There’s a yawning gap between the U.S. Environmental Protection Agency (EPA) estimate for climate-busting methane emissions in America’s industrial sector and real-world levels of leakage, according to new analysis by Cornell University and the Environmental Defense Fund.

While methane levels have been assessed at extraction points (well pads), power plants, and downstream along municipal pipelines, EDF reports in a recent blog post, there is a lack of hard data on fugitive emissions from companies that use natural gas as a fuel, or as an ingredient in industrial processes. The omission matters, EDF lead senior scientist Joseph Rudek, since methane carries 84 times the global warming potential of carbon dioxide over a 20-year span.

To begin to address the gap, an EDF/Cornell research team focused on the fertilizer industry—a good choice, explains EDF, as the few dozen ammonia fertilizer plants in operation in the United States are often found close to public roads.

Taking advantage of that proximity, the researchers mounted a high-precision methane sensor on a Google Street View car and traversed roads near six representative fertilizer plants in the midwestern U.S., hunting for telltale methane plumes resulting from “incomplete chemical reactions during fertilizer production, incomplete fuel combustion, or leaks.”

When it detected a methane plume, writes EDF, the Google car lapped the facility dozens of times to fully measure the extent of the leak. The researchers discovered that “on average, 0.34% of the gas used in the [six] plants is emitted to the atmosphere,” a measure which, when “scaled to the entire industry, suggests total annual methane emissions of 28 gigagrams—100 times higher than the fertilizer industry’s self-reported estimate of 0.2 gigagrams per year.”

EDF adds that “this figure far exceeds the EPA’s estimate that all industrial processes in the United States produce only 8.0 gigagrams of methane emissions per year.”

In addition to revealing the “huge gap between a priori estimates and real-world measurements,” writes EDF, the study “demonstrates that mobile sensing is an economical way to pinpoint significant emissions sources and enable rapid and efficient mitigation—vital for rapidly reducing the rate of warming.” Broad deployment of sensors on public vehicles like school buses and postal trucks would vastly expand capacity “to characterize sources and rates of pollution—and perhaps influence policy-makers.”

Mobile sensing is also inherently egalitarian, noted Cornell civil and environmental engineer John Albertson. “Pollution in the air doesn’t respect property boundaries, so even if you don’t have access to private land, the current revolution in sensor technologies allows us a lens into the degree of cleanliness of a factory,” he said. “With opportunistic sensing, pushing data to the cloud, doing the proper analysis, and drawing inferences, we can build environmental policy that is based on evidence.”