SEQUESTERING CARBON IN NATURAL FORESTS

Closed tbrcsts cover about 3 billion hectares. or 2OVo of the world's total land area (excluding Antarctica). Forest plantations comprise less than 1% of this area. Natural forests range from the intensrvely managed ones ofCentral Europe and Scandinavia to the wild boreal forests of Russia and Canada and the deeplungles and dry iorests of the tropics. Numerous techniques-largely drawn from the ordinary repertoire of forest management-are available to enhance our ability of these forests to sequester and store C. Although the costs of sequestering additional C in these forests may be quite low (even in comparison with intensive plantation options), increased use of natural forests for this purpose ralses a host of concems about competing fbrest uses, biological risk, and the capacity to actually measure the incremental C sequestered. The problems of poverty, expanding populations, weak institutions, incomplete scientific knowledge, and climatic change itself will challenge the world's capacity to use natural forests as part of a COt control strategy.


INTRODUCTION
Forests serve humans in many ways. Indeed, some believe that forests have intrinsic values 0r even rights beyond those conferred by human use alone. This paper focuses on one of these uses: The capacity of natural forests to sequester CO, and storc C emitted as a waste product of our industrial activities. This paper specifically discusses the mechanisms behind such C uptake and storage activities and the key points related to estimating the costs of accumulating C in natural forests. Most of these techniques are classical methods of forest management, albeit developed to achieve different objectives. While precise quantification of both the physical and economic amounts involved is possible in specific instances, it is beyond the scope ofthis paper. Instead, we provide a road map so the reader may pursue these important details independently. Dixon (this volume), Sedjo et al. (1995), Hoen and Solberg (1994 and this volume) and Richards and Stokes (1995) critically review some of the alternatives discussed in this paper and quantify some of the biophysical impacts and economic costs. '1.'!.
Definition of "Natural Forest" For the purposes ofthis discussion the term "natural forests" refers to those lands currently occupied by closed forests or being regenerated to Lhe same or similar species as removed from the site after logging. We specifically exclude agricultural and other non-forested lands which are planted to tree species. We do however, consider those cases where open areas become forested through the ordinary processes of forest succession (e.g., occupation of abandoned farm lancls or lands laid bare Lry recent glaciation). We exclude non-forested peat lands' open woodlands and savannas" This definition probably encompasses vrrtually all of the world's foresrs (Dixon et al. i994). Human intervention in the development of natural forests ranges from intensive culture approxlmatlng the care lavished on agriculturai crops ro complete diffidence. Managed forests form one end of the spectrum and wild forests the other. Examples of the former include the industriai and multiple-use forests of Scandinavia and Central Europe; examples of'the latter include the remote borcal wildernesses in Canada and Russia, and the highly diverse, inaccessible tropical jungles in Brazil and zaire. Many parks and reserves lie towards thc latter end of the spectrum, but require management intervention even ifit is not for the purposes ofproducing industrial products. All ofthese landscapes share thc quality that the species composition of the forest overstory has been largely determined by nature" Globally, closed natural forests cover approximately 3 billion ha of the world's land (Sharma, 1992).Forty-three percent of the natural forests are found at low latitudes,32Vo at the high latitudes and25Vo at mid-latitudes (Dixon et al. l994a). Low-latitude forests are currently shrinking at a ratc of approximately 15 million ha annually. and thereby are a sourceof l.6Pgof Cperyear.Brazil. Incionesia, andZaircarelargeCOrsouroes. Iftropical deforestation and land-use change were slowed or eliminated, these forests wouid become a large sink of CO, (Brown et at. 1993). At present midand high-latitude natural forests sequester over 0.6 Pg c annually (Dixon et al. 1994a). canada, Russia, USA, China. and the USA all have large natural forest CO, sinks.

How Do Natural Forests Sequester Carbon?
All natural forests store C in phytomass, forest floor litter and soils (Cole, 1995). Phytomass includes all above and below grouncl components of the dominant overstory tree species as well as all other vegetation comprising the overstory, understory, and forest floor cover (including lichens and mosses). In natural forests-particularly those at the wild end of the spectrum-the noncommercial components of the forest may represent a signlficant and even dominant component of the standing biomass. The fbrest floor ranges from a dominant pool in many northern forests to a negligible one in parts of the tropics (Brown,l99l). These facts make it extremely difficutt to estimate C stocks from commercial inventories which-when they are available-of ten record only the merchantable bole volumes of the economic species. The lack of accurate allometric rclationships to estimate root, branch and foliage biomass from stem measurements exacerbates the problem. Thc non-merchantable biomass componcnts are important not only for their contribution to total biomass, but also in their role as inputs to the othcr C pools in the ecosystem.
Several factors distrnguish C scquesratron rn natural forests from that in plantatrons, although the distinctions are not absolute. In natural forest systems where extensive forestry management is practiced, the fraction of Net Primary Production (Npp) that enters the forest floor as litter, deacl snags or as dead-root turnover in the below-ground system ls generally much higher than it is in plantation forests" Indeed, capturing this mortality as merchantabie volume is one of the major objectives of plantation management. trn this respect. the managed natural forests of Europe tend to resemble plantations, and the distinction between plantation and natural fbrests trecomes very murkvL ife-cycie dynamrcs aiso <Jistrnguish natural forests and plantatrons. plantatron rnanagement usually seeks a"normal' forest agc-class distribution wrth an equar area ln eacrr age class up to the planned rotation age. Each cohon ls harvested as lt reaches the rotatron age and is replaced wrth a new cohorr which then becomes the youngest age class. Management of natural forests gencrally includes more chaotic, stochastic processes of mortality, disturbancc and succcssion. Storing C in forests by increasing the standing above-ground biomass may incrcase the risk of devastating disiurbances associated with fire, insects or diseases. Nonetheless, a C heneflt still accrues because the increase in atmospheric concentratlons of CC, is cielayed.
The response and feedbacks ol boreal, temperate anci troprcai f'orest systems to giobal climate change may be profound (Dixon et al. 1994b;Smith er al. l99l). Scenarios developed by coupling General Circulation Models (GCMs) of global climate with vegeta_ tlon response mociels generally rmply largc shrfts rn the distribution and productivily of forest systems (Smith ct al. 199 I ;Kauppr and Posch, 1988;(Songhen and Mendelsohn, this volume;Ferez et al', this volume). uncertarnty regarcling thc potentiai redistribution of forest systems in response to global climate change complicates our ability to predict future Co, sequestration by narural forest systems (King tgg3). gven if the GCMs are only partrally corrcct, the productivrty ofexisting natural fbrests will inevitably be affected, and the current C biogcochemrstry. both pools and fluxes. will aiso change (Drxon et al. 1994a,t. For the purposes of understanding thc rnteraction between the C pools rn the atm.sphere. Nct Ecosystcm Pro<lucrivrty (NEP) may be a more usefui conccpt than NpF. NEF refers to the net change ln soil. lrttcr and nhytomass. Becausc cliflercnt o,og"o.n.*,*, processcs control each oi these pools, it is possible for increases in phytomass to be negligible but for continucd sequeslration to occur in soil and Iitter pools. Such conditions may obtain in cool. moist micro-climatic conditions bclow thc canopies of old-growth temperate fbrests where gap-phase replaccmenr operates anci stancl-repiacrng distur'bances are rare events (Harmon et al.,lgg0). In boreal forcsts, break-up of thc overstory may be accompanicd by a decrcase in both the standing phytomass-and because of increased exposure and soil warmth-decreasecl forcst floor and soil C.
Peat is formed in natural forests in waterloggcd conclitions. Trees can survive on large areas at least in the carly phases of paludification (the peatland formation orocess). Such spectalized peatland piants as Sphaghunrfflosses transfer organic rnateriai releasecl from the overstory tnto thc peat layer. Peatland fbrcsts difl'er from other natural forests because CO, fixation excecds CO, release over time frames stretching fiom centuries to millennia. The boreal anci su-barctic peatlanils comprise a C pool estimated at 455 pg that has accumulated during thc post-glacial period at an average net rare of 96 Tg/yr (Gorham, 1991 ). while this is insignificant in an overall global C budget where the flux liom fossil-fuel combustion is at least two orders of magnitude greater, the relative contribution of peatlands to national C budgets is signilicant in somc northcrn countrics.

Sequestering C in the Economic System
Management of natural forests may reguire the rcmoval of trees, either as a planned component of a larger industrial activity, or as a means of sustaining specific ecosystem conditions in a park or forest reserve. Such rcmovals ordinarily enter the economic system. While the total amount of C fixcd in forest products pools globally is less than 20 Pg over the iast 50 years (Dixon, 1994a), it rnay bre critrcal rn the C builgets of those countries where forest products production rs high and consumption is low (e.g., Frnland, Sweden and Canada). This papen exphcitly treats the impact of removing the tree from the forest, but excludes the impacts of that matcrral on other parts of the economic system. This procedure obviously ignores the potential role of forests to substitute for fossil fuels in the energy production, or to substitute for such energy-intensive materiais as concrete or steel in construction applications. Such calculations are beyond the scope of our analysis but are extensrvely treated elsewhere (e.g., Marland, this volume;Matthews, 1995).
In considering forest C seguestration, it is important to note that an effective optiotl might be to optimize thc capacity of forests te remove CO, from the atmosphere but [o store the C in forms other than piant phytomass (i.e., as a substitute for fossil fuels. or tn longlived forest products). Such a straiegy would involve harvesting forests at a comparatively young age, and resultrng tn a lower standing inventory of timber than would occur if the forests themseives wcre used to scquester C (Harmon et al., 1990). We cliscuss the rmpact of such strategics on the fbrcsts themselves, but do not expand this analysis to include the impact on C off'set in thc energy systcm or other economic activitics.

1"4" Other Policies Affect C Sequestration in Natural Forests
Policies in spheres of public affairs othcr than forestry frcquently havc a profound impact on foresrs. For example. agricultural subsidies have long encouraged clearing of forested iand in many countries-both developed and developing. Such land-use converslon release a pulse of C rnto the armosphere, anci may also reduce the annual rate of C sequestration. As another example, in Scandinavia and rn parts of eastern North America it is well understood that N deposition associated with high levels of industrial air pollution fertilizes fbrests, increases lorest growth, and induces a higher level ol'C sequestral,ion (Eriksson and Johansson, 1993;Nilsson and Wiklund,1992). As a third example. international agreements on the conservation of biological diversity may requlre countrtes to sustain conditions in forested ecosystems which do not optimize their capacity to sequester C (Victor and Salt, 1995). Policy makers in these areas, seemingly remote from forests, should consider such impacts on forests (Sharma, 1992).

CATALOG OF TECHNIQUES: TWELVE WAYS TO SEQUESTER CARBON
The remainder of this paper catalogs various techniques for sequestcring C in natural forests. The first section below dcscribes twclve possibilities which we believe collectively cover the range of alternatives. In each case we describe the biophysical response mecha-nrsm and the possible impact on C sequestration. The next section covers the problems of determining the economic cost of each practice. In dealing with both the biophysical responses and the economic costs, thc discussion fosuses on the appropriate conceptual approach for evaluatrng the technique with references to spccific examples on the technlque reported in the literature.

Biophysical Responses
We are specifically interested in the total stock of C held in natural forests at any point in time. Annual accumulation of C can be measured either as the change in C stocks from one year to the next, or as the sum of the changes in the stocks of the individual components of the ecosystem. In considering the diff'ercnt forest management techniques in different locales. one or the othcr method may be the more straightforward to apply, but they are conceptually identrcal. The rotal amount of C sequestereii in natural forests simpiy equals the storage per hectare multiplied by the forested area. Sequestration strategies logically focus both on increasing lhe sroragc per hcctare and on rncrcaslng the forested area over what it would otherwise be (Winium et al., i993).

2.1.'1. Protect Against Fires
Fires, of both natural and anthropogcnic origin, play an important role in the life cycle of many natural forests. From the point of view of C storage, fires may be broadly categorized into two types: non-stand-replacing and stand-replacing. The first are commonly associated with low-intensity but relatrvely ficquent ground fircs. and relatrvely open woodland srrucrures. They oftcn result rn uneven-aged fbrests. Somc high-latitude forest specres (e.g. Pinus ponderosa) are well adapted to this fire reglme but most low-latrtude specles are not. Such fires produce relatively low immediate C releases, and-by clearing understory debris-may both clcanse the forest tloor of pathogens and remobilize nutrient for overstory utilization, thereby increasing subsequent C uptake. As a result. managers of some forests types prescribe fires as a regular practice. When fine fuels on the forest floor build to high enough ievels, however. crown fires can be sustained. and stand-repiacing fires rnay occur with very different consequenccs.
Stand-replacing fires are dramatic events having locally catastrophic effects leading to complete mortality of the overstory. Large. intense conflagrations are the domtnant type of fire in many boreal systcms (Apps and Kurz, 1993). As an example, in 1989, more than six million ha of boreal foresr 1an area 507o larger than Swrtzerland) burned rn Northern Saskatchewan rn a singic fire season^ The Grcat Black Dragon Fire rn Northern China ancjl the Russian Far East covered more than l0 million ha, and the resulting smoke plume was easily secn in satcllite images extending far to the East for many days (Salisbury, 1987;Stocks, 199 l). A large fire burned in Borneo in 1993 for many weeks before it was finally detccted fiom satellite telemetry.
The effects of thcsc fires are threcfold: (i) C is redistributed amongst the various ecosystem pools, (ii) C is released to the atmosphere as CO, and other C compounds including CO and CH4 (Levinc, 199 l ;Crutzen and Goldammer, 1993). and (iii) the forest structure is changed as the stand age is 'reset' and seral succession is restarted. While such fires entail major and immediate C released to the atmosphere (Cofer et al., 1991;Levine, 1991) the forests where such fires occur naturally are adapted to them and, indeed, are dependent on them for regeneration, removal of pests and disease vectors, and a host of other ecological nelationships.
The C releases associated with firc go beyond thc immcdiate pulse to lnclude subsequent emrssions from the non-combusted. decomposing, ciead biorrass ieft on site. For example, Auclarr and Carter (1993) estlmate that post-fire reieases may be as high as three times the immediate release. Some of this material becomes coarse woody debris (Harmon et al., 1986), or other forest floor reservoirs (Apps and Kurz, 1993;Dixon and Krankina, 1993).
While eliminating these large C releases is a potentially attractive option. suppression of fire may aiso merely open the way for othcr pathological agents which would normally have been kept at negligible levels by the pcriodic cleansrng actron of wild fire. For cxampie. rn Alaska the increased incidence of bark hreetle and other insects has been associated with the increaseci frequency of overmature forests (Dixon and Krankina,i 993). The trarge fires in the American West during the last decade suggest that the thorougtl suppresslon of'fire may be counter-productive because lt results in a build-up of fine fuels and the eventual. inevrtable, occurrence of a much higher rntensity fire. The resulting fires have significantly greatcr damagc to sitc fertility due to changes in soil structure and organic and nutrient capital than would havc otherwise occurrcd with a series of smaller, more frequent fires.
Although fires do periodically dcvastate low-latitudc in some regions (e.g., in Indonesla), rn most moist troprcal forests neither fucl Ioading nor fuel condition are conducive to large or lntense fires. Fires may piay a role rn these forests rn unusually dry weather patterns or after large-scale mortality caused by agents such as hurrrcanes or typhoons. Anthropogenic fires occur throughout low-latrtude forests, as forest burning ls a common management tool for resource-poor farmcrs. Although the size of these controlled fires is relatively small, their large number produce a globally significant pulse of greenhouse gases (Crutzen and Goldammer, 1993). During peak burning season in some low latitude countries such as Brazil, satellite sensors detect many thousands of small fires on any given day.

Protecting Against Disease, Pest Insects, and Other llerbivores
As with fire, these types of forest disturbance can be either stand replacing or endemic. Stand replacing events have the same three characteristics as stand-replacing fires. except that direct releases of C to the atmosphere is generally smaller (it occurs as respiration of the pathogens rather than as combustion products), and the transfer ofC to the forest floor conespondingly greater. Protection against these sorts of events is, like fire, problematic. For example, insect-induced stand mortality depends on present year weather conditions, previous year's insect populations and (of some insects) stand-age, type and health status (Volney, 1995;Galinsky and Witrowski, 1995). Avoidance of conditions which arc conducive to these disturbances is the best protection. Acrial applications of chemical or biological pesticides are generally effective in sustaining forests in a living condition iong enough to harvest wood, but are generally not effective in eliminating inscct populations altogethcr.
Endemic impacts of insect, discase and herbivores, may result in direct reduction oi NPP by reducing the ner incremenr of fbrest phytomass" The effecr on NEP and C sequestration is not clear and dcpends on many factors. rncluding forest age and disturbance type (Kurz et al., 1995). In relativcly mature stands, phytomass decrements assoclated with the endemic disturbances largely rcappear as increased C on the forest floor and soils. In these cases the net C balance is determined by the relative rate of biomass replacement and fbrest floor decomposition. The type of litter created is also a f'actor. (For exampie, ciead roots, snags and other coarse woody debris associated with individual tree mortality have longcr turnover times in the cooler sheltered micro-climatic conditions below a fully developcd crown than does lcaf and branch littcr fall following attacks of defoliators). In younger, regenerating stands, the effect of herbivory in particular may cause a regeneration delay which has a temporary, and srnall, impact on C storage, tlut may have important economic repercussions. Some level of cndemic insect, disease, and/or henbrvory disturbance affects all forests. The fraction of NPF consumcd by these agenrs has nor. to our knowiedge, been cstrmated except in a few specrfic lnstanccs.
Laboratory studies havc demonstrated that insect herbivory of plants grown in high CO, environments is dramatically incrcased (Drake 1992). As the proportion of N declines relative to C in plants grown in high CO, conditions, insects and other grazers increase their consumption to compensate tbr the loss of proteln" In the future, management of natural forests as CO, srnks wili need to be adjusted to compensate for this increase in defoliatron. If insects and other forcst pests adapt a more aggressrve consumptlon pattern in a future global ciimate. net COz sequestration of future forests may be less than it is ln current conditions (Drxon et ai. 1995).

Salvage Dead and Dying Trees
Natural disturbances leave dead or dying trees that have both positive aspects (e.g., animal habitat, landscape diversity) and negative oncs (e.g., increased fucl loadings, loss of vaiuabie ttmbcr, reduceci NPP). {n some cascs. where hoth the infrastructure anctr the markets exlst, tt rs possrbic to salvage somo or all oi this deaci or dying timber to recover valuable wood, reciuce the risk of subsequent contagious disturbancc by insects, fire or pathogens associatcd with the accumulation oi logging dcbris, and rehabilitate the site for subsequcnt regeneration.
By moving decomposing C from the natural system to the economic system, it may be possible to increasc the net C storage associated with the given piece of land. The extent to which there is a net C benefit in salvage operations depends on several factors. First, is the turnover rate of C transferred to forest products (including all process wastes) greater than the C turnover ratc on site (Hendrikson, 1990)'/ Second. to what extent is site regrowth increased relative to the untreated condition'/ Third, how much fbssil fuel is used in the salvage and product nranufacture and distribution opcrations'l Natural disturbances may af'lbct very large areas. Therc arc however two significant banicrs to salvage operations: operability and marketability. Lack of railroads, highways and waterways lirnit thc access to many natural fbrests. Hazardous conditions that often s29 exlst after iarger scale disturbances also restrict salvage operations. Saivaged timber has limited uses. For example, pulp and paper mills are understandably reiuctant to acceDi wood charre<l by fires. Concern for rnadvertent importatron of insect and disease greatly restricts the markets for trmber salvaged from outbreaks of these pathogens" Bioenergy might tre a potentially vrable market opportunrty, but access and infrastructure would appear to restrict this use to hrghiy specrfic instances.

Change Rotations
Cooper ( 1983) demonstrated that converting a fbrest region of fully stocked mature stands into a maximum sustained yield fbrest dccreases the standing stock by about two thrrds, and thus ncleases CO, into the atmosphere. Harmon et al. (1990) simuiated conversion of oldgrowth forests into young fast-growing forests and concluded that on-sitc C storage does not approach old-growth capacity for at least 200 years. They estimated that the conversion of five million hectares of old-growth forests to younger plantations in western Oregon and Washington in the last 100 years has added 1.5-1.8 Pg C into the atmosphere.
Natural forests in disturbance-domrnateci systems inciude immature as well as mature stands. Thereforc, logging in natural fbrests does not always result in such a dramatic decrease of C pool as Harmon et al. ( 1990) demonstrated for the US Pacific Northwest. In some cases iogging can merely substitute the natural disturbances as the rnechanism by which CO, is released into the atmosphere. trn the Nordic countries. for exampie, the pool of C in forest vegetation has increased this century despite continuous and steadily lncreasing logging (e.g" Kauppi et al" 1995). trn contrast. trogging of'troprcal forests ln Peninsular Maiaysia has resuiteci in a dramatrc reduction rn C storage over the past 30 years (Brown, 1991).
In a steady-state system according to Dewar's (1990) model, a very long rotation can promotc C sequcstration, if the products decay quickly. But a short rotation can also serve the same purpose grven high biomass yrcld, and a iong product life-time and/or effective rcplaccmcnt of the consumptron of fossil fuels. Most short rotations systcms are "plantations" (and therefore beyond thc scopc of this paper), yet some coppice systems can be classified as "naturai forest". Dewar (1990) fbrmuiated a model that describes C storage in a forest and its timber products as functron of the fbrest growth curve. the rotation period and the C retention curves in timber products" He showed that. whcn the lbrcst is managed for a maximum yield of biomass, the contribution of asymptotic C storage liom timber products is about 2.5Dn* timcs thc contribution from living trees, where D is the characteristic decay time for conversion of timbcr products to C dioxide, and T* is the normal rotation period for maximum yield. When D/Tx <tr, as the rotation period is increased indefinitely, the asymptotlc level of C storage lncreases monotonicaily toward the value of C content of livrng rrecs at matunty. But when D/T'k > | there rs a finrte, optrmal rotatlon perioci, greater than T*, fbr which asymptotic C storage ls greater than the C content of trees at matunty.
Either lengthening or shortening the rotation age will increase the C stocks, depending on the initial conditions (e.g. age structure) of the fbrest, harvesting methods, silviculture and, in particular, the fate of harvcstcd C (Schlamadingcr and Marland, 1 995; Marland and Schlamadrnger, this volume). In some cascs, making long-lived products from wood grown s30 il i i in short-rotation fbrests can sequester C effectively. Short-rotation management of woody crops for production of broenergy fuel stocks, can repiace or offset the need for fossil fuels (Sampson et al., 1993) Although a careful analysis is needed for each specific case, as a rule of thumb that prolonging rotations in natural forests will generally contribute to C sequestration. Converting natural forests to short rotation systems can make positive sequestration contributions mainly in the longer term, that is, in the period beyond 20 to 50 years from now2 .'1.5. Control Stand Densitv Thrnnrng is wldely used to alter the size oi individuai stcms in a stand. the trmrng of therr availability, and the overall amount of merchantabie timber available. In some regions, stands are thinned thrcc to five times during the rotation in order to collcct material which otherwise would decay in the forest. When thinnings provide sufficient space, the remaining trees grow to larger dimensions more quickly and thus more valuable per cubic meter harvested. The trend of dccreasing thinnings and an increasing share of final cuttings has prevaiied in Europc srnce the 1970s because loggrng costs are higher in thinnings than in clear auttlng.
Thrnnrng affccts C sequestratron and storage rn several ways. X-ow thinnrngs utihze small stems which would otherwlse decay and reiease C back to the atmosphere. High thinnrngs which reduce overall forest growth rates may also help sequester C by providing more of the total volume in longer-lived solid wood products such as lumber. Thinning-towaste will generally bc neutral or negative in terms of carbon sequestration.

Enhance Available Nutrients
Water and nutrient availability control fbrest growth in most parts of the world (Nambiar and Sands, 1993). As a result, fertilization and irrigation will gcnerally enhance tbrest NPP. For example, in Portugal and Brazil experimental Eucalyptus plots have reachcd annual production of up to 40 Mg/halyr (Campinhos, l99 l). In tropical Australia, Fife and Nambiar ( 1995) found clear responses of N fertilization on the growth (up to 99Vo growth increase) of Pinus radiata grown in a very dry environment. The authors also concluded that improvcd N status of the trees leads to morc cfficient water utilization. In Sweden a series of fcrtilization experiments have shown clear growth response duc to N fertilization on both Pinus anrJ Picea abies (Tamm 1992;Linder, 1995). In the most polluted parts of south west Sweden biomass production in Picea abies was doubled due to liquid fertilization with irrrgation in 30 year old Picea abies (Nilsson L.O: this volume). Davey (1990) and Weetman et ai. (1987) cite evidence for posrtrve growth responses due to fertilization in both the US and Canada. Cole (1995) and Henry et al. (1994) showed a sustainable increase in basal area of 4-5 times after application of nutrient rich sewage sludge in a 45 year old Douglas fir lbrest. Allen (1995) has noted similar growth responses with liquid fertilization and inigation. Fertilization may also increase C storage in the soil (Berd6n 1994. Nilsson 1995. s31 in most cases N rs the malor nutnent eiement cxplatntng the productlon lncrease. However, other nutrients-mostly P, K, Mg and such micronutrients as B, Cu, Mn or Znlimit NPP in somc areas. For example, in New Zealand, soil fertilization with P, B or Zn <lramatically lncreases the productivity of Pinus radiata and various Eucalyptus species piantations (Nambiar and Sancls, 1993). Similar vlgorous rcsponscs to forest fertiiization are obscrveci ln Afrrca ancl Southeast Asia in stands of Tectona, Shorea and Leucaena iDrxon ct al. i994b).

2.1"7" Controf the Water Table
About I L5 million hectarcs of borcal peat lands have been drained in the world in order to lower thc water table ancl thereby promote forest growth (Gorham 199 1). Almost all of this area can be classified as "natural forests" given the dcfinition used in this paper' Although drarnage ot'forested peatlands promotes forest biomass growth, but the net C .benefit rs partly offset by increascd resplration in the exposeci peat and (Zoltai and Martikarnen, 1995). [Jndrained northern peat trands annually remove 76 Tg C from the atmosphcre, whero as the drarned peat lands releasc 8.5 Tg C annually (Gorham 1991). Zoltar and Martikaincn (i995) find that lowering water tables by 20-30 cm increases CO, emissrons t:y X50-200ok They aiso polnt out that a more lmportant. but difficult to predict, consequcnce of altcring the water tablc ls a change tn methane (CHo) productlon, an even more potent greenhouse gas than COr.
Irrigating dry forcsts would have a positive impact on C sequestration, by increasing storagc in all of the subsystems-soil, vegctation, and products. Danish (Beier ct al. 1995) and Swe<lish (Lindcr. 1995;Nilsson. 1995) cxperimcnts havc shown a large growth increase due ro rrngatron alone dcspite a rclatively hurnid climate. Water also limits NPP and NEP in many low-latitude foresr systems (Jain ct al", 1989). Loss oi forest systems and desertification is a global problem of growing significance (Sharma,tr992). trrrtgatton and other silvrcultural tcchniques can be empioyed to sttmuiatc NEP, but the economlc costs of these practices ancl the lack of appropriate infrastructure makc such efforts virtually impossible. Many trec species are well adapted to dry conditions, and reclamation of substandard soils and desertifled sites is possible systems (Jain et al., 1989). Biomass oroduction on these low-latitude sites can be l5-20lhectare. Given the large area of dryland natural fbrcsts. sttmulation of CO, productron in these ecosystcms is a giobally slgnificant opportunity (Dixon et ai., 1994b).

2.1-8. Select Useful Species and Genotypes
The natural variation in NPP between provenances of the same tree species origrnating from a relativcly small geographical area may be substantial,(e.g., Fife and Nambiar, 1995). Thus. selcction of provenanccs well adapted to the sitc and expccted climate may improve the NPP.
Tree improvement with some gencra, including Pinus, Populus, Picea, Eucalyptus, Leucaena, Tectona, has dramatically irnproved thc yield and othcr favourable commerical attributcs ovcr the last 50 years. However, it is not clear to what cxtent these increases in s32 yield stem fiom increased NPP of individual trecs, or simply reallocation of C from the noncommercral brelow-ground portions to stcmwood.
A basic factor fbr improving NPP is a proper choice of tree spccics. Forest growth has declined in several locations responding to a poor choice of spccies (e.g. the decision to plant Picea abies on hardwood sites in central Europe). Because N limits NPP in many ecosystems. the rntroduction of such N fixing species like Alnus, at high latitudes. or 'tr ueccaena. at low latrtudes, would improve the nutnent status of the site, which wor-lid rn turn xmprove (Jarn er al., [989). A large number of multrpurposetree species have been identified which can be employed to rncrease C sequestration and storage, or to improve site productivity on a sustained basis (Burley and Stewart, 1985).
Finally, enrichment planting below thc main canopy can increase NEP in cases where one or more aspects of the growing sites arc undcr utilized. This might occur in naturai stands which are not fuily stocked as a result of various non-stand-replacing disturbances. This technrque could aiso be eff'ectrvc if shadc tolerant species were planted beiow less tolerant oncs. So, for example, mixed stands of Tsuga heterophl,lh and Pseudotsuga mensiesii have higher abovc-ground productivity than do pure stands of cither species aione.

Reduce Regeneration Delays
Recovery of disturbed natural forests-whethcr by harvcst, insect, fire, disease, blowdown, drought or any of a host of other factors-dcpends on both the availability of viable seed sources anci surtability of slte conditrons. There rs often a period of time-the regeneration delay-followrng disturbance where these critena arc nof met. On some sltes, particularlylow quaiity ones" losses of organrc matcnal and nutrrent caprtal by ieaching or eroslon can both degrade the site potcntial and delay its realizatron. In somc high quality sites, more successful opportunrstrc species. such as grasses (e.g., Calamagrostis sp., ln many northern systems) may tnvade the srte and tcmporarily prevent the establishment of trccs and the longer-term accumulatron ol'phytomass. In both cases. such delays may be exaccrbated by such factors as incrcased herbivory, and changcs in soil structure or water table.
Appropriate choice of harvest timing and mcthod can prevent regeneration dclays. If regeneration is not prompt. such mitigation tcchniqucs as in-planting, site preparation, and nutrient suppiements can be ei'fectrve in speeding lull stand occupancy wlth tree spccies {Winjum et al.. n993) In extrcme cases the site may fail to recover to its original forest cover and remain in anon-rcgcnerated state (Dixon ct al., 1994b). The C consequences arc the same as ianduse changes. In less cxtreme cases, the regcncratlon dciay rntroduces a corresponding delay in C accumulation"

Select an Appropriate Harvest Method
The current techniques fbr harvesting trees varies from the ancient practice of hand felling individual trees and yarding by gravity or draft animals to highly mechanized, capital intensive methods. The fossil C consumption per unit of C harvested is low even in the s33 highly mechanized methods. Therefore, the fossil fuel use is not decisive in determining the impact of harvesttng methoci on C sequestratton.
Analogously to fires and pest outbreaks. iogging methods can be divided into stand conserving and stand-replacing ones. Clearcutting ls the most straight-forward technique lbr replacrng a stand. However. sced-tree and shelterwood systems also remove rnost of the trees with a few ieft bchind to provide the seeds and/or shelter for the new stand. Selecttve logging maintains fbrest cover while gradually replacing the stand. It can be used in some cases in a way that therc never occurs a stand replaccment phase: individual trees rather than the whole stanci arc replaceci and rotated. Normally, trccs are plantcd only in connection with stand-repiacing harvest rncthods, but this practicc is also appropriate as an enrichment technique as described above.
The choice of thc harvesting method can have impacts on C scquestration, although such differences are poorly documented in literature. Olsson ( | 995) found that clear felling high-latrtude forests resuits rn signifrcant losses of C and N from the humus layer down to 2Cl cm depth" Harvesting and the associated disturbance also influence C pools and flux in low-latrtude forests (Brown et al.,199 l). Technrques which have been emDloyed to conserve and sequester C in Malaysian forests inciude (i) preservation of non-harvestecl trees and associated vcgetatlon by selective logging (versus clearcutting), (ii) before harvesting takes placc. cutting vincs which link boles of trccs and thereby reducing the felling losses of non-target trces. and (iii) using low-impact harvesting systems which reciuce soil disturbancc (Dixon et al., 1993). Thc kcy to analyzing the impact is to measure and monltor the C pool in vegetatron and soils over a fairly long period of timc' 2.1.11. Manage Logging Residues {n soms places, loggrng residues fiom forest managcmcnt are, to a large extcnt, already being used as bioenergy (Sampson et al., 1993). In Finland and Sweden, bioenergy represents as much as l7-l8o/a of all energy consumption when pulpmill residuals are includcd . Biocnergy substitutcs directly for fbssil fuels, in essence storing the C that would have been releascd in oil, gas or coal rescrves.
Whole-tree harvesting -where branches, needlcs and possibly root and stumps are removecl--can lncrease the amount of the forest available ibr bioenergy. However. this treatment also reduces the pool of nutrients in the ecosystem, particularly in coniferous foresrs Wiklund , 1994: 1995). On sites where nutrient availability limits NPP or NEP, this may result in recluccd fixation of C (Olsson 1995). Fertilization can counteract this cffect, but the carbon costs may not be favourable.

2.1.12-Establish, Maintain and Manage Reserves
Dcforestation currcntly contributes about 2/7s of all net C emissions (Dixon et al., 1994a:' Houghton,etal., 1992.Asaconscqucncc.slowingthcrateof dcforestationwillreduceC emissions as well as incrcasing the capacity of the globe to sequcster C. Two questions arise with this strategy: can suitable human intcrvention increase thc stocks held in reserves? And what is the long-tcrm fate of thcse stocks'? Increases in the C stocks in rescrves can s34 be accomplishcd by rncreaslng Lhe NEP oi exrstrng reserves or ny sustarnrng or increasing the areal extent of such reserves. At a stand-levei. a set-aside fbrest reserve may accumulate C while in a growth phase, but as maturity is reachecl, growth decreases. resprration rncreases and net C sequestratrori slows and may even decline as stand-breakup occurs. ln many temperate and troprcal forests, the fbrest siand may be maintained indefinitely in a close<i canopy, marure srare through some form of gap-phase replacement (Botkin 1992;Shugart, 1984;Leemans and Prentice, 1985). In the absencc of stand-replacing disturbances, relatively smooth changes in C accumulation make take place in responsc to changes in nutrient or environmental conditions. In disturbance-regulated forests, such as thc boreal forest, however. C stocks rnay rlse and fall as largc C pulses are assocrated with ciisturbancc events. regeneration and the regrowth followrng rt. As discussed prcvrously. controlling these swrngs-particulariy in isoiated reserves-may be diificult.
The overriding point is this: C can retained by avoiding deforestation only if the C poois in the reserves are explicitly managecl. Management oi'c pools may or may not be compatible with other desrrable objectrves for forest. reserves such as preservation of biological diversity. Creatton of a reserve might also simpiy displace the competing land users to a dif'ferent iocation, wtth consequent impact on C stocks and sequestration rates. To compute the total impact of the reserve requires netting out these offsite effects.

2"2" Economic Costs
Most anaiysts agrec that it is most useful to describe the costs of C sequestratron ln rerms of real monetary units (e.g., US$ fixed at a specific date) per unit of C removed from the atmosphere (and perhaps per unit time). Howevcr, some confusion surrounds the exact way to measure this quantity, both for the numerator and for the dcnominator. Furthermore, most of the alternatives discussed above are continuous, in that to sequester more C the alternattve can be more intensrvely applied (e.g., rnore fertilizer means higher growth rates; more protectlon mcans lcss C lost to f irc). The optimal scale of the alternative must also be determincd.

2.2.'1" Measuring Froject Costs
The economist's concept ofcost for a specific activity refers to the value ofthe resources used if applied to their next best use. Where markets exist. market prices provide a useful guidepost of costs, so, for example, local wage rates measure the cost of labour used in a project and local machinery costs measure the costs of mechanized activities. In cases where markets are not well developed-either as a matter of government policy or as a matter of weak property or other institutions necessary for markets-the analyst must estimate the shadow prices of resource inputs in order to quantify the costs of C sequestration.
Ideally the total cost of'a project to sequestcr C should includc both the direct costs of the project and the opportunity costs ofresourccs used. A fail-safe approach to comprehensrve cost analysis is thc "with and without" approach. With this procedure the analyst determlnes the totai economlc return of'an actrvlty without the specific actlon targeted to scquesrer C" and then thc total economrc return of the actron with the C sequestertng actlvlty lnciudcci. The cost of'C sequcstratron is then the difference tn the returns. For example, the cost of sequcstering C by extending forest rotations is the difference in the nct present values of the shorter and longer rotations, while accounting for a// costs and benefits. That is, the costs of a C scquestration projcct should be the nel costs. Any incremental bcncflts assocrated with thc pro1ect not related to C sequestration should be subtracted from the prolect's dircct costs. Examples include the recreation or aesthettc benefits of natural reserves, or thc value of timber prociuccci when management practlces are intensified.
Bccause most natural fbrcst management options cxtend across a long time period, the choice of discount ratc will generally bc quite consequentlal tn detenninrng the oost of C ,iequesrratlon. Ideally the discount nate should reflcct socrety's pret'erences tbr consumptton rn one period compareci with the next. One measure, albeit rmperfcct. is the reai cost of long-tcrm government debt. In fast-growing devcloping countrics, the discount rate is logically higher than it is fbr more slowly growing developed countries. In the latter, discount rates of 2-4Va are generally justifiable, while rates as high as loo/o might be appropriate for thc lbrmer. Because of the uncertainty surrounding the choice of discount rate-aniJ its conseguences tor tho analysis-good analytrcal practice always computes the prcscnt value of project costs ibr a range of valucs.

2.2"2" Measuring C Sequestration
Most studies measurc thc amount oi"C sequestercd as an average annual figure, and one of two approaches can be fbllowed. In some cases, it will be possible to estimate the actual amount of C flxcd cach year (as, for examplc, bcing proportional to a conventional timber yield table). Alternatively, just as in the "with and without" analysis of economic costs, the amount of C fixeci as the differcncc be tween stock of C without the activity and the stock with the activity divided by thc number of years the pro"jcct is in place. These approaches assumc that the present valuc oi the damage of a untt of C (the shadow price for C) is the same regardless o[ when it is emitted. This assumption is unlikely to be true. Abatcment, adaptatron and rnitigation technologies will change over time. Damages may be nonlinear in the totai amount ol CO, in the atmosphere. All of these considerations suggcst that, just as the economic costs of a C sequestration project must be discountcd to determine their present value, thc annual C emissions should be discounted (Richards, this volumc;Richards and Stokes. 1995). Because of the complexities and ambiguities of this procedurc, however, most analysis measure the costs of C scquestration as the present vaiue of all costs (or their annualizcd equivalent) divided by the average annual amount oi C sequestered over the proJcct's iife. Becausc this procedure docs not account fbr thc temporal pattern of C relcascs, we suggest that the discounted sum of C fiows (using the same discount ratc as for costs) should also be reportcd.
Finaliy. the extant iirerature focuses on ex ante forecasts of the amount of C which would be sequestcrcd under various assumptions. As C sequcstration policies are put in placc, it will become necessary to measure the ex post quantities of C that actually are being sequestercd. Such measurcmcnts will be required to monitor progress towards targets for s36 Marginal Benefits MCr MCo Protecting the Forest. Figure I shorvs the standard analysis of the optimal level of forest protectron. net emission reductions, and to verify any multiparty agreements about C offsets or 'Joint rmpiementation pro1ects". The level of precision needeci for such measurements in natural forests currently exceeds that whrch ecosystem screntists can produce at a reasonable cost. We must either develop quick, inexpensivc methods or be satisfied with indirect estimates from modeling procedures.

Optimal Project Scale
Because the biophysically fcasible management alternatives discussed above are ordinary forest management practices. considcrable past anaiysis has centred on therr optimal application. In many cases, these analytical methods can be extended in a straightforward way to incorporate the benefits of sequestering C and the cost of emitting it. To fix ideas, it is helpful to imagine that the C sequestration policy is implemented via a tax on C emissions and a subsidy for C sequestration, both levied on a per unit basis. The twelve alternatives discussed above fall into six general categories of economic concern.
Without considering the value of C sequestration, the optimal level occurs at P*0 where the marginal cost of another unit of protection (perhaps measured as total forest area which is managed under a protection program) just equals the marginal benefits of avoided losses (e.g. the value of thc timber and resrcational values protected). Now consider the case where releasing C to the atmosphere carries a cost. Protection against fires, insects and diseases produces additional benefits equal to the upward shift from MVO to MV1. The optimal level of protection increases to P* l. Operationally this means that roads would be maintained further into the extensive margin of the forest. more people would be held on aiert status during fire season, and so on. Finally, note that the margrnal costs ofprotectron mrght aiso rise in response both to a changed climate and to effbrts to store more C in the forests themseives. While the optimai level oi protectlon is iikeiy to increase. it wiil still probably not be economically efficient to exclude devastating fires or pest events from all forested areas, even if were physically possible and ecologically desirable to do so.
Fire monitoring and management in natural forest systems is the most efficient near-'rerrn to conservc and sequester C in forest systems. Perhaps this point is best illustrated in boreal fbrest systems, where frequcnt occurrence of wildfire results in globally signrficant pulses of COr to the atmospherc (Drxon and Krankina 1992). Within the Russian Federatton, forest fire monitonng systems (e.g., aircraft. satellites) in East or West Siberia or the Far East have not proven sufficiently efflcrent to support a lire management system. C emissions from wildfire. including direct and post-fire release are estimated to be approximately 0.2PgC annually. Improvement in fbrest fire monitoring and management systems and its expansion to all forest land rn Russra can potentially reduce the are burned annually by 20Vo and thus conserve 0.05 Pg C. Dixon and Krankina ( 1993) estimate that this C can be conserved in forest systems fbr much less than $1.00 per hectare.
Rotation Length and the Optimal Amount of Growing Stock. Policies designed to sequester C in the forests themselves will logically increase the optimai forest rotation (van Kooten et al., 1995;Plantinga andBirdscy. 1995, Hoen andSolberg, 1994) and the optrmal airrount of growing stock in uncven-aged managemcnt systerns^ The troglc of this ls captured in the firsrorcier condition Ibr the optirnal rotation in the presence of a C eternality (van Kooten. et al. 1995). In comparison wrth the usual Faustmann case. the presence of the C eternality creates an addrtionai bcnefit of holding the stand longer because of the vaiuc of the C sequestercd. At the same time, the C externally increases the costs of harvesting the stand bccause of the C cmitted both from the ecosystem and from the industrial activitics involved with processing the harvested timber. Both factors work to lengthen rotation ages, and the irnpact may be quite large (van Kooten et al., 1995). By similar logic, the optimal levcl of growing stock in a selcction fbrest would increase in the presence ot a C externally.
Stocking Control. Thinning perfbrms three cconomic functions-harvesting trees that would otherwise succumb to suppression, accelcrating the timing of cash flows from the stand, and increasing the picce size of the remaining trees. Thcse benefits are purchased at the cost of a possible rcduction in final harvcst volumes. While the solution to the optimal thinning problcm is quite complex, thc presence of C externalities should gcnerally favour repeatcd light thinnings, with some anticipated loss in final harvest volumes. The repeated light thinnings would harvest trecs whosc death and decay would contribute directly to C emissions. As long as thc thinning were used for fossil fuel substitution or products with a longer lifc than that ofthe dead trces left to decay in thc forest, their harvest would reduce or delay total C emissions. Thinnings heavy enough to reduce final harvest volume should lead to higher piece sizes, and a larger fiaction ofthe final harvest going into products with long useful lives. The tradcoff betwecn the annual C rcduction services of the unthinned s38 stand and the higher amounts of C fixed when stand densities are lower must be computed for each specific case. Thinnings to waste should clearly be avoided (Hoen and Solberg, 1994), suggesting wider initial stockings than in situations where precommercial thinning is practrced.
Changes in Management Intensity. A number of management practices, including fertilization, and water control, are conceptually similar. The general problem of the
As with Figure l, the subscripts 0 refcr to thc choicc of managcmcnt intensity in the absence oi the C externally. The optimal levcl occurs whcrc the cost of another unit of management intensity MCO (e.9., thc cost of a tonnc of N fcrtilizcr) just cquals the marginal benefit of that unit of (e.g., the additional timbcr at thc cnd of thc rotation procluced as a result of fcrtilization multiplied by the value of the timber). The C eternality will increase the marginal benefit of the management intensification, but may also increase the cost to the extent that C is released by the management intcnsification (e.g., rn the production of fertilizers). The net effect is probably an increasc in managcmcnt intcnsity, although some specific circumstanccs might lead to the opposite result.
Utilization of Salvage and Logging Residues. The limit of utilization is defined by the smallest size and lowcst quality of timber that just has enough value to "pay its way out of the woods." C sequestration policies will raise the value of small, low-quality logs either for fossil fuel offsets or as a store of C in the product sector. If such policies were implemented as a C tax on cmissions and subsidy to scqucstration, forcst managers would utilize smaller stems rather than leave them as logging debris. $/ha s39 Incremental Value of Sequested Carbon Land-use. The margins of land use occur where the value of one use (e.g. for agriculture) just equals the value for another (e.g., forestry). C sequestration policies will increase the economic value of land-uses which store more C over those which store less. Hence the tntensive margin should shift inward, with more land dedicated to forestry and less to agnculture and other uses. At the same time, the benefits of'managlng remote sites will rncrease, both through the value of products stored in the economrc system via harvesting and the tncreased storage made possible by fbrest management. As a consequence. the extensrve margln of utilization will shift outwarci. Masera (this voiumel iDrovlcies an rnterestlng oxample of these effects rn Mexrco.
Defbrestatron is, in eff'ect, an outward shift of the intensrve margin of land-use. where a more lntenslve use such as agrrcuiture displaces the lcss rntensive forest use. In some cases, policies such as agricultural subsidies drive these changes. Policies targeted on sequestering C will work to offset these margin shifts, and help halt deforestation. In these circumstances it is logical also to examine how policics outside the forest sector are influencrng sucih margrn shifts.

DISCUSSION AND CONCLUSIONS
"Natural forests" comprise thosc iands currently occupied by closed forests or being regenerated to the same or similar specres as removcd from the site. The vast majority of the world's forest fall wrthin this definition (perhaps over95Vo). They range from intensively managed natural forests of Central Europe and Scandinavia to the wild forests of the Russia and Canada and deep jungles of the tropics. Many indigenous people call these forests home. Much of the world's biological diversity is harboured within them.
There are numerous techntques for enhancing the capacity ofthese forests to sequester C" While the literature orovides useful analysis of some of these techniques (e.g. changes in rotatton iength, fertilization, stocklng control, forest reserves), we have identrfied a somewhat longer list of possibilities whrch ment attcntion.
These are ordinary techniques of forest management already used to achieve other socictal objectives. Viewed in this context, CO, sequestration is but one of the multiple outputs of the forest. The C impact and cost of cach of these techniques depends heavily on local conditions, perhaps evcn more so than is the case with afforestation and plantation management options because natural forest management begins by dealing with an existing growing stock. Thc litcrature suggests a range ofcosts for C sequestration in natural forests, from cheap to dear. Forest sector 'Joint implementation" projects, mostly established or sponsored by U.S. or European electric utilities, propose to conserve or sequester several Tg of C at initial project costs of less than $5.00 per Mg C (Dixon et al., 1993). Analysis of these proJects reveals that the long-term cosis of CO2 sequestratlon in natural forests are generally low or negative (e.g., yreid net benefitsi.
Despite the fact that the biophysical rcsponses will differ widely as one moves among the various tropical, temperate and boreal forests, a few general principles seem clear: Accumulating higher levels of growing stock is apt to incrcase the biophysical risk of holding stands managed expressly with C sequestration in mind, Because of poor access and their extensive nature, natural forest are already subject to a host of biotic and abiotic stresses. and climatic chanse will to add to these: s40 lncreasrng atmospheric concentrations of co2 and climatic change will affect the amount of C stored in these forests; while it is possible in many cases to estimate ex ante the amount of c that will be sequestered under various management strategies, it will be extremely difTicult to measure and verify these quantities in the field.
It rs conceptually straightforwarci to cletermine the incremental cost of addins C sequestratlon as an objective of managing natural forests. Similarly, the literature cont;ins many reports of attempts to compute the present value of these incremental costs of C sequestered on an average annual basis. However, the temporal pattern ofC sequestration in natural forests will generally be complex, so all of the pioblems of ,.c discounting,, will Occur' As is the case with plantations, correct estrmates of C sequestration will inciude that held in the economic system, both as fossil-fuel offsets and as iong-lived products. While the economic concepts for measuring the net costs of C sequestralon programs are clear, application of these concepts to specific cases rnay face daunting empirical obstacles related to the quality and availability of data, and the high cost of studfes needed to estimare the value of such elusive benefits as the option varue of preservation" Weak physrcal and institutional infrastructure will impede rmplementing C sequestratton policies in many natural fbrests. In some places, traditional use of these forests by aboriginal people raises serious ethical questions about storing the offal from industrial soclety there' In other places, lack of physical access proscribes management intervention. Three general recommendations for policy attentron seem approprrate: ' attencl to how policies from other spheres ofsocietai concern influence the capacity of natural forests to sequester C (c.g. agncultural subrsidies encourage deforestation): ' look for opportunities to comtrine the solution of several problems into a single policy (e"g. use rnunicrpal waste water to fertilize forests): n remember that the techniques for using natural forests to sequester C are quite traditional forest management practices. As a consequence, quit" u lot is known about policy implementation, even if it is difficult.
Even in the absence of the problems associated with climatic change, the problems of poverty, weak political institutions, and expanding populations challerige our policies for the management of natural forests throughout the world. The new requiiement tbr sequestering C and responding to altered climatic regimes in forests add to this burden. Given the relative complexity of forest-sector C offset projects, a number of biogeochemical, rnstitutional and socioeconomic, monitoring and regulatory issues merit furtirer analysis before the potential of this greenhouse gas stabilization approach can be fully understood.

LITERATURE CITED
Apps' M'J' and Kurz, w.A. (1993), The Role of Canadian Forests in the Global C Balance. In M.