**1. Introduction**

High‐frequency (HF) radars were first used for ocean observations in the 1960s. Located on the coast and transmitting vertically polarized radiations, they exploit the high conductivity of sea

© 2016 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. © 2016 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

water to propagate their signals well beyond the visible or microwave‐radar horizon. They have found widespread use for mapping surface currents and monitoring sea state.

Barrick [1] suggested in 1979 that these radars could detect tsunamis by means of their orbital wave velocity as they approach the coast. Because the distribution of radars around the world was sparse until the 1990s, this concept was not pursued until the tsunami caused by the catastrophic 2004 Banda Aceh earthquake in Indonesia claimed a quarter of a million lives. Although there were no radars in place to observe that event, work began to quantify the radar tsunami response. It was not until the 2011 Tohoku (Japan) tsunami that sufficient radars were in place to capture real tsunami data, which led to development of algorithms to provide robust detection and warning. The Japan tsunami signal was observed by many HF radars around the Pacific Rim with clear results from sites in Japan, USA, and Chile [2–4]. Additional weak tsunamis have also been observed: the 2012 Indonesia tsunami [5] and the 2013 US East Coast meteotsunami [6]. A database of actual HF radar tsunami observations from both strong and weak tsunamis has been accumulated, which has been used to identify the tsunami current velocity signature in the background ocean current velocity field. An empirical method for the automatic detection of a tsunami has been developed, based on pattern recognition in the velocity time series. Over 350 HF radar stations operate from many coastal locations, see, for example, http://www.codar.com/seasonde\_world\_locations.shtml, providing continuous measurement of surface current velocities and waves. Tsunami detection software can run in a background mode, issuing a warning before the tsunami strikes the coast.

The first possible indication of a tsunami might be the seismic detection of an earthquake. However, not all subsea earthquakes produce tsunamis, and hence the magnitude of an earthquake cannot be used to forecast the detailed generation or intensity of a resulting tsunami. At present, the only operational sensor that detects a tsunami and measures its intensity is a bottom pressure sensor connected to a buoy overhead. Developed by the National Oceanic and Atmospheric Administration (NOAA), networks of these sensors called DART™ (Deep‐ocean Assessment and Reporting of Tsunami) were deployed after the 2004 event. They observe the height of the tsunami wave as it passes above them. The tsunami height measured by these buoys is then entered into numerical tsunami models [7–8] to give rough forecasts of the tsunami arrival time and intensity at coastal points around the world. As this network is located in the deep ocean, not all tsunamis are observable by DART and then are not entered into the model before coastal impact. Furthermore, the model's forecast of intensity at the coast is often coarse, so that more accurate estimates of intensity at specific locations are needed; such local variations not captured by the models are referred to as "near field." HF radars make their areal observations over this local near field and so provide an ideal solution to this need. We describe an empirical tsunami detection algorithm that can run in the background on these radars. This can detect and warn of an approaching tsunami in the near‐shore region over which these radars observe the sea surface. A total of 21 offline radar detections of tsunamis have been made to date. Many are described here and others are reported in the literature [2, 3, 5].

A tsunami's orbital velocity appears as part of the surface current as the wave approaches the coast. Tsunami periods lie typically between 20 and 50 min. A tsunami originates when there is a massive displacement of water: The spatial scales of water displacement are the spatial scales of water displacement are usually of great horizontal but small vertical dimensions These include subsea earthquakes where plates force each other upward/downward, respec‐ tively; subsea landslides along steep submerged mountainous slopes; or fast‐moving atmos‐ pheric anomalies (e.g. low‐pressure centers) that create "meteotsunamis." The sources can be thousands of kilometers from an impacted coastal area (where an HF radar might be located), or very close. As the displaced water mass leaves its source region under the influence of gravity, it becomes a freely propagating shallow‐water wave. Although the origins of meteot‐ sunamis vis‐à‐vis seismically generated tsunamis differ, the propagation and evolution of these shallow‐water waves are the same, as are the applicable detection and warning methods. Tsunami warning times are mainly dependent on the width of the adjacent continental shelf, ranging from minutes for a narrow shelf (e.g. California) to hours when the shelf is broad (e.g. New Jersey).

Some sites may be less suitable for tsunami monitoring by radar, as the tsunami signature can be masked by large, variable background currents. Tsunami detection is favored by shallow water extending far offshore and by slowly varying background current fields. We describe a method for the evaluation of a coastal site for tsunami warning based on simulated tsunami velocities superimposed on the site's measured velocities. Factors affecting radar detection of tsunamis are discussed. Difficulties that can occur in tsunami detection and methods for alleviation are described. At present, work on the evaluation of coastal sites for tsunami warning using HF radars is being performed in a partnership between Codar Ocean Sensors and NOAA.

It is often convenient to identify Codar SeaSonde® radars by their abbreviated site names used in the field. Those referred to in this article are listed here in alphabetical order, along with their geographical locations:

A087: Usujiri, Hokkaido, Japan A088: Kinaoshi, Hokkaido, Japan BELM: Belmar, New Jersey, USA BML1: Bodega Marine Lab., California, USA BRMR: Brigantine, New Jersey, USA BRNT: Brant Beach, New Jersey, USA COMM: Commonweal, California, USA ESTR: Point Estero, California, USA GCVE: Granite Cove, California, USA LUIS: Point San Luis, California, USA PREY: Point Reyes, California, USA SEA1: Seaside, Oregon, USA

water to propagate their signals well beyond the visible or microwave‐radar horizon. They have

Barrick [1] suggested in 1979 that these radars could detect tsunamis by means of their orbital wave velocity as they approach the coast. Because the distribution of radars around the world was sparse until the 1990s, this concept was not pursued until the tsunami caused by the catastrophic 2004 Banda Aceh earthquake in Indonesia claimed a quarter of a million lives. Although there were no radars in place to observe that event, work began to quantify the radar tsunami response. It was not until the 2011 Tohoku (Japan) tsunami that sufficient radars were in place to capture real tsunami data, which led to development of algorithms to provide robust detection and warning. The Japan tsunami signal was observed by many HF radars around the Pacific Rim with clear results from sites in Japan, USA, and Chile [2–4]. Additional weak tsunamis have also been observed: the 2012 Indonesia tsunami [5] and the 2013 US East Coast meteotsunami [6]. A database of actual HF radar tsunami observations from both strong and weak tsunamis has been accumulated, which has been used to identify the tsunami current velocity signature in the background ocean current velocity field. An empirical method for the automatic detection of a tsunami has been developed, based on pattern recognition in the velocity time series. Over 350 HF radar stations operate from many coastal locations, see, for example, http://www.codar.com/seasonde\_world\_locations.shtml, providing continuous measurement of surface current velocities and waves. Tsunami detection software can run in

found widespread use for mapping surface currents and monitoring sea state.

a background mode, issuing a warning before the tsunami strikes the coast.

3, 5].

74 Tsunami

The first possible indication of a tsunami might be the seismic detection of an earthquake. However, not all subsea earthquakes produce tsunamis, and hence the magnitude of an earthquake cannot be used to forecast the detailed generation or intensity of a resulting tsunami. At present, the only operational sensor that detects a tsunami and measures its intensity is a bottom pressure sensor connected to a buoy overhead. Developed by the National Oceanic and Atmospheric Administration (NOAA), networks of these sensors called DART™ (Deep‐ocean Assessment and Reporting of Tsunami) were deployed after the 2004 event. They observe the height of the tsunami wave as it passes above them. The tsunami height measured by these buoys is then entered into numerical tsunami models [7–8] to give rough forecasts of the tsunami arrival time and intensity at coastal points around the world. As this network is located in the deep ocean, not all tsunamis are observable by DART and then are not entered into the model before coastal impact. Furthermore, the model's forecast of intensity at the coast is often coarse, so that more accurate estimates of intensity at specific locations are needed; such local variations not captured by the models are referred to as "near field." HF radars make their areal observations over this local near field and so provide an ideal solution to this need. We describe an empirical tsunami detection algorithm that can run in the background on these radars. This can detect and warn of an approaching tsunami in the near‐shore region over which these radars observe the sea surface. A total of 21 offline radar detections of tsunamis have been made to date. Many are described here and others are reported in the literature [2,

A tsunami's orbital velocity appears as part of the surface current as the wave approaches the coast. Tsunami periods lie typically between 20 and 50 min. A tsunami originates when there

STV2: Fort Stevens, Oregon, USA TRIN: Trinidad, California, USA YHS2: Yaquina Head South, Oregon, USA
