A tsunami warning center (TWC) operates on the fundamental principle that seismic waves travel 20 to 30 times faster through the earth than tsunamis travel through the oceans. This speed difference allows TWC scientists to determine the tsunami-generating potential of an earthquake, or even analyze the tsunami itself, before the tsunami reaches threatened coastlines. Given that we know where tsunami-generating earthquakes are likely to occur, how fast both seismic and tsunami waves propagate, and the distribution of both seismic and sea-level sensors installed around the world, we can predict how rapidly a TWC should be able to respond to any tsunamigenerating earthquake and thus which coastlines will receive a tsunami warning in time to react and which coastlines will get little or no warning (Becker et al., EOS Trans. AGU, 2010). TWCs can forecast a tsunami and issue a warning based on earthquake characteristics alone, but near real-time sea-level measurements are necessary for the TWC to refine its forecast and expand, narrow, or cancel tsunami warnings, to not only protect more lives and property when a tsunami is more severe than initially forecast, but to also prevent unnecessary and costly evacuations when a tsunami is less intense and/or more focused than initially forecast.

To determine how soon a TWC will have useful sea-level data after an earthquake, we calculated tsunami travel times (TTTs) from likely tsunami sources, i.e., thrust faults at subduction zones, to both coastal and deep-ocean sea level sensors monitored by the Pacific Tsunami Warning Center (PTWC). For this purpose we geographically sampled the axes of deep-sea trenches with a geodesic grid to ensure equal spacing of simulated sources. We then calculated global TTT grids for each of these sources and sampled each grid at the locations of coastal and deep-ocean sea-level gauges currently monitored by the Pacific Tsunami Warning Center (PTWC). We then added each gauge's transmission interval to the TTT values sampled at the gauges to determine the maximum time required for one, two, and three gauges to detect and transmit information about a tsunami generated by these sources. We also analyzed the network available to PTWC in 2005, showing how detection times have improved since then, and a hypothetical network in which all coastlines and oceans are saturated with sensors, showing that further improvements are still possible. We also used these data to quantify how the network can be compromised by sensor outages, identify those tsunami sources most in need of additional sea-level sensors for tsunami detection, and determine where future sensors should be located. As installing one deep-ocean sensor costs about 10 times as much as installing a coastal gauge, we also determine which sensor is more cost-effective for filling these network gaps. These analyses show that for global tsunami hazard mitigation the installation of about 100 additional carefully-selected coastal sea-level gauges could greatly improve the speed of tsunami detection and characterization.