Geoscience Reference
In-Depth Information
and detection system. These forecast models make available to emergency managers in near-
real time the time of irst impact as well as the sizes and duration of the tsunami waves, and
give an estimate of the area of inundation, similar to hurricane forecasting.
The entire forecasting process has to be completed very quickly. For example, Hawaii Civil
Defense needs about 3 hours to safely evacuate the entire coastline. As most far-ield tsunamis
generated in the North Paciic take less than 7 hours to strike Hawaii, the entire forecast, includ-
ing data acquisition, data assimilation, and inundation projections, must take place within
4 hours or less. Although this sounds like a comfortable margin, in fact it is quite a short time
period compared to many other natural disasters, especially since it can take anywhere from
30 minutes to 3 hours to acquire suficient sea level data (Whitmore et al., 2008). For hurricanes,
forecasts are made days in advance of landfall and evolve spatially at scales over 100 times
slower than a tsunami. The time window for a forecast for a near-ield tsunami event is even
smaller, because the irst waves may arrive in less than 30 minutes (see the section on
Instru-
mental Detection of Near-Field Tsunamis
below).
The importance of forecasting the duration of wave arrivals, and forecasting the sizes of
each arrival, is well known; for example, the largest and most destructive wave of the tsunami
originating off the Kuril Islands on November 15, 2006, was the sixth wave to hit Crescent
City, California. This wave hit more than two hours after the irst wave arrival (Uslu et al., 2007;
Barberopoulou et al., 2008; Dengler et al., 2008).
Although time-of-arrival information has been available since the 1960s (Ambraseys, 1960),
only beginning in the 1990s (e.g., Kowalik and Whitmore, 1991; Whitmore and Sokolowski,
1996; Titov and González, 1997), with full development not completed until a decade later,
have forecast methodologies been employed to provide estimates of inundation prior to wave
arrival and of duration (see Whitmore, 2003; Mofjeld, 2009; Titov, 2009). The use of near-real-
time forecasting models is only possible because of data from the coastal and open-ocean sea
level networks. Modeling tsunamis based on seismic data alone is currently not very accurate,
as noted in the above section on
Detection of Earthquakes
. The importance of accurate fore-
casts of maximum wave height was illustrated quite clearly in the wake of the recent Chilean
earthquake on February 27, 2010.
In the United States, NOAA's WC/ATWC and PMEL have developed distinct tsunami forecast
systems (respectively, Alaska Tsunami Forecast Model (ATFM), http://wcatwc.arh.noaa.gov/
DataProcessing/earthvu.htm; and SIFT, http://nctr.pmel.noaa.gov/tsunami-forecast.html) to
provide information on tsunami arrival times, wave sizes, and event durations at the shoreline.
An advanced version of the ATFM is currently in development at the WC/ATWC.
These systems employ pre-computed, archived event scenarios, in conjunction with near-
real-time sea level observations. The PMEL system takes the forecast a step further by provid-
ing inundation distances and run-up heights that enable even more targeted evacuations.
These forecast models allow the TWCs to make more accurate tsunami wave predictions than
were possible without them, enabling more timely and more spatially reined watches and
warnings (e.g., Titov et al., 2005; Geist et al., 2007; Whitmore et al., 2008). The PTWC was able
to forecast reasonably well the observed tsunami heights in Hawaii more than ive hours in
advance of the Chilean tsunami arrival (Appendix J). The models place an additional emphasis
