Antonio Morales-Esteban

Neural networks to predict earthquakes in Chile

A new earthquake prediction system is presented in this work. This method, based on the application of artificial neural networks, has been used to predict earthquakes in Chile, one of the countries with larger seismic activity. The input values are related to the b-value, the Baths law, and the Omori-Utsus law, parameters that are strongly correlated with seismicity, as shown in solid previous works. Two kind of prediction are provided in this study: The probability that an earthquake of magnitude larger than a threshold value happens, and the probability that an earthquake of a limited magnitude interval might occur, both during the next five days in the areas analyzed. For the four Chiles seismic regions examined, with epicenters placed on meshes with dimensions varying from 0.5^ox0.5^o to 1^ox1^o, a prototype of neuronal network is presented. The prototypes predict an earthquake every time the probability of an earthquake of magnitude larger than a threshold is sufficiently high. The threshold values have been adjusted with the aim of obtaining as few false positives as possible. The accuracy of the method has been assessed in retrospective experiments by means of statistical tests and compared with well-known machine learning classifiers. The high success rate achieved supports the suitability of applying soft computing in this field and poses new challenges to be addressed.

Pattern recognition to forecast seismic time series

Earthquakes arrive without previous warning and can destroy a whole city in a few seconds, causing numerous deaths and economical losses. Nowadays, a great effort is being made to develop techniques that forecast these unpredictable natural disasters in order to take precautionary measures. In this paper, clustering techniques are used to obtain patterns which model the behavior of seismic temporal data and can help to predict medium-large earthquakes. First, earthquakes are classified into different groups and the optimal number of groups, a priori unknown, is determined. Then, patterns are discovered when medium-large earthquakes happen. Results from the Spanish seismic temporal data provided by the Spanish Geographical Institute and non-parametric statistical tests are presented and discussed, showing a remarkable performance and the significance of the obtained results.

Determining the best set of seismicity indicators to predict earthquakes. Two case studies: Chile and the Iberian Peninsula

This work explores the use of different seismicity indicators as inputs for artificial neural networks. The combination of multiple indicators that have already been successfully used in different seismic zones by the application of feature selection techniques is proposed. These techniques evaluate every input and propose the best combination of them in terms of information gain. Once these sets have been obtained, artificial neural networks are applied to four Chilean zones (the most seismic country in the world) and to two zones of the Iberian Peninsula (a moderate seismicity area). To make the comparison to other models possible, the prediction problem has been turned into one of classification, thus allowing the application of other machine learning classifiers. Comparisons with original sets of inputs and different classifiers are reported to support the degree of success achieved. Statistical tests have also been applied to confirm that the results are significantly different than those of other classifiers. The main novelty of this work stems from the use of feature selection techniques for improving earthquake prediction methods. So, the information gain of different seismic indicators has been determined. Low ranked or null contribution seismic indicators have been removed, optimizing the method. The optimized prediction method proposed has a high performance. Finally, four Chilean zones and two zones of the Iberian Peninsula have been characterized by means of an information gain analysis obtained from different seismic indicators. The results confirm the methodology proposed as the best features in terms of information gain are the same for both regions.

Computational intelligence techniques for predicting earthquakes

Nowadays, much effort is being devoted to develop techniques that forecast natural disasters in order to take precautionary measures. In this paper, the extraction of quantitative association rules and regression techniques are used to discover patterns which model the behavior of seismic temporal data to help in earthquakes prediction. Thus, a simple method based on the k-smallest and k-greatest values is introduced for mining rules that attempt at explaining the conditions under which an earthquake may happen. On the other hand patterns are discovered by using a tree-based piecewise linear model. Results from seismic temporal data provided by the Spanishs Geographical Institute are presented and discussed, showing a remarkable performance and the significance of the obtained results.

Detecting precursory patterns to enhance earthquake prediction in Chile

The prediction of earthquakes is a task of utmost difficulty that has been widely addressed by using many different strategies, with no particular good results thus far. Seismic time series of the four most active Chilean zones, the country with largest seismic activity, are analyzed in this study in order to discover precursory patterns for large earthquakes. First, raw data are transformed by removing aftershocks and foreshocks, since the goal is to only predict main shocks. New attributes, based on the well-known b-value, are also generated. Later, these data are labeled, and consequently discretized, by the application of a clustering algorithm, following the suggestions found in recent literature. Earthquakes with magnitude larger than 4.4 are identified in the time series. Finally, the sequence of labels acting as precursory patterns for such earthquakes are searched for within the datasets. Results verging on 70% on average are reported, leading to conclude that the methodology proposed is suitable to be applied in other zones with similar seismicity. Discovery of precursory patterns for medium-large earthquakes in Chile.b-values ability to predict earthquakes is confirmed.General purpose methodology: can be applied to any other area.Accuracy rate above 70%.

A fast partitioning algorithm using adaptive Mahalanobis clustering with application to seismic zoning

In this paper we construct an efficient adaptive Mahalanobis k-means algorithm. In addition, we propose a new efficient algorithm to search for a globally optimal partition obtained by using the adoptive Mahalanobis distance-like function. The algorithm is a generalization of the previously proposed incremental algorithm (Scitovski and Scitovski, 2013). It successively finds optimal partitions with k = 2 , 3 , ? clusters. Therefore, it can also be used for the estimation of the most appropriate number of clusters in a partition by using various validity indexes. The algorithm has been applied to the seismic catalogues of Croatia and the Iberian Peninsula. Both regions are characterized by a moderate seismic activity. One of the main advantages of the algorithm is its ability to discover not only circular but also elliptical shapes, whose geometry fits the faults better. Three seismogenic zonings are proposed for Croatia and two for the Iberian Peninsula and adjacent areas, according to the clusters discovered by the algorithm. HighlightsAn efficient adaptive Mahalanobis k-means algorithm is constructed.Optimal partition is obtained by the adaptive Mahalanobis distance-like function.Seismogenic zonings are proposed for Croatia and for the Iberian Peninsula.

Medium---large earthquake magnitude prediction in Tokyo with artificial neural networks

This work evaluates artificial neural networks’ accuracy when used to predict earthquakes magnitude in Tokyo. Several seismicity indicators have been retrieved from the literature and used as input for the networks. Some of them have been improved and parameterized in order to extract more valuable knowledge from datasets. The experimental set-up includes predictions for five consecutive datasets referring to year 2015, earthquakes with magnitude larger than 5.0 and for a temporal horizon of seven days. Results have been compared to four well-known machine learning algorithms, reporting very promising results in terms of all quality parameters evaluated. The statistical tests applied conclude that differences between the proposed artificial neural network and the other methods are significant.

Identifying P phase arrival of weak events

Seismic P phase arrival picking of weak events is a difficult problem in seismology. The algorithm proposed in this research is based on Empirical Mode Decomposition (EMD) and on the Akaike Information Criterion (AIC) picker. It has been called the EMD-AIC picker. The EMD is a self-adaptive signal decomposition method that not only improves Signal to Noise Ratio (SNR) but also retains P phase arrival information. Then, P phase arrival picking has been determined by applying the AIC picker to the selected main Intrinsic Mode Functions (IMFs). The performance of the EMD-AIC picker has been evaluated on the basis of 1938 micro-seismic signals from the Yongshaba mine (China). The P phases identified by this algorithm have been compared with manual pickings. The evaluation results confirm that the EMD-AIC pickings are highly accurate for the majority of the micro-seismograms. Moreover, the pickings are independent of the kind of noise. Finally, the results obtained by this algorithm have been compared to the wavelet based Discrete Wavelet Transform (DWT)-AIC pickings. This comparison has demonstrated that the EMD-AIC picking method has a better picking accuracy than the DWT-AIC picking method, thus showing this methods reliability and potential. An EMD-AIC picker has been proposed to identify micro-seismic P phase arrival.The EMD-AIC picking method works efficiently for the majority of identifications.The EMD-AIC method has a better picking accuracy than the DWT-AIC pickings.

A Novel Method for Seismogenic Zoning Based on Triclustering: Application to the Iberian Peninsula

A previous definition of seismogenic zones is required to do a probabilistic seismic hazard analysis for areas of spread and low seismic activity. Traditional zoning methods are based on the available seismic catalog and the geological structures. It is admitted that thermal and resistant parameters of the crust provide better criteria for zoning. Nonetheless, the working out of the rheological profiles causes a great uncertainty. This has generated inconsistencies, as different zones have been proposed for the same area. A new method for seismogenic zoning by means of triclustering is proposed in this research. The main advantage is that it is solely based on seismic data. Almost no human decision is made, and therefore, the method is nearly non-biased. To assess its performance, the method has been applied to the Iberian Peninsula, which is characterized by the occurrence of small to moderate magnitude earthquakes. The catalog of the National Geographic Institute of Spain has been used. The output map is checked for validity with the geology. Moreover, a geographic information system has been used for two purposes. First, the obtained zones have been depicted within it. Second, the data have been used to calculate the seismic parameters (b-value, annual rate). Finally, the results have been compared to Kohonen’s self-organizing maps.

A sensitivity study of seismicity indicators in supervised learning to improve earthquake prediction

The use of different seismicity indicators as input for systems to predict earthquakes is becoming increasingly popular. Nevertheless, the values of these indicators have not been systematically obtained so far. This is mainly due to the gap of knowledge existing between seismologists and data mining experts. In this work, the effect of using different parameterizations for inputs in supervised learning algorithms has been thoroughly analyzed by means of a new methodology. Five different analyses have been conducted, mainly related to the shape of training and test sets, to the calculation of the b-value, and to the adjustment of most collected indicators. Outputs sensitivity has been determined when any of these factors is not properly taken into consideration. The methodology has been applied to four Chilean zones. Given its general-purpose design, it can be extended to any location. Similar conclusions have been drawn for all the cases: a proper selection of the sets length and a careful parameterization of certain indicators leads to significantly better results, in terms of prediction accuracy.