permanently withdrawn from other uses, including harvest of wood products, so that there is no release of carbon (e.g., Nordhaus, 1991[a or b?]; Richards et al., 1993). This assumption must be reflected in the calculation of land costs. Finally, some studies simply do not address the release of carbon upon harvest, implicitly assuming that either the forest area will not be harvested, or that the harvest will occur so far in the future as not to be a concern (e.g., Moulton and Richards, 1990).

Several components of a forest ecosystem store carbon, including tree trunks, branches, leaves, and coarse and fine roots; soils; litter; and understory. Studies have varied significantly with respect to how they address these various components. Some have included all components in their carbon accounting (e.g., Moulton and Richards, 1990). Others have limited their analysis to above-ground carbon (Dixon et al. [Dixon, Winjum, and Krankina or Dixon, Schroeder, and Winjum], 1991). Table 9.33 provides a summary of which carbon components are included in each of the studies reviewed.

Box 8.2 in Chapter 8 [DON'T HAVE BOX 8.2. PLEASE SUPPLY.] provided a discussion and sample calculations of the various summary statistics used in carbon sequestration cost studies to capture the concept of "dollars per tonne of carbon sequestration." Van Kooten et al. (1992) demonstrate the importance of the choice of summary statistics in their analysis of the cost-effectiveness of carbon sequestration in

Canada. Their analysis of costs employs the flow summation method, while an appendix provides calculations using the levelization approach. The costs in the latter case rise by a factor of 5 to 10 relative to the former case. Table 9.34 provides a summary of the approaches employed by the carbon sequestration cost studies.

9.3.1 Costs of Carbon Sequestration

Table 9.35 summarizes the estimates of unit costs of carbon sequestration provided by the studies reviewed here. The studies fall into four general categories. One group concentrates on the potential of North America to sequester carbon (Adams et al., 1993; Moulton and Richards, 1990; New York State, 1990; Van Kooten et al., 19901992; Parks and Hardie, 1995; Richards et al., 1993). This group predominantly uses a cost levelization/discounting approach. The one exception is van Kooten et al. (1992), and they complement their flow summation approach by providing cost levelization results in their appendix. The second group considers the carbon sequestration potential of major ecological regions of the world using the average storage method (Dixon, Schroeder, and Winjum, 1991; Dixon, Winjum and Krankina, 1991; and Dixon et al., 1994). The third group is comprised of studies of the global potential and cost of carbon sequestration (Sedjo and Solomon, 1989; Nordhaus, 1991[a or b?]). Sedjo and Solomon (1989) do not provide unit cost calculations of carbon sequestration, while Nordhaus applies a discounting method. The fourth group, a set of recent studies, examines the potential for carbon sequestration in individual developing countries (Masera et al., 1994; Ravindranath and Somashekhar, 1994; Xu, 1994). These studies use either the average storage method or the flow summation method.

Among the group of studies that concentrate on North America, the estimates of carbon costs fall into a relatively narrow range. After accounting for the differences attributable to cost analysis methods, Moulton and Richards (1990) provide the lowest cost range ($9 per tonne to $41 per tonne of carbon captured). These estimates were subsequently revised to reflect refined (lower) carbon yield estimates, the elasticity of demand for agricultural land, administrative costs, and failure rates (Richards, et al. 1993). The analysis by Adams et al. (1993), which is based on a method of imputing land values through consumer welfare loss derived within a mathematical programming model of the agricultural sector, tends to confirm the revised results. Both studies suggest that the marginal cost of carbon sequestration would range from $9 to about $65 per tonne[?] of carbon captured (levelized cost basis). The New York State (1991) study is in close agreement with the previous two studies.

Parks and Hardie (1995) present a very different picture. They suggest a similar lower range on costs but a very rapid increase that approaches $90 per tonne of carbon. Several factors contribute to the difference in costs. First, Parks and Hardie recognize much less land availability than either Adams et al. (1993) or Richards et al. (1993). This means that they move into more expensive land very quickly. Also, as indicated by Tables 9.28 and 9.30, their annual land costs are estimated at 40-650 $/ha/yr and their

discount rate is 4%. This suggests a capitalized land cost over the ten-year rental contracts of 320 $/ha to 5300 $/ha, a rental cost that is higher than that used by Richards et al. (1993) for the outright purchase of land. Also, Parks and Hardie (1995) use lower carbon yield estimates and include only tree carbon in their calculations. Finally, and perhaps most importantly, their costs are annualized over only a 10-year contract period. While this may be appropriate for their analysis of a specific hypothetical government programme, it almost certainly overstates the costs of carbon sequestration in a broader context, since carbon capture continues for several decades into the future, even after the end of government land rental payments.

In their presentation of levelized costs, van Kooten et al. (1992) also provide higher estimates of carbon costs than either Adams et al. (1993) or Richards et al. (1993). This might be surprising in light of the fact that they do not include land costs in their estimates. The difference can be attributed in part to higher initial establishment costs and the low growth rates expected in the Canadian forests. Their carbon capture rates are also low because they only consider the carbon in the tree trunks and not whole ecosystem carbon.

The two studies of broad geographic/climate regions suggest that the costs of carbon sequestration may be relatively low for all three types of practices - forest plantations, forest management, and agroforestry (Dixon, Schroeder, and Winjum, 1991; Dixon et al., 1994). Although these costs are calculated using the average storage method, the costs presented here suggest lower estimates than those derived in the North American studies. It is interesting to note that Dixon, Schroeder, and Winjum (1991) find relatively little difference among the boreal, temperate, and tropical regions with respect to the carbon sequestration costs associated with forest plantations and forest management, which range from $2 per tonne to $8 per tonne and $1 per tonne to $13 per tonne respectively. In contrast, the carbon sequestration costs associated with agroforestry are considerably higher in the temperate region ($23 per tonne) than in the tropics ($5 per tonne). In the second study, however, that relation seems to be reversed: the cost in the tropical areas is $2 to $69 per tonne and the cost in North America is $1 to $6 per tonne (Dixon et al. 1994). The reversal appears to have occurred because the relative costs of initial treatment in the temperate zone were lowered in the second study and the MCS capacity of land was raised.

The two studies of global cost estimates differ significantly. The Nordhaus (1991[a or b?]) estimate of $42-$114 per tonne for the unit costs of global carbon sequestration through afforestation is much higher than those from the other two groups. This is a bit surprising, given the fact that Nordhaus's land and treatment cost figures are similar to those used by other studies. The difference in results is almost entirely attributable to how the Nordhaus study treats carbon yields. First, Nordhaus uses average carbon yield factors derived from the review of greenhouse gas policy options conducted by the U.S. EPA (1989). These figures are for carbon yields on average commercial timber land and probably substantially underestimate yields expected from

conversions of marginal agricultural land to forestry plantations (see Table 9.31 for a comparison with other studies). Second, the analysis limits the total cumulative carbon to a range of 30-50 tonnes/ha and assumes that this amount occurs over a forty-year period following plantation establishment. These figures are certainly at the low end of the expected carrying capacity of forest plantations (see, e.g., Dixon, Schroeder, and Winjum, 1991). Finally, to portray the timing of carbon capture, Nordhaus applies a logistic growth curve that has the effect of delaying carbon uptake relative to the rate given by the average flow approach or the MCS approach. Combined with a levelization approach to costs, this delay in carbon uptake contributes to an increased unit cost of carbon capture. (Richards et al., 1993, also captures this effect.)

At the other extreme, Sedjo and Solomon (1988) provide land and treatment cost figures that would suggest a cost of carbon sequestration of $7 per tonne on a cost levelization basis and $3 per tonne on a flow summation basis. This relatively low cost estimate is due to their optimistic assumption regarding carbon yield, which is based upon growth rates in the Pacific Northwest and Southeast regions of the United States. Applying these rates to a global analysis is probably unrealistic, but it does suggest that, at least in some regions, carbon sequestration should be relatively inexpensive.

The three studies of individual developing countries provide an interesting contrast. While Masera et al. (1994) estimate that carbon sequestration on forest plantations in Mexico would cost $5-$11 per tonne, Xu (1994) calculates that carbon could be stored on plantations in China at a negative cost. The latter result stems from the fact that Xu includes revenues from the sale of forestry products in the cost calculations, and that China has a largely unmet demand for timber. The interpretation of cost figures for Ravindranath and Somashekhar (1994) is unclear. While they apparently use a flow summation approach in their cost calculations, they do discuss the application of discounting, at zero and 1%, to the carbon flow, which would suggest a levelized cost method. Their costs range from $0.09 to $2.78 per tonne.

9.3.2 Potential Quantities of Carbon

The studies show a wide range of estimates of potential for carbon sequestration (Table 9.36). At one extreme, Sedjo and Solomon (1988) have estimated that if 465 million hectares of land can be secured, 2.9 Gt of carbon per year can be removed from the atmosphere in forest plantations. This is nearly one-half of current annual global levels of carbon emissions. At the other extreme, Nordhaus (1991[a or b?]) suggests that an average of only O.28 GtC per year can be captured over a period of 75 years, even in the presence of a global effort. The difference between these two estimates is almost entirely due to the estimates of carbon yields, since their assumptions on land availability are very similar. At the same time, Dixon et al. (1994) estimate that globally 1.1-2.2 GtC can be captured annually using expanded agroforestry practices alone.

Opportunities in the most northern latitudes appear somewhat limited. Dixon, Schroeder, and Winjum (1991) suggest that summing across all practices, only 2 GtC can be accumulated in the boreal regions. Averaged over their 50-year period, this yields 0.04 GtC per year. However, Van Kooten et al. (1992) suggest that forestry opportunities in Western Canada alone may provide as much as 0.13 Gt of carbon capture per year.

In the temperate regions there appear to be significant opportunities. Richards et al. (1993) suggest that in the United States alone, an aggressive tree planting programme could yield an average of 0.4 GtC per year for 100 years and a cumulative total of 49 GtC if the plantations are undisturbed for 160 years. Dixon, Schroeder, and Winjum (1991) are not as optimistic. That study suggests that across all forestry practices a total of 20 GtC could be accumulated in the entire temperate zone. Over their 50-year analysis period this averages to 0.4 GtC per year. Because the estimate of land area availability in Dixon, Schroeder, and Winjum (1991) is much higher than in Richards et al. (1993), the difference in the estimates must be due to the fact that the latter study uses much higher estimates of potential carbon accumulation per hectare.

The outlook in the tropics is even better than that in the temperate region. Dixon, Schroeder, and Winjum (1991) suggest that a cumulative total of 53 GtC could be captured across all forestry practices. Houghton et al. (1993) provide an even more optimistic estimate of the potential in the tropics, 167 GtC, though they provide no cost figures.

Two other studies that provide estimates of carbon sequestration potential without analyzing costs suggest that Australia and New Zealand could capture carbon at a rate of 0.007 Gt per year and 0.005 Gt per year respectively for 25 to 30 years (Barson and Gifford, 1990; Tasman Institute, 1994). For Mexico, India, and China it is estimated that approximately 3.5 GtC, 8.7 GtC, and9.8 GtC respectively could be accumulated (Masera et al., 1994; Ravindranath and Somashekhar, 1994; Xu, 1994).

9.3.3 Cost Curves

This discussion has suggested that there is significant variation among the studies with respect to their estimates of the unit costs and potential amounts of carbon sequestration. The studies also differ in how they present the results. Several of the reports (Sedjo and Solomon, 1989; Nordhaus, 1991[a or b?]; New York State, 1991; van Kooten et al., 1992) have presented their results as point estimates, that is, as estimates of the costs of achieving a specified amount of carbon sequestration. Other studies have developed cost curves that illustrate the increasing marginal cost of carbon sequestration as a function of the level of sequestration (Adams et al., 1993; Moulton and Richards, 1990; Dixon, Schroeder, and Winjum, 1991; Parks and Hardie, 1995; Richards et al., 1993). The cost curves provide information that is not conveyed in either