Engineers dissecting telemetry from the shuttle Columbia and videotape of its fiery re-entry now believe eyewitness accounts of debris falling away from the spacecraft as it passed above California, well before its ultimate breakup high above Texas. At the same time, investigators believe the breach that let hot gas eat its way into the shuttle's left wing probably was located at or near the leading edge of the wing or perhaps at or near a seal in the left main landing gear door.

While no options have been ruled out, many NASA insiders favor the leading edge theory as the best explanation of the telemetry released to date. Wherever the breach occurred, a member of the independent Columbia Accident Investigation Board said today it must have been fairly large to explain the resulting catastrophe.

"When you're flying in this superheated air area, what's forming around the spacecraft itself is actually called a shock front," said board member James Hallock. "It's creating a boundary layer so it's not that you have a little tiny pinhole, you need something bigger than just a pinhole for something to actually be able to penetrate within that area.

"But once it gets in there, then it starts off as being a very small thing and then it would propagate. And it becomes much like, I guess, an arc welder's torch."

While the media will no doubt focus on that description, Hallock agreed later that calling a plasma intrusion a 'jet' or "torch" was misleading. An aerodynamicist with expertise in re-entry physics said in an email "there is no such thing as 'plasma jet' inside the wing or inside the landing gear well."

"Once the plasma in the shock layer of the external flow enters into the wing structure passages, the electrons and ions recombine at very high rate due to the high pressure (larger than 1 pound per square foot), and to the catalytic effect of the walls," he wrote. "The air inside is just at low pressure and very hot."

Hot enough to eat through sensor wiring, raise temperatures in the shuttle's left main landing gear wheel well and, ultimately, to affect the aerodynamic characteristics and structural integrity of the left wing itself. Hallock said the accident board has not ruled out anything in its search for what caused the fatal breach. Among the more likely suspects are impact by high-velocity space debris; impact by a micrometeoroid; damage caused by foam insulation from Columbia's external tank that fell off during launch and struck the underside of the left wing; or some combination of these and other factors.

On another front, engineers are slowly but surely coaxing additional data from the telemetry beamed down from Columbia just before it broke apart.

In the days following the disaster, shuttle program manager Ronald Dittemore said ground systems recorded sporadic telemetry from Columbia for up to 32 seconds after commander Rick Husband's final transmission at 8:59:32 a.m. Additional analysis has now extended the time Columbia was known to be intact by five seconds or so beyond Husband's final call, showing two more right-firing rocket thrusters were commanded to fire in a futile bid to prevent aerodynamic drag on the left wing from pulling the spacecraft into a fatal "sideslip" orientation.

In addition, sources say a "roll reference message" generated by Columbia's flight computers at some point midway through that 32-second period also was seen in the telemetry. The message got as far as a computer buffer but it never reached the crew's cockpit displays. By that point, the astronauts undoubtedly knew something was amiss.

While he did not confirm the roll reference message, Hallock said "we continue to dig data out of that 32-second period of time."

Columbia was in a left bank at the time, descending at Mach 18 over northwest Texas at an altitude of more than 200,000 feet. While the exact timing is still unclear, time-stamped video from an Apache helicopter flying near Fort Hood, Texas, shows multiple contrails by 9:00:30 a.m., indicating Columbia broke completely apart less than a minute after the last valid data frame was downlinked.

Already released telemetry from the shuttle shows whatever went wrong went wrong early in the descent, with unusual temperature increases beginning just a minute and 24 seconds after the shuttle reached the region of maximum heating while approaching the coast of California.

While this page has generally avoided the "competitive speculation" fueling many media accounts of the tragedy, a fair number of agency insiders now suspect that a breach at or near the leading edge of the left wing in front of the wheel well is a reasonable explanation for the telemetry released to date.

Telemetry shows the first clear signs of elevated temperatures in the left main landing gear wheel well while Columbia was still off the coast of California. As the shuttle's descent proceeded, additional wheel well sensors either failed or recorded high temperatures while other skin temperature and hydraulic sensors positioned near the back of the wing simply failed. Wiring from those sensors was routed around the left side of the landing gear wheel well and then directly in front of it before entering Columbia's fuselage. A plasma intrusion near the wheel well could have affected the rear-wing sensor wiring as well as the sensors in the unpressurized wheel well.

"When you look at how the wires are set up within the craft itself, it leads you directly towards the wheel well, there's no question," Hallock said. "The wires that lead to those sensors that we're talking about go right in front of the wheel well. Then when you look where the temperature sensors are and you get a sense of how they're reading, you can almost triangulate (to where the problem is). That's what we're trying to do."

NASA engineers familiar with the shuttle's construçtion say a plasma intrusion directly into the wheel well, perhaps from a breach in a landing gear door seal, would have wreaked more immediate havoc than the relatively modest temperature increases seen in Columbia's telemetry. But a wheel well breach has not yet been ruled out.

Video shot by amateurs along Columbia's path shows what appear to be pieces of debris breaking away from the shuttle well before Husband's final transmission. The accident investigation team has asked the public for help locating any such debris, which could include reinforced carbon carbon leading edge panels or heat shield tiles. Investigators also are seeking FAA and military radar tapes and checking into widespread reports of sonic booms over California and Nevada. Such booms would be consistent with the passage of high-velocity debris.

Outside experts have been asked to evaluate the trajectory of any debris that might have broken free of Columbia over California and the southwest to help search teams narrow the potential debris "footprint." So far, no such debris has been located.

But there is little doubt that Columbia was shedding some sort of debris well before its breakup. Numerous witnesses reported seeing what appeared to be debris falling away from the orbiter or hearing muffled rumbling that may have been caused by torn-off shuttle components falling along the spacecraft's ground track.

"We have been pouring over the films," said Hallock. "From the timing I've seen right now, it does look like things were beginning to come from the shuttle as it approached, right about California. So it's not something that was close in, it began way out in that area.

"People are looking at the pieces (on film), we're trying to get a sense by looking at them where we have these films and trying to extract from them what we think the mass would be and then how it's moving. Some people here at (the Johnson Space Center) are trying to extrapolate that to the point of where it might land so we can pick some places and try to find it. Obviously, it would be very important to understand what those pieces are, particularly those that started falling off at the very beginning."

Near town of Twain Hart, Calif., 4,000 feet above sea level in the foothills of the Sierra Mountains, Glen Caldwell and his daughter got up early to watch Columbia fly overhead. A veteran of two previous shuttle re-entries, Caldwell knew what to expect and told his grown daughter the view would be worth the early morning effort.

The weather was excellent and "it's right on time, coming out of the west, and when it gets pretty darn high in the sky for us, we see a little, like a fireball, like a Roman candle or a fire come off this thing," Caldwell, an insurance broker, said today in a telephone interview. "Then a second later or so, we see it again. I commented at that time to my daughter, 'it looks like that thing is falling apart.'"

Walking back to his house, Caldwell watched the remainder of Columbia's entry on NASA's satellite television channel, which he was recording. All too soon, the projected ground track of the shuttle, visible on a projection TV screen in mission control, stopped updating over Texas.

"I see the ground track stop and I'm getting a little bit concerned and starting to put two and two together," Caldwell said. He later calculated that Columbia's ground track must have passed about 15 miles north of his location, giving him a good, high-elevation view of the ship's fiery path across the sky. For scale, he said imagine holding an almond at arm's length.

Relative to that almond, "what we saw come off the side of this thing were two pieces the size of a BB," he said. "The orbiter is very, very bright, very iridescent when it comes in and these two pieces when they came off were the same color and intensity. They were very, very bright, very easily discernible and very clearly separate from the orbiter. From our perspective, they disappeared very quickly into the (shuttle's plasma) trail and disappeared."

Unlike video of Columbia's debris falling over Texas, in which different pieces of wreckage fell together, "what I saw was kind of the opposite," Caldwell said. "They peeled off and the orbiter raced away from them."

Mike Dambacher, who was taking a morning walk Feb. 1 and saw the shuttle pass over by accident, did not see any signs of debris falling away. But he heard a rumbling that might have been sonic booms triggered by the passage of unseen debris.

"I'm watching this ball of flame going over," he said in a telephone interview. "I'm watching for probably 15 seconds. My first thought was, I'm watching a falling star, except it's not falling."

A few moments later, he heard a boom and "I'm thinking is that a sonic boom? We hear sonic booms occasionally because they do test flights over here. This was a very muffled sound. There as such a delay, I couldn't tie that to what I saw fly over."

Watching news reports in his home a few minutes later, Dambacher realized he had just seen the shuttle fly over. Comparing notes later with other witnesses, he said some people heard two booms, others heard one.

"The sonic boom is not heard when something goes subsonic!" the re-entry expert said in another email. "Every time a supersonic body flies over an ear at ground level, its bow shock will excite the ear with what we call a 'sonic boom.' The sonic boom is influenced by atmospheric effects, body size, rigidity, acceleration, etc. They all contribute to what we call the 'signature of the sonic boom.'"

Large craft like the space shuttle and the Concorde produce double sonic booms caused by bow shock and wake shock.

"A cluster of supersonic bodies will produce a very complicated sonic boom. Anything can happen, but it will sound more like a long thunder, like a rumble, as the cluster of bodies continues to break down. The shock waves coalesce and form a complicated acoustic pulse."

Small bodies a few inches across - including shuttle tiles - would not produce sonic booms," he said. And while higher Mach numbers yield stronger booms, higher altitudes tend to weaken the boom.

"You will not hear a sonic boom from one tile," he said. "You may from a cluster of tiles (about 20) if they go off simultaneously. But you will hear an alteration of the vehicle's 'nominal' signature if a non-structural member starts to flutter."

For example, one of the insulation blankets on the side of the shuttle.

"The human ear is very good at signature analysis of broad band noise," the physicist continued. "The abnormal noise people heard could have been related to the shedding off of (an insulation blanket)."