he space agency was warned in 1990 that the protective tiles around the shuttle's wheel wells were particularly vulnerable to damage and failure, inviting catastrophe because those tiles protect both fuel tanks and the shuttle's hydraulic system.

The study, conducted by experts at Stanford University and Carnegie Mellon and financed by NASA, also identified ice that builds up on the supercold external fuel tank as a major source of debris that could fall on the tiles and trigger a cascade of failures that could doom the spacecraft.

These two observations fit the leading theory emerging as investigators try to discover what destroyed the shuttle Columbia. If this theory is correct, it would mean that accidents involving two vulnerable areas — the foam and the tiles — combined to destroy the craft.

NASA officials said yesterday that they still believed that the object that hit the underside of the wing on takeoff was foam insulation, but there is growing speculation that it may have been mixed with ice. Video images taken about 80 seconds into the flight show the object to be white or light — the insulation itself is bright orange — fueling speculation that NASA engineers may have seriously underestimated its weight when they concluded that a blow from a block of rigid foam would pose no safety hazard to the orbiter. The insulation is applied as a shaving cream-like foam, but turns hard as a brick.

In another indication that NASA is focusing on the likelihood that foam may have broken loose from an external fuel tank and damaged the spacecraft, a team of investigators spent a second day yesterday examining records concerning the construction of the tank at the agency's Michoud Assembly Facility in New Orleans. Parts of the factory were shut down on Monday and records have been secured by a team of about 20 investigators from NASA.

"The investigative team is on site," said Marion LaNasa, communications director for Lockheed Martin Space Systems, which manufactures the tanks at the site. "They are directing the activity."

NASA officials also confirmed yesterday that the Columbia had sat on its launching pad for 39 days, more than two weeks longer than usual. That means the Columbia was on the pad for 23 days in December, when the Cape Canaveral area received four times the usual amount of rain.

The unusual weather conditions mean the foam insulation around the 15-story external tank was drenched. If that water soaked into insulation or cracks around it, it could have created significant ice when the tank was filled with supercold hydrogen and liquid oxygen before the launching, experts said yesterday.

The study by researchers at Stanford and Carnegie Mellon was particularly significant because it determined that the two wheel areas were especially vulnerable to damage. They were often hit by flying debris, became unusually hot during the shuttle's fiery re-entry through the atmosphere and covered some of the most vital — and volatile — sections of the orbiter. Its helium, oxygen and hydrogen are stored there in pressurized tanks, any of which might explode and rupture the winged spaceship.

If even a single tile were lost in this area, said an update on the 1990 study, a "zipper effect" could occur, stripping away other tiles. The study noted that once several tiles were gone, the resulting heat "could cause a burn through in the aluminum skin of the orbiter during re-entry, exposing and possible crippling some of the critical subsystems and leading to the loss of the vehicle and crew."

Yesterday, one of the authors of the study, Dr. Elisabeth Paté-Cornell, a researcher at Stanford, said NASA headquarters called two days ago to get a copy of its original 1990 study. "I don't blame them," she said of the space agency's inability to find the study. "I'm the same way."

But the research does not appear to be widely understood in NASA. Yesterday afternoon, Maj. Gen. Michael C. Kostelnik, the deputy associate administrator for the space station and the shuttle — who oversees safety issues — was asked whether the tiles around the wheel wells were considered a particular safety issue.

"Not really that I'm aware of," he said. He described the protective tiles, which shed heat as the orbiter re-enters the atmosphere, as "a very robust system."

General Kostelnik joined the agency only last year.

General Kostelnik told reporters today that the debris that hit the shuttle was larger than any seen before. But he said that it was "in the family" of previous stray pieces of insulation and that the agency did not believe there was any reason to be concerned about it.

The two researchers, Dr. Paté-Cornell and Dr. Paul S. Fischbeck of Carnegie Mellon, analyzed the orbiter's 25,000 tiles and identified which were the most likely to fail catastrophically. They did their study for NASA in 1990 and published three academic papers on their work in 1993 and 1994.

Dr. Fischbeck was dismayed about NASA's sudden request for a copy of the study the agency itself had sought and paid for. "That was discouraging," he said.

Even so, the researchers agreed that their warnings in the 1990's had helped NASA improve its maintenance and repair of the shuttle's fragile tiles.

The federal investigation is scrutinizing the left wheel area of the Columbia because it may have been hit by flying debris during the shuttle's climb into space. Moreover, several clues from re-entry, including rising temperatures, suggest that the ship's end may have started there.

The two scientists, who worked closely with NASA managers, engineers and technicians to produce their 1990 study, found the shuttle's two wheel areas were among a half-dozen dangerous tile zones. Others included the nose area and either side of the shuttle's central belly.

Fifteen percent of the shuttle's tiles contributed to 85 percent of the risk of catastrophe, they found.

The scientists began their analysis by studying how debris flying off the shuttle's external fuel tank and booster rockets — including ice and hardened foam — had damaged orbiter tiles during the shuttle's first 33 flights. They found high-, medium- and low-damage zones. Tile damage in the wheel areas was relatively high, as it was along the whole right side of the underbelly. That, they found, was because foam insulation had flown off two long pipes that run alongside the external fuel tank.

With Columbia, NASA officials have said the foam piece that hit the orbiter's wing came dislodged from around the "bipod," an attachment point that goes from the external fuel tank to the shuttle's nose.

That location has long been a source of trouble. One reason may be that when the external fuel tank is filled with its superchilled fuels, its metallic interior shrinks, which can cause its shell of insulating foam to pull away. The area around the bipod is particularly vulnerable, one NASA expert said yesterday, meaning that "it's possible that the foam there can pull away more easily from the tank."

At the moment that the foam or ice hit the wing, the shuttle was 80 seconds into flight. That is a period of maximum stress — the same stresses that contributed to the explosion of the Challenger, 73 seconds into flight, in 1986.

Dr. Paté-Cornell, who is an engineer, said places where things like the bipod penetrate the foam "you have a potential for weakness."

To the data on tile damage, the scientists added temperature readings of where the shuttle tiles became hottest and were most likely to burn through. Dr. Paté-Cornell said the wheel areas got unusually hot during re-entry because their doors created a "discontinuity" in the tiles that would disrupt the smooth flow of air over the shuttle's belly.

"It's probably more turbulent," she said.

NASA's heat map that the team obtained, she added, showed clear heat increases near the wheel doors but did not explain why.

A final ingredient of their study was to correlate the high-damage and high-heat areas with critical parts. The fear, they said, was that flying debris could damage a tile in an unusually hot area, creating a blowtorch aimed at gear below the shuttle's aluminum skin.

Dr. Fischbeck of Carnegie Mellon said that in addition to holding the wheels and landing gear, the wheel wells and adjoining areas held many fuel and oxygen tanks, as well as hydraulic lines that control the shuttle's flight surfaces.

NASA, the researchers say, took the warnings seriously.

"They did a lot of the things we suggested," Dr. Fischbeck said. "Mainly, they inspected the tiles in a different fashion. We said all tiles are not created equal in terms of risk. And so they started to inspect the ones in the most critical areas first."

Dr. Paté-Cornell said one of the 1990 report's recommendations to NASA was to "make sure that the insulation on the external tank is applied in a very secure way."

In their report four years later, the researchers estimated that such improvements could reduce the possibility of shuttle accidents attributable to tile failure by about 70 percent.