UNCERTAINTY IN LAND-USE CHANGE AND FORESTRY SECTOR MITIGATION OPTION= S: PLANTATION SILVICULTURE VERSUS AVOIDED DEFORESTATION

<= o:p>

The text that follows is a PREPRINT.=

<= o:p>

P= lease cite as:

<= o:p>

Fearnside, P.M. 2000. Uncertainty in land-use change and forestry sector mitigation options for global warmin= g: Plantation silviculture versus avoided deforestation. Biomass and Bioene= rgy 18(6): 457-468

&nb= sp;

DOI: 10.1016/S0961-9534(00)00003-9

ISSN: 0= 961-9534

Copyrig= ht: Elsevier

The original publication is available at http://else= vier.com.nl <Ingenta access> <CSA access>

<= o:p>

UNCERTAINTY IN LAND-USE CHANGE AND FORESTRY SECTOR M= ITIGATION OPTIONS FOR GLOBAL WARMING: PLANTATION SILVICULTURE VERSUS AVOIDED DEFORESTATION

&= nbsp; &nbs= p; Philip M. Fearnside

&= nbsp; &nbs= p; Department of Ecology

&n= bsp; = ; National Institute for Research

&n= bsp; = ; in the Amazon (INPA)

&= nbsp; &nbs= p; C.P. 478

&= nbsp; &nbs= p; 69011-970 Manaus, Amazonas

= &nb= sp; BRAZIL

&= nbsp; &nbs= p; Fax: 55-92-642-8909

&= nbsp; &nbs= p; PMFEARN@INPA.GOV.BR

= &nb= sp; 27 Nov. 1999

= &nb= sp; 17 Apr. 2000

Biomass and Bioenergy 18(6= ) [2000] (In press)=

&nbs= p; &= nbsp; TABLE OF CONTENTS

ABSTRACT .......................................

KEYWORDS .......................................

1. INTRODUCTION ................................

2. PROBABILITY OF SUCCESS ......................

3. EFFECT OF UNCERTAINTY REDUCTION ............

4. EFFECT OF A REQUIRED LEVEL OF CERTAINTY .....

5. CONCLUSIONS..= .................................

ACKNOWLEDGMENTS ................................

REFERENCES ....................................

FIGURE LEGENDS ................................

LIST OF TABLES

T= able 1. Deforestation versus carbon value in Brazilian Amazonia.

T= able 2. Hypothetical example of a Bayesian comparison of mitigation options.

LIST OF FIGURES<= o:p>

Fig. 1. Effect of uncertainty reduction

<= o:p>

Fig. 2. Carbon credit at different levels of uncertainty

Fig. 3. Plantati= ons for charcoal: benefit/ha and number of hectares

Fig. 4. Avoided deforestation: benefit/ha and number of hectares

F= ig. 5. Effect of different levels of required certainty on the cost of carbon sequestration, using the example parameters.
UNCERTAINTY IN LAND-USE CHANGE AND FORESTRY SECTOR MITIGATION OPTIONS FOR GLOBAL WARMING: PLANTATION SILVICULTURE VERSUS AVOIDED DEFORESTATION

&= nbsp; &nbs= p; Philip M. Fearnside

&= nbsp; &nbs= p; Department of Ecology

&n= bsp; = ; National Institute for Research

&n= bsp; = ; in the Amazon (INPA)

&= nbsp; &nbs= p; C.P. 478

&= nbsp; &nbs= p; 69011-970 Manaus, Amazonas

&= nbsp; &nbs= p; BRAZIL

ABSTRACT

How land-use change and forestry sector options can be used to mitigate global warming will depend on a vari= ety of pending decisions regarding interpretation of the Kyoto Protocol, including treatment of uncertainty. In tropical forest countries, the allocation of effort between plantation silviculture and reduction of deforestation would be influenced by the stri= ngency of requirements regarding certainty. Slowing deforestation offers much greater potential benefits, but the certainty associated with these is much lower than in the case of plantatio= ns. In the Brazilian case, deforestati= on avoidance could produce carbon benefits worth 6-45 times as much as the destructive ranching and logging uses to which the forest is now being converted. Capturing the pote= ntial value of carbon benefits from avoided deforestation will depend on increasi= ng our understanding of the deforestation process and consequent ability to re= duce the uncertainty associated with the effects of deforestation-avoidance measures. It will also depend= on whether carbon credits are defined in terms of a maximum level of uncertain= ty.

KEYWORDS: Carbon, Global warmin= g, Kyoto Protocol, Uncertainty, Climate change, Mitigation

1. INTRODUCTION<= o:p>

The danger of global warming caused by increased atmospheric concentrations of gases such as carbon dioxide (CO₂) gives urgency to finding ways= to reduce greenhouse gas emissions and/or enhance flows of atmospheric carbon = to biomass and other “sinks.”&nbs= p; The United Nations Framework Convention on Climate Change (UN-FCCC) = was negotiated in 1992 for this purpose, and has since been supplemented by the December 1997 Kyoto Protocol [1]. The Kyoto Protocol establishes emissions limits for countries that a= re members of Annex B of the Protocol (similar to Annex I of the UN-FCCC), whi= ch currently includes most of the industrialized countries. The Protocol also establishes a va= riety of ways in which these commitments can be met. These include undertaking mitigati= on projects in non-Annex B countries under the Clean Development Mechanism (CD= M) defined in Article 12 of the Protocol [2], and emissions trading defined in Article 17. The uncertainty of estimates of emissions and uptakes is likely to affect the amount of credit= or debit that would be assigned to a global warming mitigation project or to a country under the Kyoto Protocol. Decisions have not yet been made on how these issues will be handled. It is hoped that the conceptual discussion that follows will contribute to understanding the consequences of different possible decisions in this area.

The most pressing question regarding sinks in the land-use change and forestry (LUCF) sector is whether avoided deforestation can be made a viable mitigat= ion option under the Kyoto Protocol, particularly in the context of the CDM. The CDM (Article 12) applies only = to countries that have not joined Annex B of the Protocol, and therefore do not have limitations on their national emissions. At present, all countries with substantial areas of tropical forests fall into the “non-Annex B̶= 1; category, making them eligible for CDM projects but ineligible for other provisions for carbon credit under the Protocol, such as Joint Implementati= on (under Article 6) and emissions trading (under Article 17). At some time in the future, tropic= al forest countries may find it in their interests to join Annex B in order to capture major potential financial benefits from carbon credits traded under Article 17 that could be achieved by slowing deforestation [3].<= /span>

The CDM is expected to come into effect before other “flexibility mechanisms” (provisions under the Protocol allowing Annex B countries= to meet some or all of their commitments through activities in other countries). This is because the CDM includes a “banking” provision through which carbon credit can be earned as early as 2000 and later applied to offset emissions of Annex B countries du= ring the Protocol’s first commitment period (2008-2012). Decisions are still pending as to = what, if any, LUCF activities will be eligible for credit under the CDM. Decisions will probably await comp= letion in May 2000 of the Intergovernmental Panel on Climate Change (IPCC) Special Report on Land-Use Change and Forestry.&nb= sp; Decisions on these matters are likely to be taken at the Sixth Conference of the Parties (COP-6) of the UN-FCCC, to be held in November 20= 00.

Because the CDM is project based, many potential LUCF projects could be subject to “leakage,” or loss of their benefits for global climate by effe= cts beyond a given project’s geographical, temporal or subject area boundaries. For example, if a forest reserve is set up to protect carbon, people who would otherwise have been clearing in that area may simply move to some other part of the forest= to continue clearing.

The problem of uncertainty in establishing causality (attribution) is substanti= al in the case of deforestation reduction projects due to high potential for leakage (in the case of reserve protection) and to poor understanding of the social processes leading to deforestation (in the case of policy initiatives). This problem is likely to reduce substantially the carbon benefit that can be claimed under= the CDM (Article 12) and in many cases under Joint Implementation (Article 6), = but has much less potential effect on the credit that could be obtained under emissions trading (Article 17) should tropical forest countries join Annex = B of the Protocol. This is because emissions trading is based on accounting at the national level, rather than= at the project level, making any leakage that occurs within national borders irrelevant and making it unnecessary to show why a given reduction in emiss= ions has occurred. This assumes th= at Article 3.7 (the “Australia clause”) of the Protocol is interpr= eted to include deforestation as a form of land-use change to be counted in the = 1990 baseline for countries that, like most tropical forest countries, had net e= missions from land-use change in 1990 [4, pp. 318-319].

Table 1 gives an idea of the magnitude of potential gains to tropical countries s= uch as Brazil. The average sale p= rice of forested land in Brazilian Amazonia represents the net present value (i.= e., discounted profits) that could be obtained from current land uses in the region, namely selling the timber and deforesting land for cattle pasture.<= span style=3D'mso-spacerun:yes'> At the US$ 5-35 per metric ton of = carbon (t C) currently used by the United States government in planning its future expenses to meet its Kyoto commitments through “flexibility mechanisms” such as the CDM, the value of maintaining the same hectare’s stock of carbon out of the atmosphere would be 6-45 times t= he average 1998 sale price of forested land in Amazonia (see [5]). The carbon prices used in this calculation refer to tons of carbon of immediately achieved fossil fuel emissions reduction, which has the same effect as avoided deforestation in = the case of countries like Brazil with large areas of remaining forest [6]. Devaluation of the Brazilian Real = by 40% in 1999 makes the contrast between land prices and carbon value even greater today.

Brazil’s 1990 net committed emissions of 267 × 10⁶ t C from Amazoni= an deforestation ([7], updated from [8]) would be worth US$ 1.3-9.3 billion. An estimated 2.4 × 10^{9<=
/sup>
t C is emitted globally each year from tropical deforestation and related
land-use changes such as shifting cultivation and clearing of non-forest
vegetation, calculated as the annual average net committed emissions over t=
he
1981-1990 period [9]. At the =
US$
5-35/t C expected price for carbon, this would represent a loss of US$
12–84 billion in potential revenue from the carbon benefits of reduci=
ng
deforestation. Obviously, eve=
n if
only a small fraction of this potential were realized it could have major
economic benefits for tropical forest countries and major benefits for the
global environment, not only through global warming mitigation but also by
maintaining biodiversity and the roles of forest in the hydrological cycle,
watershed protection, etc.}

&n= bsp; [Table 1 here]

2. PROBABILITY OF SUCCESS

Uncertainty enters decision-making both before the fact (ex ante) and after the = fact (ex post). Ex ante<= /u> decisions regarding deforestation are clearly faced with substantially higher levels = of uncertainty than are ex post ones, as ex ante decisions invol= ve great doubt as to how deforestation will proceed in addition to doubt concerning how any changes might be attributed to different causes. While decisions on which forms of mitigation to finance will undoubtedly be made ex ante, it is likely that any granting of credits will be done ex post.=

Certainty could be a key factor determining whether carbon mitigation funds are used = in LUCF options or in other sectors, such as many energy-sector options. Within the LUCF sector, these same concerns could be critical in determining whether plantation silviculture or deforestation-avoidance options are funded.

Plantation silviculture is much safer and better-understood than deforestation avoidan= ce as a global warming mitigation option–-i.e., uncertainty is much lowe= r in the case of plantations. On t= he other hand, if successful, the potential carbon gains from investments in deforestation avoidance may be orders of magnitude greater than the best possible gains from plantations. The great disparity in the probability of success requires a Bayesian approach to interpreting such numbers.&nbs= p; As in gambling, a prize of fabulous value is misleading if the probability of winning it is minuscule.&nb= sp; The expected monetary value (EMV) of each choice is the value of the payoff for each outcome multiplied by its respective probability of occurre= nce, summed over all possible outcomes associated with each decision (e.g. [13]). A hypothetical numeric= al example representing plantation and deforestation-avoidance alternatives is given in Table 2, illustrating how deforestation avoidance can still be a rational choice even if investments in this area had a high (85%) probabili= ty of achieving nothing.

&n= bsp; [Table 2 here]

3. EFFECT OF UNCERTAINTY REDUCTION

The credit that is awarded to mitigation measures is likely to depend on the benefits to global climate being demonstrable with a specified level of certainty. The logic is simil= ar to that in analyses of the consequences of global warming impacts in terms of a “safe minimum standard” (SMS) of conservation [14]. Canada [15] has proposed that sink= s only be credited for carbon that can be shown to have been held out of the atmosphere with 95% certainty. Marland and Schlamadinger [16] provide an example of simulated plantation benefits with varying levels of certainty, demonstrating that re= ducing uncertainty results in increasing credit if a given level of certainty is required, such as the 95% minimum proposed by Canada.

The potential benefit that reducing uncertainty would yield to those selling ca= rbon credit is illustrated in Figure 1. If 95% certainty is required, the credit that can be awarded is C1 under the low certainty case, but increases to C₂ if the level of certainty is improved (reducing the coefficient of variation, or the standa= rd deviation as a percentage of the mean).&nb= sp; The difference between C₂ and C₁ represents a = gain that could be achieved by increasing the quantity and quality of data and t= heir interpretation. In many cases= this may be more cost effective than investments in expanding the scale of a giv= en project.

&n= bsp; [Fig. 1 here]

The multiplicative nature of uncertainties associated with chain calculations m= eans that the uncertainty of the final result can quickly explode to very high levels [17]. The variance of = the product of each multiplication (Z =3D X Y) can be calculated as:=

&n= bsp; s_z² =3D Y² s_x² + X² s_y2

<= /p>

where X and Y are the means bei= ng multiplied, Z is the product, and s_x², s_y<= sup>2 and s_z^{2 are
their respective variances. T=
his
applies only if the terms are mutually independent; if any autocorrelations
exist, the interaction terms that must be included in the equation usually
result in much greater uncertainty of the final result [18].}

Increasing uncertainty from chain multiplication is particularly important for estimat= ing the benefits of deforestation avoidance.&n= bsp; For example, if estimates of the number of hectares cleared, tons of biomass per hectare and carbon emission per ton of biomass each has a coefficient of variation (cv) of 20%, the cv of the result would increase to 34.6% as a result of the two multiplications. If an additional link in the chain= were added with the same 20% cv, say for methane emission per ton of carbon emit= ted, the cv of the final result would increase to 40%, or double the cv of each component; an additional multiplication would raise it to 44.7%. The same proportionalities are maintained for other levels of uncertainty. The uncertainty of the final result tends to explode if multiplication chains include factors with cv values ab= ove 30% [17]. The final result increases greatly if any factor in the chain has a high cv, regardless of t= he size of the mean value of the factor involved.

Increasing levels of uncertainty have a sharp effect on carbon credit (Fig. 2). For a cv of 20%, about 70% of the = full credit level is retained under a 95% certainty requirement, but if the cv is 40% the percentage of credit retained falls to 35%.

&n= bsp; [Fig. 2 here]

4. EFFECT OF A REQUIRED LEVEL OF CERTAINTY

The uncertainty associated with the benefits of either plantation silviculture = or deforestation avoidance is composed of two component uncertainties: uncerta= inty regarding the carbon benefit per hectare and uncertainty regarding the numb= er of hectares of either plantation establishment or avoided deforestation resulting from a given mitigation investment. Uncertainty in per-hectare benefit= s can arise, for example, due to variation in estimates of biomass of the original vegetation, biomass of the replacement landscape, soil carbon changes, uncertainty in the baseline deforestation rate, etc.

In the case of plantations for charcoal, an example is given in Figure 3 based= on average yield and biomass parameters from Brazilian plantations [6, pp. 314= and 317]. The uncertainty associa= ted with both carbon benefit per hectare and the number of hectares is assumed = to correspond to a coefficient of variation (cv) of 5%. It is assumed that the plantation = lasts for 30 years, during which the carbon stocks in biomass and soils apply, en= ding at the conclusion of this period. The fossil carbon substitution benefits that accumulate over the 30 years are permanent. For simplicity, the timing of benefits over the 30-year project life is smoothe= d to a constant yearly rate averaging 67.7 t C ha^-1, of which 23.9 t = C ha^-1 are from biomass and soils. <= o:p>

&n= bsp; [Figure 3 here]

The probability distributions of expected carbon benefits (e.g. Fig. 3C) were generated in this and other examples discussed in this paper by Monte Carlo simulations in which the area and benefit per hectare values were drawn from their respective probability distributions (e.g. Figs. 3A and 3B) and multiplied together in 2000 simulated cases. The mean carbon benefit (e.g. poin= t x̅= _p in Fig. 3C) and the benefit credited at diffe= rent levels of required certainty (e.g. point C_p for 95% certainty in= Fig. 3C) were calculated from the resulting simulated cases.

In order to compare silvicultural plantations with avoiding deforestation, assumptions must be made about the relative costs of these two options. The examples developed in the foll= owing paragraphs assume that slowing deforestation costs an average of US$ 60/ha-—a value one order of magnitude cheaper than the US$ 625/ha installation cost of plantations [6, p. 317]. Avoiding deforestation prevents an average of 200 t C ha^-1 from being emitted, rounded from the 194= t C ha^-1 average net committed emission from 1990 deforestation in Brazilian Amazonia [7, updated from 8].&nb= sp; The example in Figure 4 avoids an average of 0.18 ´ 10³km² of deforestation, and so would cost US$ 10.8 million. This amount of money= would pay for only 0.017 ´= 10⁶ha of plantatio= ns; at the 67.7 t C ha^-1 benefit level for charcoal this would offset 1= .17 ´ 10⁶t C, on average.

&n= bsp; [Figure 4 here]

A carbon ton-year approach is necessary in order to compare the carbon benefi= ts of deforestation avoidance with those of most silvicultural plantations (plantations other than those that produce fuelwood that substitutes for fo= ssil fuel). This is because the be= nefits of deforestation avoidance are effectively permanent, at least in the case = of countries like Brazil with large areas of remaining forest, whereas carbon sequestered in plantation biomass and in wood products can be expected to be released to the atmosphere at the conclusion of a given project’s life [see 2, 6]. As an illustratio= n, if one assumes a 100-yr time horizon for analysis, a 30-yr project life for plantations and a 5% annual discount rate, one ton of fossil fuel carbon th= at would have been emitted immediately (or one ton of avoided deforestation ca= rbon for countries like Brazil with large areas of remaining forest) is equal to 19.9 ton-years, while one ton of plantation carbon in biomass or soil equals 15.4 ton-years. At this rate = of discount, the carbon sequestered in plantation biomass (as distinct from substitution for fossil carbon) must be adjusted by a factor of 15.4/19.9 = =3D 0.775 to obtain the equivalent in fossil fuel carbon at the same rate of discount. This equivalence ha= s been used to express all values in metric tons of carbon (t C) equivalent to instantaneous avoidance of fossil fuel emission. The equivalence used does not take= into account the natural removal of CO₂ from the atmosphere; however, this factor would be unlikely to substantially alter the relative benefits = of plantation silviculture versus deforestation avoidance.

If pulpwood rather than charcoal were produced by the plantations, then carbon benefits would be about 30-fold lower, with an average benefit of 2.1 t C h= a^-1. Under the same assumptions regardi= ng uncertainty as those applied to the charcoal example, this per-hectare carb= on benefit would produce a total carbon benefit of only 0.036 ´ 10⁶t C, and a credit at 95% certainty of 0.032 ´ 10⁶t C.

For deforestation-avoidance initiatives, uncertainty is especially high regardi= ng the number of hectares that has been or would be saved by a given project.<= span style=3D'mso-spacerun:yes'> The shapes of the probability dens= ity curves associated with benefit per hectare (Fig. 4A) and area of avoided deforestation (Fig. 4B) are expected to be quite different: the curve repre= senting benefit per hectare would be a bell-shaped normal probability distribution (e.g. [19]), while the corresponding curve for the number of hectares of deforestation avoided would have a peak at zero to represent a large chance that nothing would be achieved and a tail on the upper side extending to ve= ry high values. The distribution= of the product of the two values (i.e., expected C benefits) would have a large peak at zero and a long upper tail representing the “jackpot” of potentially very valuable carbon benefits that might be achieved (Fig. 4C).=

The high peak of probability of zero or near-zero outcomes in the case of deforestation avoidance indicates reducing this as the greatest priority for research in this area. Effort= in understanding the causes of deforestation and developing functional models capable of generating scenarios under different policy assumptions may well produce greater dividends in making deforestation avoidance into a viable mitigation option than would effort in reducing the uncertainty surrounding carbon benefits per hectare of deforestation.

Expected carbon benefits from silvicultural plantations and avoided deforestation wi= ll have probability density functions with different shapes (Figs. 3C and 4C).= Plantations (Fig. 3C) will have nor= mally distributed benefits with low variance.&nb= sp; Although not indicated in this example, there may be a small probabi= lity of zero benefit, and also a small probability of negative benefits should t= he project produce impacts, as by stimulating the clearing of high-biomass vegetation to make way for plantations, or by expelling people who clear forests elsewhere. Avoided deforestation will have the high peak at zero and the long tail representing the potential “jackpot” at the right hand side of the graph (Fi= g. 4C). A small probability of negative benefits will also exist. The mean carbon benefit, representing the expected monetary value (a= s in Table 2), is expected to be substantially higher in the case of avoided deforestation (x̅_d &g= t; x̅= _p).

The differing shapes of the probability density curves for plantations versus deforestation avoidance (Figs. 3C and 4C) will affect the result of requiri= ng a minimum level of certainty for carbon benefits. The effect of the large probabilit= y of obtaining results with zero or near-zero values in the case of deforestation avoidance is to drastically reduce the amount of carbon credit that could be awarded for avoided deforestation (C_d in Fig. 4C) to the point w= here it could be lower than that awarded for plantations (C_p in Fig. = 3C; i.e., C_d < C_p).

In the case of a normal distribution, such as that for plantations (Fig. 3C), = the mean value coincides with what can be achieved with 50% certainty, since ha= lf of the expected outcomes will have carbon benefits greater than this value. As one increases the l= evel of required certainty (moving to the left in Fig. 3C), successively less of= the expected (mean) value can be claimed. In the case of the avoided deforestation example, the expected (mean) value corresponds to a required certainty of 27%. The progressive loss of claimable = value will not occur at the same pace as one moves towards a requirement of 100% certainty. In the case of the charcoal plantation example (Fig. 3C), a 95% certainty requirement reduces = the credit (C_p) to 88% of the mean value (x̅_p), whereas in the avoided deforestation case (= Fig. 4C), the credit (C_d) is reduced to zero, thereby losing all of t= he substantially higher expected (mean) value (x̅_d).<= span style=3D'mso-spacerun:yes'> The requirement of a specified lev= el of certainty would significantly favor plantations over avoided deforestation, with the relative advantage of plantations increasing as the required level= of certainty rises.

It is obvious that, if 95% certainty is required, any option with a substantial probability resulting in a zero value will be excluded (i.e., the credited level will be equal to zero). Deforestation avoidance options would be excluded regardless of the magnitude of the reward that could accrue should a “jackpot” be= won by a successful reduction in clearing rates. The mean or expected carbon benefi= t, which is strongly influenced by a “jackpot” in the right-hand t= ail of the probability distribution, will likewise be irrelevant to choices bas= ed on a 95% certainty criterion. The most salient advantages of deforestation avoidance would therefore count for nothing under such a system.

The cost per ton of carbon sequestered would be equal to the mean if no certain= ty requirements apply; with the example parameters given above, this is US$ 9.= 24/t C for plantations producing charcoal and US$ 0.30/t C for deforestation avoidance. In the case of deforestation avoidance, because this example assumes a probability greater than 50% that zero benefit would result (55% in this example), any certainty requirement would immediately result in zero credit (and consequently an infinite cost per ton of carbon), whereas the cost per ton of carbon for plantations would only rise gradually, exploding to infinity only as the required certainty approaches the impossible-to-meet level of 100% (Fig. 5)= .

&n= bsp; [Fig. 5 here]

The assumption that investment in deforestation avoidance has a high probabilit= y of resulting in zero benefit (the 0.55 probability that the area avoided will = be zero in Fig. 4B) is prohibitive. This assumption may be overly severe; if some gain is assured, then = the avoided deforestation curve in Figure 5 would not explode to infinity until= a higher level of required certainty is reached.

Deforestation avoidance is generally seen as having more collateral benefits than expansi= on of plantations because forests provide environmental services other than ca= rbon storage. Maintaining biodiver= sity and the hydrological cycle are particularly important in the case of Amazon= ia, and the willingness to pay for these values already represents a substantial amount [20, 21]. Giving addit= ional weight to the benefits of deforestation avoidance, as compared to expansion= of silvicultural plantations, can be justified on these grounds.

Decisions on the allocation between plantations and avoided deforestation will be critically affected by the choice of a required level of certainty; negotia= tors should be aware that demanding high levels of certainty has both positive a= nd negative consequences for climate, biodiversity and sustainable development. The potential of such a requirement making deforestation avoidance worthless as a global warming mitigation mea= sure should be regarded as a serious disadvantage. A system of weighting is needed th= at explicitly adjusts the index of benefits used in decision-making to reflect= the additional environmental and social benefits of avoided deforestation. Such an adjustment would result in= an upward shift in benefit curves such as those in Figures 4A and 4C and a downward s= hift in the cost curve for avoided deforestation in Figure 5. An additional adjustment to recogn= ize the value of a possible large “jackpot” (the right-hand tail in Fig. 4C) could be applied through expected utility (EU) approaches. While all of these adjustments wou= ld contribute something to increasing the value attached to slowing tropical deforestation, they would do little to overcome the crippling effect of imposing a required high level of certainty. Uncertainty restrictions would be especially unfavorable to deforestation-reduction projects under the CDM, w= hich requires demonstrating that project effects have “additionality” with respect to a no-project baseline.

5. CONCLUSIONS

Uncertainty regarding the magnitude and attribution of carbon benefits from efforts to mitigate global warming is expected to reduce the credit that can be assign= ed to relatively uncertain options. Reducing deforestation could yield much greater benefits for global climate than expanding plantation silviculture, but deforestation reduction= is much more uncertain. The uncertainty associated with deforestation-reduction initiatives is particul= arly high if attribution of causes is required, as under the “clean development mechanism” (CDM) that applies to projects in countries outside Annex B of the Kyoto Protocol.

Requirement of a minimum level of certainty can influence the choices of land-use change and forestry options for mitigating global warming. Requirements for high levels of ce= rtainty would reduce the credit awarded for slowing deforestation, relative to that= for expanding plantation silviculture.

Tropical forest countries could substantially increase their potential for gaining credit for reducing tropical deforestation by joining Annex B of the Kyoto Protocol, thereby eliminating the necessity of establishing the attribution= of causes for changes in tropical deforestation rates (gaining credit through emissions trading under Article 17 rather than through the CDM under Article 12).

ACKNOWLEDGMENTS<= o:p>

&n= bsp; Brazil's National Council of Scientific and Technological Development (CNPq AIs 350230/97-98 and 523980/96-5) and the National Institute for Research in the Amazon (INPA PPIs 5-3150 and 1-3160) provided financial support. <= span style=3D'color:black;letter-spacing:-.15pt'>An earlier version of this disc= ussion was presented at the 6^th F-7 Workshop, Mexico City, 17-19 Feb. 1= 999. I thank P.M.L.A. Graça, D. Lashof, W.F. Laurance, S.V. Wilson and two anonymous reviewers for comments= .

REFERENCES<= /o:p>

[1] UN-FCCC (United Nations Fra= mework Convention on Climate Change). Kyoto Protocol to the United Nations Framewo= rk Convention on Climate Change, Document FCCC/CP/1997;7/Add1 (available at http://www.unfccc.de), 1997.

[2] Fearnside PM. Forests and global warming mitigation in Brazil: Opportunities in the Brazilian forest sector = for responses to global warming under the “Clean Development Mechanism.” Biomass and Bioenergy 1999;16:171-189.=

[3] Fearnside PM. Como o efeito estufa pode render dinheiro para= o Brasil. Ciência Hoje 1999;26(155):41-43.

[4] Schlamadinger B, Marland G. The Kyoto Protocol: Provisions and unresolved issues relevant= to land-use change and forestry. Environmental Science and Policy 1998;1:313-3= 27.

[5] Fearnsi= de PM. The potential of Brazil's forest sector for mitigating global warming under= the Kyoto Protocol's "Clean Development Mechanism." pp. 634-646. In: Kinsman JD, Mathai CV, Baer M., Holt E and Trexler M, editors. Global Clima= te Change: Science, Policy, and Mitigation/Adaptation Strategies. Proceedings = of the Second International Specialty Conference, Washington, DC, 13-15 October 1998. Sewickley, PA: Air & Waste Management Association (AWMA), 1999.

[6] Fearnside PM. Global warming response opt= ions in Brazil's forest sector: Comparison of project-level costs and benefits. Biomass and Bioenergy 1995;8(5):309-322.

[7] Fearnside PM. Greenhouse gas emissions from land use = change in Brazil’s Amazon region. pp. 231-249. In: Lal R, Kimble JM a= nd Stewart BA, editors. Global Climate Change and Tropical Ecosystems. Advance= s in Soil Science. Boca Raton, FL: CRC Press, 2000.

[8] Fearnside PM. Greenhouse gases = from deforestation in Brazilian Amazonia: Net committed emissions. Clim. Change 1997;35(3):321-360.

[9] Fearnside PM. <= /span>Global warming and tropical land-use change: Greenhouse gas emissions from biomass burning, decomposition and soils in forest conversion, shifting cultivation= and secondary vegetation. Clim. Change (in press).

[10] O Diário [Mogi d= as Cruzes] "Preços das terras estão caindo, afirma FGV.&quo= t; 25 January 1998, p. 5.

[11] Walsh MJ. Contributed statement, Brazil/U.S. Aspen Global Forum on the Kyoto Accords, Snowmass, C= O, 9-11 October 1998.

[12] Brazil, INPE (Instituto Nacional de Pesquisas Espaciais). Monitoramento da Floresta Amazônica Brasileira = por Satélite/ Monitoring of the Brazilian Amazon Forest by Satellite: 1998-1999. São José dos Campos, São Paulo, Brazil: INP= E. 2000.

[13] Raiffa H. Decision Analysis: Introductory Lectures on Choices under Uncertainty. Reading, MA: Addison-Wesley. 1968.

[14] van Kooten GC, Grainger A,= Ley E, Marland G, Solberg B. Conceptual issues related to carbon sequestration: Uncertainty and time. Critical Reviews in Environmental Science and Technol= ogy 1997;27(special):S65-S82.

[15] Canada. In: UN-FCCC, Subsi= diary Body on Scientific and Technical Advice (SBSTA), Report FCCC/SBSTA/1998/Misc.6/add1, additional submissions by the Parties (availab= le at http://www.unfccc.de), 1998.

<= o:p>

[16] Marland G, Schlamadinger B. Carbon sequestered, carbon displaced and the Kyoto context. Paper presented at The Association of Pulp and Paper Industries (TAPPI) International Environmental Conference a= nd Exhibit, 18-21 April 1999, Nashville, TN.

[17] Robinson JM. On uncertaint= y in the computation of global emissions from biomass burning. Clim. Change 1989;14:43-262.

[18] Alonso W. Predicting best = with imperfect data. Jour. of the Amer. Instit. of Planners 1968;34:248-255.

[19] Laurance WF, Fearns= ide PM, Laurance SG, Delamonica P, Lovejoy TE, Rankin-de-Merona JM, Chambers JQ, Gascon C. Relationship between soils on Amazon forest biomass: A landscape-scale study. Forest Ecol. and Manage. 1999;118(1-3):127-138.

[20] Fearnside PM. Environmental services as a strategy for sustainable development in rural Amazonia. Ecol. Econ. 1997;20(1):53-70.

[21] Fearnside PM. Biodiversity as an environmental service in Brazil's Amazonian forests: Risks, value and conservation. Environ. Conserv. 1999;26(4):305-32= 1

&n= bsp;

FIGURE LEGENDS

F= ig. 1 -- Effect of uncertainty reduction: under high uncertai= nty (A) the amount that can be credited (C₁) under a 95% certainty requirement is less than in the low-certainty case (B). The amount that can be credited is determined by 95% of the area under the curve being located to the right of= the point (C₁ in the high-uncertainty case and C₂ in the low-uncertainty case). The difference between the amounts credited (C₂-C₁) represents the potential gain from improving certainty.

<= o:p>

F= ig. 2 -- Carbon credit at different levels of uncertainty (coefficient of variation =3D cv) for an example with a mean benefit of 200= t C ha^-1 of avoided deforestation emissions and a requirement of 95% certainty. The right axis giv= es the percentage of full credit awarded.

<= o:p>

Fig. 3 --<= span style=3D'mso-tab-count:1'> Silvicultural plantations for charcoal: expected carbon benefit (C_p) is derived from benefit/ha (A) and number of hectares (B). Both parameters are relatively well understood, here assumed for illustration to have cv values of 5%. This is reflected in the expected carbon benefit from plantations (C= _p).

<= o:p>

Fig. 4 --<= span style=3D'mso-tab-count:1'> Avoided deforestation: expect= ed carbon benefit from deforestation avoidance (C_d) is derived from benefit/ha (A) and number of hectares (B).= The number of hectares (B) has a high probability associated with ze= ro or near-zero hectares, but also has a long tail representing small probabilities of avoiding large areas of deforestation. This pattern is reflected in the expected carbon benefit from deforestation avoidance (C_d). Differing shapes of curves for plantations and deforestation avoidance can reverse the outcome: although plantations (Fig. 3C) have a substantially lower expected benefit than avoi= ded deforestation (x̅_p&l= t;< x&= #773;_d), a 95% certainty requirement can reduce the = benefits assigned to avoided deforestation (Fig. 4C) to a level lower than the corresponding credit assigned to plantations (C_d < C_p).

F= ig. 5 -- Effect of different levels of required certainty on = the cost of carbon sequestration, using the example parameters. Deforestation avoidance, although = much cheaper than plantations for charcoal as a mitigation option if no certainty requirement is imposed, becomes worthless if even modest levels of certainty are required.

<= o:p>

Table 1. Deforestation= versus carbon value in Brazilian Amazonia

Land price (NPV of deforestation)	US$ 150/ha	Fundação Getúlio Vargas [10]
Net committed emission	194 t C ha^-1	Fearnside [7] updated from [8]
Value of carbon	US$ 5-35/t C	Walsh [11]
Value of avoided emission	US$ 970-6790/ha	Calculated from above
Advantage of deforestation avoidance	6-45 times	Calculated from above
Value of C released by 1998 deforestation (1.74 ´ 10⁶ha)	US$ 1.7-11.8 billion	Calculated from above and Brazil, INPE [12]

<= o:p>

Table 2. Hypothetical example of a Bayesian comparis= on of mitigation options
Carbon benefit (10⁶ ton-years)	Probability of outcome	Expected ton-year value

	Deforestation avoidance	Plantations	Deforestation avoidance	Plant= ations

0	0.850	0	0.0	0
10	0.050	1	0.5	10
100	0.050	0	5.0	0
1,000	0.025	0	25.0	0
10,000	0.020	0	200.0	0
100,000	0.005	0	500.0	0

Total	1	1	730.5	10

F= ig. 1

F= ig. 2

F= ig. 3

F= ig. 4

F= ig. 5

<= o:p>