Cutting greenhouse-gas emissions is challenging enough. But what if we can't even measure the pollution we're trying to reduce?

Euan Nisbet, a University of London professor who has travelled the world testing the air for greenhouse-gas pollution, makes his way to a rocky outcrop on the eastern coast of Hong Kong Island on a sunny November afternoon.

He takes out a battery-operated pump connected to a thin tube and a plastic bag to capture traces of the wind.

"This is a good day for collecting samples," he says. "There's a good, strong breeze blowing in from the mainland. It's the breath of China."

Hooking up his air-sucking device, Nisbet says the world puts too much faith in government estimates of carbon dioxide, methane and other heat-trapping gases blamed for climate change. That's because companies and countries base emissions calculations on the raw materials that go into a factory or power plant - they don't check the pollution that comes out.

"It's like going on a diet without weighing yourself," explains Ray Weiss, a geochemistry professor at the Scripps Institution of Oceanography in California, whose article with Nisbet in the June issue of Science argues for measuring the atmosphere. "We're counting all the calories, believing the bottom-up calculations are right and not making any effort to test by getting on the scale."

As the world turns to climate treaties and emissions-trading markets to tame global warming, scientists and regulators are clashing over a key question: how do we measure the pollution we're trying to reduce?

Companies report their emissions estimates based on inputs: how much coal a plant burns; how much oil a factory consumes; how much lime is added to cement. Countries tabulate these estimates and add nationwide figures: how many vehicles drive within their borders; how much waste goes into landfills; how many methane-belching sheep graze in pastures.

Nisbet, Weiss and dozens of researchers say that bottom-up approach doesn't reveal what we really need to know - what's happening in the air. They're sounding an alarm that greenhouse gases measured in the atmosphere can be double what companies and nations estimate on the ground.

In the case of one heat-trapping gas called sulphur hexafluoride (SF6), countries report that output has plummeted since 1995 based on their counts of military radar systems, electrical equipment and factories that use it to soundproof windows, all of which can release the gas. Scientists say measuring the air shows levels have surged since at least 2000. The US Environmental Protection Agency (EPA) says getting the measurement right is crucial because SF6 traps 23,900 times more heat per molecule than carbon dioxide and lingers for up to 3200 years.

"These gases lead to climate change for centuries or millennia, so there's a finite amount you can ever put in the atmosphere," says Stephen Pacala, director of Princeton University's environmental institute. "For every release, you're taking something from some subsequent person."

Making sure the numbers add up is important for a more immediate reason: they underpin the only international climate treaty that sets mandatory limits for greenhouse gases.

That 192-country pact, the Kyoto Protocol, came into effect in 2005 and set emissions targets for 38 of the countries that signed on.

Today US$141billion ($188billion) worth of credits that help countries meet their Kyoto goals are changing hands in global emissions markets. And business is booming in offsets - the right for countries and companies that still pollute too much to claim credit for green projects elsewhere. All those efforts are based on bottom-up calculations being accurate.

"We have to get emissions numbers right early on," says Shakeb Afsah, a former World Bank environmental economist who founded data-analysis company Performeks LLC.

Traders and regulators say Europe's carbon market, called the Emissions Trading System, and the United Nations' offset market, called the Clean Development Mechanism, are making a real dent in greenhouse emissions. Bloomberg New Energy Finance says from US$570 million a year in 2004, the global carbon market could surge to US$1.4trillion worth of transactions by the end of the decade. Trade in carbon options contracts alone soared to almost US$10.6billion last year.

The 15 countries who were EU members at the time the pact was adopted in Kyoto, Japan, say the system has already helped them: they've surpassed their goal of trimming emissions in the five years to 2012 by an average of 8 per cent compared with 1990 levels. As of 2008, the now 27-nation EU reported to the UN that emissions had dropped 11.3 per cent below 1990 amounts. The EU has set internal targets in addition to Kyoto goals.

"It is working," says James Cameron, vice chairman of Climate Change Capital, a London fund manager that has invested more than US$1billion in carbon credits. "This system is only there to take tonnes of carbon out of the atmosphere. It has no other purpose," says Cameron, who won't say how much the fund has profited. "It has propelled some very, very good investments." Bloomberg data show wind farms in China are among them.

Jean-Pascal van Ypersele, vice-chairman of the Intergovernmental Panel on Climate Change, a UN-led network of scientists, says companies and countries use tried-and-true formulas to calculate emissions.

Carbon dioxide is the main greenhouse gas traded in the EU market and burning coal, oil and other fuels accounts for most of the emissions from human activities. CO2 is relatively simple to calculate, he says. Auditors and expert panels review and sign off on the numbers companies and countries report. "Emissions from burning fossil fuels are quite easy to monitor and quantify because burning 1 kilogram of carbon is going to provide 3.7 kilograms of CO2 - that's the law of chemistry," van Ypersele says. "As long as you know how much carbon is present in the fossil fuel you burn, you automatically know the amount of CO2."

If there are discrepancies between emissions-reduction figures, it may be because countries that calculate them don't all follow one set of rules, van Ypersele says.

The Kyoto Protocol requires just 38 developed countries of the 192 in the pact to submit emissions numbers using stringent guidelines. The US has refused to ratify Kyoto. And fast- growing polluters such as China and India can use looser calculations that aren't audited, the treaty says.

"That's not a technical problem," van Ypersele says. "It's a political problem."

India, the second-biggest earner of UN offsets, claims more than 550 green projects - from methane-trapping landfills to bricks made without coal-fired kilns. The UN says those projects have eliminated the equivalent of 80 million tonnes of CO2 from the air. India isn't required to verify those calculations: nobody takes a measurement device to a plant or assesses the air across polluted cities to say CO2 levels have dropped.

Matthias Jonas, a physicist at Austria's International Institute for Applied Systems Analysis, has studied emissions numbers that do adhere to UN guidelines. He says he's sceptical about the gains European countries are claiming. Estimates based on fuel consumption, forest cover and other bottom-up factors are riddled with inaccuracies, and the margins of error are too wide to be useful, Jonas says. "All the emissions we're accounting for so far under the Kyoto Protocol are based on what we think the atmosphere sees by standing on the ground," Jonas says. "Real verification would be with a measurement device sitting in the atmosphere and saying, yes, what you have estimated is true and we can confirm by measuring what the atmosphere has received."

Pieter Tans, a Boulder, Colorado-based senior scientist with the National Oceanic and Atmospheric Administration, says with billions of dollars riding on markets that are literally based on thin air, emissions trading has to be especially transparent.

A tonne of CO2 was this week trading at €11.80 ($21) on London's European Climate Exchange. "As soon as emissions become worth a lot of money, I start losing faith in self-reported numbers regardless of who signs off on them," Tans says. "We need something more objective - like checking what's actually appearing in the atmosphere. That's what really counts and requires a serious enhancement of the present capabilities."

Bart Chilton, a commissioner at the US Commodity Futures Trading Commission, says if the US creates a nationwide cap-and-trade market, carbon could become the world's most widely traded commodity.

Ethan Zindler, head of clean-energy policy at Bloomberg New Energy Finance, says results of the November midterm elections may have killed federal cap-and-trade for at least two years. Still, 10 states, including New York, New Jersey and Massachusetts, run a mandatory cap-and-trade system to cut CO2 emissions from the power industry. California has proposed statewide cap and trade.

Joseph Mason, a banking professor at Louisiana State University who specialises in financial crises, says carbon is unique because it isn't backed by a commodity such as gold or oil that can be seen and touched.

Carbon credits get their value from people having faith in them, he says. "We really are setting up a market from scratch here for this hybridised pseudo-commodity of our own making." Anything that raises doubts about the integrity of emissions reductions could potentially damage the market, he says.

Gregg Marland, a staff scientist at the US Department of Energy's Oak Ridge National Laboratory, says, "If you're selling oil, you're actually transferring something tangible. If somebody lies, somebody loses," he says. "In a CO2 transaction, you can lie and both win." The seller gets paid, and the buyer has his credit for compliance. "We're going to create a situation where both sides can win by cheating," Marland says.

United States success in reducing acid rain shows how tracking what comes out of a factory - as well as what goes in - can pay off. The EPA began requiring companies to continuously measure the sulphur dioxide and nitrogen oxides coming from their smokestacks in 1995. They reported as often as every hour in the world's first large-scale emissions-trading effort. By 2006 the US had cut sulphur dioxide emissions by 40 per cent and nitrogen oxides by almost 50 per cent, a 2007 EPA assessment found.

One recent carbon study shows how far off estimates can be. Afsah's Performeks compared emissions calculated by the Netherlands Environmental Assessment Agency with those calculated by BP. The difference between the two sets of estimates for 23 nations added up to 2 billion tonnes of CO2 - more than the amount that Russia, the world's third-largest emitter, spewed from burning fossil fuels in 2008.

The gap shows how widely results can vary depending on how calculations are tweaked and what statistics go into the number crunching, Afsah says.

Even an early supporter of bottom-up calculations says he now sees the drawbacks.

"When it comes down to it, these estimates are all guesses," says John Bosch, who retired from the EPA in 2009 after 38 years.

Bosch says he left to become a consultant because he was frustrated that regulators didn't require precise measurements. New laser-based instruments make policing emissions possible, he says. "In the real world, there are huge motivations for everyone to low-ball emissions," he says. Regulators wanted to report progress, and polluters wanted to pay less for permits. "Everybody is in the game."

Scientists say studies published this year support their call for better verification of bottom-up figures. Measurements of methane, the most common greenhouse gas after CO2, gathered in Los Angeles with a laser-based analyser were a third greater than estimates the California Air Resources Board calculated using UN guidelines, according to the January 2010 journal Atmospheric Environment. Bottom-up calculations may underestimate natural-gas pipeline leaks and miss other sources entirely, the study found.

Researchers who looked at perfluorocarbons in the air around China, Japan, the Koreas and Taiwan found that levels of the gas may be 97 per cent higher than calculations from bottom-up approaches.

Mark Lewis, managing director of global carbon research at Deutsche Bank, says traders don't have the expertise to second-guess how emissions markets have been designed. They have to believe experts have set up a system that achieves real pollution cuts. "I assume the scientists are doing their job correctly and the market can then get on and do its job, which is to deliver emissions reductions at the least cost," Lewis says.

"There's no reason why you need to understand the science or even believe in it to trade in the market."

The EU and UN markets are audited by experts who inspect logbooks, cross-check production data against fuel invoices and interview workers.

Auditing doesn't guarantee accuracy, Afsah says. When he worked at the World Bank, Afsah examined environmental data at thousands of factories. He found air measurements always deviated from bottom-up estimates - no matter how rigorously they were audited.

"Every auditor audited Enron's accounts, and it still blew up," he says. "There's no substitute for actual measurements."

Measuring gases in the air would require a global network such as the stations that report weather patterns. So far, countries haven't been inclined to shell out the money. That leaves a loose band of air checkers spread thinly around the planet.

In India, D.V. Borole, a retired government researcher, treks to a bluff in Goa on the west coast once a fortnight.

He gathers samples at one of just six sites in the nation of 1.2 billion to study heat-trapping pollutants.

Leonard Barrie, director of research at the World Meteorological Organisation, says a bare-bones atmospheric system would take about US$65 million a year to run.

Michael Woelk, chief executive of Sunnyvale, California-based measurement equipment maker Picarro, says building a worldwide network would cost roughly US$5billion. He says the money is worth it.

"There's a misunderstanding that self-reported emissions are accurate and fail-safe," he says. "They aren't."

Sloppy emissions estimates today could mean a nasty correction for the economy down the line, Woelk says.

"How can we know if cap and trade is working to reduce emissions if we don't understand the actual composition of the air?" he says. "We're essentially setting up a subprime carbon crisis."

Barrie says laser analysers are cheap and make fast, accurate, continuous measurements.

"In the next five to 10 years, we expect to see a revolution in observations," he says.

Nisbet, the professor sampling air in Hong Kong, has seen how testing the atmosphere can make a difference.

His data on methane was used in an early study that detected discrepancies between air-based measurements and bottom-up estimates.

That study found that methane from human activities in France, Germany and Britain in 2001 was 50 per cent to 90 per cent higher in the air than the countries had first reported to the UN.

Nisbet says the bottom-up calculations that underpin today's emissions markets and climate pacts have one big selling point: they're consistent.

"You can do the paperwork again and again and you'll get the same thing," he says.

"It's like if you ask someone again and again for their tax return, it'll be the same. It might be true, it might not. Who knows?"

Measuring the air has its flaws. "You don't always get the same result," Nisbet says. "But it's what's really there. It's there because we can measure it."