Tsunami Currents in Hilo Harbor
Numerical results for the three requested resolutions (2/3, 1/3 and 1/6 arc-sec) were computed and a sensitivity study to friction with values for n=0.025 (requested), 0.030 and 0.035 was performed. For the local results in the reduced domain (mandatory), the boundary condition aimed at reproducing the sea surface elevation provided at the control point was set by imposing at the upper grid boundary the same time series provided for the control point. As result a good match with the time series at control point (not shown) was obtained. For the complete scenario (optional), three nested meshes where used with decreasing resolutions of 128/3, 8/3 arc-sec and for the finer grid the same three resolutions as for the local case were used (2/3, 1/3 and 1/6 arc-sec). Spatial resolution of outer meshes was also varied to 256/3 and 16/3 in order to assess the role of “ambient” meshes resolution in model results. Two Tohoku 2011 sources, one from NOAA and another from GeoClaw, were used.
Some conclusions can be extracted from the numerical results. First for the reduced scenario, at HA125 the two initial perturbations in the u and v components of the velocity are well captured, then the simulated u component becomes too oscillatory. In general the effects of varying friction and increasing resolution are very limited, and they are only slightly felt in the maximum and minimum values of the time series (in speed and surface elevation). The observed behavior for HA126 numerical results is similar to the one described for HA125. But in this case larger differences are observed for different values of the friction parameter and increased resolution. The differences for the maximum speed maps between the 20m and 5m resolution simulation on the one hand and the 10m minus 5m resolution simulation on the other, suggest convergence of the numerical solution. Hilo TG fit is pretty good. In what respect the complete scenario, simulated from the source, the computed variables reproduce well the first two or three initial waves, in particular for the sea surface elevation in the control point. At HA125 the simulated signal is less oscillatory than in the local scenario and mesh refinement has a minor effect and depending on the sampling (see comments of other modelers), the v component of the speed is well captured or highly overestimated but for the first perturbation. Again, at HA126, larger but still limited sensitivity to mesh refinement is observed. The bad news come when trying to assess sensitivity to the ambient and intermediate mesh resolutions: Much larger sensitivity is now observed, meaning that the “ambient” resolution is important and that the outer meshes can not be very coarse, something we would like to have in order to speedup computations. On the other hand, in [14] is also shown that going beyond a certain resolution is also useless. All this, while the role if increasing inner resolution remains limited (for the three resolutions considered here). Finally, two different sources where used for the complete scenario. Both simulations surprisingly agree quite well, despite the differences in the sources. In general, time series for the velocity components obtained when GeoClaw source is imposed are more oscillatory, mainly for the u component, but both simulations produce good results.
Figure 2. Measured data (dashed line) and numerical simulation (solid line) at the harbor tide gage (top), HA25 ADCP (middle), and HA26 ADCP (bottom).
Figure 3. Maximum predicted fluid speed during entire duration of the 10-m resolution simulation.
