O desenvolvimento das cidades freqüentemente propicia o aparecimento de empreendimentos de grande porte que, juntamente com o aumento da população e dos veículos, causam impactos nos sistemas de transportes. Dado que há fortes indícios que a localização e o porte destes empreendimentos, aqui denominados de pólos geradores de viagens (ou PGVs), seja influenciado pela presença de outros PGVs, o objetivo deste trabalho foi avaliar esta hipótese em uma cidade média brasileira, através de duas vertentes da análise espacial: a estatística espacial e a modelagem espacial baseada na comparação de informações oriundas de diferentes entidades espaciais. Foram utilizados conceitos de análise exploratória de dados espaciais, tais como o índice e o diagrama de espalhamento de Moran, para a delimitação de regiões similares, divididas de modo a indicar pontos de associação espacial, em relação a uma variável analisada. Em seguida, a modelagem espacial foi feita com uma adaptação da técnica de cellular automata, usando redes neurais artificiais. Dois modelos foram propostos e aplicados em um estudo de caso na cidade de Campinas, São Paulo, com base em diferentes variáveis. No primeiro modelo, somente variáveis representativas da ocupação comercial foram consideradas, enquanto que, no segundo modelo, foram introduzidas também variáveis caracterizando a infra-estrutura viária. Os resultados permitem observar um bom desempenho dos modelos, em particular na fase de validação. As projeções futuras obtidas com os modelos, no entanto, não parecem caracterizar adequadamente o surgimento, que seria naturalmente esperado, de novos pólos comerciais ao redor dos shopping centers estudados. Essa limitação foi surpreendentemente mais evidente no caso do modelo com variáveis de infra-estrutura viária. / The development of cities often leads to the development of large business areas. Those areas, together with the growth of population and vehicles figures in the entire urban area, produce several impacts in the transportation systems. Given the evidences that the location and size of new trip generators (TG\'s) may be strongly influenced by existing TG\'s, the objective of this study was to evaluate that hypothesis in a brazilian medium-sized city. Two branches of spatial analysis were explored in the present study while dealing with different sources of spatial information. They were: spatial statistics and spatial modeling. Firstly, concepts of exploratory spatial data analyses (ESDA), like Moran\'s I index and scatterplot, were used to characterize regions of similar behavior in terms of particular variables. Secondly, spatial modeling was carried out using cellular automata concepts and artificial neural networks. Two models were proposed and applied in a case study carried out in the city Campinas, São Paulo, based on different variables. In the first model, only variables representing commercial land use were considered, while in the second model variables representing the transportation infrastructure supply were also added. The results found indicate a good performance of the models, particularly in the validation process. Their future projections, however, are apparently not able to represent the development of new commercial activities that would be naturally expected around the shopping malls investigated. Although that limitation was observed in both models, it was surprisingly more evident in the case of the model with variables of transportation infrastructure.

O desenvolvimento das cidades freqüentemente propicia o aparecimento de empreendimentos de grande porte que, juntamente com o aumento da população e dos veículos, causam impactos nos sistemas de transportes. Dado que há fortes indícios que a localização e o porte destes empreendimentos, aqui denominados de pólos geradores de viagens (ou PGVs), seja influenciado pela presença de outros PGVs, o objetivo deste trabalho foi avaliar esta hipótese em uma cidade média brasileira, através de duas vertentes da análise espacial: a estatística espacial e a modelagem espacial baseada na comparação de informações oriundas de diferentes entidades espaciais. Foram utilizados conceitos de análise exploratória de dados espaciais, tais como o índice e o diagrama de espalhamento de Moran, para a delimitação de regiões similares, divididas de modo a indicar pontos de associação espacial, em relação a uma variável analisada. Em seguida, a modelagem espacial foi feita com uma adaptação da técnica de cellular automata, usando redes neurais artificiais. Dois modelos foram propostos e aplicados em um estudo de caso na cidade de Campinas, São Paulo, com base em diferentes variáveis. No primeiro modelo, somente variáveis representativas da ocupação comercial foram consideradas, enquanto que, no segundo modelo, foram introduzidas também variáveis caracterizando a infra-estrutura viária. Os resultados permitem observar um bom desempenho dos modelos, em particular na fase de validação. As projeções futuras obtidas com os modelos, no entanto, não parecem caracterizar adequadamente o surgimento, que seria naturalmente esperado, de novos pólos comerciais ao redor dos shopping centers estudados. Essa limitação foi surpreendentemente mais evidente no caso do modelo com variáveis de infra-estrutura viária. / The development of cities often leads to the development of large business areas. Those areas, together with the growth of population and vehicles figures in the entire urban area, produce several impacts in the transportation systems. Given the evidences that the location and size of new trip generators (TG\'s) may be strongly influenced by existing TG\'s, the objective of this study was to evaluate that hypothesis in a brazilian medium-sized city. Two branches of spatial analysis were explored in the present study while dealing with different sources of spatial information. They were: spatial statistics and spatial modeling. Firstly, concepts of exploratory spatial data analyses (ESDA), like Moran\'s I index and scatterplot, were used to characterize regions of similar behavior in terms of particular variables. Secondly, spatial modeling was carried out using cellular automata concepts and artificial neural networks. Two models were proposed and applied in a case study carried out in the city Campinas, São Paulo, based on different variables. In the first model, only variables representing commercial land use were considered, while in the second model variables representing the transportation infrastructure supply were also added. The results found indicate a good performance of the models, particularly in the validation process. Their future projections, however, are apparently not able to represent the development of new commercial activities that would be naturally expected around the shopping malls investigated. Although that limitation was observed in both models, it was surprisingly more evident in the case of the model with variables of transportation infrastructure.

Hipóteses e sistemas ecológicos podem ser descritos pelos mais variados tipos de modelos teóricos. Modelos teóricos por sua vez descrevem sistemas idealizados e podem ser construídos sob diferentes abordagens. A ecologia foi profundamente influenciada por modelos ou hipóteses verbais construídos em uma abordagem predominantemente indutiva. Apesar de inspiradoras, tais hipóteses podem apresentar previsões que são logicamente falhas, uma vez que sistemas ecológicos são altamente complexos cujas trajetórias são difíceis de se prever intuitivamente. Neste sentido modelos teóricos quantitativos podem ser usados como ferramentas para traduzir hipóteses e deduzir previsões que podem ser confrontadas com dados reais. Modelos matemáticos são a forma mais tradicional e aceita de abordagem quantitativa. Apesar das vantagens analíticas da aplicação de modelos matemáticos, eles apresentam limitações para tratar de sistemas mais complexos e em escalas mais basais. Além disso, eles podem se tornar rapidamente complicados não só em termos de tratamento matemático mas também de entendimento por parte de um público mais biológico e empírico. Essa dificuldade de entendimento pode estar impedindo que estudos empíricos sejam fortemente embasados em teoria. Neste sentido, modelos computacionais pode ser uma solução promissora. Modelos computacionais podem ser criados para descrever sistemas virtuais que além de ser mais fáceis de serem desenvolvidos e entendidos por biólogos, permitem a inclusão de vários processos, variáveis e interações. Neste trabalho desenvolvemos um modelo baseado em indivíduos (IBM) para descrever comunidades com estocasticidade demográfica, interações intra e inter-específicas e dispersão. Com essa composição de modelo é possível combinar de diferentes maneiras três dos quatro processos presentes em hipóteses e teorias em ecologia, a saber: deriva, seleção e dispersão. No primeiro capítulo descrevemos os detalhes do modelo e como foi sua concepção e implementação. Ainda no neste capítulo simulamos o modelo explorando um espaço de parâmetros arbitrário, i.e. sem especificar um grupo ou sistema de estudo e analisamos o comportamento do modelo em relação à proporção de espécies persistentes ao fim da simulação e comparamos com a previsão do modelo determinístico de competição sem dispersão. No segundo capítulo aplicamos o modelo em uma versão não espacializada para avaliar dentro das premissas do modelo a consistência lógica das previsões da Hipótese da perturbação intermediária (IDH). No terceiro capítulo simulamos o modelo explorando um espaço de parâmetros baseados em dados empíricos de aves e avaliamos quais características das espécies as tornavam mais ou menos suscetíveis à extinção em paisagens com destruição de habitat. Por fim, discutimos brevemente sobre como o modelo apresentado o modelo apresentado e explorado aqui pode ser usado para diferentes propósitos e responder diferentes perguntas dentro dos contextos teóricos de cada capítulo da tese. E concluímos com algumas considerações finais sobre quais foram as contribuições de se desenvolver um modelo computacional e aplicá-lo a diferentes contextos nesta tese para a formação da doutoranda. / Ecological systems and hypothesis can be described by many different kinds of theoretical models. Theoretical models, on the other hand, are idealized descriptions of real systems that can be constructed under different approaches. Ecology was deeply influenced by verbal models or hypothesis under a inductive approach. Although inspiring, such hypothesis can be logically flawed, since ecological systems are highly complex which trajectories are difficult to predict by intuition. Accordingly quantitative theoretical models can be used as tools to translate hypotheses and deduce predictions that can be confronted with empirical data. Mathematical models are most traditional and well-accepted quantitative approach. Despite of the analytical advantages of using mathematical models , they have limitations to address the complexity of biological systems in lower scales. Furthermore, they become rapidly complicated not only in terms of mathematical treatment but also in terms of comprehension by a biological and empirical audience. This difficulties might prevent that theoretical studies predictions play its role of ground empirical studies. In this sense, computer simulation models can be a promising solution. Computer simulation models are more flexible to include various processes, variables and interaction than mathematical models. Furthermore, they create virtual systems that are easier to be developed and understood by biologists. Here, we developed an individual based model (IBM) to describe communities with stochastic demography, intra and inter-specific interactions and dispersion. With this configuration we can build models combining of different manners three of the four processes present in hypotheses and theories in ecology: drift, selection and dispersal. In the first chapter we describe the model details of implementation and conceptions. We also simulated the model to explore a broad parameter space of competing systems, without specifying a group or system of study; we then analyze the model behavior regarding the proportion of persistent species in the end of the simulation and compared the results with the predictions of deterministic model with competition, without dispersion. In the second chapter we apply the model in a non spatialized version of it to assess the logical consistency of the predictions of the Intermediate disturbance hypothesis (IDH). In the third chapter we used the model to describe fragmented landscapes. We explored a parameter parameters based on empirical data of birds and we evaluate which characteristics of the species made them more or less susceptible to extinction in landscapes with habitat destruction. Finally, we discussed briefly how the model can be used for different purposes and some of the future directions within the theoretical contexts of each chapter of the thesis. We conclude the thesis with a reflexion on how the development and exploration of computer model in this thesis contributed to the student ecological background.

No presente trabalho, fez-se a adaptação do reservatório superficial do ModSimLS para trabalhar como reservatório do solo. O modelo denomina-se IrrigaLS. Por comparação com o método do balanço hídrico, verificou-se que o IrrigaLS é apto para determinar demandas de irrigação suplementares e/ou totais. Com a adaptação, o modelo considera a produção agrícola como o resultado da interação dos fatores água-clima-planta-solo. A capacidade de simular sistemas complexos de recursos hídricos foi testada por comparação dos resultados do IrrigaLS com um modelo de simulação similar (ModSimLS). No ModSimLS, as demandas foram calculadas usando método tradicional de balanço hídrico. Verificou-se que o modelo é apto para simular sistemas complexos de recursos hídricos e apresenta algumas vantagens sobre o ModSimLS, quais sejam: determina a necessidade hídrica real da cultura; garante em períodos de seca um suprimento mínimo de água para a planta; garante uma economia de água pela inclusão da precipitação e possibilidade de variação do volume meta; diferencia as culturas com relação a sensibilidade ao déficit hídrico e calcula o balanço hídrico diário do solo. / In this work, an adaptation in the ModSimLS surface reservoir has been made in order to make it work as a soil reservoir. The model is called IrrigaLS. When compared to the water balance method, it was ascertained that IrrigaLS is suitable for determining supplementary and/or total irrigation demands. The model adaptation considers agricultural production as a result from water-weather-plant-soil factors interaction. The model capability to simulate complex water resources systems has been tested by comparing the results of IrrigaLS with a similar simulation model (ModSimLS) where the irrigation demands were computed using traditional water balance methods. It was ascertained that the model is suitable for simulating complex water resources systems and it has some advantages over ModSimLS, that are: it determines the real water needs of the culture; it guarantees a minimum water supply to the crop during dry periods, it guarantees water saving by the inclusion of rainfall and by the possibility to choose the soil moisture target; it distinguishes the crops in relation to the water deficit sensibility and it calculates the soil daily water balance.

Este estudo objetivou avaliar o efeito das mudanças no uso e cobertura do solo nos últimos 37 anos (1972 a 2008) na resposta hidrológica da bacia do rio Piracicaba, SP. As principais mudanças de uso e cobertura do solo na bacia neste período foram aumento da área urbana, redução da área de pastagem e aumento do plantio de canade- açúcar. Além disso, a bacia conta com uma peculiaridade, que é a presença do Sistema Cantareira, que desvia água à região metropolitana de São Paulo (RMSP), desde 1974. Foi utilizado o Modelo Hidrológico de Grandes Bacias (MGB), que tem a vantagem de avaliar de forma distribuída a variação da cobertura, do tipo e uso do solo, além da própria propagação no escoamento. O modelo foi ajustado à bacia do rio Piracicaba (área de 12.245 km2) para simulação de vazões diárias. Para o processo de ajuste do modelo foram utilizados dados diários de 1972 a 2008, de 12 postos fluviométricos, 61 pluviométricos e 20 meteorológicos, modelo numérico do terreno, mapas de solos e uso e cobertura do solo (para os anos: 1978, 1983, 1985, 1997 e 2003), e parâmetros que são importantes para diferenciar tipos de solo e usos e coberturas do solo. O processo de simulação das vazões foi realizado em duas etapas: calibração dos parâmetros (entre 1972 a 1990) e verificação (1991 a 2008). Para avaliação da qualidade de ajuste foi considerada a análise visual dos hidrogramas (vazões observadas e calculadas) e os valores das funções objetivo (R2, Rlog e V). Foi realizada uma análise de sensibilidade de cada parâmetro, a fim de verificar a influência de cada um. Os resultados mostram um bom ajuste entre as vazões diárias observadas e calculadas pelo modelo. Os melhores resultados das funções objetivo foram obtidos para as sub-bacias de maior área de drenagem. O Sistema Cantareira exerce forte influência na simulação das vazões diárias. As análises de sensibilidade dos parâmetros revelam que os resultados da simulação do modelo são extremamente sensíveis às variações dos parâmetros: capacidade de armazenamento de água no solo (Wm); parâmetro de forma da relação entre o armazenamento e saturação (b); parâmetros de drenagem subterrânea (KBAS) e sub-superficial (KINT); albedo e resistência superficial (rS). Cenários alternativos em relação às mudanças no uso e cobertura do solo foram testados e comparados a simulação do cenário atual da bacia do rio Piracicaba no período entre 1972 e 2008. Comparando o cenário atual ao cenário em que não ocorreu nenhuma alteração na bacia a partir de 1972, verificou-se uma redução das vazões: máxima de 2,5 m3/s; mínima de 2,1 m3/s; e média de 1,9 m3/s. Outros cenários alternativos foram testados e comparados ao cenário atual e foi verificado um aumento da vazão média de 17,2 m3/s para cenário de 100% agricultura; e maior redução da vazão média de 3,9 m3/s para o cenário de conversão da agricultura para pastagem. / This study aimed at to evaluate the effect of changes in land use and land cover over the last 37 years (1972 to 2008) on the hydrological response in the Piracicaba river basin, SP. The main changes in the land use and land cover in the basin over that period were the increase in the urban area, decrease in the pasture area, and increase in the sugarcane plantations. Moreover, the basin has a peculiarity, which is the presence of the Cantareira System, which diverts water to the metropolitan region of São Paulo (RSMP), since 1974. We used a hydrological model of Great Basin (MGB), which has the advantage of evaluating in the distributed form, the variation of the land use/land cover, besides the flow propagation. The model was adjusted to the Piracicaba river basin (12,245 Km2 in area) to simulate daily streamflow. Daily data, from 1972 to 2008, were used for the process of model fitting, from 12 streamflow stations, 61 rainfall and 20 meteorological, digital terrain model, maps of soils and land use and land cover (for the years 1978, 1983, 1985, 1997 and 2003), and parameters that are important to differentiate soil types and uses and land cover. The process of simulation of the streamflow was carried out in two stages: calibration of parameters (from 1972 to 1990) and verification (1991 to 2008). To assess the quality of fitting it was considered the visual analysis of the hydrographs (observed and calculated streamflows) and the values of objective functions (R2, Rlog and V). We performed a sensitivity analysis of each parameter, in order to verify the influence of each one. The results showed a good fitting between the daily observed and calculated streamflows. The best results of the objective functions were obtained for the sub-basins of larger drainage area. The Cantareira System strongly influences the simulation of daily streamflow. Sensitivity analysis of parameters revealed that the simulation results of the model are extremely sensitive to variations of the parameters: soil water storage capacity (Wm), form parameter of the relationship between storage and saturation (b), parameters of drainage underground (KBAS) and subsurface (KINT), albedo and surface resistance (rS). Alternative scenarios for changes in the land use and land cover were tested and compared the simulation of the scenario actual of Piracicaba river basin in the period between 1972 and 2008. Comparing the scenario with no changes in the basin since 1972 to the actual scenario, a reduction in streamflow was detected: maximum of 2,5 m3/s, minimum of 2,1 m3/s, and an average of 1,9 m3/s. Other scenarios were tested and compared to the actual scenario, and it was detected an increase of the mean streamflow of 17,2 m3/s for scenario 100% of agriculture, and a decrease of the mean streamflow of 3,9 m3/s for the scenario of conversion from agriculture into pasture.

No presente trabalho, fez-se a adaptação do reservatório superficial do ModSimLS para trabalhar como reservatório do solo. O modelo denomina-se IrrigaLS. Por comparação com o método do balanço hídrico, verificou-se que o IrrigaLS é apto para determinar demandas de irrigação suplementares e/ou totais. Com a adaptação, o modelo considera a produção agrícola como o resultado da interação dos fatores água-clima-planta-solo. A capacidade de simular sistemas complexos de recursos hídricos foi testada por comparação dos resultados do IrrigaLS com um modelo de simulação similar (ModSimLS). No ModSimLS, as demandas foram calculadas usando método tradicional de balanço hídrico. Verificou-se que o modelo é apto para simular sistemas complexos de recursos hídricos e apresenta algumas vantagens sobre o ModSimLS, quais sejam: determina a necessidade hídrica real da cultura; garante em períodos de seca um suprimento mínimo de água para a planta; garante uma economia de água pela inclusão da precipitação e possibilidade de variação do volume meta; diferencia as culturas com relação a sensibilidade ao déficit hídrico e calcula o balanço hídrico diário do solo. / In this work, an adaptation in the ModSimLS surface reservoir has been made in order to make it work as a soil reservoir. The model is called IrrigaLS. When compared to the water balance method, it was ascertained that IrrigaLS is suitable for determining supplementary and/or total irrigation demands. The model adaptation considers agricultural production as a result from water-weather-plant-soil factors interaction. The model capability to simulate complex water resources systems has been tested by comparing the results of IrrigaLS with a similar simulation model (ModSimLS) where the irrigation demands were computed using traditional water balance methods. It was ascertained that the model is suitable for simulating complex water resources systems and it has some advantages over ModSimLS, that are: it determines the real water needs of the culture; it guarantees a minimum water supply to the crop during dry periods, it guarantees water saving by the inclusion of rainfall and by the possibility to choose the soil moisture target; it distinguishes the crops in relation to the water deficit sensibility and it calculates the soil daily water balance.

Understanding the interactions of predators and their prey and their responses to environmental changes is one of the striking features of ecological research. In this thesis, spring dynamics of phytoplankton and its consumers, zooplankton, were considered in dependence on the environmental conditions in a deep lake (Lake Constance) and a shallow marine water (mesocosms from Kiel Bight), using descriptive statistics, multiple regression models, and process-oriented dynamic simulation models. The development of the spring phytoplankton bloom, representing a dominant feature in the plankton dynamics in temperate and cold oceans and lakes, may depend on temperature, light, and mixing intensity, and the success of over-wintering phyto- and zooplankton. These factors are often correlated in the field. Unexpectedly, irradiance often dominated algal net growth rather than vertical mixing even in deep Lake Constance. Algal net losses from the euphotic layer to larger depth were induced by vertical mixing, but were compensated by the input from larger depth when algae were uniformly distributed over the water column. Dynamics of small, fast-growing algae were well predicted by abiotic variables, such as surface irradiance, vertical mixing intensity, and temperature. A simulation model additionally revealed that even in late winter, grazing may represent an important loss factor of phytoplankton during calm periods when losses due to mixing are small. The importance of losses by mixing and grazing changed rapidly as it depended on the variable mixing intensity. Higher temperature, lower global irradiance and enhanced mixing generated lower algal biomass and primary production in the dynamic simulation model. This suggests that potential consequences of climate change may partly counteract each other. The negative effect of higher temperatures on phytoplankton biomass was due to enhanced temperature-sensitive grazing losses. Comparing the results from deep Lake Constance to those of the shallow mesocosm experiments and simulations, confirmed the strong direct effect of light in contrast to temperature, and the importance of grazing already in early spring as soon as moderate algal biomasses developed. In Lake Constance, ciliates dominated the herbivorous zooplankton in spring. The start of ciliate net growth in spring was closely linked to that of edible algae, chlorophyll a and the vertical mixing intensity but independent of water temperature. The duration of ciliate dominance in spring was largely controlled by the highly variable onset of the phytoplankton bloom, and little by the less variable termination of the ciliate bloom by grazing of meta-zooplankton. During years with an extended spring bloom of algae and ciliates, they coexisted at relatively high biomasses over 15-30 generations, and internally forced species shifts were observed in both communities. Interception feeders alternated with filter feeders, and cryptomonads with non-cryptomonads in their relative importance. These dynamics were not captured by classical 1-predator-1-prey models which consistently predict pronounced predator-prey cycles or equilibria with either the predator or the prey dominating or suppressed. A multi-species predator-prey model with predator species differing in their food selectivity, and prey species in their edibility reproduced the observed patterns. Food-selectivity and edibility were related to the feeding and growth characteristics of the species, which represented ecological trade-offs. For example, the prey species with the highest edibility also had the highest maximum growth rate. Data and model revealed endogenous driven ongoing species alternations, which yielded a higher variability in species-specific biomasses than in total predator and prey biomass. This holds for a broad parameter space as long as the species differ functionally. A more sophisticated model approach enabled the simulation of a continuum of different functional types and adaptability of predator and prey communities to altered environmental conditions, and the maintenance of a rather low model complexity, i.e., low number of equations and free parameters. The community compositions were described by mean functional traits --- prey edibility and predator food-selectivity --- and their variances. The latter represent the functional diversity of the communities and thus, the potential for adaptation. Oscillations in the mean community trait values indicated species shifts. The community traits were related to growth and grazing characteristics representing similar trade-offs as in the multi-species model. The model reproduced the observed patterns, when nonlinear relationships between edibility and capacity, and edibility and food availability for the predator were chosen. A constant minimum amount of variance represented ongoing species invasions and thus, preserved a diversity which allows adaptation on a realistic time-span. / Eine der großen Herausforderungen der heutigen ökologischen Forschung ist es, Veränderungen von Ökosys­temen vorher­zusagen, die mit dem Klimawandel einhergehen. Dafür sind ein umfassendes Verständnis der ver­schiedenen Steuerungsfaktoren des entsprechenden Systems und Kenntnisse zur Anpassungs­fähigkeit des Systems nötig. Auf der Grundlage dieses Wissens, können mit mathemati­schen Modellen Klima­szenarien gerechnet und Vorhersagen erstellt werden. Die vorliegende Arbeit untersuchte die Regulation des Phytoplanktons (kleine freischwebende einzellige Algen) und seiner Konsumenten (Zooplankton, tierische Kleinstlebewesen) sowie deren Wechselspiel während des Frühjahrs mit Bezug auf den Klimawandel. Als Basis dienten langjährige Daten von einem großen tiefen See (Bodensee) sowie Daten von Versuchen mit Organis­men aus einem flachen marinen Ge­wässer (Kieler Förde, Ostsee). Diese Daten wurden mit statistischen Verfahren und mathematischen Modellen ausge­wertet. In Gewässern sind Algen als Primärproduzenten die Nahrungsgrundlage für tieri­sche Organismen bis hin zu Fischen und Meeresfrüchten, und bestimmen die Wasserqualität der Ge­wässer. Daher ist es wichtig zu verstehen, welche Mechanismen die Dynamik der Algen steuern. Der Grundstein für die saisonale Entwicklung von Phyto- und Zooplankton in Gewässern un­serer Breiten wird mit dem Be­ginn des Wachstums im Frühjahr gelegt. Diese Arbeit zeigt, dass es bereits im zeitigen, noch kalten Frühjahr ein Wechselspiel physikalischer und biologischer Steuerungsmechanismen für die Algenent­wicklung gibt. Physikalische Faktoren sind die Wassertemperatur, die Globalstrahlung und die Durchmischung des Gewässers, die durch die Stärke des Windes beeinflusst wird. All diese Steue­rungsmechanismen sind eng miteinander verwoben und werden unterschiedlich stark vom Klimawan­del beeinflusst. Mit mathematischen Modellen gelang es den Einfluss einzelner Faktoren voneinander zu trennen und zu zeigen, dass Effekte durch den Klimawandel sich gegenseitig aufheben oder aber auch verstärken können. Schon geringe Änderungen an der Basis der Nahrungsnetze können weitrei­chende Auswirkungen auf höhere Ebenen habe. Wie stark diese Auswirkungen im Einzelnen sind, hängt entscheidend von der Anpassungsfähigkeit gesamter Ökosysteme und ihrer Artengemeinschaf­ten sowie einzelner Individuen ab. Beispielsweise reagiert die Algengemeinschaft auf einen starken Fraßdruck ihrer Räuber mit einer Verschiebung zu weniger gut fressbaren Algenarten. Diese weniger gut fressbaren Arten unterscheiden sich jedoch auch in anderen Eigenschaften, wie zum Beispiel der Ressourcenausnutzung, von besser fressbaren Algen. In dieser Arbeit wurden Modellansätze entwi­ckelt, die diese Fähigkeit zur Anpassung berücksichtigen. Auf dieser Grundlage und mit Einbeziehung der physikalischen Steuerungsfaktoren können Klimaszenarien gerechnet werden und Vorhersagen für den Einfluss des Klimawandels auf unsere Gewässer gemacht werden, die letztlich auch Perspektiven für Handlungsmöglichkeiten aufzeigen.

Environmental simulation models have been playing a key role in civil and environmental engineering decision making processes for decades. The utility of an environmental model depends on how well the model is structured and calibrated. Model calibration is typically in an automated form where the simulation model is linked to a search mechanism (e.g., an optimization algorithm) such that the search mechanism iteratively generates many parameter sets (e.g., thousands of parameter sets) and evaluates them through running the model in an attempt to minimize differences between observed data and corresponding model outputs. The challenge rises when the environmental model is computationally intensive to run (with run-times of minutes to hours, for example) as then any automatic calibration attempt would impose a large computational burden. Such a challenge may make the model users accept sub-optimal solutions and not achieve the best model performance. The objective of this thesis is to develop innovative strategies to circumvent the computational burden associated with automatic calibration of computationally intensive environmental models. The first main contribution of this thesis is developing a strategy called “deterministic model preemption” which opportunistically evades unnecessary model evaluations in the course of a calibration experiment and can save a significant portion of the computational budget (even as much as 90% in some cases). Model preemption monitors the intermediate simulation results while the model is running and terminates (i.e., pre-empts) the simulation early if it recognizes that further running the model would not guide the search mechanism. This strategy is applicable to a range of automatic calibration algorithms (i.e., search mechanisms) and is deterministic in that it leads to exactly the same calibration results as when preemption is not applied. One other main contribution of this thesis is developing and utilizing the concept of “surrogate data” which is basically a reasonably small but representative proportion of a full set of calibration data. This concept is inspired by the existing surrogate modelling strategies where a surrogate model (also called a metamodel) is developed and utilized as a fast-to-run substitute of an original computationally intensive model. A framework is developed to efficiently calibrate hydrologic models to the full set of calibration data while running the original model only on surrogate data for the majority of candidate parameter sets, a strategy which leads to considerable computational saving. To this end, mapping relationships are developed to approximate the model performance on the full data based on the model performance on surrogate data. This framework can be applicable to the calibration of any environmental model where appropriate surrogate data and mapping relationships can be identified. As another main contribution, this thesis critically reviews and evaluates the large body of literature on surrogate modelling strategies from various disciplines as they are the most commonly used methods to relieve the computational burden associated with computationally intensive simulation models. To reliably evaluate these strategies, a comparative assessment and benchmarking framework is developed which presents a clear computational budget dependent definition for the success/failure of surrogate modelling strategies. Two large families of surrogate modelling strategies are critically scrutinized and evaluated: “response surface surrogate” modelling which involves statistical or data–driven function approximation techniques (e.g., kriging, radial basis functions, and neural networks) and “lower-fidelity physically-based surrogate” modelling strategies which develop and utilize simplified models of the original system (e.g., a groundwater model with a coarse mesh). This thesis raises fundamental concerns about response surface surrogate modelling and demonstrates that, although they might be less efficient, lower-fidelity physically-based surrogates are generally more reliable as they to-some-extent preserve the physics involved in the original model. Five different surface water and groundwater models are used across this thesis to test the performance of the developed strategies and elaborate the discussions. However, the strategies developed are typically simulation-model-independent and can be applied to the calibration of any computationally intensive simulation model that has the required characteristics. This thesis leaves the reader with a suite of strategies for efficient calibration of computationally intensive environmental models while providing some guidance on how to select, implement, and evaluate the appropriate strategy for a given environmental model calibration problem.

Environmental simulation models have been playing a key role in civil and environmental engineering decision making processes for decades. The utility of an environmental model depends on how well the model is structured and calibrated. Model calibration is typically in an automated form where the simulation model is linked to a search mechanism (e.g., an optimization algorithm) such that the search mechanism iteratively generates many parameter sets (e.g., thousands of parameter sets) and evaluates them through running the model in an attempt to minimize differences between observed data and corresponding model outputs. The challenge rises when the environmental model is computationally intensive to run (with run-times of minutes to hours, for example) as then any automatic calibration attempt would impose a large computational burden. Such a challenge may make the model users accept sub-optimal solutions and not achieve the best model performance. The objective of this thesis is to develop innovative strategies to circumvent the computational burden associated with automatic calibration of computationally intensive environmental models. The first main contribution of this thesis is developing a strategy called “deterministic model preemption” which opportunistically evades unnecessary model evaluations in the course of a calibration experiment and can save a significant portion of the computational budget (even as much as 90% in some cases). Model preemption monitors the intermediate simulation results while the model is running and terminates (i.e., pre-empts) the simulation early if it recognizes that further running the model would not guide the search mechanism. This strategy is applicable to a range of automatic calibration algorithms (i.e., search mechanisms) and is deterministic in that it leads to exactly the same calibration results as when preemption is not applied. One other main contribution of this thesis is developing and utilizing the concept of “surrogate data” which is basically a reasonably small but representative proportion of a full set of calibration data. This concept is inspired by the existing surrogate modelling strategies where a surrogate model (also called a metamodel) is developed and utilized as a fast-to-run substitute of an original computationally intensive model. A framework is developed to efficiently calibrate hydrologic models to the full set of calibration data while running the original model only on surrogate data for the majority of candidate parameter sets, a strategy which leads to considerable computational saving. To this end, mapping relationships are developed to approximate the model performance on the full data based on the model performance on surrogate data. This framework can be applicable to the calibration of any environmental model where appropriate surrogate data and mapping relationships can be identified. As another main contribution, this thesis critically reviews and evaluates the large body of literature on surrogate modelling strategies from various disciplines as they are the most commonly used methods to relieve the computational burden associated with computationally intensive simulation models. To reliably evaluate these strategies, a comparative assessment and benchmarking framework is developed which presents a clear computational budget dependent definition for the success/failure of surrogate modelling strategies. Two large families of surrogate modelling strategies are critically scrutinized and evaluated: “response surface surrogate” modelling which involves statistical or data–driven function approximation techniques (e.g., kriging, radial basis functions, and neural networks) and “lower-fidelity physically-based surrogate” modelling strategies which develop and utilize simplified models of the original system (e.g., a groundwater model with a coarse mesh). This thesis raises fundamental concerns about response surface surrogate modelling and demonstrates that, although they might be less efficient, lower-fidelity physically-based surrogates are generally more reliable as they to-some-extent preserve the physics involved in the original model. Five different surface water and groundwater models are used across this thesis to test the performance of the developed strategies and elaborate the discussions. However, the strategies developed are typically simulation-model-independent and can be applied to the calibration of any computationally intensive simulation model that has the required characteristics. This thesis leaves the reader with a suite of strategies for efficient calibration of computationally intensive environmental models while providing some guidance on how to select, implement, and evaluate the appropriate strategy for a given environmental model calibration problem.

Some normally consolidated soft soils manifest strength sensitivity, ie these soil manifest strain softening when shear in an undrained mode. These soils, referred to as sensitive soft soils, have the typical features of strain hardening in drained shearing and strain softening in undrained shearing. The consolidation lines of these soils are also curved (concave upwards) in the semi-log space. However, under high consolidation stress or upon large shearing, these soils re-gain the features of re-constituted soil. Ground improvement methods like stone columns were reported as not effective when installed in the sensitive soft clays. But mechanism of the un-effectiveness of the stone columns remains unknown because of lack of a suitable and simple model for simulating the stress-strain behaviours of sensitive soft soils. Although these soils have a meta-stable micro-structure, models that developed for simulating structured firm soils are not suitable for simulating sensitive soft soil features. Thus, a new model was formulated. The new model can degenerate back to a Modified Cam Clay model. The ability of new model in simulating a range of behaviour was verified by using the finite difference (FD) method in solving the partial differential equations of the soil model for a range of tri-axial test conditions. The model was further implemented in coupled analysis formulation and coded into FEM program AFENA. Various cases with different soil parameters were then simulated and compared with the FD solutions for various triaxial tests so as to check the stability of the FEM code. The coupled FEA was then used to simulate the performance of geosynthetic-encased stone columns. A new stone column element and a geo-encasement element were developed and coded into AFENA. The stone column simulations were then done for both non-sensitive soils (represented by Modified Cam Clay model) and sensitive soft soil (represented by the new model). Parametric study was conducted to examine the performance of the geo-encased stone columns in both types of soils. Furthermore, two different installation methods: wished-in installation and full displacement installation were studied numerically. Cross comparison was done to investigate how the sensitive soft soil features interact with the installation method in affecting the performance of the geo-encased stone columns. A range of factors that influence the geosynthetic-encased stone columns performance installed in soft soils were also made clear.