5.4.4. Precision and Costs

Field methods to accurately quantify carbon pools exist, but the level of precision can vary by pool. The total error in measuring a given carbon pool is based on sampling error (variation among sampling units-e.g., the number of plots, within the population of interest), measurement error (error in measuring the parameter of interest-e.g., stem diameter and soil carbon) and regression error, when appropriate (e.g., error resulting from conversion of tree diameter to biomass based on a regression equation). Sampling error is usually the largest source of error (Phillips et al., 2000), and increased precision generally comes at increased costs of inventorying because of the time and cost involved in establishing the appropriate number and distribution of permanent plots. Carbon inventory in forests can be more complicated than traditional forest inventories because each carbon pool will have a different variance. The sample size for each pool can be calculated individually; based on resources available for monitoring the project and the information in Table 5-7, informed decisions can be made about which pools to measure and count. Such information can be used at the design stage to select pools to be included in the project, with significant implications for the total cost of the project and the measurement and monitoring costs per ton of carbon.

The costs of measuring and monitoring carbon offsets are a function mainly of the desired level of precision-which may vary by the type of project activities, the size of the project (areal extent, as well as whether the project area is contiguous or a dispersed bundle of small landowners), and the natural variation within the various carbon pools. For example, an increase in the coefficient of variation (a measure of the variation around the mean) within a forest stratum of about 160 percent would increase the cost of measurement by about 280 percent to maintain the same level of precision (Figure 5-6a). Stratification of the project area into more or less homogeneous units-based on vegetation type, soil type, topography or management practice-can increase the precision of carbon measurements without increasing the cost unduly by lowering the amount of variation around the mean, thereby requiring fewer plots to be within acceptable levels of precision.

Figure 5-6: Relationship between (a) coefficient of variation and (b) level of precision and total variable costs for the first inventory phase of the Noel Kempff Climate Action Project in Bolivia (data from Powell, 1999).

A few data provide some preliminary estimates of costs for measuring and monitoring of carbon in LULUCF projects in three tropical countries (Powell, 1999; Subak, 2000; Box 5-4). For the first inventory of the NKCAP, the total fixed operational costs (including human resource costs, project management, mapping, and so forth) were estimated to be about $196,000; variable costs (including labor, equipment, transport, and so forth) ranged between $230 and $281 per plot and totaled about $154,000 (625 plots). The grand total cost was about $350,000 (Powell, 1999). The precision of the inventory, based on sampling error only, was ±4 percent with 95-percent confidence (see Box 5-3). The variable costs dropped rapidly from about $108,000 for a precision level of ±5 percent to $1,000 for a level of ±30 percent; fixed costs would be the same for all levels of precision (Figure 5-6b). Estimates of the revised carbon benefits from this project for its duration based on additional measurements and data collection (Brown et al., 2000) and the additional cost to collect this information result in an estimate of about $0.10 per t C. Estimating future monitoring costs based on the first inventory is difficult, but they are likely to be less than those for the initial inventory because different sampling intensities will be used, project implementers can build on previous experience, and advances in technology will be available (see, e.g., Section 5.4.3).

The organization responsible for monitoring carbon sequestration in Costa Rica's Private Forestry Project and for acquiring remote-sensing information has an annual budget of $200,000 (Subak, 2000). Additional costs associated with the PFP relate to the costs of monitoring forests and plantations. The implementing organizations do not absorb all of the monitoring costs: They charge landowners at a rate that varies in different regions. In the Central Volcanic Range (including the upper Virilla), for example, landowners pay implementing organizations 10 percent of their annual environmental services payment for monitoring of forest protection and slightly more for monitoring of forest management but do not pay anything for monitoring of plantations (Subak, 2000). The implementing agency's unit costs will tend to be higher when monitoring of smaller landholdings, and although a stated objective of the PFP is to compensate small and medium-sized landowners, some implementers may favor larger parcels. Monitoring of the PFP is supposed to include site visits by forest engineers as well as more detailed audits of some sites. The annual visits involve making a report on the size, density, and health of the trees on the land; the more detailed audits are to assess management as well as the conditions of the trees and soil. The intention is to audit as few as 5 percent of the PFP sites. The labor costs for auditing are estimated to be $10 ha^-1 yr^-1-compared to $1 ha^-1 yr^-1 for monitoring and $2 ha^-1 yr^-1 for certification. The aggregate costs of project development, recruiting, and auditing are significant, but they have not been judged to be excessive or to reverse the cost-effectiveness of the PFP as an LULUCF project.

Currently, there are no guidelines regarding the level of precision to which pools should be measured and monitored. Setting such a level would facilitate comparison of projects and could encourage project developers to measure projects more precisely if the price of carbon was high. If the total average carbon benefit was 5 million t C with a ±30-percent confidence interval, for example, the lower bound carbon benefit would be 3.5 million t C. If LULUCF projects could only claim benefits for a lower bound of the confidence interval and if carbon was worth $10 per t C, the "loss" of carbon benefits would be 1.5 million tons-equivalent to $15 million, a value likely to greatly exceed the cost of monitoring to a ±5-percent precision level. Thus, project developers would probably chose high precision levels for their monitoring.

IPCC Homepage