|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 FOLLOWING THE RULES: A BIOECONOMIC POLICY SIMULATION OF A BRAZIL IAN FOREST CONCESSION By ALEXANDER JAMES MACPHERSON A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007
2 2007 Alexander Macpherson
3 To my wife, my children, and my parents
4 ACKNOWLEDGMENTS I thank my doctoral advisor, Dr. Douglas Cart er, for his complete support and intellectual guidance throughout my doctoral program. I would also like to thank the other members of my doctoral committee, Dr. Janaki Al avalapati, Dr. Charles Moss, Dr. Christine Staudhammer, and Dr. Daniel Zarin, for their excellent guidance th rough this tremendous challenge. I also thank the Working Forests in the Tropics program for its award of a National Science Foundation Integrated Graduate Education and Research Traineeship, which funded my doctoral education and research. This study would not have been possible without the generous support of fellow researchers and students. At the Instituto do Homem e Meio Ambiente da Amaznia (IMAZON) in Belm, Brazil, I am extremely grateful to Paulo Barreto, Brenda Brito, Marco Lentini, Mark Schulze, Denis Valle, Beto Verssimo, and Ed son Vidal for sharing their time, data, and extensive knowledge. I am very fortunate to have received the continuous guidance of Mark Schulze and Marco Lentini, which wa s critical for this projects success. I hope the results of this study are valuable to IMAZON and that our collaboration continues in the future. I would like to thank my fellow student, On il Banerjee, who gave me very helpful advice on this project. Fred Boltz of Conservation International broke a lot the ground with his research on production forests in Bolivia which gr eatly facilitated my work. I owe the largest debt of gratitude to my l oving family who shared the struggle every day. I could not have succeeded without them. My wife Natalie was w illing to sacrifice an enormous amount so that I could achieve one of my drea ms. My boys, Ewan and Lucas, have grown and learned along with me.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........8 LIST OF FIGURES................................................................................................................ .......10 ABSTRACT....................................................................................................................... ............11 CHAPTER 1 INTRODUCTION..................................................................................................................13 Brazilian Public Fore sts Management Law............................................................................13 Sustainability................................................................................................................. .........16 Economic Approach.............................................................................................................. .22 2 A MATRIX-BASED GROWTH AND YIELD MODEL OF AN EASTERN AMAZONIAN FOREST UNDER HARVEST PRESSURE.................................................33 Introduction................................................................................................................... ..........33 Methods........................................................................................................................ ..........36 Study Site..................................................................................................................... ....36 Classification of Species Groups.....................................................................................39 Growth and Yield Model.................................................................................................41 Estimation Results............................................................................................................. .....48 Growth Model.................................................................................................................48 Damage......................................................................................................................... ...48 Recruitment.................................................................................................................... .49 Discussion..................................................................................................................... ..........50 Validity....................................................................................................................... .....50 Log and Leave Scenarios.............................................................................................51 Conclusion..................................................................................................................... .........54 3 THE SUSTAINABILITY OF TIMBER PRODUCTION FROM AN EASTERN AMAZONIAN FOREST........................................................................................................72 Introduction................................................................................................................... ..........72 Objectives of the Study........................................................................................................ ...73 The Model...................................................................................................................... .........75 Merchantability Restrictions...........................................................................................75 Equilibrium Dynamics.....................................................................................................79 Brazilian Regulatory Policy............................................................................................79 Sustainability Constraints................................................................................................81 Economic Variables.........................................................................................................84
6 Waste.......................................................................................................................... .....85 Concessionaire Objectives...............................................................................................86 Results........................................................................................................................ .............88 Maximum Sustainable Yield Harvests............................................................................88 Unconstrained and Brazilian Regulatory Policy Harvests..............................................89 Harvests under Weakly Sustainable Inventories.............................................................92 Harvests under Strongly Sustainable Inventories............................................................94 Financial Returns.............................................................................................................95 Discussion..................................................................................................................... ..........98 Silviculture................................................................................................................... ...98 Role of Currently Non-Commercial Species.................................................................100 Ways to Compensate Opportunity Costs of Additional Management..........................101 Matters of Scale: Spatial and Temporal Landscape Management................................102 Conclusion..................................................................................................................... .......103 4 A POLICY SIMULATION OF A BRAZ ILIAN LOGGING CONCESSION UNDER IMPERFECT ENFORCEMENT AND ROYALTIES.........................................................116 Introduction................................................................................................................... ........116 Objectives of the Study........................................................................................................ .117 The Model...................................................................................................................... .......121 Partial RIL Adoption.....................................................................................................121 Harvesting Above the Legal Volume Limit..................................................................124 Enforcement Mechanisms.............................................................................................125 Audit pressure........................................................................................................125 Performance bonds.................................................................................................127 Economic Variables.......................................................................................................128 Government Payments and Costs..................................................................................130 Objective Functions under Imperfect Enforcement and Performance Bonds...............132 Results and Discussion.........................................................................................................135 Renewability Audits and Annual Harvest Inspections..................................................135 Performance Bonds.......................................................................................................137 Royalty Instruments.......................................................................................................137 Ad valorem royalty under audit pressure................................................................138 Revenue-based royalty under audit pressure..........................................................139 Ad valorem royalty under performance bonds.......................................................140 Revenue-based royalty under performance bonds.................................................141 Issues with differentiated royalties.........................................................................141 Performance Bonds and Firm Size................................................................................142 A Note on Market-based Enforcement Efforts..............................................................143 Conclusion..................................................................................................................... .......143 5 CONCLUSIONS..................................................................................................................158 Findings and Methodological Advances.......................................................................158 Data Limitations............................................................................................................162 Future Extensions of the Model....................................................................................163
7 Experimentation and Adaptiveness...............................................................................165 APPENDIX A FAZENDA SETE SPECIES LIST GROUPS, AND PRICES............................................168 B LIST OF VARIABLES, VECTORS, AND MATRICES....................................................178 LIST OF REFERENCES.............................................................................................................182 BIOGRAPHICAL SKETCH.......................................................................................................192
8 LIST OF TABLES Table page 2-1. Maximum likelihood estimates of transition parameters...................................................56 2-2. Proportion of population per species gr oup and size killed per tree harvested under RIL treatment.................................................................................................................. ...57 2-3. Proportion of population per species gr oup and size killed per tree harvested under CL treatment................................................................................................................... ...58 2-4. Ordinary least squares estimates of the recruitment parameters........................................59 2.5. No harvest growth matrix G0.............................................................................................60 2-6. RIL growth matrix G1........................................................................................................62 2.7. CL growth matrix G2.........................................................................................................64 3-1. Merchantability criteria by species group........................................................................105 3-2. Commercial volume (m3/stem) by species group and DBH............................................106 3-5. Components of wast e across logging systems.................................................................109 3-5. Components of wast e across logging systems.................................................................109 3-6. Solution of the MSY program at the species group-level (tree/ha and m3/ha)................110 3-7. Pre-harvest standing stock, total harvest, and recovered timber for all cutting cycles (m3/ha)........................................................................................................................... ...111 3-8. NPV of scenarios ($/ha)................................................................................................. ..112 4-1. Results under imperfect enforcemen t (audits = 14 and fine = $700/ha) and ad valorem royalties (actual RIL costs)................................................................................147 4-2. Results under imperfect enforcemen t (audits = 14 and fine = $700/ha) and ad valorem royalties (high RIL costs)..................................................................................148 4-3. Results under imperfect enforcement (a udits = 14 and fine = $700/ha) and revenuebased royalties (actual RIL costs)....................................................................................149 4-4. Results under imperfect enforcement (a udits = 14 and fine = $700/ha) and revenuebased royalties (high RIL costs)......................................................................................150 4-5. Results under performance bonds ( bond = $250/ha) and ad valorem royalties (actual RIL costs)..................................................................................................................... ....151
9 4-6. Results under performance bonds ( bond = $250/ha) and ad valorem royalties (high RIL costs)..................................................................................................................... ....152 4-7. Results under performance bonds ( bond = $250/ha) and revenue-based royalties (actual RIL costs).............................................................................................................153 4-8. Results under performance bonds ( bond = $250/ha) and revenue-based royalties (high RIL costs)...............................................................................................................154 A-1. Species groups, scientific names, common names, and economic value class (stems/ha)..................................................................................................................... ....168
10 LIST OF FIGURES Figure page 1-1. Comparative harvest profiles.............................................................................................31 2-1. Actual and predicted 10-year diam eter distributions (stems/ha.........................................67 2-2. 100-year post-harvest projections acro ss harvest system and species groups (stems/ha)..................................................................................................................... ......68 2-3. 100-year post-harvest pr ojection of basal area (m2/ha of stems > 10 cm DBH)..............69 2-4. 100-year post-harvest project ion of merchantable volume (m3/ha of merchantable stems > 50 cm DBH).........................................................................................................70 2-5. Projection of average annual increment (m3/ha/year of merchantable stems > 50 cm DBH............................................................................................................................ .......71 3-1. Commercial volume recovery (m3/ha) after initial RIL harvest of increasing intensity (15 to 40 m3/ha).................................................................................................113 3-2. Dynamics of pre-harvest standing stock and to tal harvest by species group (40year cutting cycle)................................................................................................................. ...114 3-3. Size distributions of emergent and pioneer sp ecies groups across scenarios (40year cutting cycle)................................................................................................................. ...115 4-1. Probability of being caught br eaking laws and paying fine as an increasing function of illegal behavior............................................................................................................155 4-2. Effect of increasing number of periodic audits................................................................156 4-3. Effect of increasing performance bonds..........................................................................157
11 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy FOLLOWING THE RULES: A BIOECONOMIC POLICY SIMULATION OF A BRAZIL IAN FOREST CONCESSION By Alexander James Macpherson December 2007 Chair: Douglas R. Carter Major: Forest Resources and Conservation After decades of difficult experiences with fore st concessions, many countries continue to view concessions on public lands as a strategy to secure vuln erable lands while developing economic opportunities and raising government re venue. Brazil recently enacted the Public Forests Management Law, which focuses on the expansion and improved management of public forests, including the establishment of large areas of forest concessions. Using data from one of the longer running tropical forest experiments in the Eastern Brazilian Amazon region, my study develops a matrix growth and yield model whic h captures the dynamic effects of harvest system choice on forest structure and composition. The growth and yield model is embedded within a variety of optimization models parameterized w ith economic data from the Eastern Brazilian Amazon to study economic and forest condition outcomes as a function of policies such as current Brazilian regulations, su stainability constraints, enfo rcement mechanisms, and royalty systems. Results demonstrate that current re gulations are unlikely to guarantee sustainable timber yields, even when regulations are followe d and best logging practices are implemented. Inventory-based regulatory defi nitions of total volume and species-level sustainability are proposed and analyzed in order to estimate the viability and opportunity costs of additional harvest regulations. Results show that, under the proposed definitions, sustainable timber
12 inventories are likely viable but at significant opportunity co sts. In a weak enforcement environment, audit pressure is unlikely to induce full compliance with harvest regulations, while performance bonds may potentially be more effectiv e. The use of traditio nal royalty instruments such as the ad valorem and revenue-based royalties can eff ectively generate revenues and, in the case of revenue-based instruments, modify harvest behavior, but only under very limited circumstances. When critical variables change, such as the cost to implement reduced impact logging, the outcomes under the same instrument can vary dramatically. The optimal choice between applying a royalty-base d strategy and a strategy based upon an area fee determined through a competitive or administrative process is crucial and is likely dependent upon the qualities and quantities of the forest resource, the characteristics of the logging firms, and regional institutional strength.
13 CHAPTER 1 INTRODUCTION Brazilian Public Forests Management Law In the 1960s and 1970s, government investment opened access to extensive portions of terra firme forests in inland re gions of the Brazilian Amazon, mainly through the construction of roads. Improved forest access and subsidies for migrants to occupy and clear vast unclaimed areas promoted the creation of a highly predat ory, extensive, and mi gratory logging industry (Lima et al., 2006; Stone, 1998a; Stone, 1998b; Uhl et al., 1997; Verssimo et al., 1998; Verssimo et al., 2002a). Today, in much of the Brazilian Amazon, the forest sector is dominated by insecure property ri ghts and a rent-seeking, ineffici ent bureaucracy (Merry et al., 2006). The combination of high opportunity costs of capital and weak property rights increase the risks of forest investment, which compels loggers to have short-term, profit-maximizing objectives, leaving little incentive for managing for future harvests (Merry et al., 2006). Government inefficiency and corruption add to th e disincentives of management, as oversight is often so lax or corrupt that illegal loggers gain a strategi c advantage though bribery, while loggers wishing to follow the rules often cannot get management plans approved by the government (Merry et al., 2006). The result is that an estimated 43% of the Brazilian Amazons timber supply is sourced by illegal logging (Lentini et al., 2005). Land tenure instability poses the most signifi cant constraint to in creasing the area of managed forest in the Brazilian Amazon. Roughl y 33% of the Brazilian Amazon, approximately 160 million hectares, is in terras devolutas lands that are thought to be publicly-owned but are either subject to conflicting ow nership claims or unclaimed and unsecured by a state presence. Because of the uncertain legal and ill-protected st atus of these lands, they are frequently used by loggers and ranchers for private gain (Lentini et al., 2005). The predat ory use of these public
14 lands creates a significant loss of the nonmarket benefits that s hould accrue the public. As well as maintaining ecological services and biodiversity, ther e is an enormous potential to use these forests to produce timber and non-timber forest products on a sustainable basis, generating increased governmental revenue and a wide range of social and economic be nefits. Verssimo et al. (2000) estimate there are 114 million hectares of forestlands, 28% of the Brazilian Amazon, with the potential to become sustainable multi-use production forests. In 2006, Brazil enacted the Public Forest s Management Law (Lei 11284/2006), which focuses on the expansion and improved mana gement of public forests, including the establishment of large areas of forest concessi ons. The implementation of logging concessions on public lands is viewed as an ingredient towa rd correcting many of the failures of the current timber economy, at least on public lands (Verssimo et al., 2002a; Verssimo and Barreto, 2004). The new law arose from a long debate be tween public and non-governmental research institutions, state and federal governments, the priv ate sector, and social movements. As well as asserting an overarching strategy for the sustainable multi-purpose use of public forests, the law approves the establishment of logging concession s in federal and state-administered public forests, the creation of a federal forest de velopment fund (Fundo Nacional de Desenvolvimento Florestal) to financially support sustainable resource use, and the creation of a federal forestry agency (Servio Florestal Brasileiro), which will manage the development fund and supervise the concessions. Newly declared public lands will be alloca ted across the multiple forms of Brazilian conservation and sustainable land use categories at the state and federal levels, including strict protection, extractive reserves, la nds allocated directly to trad itional communities who prove historic use of the lands, and loggi ng concessions. The law contai ns provisions that establish a
15 framework for the decentralization of public land authority to the municipal and state-levels. These efforts are ongoing and parallel to the fe deral government devolving significant authority over forest policy to state envir onmental agencies. A forest des tined to become a public forest under the new law must pass several legal tests a nd procedures, which is expected to require many years for many highly disputed areas. As is typical with public policie s, the new legislation is full of vague directives and goals to achieve in the future. Many of the procedural elements of the law have been left for rulesetting within the bureaucracy. In addition, the Public Forests Ma nagement Law (PFML) is, in a sense, a secondary law. In the general public forest case, designation und er the new law depends upon other complex legal dynamics, such as land tenure reform, which also operate under their own sets of vague directives. Many debates during the prep aration of the law focused on the government's low institutional capacity to mon itor and enforce contracts a nd appropriate environmental management and royalty payment requirements. As the PFML is implemented, the development of sound administrative procedur es, revenue collection systems based on appropriate rates, enforcement capacity, and royalty distribution pr ocedures will become crucial. There is significant controversy whether these multiple tasks can be achieved. For example, Merry and Amacher (2005) identify several potential problems w ith Brazilian concessions, many of which may be analogous to well-documented problems w ith concessions in other countries (see e.g. Gray, 2005; Repetto and Gillis, 1998). First, conc essions may allow concessionaires artificially high rents because of poorly-designed revenue systems. As well as give away public resources, this problem may additionally stifle innovation and the longer term competitiveness of the Brazilian forest industry (Merry and Amacher 2005). Second, the Brazilian government may
16 become a rent-seeking government as a result of unexpectedly high costs of administering the concession system added to the possible nonpayment of royalties. In order to increase revenue, a rent-seeking government might seek to induce hi gher harvest rates by lowering tax or royalty rates away from the first-best rates that ma ximize economic efficiency. This effect, also discussed for concessions in general in Amach er (1999), Amacher and Brazee (1997), and Merry and Amacher (2005), can lead to an accelerated a llocation of land into concessions, which risks compounding budgetary shortfalls. The interaction between logging on public a nd logging on private land is a critical additional complexity, and it is difficult to foresee how the timber economy might evolve. Studies have shown that regional timber demand in the short and medium-term could be satisfied by low-intensity logging on small holder properties, where social welfare and conservation gains are potentially large in comparison to concessi on logging (Campos and Nepstad, 2006; Lima et al., 2006; Nepstad et al., 2004). Most importantl y, in order to avoid unintended consequences concession policies need to be designed using a broader view that includes revenue systems and interactions with timber markets in the larger economy (Merry and Amacher, 2005). This introductory chapter develops these issu es in more detail, drawing upon relevant research and discussions. The section that follows conceptualizes sustainability in the context of this study. The following section establishes th e conceptual foundation underpinning this study by drawing upon the economics literature. Sustainability Given a political decision to contract out the management of public forest goods and services, fundamental questions about manage ment objectives appear. The structure and outcomes of the concession system will vary dram atically according to the objective sought. The simplest form of concession is a harvest conces sion, where the ultimate objective is to convert
17 the land to alternative uses. In the Sabah region of Malaysia, for example, the government largely treated forests as a nonrenewable res ource to be mined for foreign exchange and employment, emphasizing the conversion of land to ag ricultural uses (Gillis, 1988). On the other end of the spectrum are the objectives of sustai nable timber management (STM) and sustainable forest management (SFM). STM focuses on su staining timber yields, while SFM focuses on the sustained delivery of multiple goods and services over long periods of time (Pearce et al., 2003). There may be many perspectives on management objectives along the spectrum between these extremes. In the Brazilian case, the legal defini tion of SFM is clearly presented in the PFML. Concessionaires are legally obligat ed to practice SFM, which is de fined as forest administration in order to gain economic, social, and environm ental benefits, respecti ng the mechanisms that sustain the managed ecosystem and considering th e use of multiple timber species, multiple nontimber products and sub-products, and other goods and services of na tural forests (Lei 11284/2006). A central debate in the discus sions about SFM is whether it is appropriate to include requirements for sustained timber yields (STY) w ithin forest management requirements. On the surface, successfully im plementing STY addresses the need for long-term timber supplies from managed sources, often viewed to be a critical ingredient to a stable forest sector that effectively contributes to economic development. In the ab sence of more complete information about the broad portfolio of forest ecosystem goods and servi ces, as is typically the case, STY may also be viewed as an umbrella measure for forest pro cesses more holistically. This use of STY as a proxy for SFM is very attractive to many stakeh olders in the Amazon-region who are investing in forest management as a bulwark against the degradation and deforestat ion that dramatically reduces or destroys the bulk of the forests functions.
18 However, Luckert and Williamson (2005) write that forest economists are increasingly viewing STY management requirements as econom ically detrimental. STY requirements do not promote economic stability and development as much as once thought, as the requirements can bind private actors into making su boptimal decisions when market s for forest goods and services are volatile (Luckert and Williamson, 2005). The authors debate whether strong STY constraints should have any role within SFM, arguing that timber perhaps shoul d be viewed as one of a set of substitutable products within the forest. Viewed this way, forest managers are not constrained to harvest a given quantity over a set period of ti me. Rather, managers flexibly tradeoff forest goods and services in response to dynamic market c onditions, increasing forest sector efficiency. At the same time, forest services that are eith er irreversible or public goods may additionally be managed under a precautionary prin ciple (Luckert and Williamson, 2005). Luckert and Williamson draw heavily upon the discourse within economics concerning strong and weak sustainability, pi oneered largely by David Pearce in a series of publications (see e.g. Pearce et al. 1989). Whet her an economy is strongly or weakly sustainable depends upon decisions about the substitutability of human-ma de capital for natural capital. In a weakly sustainable economy, human-made capital can be substituted for natura l capital, the objective being to sustain non-declining utility into perpetuity (Pear ce et al., 1989). In a strongly sustainable economy, utility and natural capital it self must be non-declining in time, a position frequently advocated within ecological economics (Daly, 1996). Many varia tions on these basic viewpoints are examined in detail in Heal (1998) and Neumayer ( 1999). In their article, Luckert and Williamson (2005) assert that STY requireme nts are a form of strong sustainability (no substitution between timber and other products) while SFM is a fo rm of weak sustainability (substitution is permitted).
19 Luckert and Williamsons argument is compelling but is based upon the assumption that markets function well in the particular region unde r discussion, that the mark ets ability to assign prices to scarcity will generate an optimal a llocation of forests which will persist into the foreseeable future. This assumption may be reas onable in the case of forested countries with well-functioning markets that have experienced a forest transition and are seeing forests rebound (Rudel et al., 2005), but may not be as appropriately app lied within regions such as the Brazilian Amazon, where land tenure problems abound and rate s of deforestation and forest degradation continue to be high (Foley et al., 2007). Markets and management information on non-timber forest ecosystem goods and services are also l acking in the region, preventing managers from making well-informed decisions about trad e-offs between goods and services. It is the position of this pa per, then, that the continued examination and calculation of sustainable yield remains releva nt for forest management in the Brazilian Amazon, particularly with respect to public lands planning. The limits of such an assessment should be identified at the outset, however. As Karsenty and Gour let-Floury (2006) highlight, while estimating the timber harvests over multiple cutting cycles is valuable, reducing the assessment of sustainable management to the sustainable yields of a few, highly-valued species is problematic. Accepting the assertion that the ca lculation of sustainable yields rema ins relevant, the important discussion becomes about how STY is applied in practice. In much of the Brazilian Amazon, as a result of market failures, there is clear incentive to liqui date the entire merchant able timber stock, leaving behind commercially and ecologically degraded forests. Regulato ry constraints and/or economic incentives are necessary to induce socially desired behavior. Current forestry best practices are limited to reduced impact logging (RIL) systems which seek to minimize environmental impacts of harvest systems as compared to unplanned
20 conventional logging (CL) systems which constitute about 90% of the regions harvests (Zarin et al., 2007). RIL typically requires the followi ng practices, adapted fr om Dykstra (2002): Pre-harvest commercial tree inventories and maps Pre-harvest planning of roads, skid-trails and landings Pre-harvest vine cutting Employing directional felling, cutting stumps low to the ground, and optimal bucking of tree stems, all to reduce waste Construction of roads, landings and sk id-trails that satis fy design guidelines Ensuring that skidders remain on the skid -trails by winching logs when feasible Conducting post-harvest assessments in order to provide feedback In Brazil, firms are obliged to follow legal re strictions aimed at minimizing environmental damage and protecting the future productivity of the forest. Thes e restrictions include minimum diameter cutting limits, upper bounds on harvest intensity, the retention of seed trees and individuals of rare species, a nd protection of riparian buffers and wildlife. However, the RIL system applied within the contex t of these legal restrictions is not guaranteed to induce STY (Zarin et al., 2007). An additi onal requirement that the harvest from any given cutting cycle be at or below a sustainable volume may be an add itional policy instrument that proves appropriate in this environment. Further, regulators might consider imposing STY at the species-level. In fact, STY-type requirements are likely to be adopted within the PFML. These types of rules are policy-determined a pproximations that are expected to achieve broad objectives yet are administratively simple to apply and monitor (Boscolo and Vincent, 2003). But, as is the case with virtually any fo rm of regulation, one-size-fits-all policies are unlikely to optimally achieve multiple objectives w ithin complex landscapes. Zarin et al. (2007) write that the size of the landhol ding and whether the land is pub licly or privately-owned should
21 influence management requirements. Private smallholders (typically with properties under 500 ha) should be expected to practice RIL, but given technical, legal, and fina ncial constraints, these landowners should not be held to hi gher standards, such as sustai nable production at the forest or species-level. These landholders often view timbe r harvests as a one-tim e event that finances economic activities, often the pur chase of cattle, on the proporti on of the land the owners are legally permitted to clear (Zarin et al., 2007). Maintaining forest cove r on the residual land is challenging because of destruc tive reentry logging, fires escaped from adjacent lands, and clearings at the forest margin in an attempt to increase agricultural pr oductivity (Zarin et al., 2007). Meanwhile, forest operations on mid-sized pr ivate and public lands typically larger than 3000 ha, Zarin et al. (2007) argue, are of sufficient scale that th ey should be required to sustain total commercial volume production. Silvicultura l practices, cutting blocks, areas, and cycles should be appropriately be adjusted in respons e to forest conditions, information, and technology (Zarin et al., 2007). Sustaining species-level harvested volumes is e xpected to be particularly challenging with high-valued timber species that are heavily logged and do not have the typical inverse Jdistribution with a sizeable num ber of sub-merchantable stems poised to grow into the commercial size classes in time for future harvest entries, such ip ( Tabebuia impetiginosa ), mahogany ( Swietenia macrophylla ), cumaru ( Dipteryx odorata ), and cedrela ( Cedrela odorata ) (Schulze, 2003; Schulze et al., 2005 ; Zarin et al., 2007). Some of these species are thought to benefit from, and perhaps require, large disturban ce events for regeneration, indicating that the relatively flat size distributions seen today are likely to be from cohorts that established during a long ago disturbance (Fredericksen and Putz, 2003; Gullison et al., 1996; Snook, 1996; Zarin et al., 2007). Because of the expect ed silvicultural challenges and costs and the required spatial
22 scales of sustaining these high-valued species, Zarin et al. (2007) a dvocate that sustaining volumes at the species-level should be required on large public lands, such as those destined to become logging concessions in Brazil. Making the sustainability challenge more comp lex is the fact that logging impacts will inevitably lead to post-harvest floristic reco mposition over the short and medium terms in logged-and-left stands and permanently in rep eatedly logged stands under the relatively short cutting cycles typically used (Favrichon, 1998; Karsenty and Gour let-Fleury, 2006; Sist et al., 2003b). This recombination tends to favor fa st-growing, light-demanding species with very light, often not commercially valuable wood (Fav richon, 1998; Phillips et al., 2004; Valle et al., 2007; Van Gardingen et al., 2006). Sustaining species-level volum es over time, therefore, may be significantly more challenging than sustaining overall timber volumes in which substitutions among species to contribute to volume constraints is permitted. Economic Approach While the grey literature of consulting reports on the economics of tropical logging concessions flows relatively stead ily (for reviews of a large nu mber of studies, see e.g. Gray, 2005; Scholl, 2005), the peer-reviewed economics literature on tropical forest concessions appeared in two bursts. The first period, stim ulated by the publicati on of Repetto and Gillis (1988), saw a prominent dialogue amongst economist s on the influence of royalty instruments on economic efficiency and rent distribution (s ee e.g. Hyde and Sedjo, 1992; Vincent, 1990; Vincent, 1993). This literature is well-revi ewed by Merry and Amacher (2005), who synthesize lessons for Brazils emergent concession system and conclude that it is imperative to design concession policy instruments jointl y with instruments that influence harvests on private lands. Merry and Amacher (2005) also cau tion that the concession syst em may expand too rapidly if concession polices and government revenue obj ectives are linked, a point echoing Amacher and
23 Brazee (1997). Perhaps motivated by the emergen ce of new logging concession policies in Latin America, a more recent round of concession econ omics literature places a stronger emphasis on the dynamics of illegal logging and corruption. For example, Amacher (2006) issued a challenge to economists to better integrate illegal logging and forest sector corruption into their research efforts (for examples, see e.g. in Amacher et al., 2006; Amacher et al ,. 2007; Delacote, 2005; Palmer, 2003). In a region as vast as the Brazilian Amazon, illegal logging and forest sector corruption assumes a multitude of forms. It is important to contextualize the very specific types of illegality studied here within the larger Brazilian forest economy. Illegal logging includes activities such as logging on illegally settled lands, logging without government approved management plans, logging outside the spatial boundaries of approved plans or within lega lly reserved areas. Illegal logging also includes logging greater volumes that management regulations permit and the harvest of rare and smaller diameter trees. Ille gal logging also can include activities outside of the forest, such as the illegal acquisition of ma nagement authorizations timber transportation permits, and other types of documen tary falsifications. At the larger scale, there may be longterm agreements between timber firms and governme nt officials to facili tate unlawful forest activities. For many operators of forest products firms in the Brazilian Amazon, the legal environment is very ambiguous and uncertain (Rh odes et al., 2006). There is a large difference between operators of large ente rprises who collude with law en forcement and purposefully flout national and international law and the much larger number of small-scale operators who, in order to survive in a difficult business climate, falsify documents and pay small bribes (Rhodes et al., 2006). From the regulators perspe ctive, it is difficult to distingui sh between these very different
24 types of businesspeople (Rhodes et al., 2006). A dditionally, improved enforcement of misguided forest law might inhibit the rural poor from performing low-intensity timber operations by increasing transaction costs or, worse, by cutti ng off traditional access to forest resources, causing severe impacts on rura l livelihoods (Kaimowitz, 2003). In the limited forest economics literature that examines illegal logging, economists primarily examine the quantity of timber harveste d (see e.g. Amacher et al., 2007) or the size of timber harvested (see e.g. Boscolo and Vincent, 2000). To a lesser extent, economists examine the possibly corrupt interactions between logger and regulators (see e.g. Delacote 2005). With the exception of Boscolo and Vincent (2000) and Le ruth et al. (2001), econo mists have paid very little attention to logging tec hniques specifically or forest management more broadly. This study adopts a confined definition of ille gality. Building on the economics tradition of focusing on harvest volumes, this study examines the incentives to harvest over legal limits. While Brazilian regulations estab lish a general harvest limit on ma nagement plan authorizations, plans may in fact require harvests below this lim it, based upon forest type, firm size, and cutting cycle length. This study examines the most general case of harvest limits (the 30 m3/ha per harvest per cutting cycle entry), wh ich is likely to reflect plans au thorized for mid to large-scale firms operating under a 25 to 40 year cutting cycle. Like Leruth et al. (2001), this study assumes that the negative externalities associated with selective logging have as much to do (or more ) with management practices as with harvest intensities. In this study, the implementation of best logging practices, in the form of RIL, is treated as a proxy for forest management. Add itionally, the implementati on of RIL techniques is likely to be required on Brazilian forest concessions. Hence, the under-implementation of RIL is treated as a form of illegal logging.
25 To further motivate the conceptual approach of this study, Figure 1-1 modifies the diagrams discussed in Vincent (1990) and Hyde and Sedjo (1992) to in clude illegal logging. MC0 represents the illega l loggers marginal cost of harvesti ng timber volume V. In this highly simplified depiction, the illegal logger chooses to maximize priv ate welfare and harvest at the level of V0 at the given market clearing timber price p Next, V1 represents the harvest levels of the logger that follows current regulations. The additional marginal cost of operating legally is measured by MC1. Crucially, as is often the case with one-size -fits-all environmental regulation, the set of rules driving the legal logger to produce at V1 ar e unlikely to be perfectly designed to protect the future productivity of the forest. Simultaneousl y, the rules are unlikely to induce the logger to internalize the social costs of logging, such as reductions in biodiversity, nontimber forest products, climate and disease regulation, and water quality that might be associated with logging impacts. The additional constraints on harvests that protect future produc tivity and internalize social costs are represented by adding MC2 a nd MC3, respectively, which further reduces the optimal harvest levels. However, as Vincent (19 90) and Hyde and Sedjo ( 1992) discuss, the real rate of value growth in natural tropical forests ma y be so low that MC2 could equal zero, as there is little or no financial incentive to leave valu able timber standing or re duce damages to future stocks by reducing harvests or employing more costly harvest and silvicultural techniques. In the presence of imperfect enforcement and incentives for breaking the rules, government agencies face the challenge of how to jointly de velop enforcement and royalty systems that costeffectively induce desired harvest levels and improved forest management. Depending on the agent's preferences over risk, th e type of law, and the likelih ood of being caught, the would-be rule-breaker generates a probabilistic expectati on of gains and negative consequences of being
26 caught and punished. In a simplified version of this story, if the expect ed profits with rulebreaking are higher than alternat ive legal opportunities, the agent will break the rules (Becker, 1968). Second, given the likelihood of criminal be havior, the law enforcement agency faces a resource allocation problem and must make a vari ety of decisions with respect to the definition of the potential crime, the public enforcement effort and the level of the fi nes in order to identify an optimal solution under incentives for lawbreaking behavior (Becker, 1968; Polinsky and Shavell, 1979; Polinsky and Shavell, 2000). The enforcement problem can be seen then as a form of market failure, where the social planner seeks to equate the marginal costs of controlling externalities with the marginal benefi ts of reducing the externality. A representative loggers marginal prof it as a function of volume harvested, V is depicted as a downward sloping line in the simp le schematic of Figure 1-2, which modifies a diagram from Sutinen and Andersen (1985) to include illegal loggi ng. Under incentives for rulebreaking and failed enforcement, the logger will c hoose to operate at the level where marginal profit is zero, or V0 in Figure 1-1. The enforcem ent agency must allocate resources to push the logger toward producing at V1, the optimal producti on level under the current set of rules. But because enforcement is imperfect, the enforcement agency is unlikely to be successful. Where the fine is exogenously given at f (as is often the case with penalties for environmental crimes) and the probability of being caught pe rforming illegal harvests is given by V, where 0V the marginal expected fine is given by f V The intersection between the two lines locates the loggers harvest choice at V* under imperfect enfo rcement when incentives for illegal behavior exist, most likely at a leve l somewhere between V0 and V1.
27 Increasing fines is one option to reduce illegali ty. In 1998, for example, Brazil, increased maximum fines under its environmental crimes law (Lei 9605-98), which includes a maximum fine for deforestation and illegal logging of $700/ha. Yet, a triv ial mathematical solution to the enforcement problem when enforcement is costly is that fines should be made arbitrarily large while the probability of capture approaches ze ro (in order to minimize costs of enforcement effort), where the expected fine equates the marg inal damage of the offence. Yet, economists pointed out soon after Beckers influential 1968 article that initi ated the study of enforcement economics that fines above a firms capacity to pay are often meaningl ess. For practical purposes, the expected fine should not exceed a firms threshold for bankruptcy. Meanwhile, institutional barriers such as inadequate fundi ng, professional capacity, and corruption may be pervasive within an enforcement system. While enforcement pressure may induce regula tory compliance, royalty instruments, or taxation instruments more generally, some economist s have argued, can be selectively applied to meet other harvest objectives, such as reducing harv est levels to the V2 or V3 levels in Figure 11. Much concessions economics research has focused on the fact that few countries have priced the forest resource well and pr ovided appropriate economic incen tives to the sustainable and efficient use of public forest resources (Gray, 2005). Problems include setting fees too low and collecting fees at relativel y low rates (Gray, 2005). Additionally, researchers in the concession economics literature have been concerned with rent capture and distribution (A macher et al., 2001; Boltz, 20 03; Boscolo and Vincent, 2000; Hyde and Sedjo, 1992; Vincent, 1990). Rent capture appeals to the idea that the state is the true owner of the land and should recover the full econo mic rent associated with concession logging, the area p1pc in the regulated logger of Figure 1-1, for example. Distribution refers to how
28 much of the rent is captured by the government and how much is retained by the concessionaire, the common concession experience being that the c oncessionaire retains to o large a proportion of the rent (Repetto and Gillis, 1988). Meanwhile, even if the government is able to fully capture legal levels of rent, the logger may hide rents obtained through illegal harvests, depicted in Figure 1-1 by the area ap0p1c. However, rent distribution sa ys little about economic effi ciency (Hyde and Sedjo, 1992; Vincent, 1993). On one extreme, depending on how firms reinvest the rents, disproportionate private sector rent capture may lead to more local-scale economic gains than rent captured and dissipated by an inefficient government (Hyde a nd Sedjo, 1992). On the other extreme, private sector rent might accrue predominantly to natio nal and international elites, with few positive impacts poverty or forest sector development (Hyde and Sedjo, 1992). In this study, in addition to the opportunity to collect fine s when the concessionaire is caught breaking the rules, two type s of royalty charges are inves tigated. First, a percentage ad valorem rate is charged against the marginal profit, or the added-value, of harvesting a tree. In other words, a percentage of the difference between the price of a tree and the costs to harvest the tree is charged under the ad valorem royalty. Second, a percentage revenue-based royalty rate is charged against just the price of the harves ted tree. For either instrument, the royalty rate is applied only against trees that were legally ha rvested. As has been shown in previous work, the ad valorem royalty system as typically applied is a non-distortionary instru ment that will not influence marginal harvest decisions (Hyde and Sedjo, 1992). However, given the analysis here includes the potential of illegal logging, where no royalties are paid on the illegal portion of the harvest, the ad valorem royalty may not be nondistortionary across its feasible range, as high rates may induce illegal logging. Meanwhile, the
29 revenue-based royalty charged will reduce the relative profitability of harvesting the tree, distorting harvest decisions on the margin, creat ing the possibility of inducing lower harvest levels, which may or may not be a desirable eff ect, given the governments objectives. The same risk of high rates induci ng illegal harvests may also be present with the revenue-based royalty. Also, as it is commonly practiced, the ad valorem royalty requires relatively high monitoring costs relative to the revenue-based royalty because firm revenues a nd costs need to be monitored, rather than only the revenues. Under imperfect monitoring, the firm may seek to exaggerate costs, reducing the overall ad valorem charge. An area fee is also studied as a possible reve nue instrument. Area fees, as used in this analysis, are also non-distortionary in terms of harvest decisions, although at a larger scale they may help determine what lands are profitable fo r harvest. This occurred in the Bolivian concession system as large areas of less producti ve forests were taken out of the concession system under an annual area fee (Merry and Amac her, 2005). In practice, the area fee can be determined via a competitive or administrative process. An additional policy choice to be examined in this study is the use of harvest performance bonds. Performance bonds function differently than th e previous instruments in that, in one form of the bond, loggers deposit money with the gov ernment before harvest and then, upon execution of the harvest, are refunded a quantity proportio nal to the satisfaction of required performance measurements. The use of performance bond mech anisms is often proposed as a royalty or enforcement instrument and has been the subject of several studies (Boltz, 2003; Boscolo and Vincent, 2000; Leruth et al., 2001; Sun, 1997), yet there is very little pr actical experience with the instrument in tropical country concessions. The logger, for example, may deposit the area ap0p1c with the government in order to get final appr oval to proceed with harvests (Figure 1-1).
30 If the rule-abiding harvest is not performed, th e government will keep the appropriate proportion to compensate for losses associated with the rule-breaking. Leruth et al. (2001) find that performance bonds combined with traditional enforcement measures are more likely to be effective at protec ting forest resources than approaches that rely upon royalty instruments. They argue that the negative externalities a ssociated with logging have very little relationship with harvest volumes. Rather, the extern alities are largely a result of the quality of management prac tices, such as whether RIL or post-harvest silviculture is implemented. A timber tax is unlikely to suc cessfully function as Pi govian-type instrument which taxes the externality in order to better alig n social and private cost s (Leruth et al., 2001). In fact, Leruth et al. (2001) argue, timber taxati on can actually induce more damage to the forest resource than no taxes at all by encouraging lower overall harvests at higher rates of collateral damage. In this study, the logger will be assumed to pay royalties only on legal portion of the harvest. This shirking of payments adds to the already significant incentive to operate illegally. Revenue systems have little or no influence w ithin the area between the MC0 and MC0 + MC1 curves on Figure 1-1. In fact, there is severe discontinuity at the threshold between legal and illegal logging intensities; the royalty instantaneously goes to zero. Meanwhile, there are critical forest activities that are unaffected directly in the short-term by royalty pol icies, such as RIL or post-harvest silviculture, which may have signif icant economic and ecologi cal implications over the long-term. The position of this paper is that enforcement and royalty systems are intertwined, a point theoretically shown in Leru th (2001) and Amacher et al. (2007). In other words, command and control policies must be mixe d with royalty instruments in order to foster more sustainable harvest and management practices.
31 Figure 1-1. Comparative harv est profiles (modified from Hyde and Sedjo, 1992). MC0 represents the marginal costs of harvesting ille gally. MC1 represents the additional marginal costs of operating within the current set of rules. MC2 represents the additional marginal costs of investing to protect the futu re productivity of the stand. MC3 represents the additional marginal costs of internalizing environmental externalities.
32 Figure 1-2. Simple application of a penalty function (modified from Sutinen and Andersen, 1985)
33 CHAPTER 2 A MATRIX-BASED GROWTH AND YIEL D MODEL OF AN EASTERN AMAZONIAN FOREST UNDER HARVEST PRESSURE Introduction Matrix models of forest growth and yield are robust predictors of aggregate stand characteristics such as density, basal area, and di ameter distribution, while being mathematically tractable (Picard et al., 2002; Si st et al., 2003a; Vanclay, 2001). Matrix models are particularly advantageous in that they can be used to simu late a range of conditions based upon minimal data requirements (Gourlet-Fleury et al., 2005). A pair of observations of th e diameter distribution will often be sufficient to construct the transi tion matrix (Buongiorno and Michie, 1980; Michie and Buongiorno, 1984). Analysis may be performe d at smaller scales, as in the many studies that examine a representative fo rest area, such as a hectare or acre (Boltz, 2003; Boscolo and Vincent, 2000; Buongiorno and Michie, 1980), or at the landscape-leve l (Lin and Buongiorno, 1998; Lin and Buongiorno, 1999a; Lin and Buongiorno, 1999b) Because of analytical tractability, minimal data require ments, and scalability, matrix models are often used in forest economics, particularly in tropical contexts (Boltz, 2003; Boscolo et al., 1997; Boscolo and Vincent, 2000; Favrichon, 1998; Mendoza and Sety arso, 1986; nal, 1997; Sist et al., 2003b). Estimation of tropical forest growth and yield m odels is typically very difficult because of the lack of data on such complex forests. Ad equately representing the dynamics of species interactions, recruitment, and response to disturba nce is even more difficult when the objective is to develop management plans (Boltz and Carter 2006). The response to disturbance, usually a harvest, is often treated non-expl icitly by mixing data from contro l and logged sites to estimate growth and recruitment parameters. The main limitation of the approach of estimating a single growth matrix from data that mixes control and disturbed treatments is the poor incorporation of how the forest responds in the short and long-term to relatively high levels of disturbance.
34 For the Eastern Amazonian forest of this st udy, the harvest system employed has a strong effect on damage to the residual stand and future growth rates of the stand, although these effects may be short-term. For some species in certa in contexts, harvest disturbance can create a favorable environment for growth (Frederickse n and Putz, 2003). For others, the harvest disturbance may create an environment that is unf avorable, particularly if combined with heavy harvest pressure, as is the case with many highly valuable species in the Amazon region (Schulze, 2003). As matrix models are often used to project logged stands into the far future in which logging is performed over mu lti-decade cycles, repeated app lication of a growth matrix estimated on data collected over a relatively shor t post-harvest period creates a risk of bias, as the post-harvest growth response is projected for several decades, whereas the available evidence indicates that the post-logging growth pulse last s for a decade or less (De Graaf et al., 1999; Dekker and De Graaf, 2003; Silva et al., 1995; Va lle et al., 2007; Vida l, 2004). Estimating matrix transition parameters including trees killed inadvertently during harvest is also problematic because these trees do not obs erve any of the possible transitions. While other authors have examined the imp lications of reduced impact logging (RIL) on economic and ecological outcomes (Boscolo et al ., 1997; Boscolo and Vincent, 2000; Favrichon, 1998; Sist et al., 2003b), only Bo scolo and Vincent (2000) have developed a managementoriented matrix model that gives the forest ma nager an endogenous choice of logging technique and, hence, the consequent type and level of damage. The forest manager, often under highly imperfect information, will evaluate costs and benefits of various actions and then choose a decision path that best meets pa rticular objectives within rele vant institutiona l and ecological constraints.
35 The decision to adopt improved harvest practices resides not at the extremes of to fully adopt RIL or not, but is situated somewhere in the continuum between. Within an actual logging environment, some practices are more costly than others while some are more easily shirked than others. Simple oversight or lack of training may also cause firms to fail to adopt particular RIL guidelines. For example, in a study of two la rge certified companies in the Brazilian Amazon, Pokorny et al. (2005) note that about one-third of a set of 61 RIL guidelines were not fully implemented. The lack of sufficient monitori ng, training, and equipmen t explains many of the failures to meet guidelin es (Pokorny et al., 2005). This work seeks to advance the use of matrix models in tropical forestry policy analysis on two fronts: first, by improving the capacity of the model to capture the dynamic effects of harvest on forest structure and composition; and, second, by endogenizing the choice of harvest system that allows the manager to best meet objectives. In doing so, this study extends the multi-species uneven-aged forest management m odels of Lu and Buongiorno (1993), Lin et al. (1996), and Buongiorno et al (1995) to a tropical fo rest context. The model also improves the incorporation of harvest damage in a manner similar to Boscolo and Vincent (2000). The study also draws on and extends Boltz and Carter (2006) by using multinomial logit regression (MNL) to estimate the matrix transition probabilities. The model incorporates a recruitment function that models recruitment as a function of stand density and logging treatment type. To display the utility of the modeling appro ach, the model will be used to compare the long-run structure and composition of the forest arising from th e choice of implementing either RIL or conventional logging (CL), contrasted agai nst a baseline projection of an unlogged forest. The framework of the managers harvest system choice will be established in this chapter, but examined in more detail in subsequent chapters.
36 Methods Study Site The matrix model was estimated from data co llected from a 205 ha forest block in the Paragominas region of Par, Brazil 3S47.5W, one of the principal Amazonian logging centers during the regions logging boom period of the 1980s and 1990s. The annual rainfall in the area is on average 1700 mm, occurring most ly between December and May. While on level terrain with no obvious differences in soil types (oxisols or ultisols), the evergreen study forest displays wide structural variatio n (Schulze, 2003; Vidal et al., 19 97). Forests form a mosaic of high, medium and low stature patches, which can vary over very short di stances (Schulze, 2003; Vidal et al., 1997). High stat ure patch trees are 30-45 m tall, while in the low, early-building phase forests, trees are generally less than 10 m tall with frequent, dense vine cover (Vidal et al., 1997). The forest block of the experiment is a part of what is known as Fazenda Sete, a large, privately-owned forest fragment (2,500 ha) em bedded within a landscape of cattle-ranches, agricultural fields and forests th at have been logged and, in many areas, burned (Gerwing, 2002; Schulze, 2003). The bulk of the Fazenda Sete forest was logged in the early 1990s using CL, except for the area used in the logging experiment th at generated the data used in this and several other studies on forest ecology and manageme nt (Barreto et al., 1998; Johns et al., 1996; Schulze, 2003; Valle et al., 2007; Vidal et al., 1997; Vidal, 2 004). The forest block, which contains more than 350 tree species, was subject ed to three treatments in 1993: an unlogged 25 ha control plot, a 75 ha plot logged using CL by a typical, locally-recruited logging crew, and a plot logged using RIL by a well-trained crew (B arreto et al., 1998; John s et al., 1996; Vidal, 2004). The overall harvest intens ities were approximately the same within the two treatment
37 areas, but the amount of residual stand damage varied dramatically (Barreto et al., 1998; Johns et al., 1996; Valle et al., 2007; Vidal, 2004). The Fazenda Sete experiment has been the sour ce of several papers on the financial and ecological aspects of RIL. Johns et al. (1996) quantified the relative damage incurred by CL in comparison to RIL, also generating estimates of the benefits that w ould accrue to loggers employing RIL. Barreto et al. (1998) examined the relative costs and benefits of following best logging practices and found substantial economic benefit from implementing RIL rather than performing the typical unplanned logging operations prevalent in the region. The study site has been the source of ecological studies on vines (G erwing and Vidal, 2002; Vi dal et al., 1997) and post-harvest regeneration (Schulze, 2003; Vidal, 2004). Valle et al. (2007) use the Fazenda Sete data to calibrate the spatially-e xplicit SYMFOR model to study questions of yield regulation. The data used in this study were collected with in 24.5 ha areas within each treatment area. Within each treatment, two sampling strategies were followed. Firs t, all trees with diameter at breast heightDBH10 cm were measured within 5.25 ha (700 m x 75 m) plots within each 24.5 ha treatment area. Second, for the remaining 19.25 ha, trees of commercial species with DBH10 cm were measured, while only trees of non-commercial species with DBH25 cm were measured. This study largely uses data fr om the extensively sampled plots to derive model coefficients. In addition to pre-harvest measur ements, the plots have been measured several times over irregular intervals; this study draws on measurements from 1993 and 2003. A range of measurements were been performed, in cluding recruitment of new trees, species identification, DBH, vine density, stem form, crown form, and whether the trees were damaged or killed during logging or died of natural causes.
38 In terms of trees felled, harvest intensities in the logging treatment plots were similar. In the RIL plot, 4.6 trees/ha (2.1 m2/ha in basal area and 29.3 m3/ha in volume) were harvested. In comparison, in the CL plot, 4.8 trees/ha (2.2 m2/ha basal area and 29.9 m3/ha in volume) were harvested. Meanwhile, the number of trees accidentally killed dur ing each harvest diverged. In the RIL plot, 81.3 trees/ha (1.7 m2/ha in basal area and 18.5 m3/ha in volume) were killed inadvertently during harvest, while 86.9 trees/ha (2.2 m2/ha in basal area and 24.5 m3/ha in volume) were killed inadvertently in the CL pl ot. The damage figures indicate that, while the number of trees inadvertently k illed per tree harveste d was roughly equivalent across treatments, the basal area and volume damaged was about 30% high er in the CL plot. It is very important to note that these figures capture trees identifie d as trees killed imme diately during harvest (essentially destroyed) and do not include trees that were damaged and died in subsequent years, or at least suffered reduce post -harvest growth. These trees wh ich were likely to have been identified as damaged in Johns et al. (1996). Of the 6571 trees with DBH10 cm that were not killed during harvest, 6351 (96.7%) were used to estimate the growth model. Trees that could not be identified to species were excluded from the model estimation, as in Boltz (2006). In the simulations, however, these trees are proportionally allocated across species groups in order to provide more accurate predictions of aggregate variables such as basal area and merchantable volumes. Annual diameter increments ranged from 0.00 to 2.69 cm/year, the average rate being 0.28 cm/year with a standard deviation 0.33 cm/year. 80.7% of the post-harvest residual trees survived the 10-year period. Meanwhile, the recruitment data includes trees that grew to DBH10 cm during the 10year interval. 2627 trees (166.8 trees/ha) recr uited during this period, 2449 (93.3%) of which
39 were identified to species. The subset of trees with identified species was retained for the estimation of the recruitment function. Trees which were recruited a nd then died before observation in the tenth post-harvest year re-m easurement were excluded from the recruitment model. Classification of Species Groups Species classifications provide a simplified re presentation of the considerable ecological variation among species, which is a necessary simplification in many studies because data limitations and modeling constraints necessitate a small number of species groups (Vanclay, 2001). Two strategies are possible in this context: the analyst can make expe rt judgments of the relationship or data can be ag gregated so relationships can be estimated at the group-level (Vanclay, 2001). Some simple models aggregate species based upon economic criteria alone, a risky strategy that may bear little resemblance to true forest dynamics. It is preferable to use ecological information in the grouping decision (Vanclay, 2001). In this study, species were clustered into fi ve groups based on ecological traits, such as seed size, seedling shade tolerance, growth pot ential, wood density, and maximum adult size (see Appendix A for a complete list of species classi fications). The classification system used here provides a workable synthesis of published inform ation on the ecological tr aits of neotropical species, field observations made by three experien ced researchers with over a combined 45 years study history of eastern Amazonian forests, and pa tterns in the Fazenda Sete plot data. At one end of the spectrum were pioneer sp ecies, characterized by small seeds, early reproduction, aggressive colonization of canopy openings, rapid growth, high mortality rates, low wood density, and relatively small adult si ze (most species did not reach 100 cm DBH). Embaba species (e.g., Cecropia obtusa and C. sciadophylla) are common is this group. Pioneers include some commercial species but most are low-value lightwoods used primarily for
40 plywood. The shade-tolerant group was compos ed of species with generally large seeds, seedlings capable of prolonged survival in the shaded forest understory high density wood, and low mean and maximum growth ra tes. Commercial species in this group, such as maaranduba (Manilkara huberi), are used primarily for sawnwood and have medium to high commercial value. Two species groups fell in be tween the pioneer shade-tolerant ecological extremes. Light-demanding species, those with shade into lerant seedlings and capable of rapid growth under high light, but without cl assic pioneer characters such as copious and early seed production and small adult stature, include many plywood and a few sawnwood timber species, such as freijo branco (Cordia bicolor) and tacacazeira (Sterculia speciosa). The intermediate group includes species that have less abundant advance seedling regeneration in the forest understory than shade-tolerant species and generally lower wood density and higher mean or maximum growth rates. Intermediate s include species like muiracatiara (Astronium lecointei) and louro preto (Ocotea caudata) that combine shade-tolerant (e.g., diameter distributions approaching an inverse J-distribution) and light -demanding (e.g., rapid growth in high light) characteristics. A fifth group was created for a group of speci es with a unique suite of traits, the lightdemanding emergents. These species are char acterized by highly left-skewed diameter distributions, with very large and very old adults accounting for a la rge percentage of the standing population, and high density wood. Seedlings display shade intolerance, but grow relatively slowly compared with pioneers an d light-demanding species. Researchers have hypothesized that some emergent species ar e dependent on large-scale disturbance for replacement of adults and population persiste nce (Gullison et al., 1996; Snook, 1996). The
41 emergent group includes some of the highest value timbers, such as ip (Tabebuia impetiginosa) and jatob (Hymenaea courbaril), and some of the most difficult to manage sustainably (Schulze, 2003; Schulze et al., 2005; Zarin et al., 2007). Growth and Yield Model Despite the initial divergence in growth patte rns caused by the choice of logging treatment, at some point in the future, the growth dynamics of the forest under differe nt treatments is likely to re-converge to the dynamics of the unlogged forest, a phenomenon observed by researchers working in the Tapajs National Forest in th e Brazilian Amazon (Silva et al., 1995), in the Fazenda Sete forest studied here (Valle et al., 2007; Vidal, 2004), and in Suriname (De Graaf et al., 1999; Dekker and De Graaf, 2003). The tr ajectories resulting from each treatment can be represented by a unique cycle of transition matric es, which can be reduced to a single transition matrix (Winston, 1991). The benefit of this sing le matrix, then, is its easy incorporation into the classic economic optimization models, such as those that followed Buongiorno and Michie (1980). The Fazenda Sete growth and yield model is based upon a characterization of forest structure and composition using tree diameter distributions and sp ecies groups. The pre-harvest stand state is given by the vector tijt y y whereijty is the number of trees per ha in species group (1,..,)im and size class(1,..,) j n at time t. In this application of the model, the size classes are dominated in 10 cm DBH ranges beginning with 10-20 cm DBH. The highest range includes all trees greater than 100 cm DBH. Hence, there are 10 size classes. For example, the 10-20 cm DBH size class is indexed by 1 j 20-30 cm DBH is indexed by 2 j and so on until the highest DBH class, > 100 cm DBH, which is indexed by 10j
42 At each time t, the forest manager will choose (or by policy or management objective be constrained) to harvest trees or not. The harvest from species group i and size j at time t is given by tijth h. If the decision is to harvest at time t, the manager will harvest using a RIL or CL system, the choice of which will influence rate s of damage, growth, a nd recruitment. Let s index the choice of harvest system, where s = 0 indicates no harvest occurs, s = 1 indicates a harvest using RIL, and s = 2 which indicates a harvest using CL. The damage to the residual stand is assumed to be a function of overall harv est intensity and is represented by the mn x 1 vector, s td: 11 mn s tijtst ijh dD y (2-1) where s Dis a mn x mn matrix whose diagonal contains the logging damage coefficient for trees in each species group i and size j under harvest system type s. The damage coefficients are estimated as a percentage of trees killed per tree harvested within each species group i and size j under harvest system s. The damage vector includes tree s killed immediately upon harvest and does not include trees which were damaged during ha rvest operations and later died. When there is no harvest, 0D is an empty matrix. In addition to different rates of damage, th e model is constructed to test whether postharvest forests respond differently to harvest vo lumes and type of logging system employed. The choice of one type of harvest system over a nother is likely to resu lt in different residual stand damages and competitive environments. On one hand, some trees are damaged during harvests but not killed. One w ould expect relatively lower growth and higher mortality rates for these trees. On the other hand, harvest disturbance creates large gaps in the forest, changing the competitive environment. These changes are likely to favor some species over others. Consequently, a growth matrix for a forest logge d using a well-executed RIL system is not likely
43 to result in the same growth matrix as the hi ghly damaged forest logg ed with no planning or regard to future commercial trees. Both of thes e growth matrices are like ly to be different than the matrix derived from data on unlogged stands. Given a growth interval, the stand state at t is determined by the population, harvest, and damage at time t according to a modification of the multi-species uneven-aged matrix model presented in Lu and Buongiorno (1993), Lin et al. (1996) and Buongiorno et al. (1995): tsttsts y G y hdr (2-2) where ssGAR, in which s A is the transition matrix for treatment type s, Ris the ingrowth matrix that introduces the density depe ndent component of recruitment, and s r captures the fixed component of recruitment as a function of sp ecies group and treatment. The matrices and vectors are defined as: 1 s s ms A A A (2-3) 111 1 m mmm RR R RR (2-4) 111,,tts tts mtmtms y hr yhr y hr (2-5) Each matrix isA contains the transition probabilities for each species group i under harvest system s:
44 1 22 23 is isis isisis insinsa ba ba ba A (2-6) where ijsais the probability that a tree in species group i and size j under harvest system s will remain alive in size j during the interval t to t ,and ijsb is the probability that a tree in species group i and size 1 j under harvest system s will remain alive and grow into j from size 1 j during the interval t to t A myriad of procedures exist to estimate the growth matrix transitions. As in Boltz and Carter (2006), multinomial logit regression was us ed to estimate the transition probabilities over a 10-year period. Mortality incl udes both natural death and death from harvest damage that did not immediately kill the tree as identified in the immediate post-harvest inventory of the stands. Where ijscis the probability that a tree in species group i and diameter class j under harvest system s will die during the interval t to t the following must hold true: 1,,and 1,andijsijsijs ijsijsabcisjn acisjn (2-7) Where 1vis defined as mortality, 2v is defined as stability, and 3v is defined as upgrowth, the probability that any given tree u will transition to state v is a function of a set of parameters, v given by: 3 1Prob()vu vux u x ve Treev e (2-8)
45 The transition probabilities ar e estimated using maximum lik elihood as a function of tree attributes and harvest system employed. The first independent variab le, diameter at breast height, DBH, reflects the relative dominance of each tr ee in its immediate forest context. iSG, a series of dummy variables representing species group 1,...,4i, where 0iSG if species group is not i and 1 otherwise, is included to incorpor ate growth and mortality differences across species group as compared to the base species group, the emergent species. An interaction between tree diameter and species group, iSGDBH is included to incorporate differences across species groups and tree size. Finally, a set of dummy variables, s Twhere 0sT if the harvest system employed is not s and 1 otherwise, are included to incorporate the impact of logging technique as compared to th e base control plot treatment. It is assumed that the data do not violate th e Independence of Irrelevant Alternatives (IIA) property, which requires that the ratio of the probability of choosing between any two alternatives is independent of the presence or absence of any ot her alternative in the choice set (Hausman and McFadden, 1984). In the case of th e MNL estimation in this study, this property implies, for example, that the ratio of the upgrow th and stability probabilities for any given tree is not dependent upon whether mortality is includ ed as an outcome in the model, a reasonable assumption for the purposes of this model. The recruitment of new trees is expected to be negatively related to the total number of post-harvest trees per ha at time t, tTPH. Where k (=1,, m-1) now indexes a set of dummy variables representing species groups, an intera ction term between species group and trees per ha, ktSGTPH is added to reflect species-group level recruitment response to stand density. The fixed component of recruitment is hypothesized to be a function of species group and an
46 interaction between species group and harvest system employed, ksSGT to reflect differential responses to harvest acro ss species groups. Where it I equals the number of newly recruited trees of species group i at time t, the recruitment equation can be written as: 1112 0 1111 mmm itkktkktksks kkks I ccSGeTPHfSGTPHgSGTi (2-9) where 0,,,,,andkkksccefgare parameters to be estimated using ordinary least squares regression. To build each submatrix ikR, let i and k again index the full set of species group, ,1,...,ikm Each submatrix ikRin Equation 2-4 incorporates th e density-dependent effect of the trees of species k on the recruitment of trees of species group i, including, when ik, owneffects of species group i. The matrix entries for ,,andkkkscfgare equal to zero for the reference species group in Equation 2-9, or when ,ikm Using the estimated coefficients of Equation 29, each submatrix ikRis constructed in the following manner: 000 000kkk ikefefef R (2-10) Meanwhile, the fixed component of recruitment for species group i, which depends on species group but not overall stand density, is en tered in vector form as follows: 00 0iis isccg r. (2-11) Now, imagine two types of forest stand, one that is logged strictly according to a cutting cycle and a stand th at is logged at 0 t and never logged again. Fo r the repeatedly logged stand,
47 the length of the cutting cycle is denominated by where represents the number of growth periods between entries. As an example, the ev olution of the diameter distribution of a stand repeatedly logged using RIL (s = 1) would be determined by th e following iterative procedure: 1001,01 200 00 1 11,1 1... yGyhdr yGyr yGyr y G y hdr (2-12) To project the stand repeatedly logged using CL simply replace the RIL growth matrix and the damage and recruitment v ectors with those derived from the CL treatment. Meanwhile, where equals the number of growth peri ods projected into the future, the projections for the logged and le ft stand using RIL is determin ed iteratively by the procedure: 1001,01 200 2 1 000 0... .i i y G y hdr yGyr y G y Gr (2-13) This formulation is equivalent to the projec tion of an unlogged stand, only with the initial diameter distribution determined by the post-logged forest, rather than the distribution of the unlogged stand. As in the clas sical model of Buongiorno and Mich ie (1980), the projected stand will approach a steady state, where tt yyy and y is the equilibrium distribution, which is found using: 1 00. y IGr (2-14)
48 While it appears a strong assumption that the unlogged and logged stands will ultimately converge, it is important to note that this c onvergence may require extremely long periods of time, depending on the characteristics of ha rvest, damage, growth, and recruitment. Estimation Results Growth Model Using data from the full 73.5 ha study site, th e MNL estimation of Equation 2-9 resulted in a significant model (p < 0.001). Results are presented in Table 2-1. The Na gelkerke Pseudo-R2 = 0.331, which is relatively high for a discrete ch oice model, indicates an overall good fit to the data. However, largely because of the fact th ere are significantly more observations at lower sizes than in the higher sizes, some observati ons at higher diameter s exert relatively high leverage over the parameter estim ates. Pearsons goodness-of-fit te st returns a significant chisquare score, indicating the mode l does not fit the data well throughout the dataset, even though the global pseudo R-square measure is high, hi ghlighting the problem at higher diameters. Most of the variables in th e model proved to be significant predictors of transition probabilities. DBH was not signi ficant for the upgrowth trans ition, while it was strongly significant for stability. The first-order para meters for each of the species groups were significant. The interaction be tween species group and DBH were significant for the stability transition, while only the interaction between pioneer species and diameter was significant for the upgrowth transition. The RIL treatment was significant for the both the upgrowth and stability categories, although weakly for the stab ility transition. Meanwhile, the CL treatment was significant for the stability parameter only. Damage The damage vector was estimated from the 49 ha logged area (24.5 ha in the RIL and CL treatments, respectively). However, in the la rger plot area only the commercial trees were
49 measured within the 10-25 cm DBH range. The damage estimates within this size range are drawn from the intensively measur ed 5.25 ha smaller plots. Each element of damage vector, s d is estimated as the percentage of trees killed in species group i and size j under treatment s per tree harvested within treatment s. While a model that estimates damage according to the size or species of the tree that has been felled is mo re appropriate, the damages observed in the dataset are not recorded as arising from a specific tree be ing felled. Consequently, the vector estimated here best uses the available data. Tables 2-2 and 2-3 present the estimated damage across species groups, size, and logging treatments. It is evident from these tables that the intermediate, shade-tolerant, and emergent spec ies groups were more heavily damaged in the CL treatment. Also, larger trees incurred relatively more damage in the CL area. These comparisons reveal that management activities, such as vine-cutting and directional felling, bett er protect the residual stand. Particularly important is the reduction in residual damage to the emergent species group, as they are by a large margin the most economically valuable group. Recruitment In order to generate the data required to estimate the effects of stand density on recruitment, the three 5.25 ha treatment areas we re each divided into 10 equally-sized and shaped units, a strategy similar to th at of Buongiorno et al. (1997). TPH was then calculated for the 30 subplots, creating the data to estimate Equation 29. As Table 2-4 shows, the overall recruitment model is significant (p < 0.001) with an adjusted R-square of 0.54. As expected, recruitment is negatively related to total stand density. With th e exception of the pioneer species group, species group is also significantly related to recruitment. In the case of the pioneer species, however, a strong positive relationship emerges in the pioneer species interactio ns with harvest system type, an expected relationship as pioneer species tend to thrive in the pos t-logged forest environment.
50 No other species group and harvest system intera ction was significant; th ese interactions are excluded from the recruitment model. Additiona lly, the only species group and stand density interaction that proved significan t was the interaction for the in termediate species; the other interactions are also withheld from the model In early trials of the model, a problem emerge d with recruitment with the emergent species group. Recruitment into the em ergent class was low over the 10 year observation period, 1.1 stems/ha in the control area, 1.2 stems/ha in th e RIL area, and 0.7 stems/ha in the CL area. Because the predicted TPH can oscillate over 30 % depending on the intensity of harvests, TPH could become sufficiently high su ch that the emergent recruitment could be suppressed to a negative number, clearly impossible. A recruitmen t function that eliminated this problem could not be found. Therefore, the recruitment f unction was over-ridden for the emergent species group and replaced with the observed recruitment ju st mentioned, an unfortunate but necessary calibration. The resulting growth matrices for ea ch treatment are presented in Tables 2-5, 2-6, and 2-7. Discussion Validity Ideally, model predictions shoul d be compared to outcomes independent of the data used to estimate the model. However, no data of this sort currently exists and the best validation technique available is to assess the quality of the predictions with the actu al state of the Fazenda Sete stand after 10 years, as s hown in Figure 2-1. The predic ted stands are the result of projecting each treatment area of the site after incorporating the actual harvest and damage incurred during the experiment. The overall pr ojected diameter distri butions for the unlogged and CL stands appear to be reasonably close, while the RIL appears to under-predict at lower diameters. Examination at the species-group level bears out these observations. The under-
51 prediction at the lower diameters in the RIL simu lation is largely due to the under-prediction of newly recruited trees from the shade-tolerant gr oup. Attempts at improving the performance of the recruitment function to predict shade-tole rant recruitment under RIL were unsuccessful. Log and Leave Scenarios To exhibit the use of the model, three 100year simulations were performed based upon identical initial conditions, the average condition across the Fazenda Sete site. In the first simulation, no harvest was performed, so the pr ojection is simply that of the unlogged stand 100 years into the future. This simulation presents a richer context for the harvest scenarios. An identical harvest was simulated in the RIL and CL scenarios in order to compare the post-harvest trajectories of the three st ands. The simulated harves t volume was set at 30 m3/ha and the minimum DBH was set at 50 cm to reflect current regulations. Brazilian regulations requiring retention of at least 10 % of merchantable trees per species for seed trees were applied at the species group-level, as necessitated by the models structure. Since the initial forest contains 41.0 m3/ha of merchantable timber (non-hollow timber of good form from commercial species), the harvest was performed such that the most valuable stems were removed first, to simulate the behavi or of a profit-maximizing forest manager. Table 2-8 shows the simulated distri bution of the harvest and stem s killed by harvest damage distributed across species groups. While 5.8 stem s/ha were harvested in both the RIL and CL scenarios, 113.5 stems/ha were inadvertently kill ed in the CL harvest op eration compared to the 83.8 stems/ha killed in the RIL operation, representing a 26% reduc tion in direct harvest damage when RIL is adopted. Figure 2-2 shows the simulated trajectories at the species-level for th e three differentiated treatment types. Particularly evident is the dram atic jump in pioneer species in the post-harvest environment. The jump in pioneer s is strongest in the CL scenario which is likely due to larger
52 gaps and greater damage to understory vegeta tion created by the rela tively careless logging operation. Many, but not all, pioneer species are s hort-lived so while the surge occurs in the first 10 years of the simulation, the wave of pioneers still has not cleared fr om the system after 100 years. As the large pioneer populati on causes a strong negative effect of recruitment on other species, the projections portray a st rong floristic shift as the stand, at the species-level, recovers unevenly. Other species-groups do not shift as dramatically as the pioneer group but generally take decades to recover. Th e light-demanding group, which lose s about 20% of its population during logging, surges in the firs t 10 years to levels near or above the levels of the unlogged site, as would be expected given the groups ecological characteristics. Yet, this group suffers an increase in mortality in the fi rst few decades to drop the population below that of the unlogged site. The population of the inte rmediate species-group declined in the unlogged scenario, which indicates that the unlogged forest was unlikely to have been in equilibrium when the experiment began. While the shade-tolerant group contains a rela tively low proportion of commercial species, the commercial component of this group was heav ily harvested. Additionally, the large shadetolerant species group received heavy damage during the harvest, particularly in the CL scenario. Valuable shade-tolerant species like maar anduba can compose a large proportion of the commercial harvest, so damage to this species group can have signifi cant impacts upon future harvests. Again, after 100 year s, the population of this group is still appr oaching that of the unlogged forest from below. In contrast to the shade-tolerant species, al l emergent species have high economic values. Consequently, all permissible merchantable timbe r is logged from the emergent species group
53 first. Since the initial population distribution is relative ly flat and recruitment of these species is low, the emergent species are unlikely to rec over their initial merchantable volume within a reasonable management horizon. Aggregate projections of basa l area (Figure 2-3) shows that the stand logged using RIL will recover its basal area rapidly, within 20 to 30 years, while the CL stand requires at least 50 years to recover a total basal ar ea approximating that of the unlogge d site. Yet, as an indication of the shift in species composition, merchantable volumes do not recover within the 100-year horizon of the analysis, indicati ng a switch away from commercial species (Figure 2-4). The recovery of merchantable timber is relatively slow in both the RIL and CL cases. If the RIL logger returned in 30 years, a typical cutting cycle length in the East ern Amazon-region, only 18.3 m3/ha of merchantable timber would be availabl e, assuming the same suite of species were commercial then as now. The CL logger woul d encounter a merchant able volume of 15.6 m3/ha. The chapters that follow will examine timber yi elds in great detail, but it is important to note here that the model is depicting a dynami c that is becoming well-known, that logging induces large shifts in floristic composition wh ile commercial volumes are unlikely to recover within a typical management ti me-frame (Dauber et al., 2006; Karsenty and Gourlet-Fleury, 2006; Keller et al., 2007; Sist a nd Ferreira, 2007; Van Gardinge n et al., 2006). In this simulation, the RIL logger would need to wait about 80 year s to repeat a 30 m3/ha harvest. The CL logger would need to wait up to 100 years. Th e results call into questi on the sustainability of expected timber yields from pe rmanent production forests under the current set of regulations in Brazil. The results reaffirm findings of other recen t Eastern Amazon forest growth and yield studies. In the Tapajs Nationa l Forest, Van Gardinge n et al. (2006) estima te a maximum annual
54 commercial increment of 0.33 m3/ha/year for a forest logged us ing RIL. Meanwhile, the annual commercial increments estimated in Valle et al. (2007) for the same Fazenda Sete forest after accounting for defects is about 0.30 m3/ha/year for a forest logged using RIL and 0.20 m3/ha/year for a forest logged using CL. The average annual commercial increment estimated using this matrix model is 0.42 m3/ha/year in the first ten year s of the RIL simulation and 0.27 m3/ha/year in the first ten years of the CL simula tion (Figure 2-5). The increment declines over time for the RIL stand, while the increment for th e CL stand increases slightly from year 40 to year 70 before decreasing again. Conclusion The growth model was estimated using a discre te choice model with da ta for a 10-year-old logging experiment in one of the older logging fr ontiers of the Brazilian Amazon. To make the model tractable for economic and policy analysis the model is based upon extensions of the classic density-dependent Buongior no and Michie (1980) model that is often used in forest economics and management research. Two adaptati ons were made to the basic model. First, logging damage sensitive to the in tensity of the harvest was include d. Second, to better reflect the growth trajectory of a logge d forest, growth matrices were estimated as a function of logging type. It is expected that these modifications will improve the performance of the matrix model within a tropical forestry context. The adoption of RIL and improved forest manage ment practices is crucial for the long term sustainability of working forest s in the tropics. Using this model within an optimization framework, it will be relatively straightforw ard to allow the logger a choice over the logging system to incorporate, as a function of economi c incentives and regulati on. Understanding how incentives and regulations affect this decisionmaking is an im portant ingredient for tropical forest policy analysis.
55 If the objective is to manage forest resources sustainably, po licymakers and planners need to be explicit not only about the su stainable yields in an absolute sense, but the sustainability of components of the dynamic system. It is clear th at logging influences the floristic composition of the forest, both through selective economic pr essure on high-value timber species and through complex ecological interactions which are driven by intense harvest disturbances. While detailed examination of this issue is left for the chapter that follows, the simple simulations shown in this study show the depleti on of the higher-value timber species.
56 Table 2-1. Maximum likelihood estimates of transition parameters MLE parameter Upgrowth Stability Variable B S.E. B S.E. DBH 0.0760.056* 0.406 0.049 Pioneer 0.9100.188 0.799 0.179 Light-demanding -0.3450.104 0.234 0.090 Intermediate -0.2190.110 0.905 0.092 Shade-tolerant -0.2620.091 1.782 0.071 DBH x pioneer -0.3760.093 -0.479 0.083 DBH x light-demanding -0.0490.073* -0.274 0.063 DBH x intermediate -0.0010.068* -0.308 0.060 DBH x shade-tolerant -0.0690.065* -0.520 0.056 RIL treatment 0.6980.064 0.083 0.055x CL treatment 0.0210.067* -0.156 0.054 N = 6351 Chi-square = 4964.1 (df = 22) -2 log likelihood = 1260.4 Nagelkerke pseudo R2 = 0.33 Notes: Not significant. x Significant at p < 0.15. All other values are significant at P < 0.05.
57 Table 2-2. Proportion of popula tion per species group and size ki lled per tree harvested under RIL treatment Size Pioneer Lightdemanding Intermediate Shadetolerant Emergent 10-20 cm 0.06 0.04 0.04 0.04 0.02 20-30 cm 0.02 0.01 0.01 0.01 0.02 30-40 cm 0.02 0.02 0.01 0.01 0.00 40-50 cm 0.00 0.00 0.01 0.00 0.02 50-60 cm 0.00 0.03 0.00 0.00 0.00 60-70 cm 0.00 0.03 0.00 0.00 0.00 70-80 cm 0.00 0.00 0.00 0.00 0.00 80-90 cm 0.00 0.00 0.00 0.00 0.00 90-100 cm 0.00 0.06 0.00 0.00 0.00 >100 cm 0.00 0.00 0.00 0.00 0.00
58 Table 2-3. Proportion of popula tion per species group and size ki lled per tree harvested under CL treatment Size Pioneer Lightdemanding Intermediate Shadetolerant Emergent 10-20 cm 0.03 0.03 0.04 0.05 0.01 20-30 cm 0.03 0.03 0.04 0.04 0.03 30-40 cm 0.02 0.02 0.02 0.02 0.01 40-50 cm 0.01 0.03 0.02 0.02 0.06 50-60 cm 0.03 0.01 0.01 0.02 0.00 60-70 cm 0.00 0.00 0.00 0.01 0.00 70-80 cm 0.00 0.03 0.00 0.01 0.00 80-90 cm 0.00 0.06 0.00 0.03 0.00 90-100 cm 0.00 0.06 0.00 0.00 0.00 >100 cm 0.00 0.00 0.00 0.00 0.00
59 Table 2-4. Ordinary least squares es timates of the recruitment parameters Variable Parameter St. Dev. Constant 55.97513.893 TPH -0.1190.032 Pioneer 7.2608.457 Light-demanding 25.9845.541 Intermediate -47.45333.063 x Shade-tolerant 42.6195.541 RIL x Pioneer 56.81611.056 CL x Pioneer 86.40311.153 TPH x Intermediate 0.1300.078 + N =150 Adjusted R-square = 0.54 F = 30.41 Notes: Not significant. + Significant at P < .10. x Significant at p < 0.15 All other values are significant at P < 0.05.
60 Table 2.5. No harvest growth matrix G0 1 2 3 Speciesgroup DBH (cm) 1020 2030 > 100 1020 2030 > 100 1020 2030 > 100 1 10-20 0.30 -0.12 -0.12-0.12-0.12-0.12-0.12 -0.12 -0.12 20-30 0.38 0.45 0.00 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.32 0.00 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.49 0.00 0.00 0.00 0.00 0.00 0.00 2 10-20 -0.12 -0.12 -0.120.34 -0.12-0.12-0.12 -0.12 -0.12 20-30 0.00 0.00 0.00 0.23 0.49 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.22 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.71 0.00 0.00 0.00 3 10-20 0.01 0.01 0.01 0.01 0.01 0.01 0.60 0.01 0.01 20-30 0.00 0.00 0.00 0.00 0.00 0.00 0.19 0.61 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.19 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.71 4 10-20 -0.12 -0.12 -0.12-0.12-0.12-0.12-0.12 -0.12 -0.12 20-30 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 5 10-20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 20-30 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Notes: Species group 1 = pioneer, 2 = light-demanding, 3 = intermed iate, 4 = shade-tolerant, 5 = emergent.
61 Table 2.5. Continued 4 5 Speciesgroup DBH (cm) 10-2020-30 > 100 10-2020-30> 100 1 10-20 -0.12 -0.12 -0.12-0.12-0.12-0.12 20-30 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 2 10-20 -0.12 -0.12 -0.12-0.12-0.12-0.12 20-30 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 3 10-20 0.01 0.01 0.01 0.01 0.01 0.01 20-30 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 4 10-20 0.63 -0.12 -0.12-0.12-0.12-0.12 20-30 0.11 0.73 0.00 0.00 0.00 0.00 30-40 0.00 0.12 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.51 0.00 0.00 0.00 5 10-20 0.00 0.00 0.00 0.42 0.00 0.00 20-30 0.00 0.00 0.00 0.30 0.51 0.00 30-40 0.00 0.00 0.00 0.00 0.26 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.95 Notes: Species group 1 = pioneer, 2 = light-demanding, 3 = intermed iate, 4 = shade-tolerant, 5 = emergent.
62 Table 2-6. RIL growth matrix G1 1 2 3 Speciesgroup DBH (cm) 10-20 20-30 > 100 10-2020-30> 100 10-2020-30 > 100 1 10-20 0.20 -0.12 -0.12-0.12-0.12-0.12-0.12-0.12 -0.12 20-30 0.53 0.36 0.00 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.47 0.00 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.48 0.00 0.00 0.00 0.00 0.00 0.00 2 10-20 -0.12 -0.12 -0.120.27 -0.12-0.12-0.12-0.12 -0.12 20-30 0.00 0.00 0.00 0.36 0.42 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.35 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.64 0.00 0.00 0.00 3 10-20 0.01 0.01 0.01 0.01 0.01 0.01 0.53 0.01 0.01 20-30 0.00 0.00 0.00 0.00 0.00 0.00 0.31 0.53 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.31 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.62 4 10-20 -0.12 -0.12 -0.12-0.12-0.12-0.12-0.12-0.12 -0.12 20-30 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 5 10-20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 20-30 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Notes: Species group 1 = pioneer, 2 = light-demanding, 3 = intermed iate, 4 = shade-tolerant, 5 = emergent.
63 Table 2.6. Continued 4 5 Speciesgroup DBH (cm) 10-2020-30 > 100 10-2020-30> 100 1 10-20 -0.12 -0.12 -0.12-0.12-0.12-0.12 20-30 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 2 10-20 -0.12 -0.12 -0.12-0.12-0.12-0.12 20-30 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 3 10-20 0.01 0.01 0.01 0.01 0.01 0.01 20-30 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 4 10-20 0.57 -0.12 -0.12-0.12-0.12-0.12 20-30 0.19 0.67 0.00 0.00 0.00 0.00 30-40 0.00 0.20 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.44 0.00 0.00 0.00 5 10-20 0.00 0.00 0.00 0.34 0.00 0.00 20-30 0.00 0.00 0.00 0.45 0.42 0.00 30-40 0.00 0.00 0.00 0.00 0.40 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.92 Notes: Species group 1 = pioneer, 2 = light-demanding, 3 = intermed iate, 4 = shade-tolerant, 5 = emergent.
64 Table 2.7. CL growth matrix G2 1 2 3 Speciesgroup DBH (cm) 10-20 20-30 > 100 10-2020-30> 100 10-2020-30 > 100 1 10-20 0.26 -0.12 -0.12-0.12-0.12-0.12-0.12-0.12 -0.12 20-30 0.40 0.41 0.00 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.34 0.00 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.45 0.00 0.00 0.00 0.00 0.00 0.00 2 10-20 -0.12 -0.12 -0.120.30 -0.12-0.12-0.12-0.12 -0.12 20-30 0.00 0.00 0.00 0.25 0.44 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.24 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.68 0.00 0.00 0.00 3 10-20 0.01 0.01 0.01 0.01 0.01 0.01 0.56 0.01 0.01 20-30 0.00 0.00 0.00 0.00 0.00 0.00 0.21 0.57 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.21 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.67 4 10-20 -0.12 -0.12 -0.12-0.12-0.12-0.12-0.12-0.12 -0.12 20-30 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 5 10-20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 20-30 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Notes: Species group 1 = pioneer, 2 = light-demanding, 3 = intermed iate, 4 = shade-tolerant, 5 = emergent.
65 Table 2.7. Continued 4 5 Speciesgroup DBH (cm) 10-2020-30 > 100 10-2020-30> 100 1 10-20 -0.12 -0.12 -0.12-0.12-0.12-0.12 20-30 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 2 10-20 -0.12 -0.12 -0.12-0.12-0.12-0.12 20-30 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 3 10-20 0.01 0.01 0.01 0.01 0.01 0.01 20-30 0.00 0.00 0.00 0.00 0.00 0.00 30-40 0.00 0.00 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.00 4 10-20 0.60 -0.12 -0.12-0.12-0.12-0.12 20-30 0.12 0.69 0.00 0.00 0.00 0.00 30-40 0.00 0.14 0.00 0.00 0.00 0.00 > 100 0.00 0.00 0.47 0.00 0.00 0.00 5 10-20 0.00 0.00 0.00 0.38 0.00 0.00 20-30 0.00 0.00 0.00 0.33 0.47 0.00 30-40 0.00 0.00 0.00 0.00 0.29 0.00 > 100 0.00 0.00 0.00 0.00 0.00 0.94 Notes: Species group 1 = pioneer, 2 = light-demanding, 3 = intermed iate, 4 = shade-tolerant, 5 = emergent.
66 Table 2-8. Simulated harvest and damage across species groups and harvest system Species group RIL and CL harvest (stems/ha) Killed by damage, RIL (stems/ha) Killed by damage, CL (stems/ha) Pioneer 0.36.04.7 Light-demanding 0.014.716.6 Intermediate 2.614.019.3 Shade-tolerant 1.548.672.3 Emergent 1.4 0.50.6 Total 5.883.8113.5
67 Figure 2-1. Actual and predicted 10-year diameter distributions (stems/ha). A) Control area. B) RIL area. C) CL Area A0 50 100 150 200 250 300 350 400 45010-2020-3030-4040-5050-6060-7070-8080-9090-100>100DBH (cm)Stems/H a Actual Predicte d B0 50 100 150 200 250 300 350 400 45010-2020-3030-4040-5050-6060-7070-8080-9090-100>100DBH (cm)Stems/H a Actual Predicte d C0 50 100 150 200 250 300 350 400 45010-2020-3030-4040-5050-6060-7070-8080-9090-100>100DBH (cm)Stems/H a Actual Predicte d
68 Figure 2-2. 100-year post-harve st projections across harves t system and species groups (stems/ha). A) Pioneer. B) Light-demanding. C) Intermediate. D) Shade-tolerant. E) Emergent. F) All species. A0 20 40 60 80 100 120 140 0102030405060708090100YearStems/Ha B60 65 70 75 80 85 90 95 0102030405060708090100YearStems/Ha F200 250 300 350 400 450 500 550 0102030405060708090100YearStems/Ha No Harvest RIL CL C0 20 40 60 80 100 0102030405060708090100YearStems/Ha D60 110 160 210 260 310 0102030405060708090100YearStems/Ha E0 1 2 3 4 5 6 7 8 9 100102030405060708090100YearStems/Ha No Harvest RIL CL
69 0 5 10 15 20 0102030405060708090100 Years after loggingBasal area (m2/ha) No harvest RIL CL Figure 2-3. 100-year post-harve st projection of basal area (m2/ha of stems > 10 cm DBH)
70 0 5 10 15 20 25 30 35 40 45 50 0102030405060708090100 Years after loggingVolume (m3/ha) No harvest RIL CL Figure 2-4. 100-year post-harvest projection of merchantable volume (m3/ha of merchantable stems > 50 cm DBH)
71 0.0 0.1 0.2 0.3 0.4 0.5 102030405060708090100 Years after loggingVolume incr. (m3/ha/year) RIL CL Figure 2-5. Projection of average annual increment (m3/ha/year of merchantable stems > 50 cm DBH
72 CHAPTER 3 THE SUSTAINABILITY OF TIMBER PRODUCTION FROM AN EASTERN AMAZONIAN FOREST Introduction Current timber production best practices are limited to re duced impact logging (RIL) systems which seek to minimize environmental impacts compared to conventional logging (CL) systems which constitute about 90% of the regions harvests (Zarin et al., 2007). In Brazil, firms are obliged to follow legal restrictions ai med at minimizing environmental damage and protecting the future pro ductivity of the forest. These restrictions include minimum diameter cutting limits, upper bounds on harvest intensity, th e retention of seed trees and individuals of rare species, and protection of riparian buffers a nd wildlife. These types of simple harvest rules are policy-determined approximations that are expected to achieve br oad objectives yet be administratively simple to apply and monitor (B oscolo and Vincent, 2003). There are serious questions as to whether the curr ent set of forest management re gulations combined with best harvest practices are adequate to ensure sust ained timber production (Sch ulze et al., 2005; Valle et al., 2007; Van Gardingen et al ., 2006; Zarin et al., 2007). Meanwhile, Brazilian policy is likely to require sustainable yields as a matter of policy. Further, regulators might consider imposing sustained-yield require ments at the species-level, an added layer of complexity. Yet, while many call for sustainable timber ha rvests, few have attempted to establish quantitative management objectives that can be applied in practice in Brazil. A notable exception includes Van Gardingen et al. (2006) who examine a variety of yield regulation scenarios in the Tapajs National Forest, ar riving at the recommendation that to achieve sustainable yields for that re gions forests, loggers must not harvest more timber than accumulates at a rate of 0.33 m3/ha/year in combination with no more than one-third of the
73 commercial stock being removed dur ing any single harvest. Van Gardingen et al. (2006) also state that management must ad apt to local tenure and ecologica l conditions as well as varying management objectives. While estimates based on ecological models are extremely important, particularly in regions such as the Brazilian Amazon where re latively little is known about forest dynamics under harvest pressure, the studies typically exclude economic pers pectives. Yields such as those recommended in Van Gardinge n et al. (2006) may in fact be ecologically feasible, but if followed may not be profitable. Economic st udies of the cost-effectiveness of forest management have been limited to a few infl uential studies of RI L projects (Bacha and Rodriguez, 2007; Barreto et al., 199 8; Holmes et al., 2002). However, these studies are static in that they evaluated the costs a nd benefits over short periods of time for harvests that remove nearly all merchantable volume. Meanwhile the underlying problem is dynamic. How do forests respond to different harv est intensities and types and how do decisions and constraints in the present time influence the oppor tunities in future periods? Objectives of the Study The objective of this chapter is to quantitatively analyze th e dynamic cost-effectiveness of best logging practices and sust ainable yield constrai nts using a dynamic optimization model. While sustainability can be viewed as a func tion of perspectives a nd values (Karsenty and Gourlet-Fleury, 2006), the paper proposes two ne w operational definitions, weakly sustained inventory (WSI) and strongly sustai ned inventory (SSI). These defi nitions of sustainability focus on sustaining standing timb er inventories across cutting cycle entries, rather than on sustaining harvest yields, as is typically the case. WSI, a variation on stand-level sustainability, requires that the overall standing volume of merchantable timber at each harvest cycle be non-declining into perpetuity. SSI, a variation on species-level sustainability, requires that the standing volume
74 of merchantable timber at the species-group le vel at each harvest cy cle be non-declining in perpetuity. By using this terminology, this study draws upon the discourse within economics concerning the substitutability of capital, where a weakly sustai nable economy permits substitution between human-made and natural capital to sustain utility levels while a strongly sustainable economy must additiona lly sustain natural capital st ock (Heal, 1998; Pearce et al., 1989). Identifying the optimal harvest paths under the proposed sustainability constraints requires a two-stage optimization process. First, the harvests and stan ding timber stocks are found for a maximum sustainable economic yield problem, inde pendent of initial conditions. This stage is achieved by applying an innovati on to the classic Buongiorno a nd Michie (1980) uneven-aged forest management optimization model. The in novation incorporates an equilibrium condition that improves the models abil ity to dynamically account for commercial hollow and defective trees left standing, an important concern in tropical forestry. The results of this optimization step produce the species-group and sta nd-level timber inventories that quantify the sustainability constraints, which are then used in the simula tions to assess the cost-effectiveness of applying the constraints. After the model is further developed, results comparing a series of simulations will be presented. To form a baseline reference for th e regulatory simulations, RI L and CL harvests in which the loggers are not bound by a ny regulations are first presented. RIL and CL harvests under the current set of Brazilian forest management regulations ar e then simulated to evaluate how well the regulations induce sustainable yi elds. Harvests by a RIL logger under the WSI constraint and under the SSI cons traint are then compared to the outcomes in the baseline and regulatory scenarios. The closing discussion w ill bring together the analysis and discussion
75 above to offer insights into the challenges of de fining and applying sustaina bility constraints in practice. The Model The model used in this chapter is base d upon the density-dependent multinomial logitbased matrix model developed in Chapter 2. The basic model is a tropical forest extension of the multi-species uneven-aged matrix model presen ted in Lu and Buongior no (1993), Lin et al. (1996), and Buongiorno et al (1995). The model also more fu lly accounts for the relationship between the choice of harvest method and damage than the method first presented in Boscolo and Vincent (2000). To perform the analysis in this chapter, the model requires further development in terms of the in troduction of merchantability rest rictions, harvest regulations, economic factors, and, in certain scenarios, harves t restrictions designed to achieve sustainability objectives. Merchantability Restrictions A typical forest area in the Eastern Amazon may contain more than 300 tree species and, depending on local and regional timber markets a nd species distributions, will have anywhere from a few to over 100 commercially valuable sp ecies (Lentini et al., 200 5; Verssimo et al., 1998). The list of commercially valuable spec ies varies in time, according to species availability, markets, and milling technology, among other factors. Additionally, many commercially valuable mill-sized trees have stem defects and/or hollow stems, rendering them effectively valueless to the logger (Holmes et al., 2002; Schulze, 2003; Valle et al., 2006; Vidal, 2004). In this study, there is a clear distinction between the comm ercial status of any given tree and its merchantability. If a given tree come s from a commercial species, the tree must be sufficiently large, of good stem form, and not hollo w to be purchased and processed by a mill. A tree that passes these tests is considered merchantable.
76 If the logger selects against hollows and def ective stems, the mode l must incorporate a mechanism to ensure that the logger harv ests only non-hollow commercial stems with good form, or projected volumes will be over-estimated (Valle et al., 2006). In terms of growth, the residual hollow and defective stems are assumed to pe rsist at the same rates as other trees in the stand. The model must also account for the fact that future recruiting tr ees will also have some likelihood of becoming hollow or defective. In the model, the merchantability of any given tree is a function of three factors, the commercial status of the species, the form of the stem, and whet her the tree is hollow. Let Sbe a mn x mn diagonal matrix whose elements represent the proportion of the stems from commercial species within each species group (1,..,)im and size class(1,..,) j n The matrix Q is a mn x mn diagonal matrix whose elements represent the proporti on of trees in each species group i and size j with stem form appropriate for milling. The matrix H is a mn x mn diagonal matrix whose elements represent the proportion of stems in each species group i and size j thought to be hollow. While the model is de veloped to accommodate these estimates as a function of the species group and size, the values fo r this study (Table 3-1) were estimated at the species group-level, as in Boltz (2003). Given that large hollow stemmed trees in the unl ogged forest are likely to be very old, the merchantability formulation can be modified to allow for the possibility that the relatively younger trees that recruit into th e larger size classes during th e simulation may have a lower likelihood of being hollow simply because they are likely to be harvested before becoming very old. These adjustments are reflected in Table 3-1, as H0i represents the proportion of initially standing trees per species gr oup that are hollow, while H1i represents the proportion of newly recruited trees that are hollow.
77 Now enters a crucial behavior al distinction between the RI L and CL loggers. While both types of logger will reject stems with poor form the RIL sawyer will test target trees by making a vertical incision in the tree with the chainsaw. This cut is sufficient to give the sawyer a sense of whether the tree is hollow and the diameter of the hollow, enough information to avoid felling the tree if the hollow is significant. Meanwhile, the CL sawyer is assumed to be able to visually identify that a proportion of th e hollow trees are in fact hollow and avoid cutting them down, but will not perform the same tests as the RIL sawyer. Consequently, the CL sawyer will make mistakes and cut some, but not all, hollow trees. Let the proportion of hollow trees not identified as hollow (i.e. mistakes) be represented by, s For the RIL logger, 10 and, for the CL logger, 20,1 Hence, where s = 1 for the RIL logger, s = 2 for the CL logger, and I is the identity matrix, the perceived merchantable proportion for each species group i and size j is placed within the diagonal of the mn x mn matrix s M according to the equation: 1ss MSQIH (3-1) In this study it is assumed that 20.5 The assumed values for s imply that the CL logger will perceive a slightly inflated stock of mercha ntable timber. Table 3-1 shows the complete set of merchantability factors for the loggers. Where 0 y represents the initial unlogg ed distribution of the stan d, the initial pre-harvest perceived standing stock of merchantable timber, 0m y is calculated as: 00 m s y M y (3-2) During the first entry either of th e loggers will honor the followi ng constraint and not harvest and kill by damage more than the logger perceives is merchantable: 000mm s hd y (3-3)
78 where m s td contains the merchantable trees killed acci dentally during harvest. The CL logger will not gain revenues by felling hollow trees, only incur variable costs. Recruitment is also likely to generate new hollows or trees of poor form. Over many harvest cycles, the stand is likely to accumula te a proportion of non-merchantable trees far exceeding the proportion of non-merchantable trees in the initial stand (Va lle et al., 2007; Valle et al., 2006). In fact, Valle et al. (2007) predict th at the commercial volume of a stand logged under RIL will be almost entirely composed by defective stems after 60 years under a 30-year cutting cycle. About one-third of the commercial volume in a comparable stand logged using CL will be composed of defective stems as the CL logger is less selective and removes hollow trees (Valle et al., 2007). With th e exception of Boltz (2003), it is unclear whether similar matrix model-based economic studies adequate ly incorporate this dynamic. If 00 h, the merchantable stock,m t y at time t > 0, is comprised of the growing stock of residual merchantable stems remaining after harv est plus any upgrowth in to this merchantable stock during the cutting cycle interval. Where s G is defined in Equation 2-2 and letting *1 0 s s GGGindicate the cycle of growth matrix applications that arise from the loggers choice of harvest system, m t y at the next cutti ng cycle is given by: 2 *1 000 0 mmmi tsttstsss i y G y hdGMrGMr. (3-4) Written as a constraint upon harvests at t > 0, the logger cannot harvest and kill by damage at time t more merchantable stems than are perceived to exist at time t: mm tstt hd y (3-5)
79 Equilibrium Dynamics Now, assume that all harvests ac ross time are identical, such that *tt hhh and performed with a constant logging system accordi ng to a strict cutting cycle. Drawing on the seminal analysis of Buongiorno and Michie ( 1980), previous multi-species multi-age studies examining steady-state dynamics, such as Boltz (2003), assumed that th e overall stand must maintain an equilibrium di ameter distribution (i.e. tt yyy) in order to sustain the equilibrium harvest. Solving for a steady-st ate harvest in this model, however, requires considering the merchantable proportion of the st and as its own sub-stand and solving for an equilibrium diameter distribution of merchantable trees, or mmm ttyyy. Under this solution, the overall stand is allowed to vary fr om cutting cycle to cut ting cycle to accommodate the constant supply of merchantab le timber. To capture the dynami cs of the merchantable stand, it is necessary to modify the e quilibrium growth constraint of the Buongiorno and Michie (1980) model to impose the equilibrium conditions of the merchantable sub-stand: *2 ****1 000 0 mmi s stssss i GhdIG y GMrGMr (3-6) In words, the equation states th at the sum of harvest, harvest damage, and mortality must equal recruitment over any given cutting cycle interval at the given choices of **andmh y Brazilian Regulatory Policy The regulatory agency has available a set of si mple rules that it can use to try to induce desired logging behavior. In this chapter, the logger is assumed to perfectly comply with these rules, an assumption loosened in the next chap ter. The regulatory agency may apply diameter cutting limits, harvest volume intensity limits (at the overall stand a nd species-level), and provisions for leaving seed and ra re trees standing. The regulati ons are assumed to remain fixed
80 throughout the horizon of the scenario. The scenar io in which only these standard regulations are applied is called the Brazilian Regulatory Policy (BRP) case, and the specific parameters are drawn from a set of forest management re gulations enacted in 2006 (Instruo Normativa 05/2006). While the diameter cutting limit may be determined at the species group-level, the diameter cutting limit constraint used in this study is set at 50 cm DBH across all species-groups to reflect the most general case of the currently proposed regulations in Brazil: 4 10ij jhi (3-7) In some circumstances, this constraint can be co nsidered a merchantability constraint, as mills generally will accept logs over a ce rtain size. The commercial volume of stems greater than or equal to commercial size was calculated by appl ying the commercial volu me equations (with bark) reported in Silva et al. (1984) to the comme rcial trees in the permanent sample plot (PSP) data. Given the unlogged stand may contain very large commercial trees which will be logged upon first entry, to minimize the risk of overs tating the available commercial timber, it is necessary to reduce the volume in the largest diam eter class in future entries since insufficient time is likely to pass between cutting cycles to a llow commercial trees to grow to the very large sizes found in the largest diameter class in th e initial stand. The predicted volumes were averaged across species groups and size and placed into the mn x 1 vector the elements of which are shown in Table 3-2. During the second harvest and beyond, the commercial volume of trees > 100 cm DBH is assumed to be equal to the volume of a 110 cm DBH tree in order to account for the above concern with over-stating co mmercial volumes for very large trees. The merchantable volume at any time t is found by multiplying the tr anspose of the commercial
81 volume vector against the harvest vector at time t or 't h. The harvest intensity regulatory constraint, 30 m3/ha per entry, is written as: '30tt h (3-8) To retain seed trees, management standards obligate the logger to retain at least 10% of the stems of any given species that would othe rwise be authorized for logging, or: 1010 550.9,ijtijt jjhyit (3-9) While Brazilian forest regulations are unclear on whether the practice is permissible, forest managers may often satisfy this constraint by the retention of hollow stems or stems of poor form that are permissible to log, but in practice are left standing. In this study, the logger is able use non-merchantable trees to satisfy the constraint. Sustainability Constraints As discussed in the introduction, sustainability can be viewed as a matter of perspective and analytical judgment. In the study forest, fo r example, the rate at which the harvest volume recovers after the first harvest is, as would be expected, very much a function of the harvest intensity. Figure 3-1 shows es timates of volume growth afte r first harvests ranging from 15m3/ha to 40m3/ha, implementing RIL in each case. Hi gher harvest intensiti es lead to more rapid growth. If the sustained yield goal is simp ly to harvest the growth then the rules will be highly dependent upon the choice of first harves t and cutting cycle length, for which an unlimited number of combinations exist in practice. Additionally, the decision to harvest what is grown is essentially non-economic, a function of the growth potentia l of the stand alone, rather than the growth of value. The specific quantitative definitions of sustaina bility in this study are generated from within the bio economic system as a function of the maximum sustainable
82 economic yield, rather than the maximum sustai nable accumulation of a physical characteristic of the system, such as the commercial vo lume or number of commercial-sized trees. As mentioned earlier, sustainable yield c onstraints assume two definitions: weakly sustained inventory (WSI) and st rongly sustained invent ory (SSI). WSI, a variation on standlevel sustainability, requires that the overall standing volume of merchantable timber at each harvest cycle be non-declining into perpetuity, without respect to the maintenance of volumes of any particular species group. SSI, a variation on species-level sustainabili ty, requires that the standing volume of merchantable timber at the species-group level at ea ch harvest cycle be nondeclining in perpetuity. While it would be more appropriate to apply the SSI constraint at the individual species-level, targeting the constraint s at the group-level is a simplification required by the models design. The problem can be solved for the equilibrium harvest,* tt hhh, and size distribution, mmm ttyyythat maximizes an economic objective. This solution determines the maximum sustained economic yi eld (MSY), but this yield es timate is independent of the initial conditions of the stand, a limitation discusse d in detail in Boltz (2003). Given an initial stand size distribution and a feasible target distribution, the series of regulation harvests that lead to this steady state can be identified a nd optimized (Buongiorno, 2001). Given the initial unregulated condition of a typical forest in the region, this period of regulation may require such a long period of time that it is uninteresting for the policy context of the study. Similarities between the constrained harvests of the WSI and SSI harvests and the regulation cuts required to achieve a steady state may emerge indirectly, but modeling the regulation cuts directly is not an explicit objective here.
83 If the first harvest is performed at t = 0, where z indexes the number of cutting cycles, and the pre-harvest merchantable volume is found at any time t by the matrix operation 'm t y WSI requires that the pre-harvest st and-level merchantable timber never diminishes below the equilibrium pre-harvest merchantable volume found by solving the MSY problem: *'''asmmm zz y y y. (3-10) Note that this constraint exempts the merchantab le stock of the initial stand and proposes nothing directly about harvest-levels. Meanwhile, exempting the merchantable stock of the initial stand, the SSI constraint requires that the merchantab le timber never diminishes at the species-group level: ,, 111asnnn mmm ijijijijzijij jjj y yyzi (3-11) There may be situations in which the sustainabilit y of a given species or group of species is not a management challenge, exempting the species or group of species from the constraint. In the Eastern Amazon, for example, pioneer species with little or no commercial value thrive in the post-harvest environment and, hence, are by defi nition sustainably manage d within a repeatedly logged stand. Additionally, the surge of pioneers can out-compete the juveniles of more valuable timber species. In practical terms here, the SSI cons traint will not be applied to pioneer species. The SSI constraint is not likely to produce the biologically most rapid approach path to the maximum sustained economic yield. However, it is asserted but not pr oved here that the combination of the yield constraints and prof it maximizing behavior, applied over many cutting cycles, will iteratively yield ha rvest and standing merchantable diameters approaching the MSY distributions. Additionally, satisfying the sust ainability requirement in either case does not necessarily imply that harvest intensity is non-declining. In fact, depending on the stands
84 population structure and dynamics and the length of the cutting cycle, harvest intensity may oscillate, or even be zero if this is what is required to satisfy the i nventory requirements. Because the overall study is concerned with logging concession agreements of finite length, two simplifications of the WSI and SSI constraints are necessary for this work. For example, say a contract specifies two harvest entries under a fixed cutt ing cycle, permitting entries at t = 0 and t where is the time-step of the matrix model and is the number of time-steps between harvest entries. Assuming si milar terms will define the future contract, assuming one is issued, the terminal merchantab ility constraint could maintain the economic conditions of the stand for the first en try of the subsequent contract at 2t to that of the first contract. The modified constrai nt for the WSI case is written: 2'''mmm y y y (3-12) and, where 1 i for the pioneers species group, for the SSI case is written: ,,2 1111nnn mmm ijijijijijij jjj y yyi (3-13) Economic Variables For this analysis, the unit valu es of prices and costs are as sumed to be fixed throughout the time horizon and are expressed throughout th e study in US$2004. The price of a tree, ij p in from species group i in size j is presented in Table 3-3. Thes e values were calculated based upon the mill gate Free On Board prices in the Para gominas region, according to economic surveys of mill owners and operators (Lentini et al., 2005). Management costs Free On Board forest mill are classified as either variable or fixed cost s and are drawn from Barre to et al. (1998) and Lentini et al. (2005). Variable costs of harvesting a tree in species group i and size j under harvest system s is represented by ijsvc and includes the costs associ ated with felling, skidding
85 and log deck operations, and transportation costs incu rred relative to the level of harvest intensity and harvest system choice s. The costs are estimated at $15.64/m3 for the RIL logger and $13.48/m3 for the CL logger, and the per-tree costs ar e shown in Table 3-4. Fixed costs in $/ha, represented by s f c, are incurred regardless of harvest intensity at each cutting cycle entry for planning, capital costs, and transaction costs an d are estimated at $50.56/ha for the RIL logger and $13.91/ha for the CL logger. A constant real discount rate, r, of 9.75% was applied throughout the analysis. This rate was based upon the long term real interest rate re ported for the bulk of 2004 by Brazils National Bank of Economic and Social Development (Banco Nacional de Desenvolvimento Econmico e Social BNDES, 2007). Waste The benefits of implementing RIL, in partic ular from operational planning improvements, extend beyond the reduction of damage to the residual stand and minimizing the felling of hollow trees (Barreto et al., 1998; Boltz et al., 2003; Holmes et al., 2002). In the unplanned logging operation, the skidder operator will genera lly find felled trees by driving the skidder through the forest to logging gaps the operator visually identifies. RIL operators, in comparison, develop detailed maps that show the locations of felled trees, minimizing the loss of valuable timber, reducing expensive machine-time, and re ducing the damage incurred during the skidder operators search for felled trees. The RIL felling team will also waste relatively less commercial volume though improved felling and bucki ng techniques. In the model, wood waste of this form is incorporated after trees have been felled by deducting harvest revenues, while not deducting costs. Table 3-5 shows the rate at which deductions are applied against the RIL and CL loggers revenues to calculate the variable s waste, where s indexes the choice of harvest system employed, based upon estimates in Holmes et al. (2002).
86 Concessionaire Objectives In each of the scenarios that follow, the con cessionaire will be allowed two logging entries, the timing of which is strictly defined by the cutting cycle with the first harvest at 0 t The cutting cycle length, varies from 30 to 60 years. Fo r each cutting cycle length, seven scenarios are presented: Maximum sustainable economic yield (MSY) ha rvests for the RIL logger (from which the sustainability restrictions on the WSI and SSI cases are estimated) Unconstrained harvests (U) for the RIL and the CL logger, the baseline harvests in which the logger is not bound by any regulations Brazilian regulatory policy (BRP) harv ests for the RIL and the CL logger Weakly sustainable inventory (WS I) harvests for the RIL logger Strongly sustainable inventory (SS I) harvests for the RIL logger In each case, the logger will se ek to maximize the net presen t value (NPV) of two harvest cycles with the first harvest at t = 0. The profit at time t, t is calculated as: 1tsijijtijsijts ijwastephvchfc. (3-14) Where T equals the length of the concession contract and 1 1r represents the discount function, the net present value (NPV) of returns is then written: 0NPVT t t t (3-15) As mentioned earlier, the vector *m y that defines the pair of sustainability constraints is found by solving the MSY problem, independent of initial conditions. The objective function of the MSY scenario reflects th at the logger seeks to choos e the steady state harvest, *h, and preharvest merchantable stock *m y which maximize the economic objective while satisfying the
87 equilibrium merchantable growth constraint (Equation 3-17), the me rchantable harvest constraint (Equation 3-18), the regulatory constraints (Equations 3-19, 3-20 and 3-21), and the nonnegativity constraint (Equation 3-22): **, 0maxNPVmT t t t hy (3-16) subject to: 2 *****1 000 0mmi s stssss i GhdIG y GMrGMr (3-17) ** mm thd y (3-18) 4 10ij jh (3-19) '30tt h (3-20) 1010 550.9,ijij jjhyit (3-21) **,0mhy (3-22) In both RIL and CL cases for the U scenarios, the logger will be constrained by the growth model (Equation 3-24), the size restrictions on merchantable stems (Equation 3-25), the merchantable harvest constraint (Equation 3-26), the initial cond ition of the stand (Equation 327), and the non-negativity constraint (Equation 3-28): 0maxNPVtT t t t h (3-23) tsttsts yGyhdr (3-24) 4 10ij jh (3-25) mm ttt y hd (3-26)
88 0 actual yy (3-27) ,0tthy (3-28) The BRP logger is additionally constrained by the harvest intensity limit (Equation 3-29) and limits on harvesting all trees within any species group (Equation 3-30): '30tt h (3-29) 1010 550.9,ijtijt jjhyit (3-30) The WSI logger will be constrained by Equatio ns 3-24 through 3-30 and will be additionally constrained by the sustainability constr aint on overall merchantability stocks: 2'''mmm y y y (3-31) The SSI logger will be constrained by Equa tion 3-24 though 3-30 and wi ll be additionally constrained by the sustainability constraint on species-level stocks of merchantable timber: ,,2 1111nnn mmm ijijijijijij jjjyyyi (3-32) Recall that the constraint for pioneer sp ecies is removed in the SSI simulations. Results Maximum Sustainable Yield Harvests In order to compare the baseline (U) and policy (BRP) scenarios with the WSI and SSI scenarios, it is necessary to first solve th e MSY problem to identify the values of the sustainability constraints WSI and SSI. Table 36 shows the distribution of pre-harvest standing merchantable stock and harvests at the species group-level in terms of trees and volumes, across cutting cycle lengths. The table shows the rela tive dominance of the pioneer species group in the standing stock and harvest. While the highly-va luable emergent class has the most standing
89 merchantable volume at the solution in each case, the equilibrium harvests are very low, from about 7 to 10% of the standing stock of emergents. Meanwhile, because of the large number of trees from the pioneer group in the post-harvest stand and the fact that some of the pioneer species capable of reaching larger sizes are commercial, harvests from the pioneer group compose more than one-third of the equilibrium harvest at each cycle length. The lightdemanding, intermediate, and shadetolerant groups assume similar intermediate proportions in the optimally regulated stand. Because the large role of pioneer species in the post-harvest stand is one of the dominant practical management probl ems, the contribution of pioneer species in the maximum economic yield problem should be viewed with caution and definitely should not become a goal of management. As will be seen in subsequent sections the pioneer component of the SSI solution will be removed to address this management concern. The volumes harvested from each stand at th e aggregate and species-level provide the values for the WSI and SSI constraints, with th e exception of pioneer spec ies. Notice that the aggregate volumes harvested at the MSY solution, from 22.7 m3/ha under the 30-year cycle to 28.6 m3/ha under the 60-year cycle are not constrained by the 30 m3/ha limit prescribed by Brazilian law. Unconstrained and Brazilian Regulatory Policy Harvests Table 3-7 shows the timber volumes harvested and recovered during th e first harvest entry at t = 0 and during the second entry of the concession term for each of the scenarios. The table also shows the pre-harvest merchantable volume at the first and second entrie s, as well as at the time of the third entry, assuming the same cuttin g cycle length. Because the unconstrained CL sawyer cuts hollow trees, the total harvest volu me exceeds that of the standing merchantable stock in the first harvest, while the unconstrai ned RIL logger harvests slightly less than the merchantable timber because some of the valuable timber is lost to damage.
90 For each of the U and BRP scenarios, the first harvest for each cutting cycle was essentially the same, to remove all merchantab le timber in the unconstrained case and the most valuable permissible timber in the BRP scenario s. The value growth rate of no species was sufficient to leave permissible timber standing to grow into the next harvest period. The very high degree of waste in the CL case is evident as the RIL logger is able to recover about 16% more timber, even though the RIL logger logs slightly less. This result is consistent with Barreto et al. (1998) and Holmes (2002), and the reductio n in waste and damage, as will soon be seen, will resonate strongly though financial returns and available timber in future harvests. The benefits in adopting RIL to reduce harvest waste are significant acros s all cutting cycle lengths. Also, the benefits of reduced damages in th e first harvest resonate in future harvests. In the unconstrained scenarios, the logger who adopts RIL will have about 3-4 m3/ha more merchantable volume in the second harvest than the CL logger. The logger who adopts RIL under the BRP scenario will be able to harvest about 2 m3/ha more timber in the second harvest. While these volumes are not large in an absolute sense, especially considering the effects of discounting over long periods of time, from the futu re perspective of the lo gger, these differences are important. Assuming the harvest system adoption is consis tent between the first and second harvests, there is little difference between U and BRP scen arios in terms of available timber for the third harvest. Meanwhile, the adoption of RIL makes a signifi cant difference. This outcome is likely from the fact that in the first harvest, the BRP loggers are constrained to leave (the lowestvalued) merchantable volume standing. But, in the second harvest, this additional timber is now harvested, incurring relatively more damage in the stand, which, in a sense, permits the BRPconstrained loggers to catch up w ith their unconstrained counterparts.
91 For each cutting cycle length, the pre-harvest merchantable volume is higher for the third entry than the second. There is a pair of likely contributors to this tre nd. First, the growth surge stimulated by intense first harvests has, by the third entry, pushed through the system and is attaining commercial size. Second, b ecause the second harvest was relatively light compared to the first harvest, damage rates are re latively low, while the stand will still receive a relatively strong positive growth and regeneration effect because of the disturbance. This dynamic raises the possibility that, over time, the harvests arising in the unconstrained and Brazilian policy cases will oscillate. However, there is the possibility that the growth model overstates growth and recruitment after harvests of relatively low intensity, as the RIL and CL growth matrices were based upon harvests of about 30m3/ha. The density-dependence of recruitment is likely to reduce this risk, but is unlikely to fully remove the risk. Figure 3-2 shows the species-group composition of the total harvest for the first and second harvests and available merchantable stoc k for the first and second and the hypothetical third harvest, but only for a 40-year cutting cycle for graphical convenience. In the U scenario for the CL logger, the logger harvests more than is available because the CL sawyer fells hollow trees, which are not skidded from the forest. Fo r the U and BRP cases for both types of logger, the charts show how the composition of the harv est by species group shifts across harvests. The merchantable volume from the valuable emergent group is fully extinguished in the first harvest in the unregulated and regulated cases. As disc ussed earlier, because of regeneration challenges and relatively flat diameter distributions in the initial stand, when under periodic harvest pressure, the emergents are unlikely to ever regain volumes that approximate the initial condition. The distributions depi ct the differences between the in itial condition of the emergent species group and the post-l ogging condition (Figure 3-3).
92 The shade-tolerant, intermediate, and li ght-demanding groups are each diminished significantly from the first to second harvest, but th e intermediate and light-demanding species recover a larger proportion of th eir initial volume than do the emergent species. The dramatic shift in commercial composition toward pioneer sp ecies is clearly eviden t in the chart, with pioneers increasing from about 4% of the harves t to almost 40% in the unconstrained cases. Figure 3-3 shows the diameter distribution of pi oneer trees under each scenario with a cutting cycle of 40 years. The harvest-induced ingrowth of pioneer trees cause s waves of pioneers to push through the diameter distribution. Decreas ing harvest pressure by increasing constraints reduces the amplitude of the waves but they none-t he-less deform the distribution away from the typical inverse J-distribution. Cl early, this shift has economic implications. Future harvests are likely to be small in comparison to the first ha rvest and will simultaneously be less valuable per m3 extracted on an undiscounted basis, which has st rong implications for the sustainability of economic timber yields. Harvests under Weakly Sustainable Inventories Given RIL will be required on Brazilian public lands, the WSI and SSI scenarios were modeled for the RIL logger only. The WSI and SSI solutions are constr ucted such that the merchantable volume at the time of logging en try does not decrease below that defined by the MSY solution. For example, for a 40-year cutt ing cycle, the merchantable volume at time of logging entry cannot decrease below 35.1 m3/ha. Harvests can vary over time to accommodate this constraint. For the WSI scenarios, in order to satisfy the constraint, the first harvests are light relative to the first U and BRP harvests. By the second harvest, with the exception for the 30-year cutting cycles, the WSI harv ests are larger than the U and BRP harvests (Table 3-7). At that point, the WSI logger is s till constrained by the sustainabil ity harvest, which prevents the logger from liquidating the forest, but the harvest is such that the constrai nt is again met by the
93 hypothetical third harvest. Continuing to use th e 40 year cycle as an example, the WSI logger will encounter 75% more merchantable standing in the second entry than the BRP counterpart and about 33% more at the time of the third hypothetical entry. The WSI harvests at the species-group level te ll a slightly different story than the U and BRP harvests. The contribution of the emergent sp ecies is higher in the second WSI harvest than that of the U and BRP harvests (Figure 3-2). A lthough, it is evident that, by the third harvest, the commercial volume of emergent species is approach ing the same low levels as the scenarios not constrained by sustainability conc erns. The pioneer group is also assuming a larger role in the harvests. The future quantities of light-dema nding and shade-tolerant species under WSI also persist at relatively the same rate as under the BRP set of policies. The difference in species dynamics between the scenarios is th at intermediate species receive very light harvests relative to the available volume, implying that the WSI loggers strategy is to use the low-value intermediate species to satisfy the constraint, wh ile the emergent class is gradually eliminated. Note as an example that the idealized MSY logger applying a 40year cutting cycle harvests 24.8 m3/ha of 35.1 m3/ha standing at each entry for perp etuity. The WSI logger extracts 13.1 m3/ha of 41.0 m3/ha in the first entry, then 21.8 m3/ha of 35.1 m3/ha in the second entry. In future entries, this harvest is likely to increase gradually until the MSY solution is approximated, implying that, while the WSI logger is a constr ained profit maximizer, th e extraction path the logger is following is also one that is regulating the stand towa rd a long-term MSY equilibrium. Comparison of the dynamics of the emergent speci es under the WSI and MSY solutions at year 80 for the 40-year cutting cycle sh ows that this period of regulation may take a very long period of time. In the MSY case, while the available merchantable volume of the emergent species
94 group stands over 10 m3/ha, which permits a small but sust ainable harvest of the high-valued species, the emergents of the WS I solution stands at around 2 m3/ha. In the 40-year cutting cycle example just disc ussed, the logger may be able to extract high volumes of lower value species and satisfy the bi ological constraints but may not extract as much profit from the stand. As can be seen in Tabl e 3-7, the intensity of the first harvest rises relatively slowly as the cutting cycle lengthens. Because more valuable species tend to grow more slowly, the optimizing logge r seems to be making (slight) trade-offs between volume and value. Yet, as a result of the second harvest be ing discounted farther into the future, the second harvest becomes less valuable, more-or-less eliminating these gains. Harvests under Strongly Sustainable Inventories For each cutting cycle length, the first harves t under the SSI constraint exceeds the first harvest under the WSI constraint (Table 3-7). Al so, for the 30 and 40-year cutting cycle lengths, the second SSI-constrained harvest exceeds the s econd harvest under the WSI constraint (Table 3-7). In fact, the second SSI-constrained harves t volumes under 30 and 40-year cutting cycles exceed the second harvests of all of the scenario s which are constrained by the initial standing stock. This result is intere sting in that it highlights the importance of using an economic perspective in this type of analysis. In the U and BRP scenarios, the concessionaire harvests the most valuable timber first, caring very little a bout future growth. Under the WSI constraint, the concessionaire harvests valuable but slow grow ing timber, knowing that lower value species will grow sufficiently to satisfy the stand-level inventory volume constr aint. Under the SSI solution, this substitution is not possible, so lower value but fast-growing timber is harvested first, with very little of the valuable em ergent class being removed. In fact, the emergent species group beco mes more-or-less protected under the SSI constraint. For example, under a 40-year cutting cycles, no emergent stems are harvested in the
95 first harvest. In the se cond harvest, about .7 m3/ha is removed from a standing stock of about 10 m3/ha, approximately the same harvest rate as th e MSY solution. Figure 3-3 shows that the preharvest emergent distribution at years 40 and 80 has not changed significantly from the initial distribution, while the surge of pioneer species is muted in co mparison with other scenarios. Earlier in this chapter, it was asserted that the harvest path generated by a profitmaximizing concessionaire under th e sustainability constraints woul d approach the harvest path derived under the MSY solution. Figure 3-2 provid es visual evidence of this assertion, as the distribution of the pre-harvest merchantable volume across species groups under SSI constraints closely resembles that of the MSY solution. Financial Returns As highlighted in Barreto et al. (1998), Boltz et al. (2003), and Holm es et al. (2002), the comparison of financial returns (Table 3-8) de pends greatly upon the rela tive amounts of waste. While the CL loggers harvest more than their RIL counterparts, almost 20% less timber volume actually leaves the forest. For example, the unconstrained CL logger achieves an NPV of $479/ha when applying a 40-year cutting cycle, while the RIL logger earns an NPV of $604/ha, the difference between the two essentially being a result of loss within the forest. Meanwhile, the difference between the logge rs respective return s is compounded by the fact that the CL logger incurs marginal harvests costs by harves ting timber that produces no revenues. Given the impact of the high levels of waste on the CL logger, both the RIL U a nd BRP scenarios have higher NPVs than the unc onstrained CL logger. This observation connects this analysis to the larger question of why a logger would not adopt RIL practices if the financia l benefit is incentive compatible One possible answer to this issue is that RIL is cost-effective in some circumst ances, but not all. Putz et al. (2000) write that RIL may be not be cost-effective when loggers are restricted from accessing steep slopes or
96 when ground yarding of timber is not permitted during wet conditions. Because the terrain at the experimental forest of this study is relatively flat and the study area al so did not include any stream buffer areas that firms are re quired to set aside as reserves, th ere is a risk that benefits of RIL are over-stated because the estimated costs do not account for these factors. For example, according to a study of certification reports for certified timber firms operating in the Brazilian Amazon, the firms set aside about 9% of total fo rest area as permanen t buffer area (Schulze et al., 2007a). In cases where the benefits of adoption may otherwise be unambiguous, institutional constraints can also preclude adoption. Surv ey research on logging in the Brazilian Amazon revealed that many conventional loggers recognize that RIL, or forest management practices more generally, better protects fore st values and, under certain circ umstances, is cost effective, yet still do not adopt improved practices (Saboga l et al., 2006). Land tenure insecurity may prevent firms from pursuing improved environmenta l practices. Firms are unlikely to invest in practices that improve the economic value in futu re decades if they do not expect to have access to those forests decades in the future. Appropria te training, labor, and equipment may also be scarce (Sabogal et al., 2006). Institutional barriers, such as government corruption or inefficiency, may also exist that make operati ng conventionally or ille gally the only option for operators (Merry et al., 2006; Sa bogal et al., 2006). In any case, a type of technological path dependency can emerge where there has been histor ical distrust between loggers and government officials, particularly where government has been perceived to be indiffere nt to long term timber management (Smith et al., 2006). In this envi ronment, norms that promote poor management practices emerge, which may require long periods of time to change under improved enforcement and incentive systems (Smith et al., 2006).
97 With the exception of the SSI-constrained scenario, NPV is mostly invariant to cutting cycle length for the simple reason that, for each cutting cycle length of the unconstrained and Brazilian policy scenarios, the logger harvests all (in the unconstrain ed case) or the most valuable (in the policy case) merchantable tim ber. Lengthening the cutting cycle does not change anything with respect to th e first harvest. The discounted second harvest contributes very little to the NPV, regardless of volumes harvested. For the WSI scenarios, NPV falls then rises with cutting length because the longer periods of time between harvests permits increases in the allowable cuts in each harvest, shifting slightly toward slow-growing but more valuable species, but exchanges the gains for second harvest returns discount ed farther into the future. For the SSI scenario, the NPV increases with the length of th e cutting cycle, as the increasing length permits the concessionaire to harvest more valuable timbe r in the first harvest, knowing that the volume can recover by the next entry to satisfy the constraint. For the RIL logger under a 40 year cycle, fo llowing Brazilian policy generates about $80/ha in opportunity costs. The WSI scenar io presents opportunity costs of about $240/ha compared to the BRP RIL scenario and $320/ha to the U RIL scenario. The SSI scenario presents opportunity costs of about $360/ha comp ared to the BRP scenario and $440/ha to the U scenario. While the NPV is calculated from the present perspective of the logger, imagine the dynamic perspective of the future logger as the lo gger approaches the time of the second harvest. The WSI and SSI logger will encounter a significan tly richer stand then the counterpart U and BRP loggers. But to maintain the sustainability -constrained stands in perpetuity will always generate opportunity costs a nd risks of timber trespass. As mentioned before, the concessionaire who is governed by the SSI constraint is able to harvest more volume than the WSI-constrained conce ssionaire, but that volume is from relatively
98 low value species. Under a 30-year cutting cycle, the NPV under the SSI constraint is about half of that of the WSI concessionaire. Under a 60-year cutting cycles the NPV under the SSI constraint is just over two-thirds of the NPV under the WSI constraint. Given the SSI harvest is approaching that of the MSY solution, the trad e-off between short-term gains and achieving a dynamically optimal and, hence, sustainable path is clear. Discussion While demonstrating the dynamic benefits of RI L, model results exhibit how current best logging practices and current regulatory policies do not guarantee the sustainability of timber harvests. The study also shows the difficulty asso ciated with operationalizing sustainability in practice; sustaining timber yields at the stand-level led to significant opportunity costs while still leading to a stand that is cha nged in species composition and ec onomically degraded. Over two cutting cycles, the harvests under the proposed defi nition of sustainability at the species-level were profitable and led to a species group-leve l merchantable volume distribution approaching the distribution of the maximum sustainable economic yield. To achieve this, the SSI solution essentially protects valuable but slow-growing tim ber species from harvest, which again creates high opportunity costs and strong incentives fo r breaking the rules when enforcement is imperfect. These are clearly resu lts derived within a simulation of a representative stand, while the on-the-ground reality of the Brazilian Am azon is extremely complex. The following discussion will discuss a set of a pproaches that have been proposed to help induce sustainability and reduce opportunity costs of increased regulation. Silviculture A variety of silvicultural pr actices are currently performe d in the Brazilian Amazon. Walters et al. (2005) surveyed the practices of a wide range of landowners, from smallholders to larger industrial firms, to identify that factors that influence adoption of silvicultural practices.
99 Practices currently include single or mixed-speci es reforestation on agricultural, pasture, or secondary forest lands, enrichment planting in logging gaps, secondary forest and the reserve areas of private landholdings, and tending commerci ally valuable trees (W alters et al., 2005). The dominant reasons for adopting silviculture include guaranteeing future timber supplies, complying with regulations, and reforesting land. The principle barriers to adoption include lack of technical skills or assistance, problems with pests, and seedli ng supply difficulties. Curiously, lack of incentives is a relatively minor barrier. Because of the lack data on the dynamics of silvicultural treatm ents within this particular site, this study could not inco rporate silvicultural practices above and beyond the practices already included in RIL, such as vine-cutting. Re search on the costs and benefits of silvicultural practices that enhance regenera tion and upgrowth of valuable species, such as enrichment planting and liberation thinning, is ongoing. The financial cost-eff ectiveness of such activities, considering the length of rele vant managerial time-frames unde r reasonable discount rates, is uncertain but promising. For example, Wa dsworth and Zweede (2006) performed liberation thinning in a forest similar to the Fazenda Sete site, finding that the diameter increment of the liberated trees was 20% greater than the comparis on trees during the six post-treatment years. The extra harvest merchantable volume generated by the liberation thinning would have paid for the liberation, without consider ing the increased volumes likely to be available for future harvests (Wadsworth and Zweede, 2006). Future efforts with this model should tr y to augment the transition matrices and recruitment vectors to account for silvicultural treatments. Assuming public preferences for sustaining individual species exist, such activ ities may need to be required on public lands,
100 which may also merit public incentives. Improved analysis of the long term costs and benefits of silviculture would greatly assist in public forests planning. Role of Currently Non-Commercial Species It is possible that increasing the number of species extracted from tropical forests for commercial purposes is likely to increase the economic value of the forest, reduce impacts upon heavily harvested high-valued species, and reduc e the pressure to expand timber production into new forest lands (Barany et al., 2003; Buschbacher, 1990; Plumptre, 1996; Youngs and Hammett, 2001). The increase in species over time is likely to arise from market changes and increased wood processing efficiency improveme nts (Karsenty and Gourlet-Fleury, 2006). Critics argue that the diversification of harveste d species leads to increased rates of exploitation without diminishing pressure on high-valued sp ecies or increasing the likelihood the land will remain under forest cover (Rice et al., 2001). Ho wever, there is little empirical evidence to support either view, and the costs and benefits of valorizing new species depends in a complex fashion upon prices, regulations, and mana gement objectives (Pearce et al., 2003). As the Paragominas region was central to th e early expansion of logging in the Eastern Amazon during the 1980s, regional transportation and forest sector infrastr ucture is relatively well-developed, implying that th e number species harvested in this region is likely to be relatively high compared to most of the Brazilian Amazon. In fact, about 75% of the stems > 50 cm DBH in the Sete experimental forest ar e of commercial species. While evaluating the historical trends of timber species use is difficu lt because of ambiguities in the species names of low-valued species with few individuals, the incr ease in the number of sp ecies processed in local Paragominas mills of the study region appears negligible during the 12-year study period. Rather than the commercial species list in this region lengthening over time as the local industry
101 becomes exhausted, loggers are reentering l ogged forests to obtain smaller diameter and defective stems before land conversion. Assuming increased land security, as in perman ently secured state and national forests, to what extent would future commercialization of currently lesser known species affect economic viability of forest management? From the pr esent-oriented NPV-maximizing perspective of a concessionaire, optimistic commercialization scen arios make little difference. For example, while the detailed results are not presented here, assuming a 40-year cutting cycle managed under RIL with 20% of the non-commercial trees species becoming merchantable by the second entry, less than $2/ha is added to the NPV, although commercial harv ests clearly increase in the future. These numbers assume that prices rema in constant and that no new high-valued species will emerge in the future, an assumption that seems reasonable given th e strongly valued wood qualities of the high-valued sp ecies. The scenario also assumes that no species are decommercialized in time. Ways to Compensate Opportunity Costs of Additional Management While international transfers to protect biod iversity hotspots have been an important element in forest conservation strategies in recent decades, many commentators have expressed cautious excitement about the potential for emer ging payments for ecosystem service schemes to mobilize global resources to achieve longterm conservation and economic development (Chomitz et al., 2007; Wunder, 2006; Wunder, 2007) Recent Kyoto Protocol negotiations have looked more favorably upon forested developing c ountries to include aver ted deforestation, or compensated reduction, into the Protocol (G ullison et al., 2007). The central idea of compensated reduction is that c ountries can chose to reduce nati onal deforestation levels below historical levels and receive payments upon accomplishing targeted reductions, which would create significant incentives to reduce deforestation (San tilli et al., 2005). Under particular
102 proposals, the incentives are direct ed toward decreases in overall deforestation rates, not toward particular projects, freeing governme nts to seek the gains with the lowest marginal costs. Given that the establishment of state and national forests is a si gnificant component of the Brazilian national forest-sector strategy a nd that the carbon sequestration benefits of reduced impact logging have been demonstrated by previous studies (Boscolo et al., 1997; Boscolo and Buongiorno, 1997; Pinard and Putz, 1996; Smith and Applegate, 2004), resources could be directed at securing and improvi ng long-term management of pub lic production forests, Brazil could choose to direct in ternational payments toward grassroo ts management, silvicultural, and enforcement capacity building. Matters of Scale: Spatial and Temporal Landscape Management Considering a landscape as a summation of stand-level projects potentially yields misleading results by failing to include variables th at incorporate economic gradients across space and time (Boyland, 2006). For example, th e results presented here do not discuss the possibility of spatial and temporal zoning of timbe r harvests to meet long term objectives. With respect to sustaining populations of high-value sp ecies that, because of population structures and low recruitment rates, are difficult or perhaps impossible to manage sustainably, a reasonable conclusion of this study is that logging will need to be located within an appropriately designed and enforced reserve network. The Public Fo rests Management Law (PFML) requires that 5% of all allocated concessions be held in abso lute reserve for biodiversity preservation and monitoring, and that the relevant government agency can determine the boundaries of this reserve area. Difficult to manage species should receive consideration in this decisionmaking processes. The reserve network may also be designed temp orally. For example, a forest concession may be logged twice according to a given cutti ng cycle then put aside for productive purposes
103 for an equivalent length of time in order to regenerate timber stocks while protecting the nontimber values of the forest. Costly enforcemen t must be sustained during the period of dormant timber production. This approach would give futu re decisionmakers a policy variant of the log and leave scenario advocated by Rice et al. (2001) should land use priorities shift in the future. Public forest timber production occurs over extremely long time horizons, and adaptive management requires the building of capacity to change management objectives or prescriptions in the future. That said, decisionmakers are curr ently planning national a nd state forests in the Brazilian Amazon and need accurate projections of future timber supplies and demands (Lentini, 2007; Verssimo et al., 2006). Under current regula tions, estimates of supply should account for diminished timber volumes and values in future harvests, rather than assume optimistically that timber volumes are sustainable perpetually at maximum permitted volumes. Realistic projections of timber supply are critical in evaluating how the natural capital of the unlogged forest is converted into economic development, the key issue being whether logging in public forests, or particular areas within larger forest s, should be viewed as a one-shot event or as a repeating series of sustained harvests. Conclusion While it is again shown under the conditions of the study site that RIL is cost-effective to implement, the bulk of the financial gains are due to harvest planning improvements over conventional techniques, rather than through protec tion of the forest resource. RIL is shown to marginally increase the volumes of merchantable timber available in future harvests, but does not significantly change what is now well-known abou t logged forests; the structure and composition of the managed forests will be different than that of the primary forest. Perhaps silvicultural techniques in addition to continued developm ent of improved logging techniques will prove to
104 ameliorate this effect in the future. However, given the likelihood that certain RIL practices may be more financially beneficial than others, su ch as testing hollows or reducing wood waste and skidder-time in the forest, reveals the possibi lity that RIL may more often be partially implemented, as managers determine the relativ e costs and benefits of shirking certain costineffective activities. While volumes of merchantable timber predicte d to be available for the second and third harvests are significantly lower th an the available timber in the primary forests, the second and third harvests are projected to be profitable and ap pear to have the potential for sustainability at the stand-level and species-group level. The WSI s cenario showed that sta nd-level sustainability is likely to be possible at lower volumes a nd lower economic value, but that sustainable management by the criteria drivi ng that scenario is viable. The harvests under species grouplevel sustainability constraints we re profitable, but at the expense of harvesting very little of the most valuable timber. While this study is performed w ith the intent to provide anal ytical support to public forest planning, there is no reason that the results should apply exclus ively to public lands. The bioeconomic results apply to private lands equa lly, although the policy implications may differ across tenures, depending on over-arching manageme nt objectives for the different lands and public preferences for goods and services deliver ed from public lands. That said, perhaps the most important implication of the results presente d here, which is not original to this study (see e.g. Karsenty and Gourlet-Floury, 2006, and Van Gardingen 2006), is that the public, forest sector, decisionmakers, and other stakeholders such as non-government al organizations and certifying bodies, should reflect and revise their visions for th e likely financial and ecological outcomes of multi-cyclic management of natu ral forests in the Eastern Amazon region.
105 Table 3-1. Merchantability criteria by species group. Logging system Species group Si Qi H0i H1i s M0is M1is RIL Pioneer 0.560.910.200.051.00 0.40 0.48 Light-demanding 0.690.710.200.051.00 0.39 0.46 Intermediate 0.960.800.200.051.00 0.61 0.73 Shade-tolerant 0.420.930.200.051.00 0.32 0.38 Emergent 1.001.000.200.051.00 0.80 0.95 CL Pioneer 0.560.910.200.050.50 0.45 0.49 Light-demanding 0.690.710.200.050.50 0.44 0.48 Intermediate 0.960.800.200.050.50 0.69 0.75 Shade-tolerant 0.420.930.200.050.50 0.36 0.39 Emergent 1.001.000.200.050.50 0.90 0.98 Notes: Si = proportion of stems in species group i composed of commercial species. Qi = proportion of stems in species group i composed of stems of good form. H0i = proportion of hollow stems in species group i at initial harvest. H1i = proportion of hollow recruited stems in species group i. = proportion of hollow stems identif ied by logger under harvest system s. M0i = proportion of perceived merchantable stems for species group i under harvest system s at initial harvest. M1i = proportion of perceived merchantable stems for species group i under harvest system s at subsequent harvest.
106 Table 3-2. Commercial volume (m3/stem) by species group and DBH Species group 50-60 cm 60-70 cm 70-80 cm 80-90 cm 90-100 cm >100 cm initial harvest >100 cm subs. harvests Pioneer 3.0 4.16.08.39.611.9 11.9 Light-demanding 3.0 126.96.36.1990.221.3 13.8 Intermediate 2.9 4.46.08.19.617.0 13.8 Shade-tolerant 3.0 188.8.131.520.415.2 13.8 Emergent 3.0 184.108.40.206.819.6 13.8 Notes: Volumes estimated using volum e equations in Silva et. al (1984).
107 Table 3-3. Price/tree (price/m3 multiplied by volume/tree) by species group and size (DBH) Species group 50-60 cm 60-70 cm 70-80 cm 80-90 cm 90-100 cm >100 cm initial harvest >100 cm subs. harvests Pioneer 86.5 114.3 234.7 344.3 418.4 475.7 475.7 Light-demanding 84.5 110.8 156.4 205.4 275.4 537.5 348.2 Intermediate 79.7 113.5 153.7 224.3 261.3 490.1 397.8 Shade-tolerant 87.1 129.9 196.6 275.6 390.0 538.6 489.0 Emergent 120.3 200.2 282.1 333.9 514.4 1313.0 924.5
108 Table 3-4. Variable cost/tree (cost/m3 mu ltiplied by volume/tree) by species group and size (DBH) Harvest system Species group 50-60 cm 60-70 cm 70-80 cm 80-90 cm 90-100 cm >100 cm initial harvest >100 cm subs. harvests RIL Pioneer 46.9 64.1 93.8 129.8 150.1 186.1 186.1 Light-demanding 46.9 64.1 90.7 123.6 159.5 333.1 215.8 Intermediate 45.4 68.8 93.8 126.7 150.1 265.9 215.8 Shade-tolerant 46.9 67.3 103.2 131.4 162.7 237.7 215.8 Emergent 46.9 68.8 95.4 126.7 168.9 306.5 215.8 CL Pioneer 40.4 55.3 80.9 111.9 129.4 160.4 160.4 Light-demanding 40.4 55.3 78.2 106.5 137.5 287.1 186.0 Intermediate 39.1 59.3 80.9 109.2 129.4 229.2 186.0 Shade-tolerant 40.4 58.0 89.0 113.2 140.2 204.9 186.0 Emergent 40.4 59.3 82.2 109.2 145.6 264.2 186.0
109 Table 3-5. Components of waste across logging systems Harvest treatment Cut hollows Poor felling and bucking1 Lost trees1 wastes RIL 0.000 0.039 0.003 0.042 CL initial harvest 0.100 0.123 0.038 0.261 CL subsequent harvests 0.025 0.123 0.038 0.186 1Values found in Holmes et al. (2002).
110 Table 3-6. Solution of the MSY program at the species group-level (tree/ha and m3/ha) Cutting cycle (years) 30 40 50 60 Variable Species group (trees/ha) (m3/ha)(trees/ha)(m3/ha)(trees/ha)(m3/ha) (trees/ha)(m3/ha) Preharvest stock Pioneer 2.8 9.3 2.6 8.7 2.4 8.3 2.3 8.2 Light-dem. 1.5 5.1 1.6 5.7 1.7 6.3 1.8 6.8 Intermediate 1.4 4.8 1.6 5.6 1.8 6.6 2.0 7.5 Shade-tol. 1.3 4.3 1.5 5.1 1.7 5.8 1.8 6.4 Emergent 1.7 10.5 1.6 10.0 1.6 9.7 1.5 9.6 Total 8.7 34.0 8.9 35.1 9.2 36.8 9.5 38.6 Harvest Pioneer 2.8 9.2 2.6 8.7 2.4 8.3 2.3 8.2 Light-dem. 1.2 4.0 1.3 4.6 1.4 5.1 1.4 5.6 Intermediate 1.4 4.5 1.6 5.6 1.8 6.6 2.0 7.5 Shade-tol. 1.3 4.3 1.5 5.1 1.7 5.8 1.8 6.4 Emergent 0.0 0.7 0.1 0.7 0.1 0.8 0.1 0.9 Total 6.7 22.7 7.0 24.8 7.3 26.7 7.7 28.6
111 Table 3-7. Pre-harvest standing stock, total harvest, and rec overed timber for all cutting cycles (m3/ha) 1st entry 2nd entry 3rd entry Cutting cycle (years) Scenario Timber avail. (m3/ha) Harvest (m3/ha) Recovered (m3/ha) Timber avail. (m3/ha) Harvest (m3/ha) Recovered (m3/ha) Timber avail. (m3/ha) 30 CL U 41.0 43.432.1 220.127.116.11 12.1 RIL U 41.0 39.838.1 11.411.110.6 17.7 CL BRP 41.0 30.024.4 16.916.813.7 12.2 RIL BRP 41.0 30.028.7 18.317.917.1 17.8 RIL WSI 41.0 13.112.6 34.015.014.4 34.0 RIL SSI 41.0 14.914.3 31.519.218.3 29.0 RIL MSY* 34.0 22.721.8 34.022.721.8 34.0 40 CL U 41.0 43.432.9 10.410.38.5 18.2 RIL U 41.0 39.838.1 14.213.813.2 24.3 CL BRP 41.0 30.022.2 18.318.114.7 18.2 RIL BRP 41.0 30.028.7 20.319.819.0 24.4 RIL WSI 41.0 13.713.1 35.121.820.9 35.1 RIL SSI 41.0 15.014.3 32.922.321.4 33.6 RIL MSY* 35.1 24.823.8 35.124.823.8 35.1 50 CL U 41.0 43.432.1 13.713.310.8 22.7 RIL U 41.0 39.738.0 17.717.116.4 28.8 CL BRP 41.0 30.022.2 20.920.516.7 22.5 RIL BRP 41.0 30.028.7 23.122.521.5 28.8 RIL WSI 41.0 13.913.3 36.825.924.9 36.8 RIL SSI 41.0 19.318.4 34.316.115.4 37.8 RIL MSY* 36.8 26.725.6 36.826.725.6 36.8 60 CL U 41.0 43.332.0 17.817.113.9 26.0 RIL U 41.0 39.638.0 21.620.920.0 31.9 CL BRP 41.0 30.022.2 24.322.616.7 26.1 RIL BRP 41.0 30.028.7 26.425.624.5 31.9 RIL WSI 41.0 15.414.8 38.626.625.5 38.6 RIL SSI 41.0 19.018.2 34.316.115.4 37.8 RIL MSY* 38.6 28.627.4 38.628.627.4 38.6 MSY scenario is independen t of initial conditions
112 Table 3-8. NPV of scenarios ($/ha) Cutting cycle (years) Scenario 30 40 50 60 CL U 481.1 479.2 478.8 478.1 RIL U 606.1 603.7 602.1 600.8 CL BRP 416.7 411.9 410.3 409.6 RIL BRP 526.9 522.4 520.3 519.3 WSI 295.2 286.9 290.3 304.5 SSI 148.1 160.4 212.9 220.5 MSY* 278.9 279.4 303.4 328.7 MSY scenario is indepe ndent of initial conditions
113 0 5 10 15 20 25 0102030405060 Years after first harvestCommercial vol. recovered (m3/ha) 15 m 3/h a 35 m3/ha 30 m3/ha 25m 3 /ha 20 m 3/h a 40 m3/ha Figure 3-1. Commercial volume recovery (m3/ha) after initial RIL harvest of increasing intensity (15 to 40 m3/ha)
114 Figure 3-2. Dynamics of preharvest standing stock and total harvest by species group (40year cutting cycle). A) CL unconstrained. B) CL Brazi lian regulatory policy. C) RIL unconstrained. D) RIL Brazilian regulatory policy. E) RIL maximum sustainable yield solution F) RIL weakly sustainable inventories. G) RIL strongly sustainable inventories. A0 5 10 15 20 25 30 35 40 45Stock in Year 0 (m3/ha) Harvest in Year 0 (m3/ha) Stock in Year 40 (m3/ha) Harvest in Year 40 (m3/ha) Stock in Year 80 (m3/ha)m3/ha Pioneer Light-demanding Intermediate Shade-tolerants Emergents B0 5 10 15 20 25 30 35 40 45Stock in Year 0 (m3/ha) Harvest in Year 0 (m3/ha) Stock in Year 40 (m3/ha) Harvest in Year 40 (m3/ha) Stock in Year 80 (m3/ha)m3/ha Pioneer Light-demanding Intermediate Shade-tolerants Emergents C0 5 10 15 20 25 30 35 40 45Stock in Year 0 (m3/ha) Harvest in Year 0 (m3/ha) Stock in Year 40 (m3/ha) Harvest in Year 40 (m3/ha) Stock in Year 80 (m3/ha)m3/ha D0 5 10 15 20 25 30 35 40 45Stock in Year 0 (m3/ha) Harvest in Year 0 (m3/ha) Stock in Year 40 (m3/ha) Harvest in Year 40 (m3/ha) Stock in Year 80 (m3/ha)m3/ha F0 5 10 15 20 25 30 35 40 45Stock in Year 0 (m3/ha) Harvest in Year 0 (m3/ha) Stock in Year 40 (m3/ha) Harvest in Year 40 (m3/ha) Stock in Year 80 (m3/ha)m3/ha E0 5 10 15 20 25 30 35 40 45Stock in Year 0 (m3/ha) Harvest in Year 0 (m3/ha) Stock in Year 40 (m3/ha) Harvest in Year 40 (m3/ha) Stock in Year 80 (m3/ha)m3/ha G0 5 10 15 20 25 30 35 40 45Stock in Year 0 (m3/ha) Harvest in Year 0 (m3/ha) Stock in Year 40 (m3/ha) Harvest in Year 40 (m3/ha) Stock in Year 80 (m3/ha)m3/ha
115 Figure 3-3. Size distributions of emergent and pioneer specie s groups across scenarios (40year cutting cycle). A) Emergent speci es under CL U. B) Pioneer species under CL BRP. C) Emergent species under RIL U. D) Pioneer sp ecies under RIL BRP. E) Emergent species under RIL WSI. F) Pioneer species under RIL WSI. G) Emergent species under RIL SSI. H) Pioneer species under RIL SSI. A0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.0010-2020-3030-4040-5050-6060-7070-8080-9090100 >100Size (cm DBH)Trees/h a Year 0 Year 40 Year 80 B0 5 10 15 20 25 3010-2020-3030-4040-5050-6060-7070-8080-9090100 >100Size (cm DBH)Trees/ha Year 0 Year 40 Year 80 C0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.0010-2020-3030-4040-5050-6060-7070-8080-9090100 >100Size (cm DBH)Trees/h a Year 0 Year 40 Year 80 D0 5 10 15 20 25 3010-2020-3030-4040-5050-6060-7070-8080-9090100 >100Size (cm DBH)Trees/ha Year 0 Year 40 Year 80 E0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.0010-2020-3030-4040-5050-6060-7070-8080-9090100 >100Size (cm DBH)Trees/ha Year 0 Year 40 Year 80 F0 5 10 15 20 25 3010-2020-3030-4040-5050-6060-7070-8080-9090100 >100Size (cm DBH)Trees/ha Year 0 Year 40 Year 80 G0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.0010-2020-3030-4040-5050-6060-7070-8080-9090100 >100Size (cm DBH)Trees/ha Year 0 Year 40 Year 80 H0 5 10 15 20 25 3010-2020-3030-4040-5050-6060-7070-8080-9090100 >100Size (cm DBH)Trees/ha Year 0 Year 40 Year 80
116 CHAPTER 4 A POLICY SIMULATION OF A BRAZ ILIAN LOGGING CONCESSION UNDER IMPERFECT ENFORCEMENT AND ROYALTIES Introduction Supporters of Brazils new Public Fore sts Management Law (Lei 11284/2006) are optimistic that concessions will provide economic development opportunities and help modernize the natural forest products industry, while improving the capacity of the state to protect lands from incursions and degradation (Verssimo et al., 2002a; Verssimo and Barreto, 2004). While a logged forest may deliver lower qu antities or qualities of environmental services, it is argued that the lower ecological qualities of the forests are likely to be offset by the economic and control functions the concessions w ill perform. The overall supply of ecosystem services will be protected by the appropriate la nd being allocated toward stronger forms of protection. However, many debates during the preparati on of the Public Forests Management Law (PFML) focused on the government's low instituti onal capacity to manage the concession system and to monitor and enforce the concession contra cts and appropriate environmental management requirements. As the PFML is implemented, the development of sound administrative procedures, revenue collection systems based on appropriate rates, enforcement capacity, and royalty distribution proce dures will become crucial. Ther e is significant controversy whether these multiple tasks can be achieved. For exam ple, Merry and Amacher (2005) identify several potential problems with Brazilian concessions, many of which are analogous to well-documented problems with concessions in other countries (Gray, 2005; Repetto and Gillis, 1988). First, concessions may allow concessionair es artificially high rents because of poorly designed revenue systems. As well as give away public resource s, this problem may add itionally stifle innovation and the longer term competitiven ess of the Brazilian forest in dustry (Merry and Amacher, 2005).
117 Second, the Brazilian government may become a rent-seeking government as a result of unexpectedly high costs of ad ministering the concession system added to the possible nonpayment of royalties. In order to increases re venue, a rent-seeking government might seek to induce higher harvest rates by lowe ring tax or royalty rates away from the first-best rates that maximize economic efficiency. This effect, al so discussed for concessions in general in Amacher (1999), Amacher and Brazee (1997), and Merry and Amacher (2005), can lead to an accelerated allocation of land into conce ssions which may simply compound budgetary shortfalls. Meanwhile Leruth et al. (2001) find that royalty instruments do not reinforce or substitute for command and control enforcement approaches. The negative externali ties associated with logging have very little relations hip with harvest volumes, which is typically the target of timber tax instruments. Rather, the externalities are largely a result of the quality of management practices, such as whether RIL or post-harvest silviculture is implemented. A timber tax is unlikely to successfully function as Pigovian-type in strument which taxes the externality in order to better align social and privat e costs (Leruth et al., 2001). In fact, Leruth et al. (2001) argue, timber taxation can actually induce more damage to the forest resource than no taxes at all by encouraging lower overall harvests at higher ra tes of collateral damage. As a corrective instrument, Leruth et al. (2001) advocate a performance bond instrument combined with strengthened enforcement over royalty instrume nts. Performance bonds are also found to be effective instruments in the simulation-based st udies of Boltz (2003) and Boscolo and Vincent (2000). Objectives of the Study Within this fast-moving political-economic e nvironment, empirical efforts that examine the viability of concessions unde r imperfect enforcement can provide critical information to
118 decisionmakers. Using an optimization model parameterized with data from the Eastern Amazon, this research investigates the effectiv eness of renewability audits and performance bonds in inducing compliance with requirements regarding harvest volume limits and the implementation of RIL. The contracts emerging in larg er-scale Brazilian concessions have a maximum length of 40 years under a variable cutting cycl e that is typically 35 or more years. Hence, a typical concessionaire can be expected to plan for a si ngle harvest from each cutting block. In absence of policies that give concessi onaires equity in the forest beyond the concession contract, each entry from the concessionaires perspective can be considered a static optimization problem. This assertion is bolstered by the evidence presen ted in the previous chap ter, as the discounted value of future harvests is extremely low, rega rdless of damage incurred to the forest. The optimal harvest under current regu lations and discount rate rem oves the most valuable timber possible until constraints are bindi ng, without respect to the potential growth of any given tree. The concessionaire of this chapter will solve a harvest problem similar to the imperfect enforcement problem discussed in Amacher et al. (2007), which examined tax and fine structures when the logger was able to harvest more timber than legally allowed. The model of this chapter, however, permits the l ogger to also shirk RIL requirements as well as harvest illegal timber volumes, defined as harvested tim ber volume above the legal limit of 30 m3/ha. In order to examine the response of the concessionaire to periodic audit pressure the paper adopts the sequential harvest framework of Boscolo and Vin cent (2000), which simula tes the annual harvest of different stands with identical initial conditions over the course of a concession contract. In the model, if the concessionaire fails an audit, the concessionaire will not be able to operate on the concession for the remainder of the time horiz on of the contract. As in the Boscolo and
119 Vincent (2000) sequential harvest problem, a s econd harvest of any management unit is not permitted. The model will be used to examine the in centives for implementing reduced impact logging techniques (RIL) and observing legal harv est volume intensity regulations. In the model, the decisions to implement RIL and perform illegal harvests are seen as discrete decisions. For example, the concessionaire can fully implement RIL while logging illegal volumes. The model is developed in the form of a cl assic principal agent pr oblem. Royalty rates and enforcement pressure will be examined when the government is subject to a revenue constraint and the logger is subj ect to a participati on constraint, which requires that the logger earn a rate of return equivalent or higher than a typical rate of return to forestry on private lands. The government has at its disposal a set of harves t regulations, imperfect audit abilities, systems to generate royalty payments from the concession aire, and the possibility of charging pre-harvest performance bonds, which are returned to the concessionaire in proportion to the concessionaires compliance with regulations. Two types of royalty charges are inves tigated here. First, a percentage ad valorem rate is charged against the marginal profit, or the addedvalue, of harvesting a tree. In other words, a percentage of the difference between the price of a tree and the marginal costs to harvest the tree is charged under the ad valorem royalty. Second, a percentage revenue-based royalty rate is charged against just the price of the harvested tr ee. For either instrument, the royalty rate is applied only against trees that were legally harves ted. As has been shown in previous work, the ad valorem royalty system as typically applied is a non-distortionary instrument that will not influence marginal harvest decisions (Hyde and Se djo, 1992). However, given the analysis here
120 includes the potential of illegal logging, where no royalties are paid on the illegal portion of the harvest, the ad valorem royalty may not be non-distortionary across its feasible range under the participation constraint, as hi gh rates may induce illegal logging. Meanwhile, the revenue-based royalty charged will reduce the relative profitab ility of harvesting the tree, distorting harvest decisions on the margin, creating the possibility of i nducing lower harvest levels, which may or may not be a desirable effect, gi ven the governments objectives. The same risk of high rates inducing illegal harvests may also be pr esent with the revenue-based royalty. The area fee as used in this analysis is also non-distortionary in terms of harvest decisions, although at a larger scale the fee may help determ ine what lands are profitable for harvest. In practice, the area fee can be identified via a compet itive or administrative pr ocess. In this study, the area fee required to bind the co ncessionaires participation constraint is determined after the optimal harvest is identified. The value in calcul ating the area fee in this manner is that it can help guide decisions about what the concession aire might be willing to pay for harvest rights under a given set of royalty and enforcement me chanisms. The area fee amounts generated in this study could serve as estimates of the reservat ion price for the auction of the concession in a competitive process. Alternatively, the area fees can represent the administratively determined price for harvest rights within non-competitive environments. In the section that follows, the optimization model of the previous chapters will be modified to meet the objectives of this chapter. Results of several representative scenarios will then be presented. To conclude, these resu lts will be discussed in light of their policy implications. Further data collection and research is also suggested.
121 The Model Partial RIL Adoption To perform the analysis in this chapter, th e model requires further development to account for harvesting above the legal limit, the c hoice of logging technique, a budget-constrained government, and royalty and enforcement mechanis ms. Because the model developed here is static as opposed to the dynamic model of ear lier chapters, the subscript for time has been removed from previously-defined model components. As in Boscolo and Vincent (2000), in order to reflect the reality that the logger might imperfectly adopt RIL, the logger is ab le to choose at harvest a value, 0,1 where 0 implies no RIL practices are adopted while 1 implies the full portfolio of RIL practices are adopted. The degree to which RIL is not implemented illegally,illegal is measured by: 1illegal (4-1) This formulation implies that 0,1illegal Since the rate of RIL adoption will influence harvest damage, the damage vector from previous chapters is now written as the mn x 1 vector, d which is determined by the following linear interpolation: 121ddd (4-2) where 1d captures the damage under full implementation of RIL and 2d captures the damage under full implementation of conventional logging (CL). This formulation assumes the logger will receive the full benefit of reducing damage when RIL is fully implemented, while incremental reductions in the rate of adoption cause linear increases in the damage. When RIL is not adopted at all, the damage vector is equivalent to the
122 vector for the CL logger. Imposing the linear in terpolation on the decision is a simplification necessitated by the data and model. Within an actual logging environment, some practices are more costly than others while some are more eas ily shirked than others. Also, simple oversight or lack of training may cause firms to fail to adop t particular RIL guidelines. For example, in a study of two large certified companies in the Br azilian Amazon, Pokorny et al. (2005) note that about one-third of a set of 61 RIL guidelines were not fully implemented. The lack of sufficient monitoring, training, and equipm ent explains many of the failure s to meet guidelines (Pokorny et al., 2005). Meanwhile the returns to implementing RIL activities may have different time horizons. Given the evidence that the short-term opera tional improvements from implementing RIL can offset the costs, the firm may adopt practices that emphasize short run benefits over the long run benefit of protecting future ha rvests. For example, firms ar e unlikely to shirk developing relatively low cost but high quality inventories and maps because of the significant reduction in losses attributable to the improved ability to fi nd felled trees. Directi onal felling is a counterexample of an RIL activity that incurs costs for little immediate operational benefit while significantly reducing damage to future harvest trees. It is important to note that the static fo rmulation of the problem implies that the concessionaire will not perform a second harves t. The concessionaire, then, has no financial interest in implementing RIL to improve the futu re productivity of the forest. However, given many of the gains of implementing RIL occu r in the first harves t through planning and operational improvements, implemen ting RIL in the single harvest problem can be significantly beneficial, depending on the costs. That said by the definition of the static problem, RIL becomes relatively more costly a nd less likely to be implemented.
123 Recalling that both types of logger will reject stems with poor form, but the RIL loggers will test all trees targeted for harvest for hollowne ss and reject harvesting th e tree if it is found to be hollow, it is necessary to adjust the merc hantability matrix from earlier chapters. Let M be a mn x mn diagonal matrix where th e diagonal contains the perceived (as a function of logging system adopted) merchantable proportion of stems in each species group i and size j. Then m y is a mn x 1 vector representing the standing stock of merchantable timber, which is calculated by: m y M y (4-3) where: 121MMM (4-4) and 1M captures the perceived merchantability under full implementation of RIL and 2M captures the perceived merchantability under full implementation of CL. Where m d is a mn x 1 vector capturing the number of merchantable stems in species group i and size j killed accidentally during harvest as a function of RIL adoption, the logger will not harvest and kill by damage more than the logge r perceives is merchantable: mm hd y (4-5) The benefits of implementing RIL, in partic ular from operational planning improvements, extend beyond the reduction of damage to the residual stand and minimizing the felling of hollow trees to a general reduction of waste dur ing the forest operation (Barreto et al., 1998; Boltz et al., 2003; Holmes et al., 200 2). Again, in order to captur e the loggers choice of harvest system, a linear interpolation between the CL and RIL choice is applied: 121 wastewastewaste (4-6)
124 where 1wasteis the percentage of harv ested volume wasted under full implementation of RIL and 2wasteis the percentage of harvested volume wasted under full implementation of CL. Harvesting Above the Legal Volume Limit As in previous chapters, the regulatory agency has available a set of simple rules that it can use to try to induce desired loggi ng behavior. The regulatory agen cy may apply diameter cutting limits, harvest volume intensity limits at the overa ll stand-level, and provisions for leaving seed and rare trees standing. The regulations are as sumed to remain fixed throughout the horizon of the scenario. Whereas in the previous chapter the rules were assumed to be either perfectly enforced or perfectly un-enforced, this chapter loosens this assumption to allow the logger the possibility of breaking the harv est volume regulation. The harv est above the legal volume limit, illegalh, is measured as harvest above the permitted limit of 30m3/ha: 3'30if'30m/ha 0otherwise.illegalh h h (4-7) The standing stock of merchantable timber in the initial stand is 41.0 m3/ha, implying that 0,11illegalh. It is important to note that the harvest of tr ees and the implementation of RIL are treated as independent choice variables in this study in order to permit st ronger comparisons between the choices under different regulatory regimes. Treating the c hoices as independent, for example, permits firms to fully implement RIL while completely ignoring harvest volume regulations. The converse, fully legal harvests with no implemen tation of RIL, is also possible. In a real forest environment, the firms may often see these decisions as interconnected. For example,
125 road infrastructure may be designed and construc ted out of compliance with RIL principles in order to facilitate harvests a bove the legal volume limit. Enforcement Mechanisms Audit pressure Enforcement responsibilities on the Brazilian concessions will be divided between the government and independent auditors. The law re quires that the new fore st agency, the Servio Florestal Brasileiro (SFB), oversee the concession contracts, while the Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renovve is (IBAMA) and state environmental agencies will maintain their traditional ro les in regulating management pl ans and operations on national and state lands, respectively. Meanwhile, under a certification-like model, the law requires that concessions are independently audited at least once every three years. While on paper this system establishes checks and balances and several redundancies, the current enforcement capacity within SFB and IBAMA and within the judiciary more generally is low (Brito and Barreto, 2004; Brito and Barreto 2006; Hirakuri, 2003; Schulze et al., 2007b), and the promises of high enforcement expressed in the PFML might not be enacted in the short term. In a study of a sample of forest environmental crimes in the state of Par between 1999 and 2002, Brito and Barreto (2006) find th at a large majority of citations (72%) are issued for storage and transport violations, while relatively few are i ssued for violations in the forest (6%). Of the citations issued in the sample, very few of the accused ultimately paid a fine. Brito and Barreto (2006) find that over 60% of accused individuals or firms are never served process, meaning the alleged violators were not found in order to init iate legal proceedings. Negotiated settlements requiring payments for social services or enviro nmental reparations are a large component of the environmental judicial process in Brazil. Settlements offered to individuals cited for environmental crime in Par ranged from $5.50 $54.10/m3 roundwood with the offers to firms
126 ranging from $7.20 $217.20/m3, although the actual settled valu es varied greatly from these ranges, from $0.77 $480.50/m3 (Brito et al., 2005). Ultimately, fewer than half of the fines levies were paid (Brito a nd Barreto, 2006; Hirakuri, 2003). To reflect the relative weakness of enforcemen t in the model, the probability of being caught and punished for breaking a rule is considered an increasi ng function in the level of rulebreaking behavior, operationalized in this study by the weighted sum of illegal harvest volume and the degree to which RIL is not implemented, a variable which will be defined later in the text. While there are no empirically-based es timates of the penalty function for concession logging in Brazil, it is reasonable to assume that similar to the penalty function presented in Sun (1997), the function would take a logistic form. Where x represents an illegality factor and x represents the probabili ty of being caught and ultimately paying a fine as a function of the illegality factor, the logistic form of the penalty function follows: 1 1 x x e (4-8) Recalling that 0,1illegal and 0,11illegalh it is necessary to in troduce a variable that weighs the relative contribution of each activity to the penalty function. In this case, failure to adopt RIL and illegally harvested volume are as sumed to have equal weight in the penalty function. In order to we ight these factors equally, illegal is multiplied against where 11 This allows illegal to have the equivalent range as illegalh. Now, the illegality variable, x, can be fully defined: illegalillegalxh (4-9)
127 where the constant 11 The constant serves the purpose of re-centering the distribution of x such that if illegalh and illegal are equal to zero, then 0 x Also, when illegalh and illegal are at their maximum values, 1 x. The probability of being caught breaking the rules and paying a fine as a function of x is graphed in Figure 4-1. The probability of capture increases slowly within the lower range of x as the concessionaire harvests s lightly over the legal limit and/or shir ks few RIL-required activities. As the over-harvests or shirking of RIL-required ac tivities increases into the middle range, the likelihood of capture increases rapidly, quickly peaking as harvest damage becomes more obvious to the enforcement agent or auditor. The shape of the en forcement function is consistent with the observation that low levels of rule-br eaking behavior may not be easily observable in complex and remote forest environments, while increasing harvest intensity or poor logging technique is more readily observa ble, particularly by the real-t ime forest management remote sensing techniques expected to form an importa nt component of enforcement capacity in the concession system (Monteiro and So uza Jr., 2006; Souza Jr. et al., 2006 ). It is important to note that the probabilities here reflect the likeli hood that the rule-breaking concessionaire will be successfully prosecuted throughout the enforcemen t chain and pay a fine. Hence, the low probabilities of capture at lower levels of rule-b reaking behavior is cons istent with empirical observations with respect to the low likelihood of a rule-b reaker actually payi ng a fine (Brito and Barreto, 2004; Brito and Barreto, 200 6). So, while it is unfortunate that an empirical estimated enforcement model is unavailable, the model presented here is a reasonable approximation. Performance bonds The use of performance bond mechanisms (also referred to as a forest guarantee bond) is often proposed as a royalty instrument and has be en the subject of several studies (Boltz, 2003;
128 Boscolo and Vincent, 2000; Leruth et al., 2001; Paris et al., 1991; Sun, 1997), yet there is very little practical experience with performance bonds in tropical count ry concessions. In this study, concessionaires deposit a paymen t, denoted by the variable bond in this chapter, with the government before harvest. Upon execution of the harvest, concessi onaires are refunded a quantity proportional to the sa tisfaction of required performance measurements. When represents the proportion of the rules that have been followe d, as measured by the illegality factor, x, the concessionaire will be reimbursedbond If rules have been found to be broken, the government will keep the residual to compensate for losses. Economic Variables As in the previous chapter, a constant real discount rate, r, of 9.75% was applied throughout the analysis with in the discount function, 1 1r For this analysis, the unit values of prices and costs are assumed to be fixed in real terms throughout the time horizon and are expressed throughout the study in US$2004. The price of a tree, ij p from species group i and size j is presented in Table 3-3. These values were calculated based upon the mill gate Free On Board prices in the Paragominas region, accord ing to economic surveys of mill owners and operators (Lentini et al., 2005). Management costs for Free On Board forest mill are classified as either variable or fixed costs and are drawn from Barreto et al. (1998) and Lentini et al. (2005) Variable costs of harvesting a tree in species group i and size j under harvest system s is represented by ijsvc and includes the costs associated with felling, skidding and log de ck operations, and transportation costs incurred relative to the level of ha rvest intensity and harvest system choice s. The costs are estimated based upon $15.64/m3 for the RIL logger and $13.48/m3 for the CL logger, and the
129 per-tree costs are shown in Table 3-4. Fixed costs in $/ha, represented by s f c, are incurred regardless of harvest intensity at each cutting cycl e entry for planning and capital costs as well as transaction costs and are estimat ed at $50.56/ha for the RIL logger and $13.91/ha for the CL logger. Because the terrain at th e experimental forest of this st udy is relatively flat and did not include any slopes or stream buffer areas that firms are required to set aside as reserves, there is a risk that benefits of RIL will be over-stated b ecause the estimated costs do not account for these factors (Putz et al., 2000). To reflect this possibility, addi tional scenarios are simulated reflecting a 50% increase in variable and fixe d costs of implementing RIL. Additionally, the costs must be adjusted to account for partia l RIL implementation using the interpolation technique: 121,ijijijvcvcvcij (4-10) 121 f cfcfc (4-11) where costs indexed by 1 indicate costs of RIL operations and co sts indexed by 2 indicate costs of CL operations. In addition to fixed harvest and management plan costs, concessionaires will be expected to pay for the required independent audits. Th ese audits will be performed by the Brazilian Program of Forest Certificati on (CERFLOR) under a set of techni cal standards developed by the Brazilian National Institute of Meteorology, Standa rdization, and Industrial Quality. While this system is endorsed by the Programme for the Endor sement of Forest Certification, Brazilian firms have significantly more experience under the Forest Stewardship Council certification system, so estimates of the costs concessionaires will bear for certification are drawn from FSCcertified operations which were certified by the Brazilian certifying organi zation, Instituto de
130 Manejo e Certificao Florestal e Agrco la (IMAFLORA). Certification costs, cc, for a large concession with more than 100,000 ha include $0.46/ha upfront costs for the required preliminary evaluation and certi fying evaluation (Mauricio de Almeida Voivodic, IMAFLORA, personal communication). These costs are charged at t = 0, then recur every five years. During the years between the larger-scale evaluations, the firm must pay $0.30/ha for annual audits (Mauricio de Almeida Voivodic, IMAFLORA, personal communication). The net present value (NPV) of the certification costs ove r a 40-year concession agreement is estimated as $2.10/ha. In this model, the concessionaire will face a participation constraint in which the concessionaire will only bid a c oncession if the expect ed return is greater than or equal to alternative possibilities. It is assumed that the concessionaire s alternative opportunity is to perform forest exploitation on priv ate lands. The average gross prof itability of logging in private lands in the state of Par has b een estimated to range from 10 to 26% (Verssimo et al., 2002b). In this analysis, a minimum acceptable gross profit level of 20% is assumed. When E represents expected profit and represents the minimum acceptable gross profit, the participation constraint is expressed by: *E (4-12) While this constraint will not affect the profit-maximizing c oncessionaires marginal harvest decisions, identifying under what policies and royalty rates the constraint binds will help identify feasible policies. Government Payments and Costs Parallel to the concessionaire facing a participation constraint, the government will face a revenue constraint, in which expected reve nues must meet the costs in establishing,
131 administering, and enforcing the concessi on system. Expected government revenue,ER, is the sum of royalty payments, area fees, fines, and unreturned performance bonds. The fine is denominated by f where $700/ha f to reflect the maximum fine for deforestation and illegal logging under the Brazilian Enviro nmental Crimes Law (Lei 9605/98). The area fee is denoted by When represents the percentage charge against marginal profit per legally ha rvested tree, the government revenue collected under the ad valorem royalty system is calculated by: 1ijijijij ij E Rphvchfxbond (4-13) When represents the percentage charge agains t the price of a tree, government revenue collected under the revenue-based royalty rate is calculated by: 1ijij ij E Rphfxbond (4-14) The governments establishment costs may in clude costs associate with the government performing forest inventories, preliminary studies, granting lice nses, and conducting the proposal and auction processes. These types of costs w ill be location-specific and are likely to vary greatly. However, sufficient information ex ists to generate an approximation of the establishment costs. Inventories for privat e timber firms in the Brazilian Amazon cost $13$17/ha (Sabogal et al., 2006). Making rough extrapolations for the other likely establishment costs components leads to an estimate of $15-40/ha. In order to be conservative, the high end of this range ($40/ha) was chosen for this study. As is the case with establishment costs, very few data exist concerning the costs of concession administration a nd forest law enforcement in Brazil. For lack of an alternative figure,
132 this study assumes that administrative costs from the governments perspec tive are equivalent to certification costs borne by the concessionaire, or $2.10/ha. Th e costs associated with enforcing public forests management plans is assumed to be similar to the costs associated with enforcement in private lands. After modifying calculations in Hirakuri (2003), IBAMAs annual costs for post-harvest inspections and pre-harvest permitting and inspection for the subsequent harvest unit estimated at $2.80/ha fo r a relatively large harv est unit of 1500 ha. Extrapolations based upon annua l enforcement budgets reported in IBAMA technical reports (IBAMA, 2002) reveal similar figures. The sum of establishment a nd enforcement costs forms the government budget constraint, *$44.90/ha R In the simulations where the constrai nt is applied, the sum of expected royalties, fines, and unreturned performance bonds must exceed* R : *ERR (4-15) In reality, the government costs might predomin antly be incurred during earlier years of the concessions time horizon. In order to simplif y the accounting within th e model, however, the revenue constraint is ap plied each year equally. Objective Functions under Imperfect Enforcement and Performance Bonds Whenijh equals the entire harvest and ijh equals the legal harvest from each species group i and size j respectively, the expected profit, E from a single harvest of a representative hectare stand under the ad valorem royalty is written as: 01ijijijijijijij ijEwastephphvchvch f cccfxbondbond (4-16) Expected profit under the revenue-b ased royalty is written as:
133 01ijijijijij ijEwastephphvch f cccfxbondbond (4-17) As in Boscolo and Vincent (2000), the seque ntial harvest problem assumes that the concessions cutting blocks are homogenous in space and economic factors are homogenous in time, implying that the concessiona ire makes identical choices in each cutting block at each point in time. The expected net presen t value (NPV) over the time horizon, T of the concession is equal to: 1 1 0()...T T t tENPVEEE E (4-18) Because each cutting block is standardized to a single hectare, this formulation implies one representative hectare is logged each year The size of the concession, then, is T hectares. As discussed earlier, concessions are expected to undergo peri odic independent audits (at least once every three years accord ing to the PFML) as well as annual post-harvest inspections by IBAMA or the relevant state-level agency. In this study, if the concessionaire is caught violating the rules during an independent audit, the conces sionaire will risk losing the concession. If the concessionaire fails an audit, the concessionaire will not be able to operate on the concession for the remainder of the time horizon of the contract. If the concessionaire is caught during an annual IBAMA in spection, the concessionaire simply risks paying a fine. Periodic audits and annual harvest inspections ar e assumed to share the same penalty function. When enforcement pressure of this nature is introduced, it is necessary to consider the cumulative effects of the periodic audits. By assumption, the audits are performed independently, and the concessionaire does not know when the auditors will appear but has an
134 expectation of the number of visits they will receive. Let Q equal the expected number of randomly-timed independent audits. Upon the first audit, given a harvest decision that the leads to the level of illegal behavior, x the probability the concessionaire is caught-breaking the rules and losing the concession is given by x Because being caught breaking the rules during the second audit is dependent upon not getting caught during the first audit, the probability the concessionaire being caught breaking the rules during th e second audit is given by 1 x x The probability of being caught in the third audit is contin gent upon not getting caught during the first two audits, 21 x x and so on. Letting q index the individual audits, the probability of being caught breaking the rules across the time horizon of the contract given Q audits assumes the following negative binomial recursive structure: 1 11Q q qxx (4-19) Fully stated, the concessionaires private sequential harvest pr oblem under imperfect enforcement can be written as a function of not getting caught dur ing a periodic audit: 1 1 10max()11Q T q t qtENPVxxE h (4-20) subject to: 4 10ij jh (4-21) mm hd y (4-22) 0 h (4-23) actual yy (4-24)
135 E (4-25) E RR (4-26) where Equation 4-21 is a merchant able size constraint, Equation 422 is a merchantable harvest constraint, Equation 4-23 is a non-negativity constraint, Equation 4-24 defines the initial condition of the stand, based upon the averag e condition of the study site, Equation 4-25 represents the concessionaires participation constraint, and Equation 4-26 represents the governments budget constraint. Under performance bonds, the concessionaire seeks to maximize the expected NPV but without the term accounting for periodic audit pressure: 1 0max()T t tENPVE h (4-27) The concessionaire under performance bonds also faces the same set of constraints, Equations 421 through 4-26. Results and Discussion Renewability Audits and Annual Harvest Inspections Figure 4-2 shows how concessionaire behavior shifts as the number of audits per concession time horizon increases. In the resu lts depicted in this figure, annual harvest inspections are performed under the threat of a $700/ha fine. Additionally, a 20% ad valorem rate is charged against legally harvested trees in order to simulate the general case of valueadded taxation in Brazil. As expected, for the given audit function, incr easing the frequency of audits in addition to the perfor ming annual harvest inspections d ecreases illegal activities, but high audit frequencies do not induce full compliance with the law. As audit pressure increases,
136 illegal behavior decreases but gr adually decreases into the regi on of the penalty function where the likelihood of being caught a nd punish is very low. Concessionaire behavior with re spect to audit pressure varies greatly when different RIL costs are considered. When the costs estimat ed from the study site are used within the simulations, the logger will fully implement RIL. This result is consistent with the findings earlier in this study and in Barre to et al. (1998) and Holmes et al. (2002). Meanwhile, the logger will illegally harvest timber, regardless of aud it pressure. Increasing audit pressure has the strongest deterrent eff ect when the number of audits acr oss the concessions time horizon increases from zero to one, dropping the illegal harvest from about 10 m3/ha to about 6 m3/ha. Meanwhile, increasing the number of audits abov e one has a more gradual effect on reducing illegal harvest. The PFML, for example, requi res at least one independent audit every three years, which translates to about 14 audits over a typical 40-year concession contract. Given the parameters of this simulation, the logger under the legally required pressure would harvest 34.5 m3/ha, which is 4.5 m3/ha over the legal limit. When the variable and fixed costs of RIL ar e increased by 50% in order to reflect the potential situation where RIL is more expensive than at the study site, the logger exhibits a very different behavior. At zero audi t levels under the threat of the fi ne that arises from the annual harvest inspection, the logger chooses to harvest at the le gal limit while shirking all RIL obligations. As audits increase, RIL adoption incr eases but full adoption is unlikely. Under the legally mandated 14 audits, RIL adoption would be just over 50% when RIL costs are high. At all audit frequencies, harvest volumes are within the legal limit. In either case, the logger chooses to obey one set of rules, while breaking the rule that is most advantageous in that the ru le-breaking provides the most benefit at the least risk of getting
137 caught and losing the concession. While the re sults presented here are parameter-dependent simulations, the fact that concessi onaire behavior varies dramatic ally as a function of RIL costs is indicative of policy complexity. Varying the costs of operating within the same forest can have very different outcomes. While the scen arios modeled show the concessionaires shirking ignoring the harvest limit or shirking the RIL requirement exclusively, it is possible under different costs that both rule s are broken simultaneously. Performance Bonds While increasing audit pressure incrementally reduces illegal activities, increasing the amount of performance bonds has threshold effects; the use of bonds has no effect until a critical value is achieved that causes rapid declines in ill egal behavior. Figure 43 depicts the effects of increasing performance bonds on illegal harvest a nd RIL adoption. In this case, the annual harvest inspections and periodic audits are not in place in order to better compare the instruments. Absent inspections, audits, or a performance bond, the logger with RIL costs from the study site harvests all merchantable stock, while implementing RIL, the same result found in Chapter 3. As performance bonds increase from about $200/ha to $250/ha, the illegal harvest is eliminated. Bond amounts above $250/ ha are, in this case, redundant. When RIL costs are high and there is no enforcement or performance bond, the logger harvests the entire merchantable stock and fu lly implements conventi onal logging techniques. Bonds of about $150 to $190/ha ar e effective in inducing legal harvest volumes. Meanwhile, bonds up to about $230/ha are require d to induce full RIL implementatio n. As in the case of the lower RIL costs, bonds a bove $240/ha are redundant. Royalty Instruments While the simulations were performed for the 40-year sequential harv est problem in order to evaluate the response to incr easing periodic enforcem ent pressure, the general results will be
138 presented based upon the per-hectare values as sociated with the optimal solution of the sequential harvest problem. The ad valorem and revenue-based roya lty charges will be compared under imperfect enforcement and the pe rformance bond scenarios. For the imperfect enforcement scenario, the total number of audits is set at 14 and the fine at $700/ha to approximate the general Brazilian enforcement and taxation context. The performance bond is set at $250/ha, which is the amount shown in Figur e 4-3 to induce full comp liance to harvest and RIL regulations. Ad valorem royalty under audit pressure When the actual RIL costs are used under audi t pressure, the government needs to apply a 5% ad valorem rate in order to satisfy the budget c onstraint, while a maximum rate of 66% feasibly satisfies the concessionaires participati on constraint (Table 4-1). As discussed earlier, the ad valorem royalty is typically non-distortionary. But, in the presence of illegal logging, increasing the rate induces slight ly higher levels of illegal loggi ng, as the incentive to log lowvalue timber illegally increases slightly as pr ofits are reduced through th e royalty policy. The effect is relatively mild, as illegal harvests increase 1.6 m3/ha as the rate increases from its minimum to its maximum. Increasing the ad valorem rate additionally has no effect on the adoption of RIL; adoption is cost-effective across the feasible range of the rate. The per-hectare profit decreases from $518/ha to $200/ha as the rate increases from 5 to 66% (Table 4-1). Revenues from illegal logging in creases as illegal harvests increase, and this revenue at the high-end of the ad valorem rate is used to satisfy the participation constraint. This dynamic raises the issue that illegal l ogging is not simply induced by higher ad valorem rates, but is required, as the heavy r oyalty burden provides incentives for the logger to seek royaltyfree income. Absent this ability, the logge r will not participat e in the concession.
139 Similar to the results shown in Figure 4-2, wh en variable and fixed RIL costs are increased 50%, the form of illegal behavior shifts from illegal harvest to shirking RIL requirements (Table 4-2). Increasing ad valorem rates slightly increases RIL adopti on, a critical difference from the actual RIL cost case when increasing the rate i nduces higher illegal harvests. This effect reaffirms the earlier highlight that the royalty in struments might induce di fferent behaviors from firms with different costs, even if the fore st resource itself is the same across firms. When RIL costs are high and the ad valorem rate nears the maximum feasible of 40% under the participation constraint, the concessionaire reduces harv ests by over 50% (Table 4-2). Under the participation constraint and heavy royalty bu rden, the high cost logger is forced to harvest from only the most valuable species groups, as lower value species are no longer costeffective to harvest, even illegally. Absent th e participation constraint the logger would likely maintain the harvest at the lega l limit throughout the range of the ad valorem instrument. The trade-off between increasing ad valorem rates and decreasing pot ential area fees is clear. At the high end of the feasible ad valorem rates, the potential to collect area fees reduces to zero (Tables 4-1 and 4-2). The optimal choi ce between applying a ro yalty-based strategy and a strategy based upon an area fee de termined through a competitive or administrative process is crucial and is likely dependent upon the qualities and quantities of the forest resource, the characteristics of the logging firms, and regiona l institutional strength. This study of course assumes the harvest royalties appl ied to the legal harvest and th e potential area fee are actually collected, which is another important factor to consider. Revenue-based royalty under audit pressure When actual RIL costs are used under audit pr essure and increasing revenue-based royalty rates, illegal harvest rates rise slowly, as in the ad valorem (Table 4-3). The royalty instruments theoretical ability to reduce harves t rates as the rate increases is shown to be eliminated when the
140 concessionaire who faces actual RIL costs can hi de illegally harvested volumes. Meanwhile, when the concessionaire faces high RIL costs, the increasing royalty rate decreases harvest volumes as expected (Table 4-4). The results sh ow the ability for the revenue-based instrument to influence marginal harvest rates is highly de pendent upon RIL costs, a re sult again pointing at the contingent effectivene ss of the royalty instrument s to influence behavior. The feasible range for the reve nue-based instrument under aud it pressure is smaller than the range for the ad valorem instrument. Under actual RIL costs, the feasible range is from 5 to 36%, while under high RIL costs, th e feasible range is 5 to 26%. As in the ad valorem instrument under actual RIL costs, RIL is fully adopted under the revenue-based instrument when the concessionaire faces actual costs (Table 4-3). However, under high RIL costs, RIL adoption d ecreases slightly as the revenue -based royalty rate rises. Ad valorem royalty under performance bonds The behavior of the concessiona ire under performance bonds and ad valorem royalties is very different than the behavior of the conces sionaire under audit pre ssure. While Figure 4-3 showed the effectiveness of high performance bonds in inducing regulatory compliance, a background ad valorem rate of 20% was applied to reflect the general case of forest taxation in Brazil. In Table 4-5, it is evident that increasing the ad valorem rate above 20% has no effect on harvest behavior until the rate binds the participation constraint. At that point, as the concessionaire loses income to royalties, it beco mes increasingly attractive to log illegally and shirk royalties. Meanwhile, RIL adoption for the logger facing actual RIL costs remains full across the feasible range of the ad valorem instrument, 8 to 54%. Meanwhile, when RIL costs are high, the pe rformance bond is suffici ent to induce legal harvests throughout full adopti on the feasible range of the ad valorem instrument, 8 to 30% (Table 4-6). However, while rates of up to 20% induce harvests at the legal limit, RIL
141 implementation is abou t 90%. Increasing the ad valorem rates beyond 20% reduces harvests to about 24 m3/ha, while inducing full implementation of RIL. Again, the effectiveness of any given instrument is dependent upon costs and the rate at which the instrument is set. Revenue-based royalty under performance bonds The combined use of performance bonds a nd the revenue-based instrument appears relatively robust. Under actual RIL costs, rates across the feasible range of 5 to 35% induce full RIL adoption. However, as the rate is incr eased to the point of binding the participation constraint, harvest falls below the legal minimum, again highlighting the sensitivity of logging to the rate of the royalty instrument. When RIL costs are high, the revenue-based royalty has a relatively small feasible range, from 9 to 19% (T able 4-8). Harvest volumes are low under this instrument and high costs. At the low to middl e-range of the rate, harvest volumes stand at 23 m3/ha. At very high royalty rates, harvests fall to 12 m3/ha. Issues with differentiated royalties In principle, a royalty system such as the revenue-based scheme discussed can be designed such that the rate charged is differentiated by specie s, size, quality, or site This much discussed advantage changes the relative price of each tr ee, which can modify harvest behavior at the margin. In practice in tropical forests, however, because of the large number of timber species and administrative inadequacies and corruption, these royalt ies have proven difficult to administer and collect. An added complexity little discussed in the lit erature is the fact that timber prices and harvest costs are very heterogeneous. The rela tive profit margin for a ny given tree species is likely to differ across regions or even firms. A set of royalty rates which is well-calibrated to induce a desirable harvest distri bution from a firm with a par ticular cost function may not
142 function well if applied to anothe r firm operating in the same forest or the same firm operating in another region. For example, one possible desirable and relativel y simple harvest distri bution might be that all harvested trees are drawn fr om each species group in equal pr oportion, rather than extracting trees in descending order of thei r value until relevant constraint s bind. Experimentation with a range of revenue-based royalty rates under imperfect enforcemen t and performance bonds with actual and high RIL costs showed th at it is unlikely to find a set of differentiated royalty rates that could induce this desired distribution of the harvest across species groups while satisfying the revenue and participation constraints. Performance Bonds and Firm Size In an effort to improve the distribution of forest concessions across firm size and prevent domination by larger-scale industri al forest products firms, the PFML imposes requirements to allocate forest concessions across small (< 10,000 ha), me dium (10-40,000 ha), and large concessions (40-200,000 ha). It can be expected that firms along the cont inuum of firm size will have differing capacities to pay performance bonds. For example, the la rger, well-capitalized timber firms operating in the Brazilian Amazon, wh ich are typically certif ied and export a large proportion of their production, are likely to have the upfront capacity to pay performance bonds. Meanwhile, smaller firm and community forestry operations that are expected to bid for the smaller concessions are likely to perform relatively low intensity and impact harvests, often in collaboration with a nongovernment al organization. These operations are likely to pursue socioenvironmental goals rather than maximize harvests. As such, the smaller operation may not need to provide strong environmental performance assurances. Meanwhile, mid-sized firms on the economic marg in supply a large quantity of the timber in the Brazilian Amazon (Merry and Amacher 2005). In addition to competing with timber from
143 large firms operating on private lands and forest concessions, these firms compete intensely against the large supply of lowcost timber from legal deforestation headed for the domestic market (Merry and Amacher 2005). For this gro up, requiring the payment of performance bonds and the performance of what mi ght be relatively high cost management activities may create a crippling entry barrier to participation in the concession system. While the performance bonds may be an effective instrument to guarantee the performance of these firms, the Brazilian government should consider a phased or firmby-firm bond system in order to maximize the equitable access to concessions across firm size. A Note on Market-based Enforcement Efforts This study focuses on a small set of statecentered mechanisms to induce regulatory compliance to the exclusion of non-state mechan isms such as forest certification. The enforcement within this study al so emphasizes the states role in monitoring operations within the forest, rather than at mills, along roads, or at the ultimate destination of the timber, predominantly large domestic and international cities. Driven by the increasing leverage of consumer interest groups, nongove rnmental environmental organi zations, and improvements in monitoring technology such as publicly accessible remote sensing imagery, a broader portfolio of enforcement measures is emerging in the Brazilian Amazon and, indeed, most forested countries of the world (Kramme and Price, 2005). A state seeking to optimize the allocation of enforcement resources will consider a range of instruments including partnerships with nongovernmental sectors. Future re search in this area should incorporate th ese potentially more effective mechanisms. Conclusion In this chapter, a sequential harvest model was developed to evaluate the response of a representative concessionaire to royalty instruments unde r imperfect enforcement and
144 performance bonds when there ar e incentives to harvest timber volumes over the legal limit and shirk RIL requirements. The model was develope d in the form of a classic principal agent problem in which the government is revenue -constrained and the concessionaire faces a participation constraint. The results show that each instrument has strengths and weaknesses that are contingent upon the costs fi rms face to implement RIL. In a weak enforcement environment, the result s show that audit pressure is unlikely to induce full compliance with harvest regulations. Under clear incentives for harvesting over the legal limit, it is possible to maintain low-levels of illegal behavior hidden from regulators. It is critical to develop human technical capacity to regi ster and punish low-level illegal activities. In Brazil, the combined pressure of independent audits and remote se nsing of harvest-related forest damage should help ameliorate this problem, a ssuming the performance of the judicial system improves along with audit capacity. As found in Boltz (2003), Boscolo and Vincent (2000), Leruth et al. ( 2001), and Paris et al. (1991), the effectiveness of perfor mance bonds is attractive. In an environment with corruption and an inefficient or ineffective legal envi ronment, the payment sequence affords a strong information advantage. In the enforcement case, a citation may take years to process and fines may never be paid. In the bond cas e, the preventive fine is paid at the outset. The relative efficiency of the bond instrument over traditional enforcement will arise from the efficiency of the government to evaluate logger performance and return the appropr iate proportion of the bond. In Brazil, where loggers trust in governme nt officials is low (Merry and Amacher, 2005), the logger may be reluctant to contract unde r performance bonds fearing that the same inefficiencies and corruption might pervade the syst em as in the traditional enforcement system.
145 With respect to the pote ntial advantage of performance bonds, as with all things in political economy, the development of cr edible institutions matter. The use of traditional royalty instruments such as the ad valorem and revenue-based royalty can effectively generate revenues and, in the case of revenue-based instruments, modify harvest behavior, but only under ve ry limited circumstances. When simple but critical variables change, such as the costs to implement RIL, the outcomes under the same instrument can vary dramatically. In this model, the concessionaire is able to shift to harvest volumes over the legal limit when the royalty rate becomes too high. Meanwhile, if enforcement or bonds are eff ectively inducing overa ll compliance with rules, the concessionaires participation constraint can hinder the effectiveness of the instruments to modify harvest behavior in ways that mi ght better protect future resource and reduce externalities. Recalling the harvests determined under the sustainability constraints in Chapter 3, the royalty instruments applied here are unlikely to induce the harvests required for sustaining timber inventories at the stand a nd species-group level. In terms of capturing forest rent, these results appear to favor market-determined paym ents for harvest rights combined with more effective enforcement to induce desired harvest volumes. As the Brazilian experience with forest conc essions is only beginning, there are ample opportunities to extend this research as data from new concessions accumulates. A primary need is to develop an empirically-based penalty func tion that more effectively models enforcement success as a function of enforcement effort a nd concessionaire behavior. Also, as more concessions are implemented, it is important to continuously collect quality data on government costs. A recurrent problem identified in the concession economics literature is that the government can become rent-seeking, attempti ng to increase government receipts by lowering
146 royalty rates and expanding the land base under logging concessions. More empirical evidence is required to evaluate and mitigate this risk. Finally, inasmuch as this model can be used to provide guidance for revenue and regulatory deci sions, it is important to calibrate the model using a range of forest sites under different economic conditions.
147 Table 4-1. Results under imperfect enfor cement (audits = 14 and fine = $700/ha) and ad valorem royalties (actual RIL costs) Ad valorem rate (%) Harvest (m3/ha) RIL adoption (%) Profit ($/ha) Illegal revenues ($/ha) Royalties ($/ha) Excess profits ($/ha) 5* 34.2 100.0518.4143.153.1 390.1 10 34.3 100.0512.7145.059.2 383.0 20 34.5 100.0455.4151.3118.0 313.2 30 34.7 100.0398.4157.8176.5 243.9 40 34.9 100.0341.7164.5234.6 174.9 50 35.1 100.0285.5171.6292.3 106.4 60 35.3 100.0229.8179.3349.6 38.4 66** 35.8 100.0199.6193.0382.2 0.0 Notes: Minimum rate feasible under revenue constraint. **Maximum rate feasible under participation constraint.
148 Table 4-2. Results under imperfect enfor cement (audits = 14 and fine = $700/ha) and ad valorem royalties (high RIL costs) Ad valorem rate (%) Harvest (m3/ha) RIL adoption (%) Profit ($/ha) Illegal revenues ($/ha) Royalties ($/ha) Excess profits ($/ha) 9* 30.0 56.5276.50.046.8 142.0 10 30.0 56.5271.30.052.0 135.7 20 30.0 56.5219.30.0103.9 73.4 30 30.0 56.5167.40.0155.9 11.0 40** 14.4 58.798.00.0158.1 0.0 Notes: Minimum rate feasible under revenue constraint. **Maximum rate feasible under participation constraint.
149 Table 4-3. Results under imperfect enforcemen t (audits = 14 and fine = $700/ha) and revenuebased royalties (actual RIL costs) Revenuebased rate (%) Harvest (m3/ha) RIL adoption (%) Profit ($/ha) Illegal revenues ($/ha) Royalties ($/ha) Excess profits ($/ha) 5* 34.2 100.0518.4143.153.1 390.1 10 34.4 100.0466.5147.8106.0 327.2 20 34.7 100.0363.4157.8211.5 201.9 30 35.0 100.0261.1169.5316.3 77.4 36** 35.3 100.0200.4177.8378.8 3.2 Notes: Minimum rate feasible under revenue constraint. **Maximum rate feasible under participation constraint.
150 Table 4-4. Results under imperfect enforcemen t (audits = 14 and fine = $700/ha) and revenuebased royalties (high RIL costs) Revenuebased rate (%) Harvest (m3/ha) RIL adoption (%) Profit ($/ha) Illegal revenues ($/ha) Royalties ($/ha) Excess profits ($/ha) 5* 30.0 53.8272.20.054.7 137.9 10 30.0 53.8217.50.0109.3 72.3 20 19.7 51.6127.20.0165.4 10.7 26** 11.0 44.978.30.0144.6 0.0 Notes: Minimum rate feasible under revenue constraint. **Maximum rate feasible under participation constraint.
151 Table 4-5. Results under performance bonds ( bond = $250/ha) and ad valorem royalties (actual RIL costs) Ad valorem rate (%) Harvest (m3/ha) RIL adoption (%) Profit ($/ha) Illegal revenues ($/ha) Royalties ($/ha) Excess profits ($/ha) 8* 30.0 100.0473.60.049.7 359.3 10 30.0 100.0461.20.062.2 344.4 20 30.0 100.0399.00.0124.3 269.8 30 30.0 100.0336.90.0186.5 195.2 40 30.0 100.0274.70.0248.7 120.6 50 30.0 100.0212.60.0310.8 46.0 54* 36.5 100.0202.70.0315.1 0.0 Notes: Minimum rate feasible under revenue constraint. **Maximum rate feasible under participation constraint.
152 Table 4-6. Results under performance bonds ( bond = $250/ha) and ad valorem royalties (high RIL costs) Ad valorem rate (%) Harvest (m3/ha) RIL adoption (%) Profit ($/ha) Illegal revenues ($/ha) Royalties ($/ha) Excess profits ($/ha) 8* 30.0 89.8232.30.033.5 74.0 10 30.0 89.8223.90.041.9 64.0 20 30.0 89.8181.90.083.9 13.7 30** 23.7 100.0149.70.0114.4 0.0 Notes: Minimum rate feasible under revenue constraint. **Maximum rate feasible under participation constraint.
153 Table 4-7. Results under performance bonds ( bond = $250/ha) and revenue-based royalties (actual RIL costs) Revenuebased rate (%) Harvest (m3/ha) RIL adoption (%) Profit ($/ha) Illegal revenues ($/ha) Royalties ($/ha) Excess profits ($/ha) 5* 30.0 100.0474.50.054.9 359.2 10 30.0 100.0419.60.0109.7 293.4 20 30.0 100.0309.90.0219.4 161.7 30 30.0 100.0200.20.0329.1 30.1 35** 22.8 100.0145.90.0319.9 0.0 Notes: Minimum rate feasible under revenue constraint. **Maximum rate feasible under participation constraint.
154 Table 4-8. Results under performance bonds ( bond = $250/ha) and revenue-b ased royalties (high RIL costs) Revenuebased rate (%) Harvest (m3/ha) RIL adoption (%) Profit ($/ha) Illegal revenues ($/ha) Royalties ($/ha) Excess profits ($/ha) 9* 23.4 100.0179.90.083.6 37.9 10 23.4 100.0170.60.092.8 26.8 19* 12.2 100.095.50.0113.6 0.0 Notes: Minimum rate feasible under revenue constraint. **Maximum rate feasible under participation constraint.
155 0.00 0.25 0.50 0.75 1.00 012345678910111213141516171819202122 x illegalillegalh Figure 4-1. Probability of being caught breaking laws and paying fine as an increasing function of illegal behavior
156 A24 26 28 30 32 34 36 38 40 0481216202428323640 Number of AuditsHarvest (m3) .0 20 40 60 80 100RIL Adoption (%) Harvest RIL adoption B24 26 28 30 32 34 36 38 40 0481216202428323640 Number of AuditsHarvest (m3) .0 20 40 60 80 100RIL Adoption (%) Harvest RIL adoption Figure 4-2. Effect of increasi ng number of periodic audits. A) Actual RIL costs. B). High RIL costs.
157 A24 26 28 30 32 34 36 38 40 050100150200250300 Performance Bonds ($/ha)Harvest (m3) .0 20 40 60 80 100RIL Adoption (%) Harvest RIL adoption B24 26 28 30 32 34 36 38 40 050100150200250300 Performance Bonds ($/ha)Harvest (m3) .0 20 40 60 80 100RIL Adoption (%) Harvest RIL adoption Figure 4-3. Effect of increasing performance bonds A) Actual RIL costs. B). High RIL costs.
158 CHAPTER 5 CONCLUSIONS Findings and Methodological Advances This study used data from one of the longer running tropical forest experiments in the Eastern Brazilian Amazon region to develop a growth and yield matrix model based upon transition parameters estimated using a multinomial logit discrete choice model. This model is intended to advance the use of matrix models to study tropical forest management by, first, improving the capacity of the model to capture th e dynamic effects of harvest on forest structure and composition; and, second, by endogenizing the c hoice of harvest system that allows the manager to best meet objectives. The growth and yield model was embedded with in a dynamic forest management model in order to quantitatively analyze the dynamic cost-effectiveness of reduced impact logging (RIL) and sustainable yield constraints. Two new operational mathematical definitions for total volume and species-level sustainability were proposed. These definitions of sustainability focus on sustaining standing timber inve ntories across cutting cycle entr ies, rather than on sustaining harvest yields, as is typically the case. The resu lts of this study were mixed. While it is shown that RIL is cost-effective to implement, the bulk of the financial gains are due to harvest planning improvements over conventi onal techniques, rather than th rough protection of the forest resource. RIL is shown to marginally increase the volumes of merchantable timber available in future harvests, but does not significantly cha nge what is now well-known about logged forests; the structure and composition of the managed forest s will be different than that of the primary forest. Perhaps silvicultural techniques in addition to continued development of improved logging techniques will mitigate this effect in th e future. However, given the likelihood that
159 certain RIL practices may be more financially bene ficial than others, such as testing hollows or reducing wood waste and skidder-time in the forest there is the possibilit y that RIL may more often be partially implemented, as managers dete rmine the relative costs a nd benefits of shirking certain cost-ineffective activities. While volumes of merchantable timber predicte d to be available for the second and third harvests are significantly lower th an the available timber in the primary forests, the second and third harvests are projected to be profitable and ap pear to have the potential for sustainability at the stand-level. The Weakly Sustainable Inve ntory (WSI) scenario showed that stand-level sustainability is likely to be possible at lower volumes a nd lower economic value, but that sustainable management by the crite ria driving that scenario is vi able. The harvests under the Strongly Sustainable Invent ory (SSI) scenario were profitable but at the expense of harvesting very little of the most valuable timber. Wh ile profitable, the WSI and SSI scenarios leave valuable timber standing, creating st rong incentives for timber trespass. The fourth chapter developed a sequential harvest model in order to evaluate the response of a representative concessi onaire to royalty instruments under imperfect enforcement and performance bonds when there are incentives to harvest illegal volumes of timber and shirk RIL requirements. The model was developed in the form of a classic principal agent problem, in which the government is revenue-constrained and the concessionaire faces a participation constraint. The results show that each instrument has strengths and weakne ss that are contingent upon the costs firms face to implement RIL. In a weak enforcement environment, the result s show that audit pre ssure is unlikely to induce full compliance with harvest regulations. Under clear incentives for illegal logging, it is possible to maintain low-levels of illegal beha vior hidden from regulators. Meanwhile, the
160 effectiveness of performance bonds is attractive. However, as with all policy instruments, the development of credible institutions matters. The use of traditional royalty instruments such as the ad valorem and revenue-based royalty can effectively generate revenues and, in the case of revenue-based instruments, modify harvest behavior, but only under ve ry limited circumstances. When simple but critical variables change, such as the costs to implement RI L in this study, the outcomes under the same instrument can vary dramatically. In this model, the concessionaire is able to shift to illegal harvests when the royalty rate reduces profit ra tes to near the participation constraint. Meanwhile, if enforcement or bonds are eff ectively inducing overa ll compliance with rules, the concessionaires participation constraint can hinder the effectiveness of the instruments to modify harvest behavior in ways that might better protect future resource values and reduce externalities. Recalling the sustainability-con strained harvests of Chapter 3, the royalty instruments applied here are unlik ely to induce the harvest required for sustainability. In terms of capturing forest rent, these results appear to favor market-determined payments for harvest rights combined with more effective enfor cement to induce desired harvest volumes and distributions across species and sizes. While this study is performed w ith the intent to provide anal ytical support to public forest planning, there is no reason that the results should apply exclus ively to public lands. Many results may apply to private lands equally, alt hough the policy implicatio ns may differ across tenures, depending on over-arching management obj ectives for the different lands and public preferences for goods and services delivered from public lands. That said, perhaps one the most important implication of the results, which is no t original to this st udy (see e.g. Karsenty and Gourlet-Floury, 2006, and Van Gardingen 2006), is th at the public, forest se ctor, decisionmakers,
161 and other stakeholders such as non-governme ntal organizations and certifying bodies should reflect and revise their visions for the likely financial and ec ological outcomes of multi-cyclic management of natural forests in the Eastern Amazon region. The analysis in this study ha s shown that the restricted ha rvests which are required to sustain timber volumes or to simply follow cu rrent rules present opportunity costs. These assertions beg the questions, however, of who b ears these costs, and what are the available mechanisms to compensate for these costs? In a simple analysis, the concessionaire will face significantly lower profit margins if they follow sustainability guidelines. However, as the results under imperfect command and control en forcement show, the concessionaire may break the rules to satisfy the partic ipation constraint. Subsidies through environmental management incentives or low cost harvest rights might be dir ected toward private firms to compensate for the opportunity costs, but this raises the possibility of perhaps unfai rly shifting the burden of from the private agent to the public, which may not be welfare enhancing. Under the current language of the Public Forests Management Law (PFML), should markets for environmental services arise in the future, concessionaires have no right to receive compensation for these services in the same manner concessionaires have no claim to mineral rights that may underlie the forest concession. Even if the law is rewritten to allow such transfers, the specter of private appropriation of public resources is still present, as concessionaires are simply leaseholders of harvest and management rights on public lands. The potential solutions to these issues are probably not found at th e level of individual concessions, but, rather, at the regional-level, where harvests of public forestlands should be a component of larger economic development policies that include interactions with private sector logging, particularly logging on smallholder pr operties. Much will depend upon how natures
162 immense natural capital is converted into re gional development through direct and indirect employment and the redistribution of royalties an d fees. For example, how would forest sector employment and the regional economy through the multiplier effect be changed if the Brazilian government committed to the reduced but sustainable RIL harvests discussed in Chapter 3? Incentives and control measures should be develo ped that facilitate the sustainable multiple use of public forests, while contri buting effectively to regional de velopment, without crowding out smallholder activities which may be more welf are enhancing and less de structive of natural resources. Further research in this area would help elucidate th e costs and benefits of a wide range of policy choices. Data Limitations As with any study, there are limitations to the model that future efforts may address. The linear treatment of harvest cost s in this model can be improved significantly. For example, Bauch et al. (2007) present a C obb-Douglas harvest costs model ba sed upon data collected from a large survey within the Brazilian Amazon that potentially could be used to more accurately reflect nonlinear variable harv est costs, but the sample from which the model was estimated included few RIL operators, making it difficu lt to implement well within this study. Another crucial need is to develop an em pirically-based penalty function that more effectively models enforcement success as a func tion of enforcement effort and concessionaire behavior. This requires aud it by audit data collection, incl uding information on enforcement inputs, such as expenditures and personnel required per au dit, as well as outcomes, such as audit findings and results of any proceedings against the concessionaire should rule-breaking be identified. Also, as more concessions are implemented, it is important to continuously collect quality data on government costs. A recurrent problem identified in the concession economics literature
163 is that the government can become rent-seeki ng, attempting to increase government receipts by lowering royalty rates and expanding the land ba se under logging concessions. More empirical evidence is required to evaluate and mitig ate this risk on an ongoing basis. As the Brazilian experience with forest c oncessions is only beginning, there are ample opportunities to extend this research as data from new concessions accumulates. This data need highlights an important policy recommendation that the Brazilian government rapidly establish a forest sector social science data collection and research capacity within the new Servio Florestal Brasileiro. Because a large por tion of current Amazon region research capacity resides within nongovernmental research institutes strategic partnerships shoul d be further strengthened in order to start building long-ter m datasets on fundamental economic variables, such as harvest volumes, prices, costs, and enforcement effectiv eness, from both the private and government perspectives. Future Extensions of the Model The models developed within this study can be extended to address many issues. Several possible extensions are highly re levant to current policy debate s. First, the model can be extended to the landscape-scale for regional planning purposes. Under PFML, the Brazilian government, with assistance from nongovernmental research institutes, is actively planning the expansion of the state and national forest systems. Each forest will contain a mosaic of land uses, designed to meet multiple social, economic and ecological objectives. Some objectives will be complementary; other objectives will conflic t. Given this complexity, it is important to develop tools to aid decision-makers and stakehol ders to envision alternat ive landscape visions. One tool being developed is a forest-level optimization model that integrates geographic information systems and mathematical programmi ng techniques to design optimal landscapes according to the objectives being sought (Lenti ni, 2007). For example, a landscape that meets
164 goals for timber harvesting, community reserves and biodiversity areas can be designed. The model developed in this study can introduce a dynamic element to this planning tool. Simple dynamic rules can be generated that connect present and future harvests in order to better plan timber production on public forests over long time horizons. Given the evidence in this study and other similar studies that timber harvests un der current rules and best logging practices are unlikely to be sustainable, it is important to incorporate these dynamic decision rules to help decisionmakers avoid the uninte nded consequence of planning a status quo, boom-bust timber economy in and around public forests. In a second extension, the model can be adapte d to incorporate non-timber forest products and services, such as carbon sequestration. Rece nt Kyoto Protocol nego tiations have looked more favorably upon forested developing countri es to include averted deforestation, or compensated reduction, into the Protocol (G ullison et al., 2007). The central idea of compensated reduction is that c ountries can chose to reduce nati onal deforestation levels below historical levels and receive payments upon accomplishing targeted reductions, which would create significant incentives to reduce deforestation (San tilli et al., 2005). Under particular proposals, the incentives are direct ed toward decreases in overall deforestation rates, not toward particular projects, freeing governme nts to seek the gains with the lowest marginal costs. Given that the establishment of state and national forests is a si gnificant component of the Brazilian national strategy, and that the carbon sequestration benefits of reduced impact logging have been demonstrated (Boscolo et al., 1997; Boscolo and Buongiorno, 1997; Pinard and Putz, 1996; Putz and Pinard, 1993; Smith and Appl egate, 2004) resources could be directed at securing and improving long-term management of public produc tion forests. Given this discussion, the market value of carbon under a range of price sc enarios can be introduced to the model to
165 represent societys interest capturing global re sources in order to sustain flows of critical ecosystem services from managed forests. A social planners solu tion, which values carbon sequestration, can be compared to the private solutions found in this study to estimate the benefits and costs of di fferent policies under Kyoto. Third, the model can be used to evaluate add itional critical issues in tropical forest management. For example, the model can be used to better understand how exploiting lesserknown timber species can affect the viability of forest management in the future, an issue of critical importance but virtually unstudied using empirical data As discussed earlier, the Paragominas region of this study was central to the Eastern Amazon logging boom during the 1970s and 1980s. The commercial species list in th is region is extensive as mill operators have had sufficient time to develop markets and learn the milling requirements of various species, particular as many of the forest s of the region have been reente red and logged haphazardly for smaller-diameter trees and trees of lesserknown species before land conversion. A good exercise would be to, in a sense, step back in time in an effort to mimic the economic conditions within areas along current logging frontiers where only the species of highe st value are currently profitable. Within this context, the benefits of implementing RIL practices to protect future harvests is likely to be more pronounced than in this study. Another possibility is that the model can be used to examine th e specific costs and benefits of RIL practices in order to iden tify specific practices which are like ly to be ignored or practiced incompletely. This study could be of benefit to regulators and certifiers who are responsible for ensuring private agents are implementing best practice logging and forest management. Experimentation and Adaptiveness In the Brazilian Amazon, deforestation and fore st degradation continue s at a strong pace. The forces that drive deforestation and degrad ation will remain in place for the foreseeable
166 future. In many ways, Brazil is in the midst of an historical moment in the history of its forests. The Atlantic Forest, once clos e to disappearing is beginning to rebound, and much of the Brazilian Amazon still remains. The ratification of the PFML is just one of Brazilian societys many efforts to sustain the countrys great forests for future generations. While events move rapidly and there seems to be little time to save the forests, there is ample space for policy experimentation under the ne w law. Rather than muddle through with incremental changes to one-size-fits-all policy, policy experimentation can be performed. The experiments (or quasi-experiments) can be formal ly designed to elicit scientific information about the impacts of certain pol icies, programs, or procedures Ferraro and Pattanayak (2006) write that it is important to te st hypothesis about programs intended to protect biodiversity in the same way that hypotheses are tested in the eco logical sciences. Formal program evaluation enables the estimation of the causa l effect of interventions on outcomes (Ferraro and Pattanayak, 2006). Concessions will be allocated on a case-by-case basis. Additionally, in any given state or national forest, multiple concessions are likely to be allocated. This raises the possibility of structuring learning experiments by varying the treatments rece ived by the concessionaires. For example, the performance bond mechanism could be varied across pairs of logging firms in operating in sufficiently similar environments. A nother possibility would be to vary the auction process itself, comparing behavi or around bid systems or reserv ation prices. Concessionaire response to varying logging rules could be also be examined. In short, many possibilities exist for strengthening the learning process. Beyond formal experiments, policy and critical administrative rulings need to be adaptive to the wide range of conditions that exist in the region. The po licies and rules should also be
167 adaptive to emergent influences and new informa tion within the forest sector across sectors of the Brazilian and internationa l economies. The National Forestry Development Fund (Fundo Nacional de Desenvolvimento Florestal) establis hed by the PMFL can play a critical role in funding applied research that informs the dynamic course of concession policy and administration.
168APPENDIX A FAZENDA SETE SPECIES LIST GROUPS, AND PRICES Table A-1. Species groups, scientific names, co mmon names, and economic value class (stems/ha) Price-class Species Group Scientific name Common name No value Low value Med. Value High value Total Pioneer Apeiba burchelli Pente de macaco sulcado 2.0 0.0 0.0 0.0 2.0 Apeiba echinata Pente de macaco serrado 1.5 0.0 0.0 0.0 1.5 Apeiba sp Pente de macaco 0.3 0.0 0.0 0.0 0.3 Bagassa guianensis Tatajuba 0.0 0.0 0.3 0.0 0.3 Bellucia grossularioides Goiaba de anta 0.3 0.0 0.0 0.0 0.3 Bixa arborea Urucum da mata 0.1 0.0 0.0 0.0 0.1 Cecropia obtusa Imbaba branca 0.7 0.0 0.0 0.0 0.7 Cecropia sciadophylla Embaba 6.7 0.0 0.0 0.0 6.7 Eugenia heterochroma Goiabinha 0.9 0.0 0.0 0.0 0.9 Jacaranda copaia Para-para 0.0 1.4 0.0 0.0 1.4 Jacaratia espinosa Mamu 0.3 0.0 0.0 0.0 0.3 Laetia procera Pau jacare 0.0 5.0 0.0 0.0 5.0 Pouroma guianensis Mapatirana 3.3 0.0 0.0 0.0 3.3 Pourouma minor Embaba torm 0.1 0.0 0.0 0.0 0.1 Scheffera morototoni Morotot 0.0 1.0 0.0 0.0 1.0 Shichozolobium amazonicum Paric 0.0 0.1 0.0 0.0 0.1 Sloanea obtusa Urucurana 0.6 0.0 0.0 0.0 0.6 Vismia guianensis Lacre 1.0 0.0 0.0 0.0 1.0 Zanthoxylum rhoifolia Limozinho/ Tamanqueira 2.7 0.0 0.0 0.0 2.7 Not identified Imbaba verm elha 0.1 0.0 0.0 0.0 0.1 Not identified Pau de gafanhoto 1.2 0.0 0.0 0.0 1.2 Pioneer total 21.8 7.5 0.3 0.0 29.6
169Table A-1. Continued. Price-class Species Group Scientific name Common name No value Low value Med. Value High value Total Light-demanding Allophylus robustus Espeturana trifoliar 0.1 0.0 0.0 0.0 0.1 Balizia pedicellaris Mapucuxi vermelho 0.1 0.0 0.0 0.0 0.1 Bombax paraensis Mamorana terra firme 0.1 0.0 0.0 0.0 0.1 Byrsonima aerugo Muruc 0.3 0.0 0.0 0.0 0.3 Caraipa grandifolia Louro tamaquar 0.0 1.4 0.0 0.0 1.4 Casearia arborea Casiaria arboria 0.1 0.0 0.0 0.0 0.1 Cordia bicolor Freijo branco 0.0 9.2 0.0 0.0 9.2 Cordia scabrida Freijozinho 1.3 0.0 0.0 0.0 1.3 Enterolobium maximum Orelha de macaco 0.0 0.1 0.0 0.0 0.1 Eriotheca globosa Mamorana 0.6 0.0 0.0 0.0 0.6 Inga alba Inga vermelha 0.0 0.2 0.0 0.0 0.2 Inga capitata Inga 0.5 0.0 0.0 0.0 0.5 Inga cylindrica Ing branca 36.5 0.0 0.0 0.0 36.5 Inga pezizifera Ing cilndrica 0.1 0.0 0.0 0.0 0.1 Inga dumosa Ing maguinata 0.1 0.0 0.0 0.0 0.1 Inga edulis Inga cipo 1.4 0.0 0.0 0.0 1.4 Inga eplendens Ing facozinho 0.1 0.0 0.0 0.0 0.1 Inga falcistipula Ing estpula pequena 0.2 0.0 0.0 0.0 0.2 Inga gracilifolia Ing corao de preguia 0.1 0.0 0.0 0.0 0.1 Inga ingoides Ing folha peluda 1.4 0.0 0.0 0.0 1.4 Inga macrophylla Ing folha grande 0.1 0.0 0.0 0.0 0.1 Inga melinones Inga sulcado 0.5 0.0 0.0 0.0 0.5 Inga microcalix Ing de sangue 0.1 0.0 0.0 0.0 0.1
170Table A-1. Continued. Price-class Species Group Scientific name Common name No value Low value Med. Value High value Total Light-demanding (continued) Not identified Tento folha miuda 0.1 0.0 0.0 0.0 0.1 Light-demanding total 61.2 23.6 0.3 0.0 85.2 Intermediate Aioea att. Densiflora Louro preto folha brilhante 0.1 0.0 0.0 0.0 0.1 Aiouea sp Louro folha verticilada 0.1 0.0 0.0 0.0 0.1 Annona sp Envira sombrera 2.1 0.0 0.0 0.0 2.1 Astronium lecointei Muiracatiara 0.0 0.0 0.7 0.0 0.7 Auxemma oncocalyx Pau branco 15.9 0.0 0.0 0.0 15.9 Brosimum obovata Murure 0.0 0.7 0.0 0.0 0.7 Calophyllum brasiliense Jacareba 0.0 0.1 0.0 0.0 0.1 Carapa guianensis Andiroba 0.0 0.0 0.8 0.0 0.8 Clarisa racemosa Guariba 0.0 2.3 0.0 0.0 2.3 Clarisia ilicifolia Janit 0.1 0.0 0.0 0.0 0.1 Copaifera duckei Copaba 0.0 0.5 0.0 0.0 0.5 Dialium guianesis Juta pororoca 0.0 0.1 0.0 0.0 0.1 Franchetella sangotiana Guajararana 0.2 0.0 0.0 0.0 0.2 Himatanthus sucuuba Sucuuba 1.0 0.0 0.0 0.0 1.0 Licaria rigida Louro amarelo 0.0 0.5 0.0 0.0 0.5 Luehea speciosa Aoita cavalo 0.1 0.0 0.0 0.0 0.1 Moronobea coccinea Ananin 0.0 0.4 0.0 0.0 0.4 Nectandra cuspidata Louro tamanco 0.0 0.1 0.0 0.0 0.1 Nectandra grandis Louro puxur bravo 0.1 0.0 0.0 0.0 0.1 Nectandra pichurim Louro 0.0 0.1 0.0 0.0 0.1
171Table A-1. Continued. Price-class Species Group Scientific name Common name No value Low value Med. Value High value Total Intermediate (continued) Neoxythece robusta Guajar 0.1 0.0 0.0 0.0 0.1 Ocotea caudata Louro preto 0.0 1.0 0.0 0.0 1.0 Ocotea cernu Louro da folhona 0.1 0.0 0.0 0.0 0.1 Ocotea fragantissima Louro canela 0.0 0.1 0.0 0.0 0.1 Ocotea glomerata Louro abacate 0.0 1.2 0.0 0.0 1.2 Ocotea guianensis Louro branco 0.0 0.1 0.0 0.0 0.1 Ocotea longifolia Louro vermelho vincado 0.2 0.0 0.0 0.0 0.2 Ocotea rubra Louro vermelho 0.0 0.0 4.5 0.0 4.5 Ormosia coutinhoi Buiuu 0.0 0.7 0.0 0.0 0.7 Parahancornia amapa Amap 0.0 0.0 0.2 0.0 0.2 Platymiscium filipes Macacaba 0.0 0.0 0.3 0.0 0.3 Poecilanthe effusa Gema de ovo 5.2 0.0 0.0 0.0 5.2 Pseudopiptadenia suaveolens Timborana 0.0 3.0 0.0 0.0 3.0 Pterocarpus rohrii Mututi 1.4 0.0 0.0 0.0 1.4 Sacoglottis amazonica Uxirana 0.0 1.2 0.0 0.0 1.2 Sagotia racemosa Uxirana 0.0 0.1 0.0 0.0 0.1 Sclerobium melanocaroum Tachi vermelho 0.0 1.3 0.0 0.0 1.3 Tachigalia paniculata Tachi preto 0.0 8.9 0.0 0.0 8.9 Tetragastris altissima Breu manga 0.0 11.3 0.0 0.0 11.3 Tetragastris panamensis Barrote 0.3 0.0 0.0 0.0 0.3 Xylopia nitida Envira branca/ Envira cana 6.8 0.0 0.0 0.0 6.8 Not identified Breu amesclo 0.1 0.0 0.0 0.0 0.1 Not identified Envira sombreira 0.8 0.0 0.0 0.0 0.8 Not identified Louro cheiroso 0.1 0.0 0.0 0.0 0.1
172Table A-1. Continued. Price-class Species Group Scientific name Common name No value Low value Med. Value High value Total Intermediate (continued) Not identified Louro jandauba 0.1 0.0 0.0 0.0 0.1 Not identified Pau de colher 1.3 0.0 0.0 0.0 1.3 Not identified Pitaca 0.1 0.0 0.0 0.0 0.1 Not identified Tinteiro 4.7 0.0 0.0 0.0 4.7 Not identified Ucuuba preta 3.0 0.0 0.0 0.0 3.0 Intermediate total 43.7 33.7 6.4 0.0 83.8 Shade-tolerant Aspidosperma album Araracanga 0.0 4.5 0.0 0.0 4.5 Aspidosperma nitidum Carapanaba 0.1 0.0 0.0 0.0 0.1 Bowdichia nitida Sucupira amarela 0.0 0.0 0.3 0.0 0.3 Chaunochiton kappleri Pau vermelho 0.1 0.0 0.0 0.0 0.1 Chrysophyllum lucentifollium Abiu casca grossa 0.0 7.6 0.0 0.0 7.6 Cupania hirsuta Espeturana peluda 0.1 0.0 0.0 0.0 0.1 Cupania scrobiculata Espeturana 1.8 0.0 0.0 0.0 1.8 Dendrobangia boliviana Caferana 0.1 0.0 0.0 0.0 0.1 Diospyros dukei Caqui preto 18.8 0.0 0.0 0.0 18.8 Diospyros melinoni Caqui branco 0.1 0.0 0.0 0.0 0.1 Diospyros praetermissa Caqui 0.5 0.0 0.0 0.0 0.5 Diospyros sp Caqui casca grossa 0.1 0.0 0.0 0.0 0.1 Diospyros tectranda Caqui casca dura 0.1 0.0 0.0 0.0 0.1 Diploon venezuelana Seringarana/ Mangabarana folha pequena 0.0 0.1 0.0 0.0 0.1 Drypetes variabilis Maparan 0.6 0.0 0.0 0.0 0.6
173Table A-1. Continued. Price-class Species Group Scientific name Common name No value Low value Med. Value High value Total Shade-tolerant (continued) Ecclinusa abbreviata Abiu folha peluda 0.1 0.0 0.0 0.0 0.1 Eperua schonburgkiana Muirapiranga 0.0 0.1 0.0 0.0 0.1 Eschweilera apiculata Ripeiro 0.2 0.0 0.0 0.0 0.2 Eschweilera blanchetiana Mata mata preto 17.8 0.0 0.0 0.0 17.8 Eschweilera coriaceae Mat mat branco 4.8 0.0 0.0 0.0 4.8 Eschweilera grandeflorum Mat mat grande flor um 0.1 0.0 0.0 0.0 0.1 Eschweilera pedicelata Mat mat 0.7 0.0 0.0 0.0 0.7 Guarea kunthiana Andirobarana folha grande 0.1 0.0 0.0 0.0 0.1 Guatteria olivacea Envira preta folha grande 0.0 0.4 0.0 0.0 0.4 Guatteria poeppigiana Envira preta casca grossa 0.0 0.1 0.0 0.0 0.1 Guatteria schomburgkiana Envira preta 0.0 0.3 0.0 0.0 0.3 Guatteria schomburgkiana Envira preta casca sulcada 5.0 0.0 0.0 0.0 5.0 Guatteria schomburgkiana Envira preta folha peluda 0.0 0.5 0.0 0.0 0.5 Guatteria sp Envira vermelha 13.2 0.0 0.0 0.0 13.2 Lecythis idatimon Jatereu 38.4 0.0 0.0 0.0 38.4 Lecythis lurida Jarana 0.0 2.5 0.0 0.0 2.5 Licania heteromorpha Macucu/ macucu de sangue 0.4 0.0 0.0 0.0 0.4 Licania kunthiana Pintadinho 0.3 0.0 0.0 0.0 0.3 Lindackeria paraensis Canela de velho/folha seca 0.1 0.0 0.0 0.0 0.1 Macrolobium campestre Iperana bifoliar 20.9 0.0 0.0 0.0 20.9 Manilkara amazonica Maparajuba 0.0 0.0 0.9 0.0 0.9 Manilkara huberi Maaranduba 0.0 0.0 3.3 0.0 3.3
174Table A-1. Continued. Price-class Species Group Scientific name Common name No value Low value Med. Value High value Total Shade-tolerant (continued) Manilkara paraensis Maarandubinha 0.0 0.0 0.3 0.0 0.3 Maquira sclerophylla Muratinga 0.4 0.0 0.0 0.0 0.4 Maytenos guianensis Chichu/ Xixu 0.3 0.0 0.0 0.0 0.3 Micropholis melinoniana Currupixa 0.0 0.0 0.3 0.0 0.3 Minquartia guianensis Acariquara 0.0 0.1 0.0 0.0 0.1 Mouriria plasschaerti Muiraba 0.1 0.0 0.0 0.0 0.1 Neea sp. Joo mole 28.8 0.0 0.0 0.0 28.8 Perebea guianensis Moiratinga 0.0 6.1 0.0 0.0 6.1 Pithecolobium racemosum Angelim rajado 0.0 0.2 0.0 0.0 0.2 Pouteria cladantha Abiurana 6.0 0.0 0.0 0.0 6.0 Pouteria eugeniifolia Guajar pedra 0.1 0.0 0.0 0.0 0.1 Pouteria lasiocarpa Abiu seco 33.6 0.0 0.0 0.0 33.6 Pouteria macrophylla Abiu cutiti/ Abiu cutiti preto 2.2 0.0 0.0 0.0 2.2 Pouteria manausensis Guajar preto 3.7 0.0 0.0 0.0 3.7 Pouteria reticulata Abiurana amarela 1.2 0.0 0.0 0.0 1.2 Pouteria sp Abiu/Abiu sem casca 4.0 0.0 0.0 0.0 4.0 Protium decandrum Breu vermelho 0.1 0.0 0.0 0.0 0.1 Protium tenuifolium Breu 6.8 0.0 0.0 0.0 6.8 Rinorea guianensis Quariquarana 1.4 0.0 0.0 0.0 1.4 Sagotia racemosa Arataciu 24.9 0.0 0.0 0.0 24.9 Sandwithiodoxa egregia Guajarazinho 0.1 0.0 0.0 0.0 0.1 Sapium lanciolatum Murupita 0.5 0.0 0.0 0.0 0.5 Syzygiopsis oppositifolia Guajara bolacha 0.0 0.5 0.0 0.0 0.5 Talisia CF. intermedia Pitomba da mata 0.1 0.0 0.0 0.0 0.1
175Table A-1. Continued. Price-class Species Group Scientific name Common name No value Low value Med. Value High value Total Shade-tolerant (continued) Theobroma speciosa Cacau 2.4 0.0 0.0 0.0 2.4 Trichilia micrantha Cachu 0.2 0.0 0.0 0.0 0.2 Zizyphus itacaiunensis Maria preta 0.2 0.0 0.0 0.0 0.2 Zollernia paraensis Pau ferro/ pau santo 1.8 0.0 0.0 0.0 1.8 Not identified Abiu doce 0.1 0.0 0.0 0.0 0.1 Not identified Abiu folha grande 0.1 0.0 0.0 0.0 0.1 Not identified Abiu vermelho 0.5 0.0 0.0 0.0 0.5 Not identified Abiurana caramuri 0.1 0.0 0.0 0.0 0.1 Not identified Abiurana casca fina 0.1 0.0 0.0 0.0 0.1 Not identified Abiurana pitomba 0.2 0.0 0.0 0.0 0.2 Not identified Abiurana ucuubarana 0.1 0.0 0.0 0.0 0.1 Not identified Acariquarana 0.3 0.0 0.0 0.0 0.3 Not identified Caqui folha pequena 0.1 0.0 0.0 0.0 0.1 Not identified Caraipe 0.1 0.0 0.0 0.0 0.1 Not identified Casca grossa 0.1 0.0 0.0 0.0 0.1 Not identified Chichua casca grossa 0.2 0.0 0.0 0.0 0.2 Not identified Corao de negro 0.6 0.0 0.0 0.0 0.6 Not identified Currupixa folha miuda 0.1 0.0 0.0 0.0 0.1 Not identified Envira danta 0.7 0.0 0.0 0.0 0.7 Not identified Envira folha fina 0.1 0.0 0.0 0.0 0.1 Not identified Mata mata 0.1 0.0 0.0 0.0 0.1 Not identified Mata mata branco 0.7 0.0 0.0 0.0 0.7 Not identified Pau de remo 0.1 0.0 0.0 0.0 0.1 Shade-tolerant total 247.3 22.8 5.0 0.0 275.1
176Table A-1. Continued. Price-class Species Group Scientific name Common name No value Low value Med. Value High value Total Emergent Caryocar glabum Piquiarana 0.0 0.0 1.1 0.0 1.1 Caryocar villosum Piqui 0.0 0.0 0.4 0.0 0.4 Cedrela odorata Cedro 0.0 0.0 0.0 0.1 0.1 Couratari guianensis Tauari/ Estopeiro folha grande 0.0 0.0 0.1 0.0 0.1 Couratari oblongfolia Tauari/ Estopeiro folha peque 0.0 0.0 0.3 0.0 0.3 Dinizia excelsa Angelim pedra 0.0 0.0 0.2 0.0 0.2 Dipteryx odorata Cumaru 0.0 0.0 0.3 0.0 0.3 Hymenaea courbaril Jatob 0.0 0.0 1.2 0.0 1.2 Hymenaea palustris Juta mirim 0.0 0.0 0.7 0.0 0.7 Lecythis pisonis Sapucaia 0.0 0.5 0.0 0.0 0.5 Qualea cf. lancifolia Mandioqueiro 0.0 0.1 0.0 0.0 0.1 Tabebuia impetiginosa Ip roxo 0.0 0.0 0.0 0.4 0.4 Tabebuia serratifolia Ip amarelo 0.0 0.0 0.0 0.3 0.3 Emergent total 0.0 0.7 4.3 0.9 5.8 Unknown Brosimum aubletii Gameleira 0.1 0.0 0.0 0.0 0.1 Not identified Caniceiro 1.0 0.0 0.0 0.0 1.0 Not identified Galhudinho 0.5 0.0 0.0 0.0 0.5 Not identified Guaruta 0.3 0.0 0.0 0.0 0.3
177Table A-1. Continued. Price-class Species Group Scientific name Common name No value Low value Med. Value High value Total Unkown (continued) Not identified Pau seco 0.1 0.0 0.0 0.0 0.1 Not identified Pepino 0.1 0.0 0.0 0.0 0.1 Not identified Taquari 0.1 0.0 0.0 0.0 0.1 Not identified Not identified 1.0 0.0 0.0 0.0 1.0 Not identified Not identified 13.8 0.0 0.0 0.0 13.8 Unknown total 17.0 0.0 0.0 0.0 17.0 Overall total 391.0 87.7 11.9 0.0 490.6
178 APPENDIX B LIST OF VARIABLES, VECTORS, AND MATRICES i index for species group 1,..,im j index for size 1,.., jn s index for harvest system 0,1,2s t index for time, 0,..,tT T time horizon of optimization problems q index for randomly-timed independent audit Q the expected number of randomly-timed independent audits years between each transition in the growth model transitions in the growth model between harvest entries proportion of reduced impact logging (RIL) implemented illegal the degree to which RIL is not implemented when required 0 y 0 ij y trees/ha in species group i and size j at time 0; the initial condition of the stand t y ijty trees/ha in species group i and size j at time t y ij y equilibrium trees/ha in species group i and size j m t y m ijt y merchantable trees/ha in species group i and size j at time t m y ijy equilibrium merchantable trees/ha in species group i and size j m y ij y merchantable trees/ha in species group i and size j when a proportion, of RIL is implemented th ijth trees/ha harvested from species group i and size j at time t illegalh harvest volume over the legal limit (m3/ha) ijh trees/ha legally harv ested from species group i and size j
179 s D diagonal matrix containing logging damage coefficients for trees in species group i and size j under harvest system type s s td ijstd trees/ha killed by damage from species group i and size j under harvest system s at time t m s td m ijstd merchantable trees/ha killed by damage from species group i and size j under harvest system s at time t d ijd trees/ha killed by damage from species group i and size j when a proportion, of RIL is implemented m d m ijd merchantable trees/ha killed by damage from species group i and size j when a proportion, of RIL is implemented s G matrix of density-dependent transi tion probabilities under harvest system s s G repeating cycle of growth matr ix applications that arise from the loggers choice of harvest system s s A transition matrix for harvest system s R ingrowth matrix that captures the dens ity dependent component of recruitment s r fixed recruitment vector under harvest system s I identity matrix S diagonal matrix whose elements represent th e proportion of the stems from commercial species within each species group i and size j Q diagonal matrix who elements represent the proportion of trees in each species group i and size j with stem form appropriate for milling H diagonal matrix whose elements represent the proportion of stems in each species group i and size j thought to be hollow s M diagonal matrix who elements represent the perceived merchantable proportion for each species group i and size j under harvest treatment s M diagonal matrix who el ements represent the perceived merchantable proportion for each species group i and size j when a proportion, of RIL is implemented
180 ij predicted commercial volume (m3) per tree in species group i and size j ijsvc variable harvest cost per tree ($/tree) in species group i and size j under harvest system s ijvc variable harvest cost per tree ($/tree) in species group i and size j when a proportion, of RIL is implemented s f c fixed costs ($/ha) under harvest system s f c fixed costs ($/ha) when a proportion, of RIL is implemented r real discount rate discount function s waste proportion of commercial volume ha rvest wasted under harvest system s waste proportion of commercial volume harvest wasted when a proportion, of RIL is implemented t profit at time t E expected profit the minimum acceptable gross profit for the concessionaire s the proportion of hollow trees not identified by the sawyer as hollow x weighted sum of potentially illegal activities forms x function that generates the probabil ity of being caught and paying a fine parameter that shif ts the distribution of x weighs the relative contribution of under-imple mentation of RIL to th e illegality factor. f fine ($/ha) paid when caught breaking rules bond performance bond ($/ha) % compliance with RIL and ha rvest regulations to determ ine how much performance bond will be returned
181 ER Expected government revenue R required level of government revenue royalty charge area fee cc certification costs ($/ha)
182 LIST OF REFERENCES Amacher, G.S., 2006. Corruption: A challenge for ec onomists interested in forest policy design. Journal of Forest Economics 12, 85-89. Amacher, G.S., Brazee, R.J., 1997. Designing forest taxes with varying government preferences and budget targets. Journal of Environmenta l Economics and Management 32, 323-340. Amacher, G.S., Brazee, R.J., Witvliet, M., 2001. Royalty systems, government revenues, and forest condition: An application fr om Malaysia. Land Economics 77, 300-313. Amacher, G.S., Koskela, E., Ollikainen, M., 2007. Royalty reform and illegal reporting of harvest volumes under alterna tive penalty schemes. In pr ess Environmental & Resource Economics. Bacha, C.J.C., Rodriguez, L.C.E., 2007. Profitab ility and social impact s of reduced impact logging in the Tapajs National Forest, Br azil a case study. Ecological Economics 63, 70-77. Banco Nacional de Desenvolvimento Econmico e So cial BNDES., 2007. Manual da Taxa de Juros de Longo Prazo (TJLP). Barany, M., Hammett, A.L., Araman, P., 2003. Lesser used species of Bolivia and their relevance to sustainable forest manageme nt. Forest Products Journal 53, 28-33. Barreto, P., Amaral, P., Vidal, E., Uhl, C., 1998. Costs and benefits of forest management for timber production in eastern Amazonia. Fo rest Ecology and Management 108, 9-26. Bauch, S.C., Amacher, G.S., Merry, F.D., 2007. Co sts of harvesting, transportation and milling in the Brazilian Amazon: estimation and policy implications. Forest Policy and Economics 9, 903-915. Becker, G.S., 1968. Crime and punishment: an ec onomic approach. Journal of Political Economy 76, 169-217. Boltz, F., 2003. Optimal Management of Bolivia n Tropical Dry Forests. Ph.D. Dissertation, School of Forest Resources and Conservation, Un iversity of Florida. Gainesville, Florida. Boltz, F., Carter, D.R., 2006. Mult inomial logit estimation of a matr ix growth model for tropical dry forests of eastern Bolivia. Canadian Journal of Forest Research 36, 2623-2632. Boltz, F., Holmes, T.P., Carter, D.R., 2003. Economic and environmental impacts of conventional and reduced-impact logging in tropical South America: a comparative review. Forest Policy and Economics 5, 69-81. Boscolo, M., Buongiorno, J., 1997. Managing a tropi cal rainforest for timber, carbon storage, and tree diversity. Commonwealth Forestry Review 76, 246-254.
183 Boscolo, M., Buongiorno, J., Panayatou T., 1997. Simulating options for carbon sequestration through improved management of a lowland tropical rainforest. Environment and Development Economics 2, 239-261. Boscolo, M., Vincent, J.R., 2000. Promoting bett er logging practices in tropical forests: A simulation analysis of alternative regulations. Land Economics 76, 1-14. Boscolo, M., Vincent, J.R., 2003. Nonconvexities in the production of timber, biodiversity, and carbon sequestration. Journa l of Environmental Economics and Management 46, 251268. Boyland, M., 2006. The economics of using forests to increase carbon storage. Canadian Journal of Forest Research 36, 2223-2234. Brito, B., Barreto, P., 2004. Aplicao da Lei de Crimes Ambientais pe la Justia Federal no Setor Florestal do Par. IMAZON, Belm, Brazil. Brito, B., Barreto, P., 2006. A eficcia da aplicao da lei de crimes ambientais pelo Ibama para proteo de florestas no Par. Revi sta de Direito Ambiental 43, 35-65. Brito, B., Barreto, P., Rothman, J., 2005. Brazil' s New Environmental Crimes Law: an Analysis of its Effectiveness in Protecting the Amaz onian Forests. IMAZON, Belm, Brazil. Buongiorno, J., 2001. Quantifying the implications of transformation from even to uneven-aged forest stands. Forest Ecology and Management 151, 121-132. Buongiorno, J., Michie, B.R., 1980. A matrix mode l of uneven-aged forest management. Forest Science 26, 609-625. Buongiorno, J., Peyron, J.L., Houllier, F., Bruc iamacchie, M., 1995. Growth and management of mixed-species, uneven-aged forests in the French Jura implications for economic returns and tree diversity. Forest Science 41, 397-429. Buschbacher, R., 1990. Natural forest management in the humid tropics: ecological, social, and economic considerations. Ambio 19, 253-258. Campos, M.T., Nepstad, D.C., 2006. Smallhol ders, the Amazon's new conservationists. Conservation Biology 20, 1553-1556. Chomitz, K.M., Buys, P., De Luca, G., Th omas, T.S., Wertz-Kanounnikoff, S., 2007. At Loggerheads: Agricultural E xpansion, Poverty Reduction, and Environment in the Tropical Forests. The Worl d Bank, Washington, D.C. Daly, H., 1996. Beyond Growth. Beacon Press, Boston. Dauber, E., Fredericksen, T.S., Pena-Claros, M ., 2006. Sustainability of timber harvesting in Bolivian tropical forests. Forest Ecology and Management 214, 294-304.
184 De Graaf, N.R., Poels, R.L.H., Van Rampaey, R.S. A.R., 1999. Effect of silvicultural treatment on growth and mortality of rainforest in Surinam over long periods. Forest Ecology and Management 124, 123-135. Dekker, M., De Graaf, N.R., 2003. Pioneer and climax tree regenerati on following selective logging with silviculture in Suriname. Forest Ecology and Management 172, 183-190. Delacote, P., 2005a. Forestry sector concentratio n, corruption, and forest over-exploitation. Manuscript. Dykstra, D.P., 2002. Reduced Impact L ogging: Concepts and Issues. P. 23-39 in Applying Reduced Impact Logging to Advance Sustaina ble Forest Management, Enters, T., P.B. Durst, G.B. Applegate, P.C.S. Kho, and G. Man (eds.), FAO.
185 Gullison, R.E., Frumhoff, P.C., Canadell, J.G., Fi eld, C.B., Nepstad, D.C., Hayhoe, K., Avissar, R., Curran, L.M., Friedlingstein, P., Jones, C.D., Nobre, C., 2007. Tropical forests and climate policy. Science 316, 985-986. Gullison, R.E., Panfil, S.N., Strouse, J.J., Hubbell, S.P., 1996. Ecology and management of mahogany (Swietenia macrophylla King) in the Chimanes Forest, Beni, Bolivia. Botanical Journal of the Linnean Society 122, 9-34. Hausman, J., McFadden, D., 1984. Specificati on tests for the multinomial logit model. Econometrica 52, 1219-1240. Heal, G.M., 1998. Valuing the Future: Economic Theory and Sustainability. Columbia University Press, New York. Hirakuri, S.R., 2003. Can Law Save the Forest? Le ssons from Finland and Brazil. Center for International Forestry Research (CIFOR), Bogor Barat, Indonesia. Holmes, T.P., Blate, G.M., Zweede, J.C., Pereir a, R., Barreto, P., Boltz, F., Bauch, R., 2002. Financial and ecological indicators of reduced impact l ogging performance in the eastern Amazon. Forest Ecology a nd Management 163, 93-110. Hyde, W.F., Sedjo, R.A., 1992. Ma naging tropical forests reflectio ns on the rent distribution discussion. Land Economics 68, 343-350. IBAMA., 2002. Relatrio Tcnico: Manejo Fl orestal Sustentvel na Amaznia, Ano 2002. Ministrio do Meio Ambiente, Braslia, Brazil. Johns, J.S., Barreto, P., Uhl, C., 1996. Loggi ng damage during planned and unplanned logging operations in the eastern Amazon. Fore st Ecology and Management 89, 59-77. Kaimowitz, D., 2003. Forest law enforcement a nd rural livelihoods. Inte rnational Forestry Review 5, 199-210. Karsenty, A., Gourlet-Fleury, S., 2006. Assessing sustainability of l ogging practices in the Congo Basin's managed forests: The issue of commercial species recovery. Ecology and Society 11. Keller, M., Asner, G.P., Blate, G., McGlocklin, J., Merry, F., Pea-Claros, M., Zweede, J., 2007. Timber production in selectively logged tropical forests in South America. Frontiers in Ecology and the Environment 5, 213-216. Kramme, L.A., Price, S.P., 2005. Practical Actions to Combat Illegal Logging: a Summary of a Multi-stakeholder Dialogue on Best Practi ces for Business and Civil Society. The Forests Dialogue, New Haven, Connecticut. Lentini, M., 2007. Land Use Allocation under Multiple Objectives in Amaznia: Creating a Tool to Guide Zoning of Public Production Forests. M.S. Thesis, School of Forest Resources and Conservation, University of Florida. Gainesville, Florida.
186 Lentini, M., Pereira, D., Celentano, R., Pereir a, R., 2005. Fatos Florestais da Amaznia 2005. IMAZON, Belm, Brazil. Leruth, L., Paris, R., Ruzicka, I., 2001. The co mpiler pays principle: The limits of fiscal approaches toward sustainable forest ma nagement. IMF Staff Papers 48, 397-423. Lima, E., Merry, F., Nepstad, D., Amacher, G., Azevedo-Ramos, C., Lefebvre, P., Resque, F., 2006. Searching for sustainability: Forest po licies, smallholders, and the trans-Amazon highway. Environment 48, 26-38. Lin, C.R., Buongiorno, J., 1998. Tree diversity landscape diversity, and economics of maplebirch forests: Implications of Markovian models. Management Science 44, 1351-1366. Lin, C.R., Buongiorno, J., 1999a. Managing forest s for tree and landscap e diversity. Part I: Predictions with Markov-chain models. P. 137-157 in Multiple Use of Forests and Other Natural Resources: Aspects of Theory and A pplication, Helles, F., P. Holten-Andersen, and L. Wichmann (eds.), Kluwer Academic Publishers, Boston. Lin, C.R., Buongiorno, J., 1999b. Managing forest s for tree and landscap e diversity. Part II: Optimisation with Markovchain model. P. 158-174 in Multiple Use of Forests and Other Natural Resources: Aspects of Theory and A pplication, Helles, F., P. Holten-Andersen, and L. Wichmann (eds.), Kluwer Academic Publishers, Boston. Lin, C.R., Buongiorno, J., Vasievich, M., 1996. A multi-species, density-dependent matrix growth model to predict tree diversity and income in northern hardwood stands. Ecological Modelling 91, 193-211. Lu, H.C., Buongiorno, J., 1993. Long-term and shortterm effects of altern ative cutting regimes on economic returns and ecological diversity in mixed-species forests. Forest Ecology and Management 58, 173-192. Luckert, M.K., Williamson, T., 2005. Should sustai ned yield be part of sustainable forest management? Canadian Journal of Forest Research 35, 356-364. Mendoza, G.A., Setyarso, A., 1986. A transition matrix forest growth-model for evaluating alternative harvesting schemes in Indonesia. Forest Ecology and Management 15, 219228. Merry, F., Amacher, G., Nepstad, D., Lima, E ., Lefebvre, P., Bauch, S., 2006. Industrial development on logging frontiers in the Br azilian Amazon. Intern ational Journal of Sustainable Development 9, 277-296. Merry, F.D., Amacher, G.S., 2005. Forest taxes, timber concessions, and policy choices in the Amazon. Journal of Sustainable Forestry 20 15-44. Michie, B.R., Buongiorno, J., 1984. Estimation of a matrix model of forest growth from remeasured permanent plots. Forest Ecology and Management 8, 127-135.
187 Monteiro, A., Souza Jr., C., 2006. Imagens de Satli te para Avaliar Planos de Manejo Florestal. No. 9. IMAZON, Belm, Brazil. Nepstad, D., Azevedo-Ramos, C., Lima, E., McGrat h, D., Pereira, C., Merry, F., 2004. Managing the Amazon timber industry. C onservation Biology 18, 575-577. Neumayer, E., 1999. Weak versus Strong Sustaina bility: Exploring the Li mits of Two Opposing Paradigms. Edward Elgar, Northampton, MA. nal, H., 1997. Trade-off between structural diversity and economic objectives in forest management. American Journal of Agricultural Economics 79, 1001-1012. Palmer, C.E., 2003b. The nature of corruption in forest management. World Development 6, 110. Paris, R., Ruzicka, I., Speechly, H., 1991. Performance guarantee bonds for commercial management of natural forests: early e xperience from the Philippines. Commonwealth Forestry Review 73, 106-112. Pearce, D., Putz, F.E., Vanclay, J.K., 2003. Sustai nable forestry in the tropics: panacea or folly? Forest Ecology and Management 172, 229-247. Pearce, D.W., Markandya, A., Barbier, E., 1989. Blueprint for a Green Economy. Earthscan Publications, London. Phillips, P.D., de Azevedo, C.P., Degen, B., Thom pson, I.S., Silva, J.N.M., Van Gardingen, P.R., 2004. An individual-based spatially explic it simulation model for strategic forest management planning in the eastern Am azon. Ecological Modelling 173, 335-354. Picard, N., Bar-Hen, A., Franc, A., 2002. M odeling forest dynamics with a combined matrix/individual-based model. Forest Science 48, 643-652. Pinard, M.A., Putz, F.E., 1996. Retaining forest biomass by reducing logg ing damage. Biotropica 28, 278-295. Plumptre, R.A., 1996. Links between utilisation, product marketing, and forest management in tropical moist forests. Commonwea lth Forestry Review 75, 316-324. Pokorny, B., Sabogal, C., Silva, J.N.M., Bernardo, P., Souza, J., Zweede, J., 2005. Compliance with reduced-impact harvesting guidelines by ti mber enterprises in terra firme forests of the Brazilian Amazon. Internati onal Forestry Review 7, 9-20. Polinsky, A.M., Shavell, S., 1979. Optimal trad eoff between the probability and magnitude of fines. American Economic Review 69, 880-891. Polinsky, A.M., Shavell, S., 2000. The economic theo ry of public enforcement of law. Journal of Economic Literature 38, 45-76.
188 Putz, F.E., Dykstra, D.P., Heinrich, R., 2000. Why poor logging practices pe rsist in the tropics. Conservation Biology 14, 951-956. Putz, F.E., Pinard, M.A., 1993. Reduced-impact logging as a carbon-offset method. Conservation Biology 7, 755-757. Repetto, R., Gillis, M., 1988. Public Policies a nd the Misuse of Forest Resources. Cambridge University Press, Cambridge. Rhodes, W.M., Allen, E.P., Callahan, M., 2006. Illegal Logging: a Market -based Analysis of Trafficking in Illegal Timber. Abt A ssociates Inc., Cambridge, Massachusetts. Rice, R.E., Sugal, C.A., Ratay, S.M., Fonseca, G.A., 2001. Sustainable Forest Management: a Review of Conventional Wisdom. Conserva tion International, Advances in Applied Biodiversity Science, No 3. Washington, DC. Rudel, T.K., Coomes, O.T., Moran, E., Achard F., Angelsen, A., Xu, J.C., Lambin, E., 2005. Forest transitions: towards a global u nderstanding of land use change. Global Environmental Change-Human and Policy Dimensions 15, 23-31. Sabogal, C., Lentini, M., Pokorny, B., Silva, J. M.C., Zweede, J.C., Verssimo, A., Boscolo, M., 2006. Manejo Florestal Empresarial na Amaz nia Brasileira. CIFO R, Belm, Brazil. Santilli, M., Moutinho, P., Schwartzman, S., Ne pstad, D., Curran, L., Nobre, C., 2005. Tropical deforestation and the Kyoto Protoc ol. Climatic Change 71, 267-276. Scholl, J., 2005c. Environmental Fiscal Reform and National Forest Poli cies: an Overview of Forest Fiscal Systems in 18 Countries. Sche mmel, J.P. and D. Drager (eds.). GTZ Environment and Resource Management Unit, Eschborn, Germany. Schulze, M., 2003. Ecology and Behavior of Nine Timber Tree Species in Par, Brazil: Links between Species Life History and Forest Management and Co nservation. Ph.D. Dissertation, Department of Biology, The Pennsylvania State University. University Park, Pennsylvania. Schulze, M., Grogan, J., Vidal, E., 2007a. Forest certification in Amazonia: standards matter. Oryx. Schulze, M., Grogan, J., Vidal, E., 2007b. T echnical challenges to sustainable forest management in concessions on public lands in the Brazilian Amazon. in press Journal of Sustainable Forestry. Schulze, M., Vidal, E., Grogan, J., Zweede, J.C ., Zarin, D., 2005. Madeiras nobres em perigo: Prticas e leis atuais de manejo florestal no garantem explorao sustentvel. Cincia Hoje 214, 66-69. Silva, J.N.M., Carvalho, J.O.P., Lopes, J.C.A ., Carvalho, M.S.P., 1984. Equaoes de volume para a Floresta Nacional do Tapajs. Boletim de Pesquisa Florestal, 50-63.
189 Silva, J.N.M., De Carvalho, J.O.P., Lopes, J.D.A., De Almeida, B.F., Costa, D.H.M., De Oliveira, L.C., Vanclay, J.K., Skovsgaard, J. P., 1995. Growth and yiel d of a tropical rainforest in the Brazilian Amazon 13 years afte r logging. Forest Ecology and Management 71, 267-274. Sist, P., Ferreira, F.N., 2007. Sustainability of reduced-impact logging in the Eastern Amazon. Forest Ecology and Management 243, 199-209. Sist, P., Fimbel, R., Sheil, D., Nasi, R., Chevallier, M.H., 2003a. Towards sustainable management of mixed dipterocarp forest s of South-east Asia: moving beyond minimum diameter cutting limits. Envir onmental Conservation 30, 364-374. Sist, P., Picard, N., Gourlet-Fleury, S., 2003b. Sust ainable cutting cycle an d yields in a lowland mixed dipterocarp forest of Borneo. A nnals of Forest Science 60, 803-814. Smith, J., Applegate, G., 2004. Could payments for forest carbon contribute to improved tropical forest management? Forest Policy and Economics 6, 153-167. Smith, J., Colan, V., Sabogal, C., Snook, L., 2006. Why policy reforms fail to improve logging practices: The role of governance and norms in Peru. Forest Policy and Economics 8, 458-469. Snook, L.K., 1996. Catastrophic disturbance, lo gging and the ecology of mahogany (Swietenia macrophylla King): Grounds for listing a ma jor tropical timber species in CITES. Botanical Journal of the Li nnean Society 122, 35-46. Souza Jr., C., Verssimo, A., Micol, L., Guimar es, S., 2006. Boletim Transparncia Florestal do Estado de Mato Grosso. Nos. 1-4. IMAZON, Belm, Brazil. Stone, S.W., 1998a. Evolution of the timber indus try along an aging front ier: The case of Paragominas (1990-95). World Development 26, 433-448. Stone, S.W., 1998b. Using a geographic information system for applied policy analysis: the case of logging in the Eastern Amazon. Ecological Economics 27, 43-61. Sun, C., 1997. An Economic Analysis of Tropi cal Timber Concessions. Ph.D. Dissertation, Department of Agricultural Economics, University of Illinois. Urbana-Champaign, Illinois. Sutinen, J.G., Anderson, P., 1985. The economics of fisheries law-enforcement. Land Economics 61, 387-397. Uhl, C., Barreto, P., Verssimo, A., Vidal, E., Amaral, P., Barros, A.C., Souza Jr., C., Johns, Gerwing, J.J., 1997. Natural resource manage ment in the Brazilian Amazon. Bioscience 47, 160-169. Valle, D., Phillips, P.D., Vidal, E., Schultz, M ., Grogan, J., Sales, M., Van Gardingen, P.R., 2007. Adaptation of a spatially explicit tree-based growth an d yield model and long term
190 comparison between reduced impact and c onventional logging in Eastern Amaznia, Brazil. Forest Ecology and Management 243, 187-198. Valle, D., Schulze, M., Vidal, E., Grogan, J., Sale s, M., 2006. Identifying bias in stand-level growth and yield estimations: A case study in eastern Brazilian Amazonia. Forest Ecology and Management 236, 127-135. Van Gardingen, P.R., Valle, D., Thompson, I., 2006. Evaluation of yield regulation options for primary forest in Tapajos National Forest, Brazil. Forest Ecolo gy and Management 231, 184-195. Vanclay, J.K., 2001. Modelling Forest Growth and Yield: Applications to Mixed Tropical Forests. CABI Publishing, Wallingford, UK. Verssimo, A., Celentano, D., Souza Jr., C., Sa lomo, R., 2006. Zoneam ento de reas para Manejo Florestal no Par. 8. IMAZON, O Estado da Amaznia. Verssimo, A., Souza Jr., C., Amaral, P., 2000. Id entificao de reas com Potencial para a Criao de Florestas Nacionais na Amaznia Le gal. Brazilian Minist ry of Environment, Braslia, Brazil. Verssimo, A., Souza Jr., C., Stone, S.W., Uhl, C., 1998. Zoning of Timber Extraction in the Brazilian Amazon. Conservation Biology 12, 128-136. Verssimo, A., Barreto, P., 2004. National Forests in the Brazilian Amazo n: Opportunities and Challenges. P. 31-40 in Working Forests in the Tropics: Conservation through Sustainable Management, Zarin, D., J.R.R. Alavalapati, F.E. Putz, and M. Schmink (eds.), Columbia University Press, New York. Verssimo, A., Cochrane, M.A., Souza Jr., C., 20 02a. National forests in the Amazon. Science 297, 1478. Verssimo, A., Lima, E., Lentini, M., 2002b. P los Madeireiros do Estado do Par. IMAZON, Belm, Brazil. Vidal, E., Johns, J., Gerwing, J.J., Barreto, P., Uhl, C., 1997. Vine management for reducedimpact logging in eastern Amazonia. Fore st Ecology and Management 98, 105-114. Vidal, E.J., 2004. Dinmica de Florestas Ma nejadas e sob Explorao Convencional na Amaznia Oriental. Ph.D. Dissertation, So Carl os School of Engineering, University of So Paulo. So Paulo, Brazil. Vincent, J.R., 1990. Rent capture and the feasibility of tropical forest management. Land Economics 66, 212-223. Vincent, J.R., 1993. Managing tropical fore sts comment. Land Economics 69, 313-318.
191 Wadsworth, F.H., Zweede, J.C., 2006. Liberati on: Acceptable production of tropical forest timber. Forest Ecology and Management 233, 45-51. Walters, B.B., Sabogal, C., Snook, L.K., de Almeid a, E., 2005. Constraints and opportunities for better silvicultural practice in tropical fore stry: an interdisciplinary approach. Forest Ecology and Management 209, 3-18. Winston, W.L., 1991. Operations Research: Applications and Algorithms. PWS-Kent Publishing Company, Boston. Wunder, S., 2006. Are direct payments for enviro nmental services spelling doom for sustainable forest management in the tropics? Ecology and Society 11. Wunder, S., 2007. The efficiency of payments for environmental services in tropical conservation. Conservation Biology 21, 48-58. Youngs, R.L., Hammett, A.L., 2001. Diversity, produc tivity, profitability, sustainability, and the tao of underutilized species. Forest Products Journal 51, 29-35. Zarin, D.J., Schulze, M., Vidal, E., Lentin i, M., 2007. Beyond reaping the first harvest: Management objectives for timber producti on in the Brazilian Amazon. In press Conservation Biology.
192 BIOGRAPHICAL SKETCH Alexander Macpherson was born in Indianapolis, Indiana, bu t has lived in a variety of places since growing up and finishing high school in Plymouth, Indiana. He earned a Bachelor of Arts in Political Science and English at North Carolina State University in 1993. After earning a Master of Public and In ternational Affairs in Economic and Social Development at the University of Pittsburgh in 1995, Alexander worked at Carnegie Mellon University for six years as a planning analyst and researcher. After a pr omotion to senior analys t and, then, assistant director, Alexander returned to school to earn a Master of Science in Agricultural and Applied Economics from the University of Wisconsin in 2004. In 2004, he was awarded a National Science Foundation Integrated Graduate Education and Resear ch Traineeship to pursue his doctoral studies in Forest Resource Economics at the University of Florida. Alexander has been married to his wife Natalie for 10 years. They have two incredible boys, seven-year-old Ewan and five-year-old Lucas. After Alexa nder graduates, the Macphersons plan to return to North Caro lina to be close to family and friends.