The report details results from air sampling done last September and October at the Radford plant above an open field where piles of waste from the manufacture of weapons explosives are set afire daily.

The plumes drift directly towards an elementary school and residents a little more than a mile away, but the Army and regulators have long maintained that the pollution level is safe, based on its computer-modeled estimates.

Now, it turns out, some of those estimates were wrong.

The data shows that five substances were found at levels greater than the EPA’s models had predicted, meaning that previous health-risk analyses completed by regulators for the burns at Radford did not fully take into account the potential exposure of the surrounding population.

Arsenic, a chemical element known to cause cancer and skin lesions, was found to be emitted at rates 37 times what the previous Radford burn permit estimated. Lead — which can disrupt children’s brain development — was emitted at five times the level previously thought. Cadmium and silver were also present at levels higher than historical models had assumed.

The tests also detected levels of methyl chloride, a chemical used in refrigeration and manufacturing that is known to cause severe neurological effects, high heart rates and high blood pressure, at more than twice the levels previously thought.

The Radford tests are part of a national program to verify the mathematical emissions factors that are used in the permitting dozens of sites that burn similar materials, and so the findings could mean that pollution from the practice of open burning military waste has been underestimated across the country.

The EPA declined to make the researcher responsible for the data available for an interview, but sent a statement pointing out that the research was still in draft form and under active review. ProPublica was given the draft report by a person concerned about the health implications at Radford and other military sites where such burns are conducted.

The Army did not immediately respond to a request for comment.

ProPublica reported on the burn practices at Radford in July as part of an investigation into the environmental implications of the Department of Defense’s handling of munitions and the hazardous waste associated with them. Our investigation showed that the Defense Department and its contractors continue to burn explosives waste with no emissions controls at more than 50 sites across the country, utilizing a loophole in US hazardous waste regulations created on what was supposed to be a temporary basis for the Pentagon in the 1980s.

The burns are allowed by the EPA and state environment regulators based on permits that use computer models to estimate the rate of pollutant released and to calculate whether those emissions are going to make people sick. The burns at Radford, the single largest polluter in Virginia, have long been permitted based on analyses that showed they were safe.

But the actual emissions from the burns at Radford had never been previously measured. For the new research, the EPA, in cooperation with NASA and the Department of Defense, sampled the smoke plumes at Radford last year, gathering the first air samples of toxins released at the site since burning there began in the 1940s.

The report details chemicals measured by a drone flown through the smoke clouds directly above the burn site over the course of two weeks last fall, and provides the first ever confirmation that significant levels of volatile organic chemicals, including acetone, benzene and toluene — all substances known to cause cancer — are widely prevalent.

EPA officials stressed the good news in their draft findings: that the majority of pollutants measured, including aluminum and selenium, dioxin-like chemicals and volatile chemicals like benzene, were detected at levels less than what the computer models used for historical permits had estimated. The agency described the drone technology used to measure the plumes as “a significant advancement.”

While the research is the first of its kind to take direct measurements of the pollutant plume at the burn grounds in Radford, it still does not attempt to measure exposure to those pollutants in the surrounding community, something that state regulators tell ProPublica would be accomplished by placing ambient air monitors at schools and other public places near the burn site. People living near the plant have unusually high rates of cancer, thyroid disease and other health problems and have raised questions about a link to open burns, but so far there’s little evidence to prove or disprove this.

Perchlorate, a rocket fuel which has been recently detected in groundwater samples taken in the community surrounding the Radford plant, and which the Radford plant’s former commander had told the Roanoke Times was detected in the new round of samples, is not listed as a detected pollutant in the draft report.

In July, Virginia environment regulators, who are charged with overseeing the plants’ day-to-day operations, told ProPublica they were awaiting the sampling results to make a new risk assessment for the burn site as part of the process of renewing the Army’s burn permit there, which expired in 2015. Virginia officials told ProPublica that they are evaluating the information and “have not reached any conclusions.”