A imprecisão das estimativas de carbono estocado em florestas naturais no ciclo global de carbono vem criando uma demanda de desenvolvimento e padronização de métodos indiretos para modelagem deste ciclo e de emissões de CO2 provenientes de mudanças de uso da terra e florestas. O trabalho teve como objetivo estabelecer as relações empíricas existentes entre a biomassa e o estoque de carbono de uma Floresta Ombrófila Mista (FOM) e os dados ópticos provenientes de sensores remotos de média resolução espacial (ASTER, LiSSIII e TM) por meio de análise de regressão. Além disso, criou-se um cenário hipotético de Redução de Emissões por Desmatamento, Degradação Florestal e Aumento de Estoque de Carbono (REDD+). O estudo foi desenvolvido na Estação Experimental de São João do Triunfo, no estado do Paraná. As equações de regressão envolveram como variáveis dependentes (y): a biomassa e o carbono florestal, obtidos indiretamente do inventário florestal contínuo do Programa de Pesquisas Ecológicas de Longa Duração (PELD), e como variáveis independentes (x) as bandas espectrais e os índices de vegetação (IV). O tratamento estatístico envolveu a análise da matriz de correlação (r) entre as variáveis x e y; a análise de regressão linear simples, não linear e múltipla, com as seguintes estatísticas: R², R²aj., Syx, Syx% e dispersão dos resíduos, Por fim, elaboraram-se mapas temáticos para estas variáveis biofísicas. Como as correlações (r) entre as variáveis biofísicas e espectrais do sensor ASTER (15m) foram baixas, a imagem foi degradada para 30m e 45m. Na resolução de 30m, o uso dos dados ASTER foi superior ao seu uso na resolução original. Não houve diferenças significativas nos valores de r entre o uso das bandas ou dos IVs para predizer as variáveis biofísicas. Regressões lineares simples se mostraram mais adequadas do que as regressões não lineares (exponenciais e logarítmicas) e múltiplas para estimar as variáveis biofísicas, apresentando erros inferiores aos estabelecidos nas campanhas de inventários tradicionais (α < 5%). Os mapas gerados a partir do sensor ASTER 30m foram mais fidedignos ao retratar a distribuição espacial destas variáveis na área de estudo devido à alta correspondência destes com os valores observados no inventário (PELD). Assim, a equação de regressão de carbono florestal a partir do ASTER foi usada na criação do projeto REDD+. A estimativa de biomassa e de carbono florestal da FOM mediante uso de dados de sensores ópticos foi adequada, com possibilidades de ser expandida para extensas áreas. A metodologia, portanto, se mostrou apropriada para ao monitoramento, relatório e verificação de estoques de carbono em florestas. / The imprecision of the estimates of carbon stock in natural forests in the global carbon cycle has created a demand for development and standardization of indirect methods for modeling this cycle and CO2 emissions from land use change and forestry. The work had as objective to establish empirical relationships between biomass and carbon stock of an Araucaria Forest (FOM) and medium spatial resolution remote sensing data (ASTER, and LiSSIII TM) through regression analysis. In addition, we created a hypothetical scenario of Reducing Emissions from Deforestation and Forest Degradation and Enhanced Carbon Stocks (REDD+). The study was developed at the Experimental Station of São João do Triunfo, state of Paraná. The regression analysis involved the forest biomass and forest carbon obtained from continuous forest inventory of the Long Term Ecological Research Program (LTER) as dependent variables (y) and spectral bands and vegetation indices (VIs) as independent variables (x). The statistical analysis comprised correlation analysis (r) between the variables x and y; regression analysis from linear, nonlinear and multiple regressions with the following statistics: R², R²adj, Syx, Syx% and residual dispersion. Furthermore thematic maps were made. Correlations between the biophysical variables and the spectral ASTER data were weak therefore ASTER was scaling up to 30m and 45m. The resolution of 30m, using ASTER data was higher than its use in the original resolution. There were not significant differences in r values between use of bands or VIs to predict the biophysical variables. Linear regressions were more suitable than nonlinear regressions (exponential and logarithmic) and multiple to estimate the biophysical variables, with errors lower than established in traditional inventories campaigns (α <5%). Maps generated from ASTER 30m were more reliable in portraying the spatial distribution of these variables in the study area due to the high correlation of these with the values observed in the inventory (LTER). Thus, the forest carbon equation from ASTER data was used in the creation of REDD+. The estimated biomass and forest carbon by using optical sensors data was adequate, with possibilities to be expanded to large areas. The methodology thus proved suitable for the monitoring, reporting and verification of carbon stocks in forests.

Climate change is a daunting problem that results in actions-interactions from a number of actors in complex global systems, which require multi-level governance and a myriad of national policies. Academics and policy makers alike have been grappling with how to devise effective strategies on the international coordination of climate change policies. It is challenging because climate change problems involve actors with different positions, interests and motivation to cooperate due to the risks involved, the uncertainty and the high costs of adaptation and mitigation. Deforestation is the second largest source of GHG emissions. Success in this area can have a large impact on mitigation. This study focuses on the case of REDD+, a large scale governance experiment in climate finance and a promising cost-effective mitigation mechanism to motivate developing countries to implement policy approaches to reducing emissions from deforestation and forest degradation. REDD+ is considered a breakthrough mechanism in international cooperation under the UNFCCC regime as it was designed to be performance-based. Brazil is the world\'s largest recipient and has the most important REDD+ experiment - the Amazon Fund. The lack of developed theory in this domain led to the use of grounded theory methodology to understand the REDD+ governance process in Brazil. The \'REDD+ Governance Theoretical Framework\' emerged from the data. It is a substantive theory formed by seven major categories (Governance, Strategy, Financing, Implementation, Participation of stakeholders, Joint action and Collective learning) that are related to each other and explain the phenomenon. This study suggests that the lack of institutional arrangements to stimulate collective learning and incorporate lessons learned from the ground experience has been a major constraint on improving its governance in Brazil. Improving the effectiveness of the policy cycle may depend upon the establishment of specific arrangements focused on peer review processes involving lower-level entities responsible for implementation and experts from civil society. The establishment of a recursive learning system could solve certain policy coordination problems and create new opportunities to improve the effectiveness of the REDD+ governance process and implementation. This theory adds to the limited body of literature in the field by extending the knowledge on climate finance, stimulating discussion, and creating opportunities for further research and theoretical advances. The theoretical framework and lessons learned in Brazil from success and failure can help other developing countries to implement a national REDD+ strategy, system or regime. The theory can contribute to the international debate on the principles of good governance in official development assistance and aid effectiveness. This study provides an opportunity for policy makers and practitioners to learn about the challenges and constraints faced by Brazil when implementing an unprecedented results-based mechanism focused on mitigation. / A mudança climática é um problema desafiador resultante de ações e interações entre diversos atores em sistemas globais complexos, o que demanda governança em vários níveis e uma miríade de políticas nacionais. Acadêmicos e policy makers vêm se desafiando sobre como elaborar estratégias eficazes na coordenação internacional das políticas em mudança climática. É desafiador porque os problemas relacionados à mudança do clima envolvem atores com diferentes posições, interesses e motivação para cooperar, já que existem riscos envolvidos, alto nível de incerteza e custos de adaptação e mitigação. O desmatamento é a segunda maior fonte de emissões de gases causadores do efeito estufa. Sucesso nesta área pode ter um grande impacto em mitigação. Este estudo enfoca no caso de REDD+, um experimento de governança no financiamento do clima e um promissor mecanismo de mitigação com baixo custo para incentivar os países em desenvolvimento a implementar abordagens políticas que reduzam emissões oriundas do desmatamento e da degradação florestal. O REDD + é considerado um mecanismo inovador em acordos de cooperação internacional sob o regime da UNFCCC, pois foi idealizado para ser baseado em desempenho. O Brasil é o maior receptor do mundo e tem o mais importante experimento de REDD+ - o Fundo Amazônia. A ausência de teorias desenvolvidas nessa área levou ao uso da metodologia grounded theory para compreender o processo de governança de REDD+ no Brasil. A partir dos dados primários foi desenvolvido o \'REDD+ Governance Theoretical Framework\'. Trata-se de uma teoria substantiva formada por sete categorias (Governança, Estratégia, Financiamento, Implementação, Participação de stakeholders, Ação coletiva e Aprendizagem coletiva) que se relacionam e explicam o fenômeno. Este estudo sugere que a ausência de arranjos institucionais para estimular a aprendizagem coletiva e incorporar as lições aprendidas durante a implementação tem sido um grande obstáculo para melhorar a governança de REDD+ no Brasil. Melhorar a efetividade do ciclo político pode depender do estabelecimento de arranjos específicos com foco em processos de revisão por pares que envolvam entidades responsáveis pela implementação e especialistas da sociedade civil. O estabelecimento de um sistema de aprendizagem recursiva poderia facilitar a resolução de alguns problemas de coordenação política e criar novas oportunidades para aprimorar o processo de governança de REDD+. Esta teoria contribui para a construção do conhecimento científico focado no financiamento do clima, estimula a discussão, sugere oportunidades para novas pesquisas e avanços teóricos. O framework teórico pode ajudar outros países em desenvolvimento a implementar estratégia, sistema ou regime nacional de REDD+. As lições aprendidas no Brasil, baseadas no sucesso e fracasso, podem ser absorvidas por outros países em desenvolvimento. A teoria contribui para o debate internacional sobre os princípios da boa governança nos acordos de cooperação internacional e na eficácia da ajuda financeira. Este estudo oferece uma oportunidade para que os policy makers e os profissionais aprendam sobre os desafios e obstáculos enfrentados pelo Brasil ao implementar um inovador mecanismo de financiamento do clima baseado em resultados.

A imprecisão das estimativas de carbono estocado em florestas naturais no ciclo global de carbono vem criando uma demanda de desenvolvimento e padronização de métodos indiretos para modelagem deste ciclo e de emissões de CO2 provenientes de mudanças de uso da terra e florestas. O trabalho teve como objetivo estabelecer as relações empíricas existentes entre a biomassa e o estoque de carbono de uma Floresta Ombrófila Mista (FOM) e os dados ópticos provenientes de sensores remotos de média resolução espacial (ASTER, LiSSIII e TM) por meio de análise de regressão. Além disso, criou-se um cenário hipotético de Redução de Emissões por Desmatamento, Degradação Florestal e Aumento de Estoque de Carbono (REDD+). O estudo foi desenvolvido na Estação Experimental de São João do Triunfo, no estado do Paraná. As equações de regressão envolveram como variáveis dependentes (y): a biomassa e o carbono florestal, obtidos indiretamente do inventário florestal contínuo do Programa de Pesquisas Ecológicas de Longa Duração (PELD), e como variáveis independentes (x) as bandas espectrais e os índices de vegetação (IV). O tratamento estatístico envolveu a análise da matriz de correlação (r) entre as variáveis x e y; a análise de regressão linear simples, não linear e múltipla, com as seguintes estatísticas: R², R²aj., Syx, Syx% e dispersão dos resíduos, Por fim, elaboraram-se mapas temáticos para estas variáveis biofísicas. Como as correlações (r) entre as variáveis biofísicas e espectrais do sensor ASTER (15m) foram baixas, a imagem foi degradada para 30m e 45m. Na resolução de 30m, o uso dos dados ASTER foi superior ao seu uso na resolução original. Não houve diferenças significativas nos valores de r entre o uso das bandas ou dos IVs para predizer as variáveis biofísicas. Regressões lineares simples se mostraram mais adequadas do que as regressões não lineares (exponenciais e logarítmicas) e múltiplas para estimar as variáveis biofísicas, apresentando erros inferiores aos estabelecidos nas campanhas de inventários tradicionais (α < 5%). Os mapas gerados a partir do sensor ASTER 30m foram mais fidedignos ao retratar a distribuição espacial destas variáveis na área de estudo devido à alta correspondência destes com os valores observados no inventário (PELD). Assim, a equação de regressão de carbono florestal a partir do ASTER foi usada na criação do projeto REDD+. A estimativa de biomassa e de carbono florestal da FOM mediante uso de dados de sensores ópticos foi adequada, com possibilidades de ser expandida para extensas áreas. A metodologia, portanto, se mostrou apropriada para ao monitoramento, relatório e verificação de estoques de carbono em florestas. / The imprecision of the estimates of carbon stock in natural forests in the global carbon cycle has created a demand for development and standardization of indirect methods for modeling this cycle and CO2 emissions from land use change and forestry. The work had as objective to establish empirical relationships between biomass and carbon stock of an Araucaria Forest (FOM) and medium spatial resolution remote sensing data (ASTER, and LiSSIII TM) through regression analysis. In addition, we created a hypothetical scenario of Reducing Emissions from Deforestation and Forest Degradation and Enhanced Carbon Stocks (REDD+). The study was developed at the Experimental Station of São João do Triunfo, state of Paraná. The regression analysis involved the forest biomass and forest carbon obtained from continuous forest inventory of the Long Term Ecological Research Program (LTER) as dependent variables (y) and spectral bands and vegetation indices (VIs) as independent variables (x). The statistical analysis comprised correlation analysis (r) between the variables x and y; regression analysis from linear, nonlinear and multiple regressions with the following statistics: R², R²adj, Syx, Syx% and residual dispersion. Furthermore thematic maps were made. Correlations between the biophysical variables and the spectral ASTER data were weak therefore ASTER was scaling up to 30m and 45m. The resolution of 30m, using ASTER data was higher than its use in the original resolution. There were not significant differences in r values between use of bands or VIs to predict the biophysical variables. Linear regressions were more suitable than nonlinear regressions (exponential and logarithmic) and multiple to estimate the biophysical variables, with errors lower than established in traditional inventories campaigns (α <5%). Maps generated from ASTER 30m were more reliable in portraying the spatial distribution of these variables in the study area due to the high correlation of these with the values observed in the inventory (LTER). Thus, the forest carbon equation from ASTER data was used in the creation of REDD+. The estimated biomass and forest carbon by using optical sensors data was adequate, with possibilities to be expanded to large areas. The methodology thus proved suitable for the monitoring, reporting and verification of carbon stocks in forests.

Reducing Emissions from Deforestation and Forest Degradatin (REDD+) is a market-based approachto address tropical deforestation as a key driver of anthropogenic climate change. In Tanzania, participatory forest management (PFM) was used as a vehicle for the institutionalisation of REDD+and implementation of pilot initiatives. With the lens of political ecology, this thesis analyses the effects of the REDD+ pilot project ‘Advancing REDD+ in the Kolo Hills Forests’ (ARKFor) inKondoa District, Tanzania, on structures of access and use of forest resources for local communities. This analysis is done by using qualitative interviews with villagers living within the REDD+ project area and government actors involved in forest management as well as textual analysis of a PFMagreement and community bye-laws established within the ARKFor project. The findings suggest that REDD+ pilot activities were planned without real participation by local communities and failed to take complex conservation histories and underlying power structures into account. Community access rights were not legally secured which resulted in processes of re-centralisation of forest managementand ‘green grabbing’ after the conclusion of project activities in 2014. The study underscores that successful forest management decentralisation needs to be based on localised, longer-term adaptive processes which clash with the globally driven, neoliberal conservation logic of REDD+.

Redd distribution, redd density, and physical habitat were used to explain the distribution and abundance of young-of-the-year (YOY) brown trout (Salmo trutta) and rainbow trout (Oncorhynchus mykiss) in the Green River, Utah. The importance of variables at both a microhabitat and macrohabitat scale were assessed using stepwise regression analysis. Availability of cover (rock and vegetation) and proximity to spawning sites were the most important variables used to explain the distribution and abundance of YOY brown trout and rainbow trout. In addition, YOY brown trout and rainbow trout occupied specific microhabitats and showed patterns of use for particular depths, substrates, and cover. However, the importance of variables differed by year, indicating that variables other than those measured were also influencing their distribution and abundance. The results of my study indicate that variables at both a microhabitat and macrohabitat scale may be important in explaining the distribution and abundance of YOY trout in streams. Therefore, to better understand the habitat requirements of stream fishes and to better explain their distribution and abundance in streams future, studies may need to incorporate both physical habitat variables and variables affecting recruitment.

No description available.

This thesis analyzes the Suruí Forest Carbon Project in the context of settler colonialism. By exploring the three core principles of settler colonialism as outlined by settler colonial scholar Patrick Wolfe: access to land, elimination of the native, and the understanding that settler colonialism is a structure and not an event, I will demonstrate how each one of the three principles helped contribute to creating the context within which the Suruí Forest Carbon Project was situated. By taking this approach, I will be able to demonstrate the limits and possibilities of the project for the Suruí indigenous peoples. This analysis will allow me to present the challenges and contradictions associated with implementing REDD+ carbon credit projects in settler states such as Brazil and how, due to settler colonialism's structural limitations, these types of projects could be a possibility of providing some agency for indigenous peoples trying to find ways to assert their autonomy. The Suruí Forest Carbon Project was the first and still one of the only examples of an indigenous-led carbon emissions reduction project operating through the sale of carbon credits. During the first five years the project was operational, it drastically helped reduce deforestation levels within the Suruí's territory, leading many to deem the project a success. However, in 2015 and 2016, following the discovery of gold and diamonds on the Suruí's territory, the project's sight was eventually overrun by garimpeiros (small-scale gold miners), and in 2018 the project was suspended, leading some to consider it a failure. Therefore, I will present some of the challenges that arise when neoliberal conservation efforts, such as carbon credit projects, struggle to address factors outside their initial control, in this case, settler colonialism. Also, by analyzing the different components going into the project's creation, implementation, and suspension, I will present how carbon credit projects working directly with indigenous peoples can successfully halt deforestation for limited periods. But how settler colonialism makes these groups of people and their land vulnerable, which can help contribute to projects being undermined. Through my analysis, I will help demonstrate some factors that impact these types of projects' longevity and some things that would need to be implemented in the future to succeed in the long term. / Master of Arts / This thesis analyzes the Suruí Forest Carbon Project in the context of settler colonialism. My understanding of settler colonialism comes from settler colonial scholar Patrick Wolfe who believes that this specific type of colonialism has three core principles that help distinguish it from other colonial types and explain why anti-indigenous logics can continue. The three principles are access to land, the elimination of the native, and the understanding that settler colonialism is a structure and not an event. These three principles will serve as the core framework for my analysis. The Suruí Forest Carbon Project was the world's first indigenous-led carbon emissions reduction project operated by the indigenous peoples selling REDD+ carbon credits to buyers in order to achieve finances. The project occurred on the Suruí people's territory within the Sete de Setembro Indigenous Land, comprising a 250,000-ha site in the Amazon's "arc of deforestation" bordering the Brazilian states of Rondônia and Mato Grosso. The project was implemented on June 9, 2009, and in 2012 received its validation to sell carbon credits under the Verified Carbon Standard (VCS). Between 2009 and 2014, the project drastically helped limit the deforestation occurring within the project's site, causing many to deem it a success. However, trouble began in 2015 and 2016 following the discovery of gold and diamonds on the Suruí's territory. Shortly after this discovery, the territory began to be infiltrated by garimpeiros (small-scale gold miners), which led to increased levels of deforestation on the project's site. In 2018, the project could no longer meet the standards it needed to maintain to sell the credits and was suspended indefinitely. Therefore, based on my understanding of settler colonialism's three core principles, I will analyze the limits and possibilities of the project for the Suruí indigenous peoples to present how all three principles played a hand in creating the conditions within which the Suruí Forest Carbon Project was situated and how that impacted the indigenous peoples involved in the project ability to have agency over their forests.

Reducing Emissions from Deforestation and forest Degradation (REDD+) is a global carbon trading program intent on mitigating or reversing carbon emissions from forestry in the global south. REDD+ was negotiated at the 2005 United Nations Framework Convention on Climate Change (UNFCCC) and is coordinated by the Forest Carbon Partnership Facility (FCPF), administered by the World Bank Group. In this project, I explore REDD+ activity in Costa Rica, drawing on Michel Foucault's concept of governmentality. Costa Rica became a participant in the Forest Carbon Partnership Facility in July of 2008. Since then, indigenous peoples throughout the country have contested the program. This project is a single-case study of the Bribri contestation of REDD+ schemes, one of the larger indigenous communities in Costa Rica. Bribri argue that REDD+ disrespects their worldview and further endangers their local rights to land and forestry. This project argues that REDD+ and Bribri have different perceptions of nature, enabling disagreement on REDD+ goals. Whereas REDD+ perceives nature as commodifiable for the purposes of neoliberal climate policies, Bribri express a spiritual, harmonious relationship with nature. I conclude by noting that REDD+ can pose negative implications for indigenous life and culture. This is not only because REDD+ draws external and domestic actors to land and forestry for incentive-based purposes. But also because REDD+ defines 'rightful behavior' among forestry resources, challenging indigenous conceptions of environmental management. However, the Bribri are resisting REDD+ imposition and, particularly, the program's external governing of indigenous behavior amongst forests. / Master of Arts

Réduire la déforestation dans les pays tropicaux est un des principaux défis pour la communauté internationale dans le cadre du processus de négociations de la Convention Cadre des Nations Unies sur le Changement Climatique (CCNUCC). En effet, la déforestation est la seconde source d'émissions de gaz à effets de serre, juste derrière les émissions industrielles. Depuis 2005, un nouvel instrument international pour réduire les émissions de carbone liées à la déforestation tropicale est en négociation à la CCNUCC. Ce mécanisme, appelé REDD+ (Réduction des Emissions liées à la Déforestation et Dégradation des forêts) repose sur un système de compensation financière des pays en développement pour leurs efforts en termes de déforestation évitée. Cependant, la mise en œuvre du mécanisme REDD+ à l'échelle nationale et internationale soulève de nombreux problèmes méthodologiques et rencontre de nombreux obstacles. Le but de la thèse est double. Dans une première partie, une description et une analyse du mécanisme REDD+ est réalisée. Dans une deuxième partie, de nouvelles perspectives concernant le design du mécanisme REDD+ et sur sa mise en œuvre sont offertes, en se basant sur trois essais rédigés en format article. Le premier essai propose un modèle de théorie des jeux reflétant le processus de négociation Nord-Sud du mécanisme REDD. Il étudie les conditions régissant le partage de fonds entre les pays en développement et leurs impacts sur l'efficacité du système d'incitations. Le deuxième essai utilise un modèle en économétrie de panel pour différencier des comportements nationaux de déforestation selon la dotation relative en forêts de chaque pays. Le troisième essai s'intéresse à la mise en œuvre du mécanisme REDD+, en comparant les résultats de deux programmes de paiement pour services environnementaux pour deux types de gouvernements. Le modèle développé dans cet essai est ensuite testé dans le contexte de la déforestation en Indonésie, grâce à une base de données fournie par l'ONG Conservation International. / Curbing deforestation in tropical countries is one of the main current challenges for international community in the United Nations Framework Convention on Climate Change. Indeed, deforestation is the second leading cause of greenhouse gas emissions just behind industrial emissions. Since 2005, a new instrument to slow down CO2 emissions from tropical deforestation is under negotiations at the UNFCCC. This mechanism, called REDD+ (for Reducing Emissions from Deforestation and forest Degradation) is supported by a simple principle: it consists to reward developing countries for their efforts to avoid deforestation. However, the national and international implementations of REDD+ raise lot of methodological questions and meet several hurdles. The aims of the thesis are twofold. First, it proposes a description and an analysis of the REDD+ mechanism. Second, it is composed by three essays, which raise some questions about REDD+ design and implementation, in order to offer new perspectives on this mechanism. The first essay develops a game-theoretic bargaining model, simulating the on-going negotiation process over the REDD+ mechanism. It shows that the conditions under which developing countries are left to bargain over the allocation of the global forest fund may lead to an ineffective system of incentives. The second essay used a panel data analysis to reveal contrasted deforestation behaviors of tropical countries according to their relative endowment in forest cover. The aim of the third essay offered an illustration of REDD+ implementation, comparing the outcomes in terms of avoided deforestation and utility of two payments for environmental services designs for two types of governments. The model developed in this article is applied in the Indonesian context of deforestation, thanks to a database supplied by the NGO Conservation International.

Tropical forests store more than half of the world’s forest carbon and are particularly threatened by deforestation and degradation processes, which together represent the second largest source of anthropogenic CO2 emissions. Consequently, tropical forests are the focus of international climate policies (i.e. Reducing emissions from deforestation and forest degradation, REDD) aiming at reducing forest-related CO2 emissions. The REDD initiative lies on our ability to map forest carbon stocks (i.e. spatial dynamics) and to detect deforestation and degradations (i.e. temporal dynamics) at large spatial scales (e.g. national, forested basin), with accuracy and precision. Remote-sensing is as a key tool for this purpose, but numerous sources of error along the carbon mapping chain makes meeting REDD criteria an outstanding challenge. In the present thesis, we assessed carbon (quantified through aboveground biomass, AGB) estimation error at the tree- and plot-level using a widely used pantropical AGB model, and at the landscape-level using a remote sensing method based on canopy texture features from very high resolution (VHR) optical data. Our objective was to better understand and reduce AGB estimation error at each level using information on large canopy tree structure, distribution and spatial organization. Although large trees disproportionally contributed to forest carbon stock, they are under-represented in destructive datasets and subject to an under-estimation bias with the pantropical AGB model. We destructively sampled 77 very large tropical trees and assembled a large (pantropical) dataset to study how variation in tree form (through crown sizes and crown mass ratio) contributed to this error pattern. We showed that the source of bias in the pantropical model was a systematic increase in the proportion of tree mass allocated to the crown in canopy trees. An alternative AGB model accounting for this phenomenon was proposed. We also propagated the AGB model bias at the plot-level and showed that the interaction between forest structure and model bias, although often overlooked, might in fact be substantial. We further analyzed the structural properties of crown branching networks in light of the assumptions and predictions of the Metabolic Theory of Ecology, which supports the power-form of the pantropical AGB model. Important deviations were observed, notably from Leonardo’s rule (i.e. the principle of area conservation), which, all else being equal, could support the higher proportion of mass in large tree crowns. A second part of the thesis dealt with the extrapolation of field-plot AGB via canopy texture features of VHR optical data. A major barrier for the development of a broad-scale forest carbon monitoring method based on canopy texture is that relationships between canopy texture and stand structure parameters (including AGB) vary among forest types and regions of the world. We investigated this discrepancy using a simulation approach: virtual canopy scenes were generated for 279 1-ha plots distributed on contrasted forest types across the tropics. We showed that complementing FOTO texture with additional descriptors of forest structure, notably on canopy openness (from a lacunarity analysis) and tree slenderness (from a bioclimatic proxy) allows developing a stable inversion frame for forest AGB at large scale. Although the approach we proposed requires further empirical validation, a first case study on a forests mosaic in the Congo basin gave promising results. Overall, this work increased our understanding of mechanisms behind AGB estimation errors at the tree-, plot- and landscape-level. It stresses the need to better account for variation patterns in tree structure (e.g. ontogenetic pattern of carbon allocation) and forest structural organization (across forest types, under different environmental conditions) to improve general AGB models, and in fine our ability to accurately map forest AGB at large scale.