Upper West Wall Pulled Inwards

Beginning at about 9.5 seconds before the visible collapse initiation, slight building movement can be detected by using sensitive sub-pixel tracking methods.

These measurements show us a few very important features, or attributes, of the collapse initiation process. This information is new and the NIST did not know about it.

The detection of the earliest tendencies of movement of the building is one application of sub-pixel tracking. We can observe the NW corner get pulled eastward from fl 98 upwards over a 9.5 second interval. At the same time, the base of the antenna is moving eastward and then sags in a "hook" motion. while we see no movement along the west edge of the building.

The Sauret video is shown below. At 9.5 seconds before visible collapse the camera noticably shakes. This is when the slight movement of the antenna and northwest corner begin to be detected.

The Sauret video

The first 20 seconds are the original Sauret video.


1) The camera shakes about 9.5 seconds before collapse.
2) The antenna and the NW corner of the building begin to slowly move during or immediately after the camera shake.

Demonstration of subpixel tracking of a point on the antenna:

9 point tracking used to trace the movement of one round ball on the antenna.

Tracked points:

PROOF OF EARLY MOVEMENT: Two approaches, same results.

First analysis:



Frame 1 in analysis = Frame 370 in the deinterlaced mpg.


1) Deinterlace - unfold.
2) VideoEnhancer Resolution Upscale (*2x, *4y)
3) Deinterlace - fold.
4) Bob Doubler (Alternate Fields, No resize)

Only the first 1300 (interlaced), 2600 (deinterlaced) frames are examined.

Written out as .bmp, download yields 2600 files, totalling 10,782,860,400 bytes (10 Gb)



Camera shake is obvious (between frames 1150-1250).

Black thick line is horizontal movement of the NW corner.

Grey is raw NW corner.

Blue is static point.

(Remember there would be a time delay between event and camera if indeed the source of shaking originated from WTC1)


Two static points are used, one in the foreground (on the building on the left of frame), and one low on the East side of the building. The locations are shown in the following linked images:

The static feature fixed to WTC1...


The foreground static feature...


The following graph compares and shows the difference between both static points...

HiRes http://femr2.ucoz.com/photo/6-0-427-3

Blue is FG static point.

Grey is Building static point.

Black is the difference.


1) There is good correlation between both until near the very end of the trace, indicating that parralax effects are minimal between near and far field objects.

2) There is Westward movement of the static point on the building at the end of the trace.

3) Camera shake period should be obvious.

Static Feature Comparison (Vertical)



1) Slight gradual vertical drop of building static feature following camera shake.

I'm aware of stating movement of static features here, but I was expecting the traced point on the building to *stay* static.

Washer Horizontal Movement (Normalised to FG static point)...

HiRes http://femr2.ucoz.com/photo/6-0-431-3


1) Camera shake is between 1150 and 1250.

2) Eastward movement follows shake.

Static Point Foreground Vertical...

HiRes http://femr2.ucoz.com/photo/6-0-432-3

NW Corner Raw Vertical...

HiRes http://femr2.ucoz.com/photo/6-0-433-3

Draft NW Corner Normalised Vertical...

HiRes http://femr2.ucoz.com/photo/6-0-434-3

Original posting of first analysis at


Second analysis:

The entire Sauret clip is tracked to detect vertical and lateral pre-release movement of WTC1.




1) Several static points in the foreground (3 on the metal stick, 2 on the windows behind stick). These "static points" in the foreground are represented by the blue curves.

2) "Static point" at the 92nd floor NW corner of WTC1 (yellow).

3) Several points near the roof (washer, roof corner, window at 110 NW corner). Roof measurements are represented by the green curves. The curve for the window of the 110th floor is a very bright green and appears almost white.

4) Antenna mast along the black/white transition. The movement of the antenna is represented by the red tones.


The trackers have a problem staying exactly in position during the shaking. Many trackers indeed lost the track and were not able to measure the assigned position during the entire length of the clip. Some trackers stayed "connected" but re-calculated the "best fit" several times during the shaking. Therefore we may have different relative positions of tracked points at the end of the shaking. That deviation of the curves doesn't mean that a real displacement of the measured points occurred. Instead we can use the new relative positions as "zero movement" if we are not able to track the movement during the shaking precisely.




Prior to the shaking of the camera all curves follow the blue "static foreground". There is no measurable movement of the tower. All apparent movements are the result of the shaking of the camera. After the shaking the yellow curve (floor 92) stays with the blue curves (static foreground). This means the 92nd floor didn't change it's position relative to foreground static points until the 92nd floor was pushed westward during the collapse.

Interestingly, all measured points above 92 - roof, washer (green) antenna (red) - started to lean east immediately after the shaking (about frame 1350).



Prior to the shaking of the camera all curves follow the blue "static foreground". After the shaking all curves vary somewhat but move with the blue curves for about the next 200 frames. At about frame 1465 the antenna mast clearly started to "sag" while roof (green) and 92nd floor (yellow) stayed with the static foreground (blue).

We will have to compare the result with the calculated relations for the south tilt before we can differentiate between tilt and drop. Nevertheless, prior to any sag/tilt the entire upper part of the buildings started to creep eastward.

The east leaning (wide side of the core) is hardly explainable as induced by the south wall inward bowing if we do not measure any increasing south tilt during this interval. The same object tracking tool is used to measure south tilt as well as eastward tilt. Any south tilt would significantly shorten the measured vertical distance between roofline and any tracked point on the antenna.

Perhaps we wouldn't notice a small trapezoidal perimeter deformation towards the southeast if the antenna remained straight up. In that case the perimeter columns would bow towards the southeast, yet the total circumference of thr roofline along the perimeter must remain the same and we do not see a corresponding movement of perimeter roofline columns extending from the SW corner to the NE corner. Therefore such a hypothetical SW perimeter fold-in as the antenna remains near plumb does not match the visual record and so can be excluded as a possibility.

Once again everything points to a core-led collapse, not to a collapse initiated by instability in the south perimeter..

Interesting to note that prior to the collapse the distance between roof and 92nd floor decreases as seen in the HiRes plot of vertical displacement. After the collapse of the 98th floor the 92nd floor was sagging (compared to the "blue" static points in the foreground) until it was destroyed when the collapse reached that floor.

Frame 1641 of that long enhanced video is the frame 0 of the older set of measurements by achimspok.

These ideas were originally posted here:

The NW corner pull-in movement starts to become visible to the naked eye a second or two before the visible collapse initiates, as shown in these gifs:

Earliest detectable deformation of roof is progressively concave

The feature titled Antenna base shifts eastward 9.5s before collapse shows us that the earliest detectable antenna movement leads into visible collapse initiation as a eastward ad downward hook motion, pulling in the upper west wall as it shifts east.

We also can see both the 104th fl fire along the SW corner remain stationary as the antenna moves and the guy wire connecting the antenna to the middle of the E face lose tension, meaning the antenna is sinking into the perimeter roofline, in images linked at the bottom of this page.

60 frames per second means this covers close to two seconds leading into the more visible initiation sequence. Notice that the fire on the SW corner, west face, floor 104 shows no significant movement during this time.

High resolution version here


First, locate the release event. In some frame the velocity will begin to take off, meaning the slope of the velocity plot changes quickly. This is because there is an abrupt change in acceleration. This frame can be called the "release event" or "release moment".

Second, separate the curve into 3 regions: A pre-release region, a post-release region and the region of the release event.

Pre-release motion could be drift, creep. deformation, tilting, vibration or any slower movement which lacks a downward acceleration at a significant percent of g. A careful researcher should trace points in the pre-release region as far back as necessary to discover the earliest motion possible.

Post-release motion has an average downward acceleration of 0.5g to 1.0g. It may have moments of velocity reductions or may not. If so, the momentary downward acceleration between reductions may differ from the average acceleration and should be considered separately. Locations and magnitudes of the velocity reductions may provide valuable information about what is really happening inside the building over the first 12 feet of falling.

The release event Using the velocity graph, curvature at the release point may help determine how quickly the initial failure occurred.


Two examples

The Sauret drop curve of NW corner is shown in blue. It's corresponding velocity curve is shown in purple. We see there is an abrupt change in the slope of the velocity curve that can be traced back to frame 222. This is the release event for the NW corner.

The acceleration (slope of the velocity curve) quickly changes through frame 222. The traced point is now falling at 0.5g to 1.0g. There is one measured velocity reduction around frame 250. In frame 250 the positional data shows the NW corner has fallen about 3 ft.

In the pre-release region we can carefully study changes in the positional data as far ahead of frame 222 as we wish, looking for the earliest detectable deformations.

The Sauret drop curve of the black-white transition point on the antenna is in yellow. It's corresponding velocity curve is in light green. We see movement from frame 140, yet the velocity curve does not take off with a 0.5g to 1.0g acceleration until frame 215. The release event is around frame 215.

In the pre-release region we see considerable movement before frame 215. Over 2 feet of downward displacement is measure in the antenna between frames 130 and 215.

In the post-release region we detect one velocity reduction. How far has the traced point dropped when the reduction occurs? It happens around frame 228, when the positional data shows a 3 ft drop.