Narrow Results

Neural networks are investigated for predicting the magnitude of the largest seismic event in the following month based on the analysis of eight mathematically computed parameters known as seismicity indicators. The indicators are selected based on the Gutenberg-Richter and characteristic earthquake magnitude distribution and also on the conclusions drawn by recent earthquake prediction studies. Since there is no known established mathematical or even empirical relationship between these indicators and the location and magnitude of a succeeding earthquake in a particular time window, the problem is modeled using three different neural networks: a feed-forward Levenberg-Marquardt backpropagation (LMBP) neural network, a recurrent neural network, and a radial basis function (RBF) neural network. Prediction accuracies of the models are evaluated using four different statistical measures: the probability of detection, the false alarm ratio, the frequency bias, and the true skill score or R score. The models are trained and tested using data for two seismically different regions: Southern California and the San Francisco bay region. Overall the recurrent neural network model yields the best prediction accuracies compared with LMBP and RBF networks. While at the present earthquake prediction cannot be made with a high degree of certainty this research provides a scientific approach for evaluating the short-term seismic hazard potential of a region.

In this study, recurrence plot (RP) and recurrence quantification analysis (RQA) techniques are applied to a magnitude time series composed of seismic events occurred in California region. Using bootstrapping techniques, we give the statistical test of the RQA for detecting dynamical transitions. From our results, we find the different patterns of RPs for magnitude time series before and after the M6.1 Joshua Tree Earthquake. RQA measurements of determinism (DET) and laminarity (LAM) quantifying the order with confidence levels also show peculiar behaviors. It is found that DET and LAM values of the recurrence-based complexity measure significantly increase to a large value at the main shock, and then gradually recovers to a small values after it. The main shock and its aftershock sequences trigger a temporary growth in order and complexity of the deterministic structure in the RP of seismic activity. It implies that the onset of the strong earthquake event is reflected in a sharp and great simultaneous change in RQA measures.

In order to make certain of the nonlinear characteristics of the process of seismogeny, the characteristics of local scaling property of the interevent time series of the seismic sequences for Longmen mountain earthquake zone and Luquan area in China were studied by using the method of local scaling property wavelet analysis. It is indicated that there are clear characteristic variations of local scaling property prior to main shocks for the original series while there is no characteristic variation for the two randomly shuffled versions of the original series. The conclusions drawn are as follows: (1) there is a relationship between the characteristic variations of the local scaling property and the process of seismogeny; (2) this relationship reveals a mechanism of complex behavior in a nonlinear seismic system; and (3) the characteristic variations of the local scaling property can be taken as indicators of augmented or reduced stability of a seismic system.

Inspite of its location in one of the most active seismic zones in the world, Bhutan has no seismic design code of its own and no detailed study on the performance of buildings under expected earthquake ground excitation has been carried out. In this study, probabilistic seismic hazard analysis is first carried out to predict the design ground motions in Thimphu, Bhutan for the return periods (RPs) of 475 and 2475 years. These ground motions are then used to assess the performance of three typical RC buildings in the capital city, Thimphu. Soil–structure interaction (SSI) is incorporated at different soil sites and the effects of SSI are discussed. Adequacy of using Indian Seismic Code in Bhutan is also studied and discussed. The study suggests that the typical buildings in Bhutan could undergo moderate to severe damages under the 475 year RP and could even collapse under the 2475 year RP ground motions. This study is the first such effort in predicting the design ground motions and then assessing the performance of the general building stocks in Bhutan. The result can guide the seismic preparedness of the country through proper design and mitigation measures.

During the COVID-19 outbreak that took place in early 2020, restrictions on human activities were imposed by government-mandated lockdowns and stay-at-home orders. In this study, we analyse the impact of reduced anthropogenic activities on the detection of seismic events. We hypothesise and show that with reduced background noise levels due to the COVID-19 lockdowns, low-magnitude earthquakes are more easily detectable. We investigate the magnitudes of earthquakes recorded at the seismometers before and after COVID-19 lockdowns for two regions — Cascadia Subduction Zone and New Zealand. Gutenberg–Richter law, which gives a relationship between the number of earthquakes and their magnitudes (b-value), was applied in these two areas. Our results point to an increase in detection of smaller-magnitude earthquakes, as observed by an increased b-value during the COVID-19 period compared to those obtained in the pre-COVID time periods. Previous studies have shown that changes in b-value of an area over a sustained period of time affect the short-time probabilistic risk assessment. The variability of b-value also gives useful insights into the prevailing stress state of the region, crustal heterogeneity, pore pressure and tectonic setting of the area.

We reappraise the material needed to assess the 20th-century seismicity of Switzerland and of the adjacent areas of the Alps in terms of magnitude. For this we make use of macroseismic reports and literature and by calculating the surface-wave magnitude from the Prague formula of all significant earthquakes in the region. No attempt is made to relocate earthquake positions; instead their reliability is ranked using existing solutions and macroseismic observations. We find that for small earthquakes (M_S<4.5), which constitute the bulk of the events in the region, the calculation of M_S observed at relatively short distances, requires station and distance corrections which can be significant. Also we find that recomputed Ms estimates differ from those reported in other earthquake catalogues. From this reappraisal of M_S and from a uniform re-evaluation of the associated macroseismic data, we derive a stable correlations between M_S and felt areas that can be used to assess the magnitude of historical, pre-instrumental, events for which only isoseismal radii (r_i) and the associated intensities (I_i) are available. We examined the conversion of surface-wave magnitude into moment magnitude, a conversion that presents some interesting problems for relatively small events for which the M_S-log(M₀) scaling changes. We conclude that the rate of moment release derived from events of M_S>4.0 is small, and that it should be associated with horizontal and vertical velocity rates of less than 1 mm/yr, too small to be confirmed by GPS measurements over short periods of time so that can be used to constrain hazard assessment.

We use the record of earthquakes in central Greece since 1694 to estimate the seismically-released strain in that region over the last 300 years. The highest strain rate is seen in the Gulf of Corinth, an extensional graben which is the most prominent active geological structure. Over the 300-year period, the earthquakes account for an average extension rate of about 11 mm/yr between the Peloponnese and the island of Evvia, of which about 10 mm/yr occurs in the Gulf of Corinth. The earthquakes occurring in shorter time periods, perhaps as little as 100 years, account for extensional velocities similar to the 300 year average, though the 20th Century has been relatively quiescent. We are not confident that the historical record can be used to assess seismic strain release in areas smaller than the Gulf of Corinth as a whole. The similarity between the average rate of seismic strain release over 300 years and the short-term strain rates measured geodetically suggests that the bulk of the tectonic strain in this region is released by seismic slip on faults, and that aseismic creep is relatively unimportant. An apparent deficit in seismic strain release over a short period (as is observed for the 20th Century) is thus likely to be remedied by future earthquakes.

The potential areas of upcoming earthquakes were investigated along the Northern segment of the Sumatra–Andaman Subduction Zone according to the b-value of the frequency-magnitude distribution. After enhancing the completeness of the earthquake catalogue, two datasets, those recorded during (i) 1980–1994 and (ii) 1980–2003, were tested in order to verify the effective correlation between precursory b-values and the location of subsequent earthquakes. The results confirmed that areas with low b-values agreed well with the locations of the subsequent earthquakes in that region. Accordingly, the present-day dataset from 1980–2010 was carefully evaluated to determine the b-values across the region. Within this spatial investigation, three areas of low b-values and so potential hazards were found. These consisted of the (i) West coast of Myanmar, and (ii) North and (iii) South of the Nicobar Islands. From 2010–2012, a major earthquake with magnitude 7.5 m_b was recorded as being generated in the region South of the Nicobar Islands. Thus, attention should be paid to the remaining two until now quiescent areas, and mitigation plans should be raised for both seismic and tsunami hazards.

In this study, the geospatial frequency–magnitude distribution (FMD) b-value images of the prospect sources of upcoming earthquakes were investigated along the Indonesian Sunda Margin (ISM) that strikes parallel to and near the Indonesian Island chain. After enhancing the completeness and stability of the earthquake catalogue, the seismicity data were separated according to their seismotectonic setting into shallow crustal and Intraslab earthquakes. In order to verify the spatial relationship between the b-values and the occurrence of subsequent major earthquakes, the complete shallow crustal seismicity dataset (1980–2005) was truncated into the 1980–2000 sub-dataset. Utilizing the suitable assumption of fixed-number of earthquakes, retrospective tests of both the complete and truncated datasets supported that areas of comparatively low b-values could reasonably be expected to predict likely hypocenters of future earthquakes. As a result, the present-day distributions of b-values derived from the complete (1980–2005) shallow crustal and Intraslab seismicity datasets revealed eight and six earthquake-prone areas, respectively, along the ISM. Since most of these high risk areas proposed here are quite close to the major cities of Indonesia, attention should be paid and mitigation plans should be developed for both seismic and tsunami hazards.

In this study, I investigated qualitatively the earthquake catalogue of the Thai Meteorological Department (TMD), Thailand, with respect to the seismicity patterns of Thailand. The readymade relationships between the different magnitude scales were derived to allow their convenient interconversion. Earthquake declustering was performed in order to screen the main shocks from the foreshocks and aftershocks, reducing the 1998–2009 records from ~48,900 to 2,620 main events. Man-made changes in the seismicity rate were carefully checked for, but only some minor changes were found and these were not related to any network improvements. In order to assess the limit of the earthquake detection in the catalogue the criterion of the magnitude of completeness (Mc) was employed, revealing a high efficiency of earthquake detection at a low Mc (3.0–3.5 M_w), especially for the inland active fault zone that dominates in Southeast Asia. Thus, the TMD's catalogue is one of the alternative catalogues for seismicity investigation of inland earthquakes. Meanwhile for the area surrounding the Sumatra Island and Northern Myanmar, the TMD's network is sufficient only for earthquakes with a M_w > 5.4–6.0 M_w. Thus, some additional seismic recording stations are needed in the Southern and Northern parts of Thailand.

In order to determine the prospective areas of the forthcoming earthquake sources, the $b$-values of the frequency-magnitude earthquake distributions were analyzed spatially and mapped along the strike-slip fault system at the Thailand–Myanmar border. In order to constrain the relationship between the variation of $b$ and the following hazardous earthquake, the completeness of earthquake catalogue was manipulated into two datasets for (i) 1980–2000 and (ii) 1980–2005 and the $b$-values mapped. Utilizing the suitable assumption of 30 fixed earthquake events, the following $M_w\ge 5.0$ earthquakes illustrate a significant relation between their epicenter and the areas showing relatively low $b$-values. By utilizing the most recent earthquake data (1980–2015), five areas exhibiting low $b$-values (implying prospective earthquake sources) can be identified along the strike-slip fault system. Compared with earthquake activities evaluated previously along the strike-slip fault system, the data reveal that these five areas may potentially generate earthquakes up to 7.0$M_w$ within the coming 50 years; the recurrence of the $M_w$-5.0 earthquake is about 10 years and the probabilities of the $M_w$-5.0 earthquake are about 40–95%, respectively. Since these prospective hazardous seismic zones are located close to cities, population centers and hydropower dams, an effective mitigation plan should be developed.

Hilbert–Huang Transform (HHT) is a new analysis method for nonstationary and nonlinear signals. A simulation method based on HHT is used to generate uniform hazard ground motions, which are often needed for nonlinear time history analysis for structures in high-seismic zones. The HHT-based simulation method can reproduce the amplitude and frequency content change with time for nonstationary random processes, thus is very suitable for the simulation of earthquake ground motions, especially the near-fault ground motions with long-period pulses. Monte-Carlo method and historical earthquake records are used for the generation of a large pool of ground motions, from which the uniform hazard ground motions are selected. The regional seismicity and rupture directivity are considered. An example is given of a site near Los Angeles City Hall. The advantages and difficulties of the proposed method are also discussed.