angle faults cut the Plio-Pleistocene fill sequence north and south of La Paz, providing insight into more recent tectonic activity. Progressively increasing offsets with age indicate that faulting in upper Río Minasa and locally enhanced sedimentation in the Chuquiaguillo graben (Appendix F) continued throughout aggradation of the Altiplano sediment sequence. Because no fault traces are visible on Huari Pampa, activity of both features ceased shortly before formation of the Altiplano surface, either at the end of the Jaramillo subchron (0.99 Ma) or, more likely, the end of the Olduvai subchron (1.781 Ma).

Incision-driven flexural uplift

Zeilinger and Schlunegger (2007) suggest a positive feedback between localized incision of the Altiplano by Río La Paz and surface uplift of the Altiplano plateau to the west and the Cordillera Real to the east. They cite the slope of Altiplano margin west of Río La Paz away from the valley system as possible evidence of incision-driven flexural

uplift. Such uplift could affect parts of the Río La Paz area, particularly southeast of La Paz where the deepest incision has removed the largest amount of sediment and rock. However, the stratigraphy and surface morphology of the Altiplano margin near the La Paz and Achocalla basins suggest that the southwest-sloping surface there is due largely to long-term aggradation and pre-dates the onset of incision. Rapid distal fining and thinning of lithostratigraphic units suggests topographic gradients away from the Cordillera Real throughout the late Pliocene and Early Pleistocene. Erosional remnants of the relict Altiplano surface, such as Huari Pampa, are roughly concordant with the sloping Altiplano surface immediately west of La Paz. Similar gradients continue northwest of La Paz over a ~70-km-wide swath of glacial and fluvial surfaces deposited along the Altiplano margin since the late or possibly Middle Pleistocene that have since experienced little incision.

2.7.4. Basin aggradation

The chronostratigraphy and, therefore, aggradation rates (Appendix G, Table 2.5) are best constrained below the Chijini Tuff. Sub-tuff, average aggradation rates are 12.5 cm/ka at the section nearest the Cordillera and 11.8 cm/ka at the section farthest from it. These are the only sections with long polarity sequences measured below and above the tuff, but they provide an opportunity to quantitatively compare the preferred and alternative GMPT-correlation options discussed earlier (Table 2.5). Assuming the preferred GMPT correlation, long-term average aggradation rates above the tuff are similar (13.5 and 13.6 cm/ka, respectively) to those below it, suggesting the Altiplano surface most likely dates to the Olduvai subchron. In contrast, a Jaramillo age for the Altiplano surface (i.e. the alternative GMPT correlation) requires a reduction in the average aggradation rate above the tuff of nearly half (7.3 cm/ka at both sections) the average pre-tuff rate.

Long-term aggradation rates differ between sections, reflecting the influence of the Chuquiaguillo graben. The sections farthest north and south are thinner (Fig 2.14) and provide nearly identical average aggradation rates (Table 2.5). Assuming an Olduvai age for the Altiplano surface, average aggradation rates both below and above the tuff

are within the range of rates measured paleomagnetically for aggrading Pliocene and Early Pleistocene basins elsewhere in the Central Andes (9.3-17.0 cm/ka; Marshall et al., 1982; MacFadden et al., 1983, 1993) (Table 2.6). However, average aggradation rates are much higher at the approximate middle of the graben. They are about four times higher at the Minasa section (44 cm/ka under the preferred GPMT correlation) than at the far north and south sections, approaching the mean rate (90,0 cm/ka, Table A8) in the Miocene Corque basin of the central Bolivian Altiplano that Roperch et al. (1999) attribute to basin subsidence.

Table 2.5. Calculated fill-sequence aggradation and incision rates from sections at La Paz.

PREFERRED INTERPRETATION ALTERNATIVE INTERPRETATION

a Section Aggradation rate (cm/ka)b Incision rate Aggradation rate (cm/ka)b Incision rate c c Sub-tuff Supra-tuff Total (cm/ka) Sub-tuff Supra-tuff Total (cm/ka)

PTW* 12.50 13.51 13.65 15.83 12.50 7.30 9.28 28.50 PTE 8.61 10.61 10.16 ― 8.61 10.61 10.16 ― TNG ― 31.65 31.65 ― ― 31.65 31.65 ― MIN* ― 42.55 43.62 31.94 ― 22.99 23.56 57.50 VIS 17.19 ― 18.23 ― 17.19 ― 18.23 ― PUR* ― 25.74 26.38 20.00 ― 13.91 14.25 36.00 JKT* 11.81 13.56 13.46 41.67 11.81 7.33 8.33 75.00

See Appendix G and Table A7 for details. a Section abbreviations: PTW, Patapatani West; PTE, Patapatani East; TNG, Tangani; MIN, Minsasa; VIS, Viscachani (from Thouveny and Servant, 1989); PUR, Purapura; JKT, Jacha Kkota (including some data from Thouveny and Servant, 1989). b Underlined aggradation rates used for comparison of lengthy pre-tuff and post-tuff sequence at a given section to evaluate the two possible correlations to the GMPT. c Includes thickness of the rapidly emplaced Chijini Tuff and, therefore, yields a rate slightly higher than the average of the sub-tuff and supra-tuff rates. * Section reaching Altiplano surface.

2.7.5. Breaching of the Cordillera Real and Incision of the Altiplano

Breaching of the eastern Altiplano at La Paz by the headwaters of the Amazon River happened after the youngest sediments in the sequence documented in this study had been deposited. Age constraints on sections extending to the plateau surface (Fig. 2.14) and the local depth of the trunk stream below that surface provide minimum

incision rates. Actual incision rates may have been higher, given the unknown time that elapsed between the formation of the modern Altiplano surface and the onset of incision.

The apparent northerly decrease of minimum incision rates (Tables 2.5) likely reflects the progressively later initiation of incision as Rio La Paz extended headward. The rate at the southernmost section thus provides the most realistic minimum limit on long-term incision – ~40 cm/ka based on an Olduvai age for the Altiplano surface and ~75 cm/ka based on a Jaramillo age. Incision beginning in the late Olduvai subchron or soon after best agrees with millennial-scale erosion rates of 23.0 cm/ka for the Río La Paz basin (range = 10.0-60.0 cm/ka for individual sub-basins) determined by Zeilinger et al. (2009) using TCN. Denudation dominated by channel network incision in the Cordillera Real determined using TCN is similar (20.0-60.0 cm/ka: Safran et al., 2005), but is two orders of magnitude lower on the Altiplano surface immediately to the west (0.03-2.90 cm/ka; Hippe et al., 2012) where diffusive erosion dominates denudation.

The inferred greater incision rate at the Minasa section (Table 2.5) relative to rates at adjacent sections in the Río Kaluyo/Choqueyapu valley may reflect drainage- specific differences in erodibility. Faulting in the area (Lavenu, 1997; Lavenu et al., 2000) associated with the Chuquiaguillo graben may predispose the local sediments to greater erosion. Alignment of the main branches of the Río Minasa quebrada system along fault traces (Appendix F) attests to such preferential erosion. The dominance of gravel in the sediment sequence in this area (Fig. 2.14), rather than more resistant till, may also be important.

Breaching of the Cordillera Real south of the Illimani massif is broadly constrained by the age of the Altiplano surface, which is most likely Olduvai. Final penetration of the Altiplano west of the Cordillera Real occurred around or after ca. 1.8 Ma. Since then, Río La Paz has extended its watershed ~80 km through the Altiplano sediment sequence and underlying basement rock at a minimum rate of ~45 m/ka. Comparable, albeit smaller-scale, drainage expansion in the arid Sorbas basin in southeast Spain began ca. 100 ka and led to ~20 km of upstream incision of the basin fill (Mather et al., 2002). This much greater average rate of watershed expansion (200

age range 1993: pg.

Source " Marshall et al. (1982) al. et Marshall MacFadden et al. (1983) Roperch et al. (1999) MacFadden et al. (1993) Geologic age Gelasian late Calabrian late Zanclean-Piacenzian c Age boundaries Values stated in source Start of Matuyama to end of Olduvai of end to Matuyama of Start Start of Jaramillo to start of Brunhes of start to Jaramillo of Start End of Cochiti to start of Mammoth b

9.2 18.8 16.9 16.9 90.0 Aggradation rate Aggradation (cm/ka) 993: pg. 238 ). 640 827 295 865 5000

Range 780 1781 9100 3319 Top Age (ka) ca. 3300 MacFadden et al., 1 2608 1075 4184 Base 14100 ca. 4000 astronomically tuned version of the GMPT and using a different a pproach to estimate duration of time a 50 80 140 120 4500

(m) Thickness Calculated aggradation rates of other late Cenozoic continental fill sequences in the Central Andes. Central the in fill sequences continental late Cenozoic other Calculated of rates aggradation

. 229 238& ). represented by stratigraphic sequence (two magnetozones) ( 6 .

2 " Thickness provided sourceby covers a longer stratigraphic sequence (four magnetozones) than polarity sequence used to define Rate is likely overestimated since stratigraphic thickness spans moremagnetozones than age range stated by MacFadden et ( al. Based on a previous, non - Sequence

a b c Uquia Fm., Argentina Uquia Fm., Tarija basin, Bolivia basin, Tarija Corque, Bolivia Corque, Inchasi section, Bolivia section, Inchasi Table 86

m/ka), in spite of a much lower drop in base level (~500 m), may reflect differences in erodibility and climate, but could also suggest penetration of the Cordillera Real lagged substantially behind formation of the Altiplano surface.

2.7.6. Slope instability in the upper Río La Paz watershed

Several aspects of the geologic history of the La Paz area account for its considerable instability and are responsible for the high frequency of historic slope failures (Dobrovolny, 1962; O’Hare and Rivas, 2005; Chapter 3) and large prehistoric landslides (Dobrovolny, 1962, 1968; Anzoleaga et al., 1977; Malatrait et al., 1977; Vargas, 1977; Hermanns et al., 2012; Chapter 4). Basin fills of the eastern Altiplano were not deeply buried (<800 m) and thus are weakly to non-lithified. Breaching of the Cordillera Real by the headwaters of the Amazon River in the Pleistocene produced a near-instantaneous, base-level drop of ~4-km (i.e. from Salar de Uyuni in Bolivia to the Atlantic Ocean) on the Altiplano. Consequent deep, rapid incision of the weakly lithified sediments beneath the Altiplano, averaging at least 40 cm/ka, created slopes throughout the La Paz and Achocalla basins that were in a state of disequilibrium. Incision and attendant mass movements will continue as the drainage systems continues to adjust to its new, lower base level.

2.7.7. Paleontological implications

Lithostratigraphic and chronostratigraphic constraints of the sediment sequence at La Paz have implications for mammalian evolution and migration. Pliocene and Pleistocene fossils of the New World southern tropics are concentrated in the Andean region (Marshall et al., 1984; Hoffstetter, 1986; Marshall and Sempere, 1991; MacFadden, 2005), where they are important for studies of mammal biodiversity (MacFadden, 2005) and adaptation to high-altitude environments (Hoffstetter, 1986). Demonstration of alternation of glacial and interglacial climates in the Central Andes throughout the mid-Piacenzian warm period, the terminal Pliocene, and the Early Pleistocene constrains paleo-environmental conditions for these faunas, at least in the

vicinity of the Altiplano and high Cordillera Real. The magnetic polarity chronology refines existing chronologies of the fossil-bearing sequences that have previously been based largely on dated volcanic marker beds (Evernden et al., 1977; Lavenu et al., 1989; Marshall et al., 1982). It also fills gaps between chronostratigraphies of the late Miocene (MacFadden et al., 1990) Pliocene (MacFadden et al., 1993) and those of the Pleistocene (Marshall et al., 1982; Walther et al., 1996) (Fig. 2.16f).

Great American biotic interchange

Faunal evolution in South America occurred in isolation during the Paleogene, resulting in unique land mammal groups (Simpson, 1980). Periodic exchange of mammals with North America began in the Miocene owing to gradual closure of the Central American Seaway. This exchange culminated in the Great American Biotic Interchange (GABI) (Marshall et al., 1979) ca. 2.6 Ma (Cione et al., 2007, 2015; Woodburne, 2010). The chronologic framework provided here shows that the sediment sequence at exposed at La Paz spans the first two pulses of inter-American faunal migration as well as the period of apparently limited biotic connection that led to it (Fig. 2.16f).

Mammal-bearing sediments in the Central Andes south of La Paz contain important records of Pliocene-Pleistocene interchange. Seventy kilometres south of La Paz, the Ayo Ayo pyroclastic ash-flow tuff (2.8 ± 0.4 Ma; Lavenu et al., 1989), which probably correlates with the Chijini Tuff (Marshall et al., 1992), separates Pleistocene faunas with South and North American genera from underlying Pliocene faunas (Hoffstetter et al., 1971) comprising wholly pre-interchange mammals (MacFadden et al., 1994). The limited Plio-Pleistocene magnetostratigraphies of the Andes (Marshall et al., 1984; MacFadden et al., 1994) capture important biochronologic changes of GABI. North American mammals are absent during and prior to the Mammoth subchron (ca. 3.3 Ma) at the Inchasi section in southern Bolivia (MacFadden et al., 1993). The first appearance of Erethizon, Hippidion, and proboscideans in South America occur in the earliest Matuyama (ca. 2.5 Ma) in the Uquía Formation in northwest Argentina (Reguero et al., 2007; magnetostratigraphy by Marshall et al., 1982 and Walther et al., 1996). These and all other North American taxa are absent from the lower member of the formation dating

to the late Gauss (Reguero et al., 2007). Fossil finds in the La Paz area (e.g. Ahlfeld, 1946; Hoffstetter, 1986; Marshall et al., 1992; Montes and Mehl, 1990) may thus constitute an important, yet largely unexplored, record of biotic change linking the well studied fossil assemblages of the Argentine Pampas region (Marshall et al., 1984; Cione et al., 2015, and references therein) and those of southern North America (Woodburne, 2010, and references therein).

The Great American Biotic Interchange comprises multiple pulses of multi-taxa migration (Cione et al., 2007, 2015; Woodburne, 2010) that typically coincide with glaciations (Woodburne, 2010), suggesting conditions during Pleistocene glaciations promoted dispersion across the Isthmus of Panama. Eustatic sea-level depression would have increased the width and connectivity of the Panamanian land bridge (Woodburne, 2010), even before completion of the permanent dry-land connection (ca. 2.8 Ma: Bartoli et al., 1995). Additionally, development of savanna-like ecosystems during glaciations would have provided open corridors linking southern North American, Central America, and much of South America (Webb, 1991). The pre-GABI arrival of savanna-adapted camelids in South America at ca. 3.3 Ma (Cione and Tonni, 1995) and of savanna-adapted peccaries by at least the late Pliocene (ca. 3.7- 3.1 Ma; Prevosti et al., 2006; Woodburne, 2010) may be linked to early glaciations of North and South America (Woodburne, 2010). The correlative onset of glaciation in the high Central Andes, just before the beginning of the Mammoth subchron (Fig. 2.16), likely assisted southward migration of savanna-adapted species at the beginning of the GABI.

2.8. Conclusions

The Altiplano sediment sequence at La Paz faithfully records geomagnetic field directions suitable for identifying polarity and, in some cases, precise field positions resulting from secular variation. Sections can be reliably correlated on the basis of their polarity sequences and the presence of the Chijini Tuff (2.74 Ma). By tying lithostratigraphy to the GMPT with the dated tuff and composite polarity sequence, I was able to provide the most detailed chronology of the Bolivian Altiplano sediment sequence

to date. Correlation with the GMPT is robust for the part of the sequence below the tuff, which spans the latest Gilbert (ca. 3.7 Ma) and nearly all of the Gauss Chron (ca. 3.6- 2.74 Ma). Sediments above the tuff extend from the latest Gauss Chron until the latest part of either the Olduvai (ca. 1.8 Ma) or Jaramillo (ca. 1.1 Ma) subchron. Aggradation and incision rates, as well as the number and magnetostraigraphic position of glacial units above the tuff, support the interpretation that the Altiplano surface formed before ca. 1.78 Ma (OIS 63).

The Altiplano formed by sheet-like aggradation during glacial periods interrupted by soil forming intervals that spanned thousands to tens of thousands of years. Typical long-term aggradation rates (~12-14 cm/ka) are similar to rates in other, roughly contemporaneous basins of the Central Andes. Fining from diamictons near the Cordillera Real to silt and sand well into the basin reflects lateral facies changes linked to the distribution of glaciers in the cordillera. Dobrovolny (1952) was the first to document these facies changes, but I have placed them into a chronological framework through detailed the litho- and magnetostratigraphic work.

The glacial sequence north of La Paz provides Earth’s longest known record of low-latitude glaciation and the only record of pre-Pleistocene tropical Cenozoic glaciation. Its tills record at least 15 separate glaciations of the late Pliocene (nine) and Early Pleistocene (six), separated by interglaciations of sufficient length to produce paleosols and span geomagnetic polarity reversals. Onset of glaciation in the Central Andes generally correlates with the period of localized glacial expansion in North America, but substantially predates the earliest ice sheets yet found in many mid-latitude areas. It initiated no later than MIS MG2 and was followed by expansion during the globally recognized M2 cold peak. Older Cenozoic terrestrial glacial deposits are known only from the Patagonian Andes (>45° S), Iceland (65° N), and, possibly, east-central Canada (~53° N).

During the mid-Piacenzian warm period, under globally averaged atmospheric temperatures thought to be several degrees warmer than today, mountain ice caps developed during at least two glaciations in the tropical Andes where today alpine

glaciers are rapidly receding. The only other regions with records of glaciation during the mid-Piacenzian are Iceland, Antarctica, and possibly southeast Alaska, although the onset of tidewater glaciation in the last of these three areas is not precisely dated. Development of high elevation ice caps in the southern tropics during warm climates of the mid-Piacenzian, when similar records are lacking for much of the northern hemisphere, may relate to inter-hemispheric differences in orbital forcing (Dolan et al., 2011) or as-yet poorly understood effects of the Antarctic ice sheets, but could also simply be a result of the vagaries of geologic perseveration.

Ice cover in the Cordillera Real appears to progressively increase during successive glaciations from the end of this warm period until well into the Early Pleistocene, consistent with intensification of glaciations recorded in the marine oxygen isotope record.

The late Neogene and Early Pleistocene records at La Paz provides clues for the timing and processes of evolution of the Central Andean landscape. The development of extensive ice caps in the Cordillera Real at a time when climate was perhaps warmer than today suggests that the range had a similar or greater height in the Pliocene. High- angle faults of a graben at the northern limit of the city of La Paz were active during deposition of the Altiplano sediment sequence, but apparently not since then. Breaching of the Cordillera Real and subsequent incision of the enclosed Altiplano basin happened after ca. 1.8 Ma, triggering vertical incision and headwater extension at average rates of at least 40 mm/ka and 45 m/ka, respectively. This rapid erosion of the fill sequence has created steep-walled valleys in non-lithified to poorly lithified sediments that are responsible for the exceptional instability of slopes in La Paz. Additionally, the sequence provides an opportunity to improve age constraints on Plio-Pleistocene faunal evolutions and dispersal in the Americas, including the ‘Great American Biotic Interchange’.

Chapter 3. Urban landslides in La Paz, Bolivia, from 1995 to 2014

3.1. Introduction

Landslide inventories are a fundamental part of slope hazard characterization and risk management in urban environments, and individual case studies can provide important insights into specific events. However, for the inventory to be useful, it is necessary to document a large number of events to understand the frequency, size range, and typical behaviour of landslides in a particular area, as well as the factors controlling their occurrence (Ho and Lau, 2010; Klimeš and Rios Escobar, 2010; Smyth and Royle, 2010). An understanding of past events is thus key to reducing risk from landslides in urban areas (Alexander, 1989). Additionally, because urban development can change the frequency of hazards (Anderson, 1992) and concentrates elements at risk, longer term landslide inventories may help characterize changes in slope stability accompanying urbanization and the evolving risk landscape. Such knowledge is crucial given the typically high concentration of people and infrastructure in cities (Anderson, 1992; Wisner et al., 2003).

Despite the high incidence of landslides in the city of La Paz, Bolivia, even basic details of recent events are limited. Landslides in La Paz are grossly underreported in the international scientific literature (Shuster et al. 2002) and aside from a handful of reports are virtually unknown outside Bolivia. Some case studies are reported by the local media and in Spanish-language publications at national meetings, but these are not readily available outside of Bolivia. A few sources demonstrate the high vulnerability of, and losses suffered by, residents (O’Hare and Rivas, 2005; PREDCAN, 2009), but

geologic context necessary for appropriate physical characterization of events is lacking. An exception is a description of five, large twenty-first century landslides in La Paz (Hermanns et al., 2012), including the most damaging event in the city’s recent history – the 2011 Pampahasi ‘megalandslide’ (Fig. 3.1). This publication highlights the high levels of hazard and risk in the city. However, a better understanding of landslide processes, hazard, and risk requires consideration of the full range of events that the city has experienced.

Figure 3.1. Overview of the February 2011 Pampahasi landslide. The landslide is the largest and most damaging in La Paz’s recent history. It affected an area of ~2 km2. The headscarp below the horizon (between arrows) in the centre background is 80 m high. The entire slope involved in the landslide, down to the river in the right foreground, is 360 m high. Trees in the right middle ground are on the lower part of the landslide, but are less displaced. Terracing began immediately after the event. The pre-failure housing density was similar to that seen to the left of the landslide (May 2011 photo). The location of the view point is shown in Fig. 3.2.

In this chapter, I characterize modern landslide activity in La Paz, including its frequency, behaviour and associated impacts, by cataloguing and describing slope failures that have happened in the city between 1995 and 2014. The inventory is based on field investigation, collaboration with local experts, a review of scientific literature and media reports from Bolivia, and analysis of precipitation records. Data reported for events include their time, location, type, area, behaviour, rate, geology, and antecedent

rainfall. Societal impacts evaluated are limited to direct and some indirect impacts on local populations and on infrastructure loss; economic and secondary impacts are not systematically considered. This is the first published scientific account of many of these events and the first readily available characterizations of several others.

3.2. Setting

3.2.1. Physiography

La Paz is situated at the eastern margin of the Bolivian Altiplano along the foothills of the Cordillera Real (Fig. 3.2). Settlement is concentrated here because of the deep (300-1000 m) incision of the Altiplano by headwaters of the Amazon River. The lower elevations of La Paz (~3000-4000 m asl) relative to those of the adjacent Altiplano (4000-4500 m asl) and Cordillera Real (peaks exceeding 5500 m asl) result in a comparatively more favourable climate. However, incision has also formed high, steep slopes in the relatively weak sediments and sedimentary rocks underlying the Altiplano.

3.2.2. Geology

The sedimentary sequence underlying the Altiplano comprises more than 1 km of late Cenozoic clastic sediments ranging from clay and silt to coarse gravel and diamicton, almost entirely sourced from the Cordillera Real ~15 km northeast of La Paz (Ahlfeld, 1954a, 1946; Dobrovolny, 1962; Bles et al, 1977). It includes two laterally extensive pyroclastic deposits originating from mega-eruptions at the western limit of the Altiplano, which serve as marker beds throughout the La Paz area (Dobrovolny, 1962; Lavenu et al., 1989; Marshall et al., 1992). The ca. 5.5-Ma Cota Cota tuff (Lavenu et al., 1989; Servant et al., 1989) crops out at ~3500 m asl near the base of the sequence. It, and the fine-grained sediments below it are exposed in the southeastern part of La Paz where the late Cenozoic sequence is most deeply incised. The Chijini Tuff (2.75 Ma; Lavenu et al., 1989; Marshall et al., 1992 ; Chapter 2) is exposed in near-vertical cliffs

throughout the La Paz basin (Ahlfeld, 1954a, 1946; Dobrovolny, 1962; Anzoleaga et al., 1977) at ~3900 m asl. Sediments between these two tuffs are mainly fine-grained fluvial and lacustrine sediments (lower La Paz Formation; Dobrovolny, 1962), and are exposed throughout the southern and central parts of the La Paz basin. The sequence coarsens upward and toward the Cordillera, reflecting the influence of glaciers that periodically expanded southwestward from the Cordillera Real during the late Pliocene and Early Pleistocene (Chapter 2). North of the city, where the sequence is less deeply incised, valley slopes are formed in coarse glacial and glaciofluvial deposits that extend up to 200 m below the Chijini Tuff and up to 400 m above it (Dobrovolny, 1962; Anzoleaga et al., 1977; Chapter 2).

Incision of the Altiplano was accomplished during the past ca. 1.8 Ma or less (Chapter 2) and continues today (Safran et al., 2005; Zeilinger et al., 2008). Due to the incision, the exposed, nonlithified sediments are highly unstable (Chapter 2). The geotechnical properties and permeability of the sediments differ markedly over short vertical and lateral distances owing to their heterogeneity (Ahlfeld 1946; Dobrovolny, 1962; Bles et al., 1977). The Chijini Tuff, which is up to 14 m thick and is weakly cemented (Chapter 2), is the strongest unit in sequence. The tuff and the overlying glacial sequence typically form nearly vertical cliffs, although the latter are deeply gullied and include areas of gentler slopes. Fine-grained fluvial and lacustrine sediments of the La Paz Formation generally form gentler slopes (Dobrovolny, 1962; Anzoleaga et al., 1977). Unconformably underlying these sediments At the southern limit of La Paz the La Paz Formation unconformably overlies folded and faulted metasedimentary and metamorphic rocks of Paleozoic age (Bles et al., 1977).

The parallel nature of rivers and plateau margins suggest a possible tectonic control on the physiography of the La Paz area (Lavenu, 1977). Several known faults cut the fill sequence (Dobrovolny, 1962; Lavenu, 1977, 1978; Lavenu et al., 2000; Chapter 2), but their role in slope instability is unclear, as is whether these structures are active today. The Chuquiaguillo graben structure in northern La Paz, however, was active until the Early Pleistocene (Chapter2), and some faults of the Kenko structure immediately southwest of La Paz is thought to have been active in Late Pleistocene (Ramírez et al., 2009).

Figure 3.2. Locations of discrete failure events and ongoing slow failures in La Paz between 1995 and 2014 and their spatial relationship with past large landslides.

a) Location of study region within South America. b) Physiographiy of the Central Andes. c) Map of landslides in La Paz during the 20-year invenory period. Event numbers correspond to those in Tables 3.1 and 3.2. Note view point of Fig. 3.1 and location of Laykacota weather station (data in Figs. 3.5b and 3.6c). Terrain from ASTER DEM.

Slopes throughout much of the La Paz area are mantled by thick deposits from prehistoric large landslides (Dobrovolny, 1962; Anzoleaga et al., 1977; Fig. 3.2). Some of the largest of these old landslides probably represent single failures (Dobrobolny, 1962, 1968; Vargas, 1977). Most of these failures exclusively involved the La Paz Formation, but some also passively transported overlying Chijini Tuff and glacial sediments. At many locations, smaller more recent failures have displaced the paleolandslide deposits or previously intact La Paz Formation; Anzoleaga et al. (1977) label these in their mapping as ‘recently failed’.

Landslides have been documented since the sixteenth century. The include an event in 1582 near the Llojeta valley that claimed many lives (Santa Cruz, 1941; Sanjinés, 1948), several nineteenth century accelerations of the creeping Santa Barbara landslide (Sanjinés, 1948), and a landslide at Parque Laikacota (Sanjinés, 1948). The latter two sites are located at the east limit of the city centre. The catalogue becomes more complete toward the end of the twentieth century (Guzman, 2007a, b; O’Hare and Rivas, 2007; Roberts et al., 2010; Hermanns et al., 2012). Landslide susceptibility maps of the Llojeta (Galarza et al., 2005) and Allpacoma (Barragan et al., 2006) sub-basins indicate the extent and relative ages of events until 2007, and show that most of the surficial deposits at the southwest limit of the city are the product of mass movements.

Starting about 1970, lower portions of slopes at some sites in La Paz were benched, typically through cut and fill, to increase developable land. Consequently, anthropogenic fill is common in several parts of the city (Anzoleaga et al., 1977). Channels, including those of both perennial trunk streams and ephemeral streams in gullies, were culverted and buried in the course of development (Roberts et al., 2010; Fig. 3.3). More recently, entire slopes have been terraced to provide additional land for development or to stabilize recently failed slopes (Fig. 3.3c).

Figure 3.3. Buried culverts along the lateral margins of the 2009 Retamani landslide. See Fig. 3.2 for the location of the landside (no. 30). a) Landslide toe adjacent to the lower northern lateral margin of the landslide showing fragment of destroyed, previously buried culvert (arrowed) in the debris (March 2009 photo). b) Northern lateral margin of landslide viewed downslope toward the toe showing intact portion of surface culvert (arrowed) just beyond the edge of the landslide limit (note person for scale; March 2009 photo). c) Overview of the landslide showing locations of culverts along both lateral margins (arrowed) and locations of field photos (a, b, and d) along the northern lateral margin (April 2009 photo). d) Shallow buried culvert being installed along the northern lateral margin after the landslide (October 2010 photo).

3.2.3. Weather systems

Most moisture reaching La Paz area is sourced from the tropical Atlantic Ocean (Vuille et al., 2003). Seasonal changes in tropospheric circulation over the Altiplano favour precipitation at La Paz during the South American summer monsoon (December- March) (Zhou and Lau, 1998), when easterly flow aloft brings moisture from the Amazon basin over the Cordillera Oriental. Air masses ascend the upper eastern slope of the Cordillera Real, producing daily afternoon rainstorms (Sicart et al., 2003). During several field visits, I observed valley-confined weather systems moving up the Río La Paz valley from the southeast. Once moisture reaches the Altiplano is it recycled as convective storms producing episodic, high-intensity rainfall events (Garreaud et al., 2003; Vuille et al., 2003). In the austral winter, dry westerly flow aloft impedes advection of moisture air over the Cordillera Oriental (Vuille, 1999; Garreaud et al., 2003; Vuille et al., 2003), thus reducing both orographic precipitation in the Cordillera Real and convective precipitation

on the Altiplano. Inter-annual variability in precipitation in the Central Andes is linked to equatorial Pacific sea surface temperatures (Garreaud et al., 2003; Vuille et al., 2003). The main effect is a reduction of summer monsoon precipitation during El Niño periods and an increase in precipitation during La Niña periods (Vuille et al., 2003).

3.3. Methods

Local subject-matter experts – Marco-Antonio Guzman (Universidad Mayor de San Andrés [UMSA]) and Estela Minaya (Observatorio San Calixto [OSC]) – identified most of the pre-2010 landslides. They provided basic data, including location, date, and impacts. I identified other events in the two-decade period from other sources: review of technical literature; review of landslide inventories complied for the Llojeta (Galarza et al., 2005) and Allpacoma (Barragan et al., 2006) valleys during a multinational geoscience collaboration supported by the Canadian government (Proyecto Multinacional Andino: Geociencias para las Comunidades Andinas) and summarized in Quenta et al. (2006); searches of disaster response agency reports (e.g. http://reliefweb.int/disaster); searches of the online database of La Paz’s principle daily paper (La Razón; www.la-razon.com) between 2011 to 20141; and interpretation of high- resolution optical satellite imagery in GoogleEarthTM. I characterized the mechanisms, geology, and impacts of each failure through field investigation and discussion with the local experts during visits in each year from 2009 to 2014, as well as from consideration of ancillary datasets. The latter included published geologic maps of the La Paz area (Dobrovolny, 1962; Anzoleaga et al., 1977), satellite imagery in GoogleEarthTM, orthophoto mosaics produced by the municipality of La Paz in 1996 and 2006, discussions with affected residents, and local media reports. I retrieved daily precipitation records from SENAMHI (Servicio Nacional de Meteorología e Hidrología de Bolivia; www.senamhi.gob.bo) for the Laykacota weather station located in the city centre (16° 30’ 17” S, 68° 07’ 24” W, 3400 m asl) and calculated 14-day cumulative precipitation as a sliding total in Excel.

1Current coverage of the online repository of newspaper articles beings in December 2011.

I classified failure modes and rates according to Hungr et al. (2014). In addition to the initial failure mechanism for each landslide, I identified any different, subsequent movement types Two circumstances complicated the classification of some landslide mechanisms. First, I was unable to discriminate between some mechanisms, such as sediment falls and sediment topples, due to a lack of direct observation of slopes prior to or during failure. To reflect this uncertainty, I grouped topples and falls together. Second, urban slopes in the La Paz area are typically heavily modified almost immediately after a landslide through a combination of debris removal, terracing, and infilling (Hermanns et al., 2012; Figs. 3.1 and 3.4). These modifications preserve the gross morphology of the landslide, but complicate interpretation of landside mechanisms and, in some cases, landslide limits. Thus, landslides that I could not view in person, in photos, or in remotely sensed imagery prior to major post-landslide regrading could have included minor subsequent movement types not recorded here. For instance, re-graded landslide deposits may preserve evidence of rotational sliding in the form of steep arcuate headscarps, but debris modification at its toe may remove evidence of kinematic transformation to a small flow-type movement downslope.

For many landslides, the failure depth and the boundary between depletion and accumulation are not known, preventing reliable estimates of volumes. Consequently, I used the total area of landslides (depletion zone, transport zone and accumulation zone) rather than volume to characterize the magnitude of discrete landslide events.

Additionally, I mapped the limits of large (>10 ha) landslides north of La Paz from their surface morphology and drainage diversion using a combination of satellite imagery and elevation information in GoogleEarth™. Knowledge of their location and extent is necessary to enable comparison of landslides between 1995 and 2014 with large previous instabilities of the La Paz valley system, including those previously mapped in the city of La Paz.

100

3.4. Results

3.4.1. Landslide inventory

Slope failures in La Paz between 1995 and 2014 are of two basic types: discrete events with movement rates ranging from metres per day to metres per second that generally terminate within minutes to a few days of initiation (Table 3.1); and slow ongoing landslides that were active throughout much of the inventory period or even longer (Table 3.2). The two groups of failures are henceforth considered separately under the headings ‘landslide events’ and ‘ongoing landslides’. I identified 43 discrete landslide events during the 20-year inventory period and 13 locations of ongoing slow landsliding.

North of La Paz, I defined 19 large landslides (Fig. 3.2) varying in size from 15 to >300 ha. Of these, only the Limanpata landslide was previously recognized; Dobrovolny (1962) interprets it as a glacial feature, but Anzoleaga et al. (1977) and Vargas, (1977) correctly map its southern limit and entire area, respectively. A bulk radiocarbon age (9200 ± 250 yr BP; Rubin and Alexander, 1958) from the lake deposits directly up valley of this dam formed by this landslide indicate is occurred in the early Holocene. The ages of the other 18 these landslides are unknown, but based on their muted morphology, they certainly pre-date the inventory period considered here. Together with mapping by Anzoleaga et al. (1977), the large landslides I mapped provide a fairly complete overview of large previous failures in the La Paz area.

Spatial distribution and geology

Discrete events

The locations of most (34) of the known discrete landslide events are precisely known. Thirty-two of the 34 occurred in the southern half of the La Paz basin, south of the city centre (Fig. 3.2) where the sediments are mainly silt and sand (Dobrovolny, 1962; Anzoleaga et al., 1977; Chapter 2). Only two landslides happened farther north,

101

* * * * * * * * * * * * * * * * * * * * * g f, 2 2 1 1 1 2, 3 4 6 6, 7, 8 1 1, 3 3 1 1 6 1, 11 1 12 1 1, 11, 13 15 16 5 1 1, 3 1 1 2 2 2 6, 9, 10 1, 11 1 6, 14 16 Sources e Material

remobilized La Paz Fm. / La Paz Fm. Paz La / Fm. Paz La paleolandslide possible / remobilized Fm. Paz La / Tuff fill / Chijini paleolandslide remobilized Fm. / Paz Fm. La / Paz La Fm. Paz La remobilized / paleolandslide Fm. Fm. Pampahasi / Paz La Fm. Paz La remobilized / paleolandslide paleolandslide / Fm. Paz La / fill Fm. Paz. La Fm. Paz - La remobilized / Fm. Paz La - - - Fm. Paz La outwash glacioal / Fm. Paz - La / Fm. Paz La Fm. remobilized Purapurani / Fm. Paz La / paleolandslide Fm. Paz La remobilized / paleolandslide Fm. Paz La remobilized - Fm. - Purapurani / Fm. Paz La / paleolandslide - Fm. Paz La remobilized / outwash paleolandslide glaciofluvial / Tufff Chijini fill / Fm. deposits Paz La glaciofluvial and glaical remobilized Fm. Paz La remobilized remobilized type material unknown of fill / Fm. Paz La / paleolandslide Fm. paleolandslide Paz La intact possibly paleolandslide, paleolandslide Fm. Paz La remobilized outwash glaciofluvial Fm. Pampahasi / paleolandslide fill paleolandslide, Fm., Paz La Fm. Paz La Fm. Paz La Fm. Paz La remobilized ------1 5 3 1 1 4 2 1 2 7 3 2 2 2 1 0 1 0 74 14 10 0.4 0.3 0.4 0.3 0.8 0.1 0.9 0.3 0.8 143 (ha) 21.7 Area d - - - - - very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid very rapid Velocity class of class of Velocity extremely rapid extremely extremely rapid extremely rapid extremely rapid extremely rapid extremely rapid extremely rapid extremely rapid extremely likley very rapid very likley rapid very likely main movement main - - - - - d mud flow mud flow mud flow mud mud flow mud Subsiquent debris flow debris flow debris avalanche debris topple flow, mud Type of movement Type of - - - - - Initial topple/fall topple/fall silt topple/fall silt compound slidecompound slidecompound debris topple/fall debris topple/fall debris topple/fall silt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound slidesilt compound silt block topple/fall block silt silt block topple/fall block silt debris compound slidedebris compound é c Discrete landslide events in La Paz between 1995 and 2014.. 1995 Paz between in La Discrete events landslide Name Cumbalacion Av. Cotahuma Centenario Cuarto San Isidro Calama Cuartel Kupini II I Retamani Seguncoma Cotahuma Jucumarini San Simon 17 Calle Zone 3 las Flores de Valle Ventilla La Allpacoma Llojeta Bajo Antonia San Las Lomas Avila Federico Final Bolivar Calle Allpacoma RosalesLos Llojeta Rosal Norte II Retamani Patapatani Allpacoma Bajo Hampaturi II Retamani Salom Villa Huanu Huanuni I Retamani Armaza Calle Altotocagua Pampahausi Las Lomas Zamudio Adela Zamudio Adela Pipiripi

. a,b 1 . 3 Date 1995/06/01 1996/04/09 1997/04/09 1997/12/19 1998/10/06 1999/03/21 2000/02/02 2000/12/15 2001/01/03 2001/01/04 2001/01/- 2001/01/- 2001/01/- 2001/03/01 2001/05/04 2002/09/08 2003/03/13 2003/03/17 2003/08/01 2003 2004/04/05 2004/07/18 2005/03/16 2005/07/27 2005-2007 2007/09/29 2007-2008 2008/01/23 2008/01/25 2009/02/16 2009/10/11 2010/01/27 2010/02/04 2010/09/26 2010/10/13 2011/02/27 2012/02/01 2012-2013 2013/02/17 2013/02/17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 No. Table

102

19, 20 16 17, 18 2008 ); 2013 ); ation with map La Razón ( Razón La

de area, with most the 10, Quenta et al., ( 2007b ); 5, Guzmán 20 , Gobierno Autónomo 2007 ); 2014b ); La Razón (2012); 16,

, 5 2007 ); 4, Guzmán ( , La Razón ( 2013 ); 1 9 La Paz Fm. Paz La - Fm. Paz La remobilized UNDP (

2005 ); 9, Quenta et ( al. - 4 0.3 Aguilar ( , 2014a ); 1 2005 ); 3, 14

. very rapid very rapid 2010 ); likely very rapid very likely 2008 ); 8, Constance ( mud flow mud 13, La Razón (

2009 ); unication ); 2, O’Hare and Rivas ( 2014 ).

Order of listed materials reflects their relative contribution to the overall landsli

2012 ); 7, Mobarec et al. ( , Hardy ( comm

12 debris topple/fall

, Gobierno AutónomoMunicipal ( de La Paz silt compound slidesilt compound slidesilt compound

1977 ). . 2014b ).

Hungr et al. ( ’. 2010 ); - 2014a ); 1 after

Alto Florida Alto Rosal El Chico Monterani and rate 2007b ); 6, Hermanns et al. ( ( 11 , Roberts et al. ( from Anzoleaga et al. ( most extensive material types listed first. Material types comprising less than 10% of the landslide area (e.g. very localize d fill) are not included. Municipal de La Paz ( , La Razón ( Razón La 17 , 2013/02/- 2014/03/16 2014/03/28 41 42 43 Date format as YYYY/MM/DD Missing data denoted by ‘ Mechanisms Material type based on a combination of field investigation (denoted by ‘*’ sourcein column) and comparison of landslide loc Sites visited thein field denoted by ‘*’. Event names derived from local names of neighbourhoods or streets Sources: 1, M.A. Guzmán (personal

a b c d e f g

103

ocation with map f * * * * * * * * * * * * * e d, 1 1 1 1 1 Sources 21, 22 21, 22 1 1 1

). c 1962 Material ); 22, Dobrovolny (

paleolandslide paleolandslide paleolandslide paleolandslide paleolandslide paleolandslide remobilized La Paz Fm. Paz La remobilized paleolandslide paleolandslide / remobilized La Paz Fm. Paz La remobilized / paleolandslide) of paleolandslide slope up (directly Fm. Pampahasi paleolandslide Fm. Paz La remobilized paleolandslide 1948 nés ( 1977 ). Mechanismsfor the other landslides are best guesses based on local b

Velocity extremely slow extremely slow extremely extremely slow extremely slow extremely slow extremely slow extremely slow extremely extremely slow extremely slow extremely extremely slow extremely slow extremely slow extremely slow extremely 1977 ). b 2014 ). Mechanisms for the Santa Barbara and Los Olivios landslides are based on previous

1962 ) and Anzoleaga et al. (

(

. 1977 ). Order of listed materials reflects their relative contribution to the overall landslide area, with most the silt rotational slide rotational silt slide rotational silt 3.2 silt compound slidesilt compound slidesilt compound silt compound slidesilt compound slidesilt compound silt compound slidesilt compound silt translational slide translational silt slide translational silt silt translational slide translational silt debris rotational slide rotational debris slide translational silt slide translational silt Main failure mechanism Fig.

1962 ) and Anzoleaga et al. ( Ongoing slow landslides in La Paz showing motion between 1995 and 2014. 1995 and motion between showing La Paz in landslides Ongoing slow

. a mapping by Dobrovolny topography and relation to previous large landslide deposits together with grain size observed thein field and indicated by Dobrovolny ( from Anzoleaga et al. ( most extensive material types listed first. Material types comprising less than 10% of the landslide area (e.g. very localize d fill) are not included. 2 . 3

Locations shown in de las Carmelitas (small) las Carmelitas de (main) las Carmelitas de 31 Calle 32 Calle 33 Calle Las Lomas del Sur del Lomas Las Porveni r Santa Barbara Santa Cervezaria La Armonía Villa Olivios Los Huacollo Name Laikacota Parque 56 51 52 53 54 55 50 44 45 46 47 48 49

No. Note: a Event names derived from local names of neighbourhoods or streets. d Sources: 1, M.A. Guzmán (personal communiction); 21, Sanji b Mechanisms and rate after Hungr et al. ( c Material type based on a combination of field investigation (denoted by ‘*’ sourcein column) and comparison of landslide l Table

104

where the sediment sequence comprises glacial diamicton and glaciofluvial gravel (Chapter 2) – one of these occurred well north of the city in the upper part of the Río Irapvi valley (Hardy, 2009), and the other happened in the Río Kaluyo valley just north of the city limit.

Five events reported only in non-geoscientific literature (four events in O’Hare and Rivas, 2005; one event in UNDP, 2007) could not be precisely located. In addition, exact locations of one event mentioned in a newspaper report (La Razón, 2013) and three events reported by local experts are uncertain (Table 3.1). Their locations can be constrained only to districts or specific neighbourhoods based on their reported names or descriptions. As a consequence, the geologic context of these nine events is uncertain.

Mapping of large preexisting landslides north of La Paz indicates that the 2009 Hampaturi landslide happened in previously failed material, although its specific geologic properties are unknown because it lies outside the map coverage of Dobrovolny (1962) and Anzoleaga et al. (1977).

Twenty-eight (over 80%) of the 34 located landslides involved previously failed material (Table 3.1): six landslides occurred completely within previously failed material; five involved mostly (50-90%) previously failed material; and seven included a subordinate component (<50%) of previously failed material. Of the total area involved in landslides between 1995 and 2014, about three-quarters involved previously failed material and one-quarter material that had not previously failed.

Between 1995 and 2014, most landslides events coincided with the margins of larger, previously failed deposits. Of the 34 precisely located landslides, 26 (over 75%) occurred along or within 20 m of these margins. They happened at the contact between paleolandslide deposits and non-disturbed materials (15), at the contact between more recent landslide deposits and non-disturbed material (3), and at the margins of paleolandslide deposits that had remobilised since their initial failure (8). Two landslides (2008 Bajo Allpacoma silt compound slide and 2013 Pipiripi silt fall/topple) that occurred

105

within deposits of recent landslides, but more than 20 m from the limits of these past failures.

Only six landslide events in the inventory occurred entirely in intact material. Three were small failures in in situ La Paz Formation along steeply incised river channels; two were failures in the Chijini Tuff (1996 Cotahuma and the undated failure at Rosal Norte); and one event was a small failure of glaciofluvial gravel in the upper part of the sedimentary sequence beneath the Altiplano (2010 Altotocagua).

Ten failures involved artificial fill placed over culverted streams (Anzoleaga et al., 1977) and gullies, but the fill was only a minor portion (generally 10% or less) of the total area of each landslide. Of these ten failures, seven had lateral margins that coincided with buried culverted channels, one initiated downslope of a culverted channel, and two included elements of both (Roberts et al., 2010; Fig. 3.3).

Ongoing slow failures

The spatial distribution of the ongoing slow landslides is similar to that of the more rapid discrete failures (Fig. 3.2). All 13 are located in the central and southern part of the Río La Paz valley system, where the sediment sequence is predominantly silt and sand. Ten of them (>75%) are associated with large paleolandslide deposits mapped by Anzoleaga et al. (1977) (Fig. 3.2; Table 3.2). Six of the ten are located at the margins of paleolandslides and four are within, but less than 200 m form the margins of the paleolandslides. The other three ongoing landslides – Santa Barbara, Parque Laikacota, and Los Olivios landslides – correspond to smaller, more recent landslides mapped by Dobrovolny (1962) and Anzoleaga et al. (1977).

106

Mechanism and rate

Discrete events

The mechanism and movement rate of 38 of the 43 landslide events are known. They are unknown for five events that I could not precisely locate. Two types dominate: compounds slides (29, ~75%) (Fig. 3.4a-c) and topples/falls (9, ~25%) (Fig. 3.4d).

The main failure stage of the compound slides were typically on the order of many metres per minute. In some cases, very rapid motion was preceded by slower movement at rates of metres per day to metres per hour. Based on debris morphology and eyewitness accounts, one compound slide (2007 or 2008 Patapatani landslide; Fig. 3.4b) transformed into a debris flow and six compound slides, including all five large twenty first-century landslides described by Hermanns et al., 2012 (2002 and 2004 Allpacoma, 2003 Llojeta, 2005 Llojeta, and 2011 Pampahasi) transformed into mud flows. Local media reports (Gobierno Autónomo Municipal de La Paz, 2014a; La Razón, 2014a) of the 2014 El Rosal flow include photos of the El Rosal bridge ~0.5 km downstream of the source area, showing channel-filling deposits of what appears to be a viscous mudflow. The name applied to the event in these reports (‘mazamorra’ – a type of corn porridge) is used in parts of Latin America to denote mass flows with viscosity and behaviour similar to that of porridge (Nusser, 1887; Morum, 1936). These landslides range from surging ‘earthflows’ to ‘mud flows’ in the classification of Hungr et al. (2014). Whether some of the compound slides similarly developed into flows is uncertain because soon after the landslides, affected slopes were terraced, thus removing some geomorphic evidence of the events. Such ambiguity is illustrated by the Huanu Huanuni landslide (Fig. 3.4b). Nevertheless, several compound slides, including the 2009 and 2010 Retamani landslides and 2010 Huanu Huanuni landslide (Fig. 3.4a), did not change into flows.

Topples and falls occurred either in silt (three events), debris (four events), or material of unknown lithology (two events). Following release, they fragmented and moved extremely rapidly downslope. In one instance, a topple or fall transformed into a flow-type landslide. Hardy (2009) suggests the 2008 Hampaturi landslide was triggered

107

lational liding. b) The by by human activity. ha landslide was dominated by s nto a debris flow, the deposit of which forms roughly 80% . The slide transformed i

grading (March 2010 photo by M.A. Guzmán, used with permission). Note the vertical tree in the s on these events. ha landslide occurredpredomina ntly thein Chijini Tuff, but involved a smaller amount of overlying silt, ha area. d). The Rosal Norte silt block topple/fall, which happened between 28 May 2005 and 15 April 2007 (March 2013 photo). Initation involved compound sliding (dotted line shows approximate headscarp position) on the 2008

or Examples of landslide events in La Paz between 1995 and 2014 illustrating the range of mechanisms. of illustrating range 2014 the 1995 and Paz between inevents La landslide Examples of

2007 . 4 provides additional detail .

3 of the landslide’s overall 3 - (September 2012 photo). The 0.3 - sand, and gravel of the La Paz Fomration. Unlike nearly all other historic landlsides La in Paz, it has not beenyet modfiied Table 3.1 Figure a) The 2010 Huanu Huanuni silt compound slide the day after occurred, it showing steep rotational headscarp and shallow trans sliding farther downslope (January 2010 photo by M.A. Guzmán, used with permission). The 2 - same slope two months later after slope re - lower part of the landslide indicating failure was locally very shallow. c) The Patapatani debris compo und slide and debris flow, which occurred in slope that has since been terraced during road construction

108

by overflow from the Hampaturi reservoir a short distance up the valley. His limited description suggests a high water content that, together with the rupture of the aqueduct below the landslide, suggests a debris flow.

The 2011 Pampahasi landslide illustrates the complex nature of many landslides in La Paz and that can be expected in some future failures. Eyewitnesses indicated that rapid motion initiated along the upper part of the landslide’s southern lateral margin, ~300 m downslope of the headscarp, and spread both upslope and downslope (Hermanns et al., 2012).The event was characterized by rotational sliding, particularly in the upper region where the failure surface formed a nearly vertical, 80-m-high headscarp (Fig. 3.1), and translational sliding farther downslope including the area at its toe where several bridges crossing Río Irpavi were partially buckled. Subsequent behaviour of the four-day-long complex landslide included localized mud flows originating along the upper lateral margins and repeated toppling and collapse along the steep backscarp and headscarp.

There is a relation between the initial failure mechanism and the size and geology of the landslides in the inventory. The 26 compound slides of known composition range in area from 0.1 to 143 ha (Table 3.1) and cumulatively comprise ~75% previously failed material and ~25% intact material. In contrast, the eight topples/falls of known composition range in area from ~0.0001 to 5 ha (Table 3.1) and cumulatively comprise ~ 80% intact material and only ~20% previously failed material.

Ongoing slow failures

The observed type and degree of damage to local infrastructure (Mansour et al., 2010) and the limited surface morphology change between field visits suggest that ongoing slow landslides move at rates of centimetres to tens of centimetres per year. These low rates make it difficult to constrain the extent and behaviour of movement. Study of the Santa Barbara landslide by Dorbrovolny (1962) indicates that it is a slump. The approximate pre-urbanization limits of the Los Olivios landslide mapped by Anzoleaga et al. (1977), in combination with the local topography, suggest its motion is also largely rotational. The mechanisms of other extremely slow failures listed in Table

109

3.2 are uncertain, but sliding generally appears to be involved. Those at the lower limits and headscarps of previous landslides are most likely to be, respectively, translational and rotational.

Impacts

Discrete events

Losses are known for some landslide events, but not others. Most of the landslides occurred on developed slopes and consequently damaged or destroyed homes and roads (Table 3.3). A few landslide events affected sites free of buildings, including cliffs (2005-2007 Rosal Norte) or channel bottoms (2014 El Rosal). Only two of the 43 landslide events – the 2007 or 2008 Patapatani landslide and the 2008 Hampaturi landslide – occurred beyond the urban limits of La Paz (Fig. 3.2). Nevertheless, the Hampaturi landside caused significant damage in a rural settlement and indirectly affected residents living down valley in La Paz (Hardy, 2009).

The landslide events between 1995 and 2014 displaced nearly 12,000 people, damaged or destroyed almost 1600 homes, and claimed 21-28 lives (Table 3.3). Half of the displaced residents and almost two-thirds of the destroyed homes resulted from the 2011 Pampahasi landslide (Table 3.3). All of the deaths stemmed from the two extremely rapid failures resulting from topples or falls. Collapse of an exposure of glaciofluvial outwash caused one death (2010 Altotocagua). The largest topple between 1995 and 2014 (1996 Cotahuma) was also the deadliest landslide during the 20-year study period . It caused at least 20 deaths (O’Hare and Rivas, 2005), but as many as 27 people may have perished according to some local recollections (M. Guzmámn, personal communication, 2009). The number of landslide deaths in La Paz during the inventory period may, therefore, be as high as 28.

Additionally, over one-half million residents were without water following two events three years apart that ruptured the water line from the Hampaturi reservoir 10 km northeast of the city. The 2008 Hampaturi landslide directly below the reservoir cut the water supply of at least 272,000 people for 19 days (Hardy, 2009). The 2011 Pampahasi

110

Table 3.3. Documented losses from discrete landslide events and ongoing failures between 1995 and 2014.

No. Year Name Lossesa,b,c Other impacts Displacedc Homes Deaths 1 1995 Av. Cumbalacion several several 0 2 1996 Cotahuma 400 45 20 3 1997 Cuarto Centenario 450 60 0 4 1997 San Isidro 165 17 0 5 1998 Cuartel Calama 120 10 0 6 1999 Kupini II 250 32-72 0 7 2000 Retamani I 56 5 0 8 2000 Seguncoma 60 12 0 9 2001 Cotahuma - - - 10 2001 Jucumarini - - - 11 2001 San Simon - - - 12 2001 Calle 17 - - - 13 2001 Zone 3 - - - 14 2001 Valle de las Flores 240 35 0 15 2001 La Ventilla 130 15 0 16 2002 Allpacoma - 3 0 17 2003 Llojeta 325 45 0 11 factories destroyed, eliminating 100s of jobs 18 2003 San Antonia Bajo 135 15 0 19 2003 Las Lomas 180 25 0 20 2003 Federico Avila - - - 21 2004 Calle Bolivar Final 128 12 0 22 2004 Allpacoma - 1 - 23 2005 Los Rosales 200 25 0 24 2005 Llojeta - 2 0 25 2007 Retamani II 110 d ~20 0 26 2005-07 Rosal Norte 0 0 0 Discrete landslide events Discrete 27 2007-08 Patapatani 0 0 0 28 2008 Bajo Allpacoma 120 16 0 29 2008 Hampaturi several several 0 crop damage; >272,000 without water for 19 days 30 2009 Retamani II 625 50-60 0 31 2009 Villa Salomé 932 25 0 32 2010 Huanu Huanuni 650 72 0 33 2010 Retamani I 90 d 15 0 clothing factor damaged 34 2010 Calle Armaza 7 1 0 35 2010 Altotocagua 0 0 1 36 2011 Pampahausi 6000 1000 0 200,000-300,000 without water for several months 37 2012 Las Lomas 45 5 0 38 2012-13 Adela Zamudio 0 0 0 39 2013 Adela Zamudio 0 0 0 40 2013 Pipiripi 0 0 0 41 2013 Alto Florida - - - 42 2014 El Rosal 0 0 0 clogged Río Pasajahuira at El Rosal bridge 43 2014 Monterani Chico 50 d 10 0 Total documented impacts 11470 1575 21 continued on next page

111

44 Santa Barbara multiple >50 0 45 Parque Laikacota multiple 0 0 46 La Cervezaria multiple >10 0 47 Villa Armonía multiple >50 0 48 Los Olivios multiple >10 0 49 Huacollo multiple >10 0 50 Las Lomas del Sur multiple >10 0 51 Porveni r multiple >10 0 52 de las Carmelitas multiple >10 0

Ongoing landslides 53 de las Carmelitas multiple >10 0 54 Calle 31 multiple >5 0 55 Calle 32 multiple >5 0 56 Calle 33 multiple >5 0

Total documented impacts 185 0

Note: Sources and details for each landslide listed in Tables 3.1 and 3.2. Locations shown in Fig. 3.2. a Missing data denoted by ‘-’. b ‘Several’ denotes a likely small, but unspecified, degree of loss or damage. For the purpose of loss tallying, the number of affected units (displaced people, damaged homes, or deaths) is considered as 1 to provide a conservative estimate of overall loss. c Where a range of estimates is given, the lowest value is considered in the tallying of loss to provide a conservative estimate of overall loss. d Displaced residents quantified in sources by number of families, not number of individuals. The number of individuals is estimated by multiplying the number of families by five (a conservative estimate of the average family size suggested by counts of affected people in events for which both number of families and number of individuals are known). Numbers of displaced families provided in sources are: event no. 26 = 22 families; no. 33 = 18 families; no. 39 = 10 families. landslide near a major water treatment plant left between 200,000 (Hermanns et al., 2012) and over 300,000 (E. Minaya, personal communication, 2011) people without water for several months.

Landslides between 1995 and 2014 caused additional damages that I have not systematically quantified (Table 3.3). Examples include damage to crops (2008 Hamapturi; Hardy, 2009), closure of hospitals and schools due to landslide damage or disruption of water service, destruction of roads (2008 Hamapturi; Hardy, 2009; and 2011 Pampahasi), and loss of jobs and industrial capacity (2003 Llojeta; O’Hare and Rivas, 2005; Hermanns et al., 2012). Even events with no reported impacts on infrastructure or people involved costly clean-up. For instance, the municipality of La Paz removed debris of the 2014 El Rosal flow, which was lodged beneath a road bridge that provides access to the neighborhood of El Rosal (Gobierno Autónomo Municipal de La

112

Paz, 2014a). The debris blocked the Río Pasajahuira, greatly increasing the chance of future flooding.

Ongoing slow failures

Previous researchers have not quantified impacts at locations of ongoing slow landslides. The estimates of affected homes provided in Table 3.3 are based on the approximate extents of disturbed areas and the number of homes in these areas, as well as discussions with local residents. A minimum of 185 homes have probably been damaged, although generally not destroyed, by these slow movements between 1995 and 2014. However, the occasional destruction of buildings by apparent surging of the Santa Barbara landslide in the nineteenth century (Sanjinés, 1948) highlights the possibility of increased future damage by ongoing landslides should they accelerate. Apparently no lives have been lost, as they would have been reported by the local media and remembered by local residents.

Temporal distribution of discrete events

The year (40 events), month (39 events), and exact date (35 events) of most of the discrete landslide events are known. Three events detected in satellite imagery are constrained to periods ranging from nine to 23 months. For instance, blocky debris was noted behind the Rosal Norte subdivision (Fig. 3.4d) during fieldwork in 2010 and was determined to have been emplaced between 28 May 2005 and 15 April 2007. The Patapatani silt compound slide (Fig. 3.4c) was constrained to a 15-month period between 2007 and 2008, and the first of two debris topples/falls at Calle Adela Zamudio occurred sometime in a nine-month period between 2012 and 2013. I could not determine the date of a 2003 landslide mentioned only by UNDP (2007; at Federico Avila). Four events are constrained only to the month in which they occurred – three events mentioned by O’Hare and Rivas (2005) as occurring in January 2001 and a single event mentioned by La Razón (2013; 2013 Alto Florida compound slide).

The landslides were generally evenly distributed through the inventory period (Fig. 3.5), with one or two events in 16 of the 20 years of the record. Only 2006 had no

113

day running average of rainfall leading up to

Precipitation and dates of 43 events. Precipitation

) and are given as a 31 - nknown (grey labels) are represented by error bars in ‘a’ and vertical Time series of landslide events in La Paz from 1995 to 2014. 1995 from La Paz in events landslide Time series of

. 5 . 3 Landslide area and type (for events all of precisely known location: 34 events). b)

) Figure a data from the Laykacota weather station (location shown in Fig. 3.2 teach day of the time series. Events for which the exact date is u grey zones in ‘b’. Landslide numbers correspond to those Table in 3.1, which provides event details.

114

Figure 3.6. Distribution of landslide events and average precipitation by month.

115

a) Counts of landslides known to have occurred in a particular month (37 events); note high numbers in January, February, and March. b) Sizes of landslides as a function of month of occurrence. Landslide numbers correspond to those in Table 3.1, which provides event details. c) Monthly average precipitation at the Laykacota weather station (location shown in Fig. 3.2) between 1995 and 2014 (monthly ranges over the 20-year period are given above each bar). Most precipitation occurs between December and March.

confirmed events, although the topple or fall of Chijini Tuff at Rosal Norte may have occurred that year. Seven known events happened in one year (2001); five of them are based on scant details provided by O’Hare and Rivas (2005), with limited information on exact location, size, or impact. The other 18 years of the record each experienced between one and four landslides.

Events were largely concentrated in January, February, and March (23 of 39 events, 59%; Fig. 3.6a). The other events (16, 41%) show no clear monthly trend. Some periods between 1995 and 2014 experienced clusters of events (Table 3.1; Fig. 3.5). Five events occurred in January 2001 (those mentioned by O’Hare and Rivas, 2005) and were preceded by another event two weeks earlier. Pairs of closely spaced events happened in March 2003 (4 days apart), January 2008 (2 days apart), February 2013 (on the same day) and March 2014 (12 days apart). Two landslides occurred only days apart in late January and early February 2010 (Huanu Huanuni and Retamani I).

Of the seven topples or falls dated to the month, five occurred during January, February, March, and April (Fig. 3.6a). A single topple/fall event (2010 Altotocagua) occurred in October. There is no clear monthly pattern in the size of landslide events (Fig. 3.6b).

3.4.2. Relation to precipitation

Most precipitation falls between December and March, with a peak in January (Fig. 3.4c), overlapping the months during which landslides are most common (Fig. 3.4a). The wettest month of the 240-month period between 1995 and 2014 (January 2001, 179.8 mm) experienced the most known landslides (7 events). It was preceded by

116

the second highest December rainfall of the period (128.7 mm), during which another landslide event occurred.

Eighteen (~51%) of the 35 dated landslide events occurred on days with under 1 mm of precipitation, and 16 of these happened on days with no measurable precipitation. The other 17 dated landslides occurred on days with between 1.1 and 18.2 mm of rainfall. Thus, no clear relationship is apparent between daily rainfall and landslide initation.

A consideration of cumulative precipitation in the weeks before failure reveals clearer trends. Over half (18) of the dated landslides happened on days with cumulative precipitation of more than 40 mm during the previous two weeks (Fig. 3.7). This biweekly rainfall total was exceeded on only 1271 days (~17%) of the 7305 days of the 20-year inventory period. About one-third of the dated landslides (11 events) happened on days with cumulative precipitation of more than 70 mm during the previous two weeks (Fig. 3.7), a condition with a 4% exceedance probably (322 of 7305 days). They include several of the most damaging landslides in the inventory (2008 Hampaturi and 2011 Pampahasi). However, several other highly damaging landslides happened during relatively dry periods. Both of the fatal landslides in the inventory (1996 Cotahuma, 2010 Altotocagua) occurred following only periods of moderate two-week cumulative precipitation (23 and 35 mm, respectively).

Of the five landslides with estimated volumes of 1 Mm3 or more (Hermanns et al., 2012), three (2002 Allpacoma, 2004 Allpacoma, and 2005 Llojeta) occurred on days with 14-day cumulative precipitation of less than 20 mm. During the 14 days prior to the 2003 Llojeta landslide, 49.7 mm of rain fell. Only the Pampahasi megalandslide was preceded by a particularly wet two-week period (91.9 mm).

Both the date and area are known for 32 landslide events. Comparison of area and antecedent 14-day cumulative rainfall for these events shows little or no relation between landslide size and antecedent rainfall (Fig. 3.7). Only topples/falls show a possible weak relationship between timing and preceding cumulative rainfall, with all

117

seven events of known date occurring after 14-day periods with more than 20 mm of rain.

Figure 3.7. Comparison of size and initial failure mechanism of landslides and preceding cumulative 14-day precipitation. The plot shows the 31 events (25 compound slides, 6 topples/falls) for which both the precise location and time are known (the area, antecedent precipitation, or both are unknown for the other 12 events. The largest and most damaging event in La Paz’s recent history and the two fatal events between 1995 and 2014 are labelled. Event numbers correspond to those in Table 3.1, which provides event details.

118

3.5. Discussion

3.5.1. Landslide types

The inventory includes landslides of three general types: extremely slow slides, landslides initiating as very rapid compound slides, and extremely rapid topples or falls. The slowest landslides are creep-like failures that were active throughout much of the inventory period (Table 3.2; 13 locations) and that may, on occasion, accelerate (Sanjinés, 1948). Examples are the Santa Barbara, Parque Laikacota, and Los Olivios landslides, which are previously mapped features ~10 ha in size (Dobrovolny, 1962; Anzoleaga et al., 1977). The long-term activity of these landslides is uncertain. The Santa Barbara and Parque Laikacota landslides experienced motion as early as the nineteenth century (Sanjinés, 1948), and the Los Olivios landslide was active in the early twentieth century. Thus, at least three of the slow ongoing landslides involved either punctuated or continuous movement for many decades to several centuries. The other ten extremely slow ongoing landslides are difficult to characterize, but appear to involve rotation, translation, or both, occurring over much longer periods than observed here. Their spatial extent is also uncertain, and thus it is unclear whether they are only small landslides or are the active portions of larger paleolandslide deposits to which their locations correspond. Other sites of ongoing slow landslide motion undoubtedly exist in the La Paz basin and have not yet been identified.

The other landslide types involve discrete failures and are sufficiently rapid that they can generally be characterized in more detail (Table 3.1). They comprise landslides initiating as compound slides (29 events) and as topples or falls (nine events); five events of uncertain mechanism likely belong to one of these two groups. Most of the larger and more damaging events in La Paz between 1995 and 2014 are captured in my inventory, but an unknown number of small discrete landslides have likely gone unnoticed. Based on the sizes of landslides in the inventory (Fig. 3.7), it is likely that most events affecting areas larger than 0.1 ha are included. Events in my inventory smaller than 0.1 ha are mentioned only by local media and probably received that

119

attention due to a fatality (2010 Altotocagua) or coincidence with another, larger landslide (2013 Pipiripi). Seven small landslides (<0.1 to 0.4 ha) included in the Llojeta sub-basin landslide inventory (Galarza et al., 2005) are not included here as I could not clearly identify or characterize them from available imagery. The times of these small landslides are only constrained to periods ranging from four to 14 years (1983-1996, 1996-2004, or since 2004).

The majority of discrete landslides started as silt compound slides. Some of them transformed into mud flows farther downslope (Table 3.1). Development of distal flows may be more common than thought, because the distal limits of many compound slides were heavily modified before they could be characterized.

The compound slides commonly involve some amount of previously failed material and occur in clusters in parts of the city. Their movement rates range from a few metres per hour to many metres per minute, and in some cases continue to creep long after the cessation of the main period of movement (e.g. 2002 Allpacoma; Hermanns et al., 2012). Due to their frequent occurrence and possible large size, these landslides are responsible for most of the structural damage and displacement of residents in the city. Because most are preceded by signs of instability (Guzmán, 2007a, b; Hermanns et al., 2012) and move slower than a few metres per second, people living on or below them have so far been able to escape injury or death.

The ability of some of landslides La Paz to transform into mobile, potentially long- runout flows elevates the risk to residents far below. These populations may be particularly vulnerable to injury or death due to a lack of forewarning of an impending landslide. Unfortunately, it is difficult to forecast which initial slides might transform into mud flows or debris flows. The six compound slides between 1995 and 2014 that had mud flow components were preceded by 14-day periods with between 0 and over 90 mm of rainfall. Five of them occurred on the southwestern slopes of the city (the Allpacoma and Llojeta sub-basins) and had areas of at least several hectares. In contrast, the single debris flow generated by a compound slide (2007-2008 Patapatani) occurred north of the city on slopes developed in glacial sediments (Chapter 2). The

120

single topple/fall that appears to have generated a debris flow (2008 Hampaturi debris topple/fall) also happened well north of La Paz and following a particularly wet period with 104.9 mm of rain falling over 14 days.

Topples or falls occur from near-vertical cliffs, largely in relatively resistant, largely undisturbed units. Several such landslides have occurred in the Chijini Tuff, which due to its relatively low permeability, may concentrate groundwater flow and seepage onto slopes. Other topples/falls involved glaciofluvial gravel (2010 Altotocagua) or gravel-dominated exposures of the La Paz Formation (2014 Monterani Chico) that have been deeply incised by streams. In some cases, topples/falls have involved overlying weaker units. Although of small size (the exception being the 5-ha 1996 Cotahuma event) and relative uncommon, they are responsible for all recent landslide fatalities in La Paz. Their extremely rapid onset and high velocity prevent people from escaping to safety.

3.5.2. Controlling factors

Slope properties and processes

The geology of the sedimentary sequence underlying the Altiplano is the primary control on the location of landslides in the La Paz basin between 1995 and 2014. All 13 ongoing slow landslides and nearly all discrete landslide events happened on slopes in the southern half of the basin, which are developed in fine-grained sediments of the La Paz Formation. During the 20-year inventory period, there were only two landslides on slopes in the northern half of basin, which are underlain by coarser sediments. Although limited development north of La Paz may partly explain this disparity, it seems to largely reflect distal fining of largely glacier-derived sediments from Cordillera Real northeast of La Paz (Chapter 2).

The occurrence of most of the recent landslides in previously failed sediments suggests that these materials exert a strong secondary control on the locations of failures. Strain weakening (Glastonbury and Fell, 2008) during past failures and limited

121

compaction of fill during their recent terracing probably account for the low strength of slopes underlain by these units. Modern landslides in many urban areas are the result of long-term creep or sudden reactivations within localized parts of underlying ancient landslides (Nilsen and Turner, 1975; Coltorti et al., 1985; Noverraz et al., 1991; Moore et al., 2010 Preuss et al., 2015). Given the correspondence of most modern landslides to previously failed slopes in urban areas east of San Francisco Bay, California, Nilsen and Turner (1975) suggest using maps of ancient landslides to guide future development planning. Although the association of recent (1995-2014) landslide activity in La Paz to ancient landslides helps identify slopes that may likely experience future landslides, the large area of past landslide deposits in the city necessitates consideration of additional factors that further increase the likelihood of failure.

Concentrations of areas experiencing ongoing slow failure within or near the limits of large paleolandslides (11 of 13 cases) raise questions about the modern activities of these deposits. The municipality recognizes that some of these ancient landslides may currently be slowly moving, but does not generally consider most of them to be stability concerns. The occurrence of three-quarters of known landslide events (26 of 34) between 1995 and 2014 at or within 20 m of the margins of past large landslides (Fig. 3.2) suggests that recent landslide activity might be partially driven by concentrated differential motion along the headscarps (extension), toes (compression), and lateral margins (shear) of the paleolandslides. This specific geologic and geomorphic setting appears to be particularly prone to localized landsliding and requires further investigation, including characterization of possible, but generally unconfirmed, modern activity of old large landslide deposits.

One-third (10 of 29) of the compound slides in La Paz between 1995 and 2014 are partially delimited by buried culverted streams. One or both lateral margins of these landslides and, in some cases, the toe of the initial failure zone, closely correspond with these engineered channels, suggesting that this practice may decrease slope stability. This relationship is most obvious in Barrio Retamani (Roberts et al., 2010; Fig. 3.3) where stream culverting has been extensively used since about 1970 and where at least one lateral margin of four of the five documented landslides follows small, buried gullies. In light of the generally poor construction quality of these culverts (Fig. 3.3d), it is likely

122

that some of them have ruptured, possibly as a result of slope creep, allowing storm water to penetrate deeply into the slope, which is typically composed of already weak remobilized units or fill.

In contrast, topples and falls show no association with previously failed deposits, artificial fill, or engineered stream courses. Their occurrence seems to be largely controlled by local topography, making natural cliffs and deep excavations the most likely locations of future failures. The impervious nature of the Chijini Tuff relative to the La Paz Formation overlying it in the southern half of the La Paz basin may play an additional role by focusing water flowing out of the slope. This water could inturn saturate the slope below or initate piping.

Several landslides on the western slopes of La Paz are less than 1 km directly downslope of heavily urbanized parts of El Alto. These failures may have occurred in response to groundwater flowing onto the slopes from beneath the adjacent Altiplano. Urban waste-water drainage from El Alto is a likely contributor to instability on these slopes, although its role in specific failures is unclear. Landslides in this setting include the only fatal landslides between 1995 and 2014 (1996 Cotahuma, 2010 Altotocagua), and some of the largest landslides in recent decades (2002 Allpacoma, 2003 Llojeta), highlighting the need to further evaluate the role of El Alto waste water in destabilizing slopes in western La Paz.

Precipitation triggers

The occurrence of most discrete landslides in the wettest months of the year (December-March) rainy season demonstrates the influence of precipitation on the timing of landslides. The events are concentrated during the final three months (January- March) of the rainy season, even though precipitation totals in February and March are, respectively, similar to and much lower than the December total (Fig. 3.6). It thus appears that high cumulative precipitation is an important factor in triggering landslides, and that some slopes must be sufficiently wetted in order to fail. Landslides are similarly concentrated during the part of the austral summer rainy season in Niterói City, Brazil

123

(Smyth and Royle, 2000), and the second of two closely spaced annual rainy periods in Medellín City, Colombia (Klimeš and Rios Escobar, 2010).

Half of all dated landslides between 1995 and 2014 followed two-week periods with high (>40 mm) cumulative precipitation, even though such totals were recorded on only 17% of days of the 20-year period. However, neither comparison of landslide area and month of occurrence (Fig. 3.6b) nor comparison of landslide area and antecedent 14-day cumulative precipitation (Fig. 3.7) indicates a relationship between event type or magnitude and weather.

More rigorous analysis of precipitation patterns prior to dated landslides could identify meaningful precipitation thresholds for improving the forecasting of landslides of certain types and in certain parts of the city. Further investigation of possible precipitation effects in La Paz should include consideration of additional factors found to influence rainfall-triggering of landslides in other urban areas including precipitation intensity (Ho and Lau, 2010), severity of the previous wet season (Nilsen et al., 1976) (Nilsen and Turner, 1975), type and density of urbanization (Smyth and Royle, 2000), and El Niño–Southern Oscillation (Klimeš and Rios Escobar, 2010). The last of these effects is known to influence interannual precipitation variability along the eastern Altiplano (Garreaud et al., 2003). Varying durations of cumulative rainfall (Garcia-Urquia and Axelsson, 2015) should also be evaluated. In the meantime, slope monitoring in landslide-susceptible areas, particularly those discussed below, should be focused between January and March. General warnings of increased landslide likelihood could be issued following the exceedance of a 14-day precipitation threshold of ~40 mm.

However, further evaluation of precipitation as a trigger of landslides in La Paz should consider the spatial variability of precipitation in the La Paz area. The incursion of moist Amazonian air masses through the Cordillera Real and up the Rio La Paz valley probably results in differences in rainfall in La Paz and the adjacent Altiplano. Convective storms on the Altiplano may not produce as much rainfall in the La Paz and Achocalla basins, but they are a source of groundwater flowing from the western slopes of the two basins.

124

3.5.3. Hazard and risk hot spots and recommendations

Landslide-prone areas

Clustering of recent landslides in several areas of La Paz (Fig. 3.2) indicates areas of elevated local landslide susceptibility. Recent landslides are most common along the east side of Río Orkojahuriah in the area of Villa Armonía and Villa San Antonio (six events – nos. 3, 7, 18, 26, 30, and 33 in the inventory). Recent landslides (nos. 14, 31, and 36), including the ~2 km2 Pampahasi landslide, are clustered one valley to the east within a small area on the west side of Río Irpai. There is also a cluster of recent landslides along slopes at the west end of Aranjuez Ridge (six events – nos. 16, 17, 23, 24, 28, and 42) and ~1.5 km farther north on the slope rising from Río Choqueyapu to the Altiplano (seven events – nos. 2, 10, 19, 34, 37, 38, and 39). All four areas correspond to locations of large paleolandslides and have experienced numerous other failures prior to the inventory period. The clusters recent landslides in the Río Orkojahuriah and Río Irpavi valleys are likely influenced fluvial erosion as they occur adjacent to or along the entire slope above these river channels. In contrast, the clusters of recent landslides the southwest part of La Paz occur well up slope from Río La Paz are likely unrelated to river incision.

Detailed site investigations are necessary in all four areas to better characterize landslide susceptibility in areas of established populations, including the densely urbanized Villa Armonía and Villa San Antonio, the western slope of Río Irpavi that is currently recovering from the 2011 Pampahasi landslide, as well as sites that are likely to be developed in the future, for example slopes at the west end of Aranjuez Ridge. Investigations in these areas should also systematically determine the extent and degree of possible modern movement of paleolandslide deposits. The development of landslide susceptibility maps is a logical starting point for such investigations and should be followed by hazard and risk assessment at key locations.

In contrast, no landslides happened on gently sloping valley bottoms along Ríos Choqueyapu and Orkojahuriah between 1995-2014, and there are ongoing slow landslides at only two locations in these areas (Santa Barbara and Parque Laikacota

125

landslides) (Fig. 3.2). These typically flat areas are underlain by gravelly glaciofluvial outwash and are largely devoid of previously failed material (Dobrovolny, 1962; Anzoleaga et al., 1977; Bles et al., 1977).

The current practice of culverting and burying streams to create developable land must be reevaluated. A starting point should be an inspection of the many existing buried culverts, including those newly built or reinstalled at slopes that experienced landslides between 1995 and 2014. Subsequently, rigorous guidelines should be developed for the construction of future culverts. Regular monitoring of culverts, particularly during the rainy season, is advisable. Rigid concrete culverts are unsuitable for any area showing evidence of large-scale creep. Existing culverts in these areas should be replaced with flexible ones. Culverting and burial of stream courses in undeveloped creeping areas should be avoided.

Secondary hazards

Landslide-dammed lake impoundment and outburst floods

The 2004 Allpacoma landslide generated two landslide dams (Quenta et al., 2007), leading to several possible outburst flood scenarios (Hermanns et al., 2012). Although neither dam burst catastrophically, in part because of gradual drainage of the lower lake by piping (Quenta et al., 2007, Hermanns et al., 2012), the event highlights the potential for some landslides in La Paz to produce upstream impoundments that could drain suddenly, threatening lives and infrastructure downstream. In addition, impoundment upstream of a landslide dam could inundate property. Had the July 2004 landslide dams formed during the rainy season, the likelihood of a devastating outburst flood would have been much greater. Large, relatively long-lived lakes have previously formed in the La Paz basin and include the early Holocene landslide dams immediately south of La Paz (Achocalla earthflow; Dobrovolny, 1962, 1968; Hermanns et al., 2012)

126

and at the northern city limits (Limanpata landslide; Dobrovolny, 19682; Vargas, 1977) (locations in Fig. 3.2).

A worst-case scenario is the formation of a large landslide dam in a narrow channel upstream of a densely populated area during the rainy season. Such an event would pose a major challenge for emergency managers. A large lake could form rapidly and then overtop the dam, followed by breaching and a catastrophic downstream flood. Controlled drainage across the crest of the dam might be possible, but would have to be effected very quickly following the landslide. Additionally, any dam would likely comprise fine-grained, erosion-prone sediment that would greatly increase the chance of dam failure whether by seepage (piping in the case of the lower 2004 Allpacoma landslide dam; Quenta et al., 2007) or overtopping (cf. Costa and Schuster, 1988).

Debris fluidization

Based on the description by Hardy (2009), the 2008 Hampaturi landslide was triggered by overflow of the Hampaturi reservoir. The overflowing water entrained and mobilized debris downstream, likely partially by fluvial undercutting, producing a high- velocity debris flow in the valley below. Damage in the local rural community and disruption of the water distribution system in La Paz 10 km downstream highlight the risk that such events pose. Several other reservoirs located a short distance upvalley of the city and opposite the Limanpata landslide could pose similar, as-yet unevaluated hazards.

Water shortages

Disruption of the water distribution system can extend the impacts of landslides well beyond their geomorphic limits, affecting large additional populations. Although less

2 Dobrovolny (1968) believed the radiocarbon age he acquired on plant detritus from lacustrine sediments in the Río Kaluyo valley (9200 ± 250 yr BP; details in Rubin and Alexander, 1958) constrained the age of the Achocalla earthflow. However, the lake was impounded behind the Limanpata landslide (cf. Vargas, 1977), not an end moraine as Dobrovolny (1968, p. 133) thought. The age therefore constrains the timing of the Limanpata landslide, not the Achocalla earthflow.

127

acute than loss of lives, homes, or livelihoods, this impact represents an important element of hazard planning in La Paz due to the large number of residents affected. Such water losses also impact businesses and medical services (Hardy, 2008). Twice in recent years, landslides severed water mains servicing a large portion of La Paz (over a quarter million residents in each event). One of the landslides (2008 Hampaturi) was a relatively minor slope failure. The other event (2011 Pampahasi) was the largest landslide in La Paz in centuries and could be considered a rare event. However, geologic and historic evidence of repeated failure at the same location suggests that routing of primary water mains across this slope should be reconsidered. In mid-2011, the ruptured water main at Pamphasi was rebuilt at the same location. To improve resiliency, redundancy to water infrastructure should be considered.

3.6. Conclusions

La Paz is one of the most landslide-prone cities in the world. Between 1995 and 2014, the city experienced 43 documented landslides, an average of more than one event every six months. The landslides ranged in size from several square metres (2013 Pipiripi; La Razón, 2013) to nearly 150 ha (2011 Pampahasi). In addition, there was ongoing slow movement of slopes over this period in at least 13 locations. Together, these 56 landslides displaced nearly 12,000 people, damaged or destroyed over 1750 homes, and claimed between 21 and 28 lives. Two of them interrupted the supply of potable water to 500,000 residents. The landslides caused large, but non-quantified economic losses and had far-reaching secondary impacts.

The most common discrete landslides in La Paz are rapid to very rapid compound slides often involving previously failed material. Several of these landslides transformed into mud flows or debris flows, enhancing their mobility and increasing their runout by up to several times the length of the initial slide. The largest landslide during the inventory period (1995-2014), and in fact the largest in the area in at least the past 400 years, emphasizes the potential complexity of failures in the area. The 2011 Pampahasi megalandslide (~2 km2) was largely a slide with both rotational and

128

translational components, similar to the typical failure type in La Paz. However, it also included, localized toppling and flows in the steep, water-saturated headscarp region. Owning to its size and its location near the city centre, it was the also the most destructive landslide since the sixteenth century, displacing ~6000 people and disrupting the water access of hundreds of thousands more residents for six months (Hermanns et al., 2012, Aguilar, 2013).

The other discrete landslide events were falls and topples that almost exclusively occurred in near-vertical cliffs of relatively strong units (glaciofluvial sediments or cemented tuff) or intact steep exposures of fine-grained clastic sediments of the La Paz Formation. These extremely rapid landslides account for all of the fatalities from slope failures in La Paz in the past few decades.

Landslides are most common in the rainy season, particularly during the latter part of it (January to March). A consideration of antecedent rainfall prior to each of the dated events emphasises the influence of cumulative precipitation during the wet season. However, no strong relationship is apparent between meteorological conditions and landslide magnitude or type.

The La Paz landslide inventory highlights several causative factors that can aid mitigation of landslide hazard and risk in the city and provide guidance for additional investigations. Landslide susceptibility mapping, such as that conducted for the Allpacoma and Llojeta sub-basins, should be carried out for all of La Paz. Future research should include monitoring of paleolandslide deposits for possible current activity and more detailed investigation of the potential influence of geologic boundaries between intact and large previously failed units. Appropriate cumulative precipitation thresholds should be identified to trigger heightened vigilance for signs of instability, particularly in areas with the highest likelihood of landslides, by both the local government and the general public. Engineering of stream courses, regardless of their size, by culverting and burial should be reconsidered. The volume and transport paths of waste water from El Alto reaching the western slopes of Río Choqueyapu should be further investigated and its influence on local slope stability evaluated. Specific landslide

129

scenarios should be considered to understand the threats they pose, including the formation and failure of landslide dams, debris fluidization and consequent debris flows, and possible scenarios involving water main rupture.

130

Chapter 4. Recent slope creep at La Paz, Bolivia, from advanced spaceborne RADAR interferometry

4.1. Introduction

Mitigating losses in landslide-prone settings is complicated by the extent and number of problem areas to be considered, thus requiring prioritization of slopes that are most hazardous (Anbalagan, 1992; Guzzetti et al., 1999) and particularly those posing the highest risk (Dai et al., 2002). Remote sensing is useful in identifying and prioritizing target areas due to its ability to provide spatially continuous data over extensive areas (Mantovani et al. 1996; Metternichtet al., 2005). It is also valuable for helping understand processes and rates of landscape evolution. Technological improvements in remote- sensing systems in recent decades have increased the quality and diversity of data they can provide for landslide studies (Singhroy, 2009; Petley, 2012). Today, a wide variety of terrestrial, aerial, and spaceborne remote sensing techniques is available to detect and characterize landslide movement (Delacourt et al., 2007). These approaches are most beneficial when combined with ground-based investigations.

Slow ground deformation measured using spaceborne interferometric synthetic aperture RADAR (InSAR) provides opportunities to document ground instability over extensive areas. The technique can facilitate landslide identification (Del Ventisette et al., 2014; Bianchini et al. 2015; Piacentini et al., 2015; Oliverira et al., 2015), improve mapping of known landslides (Squarzoni et al. 2003; Peyret et al., 2008; Osmundsen et al., 2009; Yin et al., 2010; Henderson et al., 2011; Greif and Vlcko, 2012; Tofani et al., 2013; Chen et al., 2014; Del Ventisette et al., 2014; Booth et al., 2015; Novellino et al., 2015; Oliverira et al., 2015), assist in kinematics characterization (Fruneau et al., 1996;

131

Peyret et al., 2008; Necsoiu et al., 2014; Castaldo et al., 2015; Sharma et al., 2015), characterize temporal changes in movement that facilitate identification of controlling factors (Fruneau et al., 1996; Rott et al., 1999; Hilley et al., 2004; Tantianuparp et al., 2013; Tofani et al., 2013; Milillo et al., 2014; Necsoiu et al., 2014; Teshebaeva et al., 2015), and support preliminary hazard analysis (Rott and Nagler, 2006; Bianchini et al., 2012; Lu et al., 2014). Spaceborne InSAR also can assist in predicting catastrophic slope failure as part of monitoring campaigns aimed at characterising long-term failure behaviour (Petley et al., 2002; Singhroy, 2009; Mazzanti et al., 2011).

The La Paz area is particularly well suited to landslide investigations using spaceborne InSAR due in part to ubiquitous slope instability and in part to the lack of dense vegetation that would impede signal recovery. Large ancient landslide deposits (Dobrovolny, 1956, 1968; Anzoleaga et al., 1977; Hermanns et al., 2012) and historic landslide events (O’Hare and Rivas, 2005; Guzmán 2007a,b; Roberts et al., 2010; Hermanns et al., 2012) attest to long-term and continuing instability in the city and the adjacent Achocalla basin. Ground fissures and damage to urban infrastructure suggest recent slope creep at several landslide-prone localities (Hermanns et al., 2012). However, the extent, rate, timing, and nature of slope movements are poorly constrained.

Additionally, local conditions at La Paz help reduce some limitations common to spaceborne InSAR. Close alignment of the Río La Paz valley system with the roughly north- or south-directed flight paths of typical polar-orbiting SAR sensors, such as RADARSAT-2, results in typical landslide movement directions that are roughly parallel to the sensor’s orbit-perpendicular look direction. This orientation maximizes the likelihood that slower slope movements will be detected. Due to aridity and elevation, La Paz and the adjacent Altiplano have sparse (typically grassland) vegetation cover and are rarely snow-covered, minimizing land-surface changes on the scale of the RADAR wavelength and maximizing the quality of InSAR distributed targets (cf. Pritchard et al. 2013). The extensive built environment of urban areas in and around La Paz provides large numbers of ‘corner reflectors’ (Hilley et al., 2004; Cinga et al., 2014, 2015) that can be individually monitored and help to overcome atmospheric artifacts and phase decorrelation (Colesanti et al., 2003).

132

I apply state-of-the-art InSAR processing – Homogeneous Distributed Scatterer InSAR (HDS-InSAR; Eppler and Rabus, 2011) – to fine beam mode RADARSAT-2 scenes acquired over La Paz between September 2008 and the December 2011 to document the extent and character of landslide-related ground motion in the city. This processing technique increases the spatial density of displacement histories by utilizing both coherent distributed scatterers (e.g. pavement, agricultural land, and natural slopes) and persistent point scatterers (primarily built structures, fortuitously oriented rock exposures), and by applying adaptive filtering to preserve spatial resolution as much as possible while optimally suppressing the noise from surface decorrelation (Eppler and Rabus, 2011; Rabus et al, 2012). HDS-InSAR improves the spatial density of deformation characterization compared to typical InSAR techniques. It additionally includes several steps to remove other phase components, particularly the error from atmospheric water vapor variations.

In this chapter, I illustrate the use and power of the method by focusing on the district of San Antonio, which is one of the most landslide-affected areas in the city. It includes the highest concentration of twenty-first century landslides within a single neighbourhood as well as the largest landslide in the La Paz area in at least 400 years. In view of the magnitude and importance of the latter landslide, which occurred about three-quarters of the way through the RADAR acquisition period, I also apply HDS- InSAR independently to RADAR scenes acquired prior to and following its occurrence. Ground motion detected in this area and in other areas of InSAR coverage adds to knowledge of the spatio-temporal distribution of landslides in and around La Paz. Comparison of HDS-InSAR results with the local geology, paleolandslides, urban development, and recent failures documented though landslide inventories provides new insights into the controls on large-scale instability as well as the importance of slow, creep-like movements in triggering more rapid, higher risk failures. These findings provide knowledge that can be used to reduce social, physical, and economic impacts and losses in one of the most landslide-prone urban areas of the world.

133

4.2. Background

4.2.1. Application of InSAR in landslide investigation

Landslide-related InSAR applications utilize repeat-pass interferometry to characterize slow surface displacement. Differentiating the phase of sequential SAR images acquired from slightly different positions removes spatially random phase resulting from sub-pixel backscatter, provided that the surface change at the wavelength-scale of the SAR in minimal. The remaining phase difference represents the net effects of atmospheric heterogeneities, sensor noise, topography, and ground motion (Massonnet and Feigl, 1998). Removal of the first three effects through image processing isolates the line-of-sight component of landsliding.

Differential InSAR (D-InSAR) produces spatially continuous representations of line-of-sight terrain motion (interferograms; Massonnet and Feigl, 1998) in which each pixel records a complex mix of backscatter responses from a ground surface area composed of multiple targets. Consequently, the technique is best suited for homogeneous surface covers that uniformly scatter incident microwaves. Temporal changes in atmospheric and ground-surface conditions result in decorrelation between scenes, which generates noise in the interferogram (Zebker and Villasenor 1992) and hinders measurement of ground surface movement. Surface decorrelation is particularly pronounced and variable in mountainous terrain, where landslide investigations are commonly conducted. Vertical stratification of water vapour increases atmospheric artifacts (Hanssen, 2001). Additionally, climatic variability, slope aspect, and topographic shading increase the spatial-temporal variability of snow cover, soil moisture (Williams et al., 2009), and vegetation, all of which lead to decorrelation, phase bias, or both. Surface disturbance of landslide masses due to differential motion during transport may further degrade coherence (Massonnet and Feigl, 1998). Because the degree of change generally increases with time, decorrelation particularly limits detection of slow landslides; long time separation is required between acquisition pairs to quantify small

134

displacement magnitudes. Additionally, D-InSAR does not provide displacement time series that are afforded by other InSAR techniques that are useful in landslide analysis.

Persistent scatter InSAR (PS-InSAR; Ferretti et al., 2001) mitigates decorrelation by identifying and tracking pixels with backscatter responses dominated by stable-phase targets that produce strong, localized backscatter because of their angular geometries (Colesanti et al., 2003; Kampes, 2006). These pixels lack temporal speckle, enabling displacement measurement through a series of SAR images even if surrounding distributed surface scattering is decorrelated (Kampes, 2006). PS-InSAR thus provides improved characterization of landslide deformation histories (Colesanti and Wasowski, 2006). Favourably oriented rock outcrops serve as natural persistent scatterers, but their typically low spatial density limits the application of PS-InSAR in natural environments to areas with minimal vegetation (Dehls et al., 2002; Colesanti et al., 2003; Bianchini et al., 2013; Tantianuparpe et al., 2013) and near-surface bedrock (Colesanti and Wasowski, 2006; Piacentini et al., 2015), or to abundant large debris from rock slope failures (Dehls et al., 2002; Notti et al., 2010). Corner reflectors can be installed to provide artificial persistent scatterers (Frose et al., 2008; Fu et al., 2010), but they limit characterization of movement to a small number of locations and must be precisely oriented to guarantee a strong signal reflection. Thus, PS-InSAR is best suited to infrastructure features or developed landscapes (Hilley et al., 2004). Even there, however, displacement records are often lacking between built features, particularly in rural areas (Colesanti and Wasowski, 2006; Greif and Vlcko, 2012) where motions records may be based on only a few permanent scatterers per landslide (Oliverira et al., 2014). Depending on the density of appropriate point targets, PS-InSAR may provide limited results in key areas of a landslide (Necsoiu et al., 2014) or even limit detection of slope movement throughout entire regions (Cigna et al., 2015; Comerci et al, 2015). For instance, Del Ventisette et al. (2014) used a variant of the PS-inSAR technique to identify coherent point targets on only a fraction of known landslides covering 40% of landslide-affected terrain in the Italian Alps.

Small Baseline Subset InSAR (SBAS-InSAR; Berardino et al., 2002, 2003) employs an alternative approach to deal with decorrelation. The technique generates interferograms for subsets of scenes with short interferometric baselines to reduce

135

spatial decorrelation and topographic error. It conducts network analysis using least squares inversion to link the interferograms computed for separate subsets, producing a time-series that considers the largest possible number of acquisitions. The ability of SBAS-InSAR to characterize strongly non-linear motion (Necsoiu et al., 2014) makes it an especially attractive approach for investigating the temporal evolution of landslides. It has the added benefit of maintaining high-continuity coverage of displacement measurement and thus provides better spatial characterization of landslide motion than PS-InSAR (Lauknes et al., 2010). SBAS-InSAR is a particularly important advance for large-scale, spatially correlated motion such as tectonic or volcanic deformation and ground subsidence (Berardino et al., 2003). However, the indiscriminate spatial filtering it employs to improve signal-to-noise ratio comes at the expense of reduced spatial resolution. Its enhanced spatial density relative to PS-InSAR improves delineation of active landslide boundaries (Chen et al, 2014), but the technique is still limited by its inability to represent small-scale spatial variability of ground motion, which is typical of most landslides and may be particularly pronounced in the case of small landslides. Additionally, SBAS-InSAR is optimized for distributed scatterers and does not exploit the high-density stable scatterers present in developed areas and specific types of natural terrain.

A new generation of advanced InSAR algorithms combines strengths of the PS- InSAR and SBAS-InSAR approaches by utilizing both stable point scatterers and the coherent diffuse scatterers between them (Lanari et al., 2004; Parizzi and Brcic, 2011), thereby increasing the spatial density of displacement records. Greater ground-target density improves the likelihood of resolving small landslides and assists in interpreting failure mechanisms. Some approaches (Hooper, 2008) are optimized for deformation that is correlated locally in space. Techniques that can address the full spectrum, from locally correlated to uncorrelated deformation (isolated point scatterers that move differently with respect to the background), such as SqueeSARTM (Ferretti et al., 2011) and HDS-InSAR (Eppler and Rabus, 2011; Rabus et al., 2012), offer the best opportunity for detailed spatial characterization of landslides.

136

4.2.2. La Paz and surrounding area

Physiography and geology relevant to slope stability

La Paz is the administrative capital of Bolivia and together with its adjacent satellite city of El Alto form the country’s largest population and economic centre. The two cities constitute one of the highest metropolitan areas in the world, located at the east edge of the Altiplano plateau between 3200 and 4300 m asl (Fig. 4.1), adjacent to the high Cordillera Real with peaks higher than 5500 m asl. La Paz covers the floors and slopes of a valley system incised up to nearly 1 km into a late Cenozoic fill underlying the Altiplano and older basement rock of the Cordillera Real. El Alto sprawls westward across the Altiplano from the lip of the valley in which La Paz is situated.

Topography exerts a strong control on local climate. The La Paz area is in the rain shadow of the high Cordillera Real just to the east (Bookhagen and Strecker, 2008), which intercepts moisture sourced in the tropical Atlantic Ocean (Vuille et al., 2003). Some weather systems penetrate the Cordillera Real along the Río La Paz valley ~60 km southeast of the La Paz. A similar effect has been noted in Zongo Valley, 30 km north of La Paz, where precipitation falls progressively later in the day as the moist air masses move up the valley to higher elevations (Sicart et al., 2003). High-altitude air masses moving over the Cordillera Oriental and across the Altiplano produce intense convective storms (Garreaud et al., 2003). About 80% of the annual precipitation from both sources falls mainly during the austral summer (South American summer monsoon; December-March) (Zhou and Lau, 1998; Garreaud et al., 2003; Sicart et al., 2003). This pronounced seasonality results from an alternation in tropospheric circulation resulting in easterly flow in summer and westerly flow that impedes advection of moist air masses from the Amazon basin in winter (Vuille, 1999; Garreaud et al., 2003; Vuille et al., 2003).

The late Cenozoic sediment sequence underlying the Altiplano comprises more than 1 km of continental clastic sediments shed from the Cordillera Real to the east (Ahlfeld, 1954a, 1946; Dobrovolny, 1962). Key marker volcanic beds, likely sourced from the Cordillera Occidental at the western limit of the Altiplano, occur within the sedimentary sequence (Dobrovolny, 1962; Lavenu et al., 1989; Marshall et al., 1992).

137

Deposition of the sediment sequence began before eruption of the Cota Cota tuff ca. 5.5 Ma ago (Lavenu et al., 1989; Servant et al., 1989). This tuff is exposed in the southernmost part of the La Paz basin. Fine-grained alluvial and lacustrine sediments (lower La Paz Formation; Dobrovolny, 1962) dominate the lower half of the sediment sequence. These Miocene to early Pliocene sediments are exposed south of the city centre, but incision has not yet reached their equivalents north of the city. Folded and faulted metasedimentary and metamorphic bedrock of Paleozoic age underlies the La Paz Formation south of the city and includes Silurian to Devonian quartzite and shale (Sica-Sica Formation) and Cretaceous conglomerate and sandstone (Aranjuez Formation) (Bles et al., 1977).

The upper half of the late Cenozoic fill in the La Paz basin records progressive expansion of glaciers in the Cordillera Real. Late Pliocene and Early Pleistocene glacial deposits dominate the upper half of the sediment sequence in areas near the Cordillera Real (Chapter 2). The upper La Paz Formation (Chapter 2) of the La Paz and Achocalla basins and the glacial sequence nearer the cordillera include the laterally persistent, latest Pliocene Chijini Tuff (ca. 2.74 Ma: Lavenu et al., 1989; Marshall et al., 1992; Chapter 2), which in combination with magnetostratigraphy (Thouveny and Servant, 1989; Chapter 2), allows lateral correlation of units within the basin. Sediment continued to accumulate until the Early Pleistocene when the modern Altiplano surface formed (Chapter 2).

The basin fill sequence aggraded relatively rapidly, was never deeply buried, and was rapidly incised, leaving nonlithified sediment in a constant state of instability (Chapter 2). Fining of sediments away from the Cordillera Real (Chapter 2) and the typically limited lateral continuity of beds and lenses within the sequence (Ahlfeld 1946; Dobrovolny, 1962; Bles et al., 1977) result in pronounced differences in geotechnical properties and highly variable permeability. The extent and recent activity of faults cutting the fill sequence (Lavenu, 1977; Lavenu et al., 2000; Chapter 2) are poorly constrained and thus play an unknown role in slope stability. Some faults ceased to be active in the Early Pleistocene (Chapter2). However, the Kenko fault at the southern limit of El Alto shows evidence of Late Pleistocene activity (Ramírez et al., 2007). The extent of the Kenko structure across the Altiplano and its en echelon form (Dobrovonly, 1962)

138

Figure 4.1. Setting of La Paz and the surrounding area showing the extent of urban development and the distribution of landslides.

139

a) Location of study region within South America. b) Physiographiy of the Central Andes. c) Physiography and documented mass movements at La Paz. Terrain from ASTER DEM. Paleolandslides from Anzoleaga et al. (1977). Large landslides north of La Paz and recent landslides from Chapter 3. Boxs show the extent of the case study area in the western part of the San Antonio district (solid) and the entire InSAR-processed area (dashed).

suggest a tectonic origin (Lavenu, 1978; Lavenu et al., 2000). As the scarps along the only investigated segments of this structure show normal displacements with no evidence for lateral movements (Ramírez et al., 2007, 2009), the most recent activity may have involved extensional gravitational movement on pre-existing faults, as suggested by Lavenu (1978).

Mass movements

Large prehistoric landslides have been recognized in the La Paz and Achocalla basins for well over a century (Conway, 1901; Troll and Finterwalter, 1935). Systematic geologic mapping in the mid-twentieth century (Dobrovolny, 1956, 1962; Anzoleaga et al., 1977) shows that large paleolandslide deposits form much of the slopes in La Paz and play an important role in formation of the modern landscape. A few of these paleolandslides have been characterized in some detail (Dobrovolny et al., 1968 [Achocalla earth flow]; Quenta and Calle, 2005 [Pampahasi paleolandslide deposit]), but thorough investigations have not been conducted. Localized damage to infrastructure (Dobrovolny, 1962; Quenta and Calle, 2005; Fig. 4.2; Chapter 3) is attributable to creep of slopes throughout the city, but typically near the margins of paleo-landside deposits (Chapter 3). Furthermore, coincidence of most historic failures with paleolandslide deposits, particularly their margins, suggests ongoing activity of ancient landslides in the La Paz area (Chapter 3).

The first historically recorded landslide in the La Paz basin occurred in 1582 and, with a reported death toll of 2000 (Santa Cruz, 1942; Sanjinés, 1948), is the deadliest landslide in the city’s history. Frequent failures are recorded throughout the twentieth and twenty-first centuries (Guzmán, 2007a,b; O’Hare and Rivas, 2005; Roberts et al., 2010; Hermanns et al., 2012); an average of one landslide happened every six months between 1995 and 2014 (Chapter 3). Their sizes range from hundreds of cubic metres to

140

Figure 4.2. Evidence of recent instability in La Paz.

141

a) Offset of road in La Cervezaria just behind the headscarp of the Villa Salomé paleolandslide and 500 m upslope of the 2009 Villa Salomé landslide (2010 photo). b) Cracked wall a home just behind the headscarp of the Villa Salomé paleolandslide and 500 m upslope of the 2009 Villa Salomé landslide (left of the road in panel a with time-stamped plaster used by municipal staff to monitor movement (2010 photo). c) Damage to deck of railing of the pedestrian bridge in Parque Laikacota (Distrito Centro) built in 2009 and closed in 2010 due to damage (2010 photo). d) Warped road and sidewalk at the lateral margin of the Santa Barbara landslide (2010 photo). e) Cracked interior wall of a home located at the toe of paleolandslide deposit in Las Lomas (Distrito Sur) with date-stamped plaster tags for monitoring (photo 2010). f) Separation of floor in the same home as in panel e (photo 2010). g) Ruptured wall in in Bajo Seguencoma (Distrito Sur) in an area of several damaged houses (photo 2012). h) Ground fissure within the Calle 33 landslide (Distrito Sur) located at the margin of the Cotacota paleolandslide deposit (2012 photo).

millions of cubic metres (Chapter 3); at least five events, all in the twenty-first century, were larger than 1 Mm3 (Hermanns et al., 2012). The most recent large landslide (2011 Pampahasi landslide) affected ~2 km2 of suburban development and displaced ~6000 people (Hermanns et al., 2012, Aguilar, 2013).

Most landslides initiate as compound slides involving mud or debris. Some transform into debris flows or mudflows as they move downslope (Chapter 3). They move at rates ranging from several metres per hour to many metres per minute (rapid to very rapid following the classification of Cruden and Varnes, 1996, and Hungr et al., 2014). Some landslides are much slower, moving at rates of several metres per year to one metre or so per hour (slow to moderate). Consequently, impacts for most historic events are in the form of property loss and infrastructure damage, with limited life loss (O’Hare and Rivas, 2005; Mobarec et al., 2008; Quenta et al., 2008; Roberts et al., 2010; Hermanns et al., 2012; Chapter 3). Most inhabitants are able to escape the failing area or are transported on the landslide with no injury. Infrequent topples of tuff (O’Hare and Rivas, 2005) or falls of weakly cemented sediments have caused all of the deaths due to landslides in La Paz in recent decades (Chapter 3).

4.2.3. San Antonino case study area

The district of San Antonio is located directly east of La Paz’s central district (Distrito Centro) and covers the area between Ríos Orkojahuira and Irpavi, which are separated by the Pampahasi Plateau (Fig. 4.3). The case study area considered here

142

falls largely within the southern half of the San Antonio district (neighbourhoods of Villa San Antonio, Villa Armonía, San Isidro, Pampahasi, Callapa, and Irpavi), but also includes the neighbourhood of Miraflores in the eastern part of Distrito Centro (Fig.4.3). The area was urbanized, from west to east, over the course of the twentieth century. There was little development in the area as late as the early 1940s (Santa Cruz, 1942, pp. 11-12), but Miraflores underwent rapid urbanization thereafter and, by 1955, was covered with homes and roads. Most of the modern road network in Villa San Antonio and Villa Armonía was developed by about 1975, and for several decades thereafter those areas were developed and densified. This development included reclamation of gullied and previously failed slopes (Roberts et al., 2010; Chapter 3). By about 2000, most gullies were buried and culverted (Roberts et al., 2010), including Río Orkojahuira, over which Avenida Zabaleta was build (Guzman, 2007b). The Pampahasi plateau was not developed until the late twentieth century. Slopes descending east from the plateau to Río Irpavi were only sparsely developed in 1990 (Scanvic and Girault, 1989) and were still dominated by farmland in 1994 air photos. However, they supported increasing populations in the first decade of the twenty-first century.

Miraflores is underlain by gravelly glaciofluvial outwash (Dobrovolny, 1962; Anzoleaga et al., 1977; Bles et al., 1977) that provides some of the most stable building foundations in the city. East of Río Orkojahuira, slopes rise ~250 m to the Pampahasi plateau (3700-4000 m asl) and then descend again on the other side to Río Irpavi. The slopes on both sides of the plateau are mainly ancient landslide deposits of remobilized La Paz Formation that have been locally reactivated in numerous smaller landslides and mantled by alluvial fan deposits (Anzoleaga et al., 1977). The presence of intact La Paz Formation in the gulley between Villa San Antonio and Villa Armonía (Anzoleaga et al., 1977) suggests the presence of two separate large paleolandslides in the study area (Fig. 4.3). The sliding surface of the Villa Armonía paleolandslide is exposed in the south wall of this gulley (Fig. 4.4a) and shows well developed slickensides oriented roughly parallel to the slope (Fig. 4.4b). North of the gulley, sediment just beyond the toe of the Villa San Antonio paleolandslide deposit has been compressed and raised (Fig. 4.4c), suggesting recent movement. Damage to infrastructure throughout the area further indicates on-going deformation (Fig. 4.4d,e). The shape and morphology of paleolandslide deposits on the west side of the Río Irpavi valley suggest separate

143

source areas for the Pampahasi paleolandslide and another large ancient landslide farther north (Fig. 4.3). The former reached the valley bottom and deflects Río Irpavi eastward, whereas the latter terminated partway down the slope. Slow subsidence of the eastern margin of the Pampahasi plateau directly upslope of the paleolandslide north of the Pampahasi paleolandslide (Fig. 4.3a,b) suggests it has been active throughout the twenty-first century.

Figure 4.3. Overview of the case-study area, southern San Antonio district. Base image is a June 2015 Quickbird image viewed in GoogleEarth™. Paleolandslides are shown in blue (Anzoleaga et al., 1977). The limits and timing of landslide events between 1995 and 2014 are indicated by white dotted lines and associated labels (Chapter 3).

As previously noted by Scanvic and Girault (1989: pg. 19), urban densification in these areas is concerning given the existence of large ancient landslides and potential for future instability. The area directly east of Río Orkojahuira has been the site of more recent landslides than any other part of La Paz. Seven documented failures have happened since 1997 (Chapter 3); five of the seven – in 2000, 2003, 2007, 2009 and 2010 – occurred within Barrio Retamani, a densely populated neighbourhood (~300 m x

144

Figure 4.4. Features of prehistoric and recent instability in the western San Antonio district.

145

a) Sliding surface exposed at the right lateral margin of the Villa Armonía paleolandslide (2012 photo; people for scale). b) Slickensides along the Villa Armonía paleolandslide sliding surface (2012 photo; location shown in panel a) oriented roughly parallel to the slope dip direction. c) Toe of the Villa San Antonio paleolandslide with a thin zone of bulldozed sediment at the base of the exposure (2012 photo; notebook is 20 cm long). d) Tension cracks behind the head scarp (~10 m left of photo) of the 1997 San Isidro landslide (2010 photo). e) Subsidence and cracking of a new sidewalk across the upper portion of the 2009 Retamani I landslide (2012 photo; notebook is 20 cm long). f) Overview of the 2009 Retamani landslide three months after failure showing building density of the San Antonio area (2009 photo). g) Overview of Barrio Retamani showing the locations of twenty-first-century landslides (2010 photo; labeled arrows indicate year of event).

500 m) along the west bank of Río Orkojahuira (Fig. 4.4f,g). All seven failures involved rotational and translational sliding of previously failed sediments at rates of a few metres per hour to a few metres per second (Chapter 3), and are thus typical of recent landslides in the La Paz basin. The 1997 San Isidro landslide (Guzmán, 2007a) occurred within the body of the Villa Aromína paleolandslide deposit. The 1997 Cuarto Centenario landslide (Guzmán, 2007b) and the Barrio Retamani landslides happened, respectively, at the toes of the Villa Armonía and Villa San Antonio paleolandslides. Three landslides occurred on the slopes between the Pampahasi plateau and Río Irpavi in the twenty-first century, all dominantly in paleolandslide deposits (Chapter 3; Fig. 4.3). The last of these – the 2011 Pampahasi landslide (Hermanns et al., 2012; Aguilar, 2013) – was the largest and most damaging landslide in La Paz since at least the sixteenth century. It remobilized a large part of the Pampahasi paleolandslide, but its head scarp cut several tens of metres into intact La Paz Formation underlying the Pampahasi plateau. Repeated failure along this near-vertical scarp since 2011 has resulted in retrogression of the head of the landslide by several city blocks. Coincidence of the historic landslides with previously failed areas, particularly the margins of old large paleolandslide deposits are typical of historic landslide occurrence throughout the La Paz basin (Fig. 4.1). This spatial association may reflect the reduction of material strength to residual strength during the paleolandslides (Glastonbury and Fell, 2008), but the typical position at the lateral margins and toes of these old deposits also suggests periodic or on-going activity (Chapter 3).

146

4.3. InSAR processing

4.3.1. Scene details

RADARSAT-2 scenes were acquired with parameters selected to optimize landslide detection in the vicinity of La Paz (Table 4.1). RADARSAT-2’s fine beam mode provides high spatial resolution over the La Paz and Achocalla basins. The incidence angle (36.3°) and pass direction (ascending) assure acquisition of high-quality data along the westernmost (east-facing) slopes of the Río La Paz drainage system where large recent failures are most common (Chapter 3; Hermanns et al., 2012), while preventing layover or substantial foreshortening of opposing valley slopes. The repeat period (24 days) and wavelength (5.6 cm) of RADARSAT-2 make it well suited for detection of ground motion on the order of several millimetres to several decimetres per year, as was suspected on deforming slopes in the area. Forty-four of the 51 possible scenes were acquired during the 40-month imaging period; the seven missing scenes include a five-scene gap near the start of the time series (4th to 8th planned acquisitions) and two later one-scene gaps (Table H1).

Table 4.1. RADARSAT-2 scene details.

Incidence angle 36.3° Slant-range pixel spacing (m) 4.73 m Azimuth pixel spacing (m) 5.05 m Ground-range pixel spacing (m) 7.99 m a Scene footprint 54.9 x 53.3 km a Processed area 23.0 x 21.2 km a Case-study area 2.0 x 3.0 km

a Dimensions formatted as range x azimuth.

147

4.3.2. Processing chain

HDS-InSAR is a stack-based method developed by the Research and Development Group of MDA Corporation, Richmond, Canada, to characterize the coherent phase time series of both persistent point scatterers and coherent distributed scatterers (Eppler and Rabus, 2011). Specific steps in the chain remove differential phase contributions from Doppler effects, topographic distortion, baseline error, stratified atmosphere, temporal atmospheric heterogeneities, and thermal expansion, thereby isolating phase differences due to line-of-sight ground motion (e.g. Fig. 4.5). Adaptive filtering improves not only the spatial quality of the phase, but also RADAR echo intensity. The processing chain applied to the La Paz stack is an updated version of the prototype described by Rabus et al. (2012). I conducted all processing (Appendix H) at MDA Corporation in Richmond using in-house-developed programs that in some cases incorporate GAMMA scripts.

Pre-processing included resampling of single-look complex images to align with a master image (12 July 2010) and production of multi-looked images with approximately square ground spacing. A 3-arc-second Shuttle RADAR Topographic Mission (SRTM3 V2; Farr, 2007) reference digital terrain model (DTM) was projected into SAR coordinates. After pre-processing and D-InSAR analysis of the full scene (~55 x 53 km), HDS-InSAR processing was applied to a subset area (22.8 x 21.1 km) including the San Antonio study area (~ 2 x 3 km) (Table 4.1). I conducted the same processing for two temporal sub-sets of the stack of RADAR scenes spanning the period before (scenes 1 to 32) and after (scenes 33 to 44) the 2011 Pampahasi landslide.

The revised HDS-InSAR processing chain used here models and removes long- range atmospheric heterogeneities and the topography-correlated phase prior to adaptive filtering instead of after, as done by Rabus et al. (2012) in an earlier proto-type of HDS-InSAR. A continuously weighted, spatially adaptive filter then defines spectrally similar pixel clusters (HDS neighbourhoods) within a rectangular minimum-candidate area based on amplitude time series. The filter was applied to a network of interferograms to generate differential phase and coherence for each HDS (Eppler and

148

Rabus, 2011). Parametric corrections (high-frequency topographic error, linear deformation, thermal dilation, and residual phase time series) were extracted for high- quality HDS via temporal coherence maximization. I applied an optional phase- demodulation step to the La Paz stack to identify strong deformation trends from a subset of high-quality interferograms. I removed the trend from the entire network, thereby further flattening the phase prior to spatial unwrapping, and later re-added this removed trend to phase-unwrapped interferograms. Phase unwrapping utilizes a two-dimensional Minimum Cost Function (MCF; Costantini, 1998). After estimating and correcting residuals for each phase component, I inverted network interferograms by singular value decomposition (SVD) to produce line-of-sight and vertical deformation time series. Finally, I georeferenced temporally smoothed HDS time series and interpolated linear deformation maps.

4.4. Results

4.4.1. Homogeneous distributed scatterer density and quality

The entire processed area contains 9,162,500 HDS – stable reflections from both point scatterers and small neighbourhoods of spatially uniform coherent distributed scatterers. The average spatial density of motion records is 19,050/km2. The HDS have a regular gridded spacing (Fig. 4.5) due to the rectangular filtering kernel. Their density is generally greatest on east-facing slopes, intermediate on level terrain, and lowest on west-facing slopes, reflecting the influence of viewing geometry and topography on pixel ground spacing. The density is greater in areas of urban development and grassland than in areas of water, major landscape change, eucalyptus forest, some agricultural areas, and the middle parts of some deforming areas (Fig. 4.6).

149

Figure 4.5. Typical density and spacing of HDS-InSAR results exemplified by a part of the densely urbanized Santa Barbara area in Distrito Centro. Base image is a June 2015 Quickbird image viewed in GoogleEarth™. a) Satellite image showing urbanized land cover without InSAR data. b) HDS point data. c) Linear displacement map interpolated form HDS points. Displacement rates are in the line-of-sight direction. Deforming area is part of the head scarp of the Santa Barbara landslide.

4.4.2. Approximate detection thresholds

Ground motion detected by HDS-InSAR in the La Paz area includes sub-hectare- scale landslides moving as slowly ~0.5 cm/a. Smaller and slower landslides might also be detectible by the technique, although they may just not be present in the area. InSAR is most sensitive to movement closely aligned with RADARSAT-2’s look direction (14° from the equator at the latitude of La Paz), which maximizes the component of the overall vector that can be measured and makes detection of even very slow movements possible. I had expected failures oriented perpendicular (approximately north-south) to the RADAR look direction to be undetectable. However, the vertical component of movement in the head (subsidence) and toe (uplift) of the Santa Barbara landslide (no. 3 in Fig. 4.6b; ~10 ha) enable excellent characterization of its activity (see discussion for further details), despite its non-ideal failure direction.

Only a few, relatively small areas of ground motion appear to be unrelated to landsliding. Round, roughly symmetric zones on level surfaces of fine-grained sediment, such as the deposits of the now-drained lake impounded by the Limanpata

150

paleolandslide (‘X’ in Fig. 4.6b), suggest vertical subsidence (up to 3.5 cm/a), most likely due to groundwater extraction. Slow (~1.5 cm/a line-of-sight) movement directly above the entrance to a small mine (‘Z’ in Fig. 4.6b) on the slope above the Hampaturi reservoir likely records rock subsidence (~1.5 cm/a if motion is completely vertical) due to underground mining, although landsliding cannot be ruled out.

Figure 4.6. InSAR-measured line-of-sight ground motion in La Paz and the surrounding area. RADAR look direction is from left to right. Linear displacement rates are away from the sensor (negative values; red and yellow colouring) and toward the sensor (positive values; blue colouring). Terrain that is not moving in the sensor’s line-of-sight is green. InSAR coverage area shown in Fig. 4.1. a) Linear displacement map with stretched scale (-5 to + 5 cm/a), highlighting the variability of rates of ground motion and areas of greatest motion. Extents of paleolandslide deposits mapped by Anzoleaga et al. (1977) are outlined in white. Grey boxes show coverage of Fig. 4.7 (solid) and approximate coverage of panels of Fig. 4.11 (dashed).

151

Figure 4.6. continued from pervious page b). (following page) Linear displacement map with compressed scale (-1 to + 1 cm/a), emphasizing the extent of areas of ground motion and showing localities discussed in the text. Ground motion due to processes other than landsliding include subsidence of level, fine-grained valley-bottom deposits at locations of likely ground water extraction (features ‘X’ and ‘Y’) and probably rock subsidence in the roof of a shallow mine (feature ‘Z’). Important landslides are numbered: 1. Limanpata paleolandslide. 2. earth flow below Hampaturi dam, 3. Santa Barbara landslide, 4. localized deformation in the central part of a paleolandslide deposit in Sopocachi, 5. Villa San Antonio paleolandslide, 6. Villa Armonía paleolandslid, 7. Villa Salomé paleolandslide, 8. Pampahasi paleolandslide and 2011 Pampahasi meaglandslide, 9. 2010 Allpacoma landslide, 10. area of instability on slopes north of Llojeta, 11. creeping slope in northern part of Seguencoma, 12. 2010 Huanu Huanuni landslide, 13. rapid motion in a portion of the Achocalla earth flow with localize loss of HDS, 14. active sub-basin of the Achocalla earth flow, 15. Cotacota paleolandslide, 16. newly identified translational landslide, 17. newly identified rotational landslide.

152

4.4.3. Spatial extent of motion within and south of La Paz

Roughly one-third of the valley slopes within and south of La Paz show activity throughout the 40-month period of the scene stack (Fig. 4.6), which is probably representative of long-term activity during the historic period. Motion occurs in urbanized zones, rural areas, and on natural slopes. The wider linear deformation scale of Fig. 4.6a (of -5 to 5 cm/a) shows the spatial variably of displacement rates both between and within deforming areas. The compressed linear deformation scale of Fig. 4.6a (-1 to 1 cm/a) highlights boundaries of deforming areas. The extent of deforming areas ranges from only a few city blocks up to entire slopes or sub-basins. The largest and most active zone is the southern half (~16 km2) of the Achocalla basin; slopes in several other parts of the basin are also active (Fig. 4.6). Activity in the La Paz basin is largely concentrated in its southern half and in the upper headwaters at the northeast limit of the study area. The steep slope ascending from the city centre to near the airport in El Alto show little evidence of movement, despite being the site of recent landslides (Chapter 3). Glaciofluvial terraces on which the centre of La Paz and Miraflores are built are also stable and not moving.

Most of the deforming areas correspond to existing landslides mapped in the southern part of the La Paz basin and in the Achocalla basin (Dobrovolny, 1962; Anzoleaga et al., 1977; Vargas, 1977) and northeast of the city in the headwaters of the Río La Paz valley system (Fig. 4.6; Chapter 3). Deformation typically ends abruptly at the limits of previously identified landslides. In some cases, such as the Achocalla earth flow, only portions of the previously failed material (Dobrovolny 1968; Anzoleaga et al., 1977) are moving. In the case of some other paleolandslides, the entire deposit appears to be active, with the area of movement closely matching the previously mapped extent of the deposit; examples include the Cotacota paleolandslide complex southeast of La Paz (Anzoleaga et al., 1977; Vargas, 1977; no. 15 in Fig. 4.6b) and the earthflow on the west site of the Río Irpavi valley just downstream of the Hampaturi dam and ~14 km northeast of the city centre (no. 2 in Fig. 4.6b; Chapter 3). Nearly all of the Pampahasi paleolandslide deposit (no. 8 in Fig. 4.6b) active between 2008 and 2011, including the ~2 km2 portion that failed in February 2011 (Hermanns et al., 2012; Aguilar, 2013).

153

In a few instances, motion within large paleolandslides appears unrelated to their margins; it either extends beyond their limits or is restricted to the middle parts of the old landslide deposits. An oval-shaped zone of deformation (no. 4 in Fig. 4.6b) near Sopocachi, which occurs within the middle part of a much larger paleolandslide, illustrates the latter; motion is localized along the northern slope of the now culverted and buried Río Cotahuma.

In other instances ground motion closely matches the extent of much smaller (~5-40 ha), more recent landslides mapped by Anzoleaga et al. (1977) (Fig. 4.6). Nearly all of the 35 ha area of a landslide at the head of the Allpacoma valley (no. 9 in Fig. 4.6b) is creeping. This area is the site of one of the five, >1 Mm3 failures in the twenty-first century (Hermanns et al., 2012). The limits of localized motion at the northern end of Seguencoma (no. 11 in Fig. 4.6b), along the bank of Río Choquyapu and north of Bella Vista, immediately south of the 2010 Huanu Huanuni landslide (no. 12 in Fig. 4.6b), closely match ~10 ha landslides mapped by Anzoleaga et al. (1977).

Line-of-site displacement rates throughout the processed area range from several millimetres to over 10 cm per year. However, these underestimate true displacement rates as they represent only one component the overall movement vector. The Miocene to Pleistocene sediment sequence exposed at La Paz lacks well formed structural elements (Dobrovolny, 1962; Bles et al., 1977; Chapter 2), and neither ancient (Dobrovolny, 1968; Anzoleaga et al., 1977; Fig. 4.4a) nor recent (Chapter 3) landslides show evidence of structural control along their basal sliding surfaces. Aside from a small number of localized faults, discontinuities in the sediment sequence are limited to nearly horizontal bedding, which should exert little direct influence on landslide motion. I thus assume the InSAR-measured slope movements are generally slope-parallel and thus estimated true displacement rates by comparison of the line-of-sight vector and downslope vector; these range from ~0.5 cm/a to over 20 cm/a. Estimated true displacement are greatest in the active portions of the source area of the Achocalla earth flow (no. 15 in Fig. 4.6b) and in the middle of the Cotacota paleolandslide.

154

The assumption of slope-parallel movement is invalid for some landslides in the La Paz area. Metasedimentary basement rocks outcropping south of La Paz have bedding and faults with a variety of orientations (Dobrovolny, 1962; Anzoleaga et al., 1977; Bles et al., 1977) that can be expected to influence failure kinematics (Brideau and Roberts, 2014) and thus displacement vectors (e.g. Morelli et al., 2011; Böhome et al., 2014; Booth et al., 2015; Schlögel et al., 2015). Determination of true failure direction and, thus, true movement rates for these slopes requires field measurement of structural orientations and kinematic analysis. Rotational movements in the overlying, non- structurally controlled sediment will be locally steeper that the slope angle, particularly in head scarp regions where the vertical component of motion may be large (e.g. the Santa Barbara landslide).

4.4.4. Detection of previously unknown landslides

In addition to showing that a large number of paleolandslide deposits and some twentieth-century failures are currently active, InSAR results reveal several previously unidentified landslides. In previously unmapped areas south, east, and northeast of La Paz, InSAR-measured ground motion reveals numerous unrecognized landslides ranging from several hectares to several hundred hectares in area. The newly identified landslides south and east of the city occur on ridges of Paleozoic metasedimentary rocks and, where not completely eroded, overlying sediments of the La Paz Formation. In addition a few small (<20 ha), previously unidentified landslides are located in areas mapped by previous researchers.

The landslides nearest the Cordillera Real are near the northeast limit of the La Paz basin. They involve Quaternary sediments similar to those in Río Kaluyo watershed (Chapter 2). They may also involve underlying Paleozoic basement rocks, in which case landslide motion may be influenced by rock structure, and the assumption of slope- parallel movement is unrealistic. I did not recognize several of these landslides in reconnaissance mapping based on optical satellite imagery (Chapter 3). These slopes are a short distance downvalley of the Hampaturi and Incachaca reservoirs and, therefore, pose no threat to them even if they failed catastrophically. However, their

155

activity raises questions about the stability of large, recently mapped landslides directly above the two reservoirs (Chapter 3), which are just beyond the limit of the InSAR- processed area and, therefore, are of unknown activity (Fig. 4.1).

4.4.5. San Antonino case study

Of the many areas of active slope movement in the La Paz area, San Antonio district is one of the closest to the city centre (Fig. 4.6). Four separate areas of recent ground motion are evident on the slopes composed of paleolandslide deposits that descend from the Pampahasi plateau to Río Orkojahuira (Fig. 4.7). Movement rates range up to 20 cm/a (Fig. 4.7a), but decrease to near-zero at the boundaries of the paleolandslides (Fig. 4.7b). The lower, northern part of the Villa Armonía paleolandslide deposit is deforming at rates up to 10 cm/a (line-of-sight; ~20 cm/a downslope). A small region at the upper southern part of the paleolandslide is deforming at a rate of 1 cm/a (line-of-sight; ~3 cm/a downslope). Part of the uppermost Villa San Antonio paleolandslide deposit directly below the Pampahasi plateau is moving at a rate of up to 2 cm/a (line-of-sight; ~3 cm/a downslope). The upper and lateral margins of the slope movement are more abrupt than the downslope limit, where deformation decreases gradually over ~100 m. Several hundred metres farther downslope is a small (~200 x 200 m), slowly moving (0.5-1.0 cm/a line-of-sight; ~1-2 cm/a downslope) area above Río Orkojahuira in Barrio Retamani. Displacement time series from the fastest moving parts of three of the four active areas show approximately linear movement rates throughout the InSAR monitoring period (Fig. 4.8a). These temporal patterns are representative of InSAR time series of most deforming slopes in the La Paz and Achocalla basins.

Deformation on slopes east of the Pampahasi plateau is restricted to paleolandslide deposits and the intact slopes adjacent to them (Fig. 4.7b). Average displacement rates during the InSAR processing period are greatest in the western part of the Villa Salomé paleolandslide and in the northwest portion of the Pampahasi paleolandslide and its adjacent intact slope to the north (Fig. 4.7a); there slope-parallel motion ranges up to 10 cm/a. These long-term averages are dominated by measured displacements in the 30 months prior to the 2011 Pampahasi landslide. During that

156

Figure 4.7. InSAR-measured line-of-sight ground motion in San Antonio study area.

157

RADAR look direction is from left to right. Linear displacement rates are away from the sensor (negative values; red and yellow colouring) and toward the sensor (positive values; blue colouring). Terrain that is not moving in the sensor’s line-of-sight is green. InSAR coverage shown in Fig. 4.6. a) Linear displacement map with stretched scale (-5 to + 5 cm/a), highlighting variability of rates of ground motion and areas of greatest motion. The extents of paleolandslide deposits mapped by Anzoleaga et al. (1977) are outlined in white. Dashed white lines show the extent of discrete landslide events between 1995 and 2014 (Chapter 3). Tips of triangles indicate locations of HDS time series (Fig. 4.8a). b) Linear displacement map with compressed scale (-1 to + 1 cm/a), show the extent of areas of ground motion and localities discussed in the text. Dashed white lines show the extent of discrete landslide events between 1995 and 2014 (Chapter 3). Black ovals represent resettlement of populations following recent landslides.

period, deformation was concentrated in the in the portion of the Villa Salomé paleolandslide directly upslope of the 2009 Villa Salomé landslide and in the slope corresponding to the northern (up-valley) portion of the 2011 Pampahasi landslide (Fig. 4.9a). After February 2011, however, the area of deformation expands to include the entire toe of the 2011 Pampahasi landslide, the plateau surface behind the headscarp, and the area between the Pampahasi landslide and the 2009 Villa Salomé landslide (Fig. 4.9b). Displacement time series from these areas show changes in rates of motion following the 2011 megalandslide (Fig. 4.8b). Displacement of the slope located between the 2009 and 2011 landslides increased during the first three RADAR acquisitions following the 2011 event, but then appeared to return a long-term rate similar to that experienced before the 2011 landslide. Displacement similarly increased a short distance behind the head scarp on the Pampahasi plateau, but more gradually and longer after the landslide occurred. In contrast, displacement at the toe of the Pampahasi landslide, where activity was greatest before the 2011 event, temporarily slowed for several months after February 2011 before returning to approximately the pre-failure rate.

158

Figure 4.8. Comparison of displacement time-series from the San Antonio district on slopes a) west and b) east of the Pampahasi plateau the western slopes with c) precipitation records. Locations of the HDS providing displacement records are shown in Fig. 4.7a. Note the difference in cumulative displacement scales. The time-series show no evidence of seasonality. Precipitation data from San Calixto weather station in the city centre, downloaded from the Servicio Nacional de Meteorología e Hidrología [SENAMI] at www.senamhi.gob.bo). Time of discrete landslide events (in red) correspond with periods of high cumulative precipitation, whereas slow deformation revealed by InSAR shows no relationship with precipitation.

159

Figure 4.9. InSAR-measured line-of-sight ground motion of in the eastern half of the San Antonio study area a) before and b) after the 2011 Pampahasi landslide. RADAR look direction is from left to right. Linear displacement rates are away from the sensor (negative values; red and yellow colouring) and toward the sensor (positive values; blue colouring). Terrain that is not moving in the sensor’s line-of-sight is green. The extents of paleolandslide deposits mapped by Anzoleaga et al. (1977) are outlined in white. Dashed white lines show the extent of discrete landslide events between 1995 and 2014 (Chapter 3). Tips of triangles indicate locations of HDS time series (Fig. 4.8).

4.5. Discussion

4.5.1. Spatial density of HDS

Landslide investigations using C-band PS-InSAR report target densities ranging from tens to hundreds of coherent point scatters per square kilometre (Bovenga et al., 2006; Del Ventisette et al., 2014; Lu et al., 2014; Cingna et al., 2015). The densities

160

depend on the type and number of scenes, as well as types of surface cover. They are generally low in high mountain areas (e.g. 33-83/km2 in the Italian Alps [Del Ventisette et al., 2014]) and greatest in urban areas, for example in central Italy (81-236/km2; Bovenga et al., 2006) and in the Greater London administrative area (386-446/km2; Cingna et al., 2015). Using SqueeSARTM, which like HDS-InSAR combines point and distributed targets, Bianchini et al. (2015) report point scatter densities between 16 and 112/km2 for rural and urban land cover in Sicily.

The high density of HDS in La Paz enables detailed characterization of the morphology and extent of slope deformation. However, this density exceeds the true density of independent values. Neighboring HDS points are computed using overlapping adaptive windows that in many cases combine phase components from discrete measurements. Despite the spatial averaging of ground motion resulting from this oversampling, the HDS dataset considered here still enables differentiation of abrupt and diffuse spatial limits of motion (see discussion of kinematics below).

Gaps in HDS coverage and thus in the linear deformation maps (Figs. 4.7 and 4.8) are the result of removal of HDS during quality thresholding. The low phase quality of these areas is, in many cases, a result of decorrelation due to the variable nature of surface cover, particularly in the few areas of forest cover, lakes, and terrain heavily modified by anthropogenic activity. Thresholding did not remove all unreliable HDS in these neighbourhoods. For instance, meaningless InSAR results persist at the southern end of the Hampaturi reservoir (Fig. 4.6). The absence of HDS points in some rapidly moving areas, such as the middle of fast-moving parts of the Achocalla earth flow (Fig. 4.6), is the result of aliasing due to line-of-sight ground movement magnitudes beyond

the detection ability of RADARSAT-2.

4.5.2. Relation of slope deformation to the geology and physiography of La Paz

The concentration of slope deformation south of the centre of La Paz (Fig. 4.6) may reflect geologic controls, physiographic controls, or both. The combination of more

161

competent geology and less incision would logically result in fewer and smaller deforming slopes in the northern part of InSAR coverage, north of the city. The sediment sequence north of La Paz is dominated by coarse-grained, relatively high-strength sediments that have been loaded by numerous late Pliocene and Early Pleistocene glaciers extending from the Cordillera Real (Chapter 2). These units are incised about 200 m nearest the Cordillera Real and 500 m at the northern city limits. Areas of recent ground motion upvalley from La Paz are relatively uncommon and concentrated in the extreme northeast area of the InSAR coverage (e.g. no. 2 in Fig. 4.6b) where the cover of late Cenozoic sediments overlying Paleozoic rocks of the Cordillera Real is thin.

In contrast, south of the city centre, the La Paz basin sediment fill is finer grained, highly heterogeneous, and laterally variable. The resulting high permeability contrasts over short distances direct more groundwater onto slopes. Additionally, the higher clay content in fine-grained sequences may increase their susceptibility to shearing. The depth of incision ranges from up to 800 m in the southern La Paz basin to over 1000 m a few kilometres farther downstream near the southern limit of InSAR coverage. Deep incision of a relatively weak sequence with high slope-water content during prolonged wet periods may contribute to greater recent slope activity, resulting in greater occurrences of both prehistoric (Dobrovolny, 1962; Anzoleaga et al., 1977; Vargas, 1977) and recent (Chapter 3) landslides.

4.5.3. Relation to historic landslides

InSAR-measured activity of paleolandslides in the Achocalla basin and the southern part of the La Paz basin suggests that historic instability in the La Paz area is related to strength reduction of previously failed slopes (Chapter 3). It also suggests that some of these ancient landslide deposits may be the product of punctuated or sustained long-term deformation. Coincidence of some deforming areas and small modern landslides (Anzoleaga et al., 1977) further supports this inference. By contrast, large paleolandlsides at the northern limits of La Paz appear to be stable today. Ground motion detection over much of the Purapura paleolandslide is hindered by decorrelation due to the vegetation cover dominated by eucalyptus forest, but urbanized areas with

162

coherent InSAR results near the toe of the landslide are not moving. Minor deformation at the toe of the Limanpata paleolandslide (no. 7 in Fig. 4.6b) along the steep bank of Río Choqueyapu corresponds to an area of recent small debris slides and probably reflects localized fluvial undercutting as the stream adjusts to the nick point formed by the ancient landslide deposit.

Furthermore, the paleolandslides with the highest movement rates (Fig. 4.6) are those that have experienced frequent failures along their margins in recent decades (Fig. 4.1). Of the 43 landslides and 13 areas of creep responsible for infrastructure damage between 1995 and 2014 (Chapter 3), over half of those with precisely known locations correspond to the margins of paleolandslides showing extensive motion between 2008 and 2011. This relationship supports the inference in Chapter 3 that many recent landslides in La Paz, and perhaps most historic failures, are driven by differential motion at the limits of extremely slow, currently active paleolandslides.

4.5.4. San Antonino case study

Spatial patterns of ground motion

The spatial limits of InSAR-measured movement in the district of San Antonio agree well with recent infrastructure damage, indicating that InSAR reliably documents current slope movements in La Paz. For instance, warped roads and cracked walls are visible in both the active portion of San Isirdo (e.g. Fig. 4.4d) and Villa Armonía, but are lacking in the narrow, non-deforming zone between them (Fig. 4.7).

Four of the five, twenty-first century failures in Barrio Retamani occurred at the edges of a slowly deforming area measuring only ~200 m x 200 m (Fig. 4.10). In all four instances, one lateral margin of the failure roughly corresponds to the edge of the deforming zone at toe of the Villa San Antonio paleolandslide. Each of these four recent failures extends 100 m or more beyond the creeping zone into adjacent parts of the paleolandslide with no InSAR-measured motion. The 1997 Cuatro Centenario landslide, ~500 m south of Barrio Retamani (Guzmán, 2007b), similarly happened at the edge of

163

the Villa Armonía paleolandslide (Fig. 4.10). The fifth recent Retamani failure is almost 200 m north of the deforming zone at the toe of the Villa San Antonio paleolandslide (Fig. 4.10), but was triggered by construction activity. The presence of failure surfaces beneath the recently active Villa Armonía and Villa San Antonio paleolandlsides (Fig. 4.4a-c) suggests that modern large-scale deformation in some slopes of the La Paz area is accommodated by shearing along discrete basal shear zones.

Figure 4.10. Three-dimensional perspective view of Barrio Retamani and adjacent parts of the San Antonio district showing spatial relations between twenty-first century landslides and InSAR-measured deformation within the paleolandslide deposits. Four of the five twenty-first century landslides in Barrio Retamani (dashed outlines) correspond to the limits of a slowly deforming zone at the toe of the Villa San Antonio paleolandslide. Residents displaced by the 2011 Pampahasi landslide were relocated to shelters (white half circle) at the active toe of the Villa Armonía paleolandslide. Tips of triangles indicate locations of HDS time series (Fig. 4.8a). See Fig. 4.7 for general setting.

The 2009 Villa Salomé landslide is directly down slope of the most rapidly deforming portion of the Villa Salomé paleolandslide deposit (Figs. 4.8). However, it is unclear whether the slope involved in the landslide was moving before or shortly after the landslide. The lack of InSAR results for the 2009 landslide (Figs. 4.8 and 10a) most likely result from temporal decorrelation due to post-landslide clearing and terracing. However, aliasing of phase data at this location due to high local displacement velocities is also possible. This latter source of data loss is supported by high displacement rates

164

measured in 2011 (Fig. 4.9b). Spatial correspondence of the 2011 Pampahasi landslide and the Pampahasi paleolandslide, including its most active portion, suggests that the largest recent landslide in La Paz involved acceleration of a slowly creeping slope.

Post-landslide expansion of the area of slope movement beyond the lateral margins and headscarp of the 2011 Pampahasi landslide indicates that this event destabilized adjacent parts of the slope. Ground motion in these areas, particularly the radial pattern on the Pampahasi plateau behind the head scarp (Fig. 4.9), suggests dilation of material adjacent to the head scarp in response to unloading by the landslide. Motion within the lowest part of the Pampahasi landslide expanded from only the northern portion, prior to the event (Fig. 4.9a) to the entire toe region after the event (Fig. 4.9b). Thus, the 2011 megalandslide locally destabilized several parts of the western slope of the Río Irpavi valley.

Temporal patterns of ground motion

Several anomalies in the displacement time series for active slopes in the San Antonio district are likely artifacts arising from data processing. The slight increase in the long-term linear displacement trend of the Villa Armonía paleolandslide after about day 250 in the InSAR time series (Fig. 4.8) may reflect an underestimate of overall displacement during the early part of the record due to temporal gaps in scene acquisitions (only six of the first 12 planned acquisitions were made). Small differences in displacement magnitudes between acquisitions result, at least in part, to noise due to either phase changes other than ground motion or to errors introduced during processing. Although some of these differences may reflect minor changes in landslide rates, reversal of the direction of movement (e.g. in two of the three time series between scenes 16 and 17; Fig. 4.8a) cannot be explained by landsliding. These erroneous changes in displacement directions are largest and most common in slowly deforming areas (< 1 cm/a line-of-sight displacement; for example the time series of the Villa San Antonio paleolandslide (Fig. 4.8a) and behind the head scarp of the 2011 Pampahasi landslide (Fig. 4.8b). The errors are cancelled out by the larger magnitude ground deformation on more rapidly moving slopes (e.g. time series of the Villa San Armonía paleolandslide in Fig. 4.8a). The reversal in displacement direction in the final step of all

165

three time series in Figure 4.8a probably results from limited temporal coverage and is only present in some HDS time series in the west part of the San Antonio district and throughout La Paz. Similar, apparently anomalous displacements away from the sensor are evident at the end of time series from the slopes east of the Pampahasi plateau (Fig. 4.8b).

Slopes in La Paz display a relatively constant rate of InSAR-measured motion throughout the year (Fig. 4.8a, b), despite the region’s strongly seasonal precipitation (Fig. 4.8c). It is possible that a temporal smoothing window applied to each HDS time series during processing (Appendix H) has masked weather-related changes in landslide velocity. Complete masking of such signals is most likely if the velocities are small compared to overall long-term creep rates. Portions of some HDS time series show slight acceleration during long rainy periods, such as at Villa Armonía during the 2010- 2011 wet season (Fig. 4.8), but changes in motion are generally inconsistent between periods of similar precipitation (Fig. 4.8b) and are spatially inconsistent across moving areas. Reprocessing using a time-adaptive filter might reveal seasonal trends in landslide creep rates, but they would likely be small. The apparent lack of an effect of seasonal precipitation on slow slope creep contrasts markedly to the concentration of discrete rapid landslide events during the rainy season (Chapter 3).

There is no clear change in displacement rates at the active toe of the Villa San Antonio paleolandslide leading up to or following the rapid failures in Barrio Retamani in 2009 or 2010 (Fig. 4.8a). Similarly, there are no changes in displacement rates on slopes east of the Pampahasi plateau prior to the 2009 Villa Salomé landslide and the 2011 Pampahasi landslide. Although on-going creep of paleolandslide deposits may predispose some locations to more rapid failure, the timing of these recent landslides is not clearly linked to paleolandslide activity.

However, the times of the 2009 and 2010 Retamani landslides suggest a meteorological trigger, as both failures followed periods of high 14-day cumulative precipitation (Fig. 4.8b; Chapter 3). Changes in slow deformation prior to localized rapid failures cannot be ruled out. If such accelerations span a small number of scenes, they

166

are particularly likely to be masked by indiscriminate temporal smoothing of displacement time series. Alternatively, they may occur on a timescale shorter than the periodicity of RADARSAT-2 acquisitions (i.e. 24 days). Peyret et al. (2008) detected possible precipitation controls on landslide movement in the Alborz Mountains (northern Iran) based on GPS records, but saw no evidence of this relation in displacements inferred from D-InSAR. The Envisat ASAR scenes used in that study have a similar bandwidth and revisit frequency to RADARSAT-2.

In contrast, InSAR-measured displacement rates within and adjacent to the 2011 Pampahasi landslide change after the landslide (Fig. 4.8b). Deceleration of the northern toe of the landslide in the months following the landslide suggests a temporary increase in stability. Conversely, acceleration in the area between the upper part of the landslide and the 2009 Villa Salomé landslide suggests an increase in instability due to unloading of the slope below. Delayed movement of the plateau surface behind the landslide might reflect gradual debuttressing by repeated post-landslide collapse at the head scarp. At all these locations, ground motions returned to previous, long-term rates within several months of failure (Fig. 4.8b).

Similar instability on opposite slopes of the Pampahasi plateau

The slopes west of the Pampahasi plateau, including Villa San Antonio and Villa Armonía, could experience a large rapid failure similar to the 2011 landslide on the slopes east of the plateau. Both areas are underlain by fine-grained sediments of the La Paz Formation that have failed on many previous occasions (Anzoleaga et al., 1977; Hermanns et al., 2012; Chapter 3), producing remobilized deposits that cover much of the modern slopes. Rates of motion throughout the InSAR time series range from several centimetres to several decimetres per year on both sides of the plateau (Fig. 4.6). The absence of a clear change in the velocity of InSAR-measured creep prior to the 2011 megalandslide suggests that there is the potential for a future similar event in the western part of the San Antonio district with limited pre-failure warning.

167

Resettlement after recent landslides

Most of the Barrio Retamani residents displaced by the 2009 Retamani II landslide were resettled in a government-built apartment building in San Isidro (Chapter 3), directly across a narrow gulley from the 1997 San Isidro landslide (Guzmán, 2007a). Although InSAR results suggest the site of the apartment complex is not moving, gradual displacement of adjacent paleolandslide deposits (Fig. 4.7) could lead to nearby failures in the future.

Other short-term relocations of landslide-affected residents of La Paz have resulted in their continued exposure to high landslide hazard. Some evacuees from 2011 Pampahsi landslide were temporarily resettled at the deforming toe of the Villa Armonía paleolandslide (Figs. 4.8 and 4.10). The concentration of recent failures along the active margin of this paleolandslide and at similar settings at the toe of the Villa San Antonio paleolandslide illustrate that this location is hazardous.

4.5.5. Discrimination of kinematics and failure mode

Spatial differences in the direction and magnitude of line-of-site deformation provide insight into the mode of deformation and hence mechanisms of slow, on-going landslides in La Paz (Fig. 4.11). The kinematics of some previously documented landslides in the city can be further characterized or confirmed from the InSAR results. The Santa Barbara landslide east of the city centre, which has been active throughout the nineteenth (Sanjinés, 1948), twentieth (Sanjinés, 1948; Dobrovolny, 1962) and twenty-first (Chapter 3) centuries, is a rotational landslide (Dobrovolny, 1962, p. 89). Spatial differences in the rate and sense of InSAR-measured displacements (Fig. 4.11a) clearly show the rotational nature of the landslide, which includes subsidence at the head and uplift at the toe. Additionally, InSAR results show that only a portion of the landslide area mapped in detail by Dobrovolny (1962) was active between 2008 and 2011 (Fig. 4.11a).

168

Figure 4.11. Examples of landslide failure mechanisms inferred from HDS-InSAR results. Locations of landslides are shown in Fig. 4.6a). a) Santa Barbara rotational landslides with dashed lines showing landslide limits mapped by Dobrovolny (1962). b) Cotacota paleolandslide with loci of recent surface disruption (Chapter 3) identified by white circles. c) Achocalla earth flow, directly south of La Paz, with multiple zones of activity including a large region in the southeast part of the Achocalla basin bounded by parallel linear lateral margins. d) Previously unmapped landslides east of Cota Cota, including a translational slide or flow (western landslide) and a rotational slide (eastern landslide).

Most of the valley bottom in Cota Cota, which forms a large part of the Cotacota paleolandslide, (Anzoleaga et al., 1977; Vargas, 1977, p. 37) is moving (‘15’ in Fig. 4.6b; Fig. 4.11b). Based on local topography, the Cotacota paleolandslide likely moves directly to the west in its lower half and toward the west-southwest in its upper half. Spatial variability in displacement rates indicate both discrete and gradual motion boundaries surrounding a central zone of spatially regular deformation. The landslide is thus a plug flow, a rigid body of constant velocity, with motion accommodated by a zone of flowing or sliding (Craig, 1981; Savage and Smith, 1985; Brewster et al., 2005). Gradually increasing velocity from the toe of the Cotacota paleolandslide eastward along its centreline until reaching a velocity plateau (up to 6.5 cm/a line-of-sight at the middle of

169

the deposit; ~12 cm/a in the presumed direction of actual motion) indicates a zone of compression down flow of the central rigid body. The lesser degree of velocity increase from its lateral margins toward the central zone (Fig. 4.11b) suggest a broad drag zone, whereas abrupt increases in displacement rates along some parts of the flow’s lateral margins suggests concentrated lateral shear (Fig. 4.11b). The inferred shear zones correspond to sites of ongoing deformation causing infrastructure damage (Chapter 3) and ground fissuring (Fig. 4.2h) in recent decades. Greater movement rates (up to 14 cm/a line-of-sight; ~25 cm/a assuming downslope movement) in an isolated part of the source area indicate a zone of compression ~1.5 km downslope of the head scarp. Alluvial terrace and fan deposits in the upper part of the valley (Anzoleaga et al., 1977) have the same displacement rates as adjacent landslide debris, indicating that they cover the landslide surface and are being carried along passively with it.

The early Holocene Achocalla earth flow (Hermanns et al., 2012) is the largest landslide in the La Paz area and one of the largest non-volcanic landslides in the Andes. Its surface morphology (Dobrovolny, 1968; Vargas, 1977) and internal deformation (Hermanns et al., 2012) are suggestive of complex failure, which is supported by InSAR results (Fig. 4.11c). Numerous zones of recent motion, several approaching 20 cm/a, are separated by areas that are presently stationary, indicating that different parts of the landslide are moving independently of one another. The largest active portion of the paleolandslide has parallel, abrupt, linear, lateral boundaries of motion, which may be structurally controlled because they align with morphologic lineaments mapped by Lavenu (1977). In contrast, displacement rates decrease gradually at the upper limit of this area. Small areas with line-of-site movement toward the sensor (blue in Fig. 4.11c), opposite to that of the overall failure direction, which is away from the sensor (yellow and red in Fig. 4.11c), suggest superficial instability controlled by local surface morphology.

Thus, unlike recent discrete landslide events in the Upper Río La Paz valley system that are largely compound slides (Chapter 3), the generally much larger, slowly moving paleolandslide deposits include a number of flow-type landslides. The paleolandslides include some very complex mass movements, particularly the Achocalla earthflow. InSAR-measured motion also reveals that although differential movement at the limits of large slow landslides may include localized sliding, such as ongoing

170

landslides discussed in Chapter 3 [Table 3.2]), overall motion may be controlled by a different mechanism.

The kinematics of landslides that I have newly identified can be similarly interpreted from InSAR results, highlighting the utility of HDS-InSAR for expanding landslide inventories in La Paz. For instance, two landslides east of Cota Cota, beyond the limits of previous mapping (Dobrovolny, 1962; Anzoleaga et al., 1977; Vargas, 1977) show distinct deformation patterns (Fig. 4.11d). The more easterly of the two landslides is oriented obliquely toward the RADAR look direction. If the failure is purely translational, its line-of-sight motion should be dominated by movement toward the sensor. The toe is moving toward the sensor, but motion near the head scarp is away from the sensor, indicating that subsidence dominates the overall motion, consistent with rotational failure. The pattern is less pronounced than that of the Santa Barbara landslide (Fig. 4.11a), likely because its overall movement direction is closer to the sensor look direction, resulting in greater InSAR sensitivity to the horizontal component of motion. Recent terracing of the upper part of this landslide, which is first evident in optical satellite imagery acquired in 2012, suggests that deformation has already affected the road crossing it.

The morphology of the western failure (Fig. 4.11d) suggests movement toward the south, nearly parallel to that of the Santa Barbara landslide and perpendicular to the RADAR look direction. InSAR-measured motion is particularly sensitive to vertical displacement on this slope. However, the sense of movement over the entire landslide is away from the sensor, suggesting largely translational motion up to ~9 cm/a downslope. The lack of HDS in a large portion of the area is the result of phase decorrelation that may stem from line-of-sight deformation rates larger than those that can be measured using the wavelength and revisit frequency of RADARSAT-2.

4.5.6. Relationships between ancient and historic instability

Extensive, slowly deforming slopes in La Paz are located mainly in areas with long histories of instability. These areas – including and exemplified by the western

171

slopes of the San Antonio district – experienced large prehistoric failures (Dobrovolny, 1962; Anzoleaga et al., 1977; Vargas, 1977) and smaller failures over the past two centuries (Santa Cruz, 1942; Sanjinés, 1948; O’Hare and Rivas, 2005; Chapter 3) that have damaged the built environment (Chapter 3; Fig. 2a,b,e,f,h and Fig. 2d,e).

The rate (Fig. 4.6) and failure mode of paleolandslides active between 2008 and 2011 differ with distance from the Cordillera Real. Earthflows northeast of La Paz have occurred in Paleozoic bedrock mantled by remnants of coarse Plio-Pleistocene sediments and typically show activity during the InSAR monitoring period. The single failure documented in recent decades in this region (2008 Hampaturi landslide; Hardy, 2009; Chapter 3) happened at the toe of one of these active slopes. Large prehistoric landslides in the northern part of La Paz show little (Limanpata paleolandslide) or no (Purapura paleolandslide) movement during the InSAR monitoring period (Fig. 4.6). In contrast, most paleolandslides farther south are slowly moving and are the sites of most sudden failures in recent decades (Chapter 3). These include the Barrio Retamani failures (Fig. 4.7; four twentieth-century failures) and the 2010 Huanu Huanuni landslide (no. 12 in Fig. 4.6b). In the middle part of the La Paz basin (e.g. San Antonio district; Figs. 4.8 and 4.10) and the upper reaches of the Allpacoma valley (no. 9 in Fig. 4.6b), recent landsliding is dominated by rotational and translational sliding resulting in downslope movements of ~1-10 cm/a. Paleolandslides in the lower parts of the La Paz and Achocalla basin sediment sequences show evidence of ongoing flow-type behaviour in InSAR results, locally exceeding 20 cm/a (e.g. Cotacota paleolandslide; Fig. 4.11b and no. 15 in Fig. 4.6b). Fatal landsides in La Paz in recent decades, which are typically very rapid falls and collapses in the upper part of the sedimentary sequence (Chapter 3), show no evidence of displacement in the InSAR results (Fig. 4.6). These failures are poorly related to ongoing creep of weak slopes, and their location is probably determined principally by contrasts in geomechanical and hydrological properties (Chapter 3).

172

4.6. Conclusions

HDS-InSAR analysis of over three years of RADARSAT-2 scenes provides insights into the location, extent, rate, and type of landslide activity in the city of La Paz and the surrounding area. Detailed characterization of landslides is possible because HDS-InSAR provides dense measurements of spatially uncorrelated ground motion using adaptive filtering. Environmental conditions and the urban infrastructure in La Paz have allowed me to produce high-accuracy deformation maps and time series.

InSAR-defined motion patterns are consistent with previously mapped, recent and ancient landslides. They improve landslide maps dating from the mid-twentieth century by providing detailed characterization of the activity of individual failures. In areas southeast of La Paz that have not been mapped, InSAR results identify undocumented landslides and provide a characterization of their activity and mechanisms. Additionally, some landslides not recognized in previously mapped areas were also identified. The behaviour of most slopes over the 40-month InSAR acquisition period is probably representative of long-term activity.

Landslide activity is controlled by the geology of the sedimentary sequence underlying the Altiplano. Recent landsliding is most common in fine-grained sediments exposed in steep slopes of deep valleys. Slow movement occurs principally in paleolandslide deposits and typically coincides closely with their boundaries, suggesting that strength reduction of previously failed slopes is an important contributor to instability. However, it is unclear if modern creep of paleolandslide deposits is a result of recent re-mobilization or long-term, continuous creep following initial failure, which in some cases dates back to the beginning of the Holocene. A smaller number of deforming areas correspond to much smaller, undated landslides, suggesting that some landslides in the area experience slow creep for at least many decades following initial failure.

The InSAR-measured slope movements summarized here constitute an important new dataset to assist in characterizing and reducing landslide hazard and risk

173

in La Paz. Many areas of InSAR-documented movement were previously known to be sites of recent slope failure, but the InSAR results have identified additional areas of instability and have further provided movement rates that are sustained over periods of months or longer. Furthermore, they provide insight into the type and likelihood of future landslides in La Paz.

Similarities in the pattern, extent, and rate of slope creep on slopes on both sides of the Pampahasi plateau highlight their potential for future, more rapid larger scale failure. The 2011 Pampahasi landslide on the slopes east of the plateau involved an acceleration of most of the creeping Pampahasi paleolandslide and was the largest and most damaging landslide in the city’s recent history. A similar acceleration of creeping paleolandslide deposits on the west side of the plateau could be even more damaging, given the greater urban population there.

The concentration of historic landslides along the margins of paleolandslide deposits indicates a specific setting in which future failures are most likely to happen. Comparison of recent landslide events with InSAR results shows that the mechanism, rates, and possible sizes of historic landslides depend largely on the behaviour of the creeping slopes with which they are associated. At the margins of extremely slow (~1-10 cm/a), translational and rotational slides, such as those below the Pampahasi plateau, landslides are likely to be rapid to very rapid. At the margins of faster (up to over 20 cm/a) flow-type slides south of La Paz, such as in Zona Sur, future failures are likely to be small, slow to moderate slides and flows. The absence of documented failures or infrastructure damage within the main body of the Cotacota earthflow appears to result from uniform plug-like flow. The lack of recent ground motion in areas of topples and falls, which are potentially the most deadly types of landslides due to their extremely rapid rates, indicates that InSAR is of no assistance in locating their future occurrence. Detection of deforming slopes adjacent to water reservoirs 10 km upvalley of La Paz suggests that landslide-generated displacement waves may be a yet-unconsidered hazard for the city.

174

InSAR results highlight areas where further studies would help improve hazard assessment and reduce risk. Detailed, ground-based site investigations are advisable in all densely developed areas showing recent deformation. These areas include Villa Armonía and Villa San Antonio, as well as the area affected by the 2011 Pampahasi landslide and adjacent slopes underlain by the Villa Salomé paleolandslide. Several other areas are priorities for further study because of their combination of recent motion, damaging historic landslides, and increasing populations; examples include the upper part of the Allpacoma valley and slopes adjacent to Llojeta. High-frequency, geodetic, ground-based monitoring during the rainy season might detect the effects of heavy rainfall on slow creep-like slope movements that cannot be resolved with InSAR.

175

Chapter 5. General conclusions

Through an investigation of landscape evolution at La Paz, Bolivia, over timescales ranging from annual to millions of years, I have contributed to an understanding of the evolution of the eastern Bolivian Altiplano and adjacent Cordillera Real. I have also improved understanding of the factors responsible for the high instability of slopes in the city of the La Paz.

The La Paz area experienced deposition of thick glacial and proglacial sediments during repeated late Pliocene and Early Pleistocene glaciations. Glaciations were separated by interglaciations during which much of the land surface was stable and was subject to soil formation. The uppermost ~200-400 m of the sedimentary fill in the La Paz and Achocalla basins is of glacial origin, but fines away from the Cordillera Real from diamicton and gravel north of the city centre to silt, sand, and fine gravel south of the city. Similar fining trends probably characterize Miocene and early Pliocene sediments lower in the sedimentary sequence. The Altiplano surface at La Paz had formed by ca. 1.0 Ma or, more likely, ca. 1.8 Ma, after which it was deeply incised by headwaters of the Amazon River system as it breached the Cordillera Real.

Landslide activity within the Río La Paz watershed is greatest in the southern part of the La Paz basin and in the Achocalla basin due to distal fining of the Neogene and Early Pleistocene sedimentary sequence. Most of the sudden-onset (discrete) landslides in La Paz between 1995 and 2014 initiated as very rapid silt compound slides in previously failed material. These slides account for most of documented landslide damage in La Paz. Their close association with the deposits, particularly those at the margins, of large paleolandslides indicates an important geologic constraint that can help identify likely locations of future landslides. The spatial association of lateral margins of numerous landslides with buried culverted streams highlights another likely

176

setting of future rotational and translational slides. These culvert-delimited landslides are particularly common in areas of InSAR-measured creep, suggesting that slow ongoing slope movement compromises culvert integrity. The low quality of some culverts, however, makes it possible that they could also leak in non-creeping slopes. Extremely rapid falls and topples have been relatively infrequent, but account for all landslide fatalities between 1995 and 2014. They typically occurred in near-vertical, previously undisturbed slopes, particularly those formed in the weakly cemented Chijini Tuff.

The concentration of landslides between 1995 and 2014 during the last three months of the wet season (January to March) and their common occurrence on days following more than 40 mm of 14-day cumulative precipitation indicates a meteorological control on the timing of many failures. Given the marginal stability of many slopes in La Paz, earthquakes are also a potentially important, but thus far uninvestigated, landslide trigger. No landslides occurred in La Paz during or immediately after the April 2014 Iquique earthquake (M. Guzmán and E. Minaya, personal communication, 2014), which at moment magnitude 8.1 (Ruiz et al., 2014) was the strongest recent earthquake in the Central Andes. In addition to subduction-zone earthquakes hundreds of kilometres southwest of La Paz, strong earthquakes occur along the Subandean fold-thrust belt hundreds of kilometres to its north and east, including very deep events (Kikuchi and Kanamori, 1994). Strong, shallow earthquakes closer to La Paz could generate much greater local ground acceleration, potentially leading to slope failures. Several earthquakes predating seismic monitoring, and therefore of unknown magnitude, were reported in La Paz during the nineteenth and early twentieth centuries (Bles, 1977). The most strongly felt of these, which occurred in February 1947, was one of serval historic earthquakes with epicentres roughly below the Cordillera Real within 100 km of La Paz (Bles, 1977). However, relatively little is known about these nearby events, including their impact on slope stability in the upper Río La Paz valley system.

High (up to 20 cm/yr) InSAR-measured displacements during the monitoring period coincide with the locations of recent and ancient landslides. Most of the recent (1995-2014) landslides in La Paz correspond to the margins of active paleolandslides; differential movement due to large-scale, on-going creep, in addition to strength reduction in previously failed slopes, seem to control the location of recent landslides.

177

The displacement rate and failure mechanism of the active paleolandslides, as well as the nature of smaller failures generated at their margins, differ between the central and southern parts of La Paz due to vertical and lateral lithologic changes within the sedimentary sequence sequence. InSAR results show no evidence of deformation at sites of extremely rapid topples and falls that occurred between 1995 and 2014, including the only recent fatal landslides in La Paz. The failures with the greatest potential for fatalities are thus the most difficult to forecast.

In summary, slope instability in the Río La Paz watershed is the result of long- term landscape evolution involving late Pliocene and Early Pleistocene glaciations, Middle Pleistocene valley incision, and Holocene paleolandslide activity. Urban development in La Paz, particularly inappropriate stream engineering, contributes to instability on some slopes.

5.1. Recommendations for reducing landslide risk in La Paz

My research highlights some practices that could be adopted to reduce landslide risk in La Paz. Hazard avoidance is the most obvious way to mitigate landslide risk in La Paz, but rapid urban growth and the limited availability of land for settlement limit land- use options. Land-use planning in La Paz already considers landslide hazards, but information on the contemporary activity of many areas is still limited. Rates of slow- moving landslides that are typical in La Paz are below those expected to cause fatalities, but cause severe damage to roadways and buildings. Nevertheless, the distribution of slow landslides in La Paz provides insights into the potential for more rapid landslides that can cause widespread infrastructure damage, properly loss, and fatalities.

The margins of active paleolandslides are most likely to be sites of future rapid failures. They include slopes on both sides of the Pampahasi plateau and slopes adjacent to Llojeta. Future development in these areas should be limited to low-risk uses, such as parks or other open spaces. Existing housing developments in these areas should be a priority for further hazard and risk studies.

178

Several areas in the La Paz area appear to be stable based on the InSAR results and thus may be appropriate sites for future development or increased urban densification. Glaciofluvial terraces on the valley bottoms of the Río Choqueyapu and Río Orkojahuira, which include the city centre, Miraflores, and Obrajes, are stable. Slopes of the Río Choqueyapu valley north of the city centre, which experienced large failures at Purapura and Limanpata, show no recent movement. Small parts of the Achocalla basin outside areas of recent movement might be suitable for future development, although their association with recent activity of the Achocalla earthflow is unclear. High western slopes of the Río Orkojahuira valley north of Miraflores, which currently are being surveyed and graded for development, show few signs of recent ground deformation. Parts of some slowly deforming slopes might be suitable for existing and future developments; an example is the area of temporally consistent, spatially uniform displacement in the middle of the Cotacota paleolandslide. However, further development in these areas must include detailed site investigations, as well as periodic surveying to detect new instability that may arise from changed land use.

Extremely rapid topples or falls pose the greatest threat to lives in La Paz. Unlike most other recent landslides in the city, they occur on otherwise stable slopes. Assessment of their future potential requires systematic site investigation of the steepest slopes. High outcrops of Chijini Tuff should be one of the first targets for field-based assessment.

The location and extent of recent landslides (Chapter 3) and slowly creeping slopes (Chapter 4) should be considered in future efforts to reduce landslide risk in La Paz. However, this knowledge can also be used when relocating residents displaced by landslides or floods. Future sites for both short-term evacuation and long-term resettlement should not include the active margins of paleolandslide deposits. The relocation of evacuees of the 2011 Pamapahsi megalandslide to the toe of the highly active Villa Armonía paleolandslide illustrates this issue. Similarly, temporary evacuation sites in the middle of the slope affected by the 2011 megalandslide appear to have become permanent settlements. In addition to subjecting already disadvantaged evacuees to further landslide hazards, such practices may further contribute to local slope instability by introducing waste water and runoff to metastable slopes.

179

My research highlights other processes that should be considered in disaster preparedness strategies:

1. Impoundment of landslide-dammed lakes, resulting in inundation upstream and downstream outburst floods (Chapter 3); 2. Debris flows due to overtopping of reservoirs in the upper part of the Río Choqueyapu valley (Siete Lagunas), Río Chuquiaguillo valley (Incachaca reservoir), and Río Irpavi valley (Hampaturi reservoir) (Chapter 3); 3. Interruption of water supply of hundreds of thousands of La Paz’s residents due to landslides (Chapter 3); 4. Overtopping displacement waves generated by landslides on slopes above the Incachaca reservoir, 10 km northeast of La Paz (Chapter 4); and 5. Catastrophic failure of slowly creeping paleolandslides in the San Antonio district, either west of the Pampahasi plateau at Villa Armonía and Villa San Antonio or east of the plateau at the location of the 2011 megalandslide (Chapter 4).

5.2. Future research

My research raises many new questions about the late Cenozoic geology and geohazards at La Paz. I recommend that future research include the following topics.

5.2.1. Characterization of the Neogene and Pleistocene sediment sequence

1. Numerous spatial and temporal gaps remain in the lithostratigraphy and chronostratigraphy at La Paz. Priorities in filling these gaps, and thereby expanding knowledge of Pliocene and Early Pleistocene paleoenvironments along the eastern Altiplano and Cordillera Real, include: 2. Localized dense magnetostratigraphic sampling at some sections of the Purapurani Gravel, Kaluyo Gravel, and their distal fine-grained equivalents in order to constrain

180

the boundaries of short, normally magnetized subchrons of the Matuyama Chron and to confirm which such subchrons are recorded. 3. Application of alternative geochronological methods to evaluate correlation of the fill sequence above the Chijini Tuff to the Geomagnetic Polarity Timescale, particularly TCN exposure dating of paloesols overlain by till to constrain their burial age and thus the timing of glaciation (Balco et al., 2005a, 2005b). 4. Description and paleomagnetic sampling of additional stratigraphic sections along a transect extending southeast from the Patapatani West section to characterize variability of facies and thus glacial environments parallel the trend of the Cordillera Real.

5.2.2. Landslide investigations

Chapters 3 and 4 of this thesis provide a strong framework for detailed investigation of specific landslide sites in La Paz. Future landslide research should include:

1. Landslide susceptibility mapping for the La Paz and Achocalla basins. 2. Long-term, ground-based monitoring of slowly moving paleolandslides and probable sites of future discrete failures through permanent differential GPS installations, repeated differential GPS surveys, and visual inspection. 3. Statistical analysis of precipitation records preceding dated landslide events with the aim of determining possible precipitation thresholds that could be used to alert the population of the potential of rainfall-triggered landslides. 4. Detailed geologic and geotechnical characterization of landslides that pose the highest risk in densely populated areas such as Pampahasi, Villa Armonía, Villa San Antonio, and Cota Cota. 5. Detailed investigation of landslides above the Incachaca reservoir, including an assessment of the potential for catastrophic failure, modeling of possible displacement waves, and estimation of dam overtopping and the extent of downvalley flooding.

181

6. Detailed geologic and geotechnical characterization of slow-moving landslides that pose a moderate risk to populations due either to their small size, such as the Santa Barbara and the landslide in northern Seguencoma, or to limited current development, such as the Achocalla earthflow. 7. Analysis of the potential impacts of various earthquake scenarios on the already marginal stability of slopes in and around La Paz.

182

Appendix A

Paleomagnetic directions by unit and section

Table A1. Means of directional data by stratigraphic position and unit for the Patapatani West section.

a Unit Lithology χ n D I k α 95 p Position (m) Material (mean) Collected Useful Used

PTW-20 Altiplano gravel Not sampled

PTW-19 Sorata Drift 230.5 Paleosol 779 * 12 12 12 1.0 -27.0 106 4.2 N 220.5 Silt lens 189 6 6 5 3.2 -33.1 65 9.5 N 220.0 Silt lens 213 6 6 6 5.2 -36.3 72 8.0 N 214.5 Silt lens 294 6 6 5 333.6 -35.5 79 8.6 N 30 30 28 357.7 -32.1 38 4.5 N

PTW-18 Kaluyo Gravels 213.5 Silt lens 431 6 6 6 17.0 -32.1 49 9.6 N

PTW-17 Kaluyo Gravels 187.5 Clay between clasts 299 12 12 10 173.6 25.8 15 13.0 R

PTW-16 Purapurani Gravels 183.5 Sand lens 114 6 6 6 174.0 20.1 76 7.7 R

PTW-15 Calvario Drift 176.0 Diamict 73 6 0 0 Indeterminate polarity - 175.5 Silt bed 104 6 2 2 222.5 45.3 - - R 12 2 2

PTW-14 Calvario Drift 165.5 Paleosol 1054 * 6 6 6 152.3 55.0 19 16.0 R 162.0 Silt bed 700 3 3 3 161.4 37.0 319 6.9 R 9 9 9 156.1 48.9 21 11.6 R

PTW-13 Calvario Drift 161.0 Paleosol 2455 * 3 3 3 348.6 -35.7 106 12.1 N N

PTW-12 Calvario Drift 153.0 Ash lens 2532 3 3 3 357.7 -22.3 36 20.9 N N

183

PTW-11 Calvario Drift 152.5 Paleosol 1633 * 3 3 3 353.7 -42.9 228 8.2 N 143.0 Sand lens 386 6 6 6 3.7 -35.2 289 3.9 N 139.5 Silt/sand lense 168 6 6 6 350.5 -3.5 11 21.6 N 134.0 Diamict & silt lens 164 12 11 8 43.3 -22.2 22 12.2 N 114.5 Silt lens 154 6 6 5 346.4 -47.2 30 14.1 N 109.5 Silt lens 181 6 6 6 12.6 -38.1 35 11.5 N 109.0 Diamict 138 6 6 5 331.0 -62.7 125 6.9 N 108.0 Sand lens 180 6 6 6 354.9 -28.8 45 10.1 N 107.5 Sand lens 151 6 1 1 356.5 -46.8 - - N 57 51 46 4.8 -35.4 9 7.6 N

PTW-10 Chijini Tuff 103.5 Cliff-forming ash 1509 b 6 6 6 18.3 -38.6 121 6.1 N 101.0 Cliff-forming ash 573 6 6 6 10.8 -37.2 90 7.1 N 100.0 Cliff-forming ash 686 6 6 3 0.9 -27.1 364 6.5 N 98.0 Cliff-forming ash 846 6 6 5 10.3 -29.8 83 8.5 N 97.0 Cliff-forming ash 729 6 6 6 15.6 -28.7 69 8.1 N 95.5 Silt-sized ash 407 6 6 5 6.7 -44.1 70 9.1 N 36 36 31 11.5 -34.9 57 3.5 N

PTW-09 Patapatani Drift 94.5 Diamicton 119 6 6 6 0.5 -13.2 37 11.2 N 92.0 Diamicton 74 6 6 6 8.3 -29.6 28 12.9 N 91.0 Diamicton 284 6 6 6 355.9 -36.9 66 8.3 N 18 18 18 1.7 -26.7 24 7.2 N

PTW-08 Patapatani Drift 89.5 Paleosol formed in diamict 2368 * 6 6 6 6.0 -43.6 251 4.2 N 88.0 Diamicton 85 6 6 5 28.4 -27.4 21 17.2 N 12 12 11 17.1 -36.9 22 9.9 N

PTW-07 Patapatani Drift 85.5 Paleosol formed in diamict 5556 * 6 6 6 19.8 -28.0 344 3.6 N 81.5 Diamicton 155 6 4 3 347.7 -23.2 67 15.2 N 12 10 9 9.0 -27.2 27 10.2 N

PTW-06 Patapatani Drift 77.5 Paleosol formed in diamict 2641 * 6 6 4 182.4 42.5 82 10.2 R 74.5 Diamicton 125 6 6 5 166.2 14.6 143 6.4 R 70.0 Sand lens 168 6 4 3 217.0 55.6 42 19.4 R 62.0 Silt lens 162 6 6 4 182.7 1.6 31 16.7 R 24 22 16 180.7 26.8 9 12.9 R

PTW-05 Patapatani Drift 57.5 Paleosol (weakly developed 154 6 4 4 213.3 70.0 8 34.8 R 52.5 Silt lens 100 14 9 6 187.6 28.8 20 15.4 R 20 13 10 192.6 44.9 7 19.5 R

PTW-04 Patapatani Drift 38.0 Silt lens 149 8 8 6 356.3 -28.4 56 9.1 N N

PTW-03 Patapatani Drift 19.0 Silt lens 127 12 9 9 191.1 26.0 8 19.7 R 9.5 Sand lens & diamict 146 16 12 10 164.9 67.5 31 8.9 R 28 21 19 182.2 49.6 7 13.8 R

184

PTW-02 Pre-Patapatani stratified diamict 7.0 Sand lens 293 6 6 6 358.2 -32.4 173 5.1 N 4.5 Silt lens & diamict 158 12 11 8 5.4 -15.7 31 10.1 N 18 17 14 2.5 -23.0 31 7.3 N

PTW-01 Pre-Patapatani gravels 1.5 Gravel matrix 113 6 5 4 359.1 -15.2 15 24.6 N N

See Fig. 2.7 for stratigraphy and stratigraphic position of sample groups. Position, sampling height in metres above base of section; χ, mean magnetic susceptibility of collected samples (x 10-6 SI units,) n, number of samples; D and I, mean declination and inclination, respectively; k, precision parameter; α95, circle of confidence (P = 0.05); p, polarity. * Magnetic enhancement of paleosol compared to material in which if formed. a Error between 10° and 20° underlined, error greater than 20° double underlined. b Apparent magnetic enhancement at top of tuff unit.

185

Table A2. Means of directional data by stratigraphic position and unit for the Patapatani East section.

a a Unit Lithology χ n D I k α 95 p Position (m) Material (mean) Collected Useful Used

PTE-08 Calvario Drift 39.0 Fine sand lens 344 6 6 6 349.8 -24.1 76 7.7 N N

PTE-07 Calvario Drift 37.0 Silt lens 206 6 6 4 339.2 -32.3 56 12.4 N N

PTE-06 Chijini Tuff 31.5 Pumacious ash 636 6 6 6 359.7 -38.2 159 5.3 N N

PTE-05 Patapatani Drift Not sampled

PTE-04 Patapatani Drift 28.0 Silt lens 130 12 12 10 10.9 -33.2 26 9.6 N 12 12 10 N

PTE-03 Patapatani Drift Paleosol Not sampled

PTE-02 Patapatani Drift 6.5 Silt lens 152 6 6 5 146.9 19.4 46 11.4 R R

PTE-01 Patapatani Drift 2.5 Fine sand lens 131 6 6 6 185.0 19.7 25 13.7 R 2.5 Fine sand lens 131 6 6 4 211.2 32.3 163 7.2 R 12 12 10 194.6 25.4 19 11.4 R

See Fig. 2.8 for stratigraphy and stratigraphic position of sample groups. Position, sampling height in metres above base of section; χ, mean magnetic susceptibility of collected samples (x 10-6 SI units,) n, number of samples; D and I, mean declination and inclination, respectively; k, precision parameter; α95, circle of confidence (P = 0.05); p, polarity. * Magnetic enhancement of paleosol compared to material in which if formed. a Error between 10° and 20° underlined, error greater than 20° double underlined.

186

Table A3. Means of directional data by stratigraphic position and unit for the Tangani section.

a a Unit Lithology χ n D I k α 95 p Position (m) Material (mean) Collected Useful Used

TNG-13 Valley-slope coverb Variable Gravel matrix 76 6 6 3 328.0 -28.9 18 29.9 N N

TNG-12 Purapurani Gravel 193.0 Silt bed 112 6 6 4 175.2 30.7 113 8.7 R 186.0 Silt bed 128 6 6 5 180.7 28.8 50 10.9 R 183.5 Silt lens 271 6 6 6 180.1 32.5 53 9.3 R 18 18 15 179.0 30.8 65 4.8 R

TNG-11 Purapurani Gravel 179.0 Gravel matrix 286 6 6 6 349.0 -18.6 60 8.7 N N

TNG-10 Purapurani Gravel 158.0 Sand lens 439 6 6 6 165.8 25.6 90 7.1 R 142.0 Sand lens 76 6 6 5 175.8 39.8 93 8.0 R 12 12 11 169.9 32.1 47 6.7 R

TNG-9 Purapurani Gravel 141.0 Sand lens 1248 * 6 6 6 138.1 43.0 73 7.9 R 129.0 Sand lens 23 6 6 4 167.2 21.3 96 9.4 R 12 12 10 151.5 35.2 19 11.5 R

TNG-08 Calvario Drift (diamicton) 127.5 Clay lens & diamict matrix 403 * 12 12 11 187.5 26.3 23 9.8 R 125.0 Silt lens 93 6 0 0 Indeterminate polarity - 18 12 11 187.5 26.3 23 9.8 R

TNG-07 Calvario Drift (diamicton) 121.0 Diamict matrix 97 6 0 0 Indeterminate polarity - -

TNG-06 Calvario Drift (gravel) 101.0 Silt lens 42 6 6 6 181.9 42.3 47 9.8 R 108.0 Fine-sandy silt bed 78 12 12 10 187.2 16.1 35 8.2 R 105.0 Silt & sand lens 46 6 6 6 177.8 24.6 47 9.9 R 102.5 Sand lens 85 9 9 8 173.9 19.4 42 8.6 R 33 33 30 180.7 23.9 25 5.4 R

TNG-05 Calvario Drift (diamicton) 101.0 Fine sand lens 173 6 6 6 177.7 24.9 246 4.3 R 105.0 Silt & sand lens 153 6 6 2 181.4 27.9 - - R 12 12 8 178.6 25.7 255 3.5 R

TNG-04 Calvario Drift (diamicton) 104.5 Sand bed 804 * 6 6 5 342.9 -41.6 21 17.0 N 86.0 Silt lens 218 6 6 6 354.7 -11.4 76 7.7 N 74.0 Silt lens 218 6 6 5 336.3 -20.5 24 15.9 N 70.0 Silt lens 184 6 6 6 339.6 -18.6 63 8.5 N 66.0 Sand lens 113 5 5 5 4.5 -23.9 72 9.1 N 52.5 Silt lens 133 5 5 4 9.1 -37.4 44 14.0 N 49.0 Silt lens 169 6 6 6 0.3 -13.6 37 11.2 N 44.0 Silt lens 172 6 6 5 2.6 0.1 36 12.9 N 20.0 Silt lens 178 6 6 4 0.6 -2.8 38 15.1 N

187

TNG-03 Calvario Drift (gravel) 13.0 Silt lens 686 6 6 6 352.3 -29.7 85 7.3 N N

TNG-02 Calvario Drift (diamicton) 8.0 Silt bed 165 6 6 6 346.5 -33.8 425 3.3 N N

TNG-01 Calvario Drift (gravel)

4.5 Silt lens 188 6 6 6 356.3 -19.0 46 9.9 N 2.5 Fine sand lens 284 6 6 5 3.8 -35.7 62 9.8 N 12 12 11 359.5 -26.2 33 8.1 N

See Fig. 2.6 for stratigraphy and stratigraphic position of sample groups. Position, sampling height in metres above base of section; χ, mean magnetic susceptibility of collected samples (x 10-6 SI units,) n, number of samples; D and I, mean declination and inclination, respectively; k, precision parameter; α95, circle of confidence (P = 0.05); p, polarity. * Magnetic enhancement of paleosol compared to material in which if formed. a Error between 10° and 20° underlined, error greater than 20° double underlined. b Colluvium drapes incised valley slope (possible mass flow deposit). Not considered in overall statistics as this unit is not part of the Altiplano fill sequence.

188

Table A4. Means of directional data by stratigraphic position and unit for the Minasa section.

a a Unit Lithology χ n D I k α 95 p Position (m) Material (mean) Collected Useful Used

MIN-15 Kaluyo Gravel/Sorata Drift (gravel to diamicton) 410.0 Modern soil (in diamicton) 1354 * 6 6 5 1.5 -33.7 183 5.7 N N MIN-14 Kaluyo Gravel/Sorata Drift (gravel to diamicton) 406.0 Paleosol (in sand lens) 1309 * 6 6 6 354.9 -24.8 119 6.2 N N MIN-13 Kaluyo Gravel/Sorata Drift (diamicton) 402.5 Sand lens 323 6 6 6 6.1 -21.5 57 8.9 N N MIN-12 Kaluyo Gravel/Sorata Drift (gravel to diamicton) 378.0 Sand lens 111 6 6 6 350.7 -22.6 26 13.3 N N MIN-11 Kaluyo Gravel/Sorata Drift (gravel) Not sampled MIN-010 Kaluyo Gravel/Sorata Drift (gravel to diamict Not sampled MIN-09 Kaluyo Gravel/Sorata Drift (gravel to diamict Not sampled

MIN-08 Kaluyo Gravel/Sorata Drift (diamicton) 324.0 Sand lens 125 6 6 5 356.1 -24.4 77 8.8 N N MIN-07 Kaluyo Gravel/Sorata Drift (gravel) Not sampled MIN-06 Kaluyo Gravel/Sorata Drift (diamicton) Not sampled

MIN-05 Kaluyo Gravel/Sorata Drift 324.0 Silt lens 91 6 5 4 350.3 -24.8 52 12.8 N N MIN-04 Kaluyo Gravel/Sorata Drift Not sampled

MIN-03 Purapurani Gravel 278.0 Silt bed 337 6 0 0 Indeterminate polarity - 275.0 Silt bed 71 6 6 4 187.3 31.2 50 13.1 R 234.0 Silty sand lens 71 6 6 6 176.6 38.9 66 8.3 R 214.0 Sand bed 82 6 6 4 176.8 40.0 73 10.8 R 24 18 14 179.9 37.1 55 5.4 R MIN-02 Multi-lithic gravel 144.0 Paleosol (in gravel) 1926 * 6 6 5 143.7 38.2 93 8.0 R 112.0 Silt lens 102 6 6 4 175.8 24.7 212 6.3 R 10.5 Medium sand lens 107 6 0 0 Indeterminate polarity - 18 12 9 159.1 33.2 22 11.2 R

MIN-01 Chijini Tuff 0.0 Silt-sized ash 85 6 6 6 259.6 -35.8 29 12.7 N N

See Fig. 2.10 for stratigraphy and stratigraphic position of sample groups. Position, sampling height in metres above base of section; χ, mean magnetic susceptibility of collected samples (x 10-6 SI units,) n, number of samples; D and I, mean declination and inclination, respectively; k, precision parameter; α95, circle of confidence (P = 0.05); p, polarity.

189

* Magnetic enhancement of paleosol compared to material in which if formed. a Error between 10° and 20° underlined, error greater than 20° double underlined.

190

Table A5. Means of directional data by stratigraphic position and unit for the Purapura section.

Unit Lithology χ n D I k α 95 a p a Position (m) Material (mean) Collected Useful Used

PUR-15 Altiplano surface gravels Not sampled

PUR-14 Sorata Drift (diamicton) 246 Silt lens 210 8 8 8 345.8 -25.0 112 5.2 N 244 Silt lens 525 6 6 6 1.0 -29.0 317 3.8 N 14 14 14 352.2 -26.9 72 4.7 N

PUR-13 Sorata Drift (diamicton) 240 Palsosol? 1461 * 8 8 8 358.0 -30.2 231 3.7 N N

PUR-12 Sorata Drift (diamicton) Not sampled

PUR-11 Sorata Drift, (Gravel) Not sampled

PUR-10 Sorata Drift (diamicton) 192 121 6 6 6 351.6 -21.6 49 9.7 N

PUR-09 Purapurani Gravel 58.5 56 6 0 0 Indeterminate polarity -

PUR-08 Calvario Drift 58 115 6 5 4 30.1 -39.5 40 14.7 N

PUR-07 Calvario Drift 57.5 Palsosol 342 6 6 4 3.2 -38.8 60 11.9 N 57.5 Palsosol 417 6 6 6 343.5 -53.3 24 13.9 N 12 12 10 352.8 -47.9 23 10.4 N

PUR-06 Calvario Drift 54 Palsosol 1631 * 3 3 3 351.2 -28.9 402 6.2 N 23.5 Possible ash 1360 * 7 7 6 337.5 -16.4 1.5 N b 10 10 9 Mix of PCA and GC N

PUR-05 Calvario Drift 13.5 84 6 0 0 Indeterminate polarity - -

PUR-04 Calvario Drift (silt and sand) Not sampled

PUR-03 Chijini Tuff 3.5 Tuff 1727 c 6 6 6 357.6 -54.9 73 7.9 N N

PUR-02 La Paz Formation, possible pyroclastic flow 2.5 Silt bed 102 3 3 3 16.1 -29.5 235 8.1 N 2 Silt bed 91 6 6 6 355.4 -30.2 282 4.0 N 9 9 9 2.3 -30.3 65 6.4 N

PUR-01 La Paz Formation (gravel)

See Fig. 2.11 for stratigraphy and stratigraphic position of sample groups. Position, sampling height in metres above base of section; χ, mean magnetic susceptibility of collected samples (x 10-6 SI units,) n, number of samples; D and I, mean declination and inclination, respectively; k, precision parameter; α95, circle of confidence (P = 0.05); p, polarity. * Magnetic enhancement of paleosol compared to material in which if formed.

191

a Error between 10° and 20° underlined, error greater than 20° double underlined. b Remanence from intersection of Great Circles. c Unusually steep remanence direction suggest possible slumping or orientation error. Remanence directions of underlying, loose basal ash was therefore used for comparisons.

192

Table A6. Means of directional data by stratigraphic position and unit for the Jacha Kkota section.

Unit Lithology χ n D I k α 95 a p a Position (m) Material (mean) Collected Useful Used

JKT-08 Altiplano surface gravel 178.0 Silt lens 194 6 6 5 355.4 -26.4 129 6.8 N N

JKT-07 Sorata Drift (diamicton) 175.0 Silt lens 133 6 6 4 352.4 -7.4 7.9 N b N

JKT-06 upper La Paz Formation Not sampled

JKT-05 Chijini Tuff 43.5 Cemented tuff 2122 6 6 6 357.0 -37.3 108 6.5 N 42.5 Loose ash 197 6 6 6 2.3 -28.9 258 4.2 N 12 12 12 359.8 -33.1 103 4.3 N

JKT-04 39.5 Silt bed 108 6 6 4 349.1 -36.8 1239 2.6 N 32.5 Fine sand bed 280 6 6 6 344.6 -26.1 54 9.2 N 12 12 10 348.3 -30.4 63 6.1 N

JKT-03 upper La Paz Formation 22.0 Paleosol 565 * 6 6 5 354.7 -34.5 49 9.6 N N

JKT-02 upper La Paz Formation 13.0 Silt lens 262 6 6 5 353.7 -23.2 98 8.2 N N

JKT-01 upper La Paz Formation 5.5 Silt bed 157 6 5 4 187.9 38.2 95 9.5 R R

See Fig. 2.12 for stratigraphy and stratigraphic position of sample groups. Position, sampling height in metres above base of section; χ, mean magnetic susceptibility of collected samples (x 10-6 SI units,) n, number of samples; D and I, mean declination and inclination, respectively; k, precision parameter; α95, circle of confidence (P = 0.05); p, polarity. * Magnetic enhancement of paleosol compared to material in which if formed. a Error between 10° and 20° underlined, error greater than 20° double underlined. b Remanence directions obtained from intersection of Great Circles.

193

Appendix B

Clast fabrics

Figure B1. Diamicton clast fabrics from sections north of La Paz.

194

Appendix C

Tuff comparisons

All previous workers recognize a single tuff – the Chijini Tuff – throughout the southern part of the upper Río La Paz valley system. However, they disagree as to whether one (Alhfeld, 1946; Dobrovolny, 1962) or multiple tuffs (Servant, 1977; Clapperton, 1979; Thouveny and Servant, 1989) are present north of the city, where the Chijini Tuff is difficult to trace due to extensive colluvial cover (Dobrovolny, 1962) and numerous high- angle faults (Dobrovolny, 1962; Lavenu, 1977; Lavenu et al., 2000). Arguments supporting the presence of a second tuff are based largely on reported K-Ar ages (Clapperton, 1979) or magnetic intensities (Thouveny and Servant, 1989) that differ from those established for the Chijini Tuff.

Five of the six sections sampled for the present study include a thick rhyolitic tuff located 60-420 m below the Altiplano surface. The elevation of the tuff exposures generally decreases from the Patapatani West section in the north (~4260 m asl) to the Jacha Kkota section in the south (~3840 m asl). Tuff thickness at the sections ranges from 5 m (Patapatani East, Purapura, Jacha Kkota) to 10 m (Patapatani West, Minasa, Viscachani section of Thouveny and Servant, 1989), but with no apparent spatial trend. At all five of my study sections, it comprises a loose basal ash several decimetres thick overlain by cliff-forming weakly cemented tuff. Stratigraphic correlation of these sections and interpretations of the fill sequence will differ depending on whether these tuff exposures record a single unit as suggested by Alhfeld (1946) and Dobrovolny (1962), or whether an additional tuff (Servant, 1977; Clapperton, 1979; Thouveny and Servant, 1989) is present in the Río Kaluyo valley.

Multiple lines of evidence (Fig. 2.13) suggest, with a high level of certainty, that the tuff exposures considered here record the same eruption (correlation F in Fig. 2.14). No strong evidence suggests that there is another tuff along the Río Kaluyo/Choqueyapu valley or in the Achocalla basin. Tuff exposures can thus be used as a primary means of chronostratigraphic correlation between sections in the western upper Río La Paz valley system.

Magnetic intensity

Thouveny and Servant (1989) believe they had sampled the Sopari Tuff in the Río Kaluyo valley and the Chijini Tuff farther south. They note that the tuff in the Río Kaluyo valley has magnetic intensity an order of magnitude higher (2 mA/m-1) than the Chijini Tuff (0.2 mA/m-1) at their Achocalla section. Because they found both tuffs to have similar magnetic susceptibly, difference in magnetic intensity implies different intensities of the geomagnetic field at the time the magnetic intensity of the materials was acquired (cf. Opdyke and Channell, 1996).

195

The sample collection considered here shows no systematic difference between magnetic intensity of the tuffs in these two areas. At Patapatani West, where the tuff was sampled at a large number of elevations, magnetic intensity differs greatly with stratigraphic position and, to a lesser degree, among samples from the same stratigraphic position. Although the single highest magnetic intensity value I measured (1350 mA/m-1) is from the tuff at the Patapatani West section, all other sample groups for tuffs, both in the Río Kaluyo valley and farther south, have range of values from 0.2 mA/m-1 to 300 mA/m-1. In any case, magnetic intensity alone is insufficient evidence to correlate or differentiate tuffs.

Additionally, samples of the tuff from the Patapatani West section show a wide range of magnetic susceptibilities (60 to 1925 x 10-6 SI units), further indicating high variably of magnetic properties even within the same exposure. Magnetic susceptibility near the surface of the tuff (~1 m below the upper contact) exceeds all other values for the tuff at the same section by a factor of two to three. Magnetic susceptibilities of the six samples of the uppermost sample group from the tuff at this section show a systematic decrease with depth. This trend of magnetic enrichment at the upper contact of the tuff at the Patapatani West section likely reflected sub-aerial weathering over a long period (cf. Opdyke and Channell, 1996, p. 46).

Radiometric ages

The five most reliable tuff ages – those on sanidine or potassium feldspar are of roughly overlapping range (Table 2.3) – from the La Paz valley system and the Achocalla basin yield an average age of 2.74 Ma, identical to a newly acquired age on tuff exposed in Río Kaluyo (2.74 ± 0.04 Ma) where Thouveny and Servant (1989) propose a younger tuff. The agreement in ages suggests the tuff exposed throughout the valley is the same, at least at the resolution of the K-Ar (ca. 105 a) and 40Ar/39Ar (ca. 104 a) methods. Potassium-argon analysis of potassium feldspar from two closely spaced exposures at approximately the same location as the new 40Ar/39Ar age (Fig. 7 of Lavenu et al., 1989) gave ages of 2.7 ± 0.1 Ma and 2.8 ± 0.1 Ma (Lavenu et al., 1989; MB153, MB154). No stratigraphically higher tuff is exposed in the Río Kaluyo valley. Farther down valley where the tuff is indisputably the Chijini, K-Ar analysis on potassium feldspar from the tuff at the approximate location of the Purapura section gave an age of 2.8 ± 0.1 Ma (Lavenu et al., 1989; MB159); and 40Ar/39Ar analysis of sanidine in tuff exposed in the Achocalla basin ~7.5 km southeast of the Jacha Kkota section gave an age of 2.757 ± 0.053 Ma (Marshall et al., 1992; LGM03). Overlap in these ages does not preclude the possibility of two eruptive events within a short period (~200 ka), but this scenario is improbable as superimposed tuffs are not observed anywhere in the La Paz area and because closely spaced eruptions in the Cordillera Oriental large enough to emplace tuffs 10 m thick at La Paz are highly unlikely.

The K-Ar age on which the Sopari Tuff is based (1.6 ± 0.1 Ma; Lavenue et al., 1989; PH53a) came from plagioclase from an exposure of tuff in the Río Chuquiaguillo valley, just east of the Río Kaluyo valley and approximately 2 km northeast of the Minasa section. Subsequently, a 40Ar/39Ar age of 2.650 ± 0.037 Ma was obtained on sanidine from a tuff at the same location (Marshall et al., 1992; LGM01). A K-Ar age on sanidine

196

farther downstream, ~0.5 km southwest of the base of the Minasa section gives a younger, non-overlapping age (Marshall et al., 1992; LGM02) that is likely erroneous.

Radiometric dating campaigns on the Bolivian Altiplano (Evernden et al., 1977; Lavenu et al., 1989; Marshall et al., 1992) have shown that biotite ages throughout the region, and in the upper La Paz basin in particular, are typically anomalously old, which likely explains the K-Ar age of biotite reported by Clapperton (1979; 3.27 ± 0.14 Ma and 3.28 ± 0.13 Ma) on the east slope of the Río Kaluyo valley. Examples of disparate ages from different minerals include K-Ar ages on biotite and potassium feldspar (MB159 and MB160) with identical coordinates and elevation reported by Lavenu et al. (1989) and 40Ar/39Ar ages on biotite and sanidine (LGM01 and LGM02) by Marshall et al (1992) (Table 2.3). Three of the four biotite ages from Lavenu et al. (1989) and Marshall et al. (1992) fall within the Miocene and are clearly erroneous. Thouveny and Servant (1989) suggest incorporation of older biotites into the ignimbrite deposits as a cause of the old ages. Marshall et al. (1992) suggest instead that feldspar-biotite age discrepancies, specifically for the Chijini Tuff (Marshall et al., 1992, p. 11), result from differential potassium/argon loss during alteration. The biotite-feldspar age differences on the Chijini Tuff are in general agreement with results of controlled experiments by Smith et al. (2008) from the North American Cordillera in which altered biotite yields ages 1-14% older than sanidine.

Magnetic remanence

Where measured, the polarities of all tuffs in La Paz and Achocalla basins are normal, as are units directly overlying and underlying them (Fig. 2.14). The early Matuyama ages (nos. 9 and 12 in Fig. 2.13) are not possible, and the pre-late Gauss ages (nos. 1, 2, and 11 in Fig. 2.13) are unlikely, for tuffs considered here, because those five radiometric ages correspond to reversely magnetized periods. Thus, polarity data reinforce the conclusion that the tuffs were deposited in the late Gauss (3.045-2.580 Ma), although they do not improve age constraints beyond the most reliable radiometric ages.

Secular variation of Earth’s magnetic field over short periods (102 to 103 years; Opdyke and Channell, 1996) produces annual changes of the local geomagnetic field on the order of fractions of a degree. Bogue and Coe (1981) suggest that the resulting directional shifts are detectable paleomagnetically only with the passage of time on the order of 10 to 100 years. Comparison of the remanence directions recorded by rapidly deposited sediments can thus reveal whether the orientation of Earth’s geomagnetic field was significantly different due to secular variation during the span of their deposition. Units representing the same geologic instant should have the same directions, whereas those emplaced on the order of a decade to a century or more apart will typically have different remanence directions.

The remanence direction of all tuff samples is nearly identical at the Jacha Kkota, Purapura, Minasa, and Patapatani East sections (clustered at 0.4°declination, -34.4° inclination, a95 = 4.1°), suggesting they were simultaneously magnetized (i.e. in the presence of the same ancient field). Only remanence directions at the Purapura and

197

Patapatani East sections are possibly distinctly different at the 95% confidence level (Fig. 2.13a), but still closely aligned. It is possible that these magnetizations record ancient fields from widely spaced periods that happen to have similar orientations, particularly because they align near the expected time-averaged GAD field direction (Bogue and Coe, 1981). However, I view this as extremely unlikely given the infrequency of large eruptions in the Cordillera Occidental capable of reaching La Paz.

The tightly clustered mean of the remanence directions for the tuff at the Patapatani West section is 11° east of the overall mean of the other four sections (11.5°declination, -34.9° inclination, a95 = 3.5°) and distinctly different from all but the tuff at the Minasa section (Fig. 2.13a). This difference likely is the result of minor fault-block rotation. Several faults are visible along the 500 m length of discontinuous exposure of the tuff in natural slopes and road cuts on the west side of the Río Kaluyo valley, including exposures sampled at the Patapatani West section. A sample group of three samples on each side of one such fault gives different and well clustered remanence values (Fig. 2.13a).

Thouveny and Servant (1989, p. 341) similarly found that the tuffs they sampled in the Río Kaluyo valley and the Achocalla basin cannot be differentiated on the basis of directional data, although they report remanence directions only for the latter site where the tuff is known to be the Chijini. Remanence directions reported here include both the portion of the Río Kaluyo valley where Thouveny and Servant (1989) propose the Sopari Tuff (Patapatani East section) and their section in the Achocalla basin (Jacha Kkota section).

198

Appendix D

Correlations of paleosols using paleomagnetic directions

Paleosols throughout the sedimentary sequence indicate long periods of subaerial exposure, followed by deposition without substantive erosion. All paleosols identified here (Figs. 2.4-2.12 and 2.14) have a variety of features characteristic of pedogenesis, including laterally continuous, sharp, planar upper surfaces, pedon formation, clay skins on clasts and pedon surfaces, orange and red hues that gradually decrease with depth, and upward fining from the parent material interface (Brady and Weil, 2002). They also have magnetic susceptibilities that are typically twice or more that of the underlying materials they formed in, suggesting authigenic ferromagnetic mineral production. Similar ‘magnetic enhancement’ of paleosols in loess sequences is generally attributed to chemical weathering caused by increased rainfall and higher temperatures during interglaciations (Opdyke and Channell, 1996, p. 46).

The paleosols offer an opportunity to further refine magnetostratigraphic correlations because the depositional hiatuses they represent are likely recorded at adjacent sections. Identical remanences of buried soils may indicate coeval deposition of parent materials or, where similar secondary components of magnetization are present, a common period of pedogenesis.

The paleomagnetic records held by unconsolidated sediments are generally of a lower quality than those held by volcanic rocks. In sediments, remanence is complicated by the presence of multi-domain magnetic grains and coarse lithic grains whose orientation is the result of depositional agents (i.e. gravity, water, wind, ice). Where sediment grains are predominantly coarse, mechanical energies begin to outweigh the aligning influence of the geomagnetic field on ferromagnetic particles (Butler, 1992). In paleosols, in particular, weathering and leaching may produce a secondary (chemical) remanent magnetization. This generally results from alteration, translocation, and precipitation of new ferromagnetic minerals. Because these processes occur over longer periods and at varying rates, the resulting secondary remanence components may be ‘noisier’ than the primary detrital remanent magnetization, and in some cases, may partially or completely obscure the primary remanence. Primary and secondary components of magnetization can generally be isolated through alternating field or thermal demagnetization.

Consequently, statistical analysis of possible differences in magnetic directions of sediments may in some cases be less reliable than it is for volcanic rocks (cf. Bogue and Coe, 1981) or even volcanic sediments composed of high susceptibility grains of fine uniform size, such as in the Chijini Tuff (Appendix C). Despite these challenges, I was able to identify six stratigraphic levels at which paleosols in two or more sections have similar primary remanence directions (Fig. 2.15), suggesting deposition and soil formation at the same time (correlations D, E, G, H, K, and N in Fig. 2.14):

199

Correlation D

The polarity reversal below the Chijini Tuff occurs at a paleosol formed in and overlain by till at the Patapatani West section and at a wavy contact between silt beds at the Jacha Kkota section 18 km south (Fig. 2.14, correlation D). Remanence directions above and below the contacts at the two sites are not statistically different (Fig. 2.15a). Consistency in paleomagnetic records on opposite sides of this polarity reversal suggests that deposition ceased and resumed coevally at both sections.

Correlation E

A well developed paleosol is present above the polarity reversal mentioned above, but below the Chijini Tuff, at the Patapatani East and Jacha Kkota sections, although the former was not sampled (Fig. 2.14, correlation E). Remanence of the paleosol at the Jacha Kkota section is similar to that of the upper of the two paleosols at a similar magnetostratigraphic position at the Patapatani West section; however, it is distinguishable at 95% confidence from the remanence direction of both paleosols at Patapatani West (Fig. 2.15b). However, relative to the underlying polarity reversal, the paleosol at the Jacha Kkota section is most similar in stratigraphic position to the lower paleosol at Patapatani West (Fig. 2.14). It is possible that the directional data obtained for the samples of the lower Patapanti West paleosol was not fully resolved with the available demagnetization levels, and that it in fact correlates with the Jacha Kkota paleosol. Given the nearly identical depth of the paleosol below the tuff on both sides of the Río Kaluyo valley, the sub-aerial paleosurface they record likely extended north- south at least 1 km, and may have formed at the same time as the surface recorded by the paleosol at the Jacha Kkota section in the modern Achocalla basin (Fig. 2.12).

Correlations G and H

A paleosol in the upper portion of the same magnetozone as the tuff at the Patapatapani West section has similar remanence directions to the uppermost paleosol in the same magnetozone at the Purapura section (Fig. 2.14, correlation G; Fig. 2.15c). A similarly positioned paleosol at the Tangani section has nearly identical declination values to those measured for the Patapatani West and Purapura paleosols, but exhibits very shallow inclination values (Fig. 2.15c) closely aligned with present Earth’s field directions. The shallow inclination directions may be the result of present field overprinting.

Temporal coincidence of these paleosols is supported by complementary remanence directions for the overlying paleosol (Fig. 2.14, correlation H). The polarity reversal ~8 m above correlation G is marked by paleosols with statistically similar (slightly westerly) remanence directions at the Patapatani West and Tangani sections (Fig. 2.15d). Remanence from the top of the same magnetozone at the Purapura section is very different, which is likely due to erosion of sediments below the base of the granitic gravel unit (Purapurani Formation). The contact between this gravel and the underlying till

200

appears erosive at this section and the reversely magnetized till below the gravel is missing.

Correlation K

Reversely magnetized paleosols within the Purapurani Gravel at the Tangani and Minasa sections have similar, easterly remanence (Fig. 2.15e). Magnetization records of samples from both paleosols are of moderate quality (e.g. Fig. 2.4h; Minasa section, sample 612) and have typical group mean errors for non-volcanic sediments in the Altiplano sequence. Their exotic position ~10° steeper than the GAD and ~30-40° easterly [Fig. 2.15e]) suggests that the close remanence agreement reflects simultaneously acquired detrital remnant magnetization, rather than chance deposition during two separate periods with the same geomagnetic field orientations (Bogue and Coe, 1981). The tuff is absent from the Tangani section, but the height of the paleosol above its inferred stratigraphic position is the same as the height of the likely correlative paleosol above the tuff at the Minasa section.

Correlation N

A strongly oxidized paleosol 1-8 m below the Altiplano surface at all four sections reaching the Altiplano (Fig. 2.14) is present in many parts of the La Paz and Achocalla basins. Remanence of the paleosol is nearly horizontal at the Jacha Kkota section and closely aligns with the present Earth’s field direction, suggesting a possible recent overprint. Primary remanence directions at the other three sections are almost the same, and are close to the geocentric axial dipole (Fig. 2.15f); they are not significantly different at the Patapatani West and Purapura sections, and just significantly different at the Minas section.

201

Appendix E

Polarities and ages of previously defined geologic units

Detailed, overlapping magnetostratigraphy from the study sites in the La Paz and Achocalla basins shows specific polarity signatures and likely ages of the main geologic units of the Altiplano sediment sequence (Table 2.4). In several cases, these signatures provide refined age estimates for units defined by previous workers.

Patapatani Drift

The full polarity sequence (N-R-N-R-N) of the pre-Chijini glacial deposits is based on the Patapatani West section. The lowermost and uppermost polarity reversals at the Patapatani West section occur at an angular unconformity and paleosol, respectively. The other two reversals at this section and the only polarity reversal at the Patapatani East section are covered. Only the upper two magnetozones (R-N) are exposed at the Patapatani East section. These tills are not exposed at any other section and, other than Clapperton’s (1979) ~2 m section located between the Tangani section and Thouveny and Servant’s (1989) Viscachani section, have not been observed elsewhere in the area.

Other descriptions of the Patapatani Drift (Dobrovolny, 1956, 1958, 1962; Clapperton, 1979) lack evidence for more than a single pre-Chijini glacial event, largely because they only include the uppermost 2-7 m of the drift sequence. The multiple paleosols and polarity reversals documented here from newly created exposures indicate the till sequence extends farther back in time than previously thought and includes several previously unrecognized glacial events. Based on correlation with the GMPT (Fig. 2.16), deposition of the Patapatani Drift initiated shortly before the start of the Mammoth subchron (3.319 Ma, Lisiecki and Raymo, 2005; 3.330 Ma, Ogg, 2012) and ended prior to emplacement of the Chijini Tuff at ca. 2.74 Ma.

Chijini Tuff

Close agreement of the magnetic remanence directions of tuff exposures throughout the western part of the upper Río La Paz valley system suggests the Chijini Tuff records a single large eruption. The tuff is normally magnetized, in agreement with a latest Gauss age suggested by the most recent high-resolution age on feldspar (2.74 ± 0.4 Ma). All radiometric age determinations on potassium feldspar from the Río Kaluyo/Choqueyapu valley and the Achocalla basin suggest a single event centered at ca. 2.74 Ma. Comparison of glacial chronostratigraphy and the marine oxygen isotope record supports these radiometric ages; the number of normally magnetized till units overlying the tuff (two) agrees with the number of cold peaks between 2.74 Ma and the end of the late Gauss (MIS G6/G4 and G2; Fig. 2.16).

202

The age of the tuff can be further refined based on the marine oxygen isotope record. The tuff was deposited on an ice-free surface. Near the Andes, this surface was a till plain, suggesting an interglacial environment. Magnetic ‘enhancement’ near the top of the tuff (Table A1) is similar to that documented for the interglacial soils at the top of non-volcanic units in the La Paz area and for interglacial soils in the Chinese loess sequences (Opdyke and Channell, 1996, p. 46). It thus appears that the surface of the tuff was exposed subaerially for a long period prior to deposition of the overlying till at the base of the Calvario Drift sequence. Because the likely age of the tuff (2.74 ± 0.4 Ma) spans an extended warm period (G9-G7) and the subsequent strong cold peak (G6) (Fig. 2.16), the mega-eruption that produced the Chijini Tuff most probably occurred between 2.78 and 2.74 Ma.

Calvario Drift

The till sequence overlying the tuff spans two magnetozones, but those zones do not correspond to Dobrovolny’s (1962) stratigraphic division of the Calvario Drift into a lower sand/gravel unit and an upper till unit. The lower, normally magnetized part of the sequence is capped by a paleosol and ranges from 57 m thick (Patpatani West section) to 94 m thick (Tangani section) (Fig. 2.14). At both sections, the normally magnetized portion of the Calvario Drift is composed entirely of till. The overlying reversely magnetized part is a much thinner unit, ranging from 15 m thick at the Patapatani West section where it is composed of till interrupted by a paleosol, to 34 m thick at the Tangani section where it is interrupted by two paleosols directly underlying and ~5 m above a thick gravel bed (Fig. 2.14). The lower subunit is present at the Purapura section, but may have been partially eroded by the overlying gravel. Given the limited magnetostratigraphy at the Purapura section, it is unclear if the overlying granite-rich gravels are part of the upper Calvario subunit or are the basal beds of the Purapurani Gravel.

The occurrence of a magnetic reversal, with multiple paleosols both above and below it, suggests that the Calvario Drift spans a long period of time punctuated by several glacial events. On the basis of the ca. 2.74 Ma age of the Chijini Tuff, and its N-R polarity sequence, the Calvario Drift spans the end Pliocene and the earliest Pleistocene.

The polarity sequences of the Calvario Drift (N-R) and the Patapatani Drift (N-R-N-R-N) provide further evidence that that these are distinct lithostratigraphic units, as suggested by Dobrovolny (1962). The alternative interpretation, that Dobrovolny’s (1962) Patapatani Drift is the same as the Calvario Drift (Servant, 1977; Ballivian et al., 1978; Thouveny and Servant, 1989) would require both drift sequences to have the same polarity record, which they do not.

Purapurani Gravel

The thick Purapurani gravel sequence through the La Paz basin is distinguished by a dominance of well sorted, rounded to subrounded, largely granitic pebbles and cobbles.

203

At the Patapatani West and Minasa sections, the sequence is reversely magnetized at all sampling sites, indicating a Matuyama age. Thus far, only a single paleosol has been identified within the unit (Fig. 2.14c,d). The gravels perhaps record a shorter period that the drift sequence below, but it comprises at least two periods of deposition separated by a hiatus of least 104 years.

The presence of a thin, normally magnetized zone at the Tangani section and at Thouveny and Servant’s (1989) Purpura section provides the only basis for constraining the age of the gravels within the Matuyama Chron. With the preferred correlation to the GMPT, the Purapurani Gravel spans the Réunion subchron, (ca. 2.14) and, given the relative thickness of the normally magnetized zone, record at least 100 ka of deposition. Alternatively, the unit may record the Olduvai subchron, and thus record deposition at ca. 1.9 Ma.

Limited recording of this normal magnetozone likely reflects the low sampling density within the Purapurani Gravel at most sections. Based on its stratigraphic position at the nearby Tangani section, the normally magnetized magnetozone is likely ~30 to 40 m above the paleosol in the middle of the unit at Minasa section.

Kaluyo Gravel

The polarity of the Kaluyo Gravel (the lower part of Dobrovolny’s [1962] Milluni Drift), particularly near the Cordillera Real, is difficult to measure because of the coarseness of the unit and the near absence of fine lenses. North of La Paz (Patapatani West section), its polarity transitions from reversed at its base to normal at its top, but the position of the polarity reversal within the unit is poorly constrained. The polarity change marks the start of the subchron during which the Altiplano surface formed – either the Jaramillo or, more likely, the Olduvai. In the latter case, deposition of the Kaluyo Gravel probably began not long before 1.968 Ma and ended before 1.781 Ma. Since the overlying Sorata Drift and Altiplano surface gravel were deposited later in the same subchron, the Kaluyo Gravel most likely ceased deposition not long after 1.968 Ma and spanned about 100 ka.

Sorata Drift

The Sorata Drift (Bles et al. [1977]; the upper part of Dobrovolny’s [1969] Milluni Drift) is normally magnetized. It records at least one glaciation at most sections and at least three glaciations at the Minasa section. It was deposited during either the Olduvai subcron or the much shorter Jaramillo subchron. Since the Olduvai subchron contains ample cold peaks (4 pronounced ones and several smaller ones) to explain the multiple glaciations at the Minasa section, the Sorata Drift was most likely emplaced during the Olduvai subchron. Based on the number of glaciations suggested and the presence of the Kaluyo Gravel below and Altiplano surface gravel above, the Sorata Drift probably spans more than 100 ka starting after about 1.95 Ma.

204

Altiplano surface gravel

Like the Sorata Drift below it, the gravel unit directly underlying the Altiplano surface is normally magnetized. It is either >1.781 or >0.991 Ma, depending on whether this normal subchron is the Olduvai or Jaramillo. Given the presence of multiple glaciations earlier in the same magnetozone, these gravels likely occur near the end of the subchron.

La Paz Formation

The polarity sequence of the La Paz Formation measured in this study and by Thouveny and Servant (1989) depends on the location of the section. Farther from the Cordillera Real, the formation extends upward, replacing till and gravel units, and thus comprises a more complete polarity sequence than the La Paz Formation below thick glacial sequences farther north (Fig. 2.14). The upper La Paz Formation (following Dobrovolny’s [1962] nomenclature) records a bottom-up polarity sequence of R-N-R-N-R-N below the Chijini Tuff at the Viscachani section, similar to that of glacial units at the Patapatani West section. The upper two magnetozones are also represented at the Jacha Kkota section, the only other section where the La Paz Formation below the tuff was paleomagnetically sampled. Equivalent stratigraphic positions and similar relative thicknesses of magnetozones suggest likely correlation between the pre-Chijini drift sequence and the upper La Paz Formation. At the Jacha Kkota section, Thouveny and Servant (1989) show that the first ~12 m of fine-grained sediments above the Chijini Tuff are normally magnetized. These deposits and probably a small, lowermost portion of the overlying reversely magnetized fine-grained deposits appear to correlate with the post- tuff drift sequence closer to the Cordillera Real. Dobrovolny (1962) appears to have been correct in considering these parts of the Calvario Drift, because they are likely in fact time equivalents and are probably related in the same way as the Patapatani Drift and upper La Paz Formation below the Chijini Tuff are related.

The fine-grained sequence below the Chijini Tuff at the Tangani section (Thouveny and Servant, 1989; Fig. 2.14e) and the sequence below and above the tuff in the Achocalla basin (Fig. 2.14h) include nine of the ten magnetozones of the composite polarity sequence. Only the short normally magnetized interval measured up valley in granitic gravels (N4) has not yet been documented in the La Paz Formation; it is likely within the unsampled portion of the Jacha Kkota section above Thouveny and Servant’s (1989) measurements and below those reported here (Fig. 2.14h). The lower and middle La Paz Formation, which extends >500 m below sequence in Fig. 2.14, no doubt includes additional older magnetozones (cf. Thouveny and Servant’s [1989] Irpavi section).

205

Appendix F

Chuquiaguillo graben

The Minasa section contains a much thicker (~400 m), post-eruption clastic sequence than sections a short distance closer to the Andes (Patapatani West section, ~130 m) or toward the main Altiplano plain (Purapura section, ~240 m; Jacha Kota, ~125 m) (Fig. 2.14). The anomalously thick sequence at Minasa requires far more vertical accommodation space than the sequence at the other sections. The presence of the locally thick granitic gravel at the section, which record a high-energy fluvial system, suggests deposition in a confined valley. A deep, long-lived paleo-valley would likely result in erosional hiatus in the Minasa sequence, yet the Chijini Tuff and several paleosols are well preserved evidence of repeated flat-lying surfaces. Additionally, evidence of major unconformities is lacking, although the coarse and homogenous nature of the gravel sequence may hinder their recognition.

A dense network of W-E and NW-SE striking faults is present in this portion of the Río Chuquiaguillo/Orkojahuira valley (Dobrovolny, 1962; Lavenu, 1977; Lavenu et al., 2000). These high-angle faults appear to be normal, although their steep dip and poor exposure limit confident determination of the dip direction and therefore the sense of displacement (normal versus reverse). Dobrovolny (1962) proposes a graben crossing the valley at this location and possibly extending toward the northwest, but suggests deformation lasted only until deposition of the lower part of the granitic gravels (Dobrovolny, 1962, p. 72). In contrast, Lavenu et al. (2000) suggest deformation in the area occurred after 1.6 Ma based on the stratigraphic interpretations of Ballivián et al. (1978) and Thouveny and Servant (1989), which suppose a 1.6-Ma tuff they call the Sopari in the Río Chuquiaguillo valley. However, confirmation that the displaced tuff in the Minasa section is in fact the ca. 2.7-Ma Chijini Tuff (Appendix C) means that deformation, at least at the Minasa section, might have occurred over 1 Ma earlier.

Progressive, approximately syndepositional subsidence of a graben is a likely source for the necessary accommodation space and constrained alluvial system. The greater thickness of the granitic gravels (Fig. 2.14) and the overlying normal magnetozone containing interbedded gravel and till relative to other sections indicate that subsidence probably spanned these units and lasted longer and later than suggested by Dobrovolny (1962). Vertical offset between opposite slopes of the Minasa section suggest a NW- trending, unmapped, high-angle fault oblique to those mapped underlying Huari Pampa and parallel to those mapped in the Río Chuquiaguillo/Orkojahuira valley by previous authors. A decrease in the amount of displacement across the fault from ~10 m in the Chijini Tuff at the base of the section to ~0.5 m near just below the Huari Pampa surface indicate continued movement along the fault during deposition of the post-Chijini sequence. The lack of offset of the Huari Pampa surface indicates a cessation of movement prior to the creation of the modern Altiplano surface. Following the favoured chronostratigraphic interpretation presented here, graben subsidence began by ca. 2.7 Ma and continued until shortly before the end of the Olduvai subchron (ca. 1.8 Ma).

206

The minor offset (maximum 10 m) across the fault at the Minasa section indicates it had only a minor role in subsidence (~200 m or more in total). Vertical displacement was likely accommodated by cumulative movement across several faults. Roughly 2 km up the Río Chuquiaguillo/Orkojahuira valley, two identical exposures of tuff only 500 m apart are offset vertically by ~130 m, suggesting further high-angle faulting of the order of magnitude necessary to account for the thick deposits at the Minasa section.

A deepening graben in the middle part of the transect shown in Figure 2.14 could also account for the erosional unconformity in the upper part of the Patapatani West section (Fig. 2.14a). It is the only major erosional contact yet documented in the sedimentary sequences and reflects locally lowered base level within, and increased slope gradients adjacent to, the graben.

207

Appendix G

Fill-sequence aggradation and incision rates

Aggradation rates

Calculation of incision rates provides an opportunity to compare the two possible correlations of the polarity sequence to the GMPT. Aggradation rates are calculated using section thickness and ages pinned by the 2.74-Ma Chijini Tuff and by polarity reversals (Table F1). Rates are summarized in Table 2.5.

a Section abbreviations: PTW, Patapatani West; PTE, Patapatani East; TNG, Tangani; MIN, Minsasa; VIS, Viscachani (from Thouveny and Servant, 1989); PUR, Purapura; JKT, Jacha Kkota (including some data from Thouveny and Servant, 1989).

Polarity reversals below the tuff are well constrained at three sections, which thus allow confident estimation of long-term aggradation rates (~12-17 cm/ka). Two of these sections (Patapatani West and Jacha Kkota) extend above the tuff to the Altiplano surface. Post-tuff aggradation rates at these two sections are similar to pre-tuff rates (12.5 and 11.8 cm/ka, respectively) with the preferred interpretation (13.5 and 13.6 cm/ka). With the alternative interpretation, post-tuff aggradation rates are about 50 percent less (7.3 cm/ka at both sections) than the pre-tuff rates.

Incision rates

Magnetostratigraphic sections extending to the Altiplano surface (Fig. 2.14) allow quantification of average rates of incision of the sedimentary sequence. The calculated rates are based on local depth of the trunk stream below the Altiplano surface and its paleomagnetically constrained age (either late Olduvai [ca. 1.8 Ma] or late Jaramillo [ca. 1.0 Ma]). Calculated incision rates are minimum estimates as the delay between formation of the Altiplano surface and the onset of incision is unknown.

The general decrease of minimum incision rates toward the headwaters of the watershed (Table 2.5) reflects progressively later initiation of incision as the drainage expands to the northeast. Thus, the rate at the farthest south section provides the most realistic minimum limit on long-term incision. The lower incision rate limit is ~40 cm/ka based on an Olduvai age for the Altiplano surface and ~75 cm/ka based on a Jaramillo age. Incision under the former scenario is closer to millennial-scale erosion rates of 23.0 cm/ka for the Río La Paz basin (range from 10.0 to 60.0 cm/ka for individual sub-basins) determined by Zeilinger et al. (2009) using terrestrial cosmogenic nuclides (TCN). Average surface denudation rates determined using TCN is similar in the Cordillera Real (20.0-60.0 cm/ka; Safran et al., 2005), but is two orders of magnitude lower on the Altiplano surface immediately to the west (0.03-2.90 cm/ka; Hippe et al., 2012).

208

The larger minimum incision rate in the Río Chuquiaguillo valley (Minasa section; Table 2.5) relative to those in the Río Kaluyo/Choqueyapu valley may reflect drainage-specific differences in erodibility. Faulting in the area (Lavenu, 1997; Lavenu et al., 2000) associated with the Chuquiaguillo graben may predispose fills of that area to enhanced erosion. Alignment of the main branches of the Río Minasa quebrada system along fault traces (Appendix F) attest to such preferential erosion. The dominance of gravel over more resistant till in this area (Fig. 2.14) may also play a role.

209

492 790 960 Total 2500 1740 1740 2100 0 132 790

Ar age of the Supra- tuff 1740 1740 1740 1740 39 Thouveny and Time (ka) elapsed 0 0 0 Ar/ 760 360 960 360 40 Sub-tuff

Ma - Tuff 2740 2740 2740 2740 2740 2740 2740 ALTERNATIVE INTERPRETATION ALTERNATIVE Top 1000 2608 1950 1000 2740 1000 1000 Age limits Age (ka) Base 3500 3100 2740 2740 3700 2740 3100 492 790 940 960 940 Total 1700 1300 0 Thouveny and Servant, 1989). 940 132 790 940 940 940 Supra- tuff Time (ka) elapsed 0 0 0 760 360 960 360 Sub-tuff Top 1800 2608 1950 1800 2740 1800 1800

PREFERRED INTERPRETATION Tuff 2740 2740 2740 2740 2740 2740 2740 measured stratal thickness. Age limits based on 2.74 - Age limits Age (ka) field Base 3500 3100 2740 2740 3700 2740 3100 50 232 250 410 175 248 175 Total sequence aggradation and incision rates from sections at La Paz. from La sections at rates incision and aggradation sequence - 0 14 127 250 400 242 128 Supra- tuff 5 0 3 5 10 10 10 Thickness (m) Tuff 0 0 3 95 31 43 165 tuff Sub- 285 575 360 750 Depth Values used to calculate fill calculate to Values used NA NA NA mated GoogleEarth.in Thickness based on 4405 4320 4190 3970 Plateau Plateau surface jini Tuff and age of polarity reversals in the LR04 GMPT. Elevation (m) Elevation

Servant, 1989 ); PUR, Purapura; JKT, Jacha Kkota (including some data from Chi 4120 3745 3830 3220 Valley bottom a PTW PTE TNG MIN VIS PUR JKT Section abbreviations: PTW,West; Patapatani PTE, Patapatani East; TNG, Tangani; MIN, Minsasa; VIS, Viscachani (from

Section Table F1. Table

esti Elevations a

210

age range 1993, pg.

" MacFadden et al. (1983) MacFadden et al. (1993) Roperch et al. (1999) (1982) al. et Marshall Source late Calabrian late Zanclean-Piacenzian Gelasian Geologic age c Start of Jaramillo to start of Brunhes of start to Jaramillo of Start End of Cochiti to start of Mammoth Start of Matuyama to end of Olduvai of end to Matuyama of Start Age boundaries Values stated in source

b 9.2 MPT and using a different approach to estimate duration of time 18.8 16.9 90.0 16.9 Aggradation rate (cm/ka) 640 865 295 827 5000 Range

780 3319 9100 1781 Age (ka) Top ca. 3300 4184 1075 2608 14100 Base ca. 4000 MacFadden et al., 1993, Fig. 8, pg. 238 ). a astronomically tuned version of the G 50 80 120 140 4500 (m) Thickness ted aggradation rates of other late Cenozoic continental fill sequences in the Central Andes. Central the in fill sequences continental late Cenozoic other Calcula of rates aggradation ted

229 238& ). represented by stratigraphic sequence. (two magnetozones) ( " Thickness provided sourceby covers a longer stratigraphic sequence (four magnetozones) than polarity sequence used to define Rate is likely overestimated since stratigraphic thickness spans moremagnetozones than age range stated by MacFadden et ( al. Sequence Based on a previous, non -

Tarija basin, Bolivia basin, Tarija Inchasi section, Bolivia section, Inchasi Corque, Bolivia Corque, Argentina Uquia Formation, a b c

F2. Table

211

Appendix H

InSAR processing methodology

Pre-processing

Scene registration and multi-look resampling

I first optimized scenes for point targets. The use of point look-up tables ensures preservation of point targets as well as all but the smoothest (e.g. some paved roads or airport tarmac) distributed targets. I then resampled single-look complexes to pixel locations of a common master scene. Next, I resampled referenced single-look complexes to produce referenced multi-look images with roughly square ground dimension. The selected master scene for the La Paz stack (2010 July 12) provides high coherence and a short baseline with most other scenes due to its position in the middle of the stack (23rd of 44 scenes), as well as its acquisition in the driest part of the year when soil moisture and, therefore, soil dielectric constant (Wang et al., 1978) for the sensor’s frequency (5.405 GHz) are likely to be lowest and most uniform. The multi-look dimensions that I used are the highest possible resolution, providing roughly square ground-dimension pixels (2x3) and a scaled, lower resolution suitable for atmospheric modeling (8x12).

Elevation reference

I stitched and cropped 3-arc-second Shuttle RADAR Topographic Mission (SRTM3 V2; Farr, 2007) digital terrain model (DTM) tiles to fit the scene footprint. The SRTM subset for the La Paz area is free of holes and other obvious abnormalities. I projected the DTM to SAR coordinates and topographic phase calculated for the single-look and for both multi-look complexes.

Differential InSAR

Multi-look 2D interferograms

I applied differential InSAR (D-InSAR) to determine time-series phase statistics and to identify atmospheric phase contributions. For each unique acquisition pair (946 total for the 44-scene La Paz stack), I generated a two-dimensional (2D) interferogram over the entire scene area for both the intermediate (2x3) and low resolution (8x12), multi-look dimensions. Topographic phase calculated from the SRTM DTM and flat-earth phase were removed, yielding initial 2D interferograms corrected for terrain and Doppler effects. I also calculated spectral coherence for each unique image pair.

212

Baseline optimization

A coherence-maximization algorithm was applied independently to each interferogram in wrapped-phase space to identify and correct errors in each of the four components of the interferometric baseline (parallel baseline, perpendicular baseline, and the velocities of each), as well as parameters of a static atmosphere model. The algorithm improves initial estimates of orbital parameters used to generate initial 2D interferograms, but produces a network of linearly inconsistent baselines. Singular value decomposition (SVD) inversion of the network of corrected linearly inconsistent baselines and the static atmosphere parameters provide a set of linearly consistent, high-accuracy interferometric baselines. I used the improved baseline estimates to produce refined interferograms with mitigated fringe effects from inaccurate baselines and the topography-correlated phase.

Long-range atmospheric phase screen generation

A network inversion solution and smoothing function applied to the coarsest multi-look interferograms characterize large spatial-scale phase differences presumed to result from atmospheric heterogeneities. I masked out areas lacking coherence (e.g. water bodies) or of suspected ground motion (e.g. known landslides or areas of small spatial- scale phase fringes remaining after baseline optimization) to prevent their inclusion in modeling atmospheric phase. To reduce the likelihood of including unknown (a-priori to InSAR processing) ground deformation while also allowing adequate variation within the scene, I chose the scale of the smoothing function to be between the scale of deformation and half the scale of the scene (between ~3 km and 12.5 km for the La Paz stack). I removed the generated phase screens from baseline-optimized interferograms to generate atmosphere-corrected interferograms.

Three main sources influence the remaining coherent phase: high-frequency topographic variability not adequately represented in the moderate-resolution DTM, cyclic movement related to thermal expansion, and permanent ground movement. Decorrelation also remains as incoherence phase because the atmospheric screens do not contain, and thus do not remove, high-frequency noise.

Homogeneous Distributed Scatterer InSAR

I performed Homogeneous Distributed Scatterer InSAR (HDS-InSAR) only on spatial subsets of single-look images to facilitate timely processing throughout the remaining processing chain and to limit computational resources required.

HDS neighbourhood definition

Multi-looking, a form of spatial filtering within a weighted kernel, improves estimates of interferometric phase and image correlation (Parizzi and Brcic, 2011). HSD-InSAR

213

employs a continuously weighted, spatially adaptive filter to group and average contiguous pixels showing similar time-series amplitude variability. The filter employs a similarity measure derived from a non-parametric statistical test (Anderson Darling) for thick scene stacks, such as the La Paz stack, or a one-parametric statistical test (Generalized Likelihood Ratio Test assuming Rayleigh distribution) for smaller stacks (fewer than ~15 scenes; Rabus et al., 2012).

The resulting neighbourhoods display spatially homogeneous spectral behaviour over time. Due to their point nature, persistent scatterers yield single-pixel neighbourhoods. Coherent distributed scatters are characterized as spatially continuous neighbourhoods, limited in extent by the dimensions of the filtering kernel applied and represented by a point at the kernel’s centre. I applied a 13 x 21 pixel kernel to the La Paz stack, which restricts spatial averaging to a roughly square ground area measuring ~105 m on a side. The adaptive filtering improves speckle and phase, while preserving sharp boundaries and point targets.

Single-look 2D interferograms

I generated 2D interferograms for the sub-set area using refined baselines and atmospheric screen developed during full-scene multi-look processing. These interferograms included each single-look network interferogram and each multi-looked (i.e. filtered using the previously defined weighted neighbourhoods) interferogram. The later are HDS differential interferograms providing phase change for each HDS candidate.

Temporal differential coherence

For every interferometric scene pair, I compared the differential phase of each pixel to a reference phase of nearby pixels immediately outside its neighborhood candidate window (just beyond 13 x 21 pixels for the La Paz stack). I then calculated the overall differential coherence of the pixel throughout the scene stack, which provides a measure of the quality of coherence of each HDS neighbourhood by characterizing how much a pixel varies over time from adjacent pixels outside its neighbourhood.

Peak-to-peak differential coherence

Solving and correcting for residual height and, if selected, temperature dilation for each HDS leads to significant improvement of coherence prior to phase unwrapping and network inversion processing. Search functions estimate height error, linear deformation rate, and thermal dilation for each pixel of each interferogram. For each factor considered, I calculated the peak-to-peak difference, that is the difference between the highest and second highest maxima found during the search, for each pixel. I then combined coherence maps of the differential phase components into an overall differential coherence map that shows the differential phase components with the lowest

214

peak-to-peak difference for each pixel location. This map shows the quality of the residual differential phase component at each pixel location.

I considered only height error and linear deformation in the case of the La Paz stack. Temperature dilation of structures can influence phase in cases where the temperature differences between scenes are large (Monserrat et al., 2011). Given La Paz’s low latitude, temperature differs little throughout the year. Diurnal temperature variation is much greater due to the high elevation of the city, but the regular scene acquisition time (ca. 19:00 local time) minimizes day-to-day temperature effects. Therefore, I did not correct for effects of temperature dilation in this study. However, the HDS-InSAR processing chain allows correction to be done by considering an independent temperature record (Eppler and Rabus, 2011).

Phase modeling

I reapplied the aforementioned search function to atmosphere-corrected HDS interferograms, but using narrow search limits. This modeling estimates residual height corrections resulting from phase correlated with perpendicular baseline, and linear deformation rates resulting from temporally consistent phase. The model can additionally estimate and correct for thermal dilation coefficients resulting temporally cyclic phase correlated with independent temperature input data; however, I did not include this optional consideration in processing of the La Paz stack. In additional to these components, the modeled phase includes residual phase time series.

HDS selection by temporal coherence thresholding

I separated spatially coherent neighbourhoods from incoherent ones using temporal coherence of the differential phase as a measure of quality measure. I applied multiple thresholds in separate trials to the linear deformation component of the phase model; higher threshold decreases the number of HDS candidates remaining. I determined the optimal coherence level by visually comparing incremental threshold trails, and. I adopted the coherence value providing the best balance between eliminating spatially variable deformation trends among HDS and preserving high-quality-phase HDS. I extracted the phase model and network interferograms for each HDS candidate above this coherence threshold.

Phase demodulation and removal of strong motion trend

I further flattened interferometric phase through phase demodulation. I computed the approximate deformation signal from the 2D InSAR solution of a minimum network of high-quality interferograms. The remaining linear deformation rate was calculated after removal of this deformation trend from overall phase.

215

Phase demodulation has two important benefits. First, it mitigates temporal wrapping due to large temporal baselines in lengthy acquisitions stacks. Second, it improves preservation of non-linear deformation for which the one-dimensional linear deformation fit during phase modeling performs poorly.

Spatial phase unwrapping

I further flattened the demodulated phase using the phase model for residual height, linear displacement, and thermal dilation. The resulting corrections of each HDS for topographic height error relative to the reference DEM, and temperature dilation if selected, improve coherence. I then unwrapped the flattened phase with a two-step iterative variant of the minimum-cost function (Costantini, 1998) unwrapper, after which I added back the modeled phase components onto the HDS.

Re-addition of deformation trend

I added back the deformation trend removed prior to spatial phase unwrapping onto the unwrapped HDS of each network interferogram to restore the complete deformation signal.

Phase model refinement and residual height correction

Using linear least squares regression, I improved the phase model by estimating residual height corrections, and linear deformation coefficients remaining in the unwrapped data. Residual height errors determined by the refined phase model were removed from the unwrapped phase for each interferogram. I did not estimate and remove thermal dilation coefficients as I have not considered thermal effects for the La Paz stack.

Singular value decomposition solution and temporal filtering

I SVD-inverted network interferograms for each instance in the time series. I smoothed the SVD-inverted time series by temporal filtering I then calculated line-of-sight deformation and vertically projected deformation components from it. The temporal filter is a least square low-pass filter and is optimumal for irregularly sampled data, such as temporal gaps resulting from missing scenes. It produces weighted-averaging smoothing over a five-scene period spanning two scenes on either side of each time-series postion. This smoothing period is thus 120 days for RADARSAT-2 due to its 24-day return interval.

216

Output generation

To enable accurate alignment with reference terrain models, satellite imagery, and landslide observations and polygons, I converted HDS-InSAR results to geodetic coordinates. I geocoded point locations of each HDS using refined residual height errors from phase model refinement to improve the input reference DTM. I subsequently refined the geocoding table using rational functions and height corrections from the final phase model to improve geocoding in the presence of large DTM height errors. Because of the high density of data points, a linear deformation map was interpolated from HDS to provide a base image to efficiently represent long-term average line-of-sight deformation over the extensive area. Finally, I exported HDS point data and the linear deformation map to a GoogleEarth-compatible plug-in rendered at a user-defined deformation range (e.g. -1 to 1 cm/a year for slow ground motion, -25 to 25 cm/a or more for motion approaching the measurement limit of the c-band sensor).

217

Table H1. List of scenes used in InSAR processing.

Da te (yyyy-mm-dd) Scene Comment

2008-09-08 1 2008-10-02 2 2008-10-26 3 2008-11-19 Not collected 2008-12-13 Not collected 2009-01-06 Not collected 2009-01-30 Not collected 2009-02-23 Not collected 2009-03-19 4 2009-04-12 5 2009-05-06 6 2009-05-30 Not collected 2009-06-23 7 2009-07-17 8 2009-08-10 9 2009-09-03 10 2009-09-27 11 2009-10-21 12 2009-11-14 13 2009-12-08 14 2010-01-01 15 2010-01-25 16 2010-02-18 17 2010-03-14 18 2010-04-07 19 2010-05-01 20 2010-05-25 21 2010-06-18 22 2010-07-12 23 Master scene 2010-08-05 24 2010-08-29 25 2010-09-22 26 2010-10-16 27 2010-11-09 28 2010-12-03 29 2010-12-27 30 2011-01-20 31 2011-02-13 32 2011-03-09 33 2011-04-02 34 2011-04-26 35 2011-05-20 Not collected 2011-06-13 36 2011-07-07 37 2011-07-31 38 2011-08-24 39 2011-09-17 40 2011-10-11 41 2011-11-04 42 2011-11-28 43 2011-12-22 44

218

References

Acton, G.D., Okada, M., Clement, B.M., Lund, S.P. and Williams, T. 2002. Paleomagnetic overprints in ocean sediment cores and their relationship to shear deformation caused by piston coring. Journal of Geophysical Research, 107: B4 2067.

Aguilar, O.Q. 2013. Detección de cambios de uso del suelo en la zona de Callapa afectada por el megadeslizamiento (La Paz – Bolivia). In Proceedings of XIV Encuentro de Geógrafos de América Latina: Reencuentro de Saberes Territoriales Latinoamericanos, 8-11 April, Lima, Perú, 20 pp.

Ahlfeld, F. and Branisa, L. 1960. Geología de Bolivia. La Paz, Instituto Boliviano Petróleo, 245 pp.

Ahlfeld, F. 1946. Geología de Bolivia. Revista del Museo de La Plata: Sección Geología, Tomo III, p. 5-370.

Ahlfeld, F. 1954a. Reseña geológica de la Cuenca de La Paz – Parte 1. Instituto Boliviano de Ingeniería de Minas y Geología, 2(16): 11-15.

Ahlfeld, F. 1954b. Reseña geológica de la Cuenca de La Paz – Parte 2. Instituto Boliviano de Ingeniería de Minas y Geología, 2(17): 11-15.

Alexander, D. 1989. Urban landslides. Progress in Physical Geography, 13: 157-189.

Anbalagan, R. 1992. Landslide hazard evaluation and zonation mapping in mountainous terrain. Engineering Geology, 32: 269-277.

Anderson, M.B. 1992. Metropolitan areas and disaster vulnerability: a consideration for developing countries. In Environmental Management and Urban Vulnerability (Kreimer, A. and Munasinghe, M. [eds.]). World Bank discussion paper 18, World Bank, Washington, DC, p. 77-92.

219

Anzoleaga, O., Malatrait, A.M., Michalski, E.R., Bles, J.L., Otazo, N. and Walgenwitz, F. 1977. Características geomtécnicas de la cuenca de La Paz y alrededores. Informe Geológico No. 15. Plan de desarrollo Urbano para la ciudad de La Paz: Honorable Alcaldía Municipal de La Paz, La Paz, 107 pp.

Aranda, K.E. 2008. En busca del pasado perhispánico del valle de La Paz. Chachapuma – Revista de Arqueologhía Boliviana, 3: 64-65.

Arbona, J.M. and Kohl, B. 2004. City profile: La Paz-El Alto. Cities, 21: 255-265.

Baksi, A.K. and Hoffman, K.A. 2000. On the age and morphology of the Reunion event. Geophysical Research Letters, 27: 2,997-3,000.

Balco, G., Rovey, C.W. and Stone, J.O.H. 2005a. The First Glacial Maximum in North America. Science, 307: 222.

Balco G., Stone, J.O.H. and Jennings, C. 2005b. Dating Plio-Pleistocene glacial sediments using the cosmogenic-ray-produced radionuclides 10Be and 26Al. American Journal of Science, 305: 1-41.

Ballivián, O., Bles, J.L. and Servant, M. 1978. El Plio-cuaternario de la región de La Paz (Andes Orientales, Bolivia). Cahiers O.R.S.T.O.M. Série Géologie, 10: 101-113.

Barendregt, R.W., Enkin, J.E., Duk-Rodkin, A. and Baker, J. 2010. Paleomagnetic evidence for multiple late Cenozoic glaciations in the Tintina Trench, west-central Yukon, Canada. Canadian Journal of Earth Sciences, 47: 987-1002.

Barnes, J.B. and Ehlers, T.A. 2009. End member models for Andean Plateau uplift. Earth-Science Reviews, 97: 105-132.

Barragan, C., Saravia, F., Quenta, G., Terán, N., Galarza, I., Pocoaca, G. and Balderrama, J. 2006. Mapa de Inventario de Movimientos en Masa, Sub-cuenca Allpacoma. SERGEOTECHMIN. Unpublished map. 1: 75 000.

Bartoli, G., Sarnthein, M., Weinelt, M., Erlenkeuser, H., Garbe-Schönberg, D. and Lea, D.W. 2005. Final closure of Panama and the onset of northern hemisphere glaciation. Earth and Planetary Science Letters, 237: 33-44.

Bartoli, G.M., Hönisch, B. and Zeebe, R.E. 2011. Atmospheric CO2 decline during the Pliocene intensification of Northern Hemisphere glaciations. Paleoceanography, 26: PA4213.

220

Berardino, P., Fornaro, G., Lanari, R. and Sansosti, E. 2002. A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms. IEEE Transactions on Geoscience and Remote Sensing, 40: 2375-2383.

Berardino, P., Casu, F., Fornaro, G., Lanari, R., Manunta, M., Manzo, M., Pepe, A. and Sansosti, E. 2003. Small Baseline DIFSAR techniques for earth surface deformation analysis. In Proceedings ‘Fringe 2003 Workshop’, 1-5 December, Frascati, Italy. ESA Communications SP-550 (H. Lacoste, [Ed.]). Unpagninated CD ROM, 6 pp.

Bershaw, J., Grazione, C.N., Higgins, P., MacFadden, B.J., Anaya, F. and Alvarenga, H. 2010. Spatial–temporal changes in Andean plateau climate and elevation from stable isotopes of mammal teeth. Earth and Planetary Science Letters, 289: 530- 538.

Berthelet, 1986.The extraction or precious metals at the time of the Inka. In Anthropological History of Andean Polities (Murra, J., Watchel, N. and Revel, J. [Eds.]). Cambridge University Press, Cambridge, p. 69-88.

Bianchini, S., Cigna, F., Righini, G., Proietti, C. and Casagli, N. 2012. Landslide HotSpot mapping by means of persistent scatterer interferometry. Environmental Earth Sciences, 67: 1155-1172.

Bianchini, S., Cigna, F., Del Ventisette, C., Moretti, S. and Casagli, N. 2013. Monitoring landslide-induced displacements with TerraSAR-X Persistent Scatterer Interferometry (PSI): Gimigliano case study in Calabria Region (Italy). International Journal of Geosciences, 4: 1467-1482.

Bianchini, S., Ciampalini, A., Raspini, F., Bardi, F., Di Traglia, F., Moretti, S. and Casagli, N. 2015. Multi-temporal evaluation of landslide movements and impacts on buildings in San Fratello (Italy) by means of C-band and X-band PSI data. Pure and Applied Geophysics, 172: 3043-3065.

Bles, J.L. 1977. Neotectónica y sismicidad de la región de La Paz. Informe Geológico No. 7. Plan de desarrollo Urbano para la ciudad de La Paz: Honorable Alcaldía Municipal de La Paz, La Paz, 52 pp.

Bles, J.L., Alvarez, A., Anzoleaga, O., Ballivián, O., Bustillos, O., Hochstatter, H., Malatrait, A.M. and Otazo, N. 1977. Características litoestratigráficas de la cuenca de La Paz y alrededores. Informe Geológico No. 5. Plan de desarrollo Urbano para la ciudad de La Paz: Honorable Alcaldía Municipal de La Paz, La Paz, 35 pp.

221

Blow, R.A. and Hamilton, N. 1978. Effect of compaction on the acquisition of detrital remanent magnetization in fine-grained sediments. Geophysical Journal of the Royal Astronomical Society, 52: 13-23.

Bogue, S.W. and Coe, R.S. 1981. Paleomagnetic correlation of Columbia River basalt flows using secular variation. Journal of Geophysical Research, 86 (B12): 11,883-11,897.

Böhome, M., Hermanns, R.L., Oppikofer, T., Fisher, L., Bunkholt, H.S.S., Eiken, T., Pedrazzini, A., Derron, M.-H., Jaboyedoff, M., Blikra, L.H. and Nilsen, B. 2013. Analyzing complex rock slope deformation at Stampa, western Norway, by integrating geomorphology, kinematics and numerical modeling. Engineering Geology, 154: 116-130.

Boman, I. 1909. Physiography of the Central Andes: II, The Eastern Andes. American Journal of Science, 4: 373-402.

Bookhagen, B. and Strecker, M.R. 2008. Orographic barriers, high-resolution TRMM rainfall, and relief variations along the eastern Andes. Geophysical Research Letters, 35: L06403.

Booth, A.M., Dehls, J., Eiken, T., Fischer, L., Hermanns, R.L. and Oppikofer, T. 2015. Integrating diverse geologic and geodetic observations to determine failure mechanisms and deformation rates across a large bedrock landslide complex: the Osmundneset landslide, Sogn og Fjordane, Norway. Landslides, 12: 745- 756.

Bovenga, F., Nutricato, R., Refice, A. and Wasowski, J. 2006. Application of multitemporal differential interferometry to slope instability detection in urban/peri- urban areas. Engineering Geology, 88: 218-239.

Bowman, I. 1914. Results of an Expedition to the Central Andes. American Geographical Society, 46: 16-183.

Brady, N.C., and Weil, R.R. 2002. The nature and properties of soil. 32th Edition. Prentice Hall, Upper Saddle River, NJ, 960 pp.

Bräuning, A. 2009. Climate variability of the tropical Andes since the late Pleistocene. Advances in Geoscience, 22: 13–25.

222

Brewster, R. Grest, G.S., Landry, J.W. and Levine, A.J. 2005. Plug flow and the breakdown of Bagnold scaling in cohesive granular flows. Physical Review E, 72: 061301.

Brideau, M.A. and Roberts, N.J. 2014. Mass movement in bedrock. In Landslide Hazards, Risks, and Disasters (Davies, T. [Ed.]). Elsevier, Amsterdam, p. 43-90.

Butler, R.F. 1992. Paleomagnetism: magnetic domains to geologic terrains. Blackwell Scientific Publications, London, 319 pp.

Cande, S.C. and Kent, D.V. 1992. A new geomagnetic polarity time scale for the Late Cretaceous and Cenozoic. Journal of Geophsycial Research, 97(B10): 13,917- 13,951.

Cane, M.A. and Molnar, P. 2001. Closing of the Indonesian seaway as a precursor to east African aridification around 3–4 million years ago. Nature, 411: 157-162.

Castaldo, R., Tizzani, P., Lollino, P., Calò, Ardizzone, F., Lanari, R., Guzzetti, F. and Manunta, M. 2015. Landslide kinematical analysis through inverse numerical modelling and differential SAR interferometry. Pure and Applied Geophysics, 172: 3067-3080.

Catt, J.A. 1990. Field recognition, description and spatial relationships of paleosols. Quaternary International, 6: 2-95.

Channell, J. E. T., Mazaud, A., Sullivan, P., Turner, S. and Raymo, M. E. 2002. Geomagnetic excursions and paleointensities in the Matuyama Chron at Ocean Drilling Program Sites 983 and 984 (Iceland Basin). Journal of Geophsycial Research, 107(B6): 2114.

Channell, J.E.T., Labs, J. and Raymo, M.E. 2003. The Réunion Subchronozone at ODP Site 981 (Feni Drift, North Atlantic). Earth and Planetary Science Letters, 215: 1- 12.

Chen, F., Lin, H. and Hu, X. 2014. Slope superficial displacement monitoring by Small Baseline SAR interferometry using data from L-band ALOS PALSAR and X-band TerraSAR: A case study of Hong Kong, China. Remote Sensing, 6: 1564-1586.

Cigna, F., Lasaponara, R., Masini, N., Milillo, P. and Tapete, D. 2014. Persistent scatterer interferometry processing of COSMO-SkyMed StripMap HIMAGE time series to depict deformation of the historic centre of Rome, Italy. Remote Sensing, 6: 12593-12618.

223

Cigna, F., Jordan, H., Bateson, L., McCormack, H. and Roberts, C. 2015. Natural and anthropogenic geohazards in Greater London observed from geological and ERS-1/2 and ENVISAT persistent scatterers ground motion data: Results from the EC FP7-SPACE PanGeo Project. Pure and Applied Geophysics, 172: 2965- 2995.

Cione, A.L. and Tonni, E.P. 1995. Chronostratigraphy and "Land-Mammal Ages" in the Cenozoic of southern South America: principles, practices, and the "Uquian" problem. Journal of Paleontology, 69: 135-159.

Cione, A.L., Tonni, E.P., Bargo, S., Bond, M., Candela, A.M., Carlini, A.A., Deschamps, C.M., Dozo, M.T., Esteban, G., Goin, F.J., Montalvo, C.I., Nasif, N., Noriega, J.I., Ortiz Jaureguizar, E., Pascual, R., Prado, J.L., Reguero, M.A., Scillato-Yané, G.J., Soibelzon, L., Verzi, D.H., Vieytes, E.C., Vizcaíno, S.F. and Vucetich, M.A. 2007. Mamiferos continentales del Mioceno tardío a la actualidad en Argentina: cincuenta años de estudios. Asociación Paleontológica Argentina Publicación Especial 11: 257–278

Cione, A.L., Gasparini, G.M., Soibelzon, E., Soibelzon, H.L. and Tonni, E.P. 2015. The Great American Biotic Interchange: A South American perspective. SpringerBriefs in Earth System Sciences – South America and the Southern Hemisphere. Springer, 97 pp.

Clague, J.J. 1974. The St. Eugene Formation and the development of the southern Rocky Mountain Trench. Canadian Journal of Earth Sciences, 11: 916-938.

Clapperton, C.M. 1979. Glaciation in Bolivia before 3.27 Myr. Nature, 277: 375-377.

Clapperton, C.M. 1993. Late Cenozoic glacial history of the Andes, Part I: Late Miocene- Middle Pleistocene. In Quaternary Geology and Geomorphology of South America (C.M. Clapperton, Eds.). Elsevier Science, Amsterdam, p. 335-354

Colesanti, C., Ferretti, C., Prati, C. and Rocca, F. 2003. Monitoring landslide and tectonics motions with the permanent scatterers technique. Engineering Geology, 68: 3-14.

Colesanti, C. and Wasowski, J. 2006. Investigating landslides with space-borne Synthetic Aperture Radar (SAR) interferometry. Engineering Geology, 88: 173- 199.

224

Coltorti, M., Dramis, F., Gentili, B., Pambianchi, G., Crescenti, U. and Sorriso-Valvo, M. 1985. The December 1982 Ancona landslide: A case of deep-seated gravitational slope deformation evolving at unsteady rate. Zeitschrift für Geomorphologie, 29: 335–345.

Comerci, V., Vittori, E., Cipolloni, C., Di Manna, P., Guerrieri, L., Nisio, S., Succhiarelli, C., Ciuffreda, M. and Bertoletti, E. 2015. Geohazards monitoring in Roma from InSAR and in situ data: Outcomes of the PanGeo Project. Pure and Applied Geophysics, 172: 2997-3028.

Constance, P. 2005. Firm footing for steep neighborhoods. Inter-American Development Bank) (1 March 2005). (http://www.iadb.org/en/news/webstories/2005-03-01/firm- footing-for-steep-neighborhoods,5641.html).

Conway,M. 1901. The Bolivian Andes: A record of climbing and exploration in the Cordillera Real in the years 1898 and 1900. Harper & Brothers Publishers, New York., 540 pp.

Costa, J.E. and Schuster, R.L. 1988. The formation and failure of natural dams. Geological Society of America Bulletin, 100: 1054-1068.

Costantini, M. 1998. A novel phase unwrapping method based on network programming. IEEE Transactions on Geoscience and Remote Sensing, 36: 813-821.

Craig, D. 1981. Mudslide plug flow within channels. Engineering Geology, 17: 273-281.

Crespo, L.S. 1906. Monografía de la Ciudad de La Paz de Ayacucho. Vol. 2. Historia Civil, 256 pp.

Cruden, D.M. and Varnes, D.J., 1996. Landslide types and processes. In Landslides, Investigation and Mitigation (Turner, A.K. and Schuster, L.R. [Eds.]). Special Report 247, Transportation Research Board, Washington, p. 36 -75. d’Orbigny, A.D. 1842. Voyage dans l’Amérique Méridionale. Volume 3, Part 3 : Géologie.

Dai, F.C., Lee, C.F. and Ngai, Y.Y. 2002. Landslide risk assessment and management: An overview. Engineering Geology, 64: 65-87.

De Schepper, S., Head., M.J. and Groenveld, J. 2009. North Atlantic Current variability through marine isotope stage M2 (circa 3.3 Ma) during the mid-Pliocene. Paleoceanography, 24: PA4206.

225

De Schepper, S., Gibbard, P.L., Salzmann, U. and Ehlers, J. 2014. A global synthesis of the marine and terrestrial evidence for glaciations during the Pliocene Epoch. Earth-Science Reviews, 135: 83-102.

Dehls, J.F., Basilico, M. and Colesanti, C. 2002. Ground deformation monitoring in the Ranafjord area of Norway by means of the Permanent Scatterers technique. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, 24-28 June, Toronto, Canada, 1, p. 203–207.

Del Ventisette, C., Righini, G. and Casagli, N. 2014. Multitemporal landslides inventory map updating using spaceborne SAR analysis. International Journal of Applied Earth Observation and Geoinformation, 30: 238-246.

Delacourt, C., Allemand, P., Berthier, E., Raucoules, D., Casson, B., Grandjean, P., Pambrun, C. and Varel, E. 2007. Remote-sensing techniques for analysing landslide kinematics: A review. Bulletin de la Societe Geologique de France, 178: 89-100.

Dobrovolny E. 1955. Preliminary report on the geology of the upper La Paz Valley of Bolivia. U.S. Geological Survey Open-File Report 55-42, 32 pp.

Dobrovolny E. 1956. Geología del Valle Superior de La Paz, Bolivia. Bolivien Alcaldía, La Paz, 67 pp.

Dobrovolny E. 1962. Geología del valle de La Paz. Departamento Nacional de Geología del Ministerio de Minas y Petroleo, La Paz, Bolivia. Servicio Geológico de Bolivia, Boletín No. 3, 152 pp.

Dobrovolny, E. 1968. A postglacial mudflow of large volume in the La Paz Valley, Bolivia. US Geological Survey Professional Paper 600-C:130–134.

Dolan, A.M., Haywood, A.M., Hill, D.J., Dowsett, H.J., Hunter, S.J., Lunt, D.J. and Pickering, S.J. 2011. Sensitivity of Pliocene ice sheets to orbital forcing. Palaeogeography, Palaeoclimatology, Palaeoecology, 309: 98-110.

Dolan, A.M., Haywood, A.M., Hunter, S.J., Tindall, J.C., Dowsett, H.J., Hill, D.J. and Pickering, S.J. 2015. Modelling the enigmatic Late Pliocene Glacial Event — Marine Isotope Stage M2. Global and Planetary Change, 128: 47-60.

Douglas, J.A. 1914. Geological Sections through the Andes of Peru and Bolivia: Part 1. From the Coast at Arica in the North of Chile to La Paz and the Bolivian ‘Yungas’ (Journal Geological Society, 70: 1-53.

226

Dowsett, H.J. and Cronin, T.M. 1990. High eustatic sea level during the middle Pliocene: Evidence from the southeastern U.S. Atlantic Coastal Plain. Geology, 18: 435- 438.

Dowsett, H., Robinson, M., Haywood, A., Salzmann, U., Hill, D., Sohl, L., Chandler, M., Williams, M., Foley, K. and Stoll, D. 2010. The PRISM3D paleoenvironmental reconstruction. Stratigraphy, 7: 123-139.

Dwyer, G.S. and Chandler, M.A. 2009. Mid-Pliocene sea level and continental ice volume based on coupled benthic Mg/Ca palaeotemperatures and oxygen isotopes. Philosophical Transactions of the Royal Society: Mathematical, Physical and Engineering Sciences, 367: 157-168.

Ehlers, J. and Gibbard, P.L. 2007. The extent and chronology of Cenozoic global glaciation. Quaternary International, 164-165: 6-20.

Ehlers, T.A. and Poulsen, C.J. 2009. Influence of Andean uplift on climate and paleoaltimetry estimates. Earth and Planetary Science Letters, 281: 238-248.

England, P. and Molar, P. 1990. Surface uplift, uplift of rocks, and exhumation of rocks. Geology, 18: 1173-1177.

Eppler, J. and Rabus, B. 2011. Monitoring urban infrastructure with an adaptive multilooking InSAR technique. In Proceedings Fringe 2011 Workshop, 19-23 September, Frascati, Italy. ESA Communications SP-697 (L. Ouwehand, [Ed.]). Unpagninated CD ROM, 5 pp.

Evans, J.W. 1903. Expedition to Caupolican Bolivia, 1901-1902. The Geographical Journal, 22: 601-642.

Evernden, J.F., Stanislav, J.K. and Cherroni, C.M., 1977. Potassium-argon ages of some Bolivian rocks. Economic Geology, 72: 1042-1061.

Eyles, N., Day, T.E. and Gavican, A. 1987. Depositional controls on the magnetic characteristics of lodgement tills and other glacial diamict facies. Canadian Journal of Earth Sciences, 24: 2436-2458.

Eyles, C.H. and Eyles, N. 2010. Glaicial depotis – Chapter 5. In Facies Models 4 (eds James, N.P. & Dalrymple, R.W. [Eds]). Geological Association of Canada, St. John’s, p. 73-104.

227

Farr, T.G., Rosen, P.A., Caro, E., Crippen, R., Duren, R., Hensley, S., Kobrick, M., Paller, M., Rodriguez, E., Roth, L., Seal, D., Shaffer, S., Shimada, J., Umland, J., Werner, M., Oskin, M., Burbank, D. and Alsdorf, D., 2007. The Shuttle Radar Topography Mission. Reviews in Geophysics, 45, article RG2004.

Ferretti, A., Prati, C. and Rocca, F. 2001. Permanent scatterers in SAR interferometry. IEEE Transactions on Geoscience and Remote Sensing, 39: 8-20.

Ferretti, A., Fumagalli, A., Novali, F., Prati, C., Rocca, F. and Rucci, A. 2011. A new algorithm for processing interferometric data-stacks: SqueeSAR. IEEE Transactions on Geoscience & Remote Sensing, 49: 3460-3470.

Fleck, R.J., Mercer, J.H., Nairn, A.E.M. and Peterson, D.N. 1972. Chronology of late Pliocene and early Pleistocene glacial and magnetic events in southern Argentina. Earth and Planetary Science Letters, 16: 15-22.

Forbes, D. 1861. On the geology of Bolivia and southern Peru. Quarterly Journal of the Geological Society of London, 17: 7-62.

Fritz, S.C., Baker, P.A., Seltzer, G.O., Ballantyne, A., Tapia, P., Cheng, H. and Edwards, R.L. 2007. Quaternary glaciation and hydrologic variation in the South American tropics as reconstructed from the Lake Titicaca drilling project. Quaternary Research, 68: 410-420.

Froese C., Poncos V., Skirrow R., Mansour M. and Martin C.D. 2008. Characterizing complex deep seated landslide deformation using corner reflector InSAR (CR- InSAR): Little Smoky Landslide, Alberta. In Proceedings 4th Canadian Conference on Geohazards: From Causes to Management (Locat, J., Perret, D., Turmel, D., Demers, D. and Leroueil, S. [Eds.]). Presse de l’Université Laval, Quebec City, p. 287-293.

Froese, D.G., Barendregt, R.W., Enkin, R.J. and Baker, J. 2000. Paleomagnetic evidence for multiple Late Pliocene – Early Pleistocene glaciations in the Klondike area, Yukon Territory. Canadian Journal of Earth Sciences, 37: 863– 877.

Fruneau. B., Achache, J. and Delacourt, C. 1996. Observation and modelling of the Saint-Étienne-de-Tinée landslide using SAR interferometry. Tectonophysics, 265: 181-190.

Fu, W., Guo, H., Tian, Q. and Guo, X. 2010. Landslide monitoring by corner reflectors differential interferometry SAR. International Journal of Remote Sensing, 31: 6387-6400.

228

Galarza, I., Hermanns, R., Teran., N., Quenta, G., Valencia, J. and Cerritos, O. 2005. Mapa de Susceptibilidad Relativa a la Ocurrencia de Deslizamientos, Llojeta, La Paz. SERGEOTECHMIN. Map. 1: 5 000.

Gao, C., McAndrews, J.H., Wang, X., Menzies, J., Turton, C.L., Wood, B.D., Pei, J. and Kodors, C. 2012. Glaciation of North America in the James Bay Lowland, Canada, 3.5 Ma. Geology, 40: 975-978.

Garcia-Urquia, E. and Axelsson, K. 2015. Rainfall thresholds for the occurrence of urban landslides in Tegucigalpa, Honduras: An application of the critical rainfall intensity. Geografiska Annaler Series A, Physical Geography, 97: 61-83.

Garreaud, R., Vuille, M. and Clement, A. 2003. The climate of the Altiplano: Observed current conditions and mechanisms of past changes. Palaeogeography, Palaeoclimatology, Palaeoecology, 194: 5-22.

Garzione, C.N., Molnar, P., Libarkin, J.C. and MacFadden, B.J. 2006. Rapid late Miocene rise of the Bolivian Altiplano: Evidence for removal of mantle lithosphere. Earth and Planetary Science Letters, 241: 543-556.

Geirsdóttir, Á. 2011. Pliocene and Pleistocene glaciations of Iceland: A brief overview of the glacial history. In Quaternary Glaciations—Extent and Chronology: A Closer Look (Ehlers, J., Gibbard, P.L. and Hughes, P. [Eds.]). Developments in Quaternary Science, 15. Elsevier, Amsterdam, p. 199-210.

Ghosh, P., Garzione, C.N. and Eiler, J.M. 2006. Rapid uplift ofthe Altiplano revealed through 13C-18O bonds in paleosol carbonates. Science, 311: 511-515.

Glastonbury, J. and Fell, R. 2008. Geotechnical characteristics of large slow, very slow, and extremely slow landslides. Canadian Geotechnical Journal, 45: 984-1005.

Gobierno Autónomo Municipal de La Paz. 2014a. Maquinaria pesada de la Alcaldía evita el desborde del río Pasajahuira. 17 March 2014. (http://www.lapaz.bo/lapaz/index.php?option=com_content&view=article&id=160 28&catid=306&Itemid=861)

Gobierno Autónomo Municipal de La Paz. 2014b. Un deslizamiento de tierra afecta a 10 viviendas en Periférica. 28 March 2014. (http://www.lapaz.bo/index.php?option=com_content&view=article&id=16137:un- deslizamiento-de-tierra-afecta-a-10-viviendas-en- periferica&catid=365&Itemid=538).

229

Gradstein, F.M., Ogg, J.G., Schmitz, M. and Ogg, G. 2012. The geologic time scale 2012. Elsevier, 1176 pp.

Granar, L. 1958. Magnetic measurements on Swedish varved sediments. Arkiv För Geofysik, 3: 1-40.

Gregory, 1913. The La Paz (Bolivia) Gorge. American Journal of Science, 36: 141-150.

Gregory-Wodzicki, K.M. 2000. Uplift history of the Central and Northern Andes: A review. Geological Society of America Bulletin, 112: 1091-1105.

Greif, V. and Vlcko, J. 2012. Monitoring of post-failure landslide deformation by the PS- InSAR technique at Lubietova in Central Slovakia. Environmental Earth Sciences, 66: 1585-1595.

Griffing, C.Y., Barendregt, R.W., Clague, J.J., Roberts, N.J., Corbella, H., Ercolano, B. and Rabassa, J. 2012. Magnetostratigraphy of continental glacial deposits in southernmost Patagonia. American Geophysical Union, Fall Meeting, 3-7 December, San Francisco. Abstract GP13B-1120.

Guzmán, M.A. 2007a. Deslizamiento rotacional de San Isidro en la ciudad de La Paz. In Movimientos en Masa en la Región Andina: Una Guía para la Evaluación de Amenazas [Proyecto Multinacional Andino: Geociencias para las Comunidades Andinas]. Servicio Nacional de Geología y Minería, Publicación Geológica Multinacional, 4, p. 188-194.

Guzmán, M.A. 2007b. Deslizamiento rotacional de Cuarto Centenario. In Movimientos en Masa en la Región Andina: Una Guía para la Evaluación de Amenazas [Proyecto Multinacional Andino: Geociencias para las Comunidades Andinas]. Servicio Nacional de Geología y Minería, Publicación Geológica Multinacional, 4, p. 198-204.

Guzzetti, F., Carrara, A., Cardinali, M. and Reichenbach, P. 1999. Landslide hazard evaluation: A review of current techniques and their application in a multi-scale study, central Italy. Geomorphology, 31: 181-216.

Hanssen, R.F. 2001. Radar Interferometry: Data Interpretation and Error Analysis. Kluwer Academic Publishers, Dordrecht, Netherlands, 308 pp.

Hardy, S. 2009. Ruptura del abastecimiento de agua potable: Sistema Hampaturi- Pampahasi, La Paz, enero-febrero de 2008. Bulletin de l’Institut Français d’Études Andines, 38: 545-560.

230

Harland, W. B., Cox, A. V., Llewellyn, P. G., Picton, C. A. G., Smith, A. G. and Walters, R. W. 1982. A geologic time scale 1982. Cambridge Earth Science Series. Cambridge University Press, Cambridge, 131 pp.

Haywood, A.M. and Valdes, P.J. 2004. Modelling Pliocene warmth: Contribution of atmosphere, oceans and cryosphere. Earth and Planetary Science Letters, 218: 363-377.

Haywood, A.M., Valdes, P.J., Hill, D.J. and Williams, M. 2007. The mid-Pliocene warm period: A test-bed for integrating data and models. In Deep-Time Perspectives on Climate Change: Marrying the Signal from Computer Models and Biological Proxies (Williams, M., Haywood, A.M., Gregory, F.J. and Schmidt, D.N. [Eds.]). The Micropalaeontological Society Geological Society, London, p. 443-457.

Haywood, A.M., Ridgwell, A., Lunt, D.J., Hill, D.J., Pound, M.J., Dowsett, H.J., Dolan, A.M., Francis, J.E. and Williams, M. 2011. Are there pre-Quaternary geological analogues for a future greenhouse warming? Philosophical Transactions of the Royal Society: Mathematical, Physical and Engineering Sciences, 369: 933-956.

Haywood, A.M., Hill. D.J., Otto-Bliesner, B.L., Bragg, F., Chan, W.-L., Chandler, M.A., Contoux, C., Dowsett, H.J., Jost, A., Kamae, Y., Lohmann, G., Lunt, D.J., Abe- Ouchi, A., Pickering, S.J., Ramstein, G., Rosenbloom, N.A., Salzmann, U., Sohl, L., Stepanek, C., Ueda, H., Yan, Q. and Zhang, Z. 2013. Large-scale features of Pliocene climate: results from the Pliocene Model Intercomparison Project. Climate of the Past, 9: 191-209.

Heim, A. 1951. On the glaciation of South America as related to tectonics: observations 1939-1947. Eclogae Geologicae Helvetiae, 44: 171-182.

Helmens, K.F., Barendregt, R.W., Enkin, R.J., Baker, J. and Andriessen, P.A.M. 1997. Magnetic polarity and fission-track Chronology of late Pliocene–Pleistocene paleoclimatic proxy record in the tropical Andes. Quaternary Research, 48: 15- 28.

Henderson, I., Lauknes, T., Osmundsen, P., Dehls, J., Larsen, Y. and Redfield, T. 2011. A structural, geomorphological and InSAR study of an active rock slope failure development. Geological Society of London Special Publications, 351: 185-199.

Hermanns R.L., Dehls J.F., Guzmán M.-A., Roberts N.J., Clague J.J., Cazas Saavedra A. and Quenta G. 2012. Relation of recent megalandslides to prehistoric events in the city of La Paz, Bolivia. In Proceedings of 2nd North American Symposium on Landslides, 3–6 June 2012, v. 1. Banff, Canada, p. 341–347.

231

Hidy, A.J., Gosse, J.C., Froese, D.G., Bond, J.D. and Rood, D.H. 2013. A latest Pliocene age for the earliest and most extensive Cordilleran Ice Sheet in northwestern Canada. Quaternary Science Reviews, 61: 77-84.

Hilley, G.E., Bürgmann, R., Ferretti, A., Novali, F. and Rocca, F. 2004. Dynamics of slow-moving landslides from permanent scatterer analysis. Science, 304: 1952- 1955.

Ho, K.K.S. and Lau, J.W.C. 2010. Learning from slope failures to enhance landslide risk management. Quarterly Journal of Engineering Geology and Hydrogeology, 43, 33–68.

Hoffstetter, R. 1986. High Andean mammalian faunas during the Plio-Pleistocene. In High Altitude Tropical Biogeography (Vuilleumier, F. and Monasterio, M. [Eds.]). Oxford University Press, New York, p. 218-245.

Hoffstetter, R., Martinez, C., Mufioz-Reyes, J. and Tomasi, P. 1971. Le gisement d'Ayo Ayo (Bolivie), une succession stratigraphique Pliocène-Pléistocène datée par des mammifères.

Hooper, A. 2008. A multi-temporal InSAR method incorporating both persistent scatterer and small baseline approaches. Geophysical Research Letters, 35: L16302.

Hungr, O., Leroueil, S. and Picarelli, L. 2014. The Varnes classification of landslide types, an update. Landslides, 11: 167-194

Instituto Nacional de Estadística. 2012. Bolivia: Características de Población y Vivienda (Censo Nacional de Población y Vivienda 2012). Estado Plurinacional de Bolivia, 31 pp.

Irving, E. and Major, A. 1964. Post-depositional remanent magnetization in synthetic sediment. Sedimentology, 3: 135-143.

Johnson, N.M., Opdyke, N.D. and Lindsay, E.H. 1975. Magnetic polarity stratigraphy of Pliocene-Pleistocene terrestrial deposits and vertebrate faunas, San Pedro Valley, Arizona. Geological Society of America Bulletin, 86: 5-12.

Kampes, B.M. 2006. Radar Interferometry: Persistent Scatterer Technique. Springer, Dordrecht, Netherlands, 211 pp.

Kaser, G. and Osmaston, H. 2002. Tropical Glaciers. International Hydrology Services, UNESCO. Cambridge University Press, Cambridge, 207pp.

232

Kikuchi, M. and Kanamori, H. 1994. The mechanism of the deep Bolivia earthquake of June 9, 1994. Geophysical Research Letters, 21: 2341-2344.

Kirschvink, J.L. 1980. The least-squares line and plane and the analysis of paleomagnetic data. Geophysical Journal of the Royal Astronomical Society, 62: 699–718.

Klein, H. 1992. Bolivia: The Evolution of a Multi-ethnic Society. Oxford University Press, Oxford, 368 pp.

Kleiven, H.F., Jansen, E., Fronval, T. and Smith, T.M. 2002. Intensification of Northern Hemisphere glaciations in the circum Atlantic region (3.5-2.4 Ma) – ice-rafted detritus evidence. Palaeogeography, Palaeoclimatology, Palaeoecology, 184: 213-223.

Klimeš, J. and Rios Escobar, V. 2010. A landslide susceptibility assessment in urban areas based on existing data - an example from the Iguaná Valley, Medellín City, Colombia. Natural Hazards and Earth System Sciences, 10: 2067-2079.

Kull, C., Imhof, S., Grosjean, M., Zech, R. and Veit, H. 2008. Late Pleistocene glaciation in the Central Andes: Temperature versus humidity control – A case study from the eastern Bolivian Andes (17°S) and regional synthesis. Global and Planetary Change, 60: 148–164

La Razón. 2010. Cerro Retamani se viene abajo y sepulta viviendas. 4 February 2010. La Paz, Bolivia.

La Razón. 2012. 45 personas se quedan sin techo por deslizamiento. 3 February 2012. La Paz, Bolivia. (http://www.la-razon.com/index.php?_url=/ciudades/personas- quedan-techo-deslizamiento-Cotahuma_0_1553244703.html).

La Razón. 2013. Lluvias causan derrumbes en 3 barrios paceños. 19 February 2013. La Paz, Bolivia. (http://www.la-razon.com/ciudades/Lluvias-causan-derrumbes- barrios-pacenos_0_1782421753.html).

La Razón. 2014a. Mazamorra daña puente El Rosario en Bajo Llojeta. 18 March 2014. La Paz, Bolivia. (http://www.la-razon.com/ciudades/Mazamorra-dana-puente- Rosario-Llojeta_0_2017598242.html).

La Razón. 2014b. Evacuan a familias en área de derrumbe en La Paz. 30 March 2014. La Paz, Bolivia. (http://www.la-razon.com/ciudades/Emergencia-evacuan- familias-area-derrumbe-La_Paz_0_2024197709.html).

233

Lagabrielle, Y., Scalabrino, B., Suárez, M. and Ritz, J.-F. 2010. Mio-Pliocene glaciations of Central Patagonia, New evidence and tectonic implications. Andean Geology, 37: 276-299.

Laj, C. and Channell, J.E.T. 2015. Geomagnetic excursions. In Treatise on Geophysics, Volume 5: Geomagnetism, second Edition. Elsevier, p. 343-383

Lanari, R., Mora, O., Manunta, M. Mallorquí, J., Berardino, P. and Sansosti, E. 2004. A small-baseline approach for investigating deformations on full-resolution differential SAR Interferograms. IEEE Transactions on Geoscience and Remote Sensing, 42: 1377-1386.

Lauknes, T.R., Piyush Shanker, A., Dehls, J.F., Zebker, H.A., Henderson, I.H.C. and Larsen, Y. 2010. Detailed rockslide mapping in northern Norway with small baseline and persistent scatterer interferometric SAR time series methods. Remote Sensing of Environment, 114: 2097–2109.

Lavenu, A. 1977. Neotectonica de los sedimentos Plio-Cuaternarios de la cuenca de La Paz (Bolivia). Revista del Geociencia, Universidad Mayor de San Andrés. 1(1): 31-56.

Lavenu, A. 1978. Néotectonique des sediments plio-quaternaires de l’Altiplano bolivien (region de La Paz, Ayo Ayo, Umala). Cahiers ORSTOM Série Géologie, 10: 115- 126.

Lavenu, A., Bonhomme, M.G., Vatin-Perignon, N. and de Pachtere, P. 1989. Neogene magmatism in the Bolivian Andes between 16°S and 18°S: Stratigraphy and K/Ar geochronology. Journal of South American Earth Sciences, 2: 35-47.

Lavenu, A., Thiele, R., Machette, M.N., Dart, R.L., Bradley, L.-A. and Haller, K.M. 2000. Maps and database of Quaternary faults in Bolivia and Chile. U.S. Geologic Survey Open-File Report 00-283, 46 pp.

Lawrence, K.T., Liu, Z. and Herbert, T.D. 2006. Evolution of the eastern tropical Pacific through Plio-Pleistocene glaciation. Science, 312: 79-83.

Lawver, L.A. and Gahagan, L.M. 2003. Evolution of Cenozoic seaways in the circum- Antarctic region. Palaeogeography, Palaeoclimatology, Palaeoecology, 198: 11- 37.

Lembcke, E., Conway, M., Bailey, W.H., Myring, H., Minchin, J.B., Markham, C. and Holdich, T.H., 1910. The land of the Incas: Discussion.

234

Lémuz, C. and Aranda, K.E. 2008. Mapa de Áreas Arqueológicas Potenciales del Valle de La Paz. Alcalde Municipal de la Ciudad de La Paz, La Paz, Bolivia, 130 pp.

Leonard, O.L. 1948. La Paz, Bolivia: Its population and growth. American Sociological Review, 13: 448-454.

Lisiecki , L.E. and Raymo, M.E. 2005. A Pliocene-Pleistocene stack of 57 globally distributed benthic δ18O records. Paleoceanography, 20: PA1003.

Livermore, R., Nankivell, A., Eagles, G. and Morris, P. 2005. Paleogene opening of Drake Passage. Earth and Planetary Science Letters, 236: 459-470.

Loza, J.E. 1949. Historía de La Paz. Empresa Editora Universo, La Paz, Bolivia, 354 pp.

Lu, P., Catani, F., Tofani, V. and Casagli, N. 2014. Quantitative hazard and risk assessment for slow-moving landslides from Persistent Scatterer Interferometry. Landslides, 11: 685-696.

Lunt, D.J., Haywood, A.M., Schmidt, G.A., Salzmann, U., Valdes, P.J. and Dowsett, H.J. 2010. Earth system sensitivity inferred from Pliocene modelling and data. Nature Geoscience, 3: 60-64.

Lunt, D.J., Valdes, P.J., Haywood, A. and Rutt, I.C. 2008. Closure of the Panama Seaway during the Pliocene - implications for climate and Northern Hemisphere glaciation. Climate Dynamics, 30: 1-18.

MacFadden, B.J. 2005. Extinct mammalian biodiversity of the ancient New World tropics. Trends in Ecology and Evolution, 21: 157-165.

MacFadden, B.J., Siles, O., Zeitler, P., Johnson, N.M. and Campbell, K.E. 1983. Magnetic polarity stratigraphy of the middle Pleistocene (Ensenadan) Tarija formation of Southern Bolivia. Quaternary Research, 19: 172-187.

MacFadden, B.J., Anaya, F., Perez, H., Naeser, C.W., Zeitler, P.K. and Campbell, K.E. 1990. Late Cenozoic paleomagnetism and chronology of Andean basins of Bolivia - evidence for possible oroclinal bending. Journal of Geology, 98: 541- 555.

MacFadden, B.J., Anaya, F. and Argollo, J. 1993. Magnetic polarity stratigraphy of Inchasi - a Pliocene mammal-bearing locality from the Bolivian Andes deposited just before the Great American Interchange. Earth and Planetary Science Letters, 114: 229-241.

235

MacFadden, B.J., Wang, Y., Cerling, T.E. and Anaya, F. 1994. South American fossil mammals and carbon isotopes - a 25 million-year sequence from the Bolivian Andes. Palaeogeography, Palaeoclimatology, Palaeoecology, 107: 257-268.

Mahaney, W.C., Barendregt, R.W., Hamilton, T.S., Hancock, R.G.V., Tessler, D. and Costa, J.M. 2013. Stratigraphy of the Gorges moraine system, Mount Kenya: palaeosol and palaeoclimate record. Journal of the Geological Society, London, 170: 497-511.

Malatrait, A.M. 1977. Características geomorfológicas de la cuenca de La Paz y alrededores. Informe Geológico No. 6. Plan de desarrollo Urbano para la ciudad de La Paz: Honorable Alcaldía Municipal de La Paz, La Paz, 73 pp.

Mankinen, E.A. and Dalrymple, G.B. 1979. Revised geomagnetic polarity timescale for the interval 0-5 m.y. B.P. Journal of Geophysical Research, 84(B2): 615-626.

Mansour, F.M., Morgenstern, N.R. and Martin, C.D. 2010. Expected damage from displacement of slow-moving slides. Landslides, 8: 117-131.

Mantovani, F., Soeters, R. and Van Westen, C.J. 1996. Remote sensing techniques for landslide studies and hazard zonation in Europe. Geomorphology, 15: 213-225.

Markgraf, V., Baumgartner, T.R., Bradbury, J.P., Diaz, H.F., Dunbar, R.B., Luckman, B.H., Seltzer, G.O., Swetman, T.W. and Villalba, R. 2000. Paleoclimate reconstruction along the Pole-Equator-Pole transect of the Americas (PEP-1). Quaternary Science Reviews 19, 125–140.

Markham, C.R. 1874. Disappearance of land in Bolivia. Ocean Highways: The Geographical Review, 1(9): 385-386.

Marshall L.G., Butler, R.F., Drake, R.E., Curtis, G.H. and Tedford, R.H. 1979. Calibration of the Great American Interchange. Science, 204: 272-279.

Marshall, L.G., Butler, R.F., Drake, R.E. and Curtis, G.H., 1982. Geochronology of Type Uquian (Late Cenozoic) land mammal age, Argentina. Science 216, 986–989.

Marshall, L.G., Berta, L.G., Hoffstetter, R., Pascual, R., Reig, O.A., Bombin, M. and Mones, A. 1984. Mammals and stratigraphy: Geochronology of the continental mammal-bearing Quaternary of South America. Palaeovertebrata Mémoire Extraordinaire, 76 pp.

236

Marshall, L.G. and Sempere, T. 1991. The Eocene to Pleistocene vertebrates of Bolivia and their stratigraphic context – a review. In Suarez-Soruco, R. (Ed.). Fosiles y Facies de Bolivia. Vol. 1 – Vertebrados. Revista Técnica de YPFB, Santa Cruz, Bolivia, 12 (3-4), p. 631-652.

Marshall, L.G., Swisher III, C.C., Lavenu, A., Hoffstetter, R. and Curtis, G.H. 1992. Geochronology of the mamma-bearing late Cenozoic on the northern Altiplano, Bolivia. Journal of South American Earth Sciences, 5: 1-19.

Massonnet, D. and Feigl, K. 1998. Radar interferometry and its applications to changes in the Earth’s surface. Reviews in Geophysics, 36: 441-500.

Mather, A.E., Stokes, M. and Griffiths, J.S. 2002. Quaternary landscape evolution: a framework for understanding contemporary erosion, southeast Spain. Land Degradation & Development, 13: 89-109.

Mazzanti, P., Rocca, A., Bozzano, F., Cossu, R. and Floris, M. 2011. Landslide forecasting analysis by displacement time series derived from satellite InSAR data: preliminary results. In Fringe 2011 Workshop, 19-23 September, Frascati, Italy. ESA Communications SP-697 (L. Ouwehand, [Ed.]). Unpagninated CD ROM, 8 pp.

McFadden, P.L. and McElhinny, M.W. 1988. The combined analysis of remagnetization circles and direct observations in palaeomagnetism. Earth and Planetary Science Letters, 87: 161-172.

McFarren, W. 1992. The politics of Bolivia’s economic crisis: Survival strategies of displaced tin-mining households. In Unequal Burden: Economic Crises, Persistent Poverty, and Women’s Work (Beneria, L. and Feldman, S. [Eds.]). Westview Press, Boulder, CO, p. 142–158.

McKay, R., Naish, T., Carter, L., Riesselman, C., Dunbar, R., Sjunneskog, C., Winter, D., Sangiorgi, F., Warren, C., Pagani, M., Schouten, S., Willmott, V., Levy, R., DeConto, R. and Powell, R.D. 2012. Antarctic and Southern Ocean influences on Late Pliocene global cooling. Proceedings of the National Academy of Sciences, 109: 6423–6428).

Mercer, J.H. 1976. Glacial history of southernmost South America. Quaternary Research, 6: 125-166.

Mercer, J.H. 1983. Cenozoic glaciation in the southern hemisphere. Annual Review of Earth and Planetary Sciences, 11: 99-132.

237

Mesa Gisbert, C.D., Mesa, J.D. and Gisbert, T. 2012. Historia de Bolivia. Eighth edition. Editorial Gisbert, La Paz, Bolivia, 792 pp.

Metternichtet, G., Hurni, L. and Gogu, R. 2005. Remote sensing of landslides: an analysis of the potential contribution to geo-spatial systems for hazard assessment in mountainous environments. Remote Sensing of Environment, 98: 284-303.

Milillo, P., Fielding, E.J., Schulz, W.H., Delbridge, B. and Bürgmann, R. 2014. COSMO- SkyMed spotlight interferometry over rural areas: The Slumgullion Landslide in Colorado, USA. Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 7: 2919-2926.

Mobarec, R., Quenta, G., Galarza, I., Caballero, T., Valencia, J., Muñoz-Carmona, F., Cerritos, O. and Hermanns, R. 2008. El papel del conocimiento geocientífico y social en la reducción de riesgo en el Valle de Llojeta, La Paz, Bolivia. In Experiencias Andinas en Mitigación de Riesgos Geológicos [Proyecto Multinacional Andino: Geociencias para las Comunidades Andinas]. Servicio Nacional de Geología y Minería, Publicación Geológica Multinacional, 6, p. 29– 33.

Mones, A. and Mehl, J. 1990. La presencia de macroeuphractus aff. Moreni (Lydekker, 1895), en la Formación La Paz (Plioceno), Bolivia. Comunicaciones Paleontologicas del Museo de Historia Natural de Montevideo, 22: 18-31.

Monserrat, O., Crosetto, M., Cuevas, M. and Crippa, B. 2011. The thermal expansion component of persistent scatterer interferometry observations. IEEE Geoscience and Remote Sensing Letters, 8: 864-868.

Moore, R., Carey, J.M. and McInnes, R.G. 2010. Landslide behaviour and climate change: predictable consequences for the Ventnor Undercliff, Isle of Wight. Quarterly Journal of Engineering Geology and Hydrogeology, 43: 447–460.

Morelli, M., Piana, F., Mallen, L., Nicolò, G. and Fioraso, G. 2011. Iso-kinematic maps from statistical analysis of PS-InSAR data of Piemonte, NW Italy: Comparison with geological kinematic trends. Remote Sensing of Environment, 115: 1188- 1201.

Morum, S.W.F 1936. The treatment of mud-runs in Bolivia. JOurnal of the Institue of Civil Engineers, 1: 426-457.

238

Nathan, F. 2008. Risk perception, risk management and vulnerability to landslides in the hill slopes in the city of La Paz, Bolivia: A preliminary statement. Disasters, 3: 337-357.

Necsoiu, M., McGinnis, R.N. and Hooper, D.M. 2014. New insights on the Salmon Falls Creek Canyon landslide complex based on geomorphological analysis and multitemporal satellite InSAR techniques. Landslides, 11: 1141-1153.

Newell, N.D. 1949. Geology of the Lake Titicaca region, Peru and Bolivia. Geological Society of America Memoir 36, 124 pp.

Nilsen, T.H. and Turner, B.L. 1975. Influence of Rainfall and Ancient Landslide Deposits on Recent Landslides (1950-71) in Urban Areas of Contra Costa County, California. U.S. Geological Survey Bulletin 1388, 21 pp.

Nilsen, T.H., Taylor, F.A. and Dean, R.M. 1976. Natural conditions that control landsliding in the San Francisco Bay region: An analysis based on data from the 1968-69 and 1972-73 rainy seasons. U.S. Geological Survey Bulletin 1424, 35 pp.

Notti, D., Davalillo, J.C., Herrera, G. and Mora, O. 2010. Assessment of the performance of X-band satellite radar data for landslide mapping and monitoring: Upper Tena Valley case study. Natural Hazards and Earth System Sciences, 10: 1865-1875.

Novellino, A., Di Martire, D., Ramondini , M. and Calcaterra, D. 2015. Landslide- inventory maps updated by means of persistent scatterer interferometry — The Portella di Mare case study, Sicily, Italy. In Proceedings of Ninth International Symposium on Field Measurements in Geomechanics. 9-11 September, Sydney, Australia, p. 757-768.

Noverraz, F., Bonnard, C.A. and Giraud, A. 1991. Environmental impact of a large landslide near Lausanne, Switzerland. In Proceedings of the Internataionl Confernece on Slope Stability Engineering – Developments and Applications (Chandler, R.J. [ed.]), Shanklin, Isle of Wight. Thomas Telford, Vol. 1, p. 101- 106.

Nusser, Ch. 1887. Die Mazamorra in Bolivien. Globus, 52: 62-63.

Ogg, J.G. 2012. Geomagnetic polarity time scale – Chapter 5. In The Geologic Time Scale 2012 (Gradstein, F., Ogg., J.G., Schmitz, M. ang Ogg, G. Eds.), Elsevier, p. 85-113.

239

O'Hare, G. and Rivas, S. 2005. The landslide hazard and human vulnerability in La Paz City, Bolivia. The Geographical Journal, 171: 239-258.

Oliverira, S.C., Zêzere, J.L., Catalão, J. and Nico. G. 2015. The contribution of PSInSAR interferometry to landslide hazard in weak rock-dominated areas. Landslides, 12: 703-719.

Opdyke, N.O. and Channell, J.E.T. 1996. Magnetic stratigraphy. International Geophysics Series vol. 64. Academic Press, London, 346 pp.

Osmundsen, P., Henderson, I., Lauknes, T., Larsen, Y., Redfield, T. and Dehls, J. 2009. Active normal fault control on landscape and rock-slope failure in northern Norway. Geology, 37: 135-138.

Parizzi, A. and Brcic, R. 2011. Adaptive InSAR stack multilooking exploiting amplitude statistics: A comparison between different techniques and practical results. IEEE Geoscience and Remote Sensing Letters, 8: 441-445.

Passchier, S. 2011. Linkages between East Antarctic Ice Sheet extent and Southern Ocean temperatures based on a Pliocene high-resolution record of ice-rafted debris off Prydz Bay, East Antarctica. Paleoceanography, 26: PA4204.

Petley, D. 2012. Remote sensing techniques and landslides. In Landslides: Types, Mechanisms and Modeling (Clague, J.J. and Stead, D. [Eds.]). Cambridge University Press, Cambridge, UK, p. 159-171.

Petley, D.N., Bulmer, M.H. and Murphy, W. 2002. Patterns of movement in rotational and translational landslides. Geology, 30: 719-722.

Peyret, M., Djamour, Y., Rizza, M., Ritz, J. F., Hurtrez, J. E., Goudarzi, M. A., Nankali, H., Chery, J., Le Dortz, K. and Uri, F. 2008. Monitoring of the large slow Kahrod landslide in Alborz mountain range (Iran) by GPS and SAR interferometry. Engineering Geology, 100: 131-141.

Pfuhl, H. A. and McCave, I. N. 2005. Evidence for late Oligocene establishment of the Antarctic Circumpolar Current. Earth and Planetary Science Letters, 235: 715- 728.

Piacentini, D., Devoto, S., Mantovani, M., Pasuto, A., Prampolini, M. and Soldati, M. 2015. Landslide susceptibility modeling assisted by Persistent Scatterers Interferometry (PSI): An example from the northwestern coast of Malta. Natural Hazards, 78: 681-697.

240

Prada, R. 2000. Cartografía del poder colonial: Conexiones entre haciendas y ciudades, recorridos entre Ayllus y Markas. Revista Ciencia y Cultura, 7: 173-194.

PREDCAN. 2009. Atlas de las Dinámicas del Territorio Andino: Población y bienes expuestos a amenazas naturales. Comunidad Andina, Lima, 185 pp.

Preuss, J., Buss, K. and Lopez, L. 2005. The Kelso, Washington, Aldercrest-Banyon landslide. In Landslide Hazards and Planning (Schwab, J.C., Gori, P.L. and Jeer, S. [eds.]). American Planning Association, Planning Advisory Service Report 533/534, p. 113-124.

Prevosti F.J., Gasparini G.M. and Bond M. 2006. On the systematic position of a specimen previously assigned to Carnivora from the Pliocene of Argentina and its implications for the Great American Biotic Interchange. Neues Jahrbuch für Geologie und Paläontologie, 242: 133–144

Pritchard, M.E., Jay, J.A., Aron, F., Henderson, S.T. and Lara, L.E. 2013. Subsidence at southern Andes volcanoes induced by the 2010 Maule, Chile earthquake. Nature Geoscience, 6: 632-636.

Quenta, G. and Calle, A. 2005. Diagnostico de Estabilidad de la Zona Valle de las Flores. Proyecto Multinacional Andino: Geociencias para las Comunidades Andinas Report. Servicio Nacional de Geología y Tecnico de Minas, La Paz, 12 pp.

Quenta, G., Galarza, I. and Teran, N. 2006. Informe Final Estudio Caso Llojeta- Allpacoma. Proyecto Multinacional Andino: Geociencias para las Comunidades Andinas, Informe Interno, 78 pp.

Quenta, G. Galarza, I., Teran, N., Hermanns, R.L., Cazaz, A. and García, H. 2007. Deslizamiento traslacional y represamiento en el valle de Allpacoma, ciudad de La Paz, Bolivia. In Movimientos en Masa en la Región Andina: Una Guía para la Evaluación de Amenazas. [Proyecto Multinacional Andino: Geociencias para las Comunidades Andinas]. Servicio Nacional de Geología y Minería, Publicación Multinacional, 4, p. 230-234.

Quenta, G., Hermanns, R., Galarza, I., Teran, N., Cazas, A., Caballero, T. and García, H. 2008. Deslizamiento y represamiento en el valle de Allpacoma. In Experiencias Andinas en Mitigación de Riesgos Geológicos [Proyecto Multinacional Andino: Geociencias para las Comunidades Andinas]. Servicio Nacional de Geología y Minería, Publicación Geológica Multinacional, 6, p. 29– 33.

241

Rabus, B., Eppler, J., Sharma J. and Busler, J. 2012. Tunnel monitoring with an advanced InSAR technique. In Proceedings of RADAR Sensor Technology XVI, 23–25 April, Baltimore, Maryland. SPIE 8361, p. 1F-1 – 1F-10.

Ramírez, V., Minaya, E., Valencia, J., Teran, N., Gonzalez, M., Hermanns, R.L., Clague, J.J. and Cerritos, O. 2009. Sistema de fallas El Alto (BO-06). In Atlas de Deformaciones Cuaternarias de los Andes [Proyecto Multinacional Andino: Geociencias para las Comunidades Andinas]. Servicio Nacional de Geología y Minería, Publicación Geológica Multinacional, 7, p. 163-170.

Ravelo, A.C., Andreasen, D.H., Lyle, M., Lyle, A.O. and Wara, M.W. 2004. Regional climate shifts caused by gradual global cooling in the Pliocene epoch. Nature, 429: 263-267.

Reguero, M.A., Candela, A.M. and Alonso, R.N. 2007. Biochronology and biostratigraphy of the Uquía Formation (Pliocene–early Pleistocene, NW Argentina) and its significance in the Great American Biotic Interchange. Journal of South American Earth Sciences, 23: 1-16.

Retallack, G.J. 1988. Field recognition of paleosols. In Geologic Time: Principles and Applications (Reinhardt, J. and Sigles, W.R. [Eds.]). Geological Society of America Special Paper 216, p. 1-20.

Roberts, N.J., Minaya, E., Guzmán, M.A., Zabala, M., Clague, J.J. and Hermanns, R.L. 2010. Overview of recent landslides in barrio Retamani, La Paz, Bolivia. In Resúmenes Extendidos del XV Congreso Peruano de Geología, 27 September-1 October, Cusco, Perú. Incluyendo e Integrando Recursos para la Sociedad. Sociedad Geológica del Perú, Lima, p. 328-331.

Roperch, P., Hérail, G., and Fornari, M. 1999. Magnetostratigraphy of the Miocene Corque basin, Bolivia: implications for the geodynamic evolution of the Altiplano during the late Tertiary. Journal of Geophysical Research, 104 (B9): 20,415- 20,429.

Rott, H., Scheuchl, B. and Siegel, A. 1999. Monitoring very slow slope movements by means of SAR interferometry: A case study from a mass waste above a reservoir in the Ötztal Alps, Austria. Geophysical Research Letters, 26: 1629-1632.

Rott, H. and Nagler, T. 2006. The contribution of radar interferometry to the assessment of landslide hazards. Advances in Space Research, 37: 710-719.

Rubin, M. and Alexander, C. 1958. U.S. Geological Survey Radiocarbon Dates IV. Science, 27: 1476-1487.

242

Ruiz, S., Metois, M., Fuenzalida, A., Ruiz, J., Leyton, F., Grandin, R., Vigny, C., Madariaga, R. and Campos, J. 2014. Intense foreshocks and a slow slip event preceded the 2014 Iquique Mw 8.1 earthquake. Science, 345: 1165-1169.

Safran, E.B., Biermann, P.R., Aalto, R., Dunne, T., Whipple, K.X. and Caffee, M. 2005. Erosion rates driven by channel network incision in the Bolivian Andes. Earth Surface Processes and Landforms, 30: 1007–1024.

Saignes, T. 1985. De los ayllus a las parroquias de indios: Chuquiabo y La Paz. In Ciudades de los Andes: Visión histórica y contemporánea (Kingman Garcés, E. [Ed.]). Institut Français d’Études Andines, Lima, p. 53-91.

Salzmann, U., Williams, M., Haywood, A.M., Johnson, A.L.A., Kender, S. and Zalasiewicz, J. 2011. Climate and environment of a Pliocene warm world. Palaeogeography, Palaeoclimatology, Palaeoecology, 309: 1-8.

Sanjinés, A.G. 1948. Síntesis Histórica de la Vida de la Ciudad, 1548-1948. Imprentas Asociadas, La Paz, Bolivia, 86 pp.

Sansosti, E., Berardino, P., Bonano, M., Calò, F. Castaldo, R., Casu, F., Manunta, M., Manzo, M., Pepe, A., Pepe, S., Solaro, G., Tizzani, P., Zeni, G. and Lanari, R. 2014. How second generation SAR systems are impacting the analysis of ground deformation. International Journal of Applied Earth Observation and Geoinformation, 28: 1-11.

Santa Cruz, V. 1942. Historia Colonial de La Paz. Editorial Renacimiento, La Paz, Bolivia, 348 pp.

Savage, W.Z. and Smith, W.K. 1986. A Model for the Plastic flow of Landslides. U.S. Geological Survey Professional Paper 1385, 32 pp.

Scanvic, J.-Y. and Girault, F. 1989. Imagerie SPOT-1 et inventaire des mouvements de terrain: l’exemple de La Paz (Bolivia). Photo Interpretation, 28: 1-20

Schlögel, R., Malet, J.-P., Doubre, C. and Lebourg, T. 2015. Structural control on the kinematics of the deep-seated La Clapière landslide revealed by L-band InSAR observations. Landslides: 14 pp. DOI 10.1007/s10346-015-0623-0

243

Schuster, R.L., Salcedo, D.A. and Valenzuela, L. 2002. Overview of catastrophic landslides of South America in the twentieth century. In Catastrophic landslides: Effects, occurrence and mechanisms (Evans, S.G. and DeGraff, J.V. [eds.]). Boulder, Colorado, Geological Society of America, Reviews in Engineering Geology 15, p. 1-34.

Servant, M. 1977. Le cadre stratigraphique du Plio-Quaternaire de l’Altiplano des Andes tropicales en Bolivie. Recherches Françaises sur le Quaternaire, INQUA 1977, Supplément au Bulletin AFEQ 1977-1, 50: 323-327.

Servant, M. and Fontes, J.-C. 1978. Les lacs quaternaires des hauts plateaux des Andes Boliviennes premières interprétations paléoclimatiques. Cahiers ORSTOM Série Géologie, 10: 9-23.

Servant, M., Sempere, T., Argollo, J., Bernat, M., Ferand, G., and Lo Bello, P., 1989. Morphogenèse et soulèvement de la Cordillère Orientale des Andes de Bolivie au Cénozoïque. Comptes Rendus de l'Académie des Sciences, Paris 309: 417- 422.

Shackleton, N. J. 1996. New data on the evolution of Pliocene climate variability. In Paleoclimate and Evolution, With Emphasis on Human Origins, (Vrba, S.E., Denton, G.H., Partridge, T.C. and Burckle, L.H. eds.). Yale University Press, New Haven, p. 242-248.

Shackleton, N. J., Hall, M.A. and Pate, D. 1995. Pliocene stable isotope stratigraphy of ODP Site 846. Proceedings of the Ocean Drilling Program - Scientific Results, 138: 337– 356.

Sharma, J., Kubanski, M., Eppler, J., Busler, J., Francioni, M., Stead, D., Clague, J. and Brideau, M.A. 2015. Monitoring geohazards near pipeline corridors with an advanced InSAR technique and geomechanical modelling. In Proceedings of 68th Canadian Geotechnical Conference and 7th Canadian Permafrost Conference, 20-23 September, Québec City, Canada. Unpagninated CD ROM, 8 pp.

Sicart, J.E. Ribstein, P., Francou, B. and Gallaire, R. 2003. Etude des précipitations et de la fonte sur unglacier tropical: le glacier du Zongo, Bolivie, 16°S. Hydrological Sciences Journal, 48: 799-808.

Silverman, H. and Isbell, W.H. 2008. The Handbook of South American Archaeology. Springer, 1192 pp.

Simpson, G.G., 1980. Splendid isolation: the curious history of South American mammals. Yale University Press, New Haven, 266 pp.

244

Singer, B.S., Brown, L.L., Rabassa, J.O. and Guillou, H. 2014. 40Ar/39Ar chronology of Late Pliocene and Early Pleistocene geomagnetic and glacial events in southern Argentina. In Timescales of the Paleomagnetic Field (Channell, J.E.T., Kent, D.V., Lowrie, W. and Meert, J.G. [eds]). American Geophysical Union Geophysical Monograph Series 145. Washington, D. C., p. 175-190.

Singhroy, V. 2009. Satellite remote sensing applications for landslide detection and monitoring. In Landslides – Disaster Risk Reduction, Proceedings of the Frist World Landslide Fourm, 18-21 November 2008 (Sassa, K. and Canuti, P. [Eds.]). Springer-Verlag, Berlin, p. 143-158.

Smith, J.A., Seltzer, G.O., Farber, D.L., Rodbell, D.T. and Finkel, R.C. 2005. Early local last glacial maximum in the tropical Andes. Science 308: 678-681.

Smyth, C.G. and Royle, S.A. 2010. Urban landslide hazards: incidence and causative factors in Niterói, Rio de Janeiro State, Brazil. Applied Geography, 20: 95-117.

Squarzoni, C., Delacourt, C. and Allemand, P. 2003. Nine years of spatial and temporal evolution of the La Valette landslide observed by SAR interferometry. Engineering Geology, 68: 53-66.

Strecker, M. 2011. Los petroglifos de Achocalla, La Paz – patrimonio perdido. Textos Antropológicos (Carreras de Antropología y Arqueología, Universidada Meyor de San Andrés), 16: 97-124.

Takahashi, K., Ravelo, A.C., Alvarez Zarikian, C. and the IODP Expedition 323 Scientists. 2011. IODP Expedition 323 – Pliocene and Pleistocene paleoceanographic changes in the Bering Sea. Scientific Drilling Journal, 11: 4- 13.

Tantianuparpe, P., Shi, X., Zhang, L., Balz, T. and Lia, M. 2013. Characterization of landslide deformations in Three Gorges Area using multiple InSAR data stacks. Remote Sensing, 5: 2704-2719.

Tarling, D.H. 1983. Palaeomagnetism: Principles and Applications in Geology, Geophysics, and Archaeology. Chapman and Hall, London, 379 pp.

Teshebaeva, K., Roessner, S., Echtler, H., Motagh, M., Wetzel, H.U. and Molodbekov, B. 2015. ALOS/PALSAR InSAR time-series analysis for detecting very slow- moving landslides in southern Kyrgyzstan. Remote Sensing, 7: 8973-8994

245

Thouveny, N. and Servant, M. 1989. Palaeomagnetic stratigraphy of Pliocene continental deposits of the Bolivian Altiplano. Palaeogeography, Palaeoclimatology, Palaeoecology, 70: 331-334.

Tofani, V., Raspini, F., Catani, F. and Casagli, N. 2013. Persistent Scatterer Interferometry (PSI) technique for landslide characterization and monitoring. Remote Sensing, 5: 1045-1065.

Ton-That, T., Singer, B., Mörner, N-A. and Rabassa, J. 1999. Datación de lavas basálticas por el metodo 40Ar/39Ar geología del Cenozoico superior de Ia región del Lago Buenos Aires, provincia de Santa Cruz, Argentina. Revista de la Asociación Geológica Argentina, 54: 333-352.

Troll C. and Finsterwalder R. 1935. Die Karten der Cordillera Real und des Talkessels von La Paz (Bolivien) und die Diluvialgeschichte der zentralen Anden. Petermanns Geographische Mitteilungen, 81: 393-399, 445-455.

UNDP. 2007. Andean capital cities 2007: Regional catalogue of municipal management tools in risk reduction and emergency preparedness (La Paz – Bolivia). United Nations Development Program report, Regional Project of Risk Reduction in Andean Capital Cities, 74 pp.

Vargas, E. 1977. Síntesis fotogeológica de la cuenca de La Paz y alrededores. Informe Geológico No. 3. Plan de desarrollo Urbano para la ciudad de La Paz: Honorable Alcaldía Municipal de La Paz, La Paz, 47 pp.

Vatin-Perignon, N., Poupeau, G., Oliver, R.A., Lavenu, L., Labrin1, F., Keller, F. and Bellot-Gurlet, L. 1996. Trace and rare-earth element characteristics of acidic tuffs from southern Peru and northern Bolivia and a fission-track age for the sillar of Arequipa. Journal of South American Earth Sciences, 9: 91-109.

Vuille, M. 1999. Atmospheric circulation over the Bolivian Altiplano during dry and wet periods and extreme phases of the southern oscillation. International Journal of Climatology, 19: 1579–1600.

Vuille, M., Bradley, R.S. and Keimig, F. 2003. Interannual climate variability in the Central Andes and its relation to tropical Pacific and Atlantic forcing. Journal of Geophysical Research, 105: 12447–12460.

Walther A.M., Orgeira, M.J., Reguero, M.A., Verzi, D.H., Chernoglasov, C., Vilas, J.F., Sinito, A.M. and Herrero Bervera, E. 1996. Nuevo estudio paleomagne´tico de la Formación Uquía de la provincia de Jujuy, Argentina. Revista de la Asociación Geológica Argentina, 51: 355–364.

246

Wang, J., Schmugge, T. and Williams, D. 1978. Dielectric Constants of Soils at Microwave Frequencies – II. NASA Technical Paper 1238. Goddard Space Flight Center, Greenbelt, Maryland, 36 pp.

Ward, B.C., Bond, J.D. and Gosse, J.C. 2007. Evidence for a 55–50 ka (early Wisconsin) glaciation of the Cordilleran ice sheet, Yukon Territory, Canada. Quaternary Research, 68: 141–150.

Webb, S.D. 1991. Ecogeography and the Great American Interchange. Paleobiology, 17: 266-280.

Williams, C.J., McNamara, J.P. and Chandler, D.G. 2009. Controls on the temporal and spatial variability of soil moisture in a mountainous landscape: The signature of snow and complex terrain. Hydrology and Earth System Sciences, 13: 1325- 1336.

Wisner, B., Blaikie, P., Cannon, T. and Davis, I. 2003. At risk: natural hazards, people’s vulnerability and disasters. Second edition. Routledge, London, 124 pp.

Woodburne, M.O. 2010. The Great American Biotic Interchange: dispersals, tectonics, climate, sea level and holding pens. Journal of Mammalian Evolution, 17: 245- 264.

Yin, Y., Zheng, W., Liu, Y., Zhang, J. and Li. X. 2010. Integration of GPS with InSAR to monitoring of the Jiaju landslide in Sichuan, China. Landslides, 7: 359-365.

Zachos, J.C., Pagani, M., Sloan, L., Thomas, E. and Billups, K. 2001. Trends, rhythms, and aberrations in global climate 65 Ma to present. Science, 292: 686-693.

Zebker, H.A. and Villasenor, J. 1992. Decorrelation in interferometric radar echoes. IEEE Transactions on Geoscience & Remote Sensing, 30: 950-959.

Zech, R., May J.-H., Kull, C., Ilgner, J., Kubik, P.W. and Veit, H. 2010. Timing of the late Quaternary glaciation in the Andes from ~15 to 40° S. Journal of Quaternary Science, 23: 636-647.

Zeilinger, G. and Schlunegger, F. 2007. Possible flexural accommodation on the eastern edge of the Altiplano in relation to focused erosion in the Rio La Paz drainage system. Terra Nova, 19: 373–380.

247

Zeilinger, G., Kober, F., Strecker, M., Ivy-Ochs, S., Kubik, P.W. and Hippe, K. 2009. Erosion rates in the Rio La Paz drainage basin: evidence for spatially and temporally variable erosion processes. Abstract, EGU Annual Meeting, Vienna, Austria.

Zhou, J. and Lau, K.-M. 1998. Does a monsoon climate exist over South America? Journal of Climate, 11: 1020–1040.

248