provide the additional data needed to construct tsunami flood level 

 maps for various probabilities of recurrence. 



Numerical procedures can also be verified by comparing with theoreti- 

 cal results for idealized cases. Theoretical solutions exist for wave 

 refraction at coastlines with uniform topography, waves passing over 

 mathematically defined transitions from deep water to shallow water, etc. 

 Deviations between the numerical results and the theoretical solutions 

 indicate the degree of accuracy where the numerical procedures are applied 

 to more complex topography. 



A continuing program of gathering field data on tsunamis in the open 

 ocean and coastal inundation by tsunamis is needed. Because of the long 

 periods of time between the occurrence of tsunamis, the accumulation of 

 data for particular coastal points is very slow. It is necessary to 

 maintain tide gages with the capability of recording tsunamis, and also 

 to have standby plans with designated personnel to obtain field observa- 

 tions immediately after tsunamis occur. It is also desirable to maintain 

 a standby capability for dropping instrument packages into the open ocean 

 immediately after a tsunami occurs. This latter capability requires the 

 maintenance of gages and associated instrument packages in operating 

 condition over long periods of time, and the maintenance of a system for 

 placing the instrument packages quickly and on short notice, including 

 the periodic testing of the system by placing and recovering the instru- 

 ments. An air-dropped system is probably the most practical for this 

 purpose. 



Also, continuing improvements are needed in the numerical procedures 

 for simulating tsunamis. A particular area of possible improvement is 

 the treatment of boundaries of the computational grid. Errors in the 

 wave reflection from solid boundaries, and errors at open-ocean boundaries 

 where the waves must pass completely through the boundary, propagate 

 through the computational grid at each succeeding time step. These errors 

 grow with increasing time so that the solution is not accurate for long 

 periods of real time. It becomes necessary to use large time steps to 

 reduce computational errors, and consequently to use a course grid, i.e., 

 to use long real distances between grid points. This smooths out the 

 topographical variations so that wave scattering caused by small topograph- 

 ical features is not properly accounted for. Because of the limited field 

 data available, the numerical solutions cannot always be verified and 

 adjusted to match field data. 



Improvements in the numerical simulation of tsunami generation are 

 also desirable. However, this requires both accurate data on real tsunami- 

 generating mechanisms, and open-ocean tsunami data so that errors in 

 simulating tsunami generation can be separated from errors in the simula- 

 tion of nearshore propagation. 



Continued research should be carried out in areas such as shelf 

 resonance. In particular, theoretical solutions are needed for simple 

 topography to provide verification for numerical procedures where field 

 data do not exist. At the present time, the various effects on tsunami 

 propagation cannot be adequately separated in the computational procedure 

 as the available data are mainly from tide gages and visual observations 

 of maximum inundation levels. 



200 



