Geoscience Reference
In-Depth Information
date with 16 events has shown that even a single DART buoy is suficient to scale the
pre-computed wave ields appropriately for qualitatively accurate predictions.
3.
Inundation estimates using the nonlinear model, Method of Splitting Tsunami (MOST), are
developed
: Once the combinations of wave ields from the pre-computed scenarios
are constrained by the DART sea level data using the least squares it technique, the
database is queried for wave height and luid velocity time series at all sea-boundaries
of the region targeted for the inundation forecast. At each boundary point, the time
histories of heights and velocities are used to initialize the boundary conditions. The
inundation computation proceeds using the nonlinear MOST model that includes
shoaling computations of wave inundating dry topography, until inundation estimates
are obtained. The process is built on the Synolakis (1987) theory of a solitary wave
propagating over constant depth and then evolving over a sloping beach. The wave
ield of approaching waves in deep waters are assumed to be linear, so there are rea-
sonable interim estimates for the entire low including relection from the beach; i.e.,
where the constant depth and sloping regions connect. Once there is a linear solution
in the deep waters (where depths are more than 20 m), this input can be used to solve
the nonlinear evolution problem on a sloping beach (Carrier and Greenspan, 1958).
Figure 4.9 displays the SIFT tsunami predictions at two stations after the February 2010,
Chilean earthquake. One of these stations is at an open-ocean island (Midway Island) at the
northwestern end of the Hawaiian Archipelago; the other station is at the North American
coast (Santa Barbara, California). In the open ocean, SIFT-predicted amplitudes (although not
the phases) agree fairly well with the observed. However, the igure also illustrates the dif-
iculty in predicting coastal amplitudes that are very sensitive to the small-scale details of
the model's bathymetry and coastal geometry. The highest observed wave at Santa Barbara,
occurring about four hours after the irst arrival, is missed by SIFT. For comparisons of SIFT
predictions with many other observations of the Chilean tsunami, go to http://nctr.pmel.noaa.
gov/chile20100227/.
For SIFT (but not for ATFM), the ability to make accurate forecasts of tsunami waves is
predicated on the availability of DART sea level measurements. The method's accuracy is tied
directly to receiving data from the sea loor in near-real time. ATFM can utilize sea level data
from both DART and coastal stations. To date the two technologies have successfully forecast
16 tsunamis with an accuracy of about 80 percent when compared with tide gauge data. (Titov
(Titov
et al., 2005; Tang et al., 2008, 2009; Wei et al., 2008; Titov, 2009). Although these models forecast
Although these models forecast
wave height reasonably well, forecasting the inundation remains a challenge. To date, only one
of the models (ATFM) is fully operational, although the SIFT model is being transitioned. At
present, based on its review the committee found no clear process by which the forecasts' skill
is evaluated and improved, nor by which the differences in the forecast outputs are recon-
ciled. As with the ensemble model approach for hurricane forecasts, the committee considers
it beneicial to run and compare multiple model outputs. However, a process is needed that
assists watchstanders in reconciling the differences and arriving at a single forecast output
to be transmitted in the warning products. Such a process is well established in the National
