at World Trade Center, New York City.
Seismologists sometimes do their work of data acquisition and analysis against a tragic background. Usually the context is fieldwork far from home, in an area subjected to the natural but sometimes devastating effects of an earthquake. But in the present case we are in our own New York City area; that is, the Lamont-Doherty Earth Observatory of Columbia University, in Palisades, N.Y.; and the context is inhuman actions against people and the fabric of our society. As the appalling events of September 11 unfolded, we found that we had recorded numerous seismic signals from two plane impacts and building collapses from the two World Trade Center (WTC) towers, often at times different than those being reported elsewhere. Collapses of the two WTC towers generated large seismic waves, observed in five states and up to 428 km away. The North Tower collapse was the largest seismic source and had local magnitude ML 2.3. From this we infer that ground shaking of the WTC towers was not a major contributor to the collapse or damage to surrounding buildings, but unfortunately we also conclude that from the distance at which our own detections were made (the nearest station is 34 km away at Palisades, N.Y.) it is not possible to infer (with detail sufficient to meet the demands of civil engineers in an emergency situation) just what the near-in ground motions must have been.
Signals at Palisades from Impacts and Collapses
Figure 1 shows seismic signals at Palisades, N.Y. (PAL) for the impacts and collapses, which are labeled by their arrival time order. Note that impact 1 and collapse 2 relate to the north tower, and impact 2 and collapse 1 apply to the south tower. Computed origin times and seismic magnitudes are listed in Figure 1. Origin times with an uncertainty of 2 s were calculated from the arrival times of Rg waves at PAL using a velocity of 2 km/s. The collapse of 7 WTC at 17:20:33 EDT was recorded but is not shown. Three other small signals shown in Figure 1 and ones at 12:07:38 and 12:10:03 EDT may have been generated by additional collapses.
Figure 1: Seismic recordings on E-W component at Palisades for events at World Trade Center (WTC) on September 11, distance 34 km. Three hours of continuous data shown starting at 08:40EDT (12:40 UTC). Data were sampled at 40 times/s and passband filtered from 0.6 to 5 Hz. The two largest signals were generated by collapses of Towers 1 and 2. Eastern Daylight Time (EDT) is UTC minus 4 hours. Expanded views of first impact and first collapse shown in red. Displacement amplitude spectra in nm-s from main impacts and collapses shown at right. Sampling is done for 14-second time windows starting about 17 s after origin time. Note broadband nature of spectra for collapses 1 and 2. Their signals are similar with a correlation coefficient of about 0.9 as are those for two impacts.
Surface waves were the largest seismic waves observed at various stations. The presence of seismic body waves is questionable even at Palisades for the two largest collapses; they are not observed at other stations. Local magnitudes ML, like those defined originally by Richter for southern California but with distance correction factors appropriate for eastern North America [Kim, 1998], were computed for the two impacts and the three largest collapses. For collapses 1 and 2, values of ML determined from E-W components are 2.1 and 2.3. ML is 0.1 to 0.2 units smaller on the vertical, an observation that we associate later with multipath propagation.
Amplitude spectra for PAL data are shown at the right of Figure 1 for the impacts and the collapses of the twin towers. The spectra of collapses 1 and 2 are above the noise for frequencies from 0.2 to 10 Hz. The two spectra are similar, but the second shows a more pronounced peak near 1 Hz. Seismic signals from both impacts are characterized by relatively periodic motion and their spectra are above the noise only for frequencies from about 1.3 to 1.6 Hz. Those frequencies are more than 10 times the frequency of the lateral fundamental mode of each tower.
Observations in Mid-Atlantic States and New England
Lamont-Doherty operates 34 seismograph stations in seven Mid-Atlantic and New England states. The network has been in operation since the early 1970s, but the stations, types of recording, and data transmission have changed with time. Digital data are now sent via the Internet in real time to Palisades. They are supplemented by data from the U.S. National Seismic Network. The modern stations record over a broad frequency band; some like PAL sample three components of ground motion, but others, only the vertical. Information on the stations and WTC recordings is available at www.ldeo.columbia.edu/LCSN. The data were sent to the Data Management Center, Incorporated Research Institutions for Seismology (IRIS), in Seattle, Washington.
Seismic waves from Collapse 2 were recorded by at least 13 stations ranging in distance from 34 km to Lisbon, NH at 428 km. The magnitude of the event was only 2.3. The predominant signals at distances greater than 200 km are short-period surface waves, which propagate at wave speeds of about 3.5 km/s, the typical Lg group velocity observed for the largest waves from earthquakes at regional distances in eastern North America. Those observations will be published separately.
Seismic Waves in Greater New York City Area
Six stations within the greater Metropolitan New York region (Fig. 2) recorded the two tower collapses. Vertical-component records are shown in Figure 3 as a record section of distance as a function of travel time. The dotted lines indicate velocities from 1.5 to 2.5 km/s assuming propagation along straight paths from the WTC to the stations. Unlike signals at distant stations, the predominant waves are surface waves of short period (about 1 s) called Rg with group velocities between 2.3 and 1.5 km/s. GPD only recorded horizontal components.
Figure 2: Seismograph stations and topography for greater New York City area. Solid triangles indicate stations that recorded events at WTC (solid red circle); black circle, epicenter of earthquake of January 17, 2001. N.B. denotes Newark basin; H.H., Hudson Highlands; M.P., Manhattan Prong.
Figure 3: Record section of vertical-component seismograms from stations in Fig. 2 following collapse of north Tower of WTC. Zero corresponds to computed origin time of 10:28:31 EDT. Data filtered for passband 0.5 to 10 Hz. Three velocities indicated by dotted lines.
Relatively simple and similar pulses with durations of about 5 to 6 s arrive at stations BRNJ, TBR and ARNY starting at a group velocity of 2.0 km/s. The paths to each of those stations from the WTC are mostly in the low-velocity sedimentary rocks of the Newark Basin (N.B. in Fig. 2), the region of low topography west of the Hudson River and southeast of that of higher topography in the Hudson Highlands (Reading Prong). Since those paths cross the boundaries of the Basin at a high angle, the signals at those stations are relatively simple. The signals (not shown) at LSCT, a station in northwestern Connecticut, are also relatively simple, reflecting propagation over a distance of 125 km entirely within the high-velocity rocks of the Manhattan Prong (M.P. in Fig. 2). Their group velocity of about 3.0 km/s is consistent with Rg propagation in that faster, older terrain. Thus, we conclude that the pulse duration at those four stations reflects mainly that the generation of seismic energy from the collapse was delivered over 5-6 s. A portion of the pulse duration probably results from the dispersion of Rg waves.
Anderson and Dorman  observed low group velocities from quarry blasts for paths that propagate mainly though the Newark Basin, and higher velocities for paths within the Manhattan Prong. Their largest arrivals also were the short-period Rayleigh wave Rg. Short-period Rg is well excited only for surface or very shallow sources, which is the case for the WTC. Since Rg propagates mainly in the upper several kilometers of the crust, it is affected strongly by rock properties in that depth range.
Anderson and Dorman also observed strong lateral refraction of Rg waves caused by the contrast in shallow rock properties at the boundary of the high and low velocity rocks of the Manhattan Prong and Newark Basin. Waves propagated to Palisades followed paths through both provinces, resulting in multiple arrivals of Rg. On the basis of polarization analysis, several of those wave packets arrived from quite different directions than those predicted for straight-line propagation. Seismic waves at PAL and MANY also are more complex than those at the other stations of Figure 3, probably indicative of arrivals refracted through the two terrains. At MANY 10s separates two arrivals.
The constructive interference of two Rg phases at PAL may well account for the large arrivals on the E-W component even though the azimuth of the direct path from WTC to PAL is NNE. We do not interpret them necessarily as Love waves; hence, a source with a horizontal component is not required to explain them. (We verified that the components and polarities of the digital data at PAL were correct using recordings of distant earthquakes close in time to the WTC events.).
Comparison with Signals from Earthquakes, Gas Explosion and Mine Collapse
The signals at PAL from Collapse 2 and a small felt earthquake beneath the east side of Manhattan on January 17, 2001 are of comparable amplitude and ML (Fig. 4). The character of the two seismograms, however, is quite different. Clear P and S waves are seen only for the earthquake. The 7-km depth of the earthquake suppressed the excitation of short- period Rg, which is so prominent for the collapse. The difference in the excitation of higher frequencies also can be attributed to the short time duration of slip in small earthquakes compared to the combined source time of several seconds of the complex system of the towers and foundations responding to the impacts and collapses. The waves from the WTC events resemble those recorded by regional stations from the collapse of part of a salt mine in western New York on March 12, 1994 (ML 3.6). That source also lasted longer than that of a small earthquake. A truck bomb at the WTC in 1993, in which approximately 0.5 tons of explosive were detonated, was not detected seismically, even at a station only 16 km away.
Figure 4: Comparison of Palisades seismograms for collapse 2 and earthquake of 17 January 2001. Arrows at left indicate computed origin times.
An explosion at a gasoline tank farm near Newark NJ on January 7, 1983 generated observable P and S waves and short-period Rg waves (ML 3) at PAL. Its Rg is comparable to that for WTC collapse 2. Similar arrivals were seen at station AMNH in Manhattan, which is no longer operating, at a distance of 15 km. AMNH also recorded a prominent seismic arrival at the time expected for an atmospheric acoustic wave. We know of no microbarograph recordings of either that explosion or the events at the WTC. Many people asked us if the arrivals at seismic stations from the WTC events propagated in the atmosphere. We find no evidence of waves arriving at such slow velocities. Instead the seismic waves excited by impacts and collapses at the WTC are short-period surface waves, i.e. seismic waves traveling within the upper few kilometers of the crust.
Significance of Findings for On-Site Conditions
Unfortunately, no seismic recordings of ground motion are currently known to exist at or very close to the WTC. Plans are pending for an Advanced National Seismic System (ANSS; see USGS ) that calls for increased urban seismic instrumentation, including New York City, and the September 11 events show that such instrumentation can be valuable to serve a purpose that sometimes transcends strict earthquake applications. Since the main collapses, a major concern has been if strong shaking affected the structural stability of nearby buildings. Earthquakes of ML 2.3 are not known to cause any structural damage in buildings. In the eastern U.S. that threshold is believed to be close to or above ML 4 to 4.5. It is more reasonable that most of the effects of those collapses on adjacent structures and people were related to the kinetic energy of falling debris and the pressure on buildings exerted by dust- and particle- laden air mobilized by falling debris. It had, except for temperature, an effect very similar to pyroclastic ash flows that descend slopes of volcanoes. The seismic shaking associated with the impacts and the main collapses probably was small compared to those other energetic processes. The following order-of-magnitude estimates of energies involved corroborate this interpretation.
The gravitational potential energy associated with the collapse of each tower is at least 1011 J. The energy propagated as seismic waves for ML 2.3 is about 106 to 107 J. Hence, only a very small portion of the potential energy was converted into seismic waves. Most of the energy went into deformation of buildings and the formation of rubble and dust. The perception of people in the vicinity of the collapses as reported in the media seems to be in full accord with the notion that ground shaking was not a major contributor to the collapse or damage to surrounding buildings. The seismic energy of a ML 0.7 to 0.9 computed for the impacts is a tiny fraction of the kinetic energy of each aircraft, about 2 x 109 J. That associated with the combustion of 50 to 100 tons of fuel in each aircraft is roughly 1012 J, most of which was expended in the large fireballs (visible in TV images) and in subsequent burning that ignited material in each tower. Less than a millionth of the fuel energy was converted to seismic waves.
We thank critical readers W. Menke and C. Scholz. We appreciate comments by Terry Wallace, Mehmed Celebi and John Goff. Lamont-Doherty Cooperative Seismographic Network (LCSN) is part of the Advanced National Seismic System. We thank the many individuals and institutions that collaborate with us in operating the network. LCSN operation is sponsored by the U.S. Geological Survey under Grant Number 01-HQ-AG-0137. Lamont-Doherty Earth Observatory Contribution 6267.
Authors Won-Young Kim, L. R. Sykes, J. H. Armitage, J. K. Xie, K. H. Jacob, P. G. Richards, M. West, F. Waldhauser, J. Armbruster, L. Seeber, W. X. Du and A. Lerner-Lam, Lamont-Doherty Earth Observatory of Columbia University, Palisades, N.Y. 10964, USA; also Dept. Earth and Environmental Sciences, Columbia University.
Anderson, J. and J. Dorman, Local geological effects on short-period Rayleigh waves around
New York City, Bull. Seism. Soc. Am., 63, 1487-1497, 1973. Kim, Won-Young, The ML scale in eastern North America, Bull. Seism. Soc. Am., 88, 935-951,
1998. U.S. Geological Survey, An Assessment of Seismic Monitoring in the United States: Requirement for an Advanced National Seismic System, U.S. Geological Survey Circular 1188, 55 pages, 1999.
The above is an html-version of the following pdf-document:
http://www.ldeo.columbia.edu/LCSN/Eq/20010911_WTC/WTC_LDEO_KIM.pdf (local copy here).
For more see,