When the world's nations meet in Copenhagen this December to try to construct a successor to the 1997 Kyoto Protocol on climate change, one major point of discussion will be 'offsets'. These are deals that could help countries meet their targets for reducing emissions by paying for others to absorb greenhouse gases in natural carbon sinks such as rainforests, or by otherwise reducing the threat of global warming.

Any new agreement would presumably build on the existing Kyoto framework that allows certified credits from offsetting projects, such as planting trees, to be traded on the international emissions market. Voluntary carbon offsets are also becoming increasingly popular among businesses and air travellers who want to compensate for the carbon footprint of their activities.

As things stand, unfortunately, the success or failure of any such policy is largely a matter of guesswork: there has never been a global observation network capable of verifying whether the carbon dioxide emissions and offsets reported by individual countries make any sense. Carbon-cycle scientists estimate, for example, that around one-third of the CO2 from fossil fuels burned globally is taken up by land vegetation. But they have no idea what the precise fraction is, or where the carbon actually goes: in situ measurements of biosphere-to-atmosphere carbon fluxes are scarce, and ecosystem inventory data are often unavailable. In addition, monitoring efforts suffered a dire setback on 24 February when NASA's US$278-million Orbiting Carbon Observatory (OCO) crashed into the ocean minutes after launch.

Any new international climate agreement, whether it emerges at Copenhagen or later, must therefore provide for a much-improved carbon-monitoring infrastructure for verifying its effectiveness. One key element will be satellite observations, which provide large-scale mapping of greenhouse-gas emissions and land-cover changes. NASA should get the support it needs to build a cheaper copy of OCO, which could be launched as early as 2011 (see Nature 458, 8; 2009).

But equally crucial will be high-precision, in situ measurement of carbon fluxes between soils, vegetation and the atmosphere. The many international agencies that make up the Integrated Global Observing Strategy partnership should implement, without further delay, their 5-year-old plan for an Integrated Global Carbon Observation programme. A good place to start would be to expand FLUXNET, an existing surface network of some 400 carbon-measurement towers that still has huge gaps, particularly in the tropics.

At the same time, the agencies that comprise the intergovernmental Group on Earth Observations (GEO) could aim to produce globally harmonized data sets on global, national and local scales, using common algorithms, variables and units. GEO, which coordinates efforts to create the Global Earth Observation System of Systems, should also commission scientists to develop an integrated model that stitches all carbon observations together. It should then make these available for use at all levels by scientists and policy-makers alike.

A global carbon-measurement system along these lines should make international climate policies much more solid than they have been in the past. It might reveal that what we are doing is not enough, and that many offset projects fail to deliver. It might expose swindlers and profit-makers in the carbon business. Or it might prove that nature is a stronger ally than we have dared to hope. Whatever the outcome, a serious investment in carbon monitoring will be money well spent.