The “budget” of a trace gas consists of three quantities: its global source, global sink and atmospheric burden. The burden is defined as the total mass of the gas integrated over the atmosphere and related reservoirs, which usually include just the troposphere and stratosphere. The global burden (in Tg) and its trend (i.e., the difference between sources and sinks, in Tg/yr) can be determined from atmospheric measurements and, for the long-lived gases, are usually the best-known quantities in the budgets. For short-lived, highly variable gases such as tropospheric O₃ and NO_x, the atmospheric burden cannot be measured with great accuracy. The global source strength is the sum of all sources, including emissions and in situ chemical production. Likewise, the sink strength (or global loss rate) can have several independent components.

The source strength (Tg/yr) for most greenhouse gases is comprised of surface emissions. For synthetic gases where industrial production and emissions are well documented, the source strengths may be accurately known. For CH₄ and N₂O, however, there are large, not well-quantified, natural emissions. Further, the anthropogenic emissions of these gases are primarily associated with agricultural sources that are difficult to quantify accurately. Considerable research has gone into identifying and quantifying the emissions from individual sources for CH₄ and N₂O, as discussed below. Such uncertainty in source strength also holds for synthetic gases with undocumented emissions. The source strength for tropospheric O₃ includes both a stratospheric influx and in situ production and is thus derived primarily from global chemical models.

The sink strength (Tg/yr) of long-lived greenhouse gases can be derived from a combination of atmospheric observations, laboratory experiments, and models. The atmospheric chemistry models are based on physical principles and laboratory data, and include as constraints the observed chemistry of the atmosphere over the past two decades. For example, stratospheric loss rates are derived from models either by combining observed trace gas distributions with theoretically calculated loss frequencies or from the measured correlation of the respective gas with a trace gas of known vertical flux. In such analyses there are a wide range of self-consistency checks. Mean global loss rates based on a priori modelling (e.g., the CH₄-lifetime studies from OxComp described later) can be compared with empirically-based loss rates that are scaled from a gas with similar loss processes that has well-known emissions and atmospheric burden (e.g., the AGAGE (Advanced Global Atmospheric Gases Experiment) calibration of mean tropospheric OH using methyl chloroform (CH₃CCl₃); Prinn et al., 1995). Our knowledge of the current budget of a greenhouse gas provides a key constraint in modelling its future abundance. For example, in both the IS92a and SRES projected emissions of CH₄ and N₂O, we apply a constant offset to each set of emissions so that our calculated burden is consistent with the observed budget and trend during the 1990s.

IPCC Homepage