Pre-Earthquake Processes E-Book

Dimitar Ouzounov

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Serie: Geophysical Monograph Series
• Sprache: Englisch

Seitenzahl: 1063

Veröffentlichungsjahr: 2018

Beschreibung

Pre-Earthquake signals are advanced warnings of a larger seismic event. A better understanding of these processes can help to predict the characteristics of the subsequent mainshock. Pre-Earthquake Processes: A Multidisciplinary Approach to Earthquake Prediction Studies presents the latest research on earthquake forecasting and prediction based on observations and physical modeling in China, Greece, Italy, France, Japan, Russia, Taiwan, and the United States. Volume highlights include: * Describes the earthquake processes and the observed physical signals that precede them * Explores the relationship between pre-earthquake activity and the characteristics of subsequent seismic events * Encompasses physical, atmospheric, geochemical, and historical characteristics of pre-earthquakes * Illustrates thermal infrared, seismo-ionospheric, and other satellite and ground-based pre-earthquake anomalies * Applies these multidisciplinary data to earthquake forecasting and prediction Written for seismologists, geophysicists, geochemists, physical scientists, students and others, Pre-Earthquake Processes: A Multidisciplinary Approach to Earthquake Prediction Studies offers an essential resource for understanding the dynamics of pre-earthquake phenomena from an international and multidisciplinary perspective.

Leseprobe

Table of Contents

Cover

Preface

REFERENCES

Part I: Historical Development of Pre‐Earthquake Phenomena Studies

1 International Cooperation in Pre‐Earthquake Studies

REFERENCES

2 Earthquake Precursor Studies in Japan

2.1. INTRODUCTION

2.2. EARTHQUAKE PRECURSOR STUDIES IN JAPAN

2.3. FUTURE DIRECTION OF EARTHQUAKE PREDICTION

2.4. CONCLUSION

ACKNOWLEDGMENTS

REFERENCES

3 Pre‐Earthquake Observations and Their Application in Earthquake Prediction in China

3.1. INTRODUCTION

3.2. STUDIES OF PRE‐EARTHQUAKE PHENOMENA IN MAINLAND CHINA

3.3. SOME CASE STUDIES OF PRE‐EARTHQUAKE PHENOMENA OBSERVED OVER THE PAST DECADE

3.4. DISCUSSION

3.5. CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

4 Multidisciplinary Earthquake Precursor Studies in Taiwan

4.1. INTRODUCTION

4.2. FIRST MULTIDISCIPLINARY EARTHQUAKE PREDICTION RESEARCH PROGRAM IN TAIWAN (1979–1984)

4.3. INTEGRATED SEARCH FOR TAIWAN EARTHQUAKE PRECURSORS (iSTEP) PROGRAM (2002–2006)

4.4. FOLLOW‐UP iSTEP PROGRAMS

4.5. TYPES OF POSITIVE PRECURSORS IDENTIFIED FROM OBSERVATIONS OF THE 1999

M

7.6 CHI‐CHI EARTHQUAKE

4.6. POSITIVE PRECURSORS IDENTIFIED WITH SEVERAL RECENT MODERATE EARTHQUAKES IN TAIWAN

4.7. FUTURE PROSPECTS ON MULTIDISCIPLINARY EARTHQUAKE PRECURSOR STUDIES IN TAIWAN

ACKNOWLEDGMENTS

REFERENCES

5 Contributions to a History of Earthquake Prediction Research

5.1. INTRODUCTION

5.2. 600 BC TO AD 1500: THE FIRST EMPIRICAL OBSERVATIONS OF POSSIBLE EARTHQUAKE PRECURSORS

5.3. THE PERIOD 1500–1800: THE DIFFUSION OF GREEK KNOWLEDGE

5.4. THE PERIOD 1800–1920: THE BEGINNING OF THE SCIENTIFIC METHOD AND SYSTEMATIC OBSERVATION

5.5. RESEARCH STARTING IN THE MODERN AND CONTEMPORARY AGES: THE FORMER USSR AND CHINA

5.6. MODERN EARTHQUAKE PREDICTION STUDIES IN THE UNITED STATES AND JAPAN

5.7. RECENT EARTHQUAKE PREDICTION STUDIES IN EUROPE AND IN OTHER WORLD SITES

5.8. NEW PARAMETERS AND NEW MONITORING TECHNIQUES

5.9. BIOGRAPHICAL INFORMATION ON SOME SCIENTISTS INVOLVED IN EARTHQUAKE PREDICTION RESEARCH

5.10. CONCLUSIONS

REFERENCES

Part II: Latest Physical Models and Concepts of Pre‐Earthquake Processes

6 Lithosphere–Atmosphere–Ionosphere–Magnetosphere Coupling —A Concept for Pre‐Earthquake Signals Generation

6.1. INTRODUCTION

6.2. LITHOSPHERE–ATMOSPHERE INTERFACE

6.3. GEOCHEMICAL–THERMAL INTERFACE

6.4. GEOCHEMICAL–ELECTROMAGNETIC INTERFACE

6.5. THE SYNERGY OF EARTHQUAKE PRECURSORS AND INTEGRATED PARAMETERS

6.5. SUMMARY AND CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

7 Electrical Coupling Between the Ionosphere and Surface Charges in the Earthquake Fault Zone

7.1. PAST RESEARCH ON LITHOSPHERE–ATMOSPHERE–IONOSPHERE COUPLING MODELS

7.2. IONOSPHERE DYNAMICS IN THE PRESENCE OF AN ATMOSPHERIC ELECTRIC CURRENT

7.3. RESULTS

7.4. CASE STUDY A: PRESEISMIC TEC CHANGES ASSOCIATED WITH THE TOHOKU‐OKI EARTHQUAKE

7.5. CASE STUDY B: PRESEISMIC PLASMA DENSITY AND ION VELOCITY VARIATION FOR CHILE EARTHQUAKE

7.6. DISCUSSION

7.7. SUMMARY

ACKNOWLEDGMENTS

REFERENCES

Part III: Pre‐Earthquake Seismic Phenomena

8 Short‐Term Foreshocks and Earthquake Prediction

8.1. INTRODUCTION

8.2. FORESHOCK RATE

8.3. FORESHOCK PATTERNS

8.4. METHOD OF ANALYSIS AND COMPUTATIONAL NOTES

8.5. FORESHOCKS PRECEDING SMALL‐TO‐MODERATE GREEK EARTHQUAKES

8.6. DISCUSSION AND CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

9 Recent Developments in the Detection of Seismicity Patterns for the Italian Region

9.1. INTRODUCTION

9.2. SEISMICITY PATTERNS AND SEISMOTECTONIC INFORMATION

9.3. PROSPECTIVE TESTING OF THE CN ALGORITHM IN ITALY AND SURROUNDING AREAS

9.4. REDUCING SPACE–TIME UNCERTAINTY OF PREDICTIONS BY INTEGRATION OF MULTIDISCIPLINARY DATA

9.5. DISCUSSION AND CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

10 Combining Probabilistic Seismicity Models with Precursory Information

10.1. INTRODUCTION

10.2. METHODS

10.3. DATA

10.4. EXAMPLES OF APPLICATION OF THE DIFFERENTIAL PROBABILITY GAIN MODEL COMBINING METHOD

10.5. CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

Part IV: Ground Geochemical and Electromagnetic Observations of Major Seismicity

11 Continuous Monitoring of Fluid and Gas Geochemistry for Seismic Study in Taiwan

11.1. INTRODUCTION

11.2. GEOCHEMICAL ANOMALIES AS PRECURSORS TO SEISMIC EVENTS

11.3. MONITORING NETWORK OF SOIL‐GAS GEOCHEMISTRY FOR EARTHQUAKE IN TAIWAN

11.4. ANALYSIS OF DATA

11.5. SOME CLASSIC SOIL‐GAS PRECURSORS TO EARTHQUAKES IN TAIWAN

11.6. SUMMARY

ACKNOWLEDGMENTS

REFERENCES

12 Geochemical and Fluid‐Related Precursors of Earthquakes: Previous and Ongoing Research Trends

12.1. INTRODUCTION

12.2. THE ORIGIN OF FLUID‐RELATED PRECURSORS

12.3. EXPERIMENTS UNDERTAKEN THROUGHOUT THE WORLD

12.4. MECHANISMS RESPONSIBLE FOR FLUID‐RELATED ANOMALIES: CRUSTAL DEFORMATION

12.5. POSSIBLE EXPERIMENTAL APPLICATIONS

12.6. INDIRECT GEOCHEMICAL MONITORING BY SATELLITE TECHNIQUES

12.7. THE NEED TO IMPROVE SEISMIC AND CRUSTAL DEFORMATIONS MONITORING

12.8. CONCLUSIONS

DECLARATION

REFERENCES

13 Statistical Analysis and Assessment of Ultralow Frequency Magnetic Signals in Japan As Potential Earthquake Precursors

13.1. INTRODUCTION

13.2. JAPANESE ULF GEOMAGNETIC NETWORK OBSERVATION SYSTEM

13.3. DATA PROCESSING

13.4. STATISTICAL ANALYSIS

13.5. ASSESSING THE PRECURSORY INFORMATION

13.6. SUMMARY AND CONCLUSION

ACKNOWLEDGMENTS

REFERENCES

Part V: Atmospheric/Thermal Signals Associated with Major Earthquakes

14 Robust Satellite Techniques for Detecting Preseismic Thermal Anomalies

14.1. STATUS OF THE ART

14.2. ROBUST SATELLITE TECHNIQUE APPROACH AND ROBUST ESTIMATOR OF TIR ANOMALIES INDEX

14.3. LONG‐TERM ANALYSES FOR THE ASSESSMENT OF RST APPROACH POTENTIAL IN IDENTIFYING PRESEISMIC TIR ANOMALIES

ACKNOWLEDGMENTS

REFERENCES

15 Thermal Radiation Anomalies Associated with Major Earthquakes

15.1. INTRODUCTION

15.2. METHODS FOR RECOGNITION OF THERMAL RADIATION ANOMALIES

15.3. THERMAL ENERGY ASSOCIATED WITH SOME LARGE EARTHQUAKES

15.4. THERMAL ANOMALIES AND THE LITHOSPHERE–ATMOSPHERE COUPLING

15.4. SUMMARY AND CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

Part VI: Ionospheric Processes Associated with Major Earthquakes

16 Very‐Low‐ to Low‐Frequency Sounding of Ionospheric Perturbations and Possible Association with Earthquakes

16.1. INTRODUCTION

16.2. VERY‐LOW‐ TO LOW‐FREQUENCY SOUNDING OF THE LOWER IONOSPHERIC PERTURBATIONS

16.3. VLF–LF PERTURBATIONS DUE TO SOLAR–TERRESTRIAL EFFECTS AND METEOROLOGICAL EFFECTS

16.4. VLF–LF ANOMALIES IN ASSOCIATION WITH EARTHQUAKES

16.5. DISCUSSION OF THE GENERATION MECHANISM OF SEISMO‐IONOSPHERIC PERTURBATIONS

16.6. RESULTS FROM A NEW VLF–LF NETWORK IN JAPAN

16.7. OUTLOOK OF VLF–LF METHOD

ACKNOWLEDGMENTS

REFERENCES

17 Application of Total Electron Content Derived from the Global Navigation Satellite System for Detecting Earthquake Precursors

17.1. INTRODUCTION

17.2. SEISMO‐IONOSPHERIC PRECURSOR DETERMINATION FROM THE GIM TEC DATA

17.3. THREE DIMENSIONAL TOMOGRAPHIC ANALYSIS

17.4. SUMMARY

ACKNOWLEDGMENTS

REFERENCES

18 Statistical Analysis of the Ionospheric Density Recorded by the DEMETER Satellite During Seismic Activity

18.1. INTRODUCTION

18.2. THE DEMETER SATELLITE

18.3. THE OBSERVATIONS

18.4. THE STATISTICAL ANALYSES

18.5. DISCUSSION

18.6. CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

Part VII: Interdisciplinary Approach to Earthquake Forecasts/Predictions

19 Significant Cases of Preseismic Thermal Infrared Anomalies

19.1. INTRODUCTION

19.2. ILLUSTRATIVE CASES OF PRESEISMIC THERMAL INFRARED ANOMALIES

19.3. CONCLUSIONS

REFERENCES

20 Multiparameter Assessment of Pre‐Earthquake Atmospheric Signals

20.1. INTRODUCTION

20.2. MULTIPARAMETER REGISTRATION OF PRE‐EARTHQUAKE ANOMALIES

20.3. STATISTICAL ASSESSMENT OF PRE‐EARTHQUAKE ATMOSPHERIC ANOMALIES

20.4. SUMMARY AND CONCLUSIONS

ACKNOWLEDGMENTS

REFERENCES

Index

End User License Agreement

List of Tables

Chapter 03

Table 3.1 Statistical description of the multidisciplinary pre‐earthquake phenomena monitoring system operating in mainland China

Table 3.2 List of physical parameters of pre‐earthquake phenomena observed in mainland China

Table 3.3 Statistical relationship between the duration of pre‐earthquake phenomena and the magnitude of the eventual mainshock based on

Earthquake Cases in China

Chapter 06

Table 6.1 Mobility and diameter of cluster ions.

Table 6.2 Seismogenic electric field in the ionosphere estimated by different models.

Table 6.3 Model comparison

Chapter 07

Table 7.1 Possible anomalies associated with Earthquake activities.

Chapter 08

Table 8.1 Parameters of earthquake sequences, E, having significant foreshock patterns in the domains of space, time and magnitude before the earthquake events examined in the present paper.

Chapter 09

Table 9.1 Results of retrospective application of the CN algorithm in the Adriatic region, for different values of the magnitude threshold

M

0

, in the time span 1964–2004.

Table 9.2 Results of the seismic history experiment for application of the CN algorithm to the Adriatic region, performed for magnitude threshold

M

0

 = 5.4.

Table 9.3 List of target earthquakes for retrospective (1954–1997) and prospective (1998–2016) application of the CN algorithm in Italy and adjacent territory. Only the central and southern regions are considered up to 1964. Earthquakes that occurred during the period of real‐time monitoring (i.e., since 1998 for the north, centre and south regions and since 2005 for the Adriatic region) are indicated by bold typeface.

Table 9.4 Space–time volume of alarm indicated by the applied CN algorithm in Italy and its surroundings.

Chapter 10

Table 10.1 Likelihood test results for forecasts of the EAST

R

and RI reference models.

Table 10.2 Results of the likelihood tests for forecasts of four models.

Table 10.3 Intermediate‐term precursory patterns for the composite forecasting model.

Table 10.4 Correlation coefficients for the pairs of alarm functions.

Table 10.5 Comparison of likelihood test (equation 10.4) results for the composite and RI models.

Chapter 11

Table 11.1 Characteristic parameters of the geochemical precursors in Taiwan.

Chapter 12

Table 12.1 Most relevant review papers on possible precursory geofluids published in past 25 years.

Chapter 14

Table 14.1 State‐of‐art of satellite‐based methods used to identify thermal anomalies in relation with earthquake occurrence (updated after [

Tramutoli et al

., 2015a]).

Table 14.2 Main factors contributing to TIR signal variability.

Table 14.3 Reported cases of preseismic thermal anomalies identified by using the RST approach.

Table 14.4 Long‐term correlation analyses among SSTAs and earthquake occurrences.

Chapter 15

Table 15.1 List of earthquakes.

Table 15.2 Energy release comparison for the major earthquakes in Sumatra 2004 and 2005, and Tohoku, Japan, 2011.

Chapter 16

Table 16.1 Number of earthquakes identified below and above 40 km depth along each wavepath.

Chapter 18

Table 18.1 Number of earthquakes considered in past studies.

Table 18.2 Number of earthquakes meeting the limitations in this chapter.

Table 18.3 Statistical results concerning seismic ionospheric influences on the ion density close to the earthquake epicenters (

D

 = 1500 km,

d

 = 0–1000 km).

Table 18.4 Data list for Table 18.3.

Table 18.5 The statistical results concerning seismic ionospheric influences on the ion density (D = 1500 km, d = 0 − 20 km).

Table 18.6 Data list for Table 18.5.

Table 18.7 Statistical results concerning seismic ionospheric influences on the ion density close to the conjugated points of the earthquake epicenters (

D

 = 1500 km,

d

 = 0–1000 km).

Table 18.8 Data list for Table 18.7.

Table 18.9 Statistical results concerning seismic ionospheric influences on the ion density close to the conjugated points of the earthquake epicenters (

D

 = 1500 km,

d

 = 0 − 20 km).

Table 18.10 Data list for Table 18.9.

Chapter 19

Table 19.1 List of analyzed seismic events.

Chapter 20

Table 20.1 List of physical parameters used in ISTF analysis

Table 20.2 List of earthquakes (US Geological Survey) studied

Table 20.3 Summary of ACP, OLR, and dTEC anomalies observed in relation to the

M

6 Napa Valley, California,

Mw

6.4 Yujing, Taiwan, and

M

7 Kumamoto, Japan, earthquakes

Table 20.4 Contingency table for OLR analysis June 2014 to March 2015, Japan

List of Illustrations

Chapter 02

Figure 2.1 Conceptional general view of electomagnetic phenomena in possible association with earthquakes and different radiofrequency techniques to measure those electromagnetic effects.

Figure 2.2 History of seismo‐electromagnetic studies (including lithospheric and atmospheric effects, and the ionospheric signature.

Chapter 03

Figure 3.1 Conceptual model showing the relationship between the strain accumulation curve and pre‐earthquake phenomena during different stages of earthquake nucleation.

Figure 3.2 Maps showing the distribution of observation stations for monitoring pre‐earthquake phenomena in mainland China. (a) China National Digital Seismic Network. Red circles represent national seismic stations; gray circles represent regional seismic stations. (b) Other observation stations that recorded pre‐earthquake phenomena. Red circles represent geomagnetic observation stations; blue circles represent crustal deformation observation stations; gray circles represent groundwater observation stations.

Figure 3.3 Crustal Movement Observation Network in mainland China. Red stars indicate continuous GPS stations; blue circles show GPS campaign stations.

Figure 3.4 Distribution map of earthquakes in the book series

Earthquake Cases in China

during 1966–2006.

Figure 3.5 Frequency of pre‐earthquake phenomena that appeared prior to eventual mainshock earthquakes during 1986–2006, according to the book series

Earthquake Cases in China

.

Figure 3.6 Rates of pre‐earthquake phenomena for

M

5,

M

6, and

M

7 earthquakes, based on the book series

Earthquake Cases in China

.

Figure 3.7 Simplified tectonic map of the southeastern margin of the Tibetan Plateau. Red dots show the epicenters of the 2008 Wenchuan

Ms

8.0 earthquake, the 2013 Lushan

Ms

7.0 earthquake and the 2014 Ludian

Ms

6.5 earthquake; blue triangles show the locations of level‐measurement stations; pink triangles show the locations of regional groundwater‐observation stations.

Figure 3.8 Curves of observations collected at fixed baseline stations along the Xianshuihe fault. The black line and gray line show the motion of the two baselines in different directions. The data shown were collected at the following stations: (a) Xuxu, (b) Goupu, (c) Laoqianning, (d) Zhuwo. Station locations are shown in Figure 3.7.

Figure 3.9 Contours of changes (in μGal) before the 2008 Wenchuan earthquake: (a) in 1998–2000; (b) in 2000–2002. Gray lines show major active faults. Red circle shows the epicenter of the 2008 Wenchuan earthquake.

Figure 3.10 The GPS velocity field and regional shear‐strain‐rate fields prior to the Ludian earthquake. The red star represents the epicenter of the Ludian earthquake. (a) GPS site velocities relative to stable southern China during 2007–2013; (b) shear strain rate field during 2007–2011; (c) shear strain rate field during 2011–2013.

Figure 3.11 Curves of observations from fixed groundwater observation stations near the epicenter of Ludian earthquake. (After Liu

et al

. [2015b].) Red lines show the time of occurrence of the Ludian earthquake. The data shown were collected at the following stations: (a) Zhaotong, (b) Huize, (c) Ninnan, (d) Huidong, (e) Ludian, (f) Qiaojia. Station locations are shown in Figure 3.7.

Chapter 04

Figure 4.1 (a) Tectonic plates near Taiwan [

Hsieh et al

., 2014]. (b) Three‐dimensional projection of seismicity 1994–2015. CP, Coastal Plain; WF, Western Foothills; HR, Hsuehshan Range; BR, Backbone Range; ECR, Eastern Central Range; CoR, Coastal Range; LVF, Longitudinal Valley Fault; PH, Peikang High (a geometric basement high); KH, Kuanyin High (a geometric basement high); LMH, Lukang magnetization high; Manila and Ryukyu Trenches are drawn on the basis of bathymetry only.

Figure 4.2 (a) GPS horizontal and (b) vertical velocities in Taiwan: S01R (23.66°N, 119.59°E) is the reference station of the GPS network.

Figure 4.3 (a) Epicenters of

M

5.0+ earthquakes in Taiwan, 1900–2014 [

Chang et al

., 2016]. (b) Intensive shaking areas of the five most disastrous earthquakes in western Taiwan during the 20th century.

Figure 4.4 (a) Observational sites. (b) Triangulation results in eastern Taiwan from the multidisciplinary earthquake prediction research program in Taiwan, 1979–1984 [

Tsai et al

., 1983].

Figure 4.5 (a) Comparison of temporal variations of radon content at four hotsprings with seismicity in northern Taiwan [

Tsai et al

., 1983]. (b) Gravity anomaly map of Taiwan [

Yen and Hsieh

, 2010].

Figure 4.6 (a) Geomagnetic anomaly map of Taiwan [

Yen et al

., 2009]. (b) Observational stations under iSTEP. LP, Lunping (25.00°N, 121.17°E) is the ionosode station; NC, Neicheng (24.72°N, 121.68°E), LY, Liyutan (24.35°N, 120.78°E), HL, Hualien (24.07°N, 121.61°E), YL, Yuli (23.35°N, 121.29°E), TW, Tsengwen (23.25°N, 120.53°E), TT, Taitung (22.80°N, 121.06°E) and HC, Hengchun (21.94°N, 120.81°E) are the geomagnetic stations; YMSM is the Yangmingshan GPS station. [

Tsai et al

., 2006].

Figure 4.7 (a) The main component and five subcomponent projects [

Tsai et al

., 2004]. (b) Precursors of the

M

7.6 Chi‐Chi earthquakee identified as of 2006 under the multidisplinary “integrated Search for Taiwan Earthquake Precursors” (iSTEP) program [

Tsai et al

., 2006].

Figure 4.8 (a) Spatial variations of

P

‐wave arrival‐time residuals. (b) Temporal variations of

P

‐wave arrival‐time residuals during three periods before and after the 1999

M

7.6 Chi‐Chi earthquake in Taiwan [

Lee and Tsai

, 2004].

Figure 4.9 (a) Distribution of Taiwan seismicity (

M

3.4+, January 1987 to June 1999); (b) Distribution of pattern informatics (PI) index of Taiwan seismicity during 3 months before the 1999

M

7.6 Chi‐Chi earthquake [

Chen et al

., 2005]. CLP, the Chelungpu fault. The red dashed lines denote the top 5% active area of historic coarse‐graining intensity of earthquakes. Circles represent earthquakes with

M

 ≥ 6 that occurred after June 1999 and inverted triangles represent earthquakes with

M

 ≥ 6 that occurred between November 1993 and June 1999. Colored pixels (hotspots) represent areas with large seismicity change caused by both seismic activation and quiescence, indicating high probability for future large events.

Figure 4.10 (a) Distribution of Taiwan seismicity, 1994–2005. (b) Time series of the monthly earthquake occurrence, 1994–2005 [

Wu and Chen

, 2007].

Figure 4.11 (a)Temporal and spatial variations of the

b

‐value near the Chelungpu fault zone before the

M

7.6 Chi‐Chi earthquake. (b) Temporal variations of InSAR images in areas near the Chelungpu fault [

Tsai et al

., 2005].

Figure 4.12 (a) GPS stations in Taiwan (left) and groundwater‐level monitoring wells in central Taiwan (right). (b) Groundwater level records at six wells before and after the

M

7.6 Chi‐Chi earthquake with (top) barometric and precipitation records in central Taiwan [

Chen et al

., 2015a]. Abbreviations YMSM, FIVE, BANC, HISN, ILAN, SUAO, SANI, HUAL, SUNM, FLNM, PKGM, CHIA, CHEN, TMAM, HENC, and KDNM stand for the GPS receiving stations. Groundwater stations with significant anomalies related to the Chi‐Chi earthquake are denoted by red open circles, those without by blue open rectangles.

Figure 4.13 (a) Correlation between temporal variations of horizontal‐strain patterns in areas surrounding the Chelungpu fault in Central Taiwan and a groundwater‐level record before the 1999

M

7.6 Chi‐Chi earthquake [

Chen et al

., 2015a]. (b) Segments of a GPS record at San‐Yi (SANI; 24.41°N, 120.77°E) site during (left) 21 September 1998 to 20 March 1999 and (right) 21 March 1999 to 20 September 1999. Note the changes in the directions of motion in the (top) N–S and (bottom) vertical components.

Figure 4.14 (a) Geomagnetic total intensity (GTI) records at the reference station LP. (b) The difference in GTI between stations Lunping (LP) and Liyutan (LY) located near the northern end of the Chelungpu fault before and after the 1999

M

7.6 Chi‐Chi earthquake in Taiwan [

Yen et al

., 2004].

Figure 4.15 (a) Daily variations of ionospheric total electron content (TEC) in September 1999; (b) Spatial variations of TEC on 17 September 1999 [

Liu et al

., 2004b]; LT, local time; P, seismo‐ionospheric precursor of the GPS TEC.

Figure 4.16 (a) Daily lightning occurrences during 6 September 1999 to 6 October 1999 around the Chelungpu fault area. (b) Spatial patterns of hourly lightning occurrences on 17 September 1999 in the Taiwan region.

Figure 4.17 (a) Map of the Central Weather Bureau Seismic Network of Taiwan. (b) Map of the Broadband Array in Taiwan for Seismology (BATS) network.

Figure 4.18 (a) Map of continuous GPS stations in Taiwan. (b) Map of GPS geomagnetic and groundwater monitoring stations in the CWBSN

Figure 4.19 (a) Map of the CWBSN groundwater‐level monitoring stations [

Shin et al

., 2013]. (b) Map of WRA groundwater‐level monitoring stations [

WRA

, 2001]. TUN, MWA, DON, LIU, NAU, CHI, and GW are groundwater‐level monitoring station locations.

Figure 4.20 (a) Map of the Central Geological Survey gas and fluid geochemical monitoring stations. (b) Automatic fluid geochemical monitoring stations on the Chukou fault [

Yang et al

., 2015].

Figure 4.21 (a) Map of gamma‐ray monitoring stations operated by the Institute of Earth Sciences [

Fu et al

., 2015]. (b) Map of geoelectric monitoring stations operated by the National Central University [

C. C. Chen et al

., 2015]. YMSG, DHUG, CCUG, and KTPG denote gamma‐ray monitoring stations.

Figure 4.22 The three‐component geomagnetic stations in Taiwan operated by the Institute of Earth Science and National Chung Cheng University. YMM, TCD, CCU, HLG, and DNA are three‐component geomagnetic stations.

Chapter 05

Figure 5.1 Tromometric measurements (angular displacements by a pendulum‐based instrument) recorded in Bologna are in black. Measurements recorded in Florence, which is 100 km away, are indicated in blue. Before the seismic event characterized by

M

5.2, the time series are uncorrelated [

Ferrari et al

., 2000].

Figure 5.2

V

p

/

V

s

measurements studied by

Whitcomb et al

. [1973] in California. Before a

M

6.7 event, a significant variation was observed in the period 1967–1969.

Chapter 06

Figure 6.1 Thermal infrared (TIR) anomaly registered 5 days before the L’Aquila

M

6.8 earthquake of 6 April 2009 (yellow and red). Blue circle, the earthquake preparation zone zccording to

Dobrovolsky et al

. [1979]; red circle, the critical area according to

Bowman et al

. [1998]. The earthquake epicenter indicated by the red cross.

Figure 6.2 From top to bottom: seismic activity in the Kobe area for the period 1990–1995;

b

‐value variations for the same period of time; fractal dimension

D

2

for seismic activity in the Kobe area; radon activity (in water) for the period from November 1993, the time correspondence is indicated by arrows.

Figure 6.3 Schematic presentation of the geochemical–thermal interface.

Figure 6.4 Dependence of the possible air heating Δ

T

on the initial relative humidity

H

0

for different initial air temperatures

T

 = 5 (1), 10 (2), 15 (3), 20 (4), 25 (5), 30 (6), and 35 (7) °C under

f

 = 40 (cm

3

⋅s)

−1

and

τ

 = 10 h.

Figure 6.5 Schematic presentation of the geochemical–electromagnetic interface: VLF, very low frequency; EF, electric field; AGW, acoustic gravity waves.

Figure 6.6 (a) Bottom panel: schematic conception of atmosphere–ionosphere coupling through the global atmospheric electric circuit; left panel: for a condition of increased air conductivity; right panel: for a condition of decreased air conductivity. IP, ionosphere potential. Upper panel: the differential maps obtained from the global ionosphere maps GPS total electron concentration data for the period before the Wenchuan earthquake on 12 May 2008; left panel: two‐dimensional distribution obtained on 3 May 2008; right panel: two‐dimensional distribution obtained on 9 May 2008. (b) three‐dimensional presentation of the left panel of Figure 6.6a for the case of increased air conductivity.

Figure 6.7 Latitudinal (geomagnetic latitude) profiles of morning 10 am local time half‐orbits of the DEMETER satellite for 4–5 May 2008. The two‐hump curve (red) represents the satellite orbit closest to the epicenter Δ

long

 = 5.5°, orbit 20515.

Figure 6.8 (a) Schematic presentation of the atmospheric boundary layer and its daily variations: CBL and NBL are the convective and nocturnal boundary layers respectively. (b) Daily variations of aerosol distribution after

Eresmaa et al

. [2012]. (c) Upper border of the atmospheric boundary layer recorded by radon in atmosphere.

Figure 6.9 Temporal dynamics of seismic activity, radon release, and variations of atmospheric and ionospheric parameters before the L’Aquila earthquake: OLR, outgoing longwave radiation.

Figure 6.10 Upper panel: correction of chemical potential time series around the time of the

M

6.7 earthquake near the east coast of Kamchatka on 20 March 2016. Lower panel: histogram of main maximum temporal position of the chemical potential in relation to the day of the main shock for 11 years of observation in Kamchatka area for earthquakes with magnitude

M

 > 6 (red), and the same for the main minimum chemical potential (blue).

Chapter 07

Figure 7.1 Diagram to demonstrate electric coupling between lithosphere–atmosphere–ionosphere.

Figure 7.2 As rocks are subjected to forces of stress in an earthquake preparation region, deformation of the lattice structure in rocks can produce electronic charge carriers (positive holes, h

•

) and electric currents (

J

rock

). The positive holes diffuse from the highly stressed region into the unstressed region, repel each other electrostatically, and push toward the rock surface. The field‐ionization (O

2

 → O

2

+

+ e

−

) occurring at the surface, in particular at sharp points, can produce the O

2

+

ions. The surface charge density

Σ

surf

and total charge amount

Q

are accumulated by positive charges from the rock surface and O

2

+

ions in the air. The upward electric field

E

air

associated with the positive surface charges will drive the upward current

J

air

[

Kuo et al

., 2014].

Figure 7.3 (a) Model with two charge layers imbedded in the lithosphere. (b) Model with dynamo (battery) current source.

Figure 7.4 The effect of Pedersen and Hall conductivity. The rotated coordinates (

x

′,

y

′,

z

′) are shown by the black line with arrows. The positive direction of the

x

′ axis and

y

′ axis are, respectively, in the eastward and northward direction in the local coordinates of Earth. For an upward current density

J

, the parallel and perpendicular components of current density

J

are along the

z

′ axis and

y

′ axis, respectively. The perpendicular (a) upward and (b) downward current

J

⊥

leads to the presence of a Pedersen electric field (

E

P

) along the + 

y

′ (−

y

′) axis and the Hall electric field (

E

H

) along the + 

x

′ (−

x

′) axis [

Kuo et al

., 2014].

Figure 7.5 The vector electric field (white arrows) and contours of the electric field magnitude produced by (a) upward and (b) downward currents at magnetic latitude 15° and by (c) upward and (d) downward currents at magnetic latitude 30°. The Hall conductivity is assumed to be ten times the Pedersen conductivity 10

σ

P

 ≈ 

σ

H

in the E layer of the ionosphere. (e) The magnitude of the electric field is plotted as a function of the current for magnetic latitude 15° and 30° in the daytime. (f) The magnitude of the electric field is plotted as a function of the current with magnetic latitude 15° and 30° in the nighttime [

Kuo et al

., 2014]. mLat, magnetic latitude.

Figure 7.6 The velocities and contour plots of electron density in the meridional plane above the current source region at magnetic latitude 15° for (a) upward currents (westward electric field) and (b) for downward currents (eastward electric field), and at 30° for (c) upward and (d) downward currents. Black lines attached to a black dot indicate velocities of ionospheric plasma, where the direction of motion is away from the dot [

Kuo et al

., 2014].

Figure 7.7 TEC variations (%) for the case with upward (downward) current are shown in the left (right) column. The panels from up to down are results for magnetic latitude 7.5°, 15°, 22.5°, and 30°, respectively. The open circle indicates the location of the current source (earthquake epicenter) [

Kuo et al

., 2014].

Figure 7.8 Maximum ΔTEC (%) varies with source current density

J

max

(nA m

−2

) where the solid (dashed) lines are for ΔTEC at latitude 30°(15°). The blue (black) lines are for daytime (nighttime) ionospheric conditions [

Kuo et al

., 2014]. mLat, magnetic latitude.

Figure 7.9 Double plasma bubbles are triggered by zonal westward electric field (upward current) at low magnetic latitude 15°: (a) contour plots of electron density

n

e

in the equatorial plane at different times for the nighttime case; (b) contour plots of ΔTEC during applied westward electric field; (c) contour plots of TEC during formation of plasma bubble [

Kuo et al

., 2014].

Figure 7.10 A single plasma bubble is triggered by zonal eastward electric field (downward current) at low magnetic latitude 15°: (a) contour plots of electron density

n

e

in the equatorial plane at different times for the nighttime case; (b) contour plots of ΔTEC during applied westward electric field; (c) contour plots of TEC during formation of plasma bubble [

Kuo et al

., 2014].

Figure 7.11 Maximum current density linearly increases from zero to its maximum value in the 40‐min period (UT 05:06–05:46) before the main shock [

Kuo et al

., 2015].

Figure 7.12 An ionospheric anomaly caused by downward current at magnetic latitude 30°; (a) the downward current leads to the presence of an eastward electric field and the resulting

E

 × 

B

motion enhances the ionospheric plasma density; (b) contour plots of ΔTEC in TECU where the open circle indicates the source region; (c) contour plots of electron density

n

e

in the meridional planes; (d) contour plots of electron density variations Δ

n

e

in the meridional planes; (e) temperature variations in the meridional planes [

Kuo et al

., 2015].

Figure 7.13 The observed results of ΔTEC from the Japanese GEONET where color code indicates the magnitude of TEC in a time sequence of (a) 21 min, (b) 10 min, and (c) 1 min before the main shock of the earthquake. The corresponding TEC contour lines from our simulation results are plotted in (d), (e), and (f). The corresponding ΔTEC from our simulation results are plotted in (g), (h), and (i) [

Kuo et al

., 2015].

Figure 7.14 Comparison of modeling results (red dots) with observed ΔTEC (blue dots) in TECU at UT 05:45, 1 min before the main shock: (a) profile at longitude 139°, (b) profile at longitude 140°, (c) profile at longitude 141°, (d) profile at latitude 36°, (e) profile at latitude 38°, and (f) profile at latitude 40° [

Kuo et al

., 2015].

Figure 7.15 Variations of space environments and the ionosphere before and after the Chile earthquake. (a) The geomagnetic index Kp and the solar activity index F10.7. (b and c) Temporal profile of electron density and ion velocity respectively. The red line represents the observations, and the blue and dotted lines the upper and lower bounds and the median. The pink and the light blue bars denote the percentage variations of electron density (|∆

N

e

|

N

e

) and perpendicular velocity (|∆

V

⊥

|

V

⊥

). The geomagnetic latitude of the epicenter is 25.74°S. (d) Comparison between the simulation (blue) and the observation (black) results before the 2010 Chile earthquake. The dots denote the perpendicular velocity and the density variations. The black (blue) best‐fit line passes through the origin from DEMETER observations (SAMI3 simulations), and

m

is the slope.

Figure 7.16 Comparison of increased anomalies in 2008, 2009, and 2010 (lower to upper panels). From left to right the panels display anomalies of electron density, parallel velocity, and perpendicular velocity. The red bars show the values of ∆

N

e

, ∆

V

⊥

and ∆

V

ıı.

Figure 7.17 Variations of space environments and the ionosphere before and after the 2009 Samoa Islands earthquake. The geomagnetic latitude of the epicenter is 16.86°S.

Figure 7.18 Variations of space environments and the ionosphere before and after the 2009 Kermadec Islands earthquake. The geomagnetic latitude of the epicenter is 29.26°S.

Figure 7.19 Relations between perpendicular velocity and density variation with different imposed electric fields at the epicenter latitude. The dots denote perpendicular velocity and density variation. The lines are best‐fit lines and

m

is slope.

Figure 7.20 The ionospheric density and velocity distributions caused by an imposed eastward electric field (20 mV m

−1

). (a) Density and velocity distributions in the meridional plane. The two left‐hand panels are the density without and with an electric field. The two right‐hand panels are the density variation and the velocity perpendicular to the geomagnetic field. (b) Relations between perpendicular velocity and density variations at a different altitude. From top to bottom panels the data represent 670 km, 400 km and 200 km.

Figure 7.21 Simulation results with an eastward electric field (

E

 = 20 mV m) at geomagnetic latitudes 5°, 10°, 15°, 20°, 25°, and 30°. From top to bottom the panels are electron density, density variation, parallel velocity, and perpendicular velocity.

Figure 7.22 Relation between perpendicular velocity and density variation with an imposed electric field (20 mV m

−1

) at different latitudes. The dots correspond to the data points of the perpendicular velocity and the density variation. The lines are their best‐fit lines and

m

is the slope.

Chapter 08

Figure 8.1 Earthquake epicenter plots in the area of Athos, northern Aegean Sea, Greece, for two time periods: (top) from 1 February 2011 to 21 December 2011 and (bottom) from 22 December 2011 to 4 March 2012. Circles are radii of 50 and 20 km from the epicenter (target point) of the 4 March 2012 main event (star). After 21 December 2011 the seismicity was significantly clustered close to the main event epicenter.

Figure 8.2 Time variation of the average Euclidean distance,

D

, of epicenters from that of the main event of 4 March 2012 (target point). At each point of time,

D

was calculated as the average of five events.

Figure 8.3 Time variation of the average Euclidean distance,

D

, of epicenters from the target point. At each point of time,

D

was calculated with the backward technique as the average of 30, 50, 100, or 150 events starting from the origin time of the 4 March 2012 main event and using a step of 1 event.

Figure 8.4 (Top) Magnitude–frequency (G–R) distribution of earthquakes occurring in the background seismicity period (BGS) and the foreshock period (FOR) before the main event of 4 March 2012:

M

c

, cut‐off magnitude;

N

b

, number of events in BGS;

N

f

, number of events in FOR;

r

b

, seismicity rate in BGS;

r

f

, seismicity rate in foreshock period;

N

min

, minimum number of events;

b

ML

and

b

GR

are calculated by the maximum‐likelihood and the weighted least‐squares methods, respectively. In FOR the

b

ML

remains lower than that in BGS for different

M

c

levels (bottom). Calculation performed for

N

min

 = 30 with a magnitude step of 0.10:

CF

, confidence interval;

P

b

, Utsu probability test;

P

z

,

z

probability‐test.

Figure 8.5 Cumulative number of earthquake events of magnitude

M

 ≥ 

M

c

 = 1.40 occurring in the target area (

R

 = 20 km) around the epicenter of the 4 March 2012 main event (target point). See Figure 8.4 for notation.

Figure 8.6 Time variation of (upper) mean magnitude, <

M

>, and (lower)

b

ML

of earthquake events of

M

 ≥ 

M

c

 = 1.40 occurring in the target area (

R

 = 20 km) around the target point. Both parameters were calculated with the backward technique by taking window

w

 = 30 events starting from the origin time of the 4 March 2012 main event and using a step of 1 event.

Figure 8.7 Cumulative number,

N

, of earthquake events of magnitude

M

 ≥ 

M

c

 = 1.40 occurring until the end of April 2012 in the target area (

R

 = 20 km) around the target point. Vertical blue and red dotted bars show the origin times of the strong foreshock of 14 February 2012 and of the main shock, respectively. See Figure 8.4 for notation.

Figure 8.8 Cumulative number of earthquake events of

M

 ≥ 

M

c

 = 1.40 occurring in the target area (

R

 = 20 km) around the target point in the BGS and FOR periods; see Figure 8.4 for notation. The 14 February 2012 earthquake coincides with the right‐hand vertical axis of both panels.

Figure 8.9 (Top) Magnitude–frequency (G–R) distribution of earthquakes occurring in the BGS and FOR periods before the

Mw

5.1 earthquake of 14 February 2012. See Figure 8.4 for notation. In the FOR period the

b

ML

remained lower than that in the BGS period for different

M

c

levels (bottom). Calculation performed for

N

min

 = 30 with a magnitude step of 0.10.

Figure 8.10 Time variation of (top) mean magnitude, <

M

>, and (bottom)

b

ML

of earthquake events of

M

 ≥ 

M

c

 = 1.40 occurring in the target area (

R

 = 20 km) around the target point. Both parameters were calculated with the backward technique by considering

w

 = 30 events starting from the origin time of the

Mw

5.1 earthquake of 14 February 2012 and using a step of 1 event.

Figure 8.11 Earthquake epicenter plots in the Polyphyto area, northern Greece, for two time periods: (top) from 1 February 2011 to 7 November 2012 and (bottom) from 8 November 2012 to 3 July 2013. Circles are radii of 50 and 20 km from the epicenter of the 4 March 2013 main event (star). After 7 November 2012 the seismicity was clustered close to the main event epicenter at the southwestern tip of the Polyphyto artificial lake.

Figure 8.12 Time variation of the average Euclidean distance,

D

, of epicenters from that of the main event of 3 July 2013. At each point of time

D

was calculated as the average of five events.

Figure 8.13 Time variation of the average Euclidean distance,

D

, of epicenters from that of the main event of 3 July 2013. At each point of time

D

was calculated with the backward technique as the average of

n

events (

n

was equal to 30, 50, 100, or 150) starting from the origin time of the 3 July 2013 main event and using a step of 1 event.

Figure 8.14 (Top) Magnitude–frequency (G–R) distribution of earthquakes occurring in the background seismicity period (BGS) and the foreshock period (FOR) before the main event of 3 July 2013. Notation as in Figure 8.6. In the foreshock period the

b

ML

remains lower than that in the background period for different

M

c

levels (bottom). Calculation performed for

N

min

 = 30 with a magnitude step of 0.10.

Figure 8.15 Cumulative number of earthquake events of magnitude

M

 ≥ 

M

c

 = 1.10 occurring in the target area (

R

 = 25 km) around the epicenter of the main event of 4 March 2012. See Figure 8.4 for notation.

Figure 8.16 Time variation of (top) mean magnitude, <

M

>, and (bottom)

b

ML

of earthquake events with magnitude

M

 ≥ 

M

c

 = 1.10 occurring in the target area (

R

 = 25 km) around the epicenter of the main event of 3 July 2013.

Chapter 09

Figure 9.1 Regionalization identified for the application of the CN algorithm in Italy, defined according to the seismotectonic model [

Meletti et al

., 2000 and references therein]. The magnitude thresholds

M

0

for each region are indicated. In the diagrams of the TIP, the black boxes represent periods of alarm, while each triangle indicates the occurrence of a strong event (

M

 ≥ 

M

0

), together with its magnitude. Two numbers above a triangle indicate the occurrence of two target earthquakes within the same prediction time window (2 months). Full red triangles indicate failure to predict.

Figure 9.2 Regionalization identified for the application of the CN algorithm to the Adria plate (blue polygon), defined based on the seismotectonic model (thin lines) proposed by

Meletti et al.

[2000]. Red dots mark the epicenters of main shocks with

M

 ≥ 3.0 reported in the UCI catalog in the time interval 1900–2005. Yellow stars mark the epicenters of target events, i.e., earthquakes with

M

 ≥ 

M

0

 = 5.4 that have occurred since 1964. The diagram shows the temporal distribution of TIP obtained for the Adriatic region. Key to symbols is the same as in Figure 9.1.

Figure 9.3 Distribution of the number of events versus magnitude for earthquakes that have occurred within the Adriatic region, as reported in the UCI catalog for the time interval 1900–2005. Red dots correspond to the target earthquakes with

M

 ≥ 

M

0

 = 5.4, whereas the blue squares correspond to the events in the magnitude range considered for the computation of CN functions.

Figure 9.4 Error diagram [

Molchan

, 1990] for the results of the applied CN algorithm in Italy detailed in Table 9.4:

η

, percentage of failure to predict;

τ

, total space–time volume occupied by alarms. The diagonal line corresponds to the results of a random guess (gain

G

 = 1). The blue dots correspond to the results obtained for the three experimental time intervals considered in Table 9.4. The prediction results, obtained considering together all retrospective and real‐time testing, are indicated by the red circle, and the curves corresponding to 95% and 99% confidence levels are shown (the CN algorithm score is above 99%, see text).

Figure 9.5 Example of an integrated alert map for the Italian territory, obtained from geodetic and seismicity data: (a) GNSS anomalies at a given time (green, polygons with normal behavior with respect to the mean tectonic deformation; red, anomalous polygons; grey, areas where no information is available. (b) Map of territory monitored (grey, yellow and blue areas) and alerted (yellow and blue areas) by the applied CN algorithm, based on the analysis of seismicity patterns (as in Figure 9.1). (c) Joint map of seismic and geodetic anomalies. Different colors correspond to different combinations of seismic and geodetic anomalies.

Figure 9.6 Time‐dependent scenarios of ground shaking associated with an alarm in each of the regions (shown in Figures 9.1 and 9.2) monitored by the applied CN algorithm: (a) Northern, (b) Central, (c) Southern, (d) Adriatic. Peak ground velocity (PGV) is computed using simultaneously all of the possible sources within the alarmed area and for frequencies up to 10 Hz. Only values of PGV ≥ 15 cm s

−1

are provided in the maps, which correspond to IMCS ≥ IX [

Indirli et al

., 2011].

Figure 9.7 Example of ground‐shaking scenario associated with a seismogenic node: (a) map of earthquake prone nodes identified for

M

N

 ≥ 6.0 (light blue) and

M

N

 ≥ 6.5 (dark blue) according to

Gorshkov et al.

[2002, 2004]. The borders of the Adriatic region (red polygon) and the selected node D70 are outlined in the map. (b) Peak ground velocity (PGV), computed for the potential seismic sources associated with the selected node (large green circle). The maximum observed value is 2.1 cm s

−1

(http://itaca.mi.ingv.it), well below the values reported in the figure.

Chapter 10

Figure 10.1 Schematic error diagram for a prediction method. The trajectory (thick line) is formed by points corresponding to different values of the alarm function. The diagonal (dashed line) corresponds to arbitrary predictions.

Figure 10.2 Error diagrams for binary and discrete alarm functions. In the binary case (a) the alarm function has only two values (“yes”/“no”) and the trajectory has two slopes. In the discrete case (b) several segments with specific slopes correspond to several alarm function values. Differential probability gain

g

(

A

) for each discrete value of the alarm function is equal to the slope of the corresponding segment. The continuous alarm function may be discretized using a set of thresholds.

Figure 10.3 The area of the test for the EAST, EAST

R

, and EAST*EEPAS models. The CSEP California testing region is outlined by the dashed line and the heavier shaded area outlined by the thick line was used for adapting parameters of the EAST model. Dots mark epicenters of

M

 ≥ 4 events during the period January 2011 to March 2016.

Figure 10.4 Error diagram for EAST and EAST

R

forecasts for the period January 2011 to March 2016 with respect to the RI model. The plots for EAST forecasts (thin line) and EAST

R

forecasts (thick line) are close to each other. The dashed diagonal line corresponds to unskilled forecasts. The shaded area indicates the zone in which the forecast of the tested models outperforms the forecast of the reference model at a level of significance

α

 = 1%.

Figure 10.5 Differential probability gain functions for the EAST model with respect to the EEPAS model showing estimation of the differential probability gain functions,

, of the EAST forecast model with respect to the EEPAS model for California from January 1984 to June 2009. (a–c) Error diagrams in which we smooth the trajectory (thin line) by a set of segments (thick lines; see section 10.2.2.2). The

value is the local slope of these segments. (d–f) Differential probability gain

as a function of the alarm function

A

EAST

of the EAST models.

Figure 10.6 Quasi‐prospective evaluation of the EAST*EEPAS, EAST

R

, and EEPAS models. Tests were carried out in (a) the California CSEP testing region and in (b) the reduced region indicated in Figure 10.3. The forecasts of the EAST*EEPAS (thick lines), EAST

R

(thin lines), and EEPAS (thin dashed lines) models are compared with respect to the RI reference model. The dashed diagonal line corresponds to an unskilled forecast. See Figure 10.4 for explanation of shaded area. For all the models we consider single rate values obtained by summing the expected rates of

M

 ≥ 4.95 target earthquakes.

Figure 10.7 Three‐month forecasts of the (a) EAST, (b) EASTR, (c) EEPAS, and (d) EAST*EEPAS models for

M

 ≥ 4.95 earthquakes from April to June 2010 in northern Baja, California along the USA–Mexico border. Circles correspond to

M

 ≥ 4.95 earthquakes that occurred in this area during this period. For the EAST model the color map varies from zero to the maximum of the alarm function

A

EAST

(equation 10.14). For other models, the same color bar is used to represent the forecast rates of

M

 ≥ 4.95 earthquakes. Note the higher contrast for the EAST*EEPAS forecasts and the increase in event rate in zones where both the EAST

R

and EEPAS models have high event rates. Straight line is the USA–Mexico border.

Figure 10.8 Expected rate distributions of EAST*EEPAS and EAST

R

 + EEPAS models. (a) Cumulative distribution functions of the rate values (thick line for EAST*EEPAS and thin line for EAST

R

 + EEPAS). Dots show the rates in the space–time region where target earthquakes have occurred during the quasi‐prospective test from July 2009 to December 2011. (b) Linear‐scale and (c) logarithmic‐scale rates of the EAST*EEPAS model with respect to rates of the EAST

R

 + EEPAS model. Open circles correspond to the entire CSEP testing region; black dots correspond to the reduced region (Figure 10.3). The

x

 = 

y

line is shown for direct comparison. The EAST*EEPAS model has remarkably higher rate values in the limit of large rate (b). These higher rates are compensated in the model by lower rates in the limit of small rate.

Figure 10.9 Level of noise in combined forecast models using the differential probability gains approach. The EAST*EEPAS model is successively combined with 10 random rate‐based models for

M

 > 4.95 target earthquakes (see text). (a) Differential probability gain

g

(

A

) estimated during the learning period from January 1984 to June 2009. (b) For the same period, a comparison of two consecutive generation forecasts before and after the seventh iteration using an error diagram. This iteration exhibits the largest deviation from the diagonal. (c) Evaluation of the initial (thin line) and final (thick line) generation forecasts. In these diagrams, the dashed diagonal line corresponds to an unskilled forecast. See Figure 10.4 for explanation of shaded area.

Figure 10.10 Estimation of the differential probability gains for four seismicity precursory patterns. (a) The “Gamma” pattern relative to the RI model; (b) the “Activity” pattern relative the intermediate model,

g

× RI, obtained at the first step; (c) the “Acceleration” pattern relative to the RI model; (d) the “b‐macro” pattern relative to the RI model. Panels at the top show error diagrams for the patterns relative the reference models (RI or

g

× RI). Panels at the bottom show the differential probability gain values as a function of the corresponding alarm function. Calculations were made for the CSEP California testing region using the ANSS earthquake catalog for 1960–1984.

Figure 10.11 Error diagrams of the composite earthquake expectation rates model based on four intermediate‐term precursors. Error diagrams for the composite model relative to the RI model: (a) learning period 1960–1984; (b) testing period 1985–2016. The shaded area indicates the zone in which the difference between forecast of the forecasts of the tested model and underperforms the forecast of the reference model at a level of significance

α

 = 1% are not statistically significant; successful forecasts within this zone should be considered as random.

Figure 10.12 Composite model and RI forecasts for the interval 1 April 2010 to 30 June 2010. Red indicates highest rate cells in which 50% of target earthquakes during the 3‐month forecasting interval can be expected. For (a) the composite model 50% of target earthquakes are expected in 35 cells (out of 7682), for (b) RI 50% of target earthquakes are expected in 624 cells. Solid line shows the SCEP California testing region. Circles represent epicenters of target earthquakes with magnitude

M

 ≥ 5 that occurred during the forecasting interval. The largest circle corresponds to the El Mayor‐Cucapah

M

7.3 earthquake of 4 April 2010.

Chapter 11

Figure 11.1 Tectonic settings and the earthquakes that have occurred around Taiwan from 1 September to 31 December 1999. (a) Simplified tectonic features around Taiwan. (b) Distributions of the epicenters of

M

L

3.0 earthquakes recorded around Taiwan: red stars indicate the epicenters of the main shocks discussed that occurred during this period; shaded circles indicate earthquakes that may have shaken the monitoring stations; earthquake data are taken from the website of Central Weather Bureau, R.O.C. (http://www.cwb.gov.tw); CL, Chung‐lun; KZL, Kuan‐tze‐ling; SHTY, Suei‐ho‐tong‐yuan. (c) Geological profile along the A‐B cross‐section shown in Figure 11.1b.

Figure 11.2 Variation of (a) SO

4

2

−

and (b) NO

3

−

of the Puli groundwater before and after the 1999 Chi‐Chi earthquake. (c) Precipitation is the daily amount obtained in the Puli area [

Song et al

., 2003].

Figure 11.3 Variation of (a) Cl

−

and (b) SO

4

2

−

of the Kuan‐tze‐ling hot spring before and after the 1999 Chi‐Chi earthquake. The vertical lines indicate the major earthquakes that occurred during this period [

Song et al

., 2006].

Figure 11.4 The three components, mantle CO

2

, crustal CO

2

. and crustal CH

4

can be identified in the studied area. Clearly, CO

2

originated from two different sources, namely the mantle and crust. However, the CH

4

from natural gases in southwestern Taiwan mainly originates in the crust. Hence, all gas compositions are clearly explained by mixing the different components. CL1, Chung‐lun hot spring; CL3, Chung‐lun mud pool; KZL, Kuan‐tze‐ling hot; SHTY, Suei‐ho‐tong‐yuan fire and water spring.

Figure 11.5 A three‐component plot of helium isotopes of the gas compositions in southwestern Taiwan: ASW, air saturated water; C, crust; M, mantle. The gas compositions of the Chung‐lun hot spring show the most diverse variations. At least three components are necessary to account for the gas compositions. It is worth noting that the gas composition shows significant variations before the major earthquakes, which occurred on 21 September (

M

7.3), 22 October (

M

6.4), and 15 November (

M

5.1).

Figure 11.6 Variations of CO

2

/CH

4

ratios of bubbling gases from the Chung‐lun mud‐pool and associated related earthquake events. Twenty‐three anomalous peaks (marked with letters) can be recognized [

Yang et al

., 2006].

Figure 11.7 Distribution of soil‐gas anomaly sites in the study area for (a) He, (b) N

2

, and (c) Rn. Cross symbols indicate the sampling sites for the soil‐gas survey in this study; bold cross symbols indicate the sites that exhibit anomalous gas concentration; dashed lines indicate the fault distribution; star indicates the location of the station that was chosen based on high concentrations of both major and trace gases [

Fu et al

., 2008].

Figure 11.8 Variation of the soil‐gas data, signal trend and rolling average of soil Rn concentration, threshold values, seismic events, and rainfall at Chihshang station (CS) from 1 December 2012 to 1 December 2013. The horizontal dashed line indicates the threshold value (+1.5σ); black arrows indicate the highest anomalous points; grey arrows indicate the

M

L

 ≥ 5.5 seismicity events.

Figure 11.9 Seismicity distributions and Group I earthquakes with epicenter distances of ≤ 30 km; Group II earthquakes with epicenter distances between 60 and 120 km; and Group III earthquakes with epicenter distances ≥ 120 km. The dashed circle shows an observed precursory radon anomaly that indicates with high probability an impending earthquake is likely to be located in the area. It could be considered a sensitive zone for probable earthquakes of the CL station.

Figure 11.10 Variations in concentrations of soil gases at the Chung‐lun station from 25 October to 17 November 2002, showing a precursory anomaly before the local earthquake on 8 November 2002. Each colored line represents a different soil‐gas component measured.

Figure 11.11 Variations of Rn, CO

2

, and rainfall at the Hsinhua monitoring station and their association with twin earthquakes on 26 December 2006. The vertical yellow bands indicate the increasing trend in gas concentrations during the observation period [

Walia et al

., 2009].

Figure 11.12 (a) The distribution of seismicity and stress transfer direction from north to south along the Longitudinal Valley system as shown by different arrow colors and rectangles. The focal mechanism (the beach ball) represents the Fanglin earthquake occurring on 21 May 2014. (b) The temporal distribution of earthquakes and gamma‐ray variations from 10 March to 27 May 2014 at Donghua station in eastern Taiwan. The red arrow and lines indicate the duration of gamma‐ray anomalies. The black arrows indicate the earthquake swarms. The accumulated number of earthquakes at different distances from the hypocenter is shown as a histogram [

Fu et al

., 2015].

Figure 11.13 A simplified tectonic map of Taiwan and distribution of earthquake detectable zones at each soil‐gas station shown by different circles. The triangles represent the location of the soil‐gas stations: TPT, Tapingti; GK, Gukeng; CL, Chung‐lun; PT, Pingtung; CS, Chihshang. The focal mechanism (the beach ball) represents the Jiasian earthquake occurring on 4 March 2010.

Figure 11.14 Variation of the soil‐gas data, seismic events, and rainfall at Tapingti station from 1 February to 20 March 2010. Horizontal dashed lines indicate the annual average of soil gas and threshold value (+1.5σ). Rainfall and earthquake occurrence and magnitude data are also plotted for comparison. The triangles, stars, and circles represent different distance zones from the earthquake hypocenters.

Figure 11.15 Variation of the soil gas data, seismic events, and rainfall at Gukeng station from 1 February to 20 March 2010. Horizontal dashed lines indicate the annual average of soil gas and threshold value (+1.5σ). Rainfall and earthquake occurrence and magnitude data are also plotted for comparison. The triangles, stars, and circles represent different distance zones from the earthquake hypocenters.

Figure 11.16 Variation of the soil gas data, seismic events, and rainfall at Chung‐lun station from 1 February to 20 March 2010. Horizontal dashed lines indicate the annual average of soil gas and threshold value (+1.5σ). Rainfall and earthquake data are also plotted for comparison. The triangles, stars, and circles represent different distance zones from the earthquake hypocenters.

Figure 11.17 Variation of the soil gas data, seismic events, and rainfall at Chihshang station from 1 February to 20 March 2010. Horizontal dashed lines indicate the annual average of soil gas and threshold value (+1.5σ). Rainfall and earthquake data are also plotted for comparison. The triangles, stars, and circles represent different distance zones from the earthquake hypocenters.

Figure 11.18 Variation of the soil gas data, seismic events, and rainfall at Pingtung station from 1 February to 20 March 2010. Horizontal dashed lines indicate the annual average of soil gas and threshold value (+1.5σ). Rainfall and earthquake data are also plotted for comparison. The triangles, stars, and circles represent different distance zones from the earthquake hypocenters.

Chapter 12

Figure 12.1 Oil‐flow‐rate variations in a 10‐years time series recorded in the Sinai area in a well 3 km deep before two

Mw

 > 5.5 earthquakes occurred within a radius of 150 km. Crustal deformation preceding shocks has induced an increase in spontaneous oil flow from a deep hydrocarbon well. Similar phenomena may affect groundwaters and thermal waters. Confined aquifers may act as natural strainmeters.

Figure 12.2 Helium/argon concentration in a thermal spring in Japan following crustal deformation induced by earth tides. Crustal deformation may induce variations in deep fluid flow rate or emissions. This phenomenon is utilized in monitoring activities in earthquake precursors research.

Chapter 13

Figure 13.1 The ULF geomagnetic observation network in the Kanto–Tokai region of Japan. The radius of the circles is 80 km.

Figure 13.2 A typical spectrum of geomagnetic field (

Z

component) observed at the KYS station (Boso) on 3 February 2001.

Figure 13.3 Statistical results of ULF magnetic anomalies by superposed epoch analysis (SEA) at the Izu (SKS) and Boso (KYS) stations during 2000–2010.

Figure 13.4 Spatial distributions of major earthquakes with

E

s

 > 10

8

around the KAK station during 2001–2010. The blue triangle indicates the location of the KAK station. The red and black open circles indicate earthquakes in regions A and B, respectively.

Figure 13.5 Statistical results of ULF magnetic anomalies by superposed epoch analysis (SEA) at the KAK station during 2001–2010. The red and the black lines demonstrate the results of 5‐day counts for regions A and B, respectively. The blue and green lines with cross markers show random_mean and random_mean + 2

σ

, respectively.

Figure 13.6 A schematic illustration of the alarm implementation scheme. Figures from the top to the bottom (a–c) show a series of anomaly, alarm, and earthquake events, respectively. The units of the horizontal axis are days.

Figure 13.7 Molchan’s error diagrams for predictions based on the magnetic anomalies observed at the SKS (a) and KYS (b) stations.

Δ

is set to 1 day and

L

to 5 days;

R

 = 100 km and

E

s

 = 10

8

for the SKS station, and

R

 = 150 km and

E

s

 = 10

8

for the KYS station. The arrows indicate the locations of the maximum

PG

and

D

values, respectively.

Figure 13.8 Molchan’s error diagram for predictions based on the magnetic anomalies observed at the KAK station.

Δ

is set to 11 days and

L

to 5 days;

R

 = 100 km and

E

s

 = 10

8

.

Figure 13.9 (a) Area skill scores for predictions with different

Δ

and

L

and (b) Molchan’s error diagram of the optimal prediction strategy (

Δ

 = 8 days and

L

 = 1 day). The arrows indicate the locations of the maximum

PG

and

D

values, respectively.

Chapter 14

Figure 14.1 Examples of SSTAs identified in the period 09–23 July 2008. Significant Thermal Anomalies (STAs) with ⊗

∆T

(

r

,

t

’) ≥ 4 (depicted in different colors according with corresponding RETIRA values) appear in the Peloponnesus area before and after

M

 ≥ 4 seismic events. In addition to the magnitude of earthquakes, the temporal gap from the first appearance of STAs is also indicated by a number (

N

) in parentheses (+/−

N

means that the earthquake occurred

N

days after/before the first appearance of STAs). The red contoured box indicates the limits of analyzed SEVIRI TIR scenes (thermal anomaly map area).

Chapter 15

Figure 15.1 Earth atmospheric energies.

Figure 15.2 Application of the robust satellite technique to seismically active areas shows space–time correlation between thermal anomalies and earthquakes occurrence. In the case of the

Mw

6.3 L’Aquila earthquake (6 April 2009, (

Lisi et al.

[2010] and

Pergola et al.

[2010] for NOAA/AVHRR and EOS/MODIS) and geostationary (MSG/SEVIRI;

Genzano et al.

[2009]) sensors.

Figure 15.3 Thermal radiation anomalies (TRA) associated with

M

7.3 Van, Turkey, eartquake of 23 October 2011. (a) US Geological Survey shake map. (b) Nighttime TRA maps observed on 19 October 2011, 4 days in advance. Epicenter is marked with red star, tectonic plate boundaries by heavy red line, and major faults by feint brown lines. (c) Yearly time series of nighttime OLR over the epicentral area: (red) anomalous values; (gray) 2011 mean value OLR; (black) 2006–2011 mean value; (blue) 2010 anomalous trend with no major seismicity for comparison. (d)

M

 > 4 seismicity (European‐Mediterranean Seismological Centre catalog) near the epicentral area, 2011.

Figure 15.4 Thermal radiation anomalies (TRA) associated with

M

6.9 Aegean Sea, Greece, earthquake of 24 May, 2014. (a) US Geological Survey shake map. (b) Nighttime TRA map observed on 14 May 2015, 10 days in advance. Epicenter is marked with red star, tectonic plate boundaries by heavy red line, and major faults by feint brown lines. (c) Yearly time series of nighttime OLR over the epicentral area: (red) anomalous values; (gray) 2014 mean value OLR; (black) 2006–2014 mean value; (blue)2013 anomalous trend with no major seismicity for comparison. (d)

M

 > 4 seismicity (European‐Mediterranean Seismological Centre catalog) near the epicentral area, 2014.

Figure 15.5 Thermal radiation anomalies (TRA) associated with the

M

6 Napa Valley, California, earthquake of 24 August 2014. (a) US Geological Survey shake map. (b) Nighttime TRA anomaly map observed on 22 August 2014; 2 days before. Epicenter is marked with red star, tectonic plate boundaries by heavy red line, and major faults by feint brown lines. (c) Yearly time series of nighttime OLR over the epicentral area: (red) anomalous values; (gray) 2014 mean value OLR; (black) 2006–2014 mean value; (blue) 2013 anomalous trend with no major seismicity for comparison; and (d)

M

 > 4 seismicity (European‐Mediterranean Seismological Centre catalog) near the epicentral area, 2014.

Figure 15.6 Anomalous daily maps of surface latent heat flux. (a) 15 December 21004, 11 days before

M

9.3 Sumatra earthquake of 26 December 2004. (b) 6 March 2005, 20 days before

M

8.7 Sumatra earthquake of 28 March 2005. (c) 5 March 2011, 6 days before

M

9 Tohoku, Japan, earthquake of 3 November 2011. Epicenters are marked with red star, tectonic plate boundaries by heavy red line, and major faults by feint brown lines.

Figure 15.7 Principle diagram for generation of pre‐earthquake thermal radiation anomalies according to the lithosphere–atmosphere–ionosphere coupling concept [

Pulinets and Ouzounov

, 2011, for more details see Chapter 6].

Chapter 16

Figure 16.1 Positions of the VLF receiver in Petropavlovsk‐Kamchatsky (PTK) and the radio transmitters JJY, JJI, NWC (Australia), and NPM (Hawaii).

Figure 16.2 Results of statistical analysis for three receiving stations (Kochi, Moshiri, and Petropavlovsk‐Kamchatsky).

N

is the number of days with the chosen parameters for

D

st

, electron and proton fluxes;

N

i

is the number of days in the same interval with the data exceeding the corresponding 2

σ

criterion.

Figure 16.3 The anomalies in the NWC signal recorded at the YSH and YUK stations during the passage of several tropical cyclones in August 2012. In the bottom panel horizontal grey bars on the abscissa show the periods when the tropical cyclones crossed the sensitivity zones of the paths under consideration, and blue vertical bars refer to times of earthquake occurrence. The epicenters of earthquakes with

M

 > 6 that took place in the Pacific region during the period of analysis are shown in the top panel by large solid brown circles.

Figure 16.4 The relative locations of two Japanese VLF–LF transmitters (with call signs of JJY (Fukushima) indicated by a blue diamond and JJI (Miyazaki) (not shown) and VLF–LF receiving stations (Moshiri (MSR), Chofu (CHF), Kasugai (KSG) and Kochi (KCH) shown with black dots). The wave‐sensitive area defined by the Fresnel zone for the propagation path of JJY–MSR and also that for the propagation path of NLK (Jim Creek, USA) to CHF are plotted (thin black ellipses). Further, the great‐circle paths (thin red lines) and the corresponding wave‐sensitive areas (thin black lines) are indicated for the paths NLK–KSG and NLK–KCH. The epicenters of the main shock is indicated by the red star.

Figure 16.5 Locations of the two Japanese VLF–LF transmitters (JJY and JJI, red triangles) and two observing stations (Petropavlovsk‐Kamchatsky (PTK) and Yuzhno‐ Sakhalinsk (YSH), green dots). The wave‐sensitive areas (elliptic zones) for the propagation paths of JJY–YSH, JJY–PTK, JJI–YSH, and JJI–PTK are plotted. Further, earthquake main shocks and aftershocks are plotted, with sizes being proportional to magnitude.

Figure 16.6 Temporal evolution of the propagation characteristics for the three propagation paths: (a) NLK–CHF, (b) NLK–KSG, and (c) NLK–KCH. In each figure, the top panel refers to the average nighttime amplitude (trend), and the bottom panel to the dispersion. All of these values are normalized by their corresponding standard deviations (

σ

). A clear anomaly is seen on 5 and 6 March. The distance (

d

) of the earthquake epicenter to each propagation path is indicated on the top left in each figure.

Figure 16.7 Temporal evolution of the propagation characteristics for the wavepath of JJI–PTK. The top panel refers to the average nighttime amplitude (trend; horizontal broken line indicates −2

σ

level); middle panel plots the dispersion (horizontal broken line, +2

σ

level). Both trend and dispersion are normalized by their standard deviations (

σ

). The bottom panel indicates the temporal evolution of the seismic activity.

Figure 16.8 Superimposed epoch analysis for (a) the normalized trend (trend*), and (b) the normalized dispersion (

D

) (dispersion*). The red line refers to shallow earthquakes (depth < 40 km) and the blue line refers to earthquakes with depth > 40 km. The abscissa indicates the day with respect to the earthquake day (0), with minus indicating days before plus days after.

Figure 16.9 Schematic illustration of lithosphere–atmosphere–ionosphere coupling and three channels: chemical (+electric field), acoustic, and electromagnetic. (After

Hayakawa et al

. [2004] and

Hayakawa

[2009b, 2011].)