@article{oai:nagasaki-u.repo.nii.ac.jp:00000223,
 author = {Nakagawa, Kei and Yu, Zhi-Qiang and Berndtsson, Ronny and Kagabu, Makoto},
 issue = {15},
 journal = {Environmental Earth Sciences},
 month = {Jul},
 note = {For a better understanding of possible physical links between geophysical observables and earthquake characteristics, it is important to analyze statistical spatiotemporal patterns in nature related to such events. For this purpose,characteristic changes in groundwater level (GWL) were observed before and after the 2016 Kumamoto earthquake in Japan. Previous research has shown that self-organizing maps (SOM) can be used to classify complex patterns of GWL-change during different parts of the earthquake sequence. In this study, we used before and after earthquake GWL data as input vectors to SOM. In total, 64 observed GWLs were classified into 12 different clusters. Most shallow wells displayed GWL difference that was small during the foreshock (first earthquake) and large during the main-shock (second earthquake). Upstream deep wells showed relatively large difference in water level from 1 to 2 days after the earthquakes. The GWL rapidly increased just after the earthquake, then tended to gradually decrease from September. Most of the shallow wells in the unconfined aquifer rapidly recovered to initial GWLs within several hours to several days, because of hydrostatic pressure. However, most of the deep wells in the confined aquifer needed longer time to recover, in some cases several weeks to several months. These findings are important for the physical understanding of earthquake effects on the groundwater environment, disaster prevention, and possibility for development of earthquake precursors., Environmental Earth Sciences, 78(15), art.no.455; 2019},
 title = {Analysis of earthquake-induced groundwater level change using self-organizing maps},
 volume = {78},
 year = {2019}
}