Earthquake Prediction Research Articles

Fluid-injection control on energy partitioning during the earthquake cycle

During an earthquake, the elastic energy stored in the Earth is released as frictional energy and radiated energy in the form of seismic waves. The partitioning of energy released during an earthquake gives an indication of the overall size of the earthquake and its potential for damage to man-made structures. Here, we perform an energy analysis of fluid-injection-induced earthquakes using a single-degree-of-freedom spring poroslider and rate-and-state friction. We show that seismic radiation can be modeled within the single-degree-of-freedom spring-slider by adding a radiation damping term. We then use it to study fluid injection and assess its effects on the energy partitioning during induced and triggered earthquakes. We find that: (1) seismic efficiency, stress drop, and total slip are directly influenced by the rate of increase in pore pressure, (2) the ratio of elastic energy stored in the skeleton to injection energy is low and is influenced by the rate of fluid injection, (3) seismic injection efficiency is also low and is lower for induced earthquakes compared to triggered ones, and (4) fluid injection leads to bigger and potentially more damaging earthquakes overall.

Spatio-Temporal Variation Seismicity Pattern in East Java Between 2002 and 2022 Based on the b-value and Seismic Quiescence z-value

East Java is one of the seismically active regions in part of the Sunda Arc whose seismicity pattern must be studied. Therefore, a spatio-temporal analysis of seismicity between 2002 - 2022 in East Java has been conducted. This analysis aims to investigate the seismotectonic conditions in the East Java region over the last twenty years, especially the seismicity patterns associated with significant earthquakes in East Java. The analysis was carried out by quantifying the spatial and temporal variations of seismicity parameters using b-value and z-value. The used data used are earthquake data from the USGS earthquake catalog during 2002 - 2022 in the East Java region. The analysis was performed using the ZMAP 6.0 program code using the maximum likelihood method. The results of the temporal analysis of b-values show that almost all significant earthquakes in East Java were characterized by a decrease in b-values of up to 0.2. Spatially, the significant earthquake on 10 April 2021 was associated with a low b-value (0.9). Temporal z-value analysis shows that the significant East Java earthquake was preceded by a quiet period (positive z-value) between 4 - 5 years (2003 - 2008 and 2012 - 2016). Spatial analysis of the z-value in the last 7 shows an increase in the z-value before the significant earthquakes occurred on 16 July 2019, 18 March 2020, and 10 April 2021. Therefore, the results of this study indicate that the b-value and z-value potentially can be used as precursors of the significant East Java earthquake. HIGHLIGHTS This is the first time a correlation of b-value and z-value as seismicity parameters has been carried out in East Java, Indonesia, as an attempt to determine the significant earthquake precursors The significant earthquakes in East Java during 2002 - 2022 were marked by a decrease in b-value of up to 0.2 The quiet period in East Java lasted between 4 - 7 years, i.e. 2002 - 2009 and 2012 - 2016, characterized by a positive z-value peaked until +3.5 The z-value at the location of a significant earthquake epicenters show a positive value 2 - 3 years prior to the main quake GRAPHICAL ABSTRACT

CORRELATION OF RADON AND SEISMIC ACTIVITY IN THE BAIKAL RIFT ZONE ACCORDING TO EMANATION MONITORING DATA

Long series of the emanation monitoring organized at Tyrgan (2018–2023), Buguldeika (2020–2023) and Maksimikha (2022–2023) stations in the central part of the Baikal Region (the Eastern Siberia) are studied to reveal the relationship of the subsurface radon volumetric activity (RVA) with the seismicity of the Baikal rift zone (BRZ). The comparison of the weekly averaged RVA values at different stations and the degree of dependence of this parameter on the logarithm of the total earthquakes energy (LgΣE) showed that during the periods of seismic activity the impact of tectonic forces makes significant corrections in the character of emanations, which in general is determined by non-tectonic factors. It has been established that seismic activity represented by the enough strong for the Baikal Region earthquakes with K≥13.3 correspond to synchronicity intervals of RVA at different monitoring stations, as well as coincidence of maximum values of the logarithm of the total earthquakes energy with local maximums of the correlation coefficient between RVA and LgΣE parameters. The area, including the earthquake epicenters that generated such an emanation response at monitoring stations, is in the form of an ellipse extended in the northeastern direction – along the main fault structures of the BRZ. The obtained results, besides confirming the previously proposed model of radon field formation in the Baikal Region under the influence of external and internal forces, provide the for the identification of further stable precursors of strong earthquakes based on a comprehensive analysis of data from a branched emanation monitoring network.

Study of the seismic activity of the Almalyk-Angren industrial zone based on lineament analysis

In this work, an automated lineament analysis was carried out to search for earthquake precursors for the territory of the Almalyk-Angren industrial zone in Uzbekistan. The seven events with a magnitude of about 3 were selected for analysis. The Landsat 8 satellite images were processed using the automated lineament detection method in the LEFA software. The processing steps included detecting line elements in raster images, calculating the characteristics of the spatial distribution of line elements, and combining collinear linear elements into lineaments. The analyses of the cyclicality of precursors before and after earthquakes were based on the study of the distribution of the lineament trend in the study area using rose diagrams and lineament density maps. The results showed a change in the dynamics of the lineament structure. The statistics of the number of lineaments showed that their increase begins almost 20 days before the event, reaches its maximum about 1 – 2 days before the earthquake, decreases starting from 14 days after the earthquake, and has a minimum value of 1 – 2 months. The main trends observed in the lineament map showed the dominant trend in NS, WE, NW-SE directions.

Monitoring the Workflow of Earthquake by Using Machine Learning

This study explores the innovative application of machine learning techniques in monitoring and analyzing the workflow associated with earthquakes. The primary objective is to enhance the accuracy and efficiency of earthquake detection, prediction, and response mechanisms the development of machine learning technology enables more robust real-time earthquake monitoring through automated implementations. However, the application of machine learning to earthquake location problems faces challenges in regions with limited available training data. This paper proposes a machine learning-based approach to monitor seismic workflows efficiently. We apply various Machine Learning algorithms, including deep learning neural networks and decision trees, to process and interpret vast amounts of seismic data. These algorithms were trained on historical earthquake records to identify patterns and anomalies that precede seismic events. Our approach not only focused on the detection of seismic activities but also on the prediction of potential aftershock locations and magnitudes. Earthquake monitoring and response systems play a crucial role in mitigating potential damages and ensuring rapid response to seismic events. Traditional methods rely on sensor networks and manual analysis, often leading to delays in detection and response. The system leverages real-time data from seismic sensors, incorporating advanced machine learning algorithms for data analysis and pattern recognition. By employing deep learning models, such as convolution neural networks (CNNs) and recurrent neural networks (RNNs), the system can detect and classify seismic activities accurately. Additionally, anomaly detection techniques enable the identification of unusual seismic patterns that might indicate impending earthquakes or aftershocks. The system integrates geographical information systems (GIS) to map the seismic activities spatially, aiding in the visualization and understanding of seismic patterns across regions. The machine learning models are trained on historical seismic data to enhance their predictive capabilities and adaptability to varying seismic behaviors. This proposed system offers a real-time monitoring solution that can assist in early earthquake detection, accurate event classification, and timely response coordination. Its ability to continuously learn from incoming data ensures adaptability to changing seismic behaviors, thereby improving overall efficiency and reliability in earthquake monitoring workflows.

Multifractal detrended fluctuation analysis of soil radon in the Kachchh Region of Gujarat, India: A case study of earthquake precursors

In this paper, the scaling properties of earthquake-induced soil radon emissions in the Kachchh region of Gujarat, India, were demonstrated through the application of multifractal detrended fluctuation analysis (MFDFA). The analysis considered soil radon data recorded at four stations, namely Rampar, Rapar, Pragpar and Paddhar, over a period of one year (January 1, 2020 to December 31, 2020). At these stations, soil radon along with soil pressure and temperature are recorded continuously at 10-minute intervals. In the first step, periodic oscillations such as diurnal and semi-diurnal periodicities of the raw soil radon data were removed by applying empirical mode decomposition (EMD). The aperiodic soil radon at all measuring stations was used as input for the MFDFA analysis. In order to obtain the exact source of the multifractality, the MFDFA is also applied to the shuffled and surrogate data of the aperiodic soil radon time series at all measuring stations. It can be observed that the fluctuation function (Log (Fq(s))) for all measuring stations increases linearly with increasing scale value(s). In contrast, the generalized Hurst exponent (H(q)) of the shuffled radon values is about 0.5, the H(q) value of the surrogate and observed radon series is greater than 0.5 at all measuring stations. The scaling exponent τqof shuffled Radon shows an approximately linear trend, while surrogate and soil observed radon show a non-linear trend. At all stations, the shuffled series shows a narrower spectrum compared to the surrogate and observed soil radon time series. This shows that observed soil radon has a long-range correlation. Three local earthquakes of June 14, 2020 (M=5.3), August 28, 2020 (M=4.1) and November 1, 2020 (M=4.1) near monitoring stations are considered for analysis in this study. The multifractal spectrum of soil radon is also examined during the seismically quiet and disturbed period. A broader spectrum is obtained during the disturbed period compared to the quiet period. This may be due to heterogeneity characterized by anomalous emission of soil radon gas before the occurrence of earthquakes during the monitoring period.

Spring water anomalies before two consecutive earthquakes (Mw 7.7 and Mw 7.6) in Kahramanmaraş (Türkiye) on 6 February 2023

Abstract. Understanding earthquake phenomena is always challenging. The search for reliable precursors of earthquakes is important but requires systematic and long-time monitoring employing multi-disciplinary techniques. In search of possible precursors, we obtained commercially bottled spring waters dated before and after the earthquakes of 6 February 2023. Hydrogeochemical precursors have been detected in commercially bottled natural spring waters (Ayran Spring and Bahçepınar Spring), which are at a distance of about 100 and 175 km from the epicenters of the Mw 7.7 and Mw 7.6 Kahramanmaraş (Türkiye) earthquakes of 6 February 2023, respectively. The available water samples cover the period from March 2022 to March 2023. The pre-earthquake anomaly is characterized by an increase in electrical conductivity and major ions (Ca2+, Mg2+, K+, Na+, Cl−, and SO42-) compared to the background for Ayran Spring water samples. The pre-earthquake anomaly lasted for at least 6 months. The anomaly in major ions sharply declined and the ion content approached the background values about 2 weeks after the earthquakes. Although only 6.5 km away from the Ayran Spring, the bottled water samples of the Bahçepınar Spring did not show any anomalies in electrical conductivity; therefore, the samples were not analyzed for ion content. Bahçepınar water is collected from shallow boreholes dug into alluvial deposits, which we believe are decoupled from the basement rocks, and this may be the reason for the lack of abnormal water chemistry prior to the earthquakes. This attests to the fact that sampling locations are very important in the detection of possible earthquake precursors. Results of the Ayran spring water samples indicate that spring water chemical anomalies of discrete samples may provide valuable information on pre-earthquake crustal deformation. Monitoring of spring waters, along with other monitoring techniques in a multidisciplinary network, and for a sufficiently long time, could potentially enable obtaining reliable proxy indicators of pre-earthquake crustal deformation.

Evaluation of Return Period and Occurrence Probability of the Maximum Magnitude Earthquakes in Iraq and Surroundings

It has long been clear that earthquake prediction is important from both social and economic perspectives; therefore, the practical objective of today’s earthquake seismology researchers is an effective earthquake prediction program. The purpose of this study is to estimate earthquake probabilities and return periods using an updated earthquake catalogue (1900-2019) for Iraq and its surroundings. Weibull’s formula and inverse Weibull’s formula were employed to calculate the period of return and the occurrence probability of the maximum magnitude earthquake. The return period for earthquakes magnitudes 5 and 7Mw was 1.1 and 10.54 years, respectively, while the occurrence probability was 93.79% and 9.5%, respectively. The greatest magnitude is 7.7, with a 121-year return period and likelihood of approximately 0.82%. The probability of exceedance increased as the time period increased. The return duration was greater for earthquakes of higher magnitudes.

Are Foreshocks Fore‐Shocks?

AbstractForeshocks are spatially clustered seismic events preceding large earthquakes. Since the dawn of seismology, their occurrence has been identified as a possible mechanism leading to further crustal destabilization, hence, to major failures. However, several cases occurred without any previous anomalous seismic activity, so that the hypothesis of foreshocks as reliable seismic precursors fails to pass statistical tests. Here, we perform an all‐round statistical comparative analysis of seismicity in Southern California to assess whether any differences can be identified between swarms and foreshocks. Our results suggest that extremely variable seismic patterns can forerun mainshocks, even though they tend to be preceded by clusters with more numerous events spread over larger areas than swarms and with a wider range of magnitudes. We provide a physical explanation of such dissimilarity and conclude, despite it, that foreshocks can hardly be reliable short‐term precursors of large earthquakes in California.

A New Algorithm for Earthquake Prediction Using Machine Learning Methods

A New Algorithm for Earthquake Prediction Using Machine Learning Methods

Predicting Stick-Slips in Sheared Granular Fault Using Machine Learning Optimized Dense Fault Dynamics Data

Predicting earthquakes through reasonable methods can significantly reduce the damage caused by secondary disasters such as tsunamis. Recently, machine learning (ML) approaches have been employed to predict laboratory earthquakes using stick-slip dynamics data obtained from sheared granular fault experiments. Here, we adopt the combined finite-discrete element method (FDEM) to simulate a two-dimensional sheared granular fault system, from which abundant fault dynamics data (i.e., displacement and velocity) during stick-slip cycles are collected at 2203 “sensor” points densely placed along and inside the gouge. We use the simulated data to train LightGBM (Light Gradient Boosting Machine) models and predict the gouge-plate friction coefficient (an indicator of stick-slips and the friction state of the fault). To optimize the data, we build the importance ranking of input features and select those with top feature importance for prediction. We then use the optimized data and their statistics for training and finally reach a LightGBM model with an acceptable prediction accuracy (R2 = 0.94). The SHAP (SHapley Additive exPlanations) values of input features are also calculated to quantify their contributions to the prediction. We show that when sufficient fault dynamics data are available, LightGBM, together with the SHAP value approach, is capable of accurately predicting the friction state of laboratory faults and can also help pinpoint the most critical input features for laboratory earthquake prediction. This work may shed light on natural earthquake prediction and open new possibilities to explore useful earthquake precursors using artificial intelligence.

Investigating ultra-low frequency emissions and total electron content anomalies as earthquake precursors in Sumatra (2019–2020)

Ultra-low frequency (ULF) emissions and total electron content (TEC) anomalies are potential earthquake precursors for short-term earthquake prediction. This study comprehensively examines ULF emissions and TEC anomalies associated with the Sumatra earthquakes spanning 2019–2020. Precursor ULF emissions are identified through the polarization power ratio calculations involving Z-component and H-component geomagnetic data, meticulously recorded by the Magnetik Acquisition Data System (MAGDAS). Employing the Single Station Function Transfer technique, ULF anomalies directed toward the epicenter of earthquakes are recognized as reliable precursors. Among the 29 earthquakes scrutinized, 25 were heralded by ULF emissions acting as precursors. Among these, 15 instances solely exhibited ULF emissions, unaccompanied by TEC anomalies, while the remaining ten were linked to TEC anomalies. The investigation of TEC anomalies employs advanced autocorrelation techniques rooted in the analysis of correlation coefficient deviations. TEC anomalies are meticulously acquired from the observations of the Sumatran GPS Array (SuGAr) stations. Furthermore, this study illuminates a positive correlation (R = 0.65) between the lead time of ULF emissions and TEC anomalies. Moreover, the frequency of ULF emissions and TEC anomalies also showcases a direct relationship with earthquake magnitude. Heightened frequencies of ULF emissions and TEC anomalies as earthquake precursors correspond to more significant magnitudes in ensuing earthquakes. The outcomes of this research significantly augment the understanding of employing ULF emissions and TEC anomalies as effective earthquake precursors in the context of Sumatra.

Earthquake prediction from seismic indicators using tree-based ensemble learning

Earthquake prediction from seismic indicators using tree-based ensemble learning

Monitoring the workflow of Earthquake by using Machine Learning

The main goal is to improve the accuracy and efficiency of earthquake detection, prediction and response through automation and, over time, to improve machine learning techniques for earthquake monitoring. However, the application of machine learning to earthquake location problems faces challenges in areas with limited data. This article presents an engine-based approach to monitor earthquake performance. We use a variety of Machine Learning algorithms, including deep learning neural networks and decision trees, for processing and interpret large amounts of seismic data. These algorithms are trained on historical earthquake records to identify patterns and anomalies prior to seismic events. Our approach focuses not only on detecting seismic events, but also on predicting the location and magnitude of earthquakes. Earthquake monitoring and response systems play an important role in reducing potential losses and providing rapid response to seismic events. Conventional methods rely on sensor networks and manual analysis, often with delays in detection and response. The system receives real-time data from seismic sensors, incorporating advanced machine learning algorithms for data analysis and pattern recognition. The system can detect and describe seismic events using deep learning models such as convolutional neural networks (CNN) and random neural networks (RNN).In addition, anomaly detection techniques allow the identification of unusual seismic patterns that may indicate impending earthquakes or earthquakes. The system integrates geographic information systems (GIS) to comprehensively map seismic activity, helping to visualize and understand seismic patterns in the region. Machine learning models are trained on historical seismic data to improve their predictive capabilities and adapt to different seismic behaviors. This proposed system provides a real-time monitoring solution that can assist in early earthquake detection, accurate event classification, and timely response coordination. The ability to continuously learn from incoming data enables adaptation to change in seismic activity, thereby increasing the efficiency and reliability of earthquake monitoring operations.

CO Emissions Associated with Three Major Earthquakes Occurring in Diverse Tectonic Environments

Significant amounts of gases are emitted from the earth’s crust into the atmosphere before, during, and after major earthquakes. To understand the relationship between gas emissions, earthquakes, and tectonics, we conducted a thorough investigation using satellite data from AQUA AIRS. We focused on three major earthquakes: the 12 May 2008 Wenchuan MW 7.9 earthquake in China’s intra-continental plate, the 26 December 2004 Sumatra-Andaman MW 9.1 earthquake in Indonesia Island, and the 4 April 2010 Baja California MW 7.2 earthquake in Mexico’s active plate margin. Anomalies in the total column (TotCO) and multiple layers (CO VMR) of carbon monoxide were observed along fault zones, with peak values at the epicenter areas. Furthermore, temporal anomalies of TotCO and CO VMR appeared in the month of the Wenchuan earthquake in the intra-continent, three months prior to the Sumatra-Andaman earthquake and one month before the Baja California earthquake in the active plate margins, respectively. Notably, the duration of CO anomalies before earthquakes in active plate margins was longer than that in the intra-continental region, and the intensity of the CO anomaly in active plate margins was higher than that in the intra-continental region. The results show a profound correlation with both seismic and tectonic activities, which was particularly evident in the earthquake’s magnitude, rupture length, and the tectonic settings surrounding the epicenter. Furthermore, the type of the fault at which the earthquake occurred also played an important role in these CO anomaly variations. These findings support the identification of earthquake precursors and may help improve our understanding of earthquake forecasting and tectonics.

Unraveling the Seismic Signal Anomaly at Mount Rinjani Station: In-Depth Exploration with the Detrended Fluctuation (DFA) Analysis Method in Connection with the 2018 Big Earthquake on Lombok Island, West Nusa Tenggara

We analyzed the seismic signals of the Mount Rinjani station to identify possible seismic anomalies before a series of major earthquakes on Lombok Island in 2018. We observe anomalies before earthquake events Mw >= 5.90. This research applies the detrended fluctuation analysis (DFA) method to investigate possible earthquake precursors associated with the analyzed earthquake. The results showed a relationship between scaling seismic signals (α) that exceeds the threshold of α values during the analysis period, and there is a deviation of the root mean square fluctuations in the corresponding scaling that should be rising but constant. The value of α with constant root mean square fluctuations occurs one to three days before the preliminary earthquake Mw 6.40 (28/07/2018) UTC and eight to 10 days before the main earthquake Mw 6.90 (05/08/2018). The Mw 6.90 earthquake had the most magnitude and could significantly contribute to the appearance of seismic anomalies.

Analysis of VLF Signal Noise Changes in the Time Domain and Excitations/Attenuations of Short-Period Waves in the Frequency Domain as Potential Earthquake Precursors

In this paper, we complete pioneering research that indicates the very low frequency (VLF) signal amplitude and phase noise reductions, and short-period wave excitations and attenuations as new potential earthquake precursors. We consider changes in the VLF signal broadcast in Italy by the ICV transmitter and recorded in Serbia that start a few tens of minutes before earthquakes. The sampling interval of the analyzed data is 0.1 s. The main objectives of this study are (1) to complete this research in the time and frequency domains during the periods of the four earthquakes analyzed in the previous studies, and (2) to define the parameters of the VLF signal amplitude and phase in both domains that should be further examined in statistical analyses of the aforementioned potential earthquake precursors. In the first part of this study, we analyze the ICV signal amplitude in the frequency domain during the period around three earthquakes that occurred in November 2010 near the considered signal propagation path. Here, we apply the Fourier transform to the relevant recorded data. In the second part, we compare characteristics of the signal amplitude and phase noise reductions in the time domain, and wave excitations and attenuations in the frequency domain. The results of these comparisons indicate the parameters that should be analyzed in subsequent studies to confirm the connection of the considered VLF signal changes with seismic activity before earthquakes, and potentially establish procedures for their detection are: (a) the start and end times of the noise reductions in the time domain and the excited/attenuated waves in the frequency domain, (b) the differences in the corresponding times, and (c) the wave periods of wave excitations of both the signal amplitude and phase.

The shape of the Himalayan “Arc”: An ellipse pinned by syntaxial strike-slip fault tips

Trans-Himalayan geodetic data show that, between both syntaxes, India/Asia convergence is steadily oriented ≈ N20°E. However, surface faulting near both syntaxes, along the 2005 and 1950 earthquake ruptures, imply long-term thrusting directed ≈ 130° apart, and post-LGM (last Glacial Maximum) shortening rates of ≈ 5 to 6 mm/y, ≈ 2 to 3 times slower than in Nepal (≈ 15 to 20 mm/y). Syntaxial earthquakes' return-time are also ≈ 3 times longer (>2,000 y) than in Nepal (≈ 700 y). In a structural frame centered halfway between the syntaxial cusps, the tectonic features of the range show remarkable symmetry. In map view, the overall shapes of the Main Front Thrust (MFT) and the Main Central Thrust (MCT) closely fit ellipses, with major-to-minor axis ratios of ≈ 2.5 to 3. This suggests that the range growth atop subducting India is "pinned" by the strike-slip faults that bound it to the east and west. Discrete Element Modeling corroborates a late-Tertiary elliptical range growth. This accounts for the ≈ 65° angles and twofold to threefold decrease in active thrusting between Nepal and the syntaxes, for the maximum Himalayan heights (≥8,000 m), larger magnitudes (≥8), and shorter return-time (≈ 700 y) of great earthquakes in Nepal, for the existence of two 500- to 600 km-long, south-concave mountain ranges north of both syntaxes and for the ≈ 9 mm/y, N100 to 110°E extension across southern Tibet. It also suggests that predictions of impending or frequent great earthquakes in the eastern- and westernmost Himalayas may be overstated.

For the Classification of Anomalous Geophysical Fields that Existed Prior to an Earthquake

The consolidated paper presents work carried out in the sphere of earthquake problems. On the base of theoretical and experimental studies, it is shown earthquake prediction possibility. There are discussed earthquake indicators and triggering exogenous factors in the example of the Caucasus region. Because the earthquake preparation process causes anomalous changes with complex characteristics in various geophysical fields, it is given scientifically proven suggestions for the classification of these fields as earthquake precursors, indicators, and triggering factors. It offers a short-term plan for future work in the earthquake prediction direction.

Reevaluating Earthquake Fatalities in the Taiwan Region: Toward More Accurate Assessments

Abstract Earthquakes occurring in the Taiwan region have the potential to cause significant damage and loss of life. To ensure an effective emergence response and mitigate the impact of seismic events, it is imperative to develop an earthquake fatality prediction model that incorporates the Taiwan region’s seismic background, geological context, and socioeconomic level. This research article analyzes 35 historically damaging earthquakes in the Taiwan region, considering their fatality records, macroseismic intensities (modified Mercalli shaking intensity scale), and population exposure, to propose a specific seismic fatality ratio applicable to the Taiwan region. The model incorporates adjustments based on the gross national income per capita to address temporal and spatial disparities resulting from socioeconomic development. The predicted results highlight the superior performance of this fatality estimation model compared to other existing models. Furthermore, the earthquake fatality estimation model for the Taiwan region differs significantly from the models employed for South China and Qinghai–Tibet, suggesting variations in geological background, earthquake-resistant structures, and seismic resilience of society between the Taiwan region and mainland China. Moreover, a comparison with the U.S. Geological Survey model for the Taiwan region reveals that previous evaluations of the Taiwan region earthquakes greatly underestimated the number of fatalities.