Characteristics of the 2011 Tohoku Tsunami and introduction of two level tsunamis for tsunami disaster mitigation

Characteristics of the 2011 Tohoku Tsunami have been revealed by collaborative tsunami surveys extensively performed under the coordination of the Joint Tsunami Survey Group. The complex behaviors of the mega-tsunami were characterized by the unprecedented scale and the low occurrence frequency. The limitation and the performance of tsunami countermeasures were described on the basis of tsunami surveys, laboratory experiments and numerical analyses. These findings contributed to the introduction of two-level tsunami hazards to establish a new strategy for tsunami disaster mitigation, combining structure-based flood protection designed by the Level-1 tsunami and non-structure-based damage reduction planned by the Level-2 tsunami.     

Subterranean electrical structure of Kozu-shima volcanic island, Japan

Following the electric current injection experiment carried out in 2009, a VLF-MT (Very Low Frequency Magnetotelluric) survey has been conducted in Kozu-shima Island to obtain further information on the subterranean electrical structure that might help understanding the results of our monitoring of geoelectric potentials. A number of VAN-type pre-seismic geoelectric potential anomalies were observed in 1997-2000, even showing a remarkable "Selectivity". However, similar pre-seismic anomalies were not observed during the Izu-Island volcano-seismic swarm 2000. All these observations would require extremely high degree of heterogeneity in the subterranean electrical structure of the volcanic island and its possible time changes. Several correlations between the results of this survey and the volcanic geology of the island and ground water distribution were found. Further investigation is needed for a complete explanation of the observed phenomena.     

基于熵学习机的恒星光谱分类

数据挖掘被广泛应用于恒星光谱分类。为了提高传统光谱分类方法性能,提出熵学习机(Entropy-based Learning Machine, ELM)。在该方法中,熵用来刻画分类的不确定性。为了得到理想的分类结果,分类的不确定性应最小,基于此,可得ELM的最优化问题。ELM在处理二分类问题和稀有光谱发现等方面具有一定优势。SDSS中K型、F型、G型恒星光谱数据集上的比较实验表明：ELM在进行恒星光谱分类时,其分类性能优于k近邻(k Nearest Neighbor)和支持向量机(Support Vector Machine)等传统分类方法。