One of Georgia's largest chemical plants is under scrutiny by state regulators over its acknowledged under-reporting of millions of pounds of toxic releases dating back to at least 2003.

DSM Chemicals North America, based in Augusta, uses large amounts of cyclohexane in the manufacture of caprolactam -- the raw material for nylon.

According to documents obtained through Georgia's Open Records Act, the company reported to regulators that it released 1,178,713 pounds of the flammable vapor from 2003 to 2006, when it actually released 10,750,923 pounds -- accounting for more than a third of the cyclohexane released by all industries in the country during that time.

The chemical can cause dizziness, nausea, dry skin and other ailments and is categorized as a volatile organic compound that contributes to smog, ground-level ozone and respiratory problems, according to the U.S. Occupational Safety & Health Administration.

Experts are unsure how the releases would have affected air quality or human health.

"It's hard to say without some engineering assistance to say what the stack emissions equate to with some sort of modeling," said Randy Manning, Georgia's environmental toxicology coordinator. "We don't have an estimate of what their emissions equated to in ambient air concentrations."

In general, the chemical is only mildly irritating to the eyes and mucous membranes and skin, and lingers in the atmosphere for only a few days.

"When you look at toxicity data, it's not very toxic," Mr. Manning said. "But can we jump from there to say millions of pounds aren't a concern? I can't make that jump."

Georgia's Environmental Protection Division raised questions about the Columbia Nitrogen Drive plant's emissions in a May 15, 2007, notice of violation issued after inspectors evaluated performance test results provided by DSM.

In the notice, EPD asked "if emissions in prior years were under-reported, if there were excessive emissions and if previous compliance certificates were accurate."

In a June 15, 2007, response, company officials said they thought a flare on the factory's smokestack was incinerating at least 95 percent of the excess cyclohexane, but it wasn't -- and the problem apparently escaped detection for several years.

As far back as 2003, the company noticed a decrease in production efficiency and tried to find the source of cyclohexane losses, wrote Beth Connell, DSM's safety, health and environmental manager, in the company's response.

Engineers eventually focused on a flare high above the factory where chemical vapors are channeled as part of a safety program to avoid explosions.

On July 5, 2007 -- three weeks after receiving EPD's notice of violation -- DSM filed amended emissions reports with the U.S. Environmental Protection Agency's Toxic Release Inventory, a public database. The reports acknowledged under-reporting both emissions of cyclohexane and a related chemical, cyclohexanol.

"These revisions are the result of DSM's recent discovery of inadvertent data quality errors that led to the recalculation of cyclohexane and cyclohexanol emissions from two process units," Ms. Connell wrote.

According to the revised forms: - During 2003, DSM reported cyclohexane emissions of 29.98 pounds. The actual emissions were 4,948,651 pounds. The same year, DSM reported cyclohexanol emissions of 15,210 pounds. Actual emissions were 119,629 pounds.

- During 2004, DSM reported cyclohexane emissions of 566 pounds. Actual emissions were 2,098,435 pounds. Cyclohexanol emissions, reported at 15,274 pounds, were actually 36,526 pounds.

- During 2005, DSM reported cyclohexane emissions of 214 pounds. The actual emissions were 1,978,159 pounds. Cyclohexanol emissions, reported at 15,276 pounds, were actually 36,703 pounds.

- During 2006, DSM reported cyclohexane emissions of 1,177,903 pounds. Actual emissions were 1,725,678 pounds. Cyclohexanol emissions, reported at 102,034 pounds, were adjusted down that year to 60,644 pounds.

Last fall, the company invested more than $1 million in a catalytic oxidizer and other equipment to ensure the destruction of chemical vapors, Ms. Connell said.

Because the stack through which the 2003-06 emissions escaped empties into the air high above the plant, the vapors didn't pose an explosion risk, Ms. Connell said.

"It was so diluted that the emissions would not burn," she said. "There was not enough material to ignite."

Accidents involving the chemical are rare, but they have occurred: In 1974, leaking cyclohexane at a plant in Flixborough, England, ignited and the blast killed 28 workers.

Currently, DSM is in compliance with its operating permits and the new equipment installed to correct the problems has been tested, said Lewis Hays, an environmental compliance engineer in EPD's Stationary Source Compliance Program.

"At this point we have no reason to believe their tests are in error because they used an outside testing firm. Those tests indicated they were destroying at least 98 percent of the (volatile organic compound) emissions," Mr. Hays said.

State regulators say the company will face enforcement actions and might have to pay fines associated with violations related to the cyclohexane emissions.

"We are evaluating all that data and trying to determine which rules were violated, and we're still getting additional information from them," Mr. Hays said.

The usual process involves a consent order negotiated between EPD and a violator. Such orders typically can include monetary penalties and stipulations for compliance. Once such an order is executed, it becomes public.

Lisa Kirby, DSM's vice president for administration, said the excess emissions and pending regulatory actions have been disclosed to DSM's parent company and its stockholders.

With last fall's addition of the new equipment, which operated three months that year, the plant's 2007 cyclohexane emissions fell to 929,155 pounds -- from 1,725,678 pounds in 2006 -- and a much more significant decline is expected for 2008, when the new equipment will have been in operation for a full year, she said.

Dr. Michael Chang, a professor at Georgia Tech's Center for Urban and Regional Ecology who spent four years studying Georgia's air pollution trends, said the emissions probably wouldn't have worsened ozone levels in Augusta. "There's already a good supply of VOCs (volatile organic compounds) available from other sources, so it would probably be relatively marginal," he said.