**7. Discussion and conclusion: the predictability of meteorological tsunami**

The recent studies on meteorological tsunami presented that much more larger-scale motions played an important role in forming the atmospheric structures to generate meso-β or γ-scale disturbances at sea level. Some latest papers pointed the importance of the (sub-) seasonal scale variation propagating from lower to mid-latitude.

Those findings in larger-scale motion will provide the capability for the prediction of the meteorological tsunamis from atmospheric and oceanographic aspects. It should be noticed that there have been much advances in technologies on observation, numerical modelling and others to make progresses scientifically in meteorological tsunamis. Especially, the highperformance computing technology brings us huge benefits to resolve the pressure disturbance in using numerical model such as the weather forecasting and research (WRF) model [46]. **Figure 14** illustrates a framework on the prediction of the meteorological tsunami. In obtaining a first guess, it is better to start the weather forecast in the synoptic scale with the duration of 2 or 3 days to make a first guess on the stability structure related to the wave duct or wave CISK. And then the downscaling run executes to resolve the atmospheric (mainly pressure) disturbance at sea level.

**Figure 14.** A framework on the prediction of the meteorological tsunami.

The infrared images of the satellite remote sensing have useful information if the sequence of the wavy clouds in the mid-troposphere moves fast over the ocean satisfying the Proudman resonance. A new geostationary meteorological satellite named Himawari-8 stated the operation in Asia-Australian region since July 2015. The Himawari-8 has 16-image channels of visible, near infrared, infrared bands, with the horizontal resolution of 0.5 (visible) to 2 km (infrared). The scan interval is every 10 min for global mode, and every 2.5 min for Japan area. The target such as typhoon or hurricane also scans for every 2.5 min for necessity. It will be much easier to see the propagation of the atmospheric wave using satellite data.

Although the mesoscale model can resolve the pressure disturbance to a considerable extent, it is not easy to apply directly into the ocean models as a surface boundary condition, because of the difficulty in giving enough accuracy as compared to the in situ observation. Even small errors of the location, wave length, frequency, the propagation direction, etc. might affect the amplitude in the inlet. Indeed, Renault et al. (2011) [47] is the first to succeed the meteotsunami simulation by coupling of the WRF and ROMS (regional ocean modelling system) [48]. But the results of the pressure disturbance from the meteorological model should be treated very carefully. If the powerful computing resource is available, it is desirable to execute the ensemble run using various initial conditions, cloud physics parameter and boundary layer options for checking the uncertainty of the model behaviour.

The enhancement of the in situ observation network is one of the ways to provide a short-term prediction or real-time warning, in which the lead time is shorter than 1 h in many situations. It is convenient to install the automated station on isolated islands around the targeted area such as Adriatic Sea [49]. However, it seems not to be applicable such as US East Coast. The high-frequency (HF) radar extends the possibility for monitoring the meteorological tsunami propagation. Lipa et al. (2014) [50] detected the wave up to 23 km offshore, 47 min before reaching the shoreline by the analysis of HF radar signals on the June 2003 event on the US East Coast.

Indeed, further developments are required to get more accurate outputs for linking each processes, but we can believe that the day will come to predict the meteorological tsunami in a practical manner.
