As distribuições de abundância das espécies (SADs) são consideradas uma maneira informativa e sintética de descrever a diversidade biológica, e revelam um dos padrões mais bem estabelecidos da ecologia, que é a predominância de espécies raras nas comunidades biológicas. A generalidade deste padrão o torna relevante para a análise de outros parâmetros das comunidades. Além disso, ele levou ao desenvolvimento dos modelos teóricos de SAD. Os modelos estatísticos de SAD, por sua vez, passaram a ocupar um papel central na biologia, como principio empírico que sustenta várias teorias ecológicas. Preston sugeriu em 1948 que as SADs das comunidades naturais teriam uma distribuição aproximadamente Lognormal, porém apareciam truncadas devido ao efeito do tamanho amostral, cuja forma real só seria revelada, portanto, em amostras grandes. Esta ideia vai de encontro à propriedade estatística da consistência, que implica que o acúmulo de evidência causado pelo aumento do tamanho amostral favorece a aproximação ao modelo verdadeiro, bem como às estimativas de seus parâmetros. Além do efeito do tamanho amostral, os padrões de distribuição de abundância podem diferir dependendo da medida de abundância utilizada. Estudos sugerem que a biomassa seria uma medida de abundância mais adequada para revelar a distribuição subjacente de uma comunidade biológica em amostras ou escalas menores (differential veiling). Neste contexto, nosso objetivo principal neste trabalho foi investigar e discutir a vantagem de considerarmos a biomassa como medida de abundância nos modelos teóricos de distribuição de abundância das espécies. Avaliamos, então, o efeito do tamanho da amostra sobre a qualidade dos ajustes dos modelos sob duas diferentes medidas de abundância: biomassa e número de indivíduos. Simulamos amostras de diferentes tamanhos a partir de amostragens aleatórias de uma base de dados empírica e as ajustamos a diferentes modelos de SADs . Com isso, pudemos avaliar a eficiência das amostras - com cada medida de abundância - em revelar o modelo correto, bem como a acurácia e precisão das estimativas dos parâmetros desses modelos. Para o presente estudo, utilizamos um grupo com reconhecido potencial indicador e relevância para estudos ecológicos, os besouros Scarabaeinae. Os resultados encontrados neste estudo mostram que a maior eficiência da biomassa para revelar a distribuição subjacente não é tão geral quanto sugerem os trabalhos anteriores Os critérios de análise utilizados não favoreceram consistentemente a biomassa como medida mais eficiente em revelar o modelo correto e apresentar maior acurácia e precisão nas estimativas de seus parâmetros. Assim, a afirmativa de que a SAD de biomassa não é oculta (veiled) em escalas e amostra menores não parece ser geral. Os resultados obtidos no presente estudo também indicam que o padrão de differential veiling entre as SADs de biomassa e número de indivíduos podem variar dependendo da escolha do modelo e do conjunto de dados avaliados. No entanto, se a causa do differential veiling entre as SADs de biomassa e número de indivíduos for principalmente devido ao efeito de escala, uma explicação alternativa é que a escala do presente estudo teria que ser ampliada para podermos demonstrá-lo. Considerando então, os efeitos de amostragem apontados pelos nossos resultados sobre a medida de diversidade, destacamos que, apesar da importância do assunto, há uma escassez de estudos que investiguem o uso da biomassa como medida de abundância nas SADs, cujas implicações de diferentes efeitos sobre esta medida destacam a necessidade de estudos adicionais que possibilitem isolar o efeito de escala do efeito de tamanho amostral / The species abundance distributions (SADs) are considered an informative way to describe the biological diversity revealling one of the most wellestablished patterns in ecology: the predominance of rare species in biological communities. The generality of such pattern made it relevant to the analysis of other biodiversity parameters an to induce the development of theoretical models of SAD. On the other hand statistical models of SAD occupied a central role in biology as an empirical principle that underlies many ecological theories. Preston suggested in 1948 that SADs follow an approximately lognormal distribution, but that may appear truncated by the effect of sample size, being completely revealed only in large samples. This idea is associated with the statistical property of consistency, which implies that the accumulation of evidence - represented by the increase in sample size - approaches the samples distribution to the true model, as well as the estimates of the parameters to their real values. Beyond the effect of sample size, the SADs may differ depending on the measure of abundance adopted. Studies suggest that biomass could be a more efficient measure of abundance to reveal the underlying distribution of biological communities in smaller samples or scales (differential veiling). In this context, our aim in this study was to investigate and discuss the advantage of considering biomass in theoretical models of SAD. Thus we evaluated the effect of sample size on the quality of models fitness under two different measures of species abundance: biomass and number of individuals. We simulated samples of different sizes taken from an empirical database of Scarabaeinae beetles - recognized as a potential and relevant indicator in ecological studies. Further we adjusted the simulated samples to different SAD models and evaluated the efficiency of each kind of abundance index to reveal the correct model, as well as the accuracy and precision of the parameters estimates. Our results show that the efficiency of biomass to reveal the underlying distribution is not as general as suggested by previous works. According to our analysis criteria, biomass was not consistently as far more efficient in revealing the correct model or in providing greater accuracy and precision in parameters estimates than the number of individuals. Thus, the statement that the SAD biomass is not veiled on small scales or samples does not seem to be general. Moreover our results also indicate that the effect of differential veiling on SADs using biomass and number of individuals depends on the choice of the evaluated model and data set. However, if the cause of the differential veiling - between the biomass and number of individuals SADs - is mainly due to scale effects. An alternative explanation is that the scale of this study was not wide enough to show it. Considering the sampling effects studied here in biodiversity evaluation we highlight the lack of studies investigating the use of biomass as a measure of abundance in SADs which the implications of different effects on this measure indicate the urgent call by further studies on this subject, enablying us to isolate the effect of scale from the sample size

As distribuições de abundância das espécies (SADs) são consideradas uma maneira informativa e sintética de descrever a diversidade biológica, e revelam um dos padrões mais bem estabelecidos da ecologia, que é a predominância de espécies raras nas comunidades biológicas. A generalidade deste padrão o torna relevante para a análise de outros parâmetros das comunidades. Além disso, ele levou ao desenvolvimento dos modelos teóricos de SAD. Os modelos estatísticos de SAD, por sua vez, passaram a ocupar um papel central na biologia, como principio empírico que sustenta várias teorias ecológicas. Preston sugeriu em 1948 que as SADs das comunidades naturais teriam uma distribuição aproximadamente Lognormal, porém apareciam truncadas devido ao efeito do tamanho amostral, cuja forma real só seria revelada, portanto, em amostras grandes. Esta ideia vai de encontro à propriedade estatística da consistência, que implica que o acúmulo de evidência causado pelo aumento do tamanho amostral favorece a aproximação ao modelo verdadeiro, bem como às estimativas de seus parâmetros. Além do efeito do tamanho amostral, os padrões de distribuição de abundância podem diferir dependendo da medida de abundância utilizada. Estudos sugerem que a biomassa seria uma medida de abundância mais adequada para revelar a distribuição subjacente de uma comunidade biológica em amostras ou escalas menores (differential veiling). Neste contexto, nosso objetivo principal neste trabalho foi investigar e discutir a vantagem de considerarmos a biomassa como medida de abundância nos modelos teóricos de distribuição de abundância das espécies. Avaliamos, então, o efeito do tamanho da amostra sobre a qualidade dos ajustes dos modelos sob duas diferentes medidas de abundância: biomassa e número de indivíduos. Simulamos amostras de diferentes tamanhos a partir de amostragens aleatórias de uma base de dados empírica e as ajustamos a diferentes modelos de SADs . Com isso, pudemos avaliar a eficiência das amostras - com cada medida de abundância - em revelar o modelo correto, bem como a acurácia e precisão das estimativas dos parâmetros desses modelos. Para o presente estudo, utilizamos um grupo com reconhecido potencial indicador e relevância para estudos ecológicos, os besouros Scarabaeinae. Os resultados encontrados neste estudo mostram que a maior eficiência da biomassa para revelar a distribuição subjacente não é tão geral quanto sugerem os trabalhos anteriores Os critérios de análise utilizados não favoreceram consistentemente a biomassa como medida mais eficiente em revelar o modelo correto e apresentar maior acurácia e precisão nas estimativas de seus parâmetros. Assim, a afirmativa de que a SAD de biomassa não é oculta (veiled) em escalas e amostra menores não parece ser geral. Os resultados obtidos no presente estudo também indicam que o padrão de differential veiling entre as SADs de biomassa e número de indivíduos podem variar dependendo da escolha do modelo e do conjunto de dados avaliados. No entanto, se a causa do differential veiling entre as SADs de biomassa e número de indivíduos for principalmente devido ao efeito de escala, uma explicação alternativa é que a escala do presente estudo teria que ser ampliada para podermos demonstrá-lo. Considerando então, os efeitos de amostragem apontados pelos nossos resultados sobre a medida de diversidade, destacamos que, apesar da importância do assunto, há uma escassez de estudos que investiguem o uso da biomassa como medida de abundância nas SADs, cujas implicações de diferentes efeitos sobre esta medida destacam a necessidade de estudos adicionais que possibilitem isolar o efeito de escala do efeito de tamanho amostral / The species abundance distributions (SADs) are considered an informative way to describe the biological diversity revealling one of the most wellestablished patterns in ecology: the predominance of rare species in biological communities. The generality of such pattern made it relevant to the analysis of other biodiversity parameters an to induce the development of theoretical models of SAD. On the other hand statistical models of SAD occupied a central role in biology as an empirical principle that underlies many ecological theories. Preston suggested in 1948 that SADs follow an approximately lognormal distribution, but that may appear truncated by the effect of sample size, being completely revealed only in large samples. This idea is associated with the statistical property of consistency, which implies that the accumulation of evidence - represented by the increase in sample size - approaches the samples distribution to the true model, as well as the estimates of the parameters to their real values. Beyond the effect of sample size, the SADs may differ depending on the measure of abundance adopted. Studies suggest that biomass could be a more efficient measure of abundance to reveal the underlying distribution of biological communities in smaller samples or scales (differential veiling). In this context, our aim in this study was to investigate and discuss the advantage of considering biomass in theoretical models of SAD. Thus we evaluated the effect of sample size on the quality of models fitness under two different measures of species abundance: biomass and number of individuals. We simulated samples of different sizes taken from an empirical database of Scarabaeinae beetles - recognized as a potential and relevant indicator in ecological studies. Further we adjusted the simulated samples to different SAD models and evaluated the efficiency of each kind of abundance index to reveal the correct model, as well as the accuracy and precision of the parameters estimates. Our results show that the efficiency of biomass to reveal the underlying distribution is not as general as suggested by previous works. According to our analysis criteria, biomass was not consistently as far more efficient in revealing the correct model or in providing greater accuracy and precision in parameters estimates than the number of individuals. Thus, the statement that the SAD biomass is not veiled on small scales or samples does not seem to be general. Moreover our results also indicate that the effect of differential veiling on SADs using biomass and number of individuals depends on the choice of the evaluated model and data set. However, if the cause of the differential veiling - between the biomass and number of individuals SADs - is mainly due to scale effects. An alternative explanation is that the scale of this study was not wide enough to show it. Considering the sampling effects studied here in biodiversity evaluation we highlight the lack of studies investigating the use of biomass as a measure of abundance in SADs which the implications of different effects on this measure indicate the urgent call by further studies on this subject, enablying us to isolate the effect of scale from the sample size

An important goal of ecological studies is to identify and explain patterns or variation in species assemblages. Ecologists have discovered that global variation in the number of species in an assemblage relates strongly to climate, area, and topographic variability in terrestrial environments. Is the same true for other characteristics of species assemblages? The focus of this thesis is to determine whether species assemblage structure, defined primarily as the body mass frequency distributions and species abundance distributions relate in convergent ways to a set of a few environmental variables across broad spatial scales. First, I found that for mammals and trees most of their geographic variation across North and South America in assemblage structure is statistically related to temperature, precipitation, and habitat heterogeneity (e.g. different vegetation types) in convergent ways. I then examined bird assemblages across islands and continents. Despite the evolutionary and ecological differences between island and continental assemblages, I found that much of the variation in bird assemblage structure depends on temperature, precipitation, land area, and island isolation in congruent patterns in continent and island bird assemblages. Frank Preston modeled species richness based on the total number of individuals and the number of individuals of the rarest species. Building on Preston’s model, Chapter 2 hypothesized that gradients of diversity correlate with gradients in the number of individuals of the rarest species, which in turn are driven by gradients in temperature and precipitation. This hypothesis assumes that species abundance distributions relate to temperature and precipitation in similar ways anywhere in the world. I found that both the number of individuals of the rarest species (m) and the proportion of species represented by a single individual in samples of species assemblages (Φ) were strongly related to climate. Moreover, global variation in species richness was more strongly related to these measures of rarity than to climate. I propose that variation in the shape of the log-normal species abundance distribution is responsible for global gradients of species richness: rare species (reflected in m and Φ) persist better in benign climates. Even though body mass frequency distributions of assemblages show convergent patterns in relation to a set of a few environmental variables, the question remains as to what processes are responsible for creating the geographical variation in the body-size distribution of species. Several mechanisms (e.g. heat conservation and resource availability hypotheses) have been proposed to explain this variation. Chapter 5 tested and found no empirical support for the predictions derived from each of these mechanisms; I showed that species of all sizes occur across the entire temperature gradient. In conclusion, assemblage structure among various taxonomic groups across broad spatial scales relate in similar ways to a set of a few environmental variables, primarily mean annual temperature and mean annual precipitation. While the exact mechanisms are still unknown, I hypothesize several to explain the patterns of convergent assembly. Résumé Un but important de l'écologie est d'identifier et d'expliquer la variation de premier ordre dans les caractéristiques des assemblages d'espèces. Un des patrons ayant déjà été identifié par les écologistes, c'est que la variation mondiale de la richesse en espèces est liée à la variation du climat, de l'aire et de la topographie. Est-ce que d'autres caractéristiques des assemblages d'espèces peuvent être reliées à ces mêmes variables? Le but de cette thèse est de déterminer si la structure des assemblages d'espèces, ici définie comme la distribution des fréquences de masse corporelle ainsi que la distribution d'abondances des espèces, est reliée de manière convergente à un petit ensemble de variables environnementales, et ce, partout dans le monde. D'abord, j'ai déterminé que, pour les mammifères et les arbres, la majorité de la variation géographique dans la structure des assemblages d'espèces est reliée statistiquement à température, précipitation, et l’hétérogénéité du couvert végétal , et ce, de manière convergente pour l'Amérique du Nord et du Sud. Je me suis ensuite penché sur l'assemblage des oiseaux sur les îles et les continents. Malgré les larges différences évolutives et écologiques qui distinguent les îles des continents, je démontre que la majorité de la variation dans la structure des assemblages d'oiseaux dépend de la température, la précipitation, la superficie et l’isolation de façon congruente sur les îles et les continents. Frank Preston a modélisé la richesse en espèces d'une localité, basée sur le nombre total d'individus ainsi que le nombre d'individus de l’espèce la plus rare. En s'appuyant sur les modèles de Preston, Chapître 3 propose une nouvelle hypothèse voulant que les gradients de diversité dépendent des gradients du nombre d'individus de l’espèce la plus rare. Celle-ci dépend des gradients de température et de précipitation. Cette hypothèse repose sur le postulat que la distribution d’abondances des espèces dépend de la température et la précipitation, et ce, de la même manière n’importe où au monde. J’ai mis en évidence que le nombre d’individus de l’espèce la plus rare (m), ainsi que la proportion d’espèces représentées par un individu unique () dans des échantillons locaux étaient fortement reliés au climat. D’ailleurs, la variation globale de la richesse en espèces était plus fortement reliée à ces indices de rareté qu’au climat. Je propose que la variation dans la forme de la distribution log-normale d’abondances d’individus soit responsable des gradients mondiaux de richesse en espèces. En d’autres mots, les espèces rares (indiquées par m et ) persistent mieux dans des climats bénins. Malgré que la distribution des fréquences de masse corporelle des assemblages d'espèces soit liée de manière convergente à seulement quelques variables environnementales, la question demeure à savoir quels processus sont responsables des gradients géographiques de variation en masse corporelle des espèces. Plusieurs mécanismes ont été proposés pour expliquer cette variation. Dans Chapitre 5, j'ai testé les prédictions dérivées de chacun de ces mécanismes sans trouver de support empirique pour aucun. Je démontre aussi que des espèces de toutes tailles se retrouvent sur le gradient de température en entier. En conclusion, la structure des assemblages d'espèces, pour différents groupes taxonomiques et à travers le monde, est liée de façon similaire à un petit nombre de variables environnementales. Bien que les mécanismes soient encore inconnus, j'en propose plusieurs pouvant expliquer ces patrons d'assemblages convergents.

An important goal of ecological studies is to identify and explain patterns or variation in species assemblages. Ecologists have discovered that global variation in the number of species in an assemblage relates strongly to climate, area, and topographic variability in terrestrial environments. Is the same true for other characteristics of species assemblages? The focus of this thesis is to determine whether species assemblage structure, defined primarily as the body mass frequency distributions and species abundance distributions relate in convergent ways to a set of a few environmental variables across broad spatial scales. First, I found that for mammals and trees most of their geographic variation across North and South America in assemblage structure is statistically related to temperature, precipitation, and habitat heterogeneity (e.g. different vegetation types) in convergent ways. I then examined bird assemblages across islands and continents. Despite the evolutionary and ecological differences between island and continental assemblages, I found that much of the variation in bird assemblage structure depends on temperature, precipitation, land area, and island isolation in congruent patterns in continent and island bird assemblages. Frank Preston modeled species richness based on the total number of individuals and the number of individuals of the rarest species. Building on Preston’s model, Chapter 2 hypothesized that gradients of diversity correlate with gradients in the number of individuals of the rarest species, which in turn are driven by gradients in temperature and precipitation. This hypothesis assumes that species abundance distributions relate to temperature and precipitation in similar ways anywhere in the world. I found that both the number of individuals of the rarest species (m) and the proportion of species represented by a single individual in samples of species assemblages (Φ) were strongly related to climate. Moreover, global variation in species richness was more strongly related to these measures of rarity than to climate. I propose that variation in the shape of the log-normal species abundance distribution is responsible for global gradients of species richness: rare species (reflected in m and Φ) persist better in benign climates. Even though body mass frequency distributions of assemblages show convergent patterns in relation to a set of a few environmental variables, the question remains as to what processes are responsible for creating the geographical variation in the body-size distribution of species. Several mechanisms (e.g. heat conservation and resource availability hypotheses) have been proposed to explain this variation. Chapter 5 tested and found no empirical support for the predictions derived from each of these mechanisms; I showed that species of all sizes occur across the entire temperature gradient. In conclusion, assemblage structure among various taxonomic groups across broad spatial scales relate in similar ways to a set of a few environmental variables, primarily mean annual temperature and mean annual precipitation. While the exact mechanisms are still unknown, I hypothesize several to explain the patterns of convergent assembly. Résumé Un but important de l'écologie est d'identifier et d'expliquer la variation de premier ordre dans les caractéristiques des assemblages d'espèces. Un des patrons ayant déjà été identifié par les écologistes, c'est que la variation mondiale de la richesse en espèces est liée à la variation du climat, de l'aire et de la topographie. Est-ce que d'autres caractéristiques des assemblages d'espèces peuvent être reliées à ces mêmes variables? Le but de cette thèse est de déterminer si la structure des assemblages d'espèces, ici définie comme la distribution des fréquences de masse corporelle ainsi que la distribution d'abondances des espèces, est reliée de manière convergente à un petit ensemble de variables environnementales, et ce, partout dans le monde. D'abord, j'ai déterminé que, pour les mammifères et les arbres, la majorité de la variation géographique dans la structure des assemblages d'espèces est reliée statistiquement à température, précipitation, et l’hétérogénéité du couvert végétal , et ce, de manière convergente pour l'Amérique du Nord et du Sud. Je me suis ensuite penché sur l'assemblage des oiseaux sur les îles et les continents. Malgré les larges différences évolutives et écologiques qui distinguent les îles des continents, je démontre que la majorité de la variation dans la structure des assemblages d'oiseaux dépend de la température, la précipitation, la superficie et l’isolation de façon congruente sur les îles et les continents. Frank Preston a modélisé la richesse en espèces d'une localité, basée sur le nombre total d'individus ainsi que le nombre d'individus de l’espèce la plus rare. En s'appuyant sur les modèles de Preston, Chapître 3 propose une nouvelle hypothèse voulant que les gradients de diversité dépendent des gradients du nombre d'individus de l’espèce la plus rare. Celle-ci dépend des gradients de température et de précipitation. Cette hypothèse repose sur le postulat que la distribution d’abondances des espèces dépend de la température et la précipitation, et ce, de la même manière n’importe où au monde. J’ai mis en évidence que le nombre d’individus de l’espèce la plus rare (m), ainsi que la proportion d’espèces représentées par un individu unique () dans des échantillons locaux étaient fortement reliés au climat. D’ailleurs, la variation globale de la richesse en espèces était plus fortement reliée à ces indices de rareté qu’au climat. Je propose que la variation dans la forme de la distribution log-normale d’abondances d’individus soit responsable des gradients mondiaux de richesse en espèces. En d’autres mots, les espèces rares (indiquées par m et ) persistent mieux dans des climats bénins. Malgré que la distribution des fréquences de masse corporelle des assemblages d'espèces soit liée de manière convergente à seulement quelques variables environnementales, la question demeure à savoir quels processus sont responsables des gradients géographiques de variation en masse corporelle des espèces. Plusieurs mécanismes ont été proposés pour expliquer cette variation. Dans Chapitre 5, j'ai testé les prédictions dérivées de chacun de ces mécanismes sans trouver de support empirique pour aucun. Je démontre aussi que des espèces de toutes tailles se retrouvent sur le gradient de température en entier. En conclusion, la structure des assemblages d'espèces, pour différents groupes taxonomiques et à travers le monde, est liée de façon similaire à un petit nombre de variables environnementales. Bien que les mécanismes soient encore inconnus, j'en propose plusieurs pouvant expliquer ces patrons d'assemblages convergents.

An important goal of ecological studies is to identify and explain patterns or variation in species assemblages. Ecologists have discovered that global variation in the number of species in an assemblage relates strongly to climate, area, and topographic variability in terrestrial environments. Is the same true for other characteristics of species assemblages? The focus of this thesis is to determine whether species assemblage structure, defined primarily as the body mass frequency distributions and species abundance distributions relate in convergent ways to a set of a few environmental variables across broad spatial scales. First, I found that for mammals and trees most of their geographic variation across North and South America in assemblage structure is statistically related to temperature, precipitation, and habitat heterogeneity (e.g. different vegetation types) in convergent ways. I then examined bird assemblages across islands and continents. Despite the evolutionary and ecological differences between island and continental assemblages, I found that much of the variation in bird assemblage structure depends on temperature, precipitation, land area, and island isolation in congruent patterns in continent and island bird assemblages. Frank Preston modeled species richness based on the total number of individuals and the number of individuals of the rarest species. Building on Preston’s model, Chapter 2 hypothesized that gradients of diversity correlate with gradients in the number of individuals of the rarest species, which in turn are driven by gradients in temperature and precipitation. This hypothesis assumes that species abundance distributions relate to temperature and precipitation in similar ways anywhere in the world. I found that both the number of individuals of the rarest species (m) and the proportion of species represented by a single individual in samples of species assemblages (Φ) were strongly related to climate. Moreover, global variation in species richness was more strongly related to these measures of rarity than to climate. I propose that variation in the shape of the log-normal species abundance distribution is responsible for global gradients of species richness: rare species (reflected in m and Φ) persist better in benign climates. Even though body mass frequency distributions of assemblages show convergent patterns in relation to a set of a few environmental variables, the question remains as to what processes are responsible for creating the geographical variation in the body-size distribution of species. Several mechanisms (e.g. heat conservation and resource availability hypotheses) have been proposed to explain this variation. Chapter 5 tested and found no empirical support for the predictions derived from each of these mechanisms; I showed that species of all sizes occur across the entire temperature gradient. In conclusion, assemblage structure among various taxonomic groups across broad spatial scales relate in similar ways to a set of a few environmental variables, primarily mean annual temperature and mean annual precipitation. While the exact mechanisms are still unknown, I hypothesize several to explain the patterns of convergent assembly. Résumé Un but important de l'écologie est d'identifier et d'expliquer la variation de premier ordre dans les caractéristiques des assemblages d'espèces. Un des patrons ayant déjà été identifié par les écologistes, c'est que la variation mondiale de la richesse en espèces est liée à la variation du climat, de l'aire et de la topographie. Est-ce que d'autres caractéristiques des assemblages d'espèces peuvent être reliées à ces mêmes variables? Le but de cette thèse est de déterminer si la structure des assemblages d'espèces, ici définie comme la distribution des fréquences de masse corporelle ainsi que la distribution d'abondances des espèces, est reliée de manière convergente à un petit ensemble de variables environnementales, et ce, partout dans le monde. D'abord, j'ai déterminé que, pour les mammifères et les arbres, la majorité de la variation géographique dans la structure des assemblages d'espèces est reliée statistiquement à température, précipitation, et l’hétérogénéité du couvert végétal , et ce, de manière convergente pour l'Amérique du Nord et du Sud. Je me suis ensuite penché sur l'assemblage des oiseaux sur les îles et les continents. Malgré les larges différences évolutives et écologiques qui distinguent les îles des continents, je démontre que la majorité de la variation dans la structure des assemblages d'oiseaux dépend de la température, la précipitation, la superficie et l’isolation de façon congruente sur les îles et les continents. Frank Preston a modélisé la richesse en espèces d'une localité, basée sur le nombre total d'individus ainsi que le nombre d'individus de l’espèce la plus rare. En s'appuyant sur les modèles de Preston, Chapître 3 propose une nouvelle hypothèse voulant que les gradients de diversité dépendent des gradients du nombre d'individus de l’espèce la plus rare. Celle-ci dépend des gradients de température et de précipitation. Cette hypothèse repose sur le postulat que la distribution d’abondances des espèces dépend de la température et la précipitation, et ce, de la même manière n’importe où au monde. J’ai mis en évidence que le nombre d’individus de l’espèce la plus rare (m), ainsi que la proportion d’espèces représentées par un individu unique () dans des échantillons locaux étaient fortement reliés au climat. D’ailleurs, la variation globale de la richesse en espèces était plus fortement reliée à ces indices de rareté qu’au climat. Je propose que la variation dans la forme de la distribution log-normale d’abondances d’individus soit responsable des gradients mondiaux de richesse en espèces. En d’autres mots, les espèces rares (indiquées par m et ) persistent mieux dans des climats bénins. Malgré que la distribution des fréquences de masse corporelle des assemblages d'espèces soit liée de manière convergente à seulement quelques variables environnementales, la question demeure à savoir quels processus sont responsables des gradients géographiques de variation en masse corporelle des espèces. Plusieurs mécanismes ont été proposés pour expliquer cette variation. Dans Chapitre 5, j'ai testé les prédictions dérivées de chacun de ces mécanismes sans trouver de support empirique pour aucun. Je démontre aussi que des espèces de toutes tailles se retrouvent sur le gradient de température en entier. En conclusion, la structure des assemblages d'espèces, pour différents groupes taxonomiques et à travers le monde, est liée de façon similaire à un petit nombre de variables environnementales. Bien que les mécanismes soient encore inconnus, j'en propose plusieurs pouvant expliquer ces patrons d'assemblages convergents.