Analysis of existing tsunami scenario databases for optimal design and efficient real-time event matching

Pre-computed tsunami scenario databases constitute a traditional basis to the production of tsunami forecasts in real time, achieved through a combination of properly selected Green’s functions-like objects. The considered case-study database contains water elevation fields and waveform signals produced by an arrangement of evenly-spaced elementary seismic sources, covering fault areas relevant in determining the Portuguese tsunami hazard. This work proposes a novel real-time processing for the tsunami forecast production, aiming at the accuracy given by traditional methods but with less time cost. The study has been conducted on the Gorringe Bank fault (GBF), but has a general validity. First, the GBF database is analysed in detail, seeking for remarkable properties of the seismic sources, in terms of frequency content, cross-correlation and relative differences of the fields and waveform signals. Then, a reference forecast for a seismic event placed on the GBF is given, by using all the traditionally available subfaults. Furthermore, a novel processing algorithm is defined to produce approximate forecasts, through a strategic exploitation of the information obtainable by each of the seismic sources, taken in minor number. A further focus on sensible locations is provided. Remarkable results are obtained in terms of physical properties of the seismic sources and time-gain for the forecast production. Seismic sources at depth produce longwave dominated signals, allowing for an optimisation of the database content, in terms of sources required to properly represent seismogenic areas at certain depths. In terms of time cost, an overall improvement is obtained concerning the forecast production, since the proposed strategy gives highly accurate forecasts, using half of the seismic sources used by traditional forecasting methods, which reduces the required accesses to the database.