Biodiversity is distributed unevenly across geographic space and the tree of life. A key task of biology is to understand the ecological and evolutionary processes that generate these patterns. I investigate how the structure and geometry of a landscape, for example the sizes and arrangements of islands in an archipelago, affects processes contributing to the generation and conservation of biodiversity patterns. In the first chapter, I integrate two disparate bodies of theory, ecological neutral theory and network theory into a powerful new framework for investigating patterns of biodiversity in a complex landscape. I examine the consequences of network structure, such as size, topology, and connectivity, for diversity patterning across the metacommunity. The second chapter focuses on how the position of a node within a network controls local community (node) diversity. Network statistics, such as node centrality, are found to predict diversity patterns with more central nodes accumulating the most diversity. In the third chapter, I use the theory to evaluate how well fundamental concepts in conservation biology perform when neutral metacommunity processes generate diversity patterns. I find that contemporary diversity patterns are poor predictors of the long-term capacity of a network to support diversity, challenging a host of conservation concepts and applications. In the fourth chapter, I consider biodiversity dynamics in a network with a different model of speciation, where spatial structure is needed for divergence. In this case, speciation hotspots form where the dispersal properties of an organism and the spatial structure of the landscape coincide. In the final chapter I study the biodiversity of a natural structured metacommunity, the ants of the Fijian archipelago. I used a variety of collecting techniques to inventory the ant species occurring across a system of islands in the southwest Pacific. Approximately 50 new species were discovered, and the distributions of the ant species across the islands are firmly established. Radiations are observed in the genera Pheidole, Camponotus, Lordomyrma, Leptogenys, Cerapachys, Strumigenys, Poecilomyrma, and Hypoponera. / text

Les données disponibles au sujet des villes ne cessent de croître en quantité et en précision. Cependant, malgré l'explosion de la quantité d'information disponible, notre compréhension des processus qui régissent les villes et le phénomène d'urbanisation restent mal compris. Dans cette thèse, nous nous proposons d'étudier les villes en adoptant une démarche inspirée de la physique statistique. Dans un premier temps, nous présentons un modèle stochastique et hors-équilibre de croissance des villes qui décrit la structure du réseau de mobilité. Ce modèle conduit à une prédiction sur la croissance du nombre de centres d'activités avec la population. Cette prédiction est vérifiée de façon indépendante sur des données concernant les villes américaines et espagnoles. Dans le cadre de ce modèle, nous sommes également capables de prédire la valeur de l'exposant des lois d'échelle qui relient la longueur totale des navettes, la longueur totale du réseau viaire, le retard total dû aux embouteillages, la quantité de dioxyde de carbone émis, la surface totale des villes à leur population. Ces prédictions sont elles aussi vérifiées sur des données concernant les villes américaines. Dans une troisième partie distincte, nous nous intéressons à la ségrégation résidentielle. En proposant une nouvelle définition de ce qu'est la ségrégation, nous dérivons naturellement une mesure d'attraction/répulsion entre les différentes catégories. Nous présentons de surcroît une méthode qui permet de diviser de façon non-ambigue et reproductible la distribution des revenus en un nombre discret de classes. Enfin, nous revisitons la dichotomie traditionnelle entre centre-ville et banlieue en construisant une mesure adaptée aux villes anisotropes et polycentriques. Finalement, dans un quatrième temps, nous reproduisons succinctement les résultats que nous avons obtenus dans le cadre de l'étude empirique et théorique des réseaux spatiaux. Dans cette thèse, nous avons essayé de démontrer que la complexité des villes est -- presque paradoxalement -- mieux comprise par des approches simples telles que l'on en trouve en physique. Les méthodes qui sont propres à cette dernière, c'est-à-dire chercher de la structure dans les données, essayer d'isoler les processus les plus importants, construire des modèles simples et ne garder que ceux dont les prédictions sont en accord avec les données, sont en effet pertinentes pour l'étude des systèmes urbains. / The amount of data that is being gathered about cities is increasing in size and specificity. However, despite this wealth of information, we still have little understanding of the processes that drive cities. In this thesis we apply some ideas from statistical physics to the study of cities. We first present a stochastic, out-of-equilibrium model of city growth that describes the structure of the mobility pattern of individuals. The model explains the appearance of secondary subcenters as an effect of traffic congestion. We are also able to predict the sublinear increase of the number of centers with population size, a prediction that is verified on American and Spanish data. Within the framework of this model, we are further able to give a prediction for the scaling exponent of the total distance commuted daily, the total length of the road network, the total delay due to congestion, the quantity of CO2 emitted, and the surface area with the population size of cities. Predictions that agree with data gathered for U.S. cities. In the third part, we focus on the quantitative description of the patterns of residential segregation. We propose a unifying theoretical framework in which segregation can be empirically characterised. We propose a measure of interaction between the different categories. Building on the information about the attraction and repulsion between categories, we are able to define classes in a quantitative, unambiguous way. Finally, we revisit the traditional dichotomy between poor city centers and rich suburbs; we provide a measure that is adapted to anisotropic, polycentric cities. In the fourth and last part, we succinctly present the most important theoretical and empirical results of our studies on spatial networks. Throughout this thesis, we try to convey the idea that the complexity of cities is almost paradoxically better comprehended through simple approaches. Looking for structure in data, trying to isolate the most important processes, building simple models and only keeping those which agree with data, constitute a universal method that is also relevant to the study of urban systems.

No description available.

O crescimento acelerado da produção científica brasileira nos anos recentes foi acompanhado pela expansão das colaborações científicas domésticas. Neste estudo, olhamos mais atentamente esse assunto na tentativa pioneira de identificar padrões espaciais da produção e colaboração científica no Brasil, e avaliar o papel da proximidade geográfica na determinação das interações entre os pesquisadores brasileiros. Por meio de uma base única composta por mais de um milhão de pesquisadores registrados na Plataforma Lattes e de sete milhões de publicações científicas, coletamos e consolidamos informações sobre as colaborações científicas inter-regionais em termos de redes de coautorias entre 1.347 municípios brasileiros ao longo do período compreendido entre 1990 e 2010, o que permitiu uma abrangência de dados e perspectiva de análise inéditas na literatura. Os efeitos da distância geográfica nas redes de colaboração são mensurados para as diferentes áreas do conhecimento por meio da estimação de modelos de interações espaciais. Os principais resultados sugerem fortes evidências de um processo de desconcentração espacial da produção científica nos últimos anos associado à expansão das redes de colaboração e ao aumento da participação de autores das regiões cientificamente menos tradicionais, tais como Sul e Nordeste. Ademais, também encontramos evidência de que a distância ainda desempenha papel crucial na determinação da intensidade dos fluxos de conhecimento nas redes de colaboração científica no Brasil, embora a magnitude do efeito varie entre as redes das diferentes áreas do conhecimento. Por exemplo, verificamos que o distanciamento de 200 quilômetros entre dois pesquisadores implica na redução média de 22% ou 45% na probabilidade de haver colaboração entre eles, caso eles sejam de Linguística, Letras e Artes ou Ciências Exatas e da Terra, respectivamente. / Recent years have witnessed an accelerated growth of Brazilian scientific production that was accompanied by an expansion of domestic research collaboration. In this paper we look more closely at this issue in a pioneering attempt to identify spatial patterns of research production and collaboration in Brazil, and to measure the role of geographical proximity in determining interaction between Brazilian researchers. Using a unique database comprised of over one million researchers registered in the Brazilian Lattes Platform and seven million scientific publications, we collect and consolidate information on interregional research collaboration in terms of co-authorship networks among 1,347 Brazilian cities over the period between 1990 and 2010, which enabled a range of data and analysis perspective unprecedented in literature. The effects of geographical distance on research collaboration are measured for different knowledge areas under the estimation of spatial interaction models. The main results suggest strong evidence of spatial de-concentration of scientific production in the last years with expansion of research collaboration networks and an increased participation of authors in scientifically less traditional regions, such as South and Northeast. Moreover, we also find evidence that distance still plays a crucial role in determining the intensity of knowledge flows in scientific collaboration networks in Brazil, although the magnitude of such effects varies among networks of different knowledge areas. For instance, we found that the distancing of 200 kilometers between two researchers implies an average reduction of 22% or 45% on probability of collaboration among them, if they are of Linguistics, Letters and Arts or Exact and Earth Sciences, respectively.

Die Netzwerktheorie ist eine wirksame Methode, um die Struktur realer Systeme, z.B. des Klimasystems, zu beschreiben und zu klassifizieren. Der erste Teil dieser Arbeit nutzt diese Diskriminanzfähigkeit um die Ost- und Zentralpazifischen Phasen von El Niño und La Niña mittels eines Index basierend auf der Evaluation zeitlich entwickelnder Klimanetzwerke zu unterscheiden. Nach dem Studium der klimatischen Einflüsse dieser unterschiedenen Phasen verlegt die Arbeit ihren Schwerpunkt von der Klassifikation einzelner klimatischer Schichten auf den generelleren Fall interagierender Netzwerke. Hier repräsentieren die Teilnetzwerke entsprechende Variabilitäten in Ozean und Atmosphäre. Es zeigt sich, dass die Ozean-Atmosphären-Wechselwirkung einer hierarchischen Struktur folgt wobei makroskopische Netzwerkmaße einzelne Atmosphärenschichten bezüglich ihrer Wechselwirkung mit dem Ozean unterscheiden. Der zweite Teil dieser Arbeit untersucht den Einfluss der räumlichen Einbettung von Knoten auf topologische Netzwerkeigenschaften. Hierzu werden Nullmodelle eingeführt, welche zufällige Surrogate eines gegebenen Netzwerks erzeugen, sodass globale und lokale räumliche Eigenschaften erhalten bleiben. Diese Modelle erfassen die makroskopischen Eigenschaften der studierten Netzwerke besser als bisherige Standardmodelle zur Erzeugung von Zufallsnetzwerken. Abhängig von der Performanz der vorgeschlagenen Modelle können gegebene Netzwerke schlussendlich in verschiedene Klassen eingeteilt werden. Die Arbeit schließt mit einer Erweiterung der bisherigen Netzwerkklassifikatoren um eine zweidimensionale Metrik, welche Netzwerke auf Basis ihrer Komplexität unterscheidet. Es wird gezeigt, dass Netzwerke des gleichen Typs dazu neigen in individuellen Bereichen der resultierenden Komplexitäts-Entropie-Ebene zu liegen. Die eingeführte Methode ermöglicht auch die objektive Konstruktion von Klimanetzwerken indem Schwellwerte gewählt werden, die die statistische Komplexität maximieren. / Complex network theory provides a powerful tool to quantify and classify the structure of many real-world complex systems, including the climate system. In its first part, this work demonstrates the discriminative power of complex network theory to objectively classify Eastern and Central Pacific phases of El Niño and La Niña by proposing an index based on evolving climate networks. After an investigation of the climatic impacts of these discriminated flavors, this work moves from the classification of sets of single-layer networks to the more general study of interacting networks. Here, subnetworks represent oceanic and atmospheric variability. It is revealed that the ocean-to-atmosphere interaction in the Northern hemisphere follows a hierarchical structure and macroscopic network characteristics discriminate well different parts of the atmosphere with respect to their interaction with the ocean. The second part of this work assesses the effect of the nodes’ spatial embedding on the networks’ topological characteristics. A hierarchy of null models is proposed which generate random surrogates from a given network such that global and local statistics associated with the spatial embedding are preserved. The proposed models capture macroscopic properties of the studied spatial networks much better than standard random network models. Depending on the models’ actual performance networks can ultimately be categorized into different classes. This thesis closes with extending the zoo of network classifiers by a two-fold metric to discriminate different classes of networks based on assessing their complexity. Within this framework networks of the same category tend to cluster in distinct areas of the complexity-entropy plane. The proposed framework further allows to objectively construct climate networks such that the statistical network complexity is maximized.

Nos últimos anos, o estudo de propriedades de redes espaciais urbanas conhecidas como centralidades, tem sido utilizado com frequência para descrever fenômenos de ordem sócio-econômica associados à forma da cidade. Autores têm sugerido que centralidades são capazes de descrever a estrutura espacial urbana (KRAFTA, 1994; ANAS et al., 1998) e, portanto através do estudo de centralidades, é possível reconhecer os espaços que mais concentram fluxos, os que possuem os maiores valores de renda da terra, os mais seguros, entre outros aspectos que parecem estar diretamente relacionados com o fenômeno urbano. A hipótese dessa pesquisa admite que centralidades em redes espaciais urbanas desempenham um papel fundamental na formação da estrutura espacial urbana e na maneira como são organizados os usos do solo da cidade. Assim, essa pesquisa investiga qual modelo de centralidade, processado sobre diversas formas de se descrever o espaço urbano na forma de uma rede, é capaz de apresentar resultados mais fortemente correlacionados com a distribuição espacial de atividades econômicas urbanas. Nessa pesquisa são avaliados cinco modelos de centralidade, aplicados sobre diferentes redes espaciais urbanas com a intenção de se verificar qual deles apresenta maior correlação com a ocorrência de atividades econômicas. Para realizar tal exercício, esses modelos são aplicados sobre três tipos de redes espaciais urbanas (axial, nodal e trechos de rua) – oriundas da configuração espacial de três cidades brasileiras – processados de forma geométrica e topológica. Os modelos de centralidade aplicados são conhecidos como centralidade por Alcance (SEVTSUK; 2010), centralidade por Excentricidade (PORTA et al.; 2009, 2011), centralidade por Intermediação (FREEMAN, 1977), centralidade por Intermediação Planar (KRAFTA, 1994) e centralidade por Proximidade (INGRAM, 1971). O coeficiente de correlação Pearson (r) é utilizado como ferramenta capaz de descrever qual modelo de centralidade, associado a qual tipo de representação espacial e qual modo de processamento de distâncias melhor se correlaciona com a distribuição de atividades econômicas urbanas nessas cidades. As evidências encontradas nessa pesquisa sugerem que os modelos de centralidade por Alcance, centralidade por Excentricidade e centralidade por Intermediação Planar destacam-se em comparação com os demais modelos processados. Além disso, os valores de correlação Pearson (r) mais relevantes foram obtidos quando os modelos de centralidade foram processados considerando-se a geometria da rede formada por trechos de rua, indicando que um tipo de representação espacial mais desagregada e processada de forma geométrica seria mais capaz de apresentar os melhores valores de correlação para a compreensão do fenômeno urbano estudado. / In recent years, the study of urban spatial networks has been often used to describe urban phenomena associated with the shape of the city. Researches suggested that centralities are able to describe the urban spatial structure (KRAFTA, 1994; ANAS et al., 1998) and then it is possible to recognize the spaces with more flows, which have the highest values of land revenue, the safest, among other aspects related to urban phenomenon. The hypothesis of this research accepts that centrality in urban spatial networks play a key role for the urban spatial structure and the way of land uses is organized. Thus, there would be some measures of centrality in urban spatial networks that would be more associated with economic activities occurring in the city. The research will evaluate five measures of centrality applied on three types of urban spatial networks (axial map, node map and segment map). Therefore we will use five models of centrality in urban spatial networks known as reach (SEVTSUK, MEKONNEN, 2012), straightness (PORTA et al., 2006b), betweenness (FREEMAN, 1977), planar betweenness (KRAFTA, 1994) and closeness (INGRAM, 1971) in order to determine which this most highly correlated with the occurrence of economic activities. The relationships between these measures of centrality and locations of economic activities are examined in three Brazilian cities, using as methodology the Pearson correlation coefficient (r). The highest correlation between the results of centrality in urban spatial networks and the location of economic activities will suggest which centrality measure, way of to describe urban space like a network and distance processing method (euclidian or topologic) is more associated with the occurrence of these activities in the city. The results indicate that Reach, Straightness and Planar Betweenness are most outstanding models of centrality. In addition, Pearson correlation coefficients (r) most relevant were obtained when models of centrality are processed considering euclidian paths in the street segments network, suggesting that this type of spatial network and distances processing method generates centralities with more significant correlation values within the urban phenomenon studied.

Nos últimos anos, o estudo de propriedades de redes espaciais urbanas conhecidas como centralidades, tem sido utilizado com frequência para descrever fenômenos de ordem sócio-econômica associados à forma da cidade. Autores têm sugerido que centralidades são capazes de descrever a estrutura espacial urbana (KRAFTA, 1994; ANAS et al., 1998) e, portanto através do estudo de centralidades, é possível reconhecer os espaços que mais concentram fluxos, os que possuem os maiores valores de renda da terra, os mais seguros, entre outros aspectos que parecem estar diretamente relacionados com o fenômeno urbano. A hipótese dessa pesquisa admite que centralidades em redes espaciais urbanas desempenham um papel fundamental na formação da estrutura espacial urbana e na maneira como são organizados os usos do solo da cidade. Assim, essa pesquisa investiga qual modelo de centralidade, processado sobre diversas formas de se descrever o espaço urbano na forma de uma rede, é capaz de apresentar resultados mais fortemente correlacionados com a distribuição espacial de atividades econômicas urbanas. Nessa pesquisa são avaliados cinco modelos de centralidade, aplicados sobre diferentes redes espaciais urbanas com a intenção de se verificar qual deles apresenta maior correlação com a ocorrência de atividades econômicas. Para realizar tal exercício, esses modelos são aplicados sobre três tipos de redes espaciais urbanas (axial, nodal e trechos de rua) – oriundas da configuração espacial de três cidades brasileiras – processados de forma geométrica e topológica. Os modelos de centralidade aplicados são conhecidos como centralidade por Alcance (SEVTSUK; 2010), centralidade por Excentricidade (PORTA et al.; 2009, 2011), centralidade por Intermediação (FREEMAN, 1977), centralidade por Intermediação Planar (KRAFTA, 1994) e centralidade por Proximidade (INGRAM, 1971). O coeficiente de correlação Pearson (r) é utilizado como ferramenta capaz de descrever qual modelo de centralidade, associado a qual tipo de representação espacial e qual modo de processamento de distâncias melhor se correlaciona com a distribuição de atividades econômicas urbanas nessas cidades. As evidências encontradas nessa pesquisa sugerem que os modelos de centralidade por Alcance, centralidade por Excentricidade e centralidade por Intermediação Planar destacam-se em comparação com os demais modelos processados. Além disso, os valores de correlação Pearson (r) mais relevantes foram obtidos quando os modelos de centralidade foram processados considerando-se a geometria da rede formada por trechos de rua, indicando que um tipo de representação espacial mais desagregada e processada de forma geométrica seria mais capaz de apresentar os melhores valores de correlação para a compreensão do fenômeno urbano estudado. / In recent years, the study of urban spatial networks has been often used to describe urban phenomena associated with the shape of the city. Researches suggested that centralities are able to describe the urban spatial structure (KRAFTA, 1994; ANAS et al., 1998) and then it is possible to recognize the spaces with more flows, which have the highest values of land revenue, the safest, among other aspects related to urban phenomenon. The hypothesis of this research accepts that centrality in urban spatial networks play a key role for the urban spatial structure and the way of land uses is organized. Thus, there would be some measures of centrality in urban spatial networks that would be more associated with economic activities occurring in the city. The research will evaluate five measures of centrality applied on three types of urban spatial networks (axial map, node map and segment map). Therefore we will use five models of centrality in urban spatial networks known as reach (SEVTSUK, MEKONNEN, 2012), straightness (PORTA et al., 2006b), betweenness (FREEMAN, 1977), planar betweenness (KRAFTA, 1994) and closeness (INGRAM, 1971) in order to determine which this most highly correlated with the occurrence of economic activities. The relationships between these measures of centrality and locations of economic activities are examined in three Brazilian cities, using as methodology the Pearson correlation coefficient (r). The highest correlation between the results of centrality in urban spatial networks and the location of economic activities will suggest which centrality measure, way of to describe urban space like a network and distance processing method (euclidian or topologic) is more associated with the occurrence of these activities in the city. The results indicate that Reach, Straightness and Planar Betweenness are most outstanding models of centrality. In addition, Pearson correlation coefficients (r) most relevant were obtained when models of centrality are processed considering euclidian paths in the street segments network, suggesting that this type of spatial network and distances processing method generates centralities with more significant correlation values within the urban phenomenon studied.

Nos últimos anos, o estudo de propriedades de redes espaciais urbanas conhecidas como centralidades, tem sido utilizado com frequência para descrever fenômenos de ordem sócio-econômica associados à forma da cidade. Autores têm sugerido que centralidades são capazes de descrever a estrutura espacial urbana (KRAFTA, 1994; ANAS et al., 1998) e, portanto através do estudo de centralidades, é possível reconhecer os espaços que mais concentram fluxos, os que possuem os maiores valores de renda da terra, os mais seguros, entre outros aspectos que parecem estar diretamente relacionados com o fenômeno urbano. A hipótese dessa pesquisa admite que centralidades em redes espaciais urbanas desempenham um papel fundamental na formação da estrutura espacial urbana e na maneira como são organizados os usos do solo da cidade. Assim, essa pesquisa investiga qual modelo de centralidade, processado sobre diversas formas de se descrever o espaço urbano na forma de uma rede, é capaz de apresentar resultados mais fortemente correlacionados com a distribuição espacial de atividades econômicas urbanas. Nessa pesquisa são avaliados cinco modelos de centralidade, aplicados sobre diferentes redes espaciais urbanas com a intenção de se verificar qual deles apresenta maior correlação com a ocorrência de atividades econômicas. Para realizar tal exercício, esses modelos são aplicados sobre três tipos de redes espaciais urbanas (axial, nodal e trechos de rua) – oriundas da configuração espacial de três cidades brasileiras – processados de forma geométrica e topológica. Os modelos de centralidade aplicados são conhecidos como centralidade por Alcance (SEVTSUK; 2010), centralidade por Excentricidade (PORTA et al.; 2009, 2011), centralidade por Intermediação (FREEMAN, 1977), centralidade por Intermediação Planar (KRAFTA, 1994) e centralidade por Proximidade (INGRAM, 1971). O coeficiente de correlação Pearson (r) é utilizado como ferramenta capaz de descrever qual modelo de centralidade, associado a qual tipo de representação espacial e qual modo de processamento de distâncias melhor se correlaciona com a distribuição de atividades econômicas urbanas nessas cidades. As evidências encontradas nessa pesquisa sugerem que os modelos de centralidade por Alcance, centralidade por Excentricidade e centralidade por Intermediação Planar destacam-se em comparação com os demais modelos processados. Além disso, os valores de correlação Pearson (r) mais relevantes foram obtidos quando os modelos de centralidade foram processados considerando-se a geometria da rede formada por trechos de rua, indicando que um tipo de representação espacial mais desagregada e processada de forma geométrica seria mais capaz de apresentar os melhores valores de correlação para a compreensão do fenômeno urbano estudado. / In recent years, the study of urban spatial networks has been often used to describe urban phenomena associated with the shape of the city. Researches suggested that centralities are able to describe the urban spatial structure (KRAFTA, 1994; ANAS et al., 1998) and then it is possible to recognize the spaces with more flows, which have the highest values of land revenue, the safest, among other aspects related to urban phenomenon. The hypothesis of this research accepts that centrality in urban spatial networks play a key role for the urban spatial structure and the way of land uses is organized. Thus, there would be some measures of centrality in urban spatial networks that would be more associated with economic activities occurring in the city. The research will evaluate five measures of centrality applied on three types of urban spatial networks (axial map, node map and segment map). Therefore we will use five models of centrality in urban spatial networks known as reach (SEVTSUK, MEKONNEN, 2012), straightness (PORTA et al., 2006b), betweenness (FREEMAN, 1977), planar betweenness (KRAFTA, 1994) and closeness (INGRAM, 1971) in order to determine which this most highly correlated with the occurrence of economic activities. The relationships between these measures of centrality and locations of economic activities are examined in three Brazilian cities, using as methodology the Pearson correlation coefficient (r). The highest correlation between the results of centrality in urban spatial networks and the location of economic activities will suggest which centrality measure, way of to describe urban space like a network and distance processing method (euclidian or topologic) is more associated with the occurrence of these activities in the city. The results indicate that Reach, Straightness and Planar Betweenness are most outstanding models of centrality. In addition, Pearson correlation coefficients (r) most relevant were obtained when models of centrality are processed considering euclidian paths in the street segments network, suggesting that this type of spatial network and distances processing method generates centralities with more significant correlation values within the urban phenomenon studied.

O crescimento acelerado da produção científica brasileira nos anos recentes foi acompanhado pela expansão das colaborações científicas domésticas. Neste estudo, olhamos mais atentamente esse assunto na tentativa pioneira de identificar padrões espaciais da produção e colaboração científica no Brasil, e avaliar o papel da proximidade geográfica na determinação das interações entre os pesquisadores brasileiros. Por meio de uma base única composta por mais de um milhão de pesquisadores registrados na Plataforma Lattes e de sete milhões de publicações científicas, coletamos e consolidamos informações sobre as colaborações científicas inter-regionais em termos de redes de coautorias entre 1.347 municípios brasileiros ao longo do período compreendido entre 1990 e 2010, o que permitiu uma abrangência de dados e perspectiva de análise inéditas na literatura. Os efeitos da distância geográfica nas redes de colaboração são mensurados para as diferentes áreas do conhecimento por meio da estimação de modelos de interações espaciais. Os principais resultados sugerem fortes evidências de um processo de desconcentração espacial da produção científica nos últimos anos associado à expansão das redes de colaboração e ao aumento da participação de autores das regiões cientificamente menos tradicionais, tais como Sul e Nordeste. Ademais, também encontramos evidência de que a distância ainda desempenha papel crucial na determinação da intensidade dos fluxos de conhecimento nas redes de colaboração científica no Brasil, embora a magnitude do efeito varie entre as redes das diferentes áreas do conhecimento. Por exemplo, verificamos que o distanciamento de 200 quilômetros entre dois pesquisadores implica na redução média de 22% ou 45% na probabilidade de haver colaboração entre eles, caso eles sejam de Linguística, Letras e Artes ou Ciências Exatas e da Terra, respectivamente. / Recent years have witnessed an accelerated growth of Brazilian scientific production that was accompanied by an expansion of domestic research collaboration. In this paper we look more closely at this issue in a pioneering attempt to identify spatial patterns of research production and collaboration in Brazil, and to measure the role of geographical proximity in determining interaction between Brazilian researchers. Using a unique database comprised of over one million researchers registered in the Brazilian Lattes Platform and seven million scientific publications, we collect and consolidate information on interregional research collaboration in terms of co-authorship networks among 1,347 Brazilian cities over the period between 1990 and 2010, which enabled a range of data and analysis perspective unprecedented in literature. The effects of geographical distance on research collaboration are measured for different knowledge areas under the estimation of spatial interaction models. The main results suggest strong evidence of spatial de-concentration of scientific production in the last years with expansion of research collaboration networks and an increased participation of authors in scientifically less traditional regions, such as South and Northeast. Moreover, we also find evidence that distance still plays a crucial role in determining the intensity of knowledge flows in scientific collaboration networks in Brazil, although the magnitude of such effects varies among networks of different knowledge areas. For instance, we found that the distancing of 200 kilometers between two researchers implies an average reduction of 22% or 45% on probability of collaboration among them, if they are of Linguistics, Letters and Arts or Exact and Earth Sciences, respectively.

Im folgenden wird ein Verfahren dargestellt welches die Möglichkeit bietet komplexe räumliche Zusammenhänge zwischen Niederschlagsereignissen quantitativ in Klimanetzwerke zu fassen und diese auf vielfältige Arten und Weisen zu analysieren. In dem Maße wie synchronisiert Niederschlagsereignisse zwischen Raumpunkten auftreten, in dem Maße sind diese Raumpunkte in Event Synchronization Klimanetzwerken verbunden. Zum einen wird das bestehende Ähnlichkeitsmaß der Ereignissynchronisation verbessert und erweitert, und zum anderen werden verschiedene, zum Teil neue, statistische Methoden zur Netzwerkanalyse vorgestellt und erläutert. Klimanetzwerke sind räumlich eingebettete Netzwerke und die statistisch zu zeigende Abhängigkeit der Ähnlichkeit vom räumlichen Abstand führt zu einer vom Raum nicht unabängigen Netzwerkstruktur. Dies ist in einer Vielzahl von Fällen ein ungewünschter Effekt und es wird eine Methodik entwickelt wie dieser statistisch quantifiziert werden kann. Des weiteren werden zwei weitere neue Netzwerkstatistiken vorgestellt. Einerseits das neue Netzwerkmaß Directionality und andererseits eine Netzwerkreduktion welche Klimanetzwerke auf Klimanetzwerke mit weitreichenden Verbindungen reduziert. Dieser neue Ansatz steht gewissermaßen im Gegensatz zur klassischen Klimanetzwerkkonstruktion die vor allem zu kurzreichweitigen Verbindungen führt. Das neue Netzwerkmaß Directionality gibt für jeden Raumpunkt des Netzwerks eine dominante Raumrichtung der Netzwerkverbindungen an und kann dadurch z.B. für bestimmte Event Synchronization Klimanetzwerke Isochronen abbilden. / In the following an approach to the analysis of spatial structures of precipitation event synchronizations is presented. By estimating the synchronicity of precipitation events between points in space, a spatial similarity network is constructed. These Climate Networks can be analyzed statistically in various ways. However, the similarity measure Event Synchronization that will be presented, as well as the concept of Climate Networks, is more general. Climate Network precipitation analyses are done in the applications part in order to present improvements to existing methodologies, as well as novel ones. On one hand, the existing similarity measure Event Synchronization will be refined and extended to a weighted and continuous version, and on the other hand, new methods for statistical analyses of Climate Networks will be presented. Climate Networks are spatially embedded networks and the probability of a link between two nodes decreases with the distance between these nodes. In other words, Climate Network topologies depend on the spatial embedding. Often this effect is distracting and should be considered as a bias in Climate Network statistics. This thesis provides a methodology to estimate this bias and to correct network measures for it. Furthermore, two novel graph statistics are introduced. First, the novel network measure Directionality, and second, a network coarse-graining approach that reduces Climate Networks to Climate Networks of teleconnections, i.e., long-ranged interrelations. This new approach is in contrast to existing Climate Network construction schemes, since commonly most links are short. The novel network measure Directionality provides a dominant direction of links in the embedding space. For undirected Event Synchronization networks this measure is applied for the estimation of Isochrones, i.e., lines of synchronous event occurrences.