126.96.36.199. Uncertainty, Precision, Accuracy, and Costs
There is a widespread perception that accounting for carbon in the biosphere
is inherently more difficult than accounting for carbon emitted by the burning
of fossil fuels. This perception is only partly true. Determining the amount
of carbon held in vegetation and soils does not require measurement of all of
the carbon in all areas. Well-established statistical sampling techniques such
as stratification and random sampling can be used to determine the carbon stocks
of the biosphere as precisely and accurately as would be possible with complete
Neither accuracy (the absence of bias) nor precision (how well a measurement
can be repeated) is compromised when appropriate statistical methods are used.
Nevertheless, in statistical sampling there is a tradeoff between precision
and effort (cost). A greater number of samples reduces error (i.e., increases
precision) in the estimate but also requires more effort. In fact, each successive
increment in precision requires a proportionately greater increase in effort.
The decision about whether or how precisely to measure a carbon sink will depend
on the effort required for that precision and the magnitude of the expected
sink. Small changes in carbon or changes that require a large effort for measurement
may not be worth the effort. Table 2-7 gives
the costs of different methods in US$ per hectare for project-level measurements.
Costs per ton of carbon are not given because the costs vary from country to
country not only with the desired precision but with the types of land-use change
and the magnitude of the change in carbon per unit area. Thus, costs per ton
of carbon at the national level must be determined for each country individually.
Similarly, the combination of methods most appropriate for national-level estimates
of change will vary for each country; we do not attempt to suggest a single
There are at least two significant generic problems with the estimation of
change in terrestrial biospheric carbon stocks. The first problem relates to
resolution (i.e., the smallest detectable change). Because the rate of change
of most biospheric pools is slow, particularly in relation to the size of the
pool, resolvable changes in stock are typically not easily obtained for the
The second problem is practical. Most countries do not have the established
infrastructure required for regular measurement of biospheric carbon (although
all Annex I countries have a regular forest inventory in place). Where no infrastructure
exists, measurement of carbon to the required degree of precision and accuracy
is an expensive and logistically complex exercise. Most of the developed countries,
as well as some less developed countries, have at least part of the required
infrastructure already in place: certifiable analytical laboratories equipped
to measure the carbon content of soils and biomass; a national forest and soil
inventory system; accurate soil and vegetation maps on which to base the sample
stratification; trained field, analytical, and statistical staff; and a physical
infrastructure that allows access to remote sites. Even where this capacity
exists, the incremental cost of performing a national-scale carbon inventory
may be substantial. Australia, for example, is investing an additional $5 million
annually in anticipation of upgrading its carbon accounting system for the Kyoto
Protocol. The costs may be greater in countries in which the inventory infrastructure
is less well developed. The use of models and stratified and multi-objective
sample programs may reduce these costs, however.
The cost of conducting biospheric carbon inventories depends on the size of
the area inventoried-but more on the range of ecological conditions within it
because the spatial scale over which soil and biomass carbon varies is quite
small: a few tens to hundreds of meters. The sample size needed to achieve the
desired precision may be similar for a small country and a distinct region within
a large country. Both need approximately the same analytical equipment and statistical
treatment. Thus, the cost to a country will depend more on the range of different
bio-geophysical regions that exist within its borders than on its actual size.
The total inventory costs for a single project will be substantially less than
the inventory costs for an entire country, but the cumulative cost of inventorying
several projects soon reaches the level of a national inventory. This is because
the costs of forest inventory vary considerably depending on the desired accuracy,
the accessibility of the forest, the availability of pre-information, the degree
of automation employed, the availability of allometric relations, and so forth.
At a typical project scale of tens of thousands of hectares, the costs are on
the order of US$11-18 ha-1 (Nabuurs et al., 1999). The typical cost of
a national forest inventory, on the other hand, is on the order of US$0.05-0.6
ha-1. Costs are declining with the increased use of automated data collection
and analysis (see also Table 2-7).
The methods for determining changes in carbon storage are described briefly
in the following sections. These methods are divided into methods for the measurement
of stocks, the measurement of flux, and the measurement of area, as well as
models. Table 2-7 summarizes the characteristics
of different methods.