It is hardly surprising that climate discussions tend to gloss over uncertainties in data on greenhouse-gas emissions. Governments are struggling towards an international agreement to reduce those emissions, and their focus is necessarily on coming up with specific, enforceable targets. But the fact is that scientists' ability to measure emissions and verify that countries are following through on their commitments is far from adequate. And that is unlikely to change without the full engagement of governments and scientists.

There are a number of reasons to be sceptical about current emissions data (see page 18). Some are a matter of human frailty: it is often in the best interests of both companies and governments to underestimate emissions, and thus to overstate the effectiveness of a given technology or policy in reducing them. That temptation will only increase as countries ramp up climate commitments. Other reasons, however, hinge on the uncertainties inherent in even the best emissions statistics.

In the energy sector, for example, the statistics rely on imperfect estimates of, say, the amount of fuel moving through a pipe or leaking out of a valve. Emissions also vary from one chunk of coal to another, according to their origins, as well as among different kinds of crude oil. Outside the energy sector, the challenges are even more complex. Agricultural emissions vary from crop to crop and from farm to farm, and there is no single equation for estimating carbon uptake by forests. The further one digs down into the data, the more uncertainties one encounters.

So what should be done? The US National Research Council tackled this question in a report issued in March, laying out a roadmap that could — and should — be implemented relatively cheaply over the next few years. The first task is to improve emissions inventories. At present, industrialized nations are required to report their annual emissions data to the United Nations each year, but these data need to be broken down by time, region and, as far as possible, by facility. Major developing countries would then need to be phased into this same system.

These inventories are calculated using government and industry data. However, given the high economic stakes, even the most thorough such reporting will not suffice as evidence about which nations are and are not living up to their commitments. These emissions numbers will need to be independently verified with an expanded network of atmospheric measurements. To accomplish this, governments should extend and improve their verification efforts by increased monitoring of major facilities, cities and agricultural areas, as well as by supporting a global monitoring capability from space.

Europe is making progress in this direction with a planned ground-based Integrated Carbon Observation System that could be rolled out within a few years, but efforts are needed around the world. Where satellites are concerned, the outlook is encouraging, although frustratingly slow. Japan is collecting initial data from GOSAT, its satellite for observing greenhouse gases; NASA, meanwhile, is pushing forward with a second Orbiting Carbon Observatory after losing the first during launch last year.

All of these efforts need to be continued, strengthened and expanded. The specifics of who pays and how the data-gathering is managed will doubtless be hammered out over time, and may well have to be included in whatever international climate treaty finally emerges. But regardless of how the details play out, it's clear that those policing any such treaty will need a much more sophisticated monitoring system than the ad-hoc version in use today.