Tropische Regenwälder sind durch anthropogene Aktivitäten gefährdet und wurden als Kippelement identifiziert. Ein Kippen in einen neuen Zustand könnte tiefgreifende Auswirkungen auf das globale Klima haben, sobald die Vegetation von einem bewaldeten in einen Savannen- oder Graslandzustand übergegangen ist. Waldbrände können die Grenze zwischen Savanne und Wald verschieben und somit das dynamische Gleichgewicht zwischen diesen beiden möglichen Vegetationszuständen unter sich änderndem Klima beeinträchtigen. In der vorliegenden Doktorarbeit wurde ein neues Erdsystemmodell entwickelt und angewendet, um explizit die Auswirkungen von Feuer, Klimawandel und Landnutzung auf eine potenzielle tropische Hysterese abzuschätzen. In den ersten beiden Teilen der Arbeit wurde das Vegetationsmodell LPJmL vor allem in Hinblick auf Feuersimulation verbessert und anschließend biophysikalisch an das Erdystemmodell CM2Mc gekoppelt. Im dritten Teil dieser Arbeit wurde das resultierende Modell CM2Mc-LPJmL schließlich angewendet, um wichtige biophysikalische Feuer-Vegetations-Klima-Rückkopplungen und einen potentiellen Kipppunkt bzw. eine Hysterese der tropischen Wälder zu untersuchen. Die Ergebnisse der Experimente zeigten, dass eine alleinige Klima Störung nicht zu einem großflächigen Kipppunkt tropischer Wälder führt. Andererseits führte die vollständige Entwaldung bei einer erhöhten CO2-Konzentration von über 450 ppm und die Wirkung von Waldbränden zu einer Verschiebung großer Teile des Amazonas Regenwaldes in einen stabilen Graslandzustand. Die Leistung dieser Arbeit ist die Entwicklung eines neuen Erdsystemmodells, das die Vorteile des umfassenden dynamischen Vegetationsmodells LPJmL und eines prozessbasierten Feuermodells mit dem geringen Rechenaufwand von CM2Mc verbindet. Diese Doktorarbeit untersuchte zum ersten Mal den expliziten Einfluss von Feuer auf tropische Kipppunkte und auf eine mögliche vegetative Erholung in einem umfassenden feuerfähigen Erdsystemmodell. / Tropical rain forests are endangered by anthropogenic activities and are recognized as one of the terrestrial tipping elements. An ecosystem regime change to a new state could have profound impacts on the global climate, once the biome has transitioned from a forest into a savanna or grassland state. Fire could potentially shift the savanna-forest boundary and hence impact the dynamical equilibrium between these two possible vegetation states under a changing climate. In this thesis, a new Earth system model was developed and applied to explicitly estimate the impact of fire, climate change and land-use on a potential tropical tipping point and hysteresis. The first part of this thesis describes the improvement of simulating fire within the dynamic global vegetation model (DGVM) LPJmL (Lund-Potsdam-Jena-managed-Land). In the second part, the improved LPJmL model was biophysically coupled to the Earth system model CM2Mc, which involved numerous changes in the original LPJmL model. In the third part of this thesis, the resulting model CM2Mc-LPJmL was finally applied to investigate important biophysical fire-vegetation-climate feedbacks and a potential tipping point and hysteresis of tropical forests. The results of the modeling experiments indicated that a sole climate disturbance does not lead to a large-scale tipping of tropical forests into a savanna or grassland state. On the other hand, complete deforestation alongside elevated CO2 above 450 ppm and the impact of fire led to a shift of large parts of the Amazon into a stable grassland state. The contribution of this thesis is the development of a new Earth system model, including the advantages of the comprehensive dynamic vegetation model LPJmL, a process-based fire model and the low computation cost of CM2Mc. This thesis studied for the first time the explicit impact of fire on tropical tipping points and a possible vegetation recovery in a comprehensive fire-enabled Earth system model.

The Arctic region shows a particularly high susceptibility to climate change, which historically manifests in an amplification of the near-surface warming in the Arctic relative to the global mean. This Arctic amplification (AA) has impacts on the climate system also beyond the northern polar regions, which highlights the importance to adequately represent it in numerical models. While state-of-the-art climate models widely agree on the presence of AA, they simulate a large spread in the magnitude of Arctic-amplified warming. This thesis addresses the need to evaluate the performance of global climate models in projecting AA and its most important drivers. For the latter, the focus is on the three amplifying climate feedbacks (ACFs) that largely drive the meridional warming structure leading to AA. The ACFs include the sea-ice-albedo feedback (SIAF), the Planck feedback, and the lapse-rate feedback (LRF). These feedbacks arise from the relevant changes in Arctic sea ice, near-surface temperatures, and the deviation from the near-surface temperature change through the atmosphere, respectively. In the thesis, two observational constraints are presented to narrow the range of climate models of the sixth Coupled Model Intercomparison Project (CMIP6) regarding their projection of AA and the ACFs in both past and future climate. While for the past, the models representation of near-surface processes can often be directly evaluated against observations, it is particularly the LRF that is difficult to constrain as it incorporates the entire atmospheric warming structure. As a consequence, the historical constraint focuses on the LRF, while the future constraint gives a prediction range for the evolution of AA and all three ACFs through the 21st century. The main results are highlighted in the view of the changing atmospheric energy budget (AEB) of the Arctic under anthropogenic climate forcing. The AEB provides a framework to address Arctic climate change at large scales, and further helps to decide on the relevant aspects that provide appropriate metrics for constraining both AA and the ACFs. In other words, the perspective of a changing Arctic AEB highlights important alterations of the energetics under climate change, that further link to changes in climate aspects that partly explain the inter-model spread in simulated AA and the ACFs. The main results of the cumulative thesis are formulated on the basis of three published research papers, papers I, II, and III. Paper I addresses the Arctic AEB which is typically characterised by an equilibrium between net radiative cooling and advective heating, and mostly an absence of convection. This radiative-advective equilibrium (RAE) approximates well the energy budget and thermal structure of the Arctic atmosphere. The main outcome of paper I is that with continuous warming as simulated by CMIP6 models in an idealised setup, a deviation from the RAE increasingly develops, resulting from sea ice retreat and increased ocean-to-atmosphere heat fluxes. These changes are further concomitant with a depletion of the typical surface-based temperature inversion and a decrease in advective heating, which is byword for the convergence of atmospheric energy transport in the Arctic. Since the RAE currently explains much of the basic thermal structure of the Arctic atmosphere, those changes have the potential to further mediate the LRF. Paper II builds on paper I and evaluates the performance of climate models in representing the key aspects of the Arctic LRF in CMIP6 historical simulations that have the best estimates of the transient climate forcings during the observational period. In particular it is found that CMIP6 models that realistically simulate both the lower thermal structure of the atmosphere and the poleward atmospheric energy transport are more trustworthy in informing about the LRF and how much it contributed to Arctic warming during the past few decades. The evaluation is based on observations of surface-based temperature inversions during the year-long Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition, and atmospheric energy transport convergence computations from reanalyses. Paper III expands the constraint approach of paper II and carries out an emergent constraint (EC) on future AA and the ACFs that further elaborates on the physical relationships between the constraining metrics and future climate projections. Previous work has highlighted that parts of the inter-model spread in simulated AA is explained through the spread in contemporaneous sea ice loss across climate models. The thesis confirms this link by showing that CMIP6 models with a stronger climatological sea ice loss project a stronger AA in the future under the assumption of a high emission scenario. By further linking the degree of future ice loss to the current-climate sea ice amount in CMIP6 models, paper III facilitates an EC on the future evolution of AA and the ACFs. In particular, models with a lower contemporary sea ice amount project a larger magnitude of AA by setting the stage for stronger climatological ice loss and near-surface warming, linking to the relevant ACFs. From the corresponding prediction it is evident that AA is expected to continue at a warming rate that is more than twice or three times larger than global-mean warming. Furthermore, the three ACFs continue to contribute to Arctic warming, with the SIAF leading the warming contribution response. Lastly, the consideration of statistically strong and physically plausible relationships across climate models makes the EC a valuable technique to constrain climate model simulations in conjunction with observations. This thesis highlights the potential of combining the advantages of both presented constraints: Using multiple process-relevant aspects instead of one singular metric (paper II), but considering the mechanistic couplings between these metrics and the climate projection of interest (paper III) will improve our model-evaluation techniques and further help guiding the design of future climate simulations.:Summary of the dissertation List of papers Author’s contribution Supervision statement 1 Introduction 2 Research focus 3 The Arctic atmospheric energy budget 3.1 The atmospheric column model 3.2 The annual atmospheric energy budget 4 Arctic amplification and climate feedbacks 4.1 Amplifying climate feedbacks 4.2 A comment on process coupling 5 Methods and data 5.1 Energy budget equations 5.2 Quantifying Arctic amplification and climate feedbacks 5.3 Climate model data 5.3.1 CMIP6 idealised simulations 5.3.2 CMIP6 historical simulations 5.3.3 CMIP6 ssp585 simulations 5.4 Observational constraints 5.4.1 Constraint on historical Arctic lapse-rate feedback 5.4.2 Constraint on future Arctic amplification and relevant climate feedbacks 6 Results 6.1 Paper I - Deviations from the Arctic radiative-advective equilibrium under anthropogenic climate change 6.2 Paper II - Constraining the Arctic lapse-rate feedback during past decades by contemporary observations 6.3 Paper III - Constraining future Arctic amplification and the relevant climate feedbacks based on the recent sea ice climatology 7 Summary and outlook References Lists Acknowledgements Appendix A: Paper I Appendix B: Paper II Appendix C: Paper III

Uma questão que ainda instiga discussões na literatura ecológica é como explicar a coocorrência dinâmica e milenar de formações florestais e campestres sob um mesmo regime climático que tende a favorecer as primeiras, como ocorre atualmente com mosaicos florestacampo no sul do Brasil. A partir de meados do século XX, têm-se evidenciado um fenômeno mundial de avanço de elementos lenhosos sobre áreas abertas. Neste sentido, a modelagem dos processos ecológicos envolvidos na manutenção de ambas as formações numa escala de paisagem permite o esclarecimento dos mecanismos que atuam na manutenção dessa coexistência até o presente e permite prever estados futuros diante dos prognósticos de drásticas alterações climáticas globais já nas próximas décadas. Para tanto, desenvolveu-se um modelo espacialmente explícito (2D-aDGVM) que agrega um Modelo Adaptativo Global de Dinâmica de Vegetação (aDGVM) e ainda inclui heterogeneidades topográficas, propagação do fogo e dispersão de sementes. Este modelo busca satisfazer a necessidade de modelagem mais realista de processos biofísicos, fisiológicos e demográficos na escala de indivíduos e relacionados de forma adaptativa às variações ambientais e aos regimes de distúrbios, ao mesmo tempo que agrega importantes processos ecológicos espaciais, até então pouco ou nada abordados por esse grupo de modelos numa escala de paisagem. Com este modelo, avaliaram-se os efeitos das variações topográficas da radiação solar incidente e destas nos mecanismos de interação (feedbacks) positiva e negativa que surgem daqueles processos na escala de indivíduos e que definem localmente os limites da coexistência entre elementos arbóreos e herbáceos. Ainda, foram analisados os efeitos do aumento da temperatura, precipitação e CO2 atmosférico, desde o período pré-industrial até projeções futuras para as próximas décadas, na performance das diferentes fisiologias envolvidas, bem como no balanço daquelas interações entre as mesmas e, finalmente, na sensibilidade da dinâmica dos mosaicos floresta-campo. Os resultados evidenciaram que, sob o regime climático vigente, uma coexistência relativamente estável entre floresta e campo numa mesma paisagem é mantida por uma alta freqüência de distúrbios, que por sua vez, resulta do forte feedback positivo do acúmulo de biomassa inflamável da vegetação campestre na intensidade do fogo, proporcionado pela condição altamente produtiva do atual clima mesotérmico. Por outro lado, intensificadas pela declividade do terreno, as heterogeneidades espaciais afetaram o balanço dessas interações, interferindo nos padrões espaço-temporais relacionados ao comportamento do fogo e dependentes da densidade de elementos arbóreos. Ainda, tanto esses efeitos observados na escala das manchas de vegetação, como o arranjo espacial inicial das mesmas na paisagem, afetaram as taxas de expansão florestal. Em outras palavras, a manutenção da coexistência de duas formações vegetais constituídas por elementos de inerente assimetria competitiva é possível pela manutenção de uma maior conectividade daquela que propicia o distúrbio, superando a vantagem da outra, que por sua vez é dependente da densidade dos indivíduos. Numa escala de paisagem, isto causa a manutenção de uma baixa conectividade entre as manchas florestais, propiciando sua relativa estabilidade num contexto de dispersão predominante a curtas distâncias. Contudo, embora ambos os sistemas tenham apresentado incremento no crescimento, produtividade e fecundidade, observou-se uma sensibilidade maior no sentido de aumento das taxas de avanço florestal em resposta às projeções climáticas futuras, principalmente nos próximos 90 anos, mesmo na presença do fogo. Isto seria proporcionado pela vantagem fotossintética das árvores-C3 sobre gramíneas-C4 na presença do fogo sob altas concentrações de CO2 atmosférico. Por fim, uma abordagem mais sistêmica dos mosaicos como estados alternativos mostrou ser adequada para o entendimento dos mecanismos que propiciam essa coexistência dinâmica na paisagem. / A longstanding problem in ecology is how to explain the coexistence over thousands of years of forests and natural grasslands under the same climatic regime, which favors the first, such as in forest-grasslands mosaics in South Brazil. Since the middle of the 20th century, a worldwide bush encroachment phenomenon of woody invasion in open vegetation has been threatening this relatively stable coexistence. In this sense, modelling ecological processes that arbitrate the maintenance of both vegetation formations at the landscape scale allows a better understanding of the mechanisms behind the maintenance of this coexistence, as well as predictions of future states under projections of drastic climate change over the next decades. For this, we developed a bidimensional spatial explicit model (2D-aDGVM) that aggregates an adaptive Global Vegetation Model (aDGVM), which includes topographic heterogeneity, fire spread and seed dispersal. The model aims at fulfilling the need for a more realistic representation of biophysical, physiological and demographical processes using an individualbased approach as it adapts these processes to environmental variations and disturbance regimes. In addition, the model includes important spatial ecological processes that have gained less attention by such models adopting a landscape-scale approach. Therefore, we evaluated the effect of topographic variations in incoming solar radiation on positive and on negative feedbacks that rise from those individual-based processes, and which in turns define the limiting thresholds upon which woody and grassy forms coexist. Additionally, the effects of increasing temperature, rainfall and atmospheric CO2 levels on the performance of distinct physiologies (C3-tree and C4-grass) were analyzed, as well as the sensitivity of forestgrassland mosaics to changes in climate from the preindustrial period to the next decades. Results showed that a relatively stable coexistence of forests and grasslands in the same landscape was observed with more frequent fires under the present climatic conditions. This was due to strong positive feedbacks of the huge accumulation of flammable grass biomass on fire intensity promoted by the high productivity of the present mesic conditions. On the other hand, spatio-temporal density dependent processes linked to fire and enhanced by slope at the patch scale, as well as the initial spatial arrangement of vegetation patches affected the rate of forest expansion at the landscape scale. The persistence of coexisting vegetation formations with an inherent asymmetry of competitive interactions was possible when the higher connectivity of the fire-prone patches (grassland) affected negatively the performance of the entire fire-sensitive system (forest). This was possible by overcoming its local densitydependent advantage, or by maintaining it with a low connectivity, which is expected to reduce the rate of coalescence of forest patches in a scenario of predominantly short distance dispersal. Despite the increments in biomass production, stem growth and fecundity that were observed in both grassland and forest, climate change increased the rates of forest expansion over grasslands even in presence of fire, and mainly over the next 90 years. This was attributed to a high photosynthetic advantage of C3-trees over C4-grasses in presence of fire under higher atmospheric CO2 levels. Finally, in the face of the general observed tendency of forest expansion over grasslands, the ancient grasslands have persisted as alternative ecosystem states in forest-grassland mosaics. In this sense, exploring this dynamic coexistence under the concept of alternative stable states have showed to be the most appropriate approach, and the outcomes of this novel perspective may highlight the understanding of the mechanisms behind the long-term coexistence.

Dans les systèmes lotiques, la végétation aquatique se développe en formant des taches générées par des rétroactions échelle-dépendantes. Les plantes modifient l'environnement physique (i.e. organismes ingénieurs), induisant des rétroactions positives dans les taches et négatives à côté, ce qui conduit à la formation de patrons réguliers. Ces rétroactions échelle-dépendantes ne permettent d'expliquer que l'expansion latérale des taches, mais pas leur développement longitudinal. L'objectif était d'étudier les processus qui induisent des rétroactions pour les plantes et les conséquences pour la dynamique des taches. Des mesures de l'hydrodynamique, des caractéristiques des sédiments et de la morphologie des plantes ont été faites in situ le long de taches de longueur croissante. Les résultats ont démontré qu'une longueur minimale est nécessaire pour induire une réduction de la vitesse du courant et une accumulation de sédiments fins dans les taches. L’ensemble conduit à des changements des concentrations en nutriments dans l'eau interstitielle au delà d’une certaine longueur de tache, consistant en une accumulation d'ammonium et une diminution des nitrates. La hauteur des plantes est liée à la longueur de la tache selon un modèle quadratique, suggérant l’existence d’une rétroaction négative au delà d’une longueur seuil, probablement due à la concentration élevée en ammonium qui peut être toxique pour les plantes. Les longueurs au delà desquelles ont lieu des changements des processus biogéochimiques et des rétroactions négatives sont plus faibles dans l’écosystème avec le niveau de nutriments le plus élevé. Enfin, les modifications de l'habitat induites par les taches dépendent des caractéristiques des plantes et des taches. Ces modifications induites par les plantes ont des effets en cascade sur les processus biogéochimiques et la croissance des plantes, avec des conséquences pour la dynamique des taches et le fonctionnement de l'écosystème / Submerged aquatic vegetation often grows in lotic systems in patches generated by scale-dependent feedbacks. As ecosystem engineers, plants modify the physical environment triggering positive feedbacks within the patch and negative feedbacks alongside the patch, resulting in regular pattern formation. These scale-dependent feedbacks enable to explain only the lateral expansion of patches, but not their longitudinal development. The objective was to study the processes that trigger positive and negative feedbacks for plants along patches and the consequences for patch dynamics. In situ coupled measurements of hydrodynamics, sediment characteristics, and plant morphology were performed along patches of increasing length. The results demonstrated that a minimum patch length was needed to induce in-patch velocity reduction and fine sediment accumulation. As a consequence of these modifications, patch length influenced the nutrient concentrations in interstitial water of the in-patch sediment, this effect being observed only over a certain threshold length. Over this threshold length, the sediment presented an accumulation of ammonium and depletion of nitrates. Plant height was related to patch length by a quadratic relationship, suggesting that negative feedbacks occur over a certain patch length, probably due to the high ammonium concentration that can be toxic for plants in the range measured. The threshold lengths over which patches influence the biogeochemical processes and negative feedbacks occur were reduced in the ecosystem presenting the highest nutrient level. The results also demonstrated that the physical habitat modifications induced by patches depend on the plant traits and patch characteristics. The plant-induced modifications of the physical habitat have cascading effects on the biogeochemical processes and plant growth, which depended on the environmental conditions, with consequences for patch dynamics and ecosystem functioning

Uma questão que ainda instiga discussões na literatura ecológica é como explicar a coocorrência dinâmica e milenar de formações florestais e campestres sob um mesmo regime climático que tende a favorecer as primeiras, como ocorre atualmente com mosaicos florestacampo no sul do Brasil. A partir de meados do século XX, têm-se evidenciado um fenômeno mundial de avanço de elementos lenhosos sobre áreas abertas. Neste sentido, a modelagem dos processos ecológicos envolvidos na manutenção de ambas as formações numa escala de paisagem permite o esclarecimento dos mecanismos que atuam na manutenção dessa coexistência até o presente e permite prever estados futuros diante dos prognósticos de drásticas alterações climáticas globais já nas próximas décadas. Para tanto, desenvolveu-se um modelo espacialmente explícito (2D-aDGVM) que agrega um Modelo Adaptativo Global de Dinâmica de Vegetação (aDGVM) e ainda inclui heterogeneidades topográficas, propagação do fogo e dispersão de sementes. Este modelo busca satisfazer a necessidade de modelagem mais realista de processos biofísicos, fisiológicos e demográficos na escala de indivíduos e relacionados de forma adaptativa às variações ambientais e aos regimes de distúrbios, ao mesmo tempo que agrega importantes processos ecológicos espaciais, até então pouco ou nada abordados por esse grupo de modelos numa escala de paisagem. Com este modelo, avaliaram-se os efeitos das variações topográficas da radiação solar incidente e destas nos mecanismos de interação (feedbacks) positiva e negativa que surgem daqueles processos na escala de indivíduos e que definem localmente os limites da coexistência entre elementos arbóreos e herbáceos. Ainda, foram analisados os efeitos do aumento da temperatura, precipitação e CO2 atmosférico, desde o período pré-industrial até projeções futuras para as próximas décadas, na performance das diferentes fisiologias envolvidas, bem como no balanço daquelas interações entre as mesmas e, finalmente, na sensibilidade da dinâmica dos mosaicos floresta-campo. Os resultados evidenciaram que, sob o regime climático vigente, uma coexistência relativamente estável entre floresta e campo numa mesma paisagem é mantida por uma alta freqüência de distúrbios, que por sua vez, resulta do forte feedback positivo do acúmulo de biomassa inflamável da vegetação campestre na intensidade do fogo, proporcionado pela condição altamente produtiva do atual clima mesotérmico. Por outro lado, intensificadas pela declividade do terreno, as heterogeneidades espaciais afetaram o balanço dessas interações, interferindo nos padrões espaço-temporais relacionados ao comportamento do fogo e dependentes da densidade de elementos arbóreos. Ainda, tanto esses efeitos observados na escala das manchas de vegetação, como o arranjo espacial inicial das mesmas na paisagem, afetaram as taxas de expansão florestal. Em outras palavras, a manutenção da coexistência de duas formações vegetais constituídas por elementos de inerente assimetria competitiva é possível pela manutenção de uma maior conectividade daquela que propicia o distúrbio, superando a vantagem da outra, que por sua vez é dependente da densidade dos indivíduos. Numa escala de paisagem, isto causa a manutenção de uma baixa conectividade entre as manchas florestais, propiciando sua relativa estabilidade num contexto de dispersão predominante a curtas distâncias. Contudo, embora ambos os sistemas tenham apresentado incremento no crescimento, produtividade e fecundidade, observou-se uma sensibilidade maior no sentido de aumento das taxas de avanço florestal em resposta às projeções climáticas futuras, principalmente nos próximos 90 anos, mesmo na presença do fogo. Isto seria proporcionado pela vantagem fotossintética das árvores-C3 sobre gramíneas-C4 na presença do fogo sob altas concentrações de CO2 atmosférico. Por fim, uma abordagem mais sistêmica dos mosaicos como estados alternativos mostrou ser adequada para o entendimento dos mecanismos que propiciam essa coexistência dinâmica na paisagem. / A longstanding problem in ecology is how to explain the coexistence over thousands of years of forests and natural grasslands under the same climatic regime, which favors the first, such as in forest-grasslands mosaics in South Brazil. Since the middle of the 20th century, a worldwide bush encroachment phenomenon of woody invasion in open vegetation has been threatening this relatively stable coexistence. In this sense, modelling ecological processes that arbitrate the maintenance of both vegetation formations at the landscape scale allows a better understanding of the mechanisms behind the maintenance of this coexistence, as well as predictions of future states under projections of drastic climate change over the next decades. For this, we developed a bidimensional spatial explicit model (2D-aDGVM) that aggregates an adaptive Global Vegetation Model (aDGVM), which includes topographic heterogeneity, fire spread and seed dispersal. The model aims at fulfilling the need for a more realistic representation of biophysical, physiological and demographical processes using an individualbased approach as it adapts these processes to environmental variations and disturbance regimes. In addition, the model includes important spatial ecological processes that have gained less attention by such models adopting a landscape-scale approach. Therefore, we evaluated the effect of topographic variations in incoming solar radiation on positive and on negative feedbacks that rise from those individual-based processes, and which in turns define the limiting thresholds upon which woody and grassy forms coexist. Additionally, the effects of increasing temperature, rainfall and atmospheric CO2 levels on the performance of distinct physiologies (C3-tree and C4-grass) were analyzed, as well as the sensitivity of forestgrassland mosaics to changes in climate from the preindustrial period to the next decades. Results showed that a relatively stable coexistence of forests and grasslands in the same landscape was observed with more frequent fires under the present climatic conditions. This was due to strong positive feedbacks of the huge accumulation of flammable grass biomass on fire intensity promoted by the high productivity of the present mesic conditions. On the other hand, spatio-temporal density dependent processes linked to fire and enhanced by slope at the patch scale, as well as the initial spatial arrangement of vegetation patches affected the rate of forest expansion at the landscape scale. The persistence of coexisting vegetation formations with an inherent asymmetry of competitive interactions was possible when the higher connectivity of the fire-prone patches (grassland) affected negatively the performance of the entire fire-sensitive system (forest). This was possible by overcoming its local densitydependent advantage, or by maintaining it with a low connectivity, which is expected to reduce the rate of coalescence of forest patches in a scenario of predominantly short distance dispersal. Despite the increments in biomass production, stem growth and fecundity that were observed in both grassland and forest, climate change increased the rates of forest expansion over grasslands even in presence of fire, and mainly over the next 90 years. This was attributed to a high photosynthetic advantage of C3-trees over C4-grasses in presence of fire under higher atmospheric CO2 levels. Finally, in the face of the general observed tendency of forest expansion over grasslands, the ancient grasslands have persisted as alternative ecosystem states in forest-grassland mosaics. In this sense, exploring this dynamic coexistence under the concept of alternative stable states have showed to be the most appropriate approach, and the outcomes of this novel perspective may highlight the understanding of the mechanisms behind the long-term coexistence.

Uma questão que ainda instiga discussões na literatura ecológica é como explicar a coocorrência dinâmica e milenar de formações florestais e campestres sob um mesmo regime climático que tende a favorecer as primeiras, como ocorre atualmente com mosaicos florestacampo no sul do Brasil. A partir de meados do século XX, têm-se evidenciado um fenômeno mundial de avanço de elementos lenhosos sobre áreas abertas. Neste sentido, a modelagem dos processos ecológicos envolvidos na manutenção de ambas as formações numa escala de paisagem permite o esclarecimento dos mecanismos que atuam na manutenção dessa coexistência até o presente e permite prever estados futuros diante dos prognósticos de drásticas alterações climáticas globais já nas próximas décadas. Para tanto, desenvolveu-se um modelo espacialmente explícito (2D-aDGVM) que agrega um Modelo Adaptativo Global de Dinâmica de Vegetação (aDGVM) e ainda inclui heterogeneidades topográficas, propagação do fogo e dispersão de sementes. Este modelo busca satisfazer a necessidade de modelagem mais realista de processos biofísicos, fisiológicos e demográficos na escala de indivíduos e relacionados de forma adaptativa às variações ambientais e aos regimes de distúrbios, ao mesmo tempo que agrega importantes processos ecológicos espaciais, até então pouco ou nada abordados por esse grupo de modelos numa escala de paisagem. Com este modelo, avaliaram-se os efeitos das variações topográficas da radiação solar incidente e destas nos mecanismos de interação (feedbacks) positiva e negativa que surgem daqueles processos na escala de indivíduos e que definem localmente os limites da coexistência entre elementos arbóreos e herbáceos. Ainda, foram analisados os efeitos do aumento da temperatura, precipitação e CO2 atmosférico, desde o período pré-industrial até projeções futuras para as próximas décadas, na performance das diferentes fisiologias envolvidas, bem como no balanço daquelas interações entre as mesmas e, finalmente, na sensibilidade da dinâmica dos mosaicos floresta-campo. Os resultados evidenciaram que, sob o regime climático vigente, uma coexistência relativamente estável entre floresta e campo numa mesma paisagem é mantida por uma alta freqüência de distúrbios, que por sua vez, resulta do forte feedback positivo do acúmulo de biomassa inflamável da vegetação campestre na intensidade do fogo, proporcionado pela condição altamente produtiva do atual clima mesotérmico. Por outro lado, intensificadas pela declividade do terreno, as heterogeneidades espaciais afetaram o balanço dessas interações, interferindo nos padrões espaço-temporais relacionados ao comportamento do fogo e dependentes da densidade de elementos arbóreos. Ainda, tanto esses efeitos observados na escala das manchas de vegetação, como o arranjo espacial inicial das mesmas na paisagem, afetaram as taxas de expansão florestal. Em outras palavras, a manutenção da coexistência de duas formações vegetais constituídas por elementos de inerente assimetria competitiva é possível pela manutenção de uma maior conectividade daquela que propicia o distúrbio, superando a vantagem da outra, que por sua vez é dependente da densidade dos indivíduos. Numa escala de paisagem, isto causa a manutenção de uma baixa conectividade entre as manchas florestais, propiciando sua relativa estabilidade num contexto de dispersão predominante a curtas distâncias. Contudo, embora ambos os sistemas tenham apresentado incremento no crescimento, produtividade e fecundidade, observou-se uma sensibilidade maior no sentido de aumento das taxas de avanço florestal em resposta às projeções climáticas futuras, principalmente nos próximos 90 anos, mesmo na presença do fogo. Isto seria proporcionado pela vantagem fotossintética das árvores-C3 sobre gramíneas-C4 na presença do fogo sob altas concentrações de CO2 atmosférico. Por fim, uma abordagem mais sistêmica dos mosaicos como estados alternativos mostrou ser adequada para o entendimento dos mecanismos que propiciam essa coexistência dinâmica na paisagem. / A longstanding problem in ecology is how to explain the coexistence over thousands of years of forests and natural grasslands under the same climatic regime, which favors the first, such as in forest-grasslands mosaics in South Brazil. Since the middle of the 20th century, a worldwide bush encroachment phenomenon of woody invasion in open vegetation has been threatening this relatively stable coexistence. In this sense, modelling ecological processes that arbitrate the maintenance of both vegetation formations at the landscape scale allows a better understanding of the mechanisms behind the maintenance of this coexistence, as well as predictions of future states under projections of drastic climate change over the next decades. For this, we developed a bidimensional spatial explicit model (2D-aDGVM) that aggregates an adaptive Global Vegetation Model (aDGVM), which includes topographic heterogeneity, fire spread and seed dispersal. The model aims at fulfilling the need for a more realistic representation of biophysical, physiological and demographical processes using an individualbased approach as it adapts these processes to environmental variations and disturbance regimes. In addition, the model includes important spatial ecological processes that have gained less attention by such models adopting a landscape-scale approach. Therefore, we evaluated the effect of topographic variations in incoming solar radiation on positive and on negative feedbacks that rise from those individual-based processes, and which in turns define the limiting thresholds upon which woody and grassy forms coexist. Additionally, the effects of increasing temperature, rainfall and atmospheric CO2 levels on the performance of distinct physiologies (C3-tree and C4-grass) were analyzed, as well as the sensitivity of forestgrassland mosaics to changes in climate from the preindustrial period to the next decades. Results showed that a relatively stable coexistence of forests and grasslands in the same landscape was observed with more frequent fires under the present climatic conditions. This was due to strong positive feedbacks of the huge accumulation of flammable grass biomass on fire intensity promoted by the high productivity of the present mesic conditions. On the other hand, spatio-temporal density dependent processes linked to fire and enhanced by slope at the patch scale, as well as the initial spatial arrangement of vegetation patches affected the rate of forest expansion at the landscape scale. The persistence of coexisting vegetation formations with an inherent asymmetry of competitive interactions was possible when the higher connectivity of the fire-prone patches (grassland) affected negatively the performance of the entire fire-sensitive system (forest). This was possible by overcoming its local densitydependent advantage, or by maintaining it with a low connectivity, which is expected to reduce the rate of coalescence of forest patches in a scenario of predominantly short distance dispersal. Despite the increments in biomass production, stem growth and fecundity that were observed in both grassland and forest, climate change increased the rates of forest expansion over grasslands even in presence of fire, and mainly over the next 90 years. This was attributed to a high photosynthetic advantage of C3-trees over C4-grasses in presence of fire under higher atmospheric CO2 levels. Finally, in the face of the general observed tendency of forest expansion over grasslands, the ancient grasslands have persisted as alternative ecosystem states in forest-grassland mosaics. In this sense, exploring this dynamic coexistence under the concept of alternative stable states have showed to be the most appropriate approach, and the outcomes of this novel perspective may highlight the understanding of the mechanisms behind the long-term coexistence.

L’adaptation des individus à un environnement dépend d’une combinaison de caractères adaptatifs, les traits d’histoire de vie, qui impactent la valeur sélective. Pour comprendre comment les organismes s’adaptent à leur environnement, on peut étudier quelles sont les traits composants la valeur sélective et comment ils dépendent de l’environnement biotique et abiotique. Au cours de cette thèse, je me suis intéressé aux composantes de la valeur sélective dans un environnement saisonnier et à ses conséquences sur la dynamique évolutive des traits quantitatifs.Pour cela, j’ai utilisé une approche de modélisation mathématique d’une évolution expérimentale de l’espèce modèle Saccharomyces cerevisiae en cultures successives en batch. La levure de boulanger S. cerevisiae ici étudiée présente un cycle de vie respiro-fermentaire : en présence de glucose, elle le consomme par fermentation tout en produisant de l’éthanol, qui sera consommé dans un deuxième temps par respiration. Les souches de levures évoluent au cours de cycles successifs de fermentation-respiration. A intervalles de temps réguliers, des cellules sont transférées dans un nouveau milieu contenant du glucose où elles effectuent un nouveau cycle. J’ai développé un modèle mathématique d’équations différentielles pour étudier quels sont les traits sélectionnés dans les différentes saisons dans ce dispositif expérimental et comment l’environnement abiotique, l’environnement biotique et les relations entre les traits, impactent leur évolution.Dans un premier temps, j’ai développé et paramétré un modèle d’équations différentielles décrivant la dynamique d’une population multi-souches au cours d’un batch (chapitre 1). J’ai ensuite proposé une décomposition de la valeur sélective et étudié quels traits sont sous sélection, et comment les pressions de sélection changent avec la composition de la population (chapitre 2). Deux types de traits sélectionnés ont pu être mis en évidence : les traits d’histoire de vie, liés au taux de croissance et à la mortalité, et les traits de transition, qui correspondent à la façon dont les souches réagissent aux changements de l’environnement. J’ai également montré que l’importance de chacune des composantes de la valeur sélective est lié à ces traits et à des traits non sélectionnés, via la longueur des différentes saisons. Au cours de l’évolution, ces composantes sont modifiées ce qui modifie la force de la sélection sur chaque trait. Ce phénomène de boucles de rétroaction éco-évolutives permet de mieux comprendre pourquoi la valeur sélective est fréquence-dépendante.Dans un second temps, j’ai utilisé des simulations d’un modèle de dynamique adaptative pour montrer que l’existence d’un trade-off entre deux traits dans la population ancêtre pouvaient entraîner l’émergence d’autres relations entre un trait sélectionné et un trait non-sélectionné au cours de l’évolution (chapitre 3).Enfin, pour mettre en regard les prédictions issues de modèles théoriques et des observations expérimentales, j’ai analysé deux jeux de données à travers le prisme de mon modèle mathématique (chapitre 4). Le premier jeu de données concerne le phénotypage de souches évoluées en batch successifs et leurs ancêtres. L’estimation des paramètres du modèle pour chacune des souches du jeu de données et leur analyse montrent que les traits liés à l’éthanol, sa consommation et sa production ont été principalement sélectionnés. Le second jeu de données, obtenu à partir de compétitions entre plusieurs couples de souches aillant des traits d’histoire de vie contrastés, a permis de mettre en évidence des différences de valeur sélective entre souches et de les relier avec des différences de traits phénotypiques, en cohérence avec les prédictions théoriques. / Adaptation of species to their environment involves combinations of traits, and in particular life history traits, that influence an organism's selective value. To understand the complexity of adaptation, it is appropriate to decipher the contributions of traits to fitness in the presence of different biotic and abiotic environments. In this thesis, I have investigated fitness components when the environment is seasonal, revealing how such components drive the evolutionary dynamics of quantitative traits.My work is based on the mathematical modeling of experimental evolutions in successive batch cultures of Saccharomyces cerevisiae (baker's yeast). The life cycle of this yeast species is of the respiration-fermentation type: (i) in the presence of glucose, it grows by fermentation, transforming glucose into ethanol; (ii) once glucose has been consumed, it grows by respiration, consuming this time ethanol. This sequence corresponds to the two « seasons » in a batch culture and leads to a cycle of successive batches if cells are periodically transferred into fresh medium. By using differential equations for the time courses, my thesis work shows how growth dynamics and environmental features (abiotic or biotic) generate selection pressures on the different traits during these successive seasons, thereby determining evolutionary trajectories.To describe batch dynamics, I first developed and calibrated a set of differential equations describing the growth dynamics of a population of yeast cells throughout a batch, allowing for one or multiple strains to be present (Chapter 1). Based on this model where cells divide without changing genotype, I then showed that a strain's fitness can be understood in terms of just a few components that are easily specified mathematically. I was then able to determine which traits were under selection and how the corresponding selection pressures were affected by the abundances of each strain in the yeast population (Chapter 2). Selected traits were found to be of two types: life history traits associated with growth and mortality rates, and “transition” traits that correspond to the way a strain reacts to environmental change. I also showed that the contributions of the different fitness components are tied to both selected and non-selected traits via the lengths of seasons. Thus, during population dynamics arising across successive batches, these components change, modifying the selection pressure on each trait. One therefore has a feedback loop, revealing why fitness is frequency-dependent in this system.Next, using the fitness decomposition, I studied adaptive dynamics in successive batch cultures. In such a framework where genotypic changes were allowed, and assuming that there was a trade-off between two traits, I showed that adaptive evolutionary dynamics could lead to the emergence of new relations between selected and non-selected traits (Chapter 3).Furthermore, in order to compare my theoretical predictions to experimental results, I used mathematical and statistical models to analyze two datasets (Chapter 4). The first dataset provides trait measurements in “evolved” strains, i.e., strains obtained after evolution across successive batches, as well as of those same traits in the “ancestral” strains at the origin of the experimental evolution. Parameters inference for the different strains showed that selection had operated mainly on ethanol-related traits (production and consumption). A second dataset was obtained from batch experiments putting strains in competition with one another; the analysis showed that my theoretical modeling well predicted the roles of the different traits for determining the relative fitness of the strains.

La technologie OFDMA (Orthogonal frequency division multiple access) a été adoptée par les systèmes de télécommunications de 4ème génération (4G) comme technique de transmission et d'accès multiple pour ses performances supérieures en termes d'efficacité spectrale. Dans ce type de systèmes, l'adaptation dynamique du débit en fonction de la qualité du canal CQI (Channel Quality Indicator) constitue une problématique de recherche d'actualité qui attire l'attention de plusieurs acteurs académiques et industriels. Ce problème d'adaptation dynamique est encore plus complexe à gérer dans des environnements multi-utilisateurs hétérogènes et à ressources limitées tels que les systèmes OFDMA comme WiMAX Mobile et Long-term Evolution (LTE). Dans cette thèse, nous nous intéressons au problème d'allocation de ressources de l'information de feedback relative au CQI dans le cadre de systèmes OFDMA multi-porteuses multi-utilisateurs. Dans le but de réduire la charge (overhead) du feedback, nous proposons une méthode de prédiction du CQI basée sur l'exploitation de la corrélation temporelle de ce dernier et d'une solution inter-couches. L'objectif est de trouver des schémas d'allocation de ressources adaptatifs respectant les contraintes de qualité de service (QoS) applicatives.Nous proposons en premier lieu un algorithme de réduction de feedback PBF (Prediction Based Feedack) qui permet à la station de base (BS) à prédire certaines occurrences du CQI en se basant sur l'algorithme des moindres carrés récursif RLS (Recursive least-square). Les résultats de simulation montrent que l'outil de prédiction du CQI réduit sensiblement l'overhead du feedback et améliore par conséquent le débit de la liaison montante. Nous proposons, par la suite, une version opportuniste de PBF pour atténuer les éventuels effets de sur et sous estimations liées à l'algorithme de prédiction. Dans ce mécanisme, nous exploitons les informations inter-couches pour améliorer les performances des mécanismes de feedbacks périodiques dont PBF fait partie. L'approche opportuniste améliore sensiblement les performances du système pour les cas de mobilité élevée comparés aux cas de faible mobilité.Dans un second temps, nous proposons une plateforme (FEREP : feedback resource allocation and prediction) basée sur une approche inter-couches. Implémentée au niveau de la station BS, FEREP intègre les fonctionnalités de prédiction, d'adaptation dynamique du CQI et d'ordonnancement des demandes de feedback. Elle comporte trois modules. Le module FWA (feedback window adaptation) gère dynamiquement la fenêtre de feedbacks de chaque station mobile (MS) en se basant sur les messages ARQ (Automatic Repeat Request) reçus qui reflètent l'état actuel des canaux respectifs. Le module PBFS (priority-based feedback scheduling) effectue ensuite l'ordonnancement des feedbacks en tenant compte de la taille de la fenêtre de feedback, du profil de l'utilisateur sous la contrainte de la limitation des ressources globales du systèmes réservées au feedback. Afin de choisir les paramètres de transmission MCS (modulation and coding schemes), le module PBF (prediction based feedback) est utilisé pour les utilisateurs dont le feedabck n'a pas pu être ordonnancé dans la trame courante. Les résultats de simulation ont montré un gain significatif des performances de FREREP en comparaison à un mécanisme de référence, en particulier, sous de fortes contraintes de limitation des ressources du feedback.Le protocole ARQ génère un accusé de réception uniquement si l'utilisateur est sélectionné par l'ordonnanceur pour envoyer des données sur la liaison descendante. Dans le cas où la fréquence d'ordonnancement des utilisateurs sur le lien descendant est réduite, les messages ARQ s'en trouvent également réduits, dégradant par conséquent les performances de la plateforme FEREP proposée ci-dessus. En effet, dans ce cas la signalisation ARQ devient insuffisante pour adapter efficacement la fenêtre de feedback de chaque utilisateur. Pour pallier à ce problème, nous proposons l'algorithme DCRA (dynamic CQI resource allocation) qui utilise deux modes d'estimation de la fenêtre de feedback. Le premier est un mode hors-ligne basé sur des études empiriques permettant d'estimer la fenêtre moyenne optimale de feedback en utilisant les profils applicatif et de mobilité de l'utilisateur. Notre analyse de performance par simulation montre que la fenêtre de feedback peut être estimée en fonction de la classe de service des utilisateurs et de leurs profils de mobilité pour un environnement cellulaire donné. Le second mode de fonctionnement de DCRA effectue une adaptation dynamique de la fenêtre en temps réel dans le cas où la signalisation ARQ est suffisante. Une étude comparative avec les mécanismes DFS (deterministic feedback scheduling) et OFS (opportunistic feedback scheduling), a montré que DCRA arrive à réaliser un meilleur gain en ressources montantes grâce à la réduction de l'overhead des feedbacks, sans pour autant trop dégrader le débit descendant des utilisateurs. Du point de vue des utilisateurs, DCRA améliore les contraintes de QoS tels que le taux de perte de paquets et réduit la consommation énergétique des terminaux grâce à la réduction de feedback. / Orthogonal frequency division multiple access (OFDMA) technology has been adopted by 4th generation (a.k.a. 4G) telecommunication systems to achieve high system spectral efficiency. A crucial research issue is how to design adaptive channel quality indicator (CQI) feedback mechanisms so that the base station can use adaptive modulation and coding (AMC) techniques to adjust its data rate based on the channel condition. This problem is even more challenging in resource-limited and heterogeneous multiuser environments such as Mobile WiMAX, Long-term Evolution (LTE) networks. In this thesis, we consider CQI feedback resource allocation issue for multiuser multicarrier OFDMA systems. We exploit time-domain correlation for CQI prediction and cross-layer information to reduce feedback overhead for OFDMA systems. Our aim is find resource allocation schemes respecting the users QoS constraints.Our study begins with proposing prediction based feedback (PBF) which allows the base station to predict the CQI feedbacks based on recursive least-square (RLS) algorithm. We showed that it is useful to use channel prediction as a tool to reduce the feedback overhead and improve the uplink throughput. Then, we propose an opportunistic periodic feedback mechanism to mitigate the possible under and over estimation effects of CQI prediction. In this mechanism, we exploited the cross-layer information to enhance the performance of periodic feedback mechanisms. The opportunistic mechanism improves the system performance for high mobility cases compared to low mobility cases.For OFDMA systems with limited feedback resource, we propose an integrated cross-layer framework of feedback resource allocation and prediction (FEREP). The proposed framework, implemented at the BS side, is composed of three modules. The feedback window adaptation (FWA) module dynamically tunes the feedback window size for each mobile station based on the received ARQ (Automatic Repeat Request) messages that reflect the current channel condition. The priority-based feedback scheduling (PBFS) module then performs feedback allocation by taking into account the feedback window size, the user profile and the total system feedback budget. To choose adapted modulation and coding schemes (MCS), the prediction based feedback (PBF) module performs channel prediction by using recursive least square (RLS) algorithm for the user whose channel feedback has not been granted for schedule in current frame. Through extensive simulations, the proposed framework shows significant performance gain especially under stringent feedback budget constraint.ARQ protocol receives users acknowledgement only if the user is scheduled in the downlink. The reduction in users scheduling frequency also reduces the rate of ARQ hints and degrades the performance of above contributions. In this case, it is difficult to exploit the ARQ signal to adapt the feedback window for that user. To address this issue, we propose a cross-layer dynamic CQI resource allocation (DCRA) algorithm for multiuser multicarrier OFDMA systems. DCRA uses two modes for feedback window estimation. The first one is an off-line mode based on empirical studies to derive optimal average feedback window based on user application and mobility profile. Our experimental analysis shows that the feedback window can be averaged according to users service class and their mobility profile for a given cell environment. DCRA performs a realtime dynamic window adaptation if sufficient cross-layer hints are available from ARQ signaling. DCRA increases uplink resource by reducing feedback overhead without degrading downlink throughout significantly compared to deterministic feedback scheduling (DFS) and opportunistic feedback scheduling (OFS). From the users perspective, DCRA improves QoS constraints like packet loss rate and saves users power due to feedback reduction.

Do changes in how cross-equatorial energy transport is partitioned between the ocean and atmosphere impact the hemispheric climate response to forcing? To find out, we alter the cross-equatorial ocean heat transport in a state-of-the-art GCM and ascertain how changes in energy transport and its partitioning impact hemispheric climate and precipitation sensitivity following abrupt CO2-doubling. We further evaluate the applicability our results in CMIP6-class ESMs, where AMOC facilitates the northward cross-equatorial ocean heat transport. In our experiments, changes in ocean cross-equatorial energy transport trigger compensating changes in atmospheric energy transport through changes in the Hadley cells and a shift in the Intertropical Convergence Zone. However, the climate sensitivity in each hemisphere is linearly related to the ocean heat transport convergence, not atmospheric energy transport convergence, due to the impact of ocean heating on evaporation and atmospheric specific humidity. Similarly, we also find that ocean heat transport convergence controls the hemispheric precipitation sensitivity through the impact of ocean heating on surface evaporation. This relationship is also evident in CMIP6 models, where we find differences in hemispheric precipitation sensitivity to be related to the Atlantic Meridional Overturning Circulation (AMOC). Changes in the AMOC control hemispheric differences in upper ocean heat content, which then affect how the hydrologic cycle responds to CO2 forcing in each hemisphere. These results suggest that ocean dynamics impact the hemispheric climate response to CO2 forcing, particularly how much regional precipitation changes with warming. / Graduate