14:30 〜 14:45
[HDS10-04] Numerical Experiments on Tsunami Inundation Estimation Using a Data Assimilation Approach
キーワード：津波、データ同化、浸水
In recent years, the development of seismic and tsunami observation networks in oceanic regions has advanced globally. In particular, Japan has established high-density offshore observation networks such as S-net, DONET, and N-net. Tsunami data assimilation approach, one of the real- time tsunami forecasting methods, can estimate tsunami propagation using time-series pressure data recorded during a tsunami at the offshore stations without source estimates.
Previous studies have thoroughly investigated the performance of tsunami data assimilation for large-scale tsunami propagation, whereas little attention has been given to its performance in tsunami inundation estimation. The objective of this study is to evaluate the effectiveness of the tsunami data assimilation in estimating tsunami inundation characteristics in real-time.
We focused on the Hokkaido coasts in the Pacific Ocean side and adopted a source model for the large earthquake that occurred along the Kuril Trench in the 17th century. The first simulation estimated tsunami propagation and inundation using the source model. Furthermore, the estimated changes in water surface level at the offshore stations (S-net) were converted into pressure changes with hydrostatic approximation to be used as time-series pressure data in the second simulation. In the second simulation, the tsunami propagation and inundation were estimated by assimilating the time-series data at the offshore stations without noise. According to the first simulation results, the tsunami arrival time at Kushiro was approximately 20 minutes, where severe tsunami inundation might occur due to its large low-lying coastal area. Based on this result, we tested three cases in which the time-series data were assimilated up to 5, 10, and 15 minutes after the earthquake.
We compared the tsunami inundation characteristics at three representative regions, Hanasaki, Kushiro, and Tokachi, in the second simulation with those in the first simulation. The data assimilation approach reasonably reproduced the maximum water surface level during the primary wave at nearshore stations of the target regions. The reproduction increased as the data assimilation time increased. At the case where the time-series data were assimilated up to 5 minutes, the inundation area was reasonably or moderately estimated in Hanasaki and Tokachi but was significantly overestimated in Kushiro. However, similar to the nearshore water surface level, the reproduction of inundation tended to increase as the data assimilation time increased. In contrast, we found that increasing the data assimilation time does not necessarily increase the reproduction of inundation in other coastal areas. This finding may suggest that the optimal data assimilation time is influenced by various factors such as characteristics of tsunamis, the locations of offshore observational stations, and nearshore topography in inundation zone. Further numerical experiments will be necessary to quantify the capability of tsunami data assimilation for estimating inundation.
Previous studies have thoroughly investigated the performance of tsunami data assimilation for large-scale tsunami propagation, whereas little attention has been given to its performance in tsunami inundation estimation. The objective of this study is to evaluate the effectiveness of the tsunami data assimilation in estimating tsunami inundation characteristics in real-time.
We focused on the Hokkaido coasts in the Pacific Ocean side and adopted a source model for the large earthquake that occurred along the Kuril Trench in the 17th century. The first simulation estimated tsunami propagation and inundation using the source model. Furthermore, the estimated changes in water surface level at the offshore stations (S-net) were converted into pressure changes with hydrostatic approximation to be used as time-series pressure data in the second simulation. In the second simulation, the tsunami propagation and inundation were estimated by assimilating the time-series data at the offshore stations without noise. According to the first simulation results, the tsunami arrival time at Kushiro was approximately 20 minutes, where severe tsunami inundation might occur due to its large low-lying coastal area. Based on this result, we tested three cases in which the time-series data were assimilated up to 5, 10, and 15 minutes after the earthquake.
We compared the tsunami inundation characteristics at three representative regions, Hanasaki, Kushiro, and Tokachi, in the second simulation with those in the first simulation. The data assimilation approach reasonably reproduced the maximum water surface level during the primary wave at nearshore stations of the target regions. The reproduction increased as the data assimilation time increased. At the case where the time-series data were assimilated up to 5 minutes, the inundation area was reasonably or moderately estimated in Hanasaki and Tokachi but was significantly overestimated in Kushiro. However, similar to the nearshore water surface level, the reproduction of inundation tended to increase as the data assimilation time increased. In contrast, we found that increasing the data assimilation time does not necessarily increase the reproduction of inundation in other coastal areas. This finding may suggest that the optimal data assimilation time is influenced by various factors such as characteristics of tsunamis, the locations of offshore observational stations, and nearshore topography in inundation zone. Further numerical experiments will be necessary to quantify the capability of tsunami data assimilation for estimating inundation.