Os modelos de distribuição de espécies (MDEs) utilizam dados de ocorrência de campo e variáveis ambientais para indicar locais adequados para a ocorrência de uma espécie. Apesar dos inúmeros trabalhos que avaliam diversos aspectos deste tipo de modelagem, a aplicação desses métodos para predição do distribuição potencial de espécies de aves em escala local para o Parque Estadual da Serra do Mar (PESM) São Paulo, ainda não foi avaliada. Este trabalho analisa o potencial dos MDEs para o referido parque a partir das variáveis ambientais disponíveis para modelagem da área. Para isso, foram desenvolvidos modelos para 23 espécies de aves a partir de dados coletados em levantamento sistemático da avifauna do Núcleo São Sebastião um dos 8 núcleos que integram o PESM. Numa primeira etapa foi utilizado o algoritmo Maxent, por ser um dos mais utilizados segundo a literatura e por sua habilidade em realizar predições a partir de informações incompletas. A seguir a performance do Maxent foi comparada com as performances dos algoritmos GARP e Enviromental Distance. O parâmetro utilizado para analisar o desempenho e definir os melhores modelos preditivos foi a analise do valor de AUC. Os modelos ainda foram avaliados através da matriz de confusão, para estabelecer a taxa de omissão, e realizar o Teste Binomial Duas proporções do BioEstat 5.0 a fim de estabelecer a probabilidade dos acertos destes modelos serem diferentes do acaso. E finalmente foi determinada a área mínima estimada, baseado na premissa que um bom modelo deve prever a menor área possível com o maior numero de acertos ao mesmo tempo. Apesar de existir um número reduzido de camadas disponíveis para modelagem da área em escala local 8 mapas - O Maxent, teve o melhor desempenho entre os algoritmos testados, esse algoritmo se mostrou capaz de gerar mapas estatisticamente bons e biologicamente confiáveis para aves, nas áreas estudadas. São necessários cuidados específicos com relação a escolha das espécies a serem modeladas, assegurando que os camadas disponíveis representem premissas das necessidades ecológicas dessas espécies. Estudos posteriores sobre as respostas do algoritmo a diferentes quantidades e padrões de distribuição dos dados de ocorrência podem elucidar a aplicabilidade do algoritmo como ferramenta para auxiliar nos trabalhos de conservação. / Species Distribution Models (SDM) use Field occurrence data and environmental variables to indicate adequate places for the presence of a species. Despite a number of papers evaluate various aspects on these methodologies, the application of such models in order to predict the potential niche for bird species in Parque Estadual da Serra do Mar (PESM) in Sao Paulo, hasnt been analyzed yet. This paper examines the potential of SDM for the referred park from available variable to the modelling of the area. In order to do so, models were developed for 23 bird species form data collected in a systematic survey in the avifauna in Núcleo Sao Sebastião which is one of the 8 centers that are part of PESM. At first, Maxent algorithm was used as it is one of the most used algorithms according to current literature and due to its ability to accomplish predictions from incomplete information. Maxent performance was then compared to GARP and Environmental Distance algorithms performances. The parameter used to analyze skill and define the best predicting models was AUC value analyses. They were also evaluated though the confusion matrix as to establish the omission rate and take the binomial test of two proportions in order to define the correctness of such models as being different from chance and finally the estimated minimum area was determined based on the premise that a good model must predict the smallest possible area with the biggest number of correctness allowed. Although there are a few number of layers for modelling the area in local scale (just 8 maps), maxent results to be potentially useful for generating maps for potential distribution of the birds from the park. It was the best result shown by the 3 algorithms methods tried out. More studies are necessary as to evaluate the algorithms behavior from a smaller number of records occurrences well as different patterns en these records for the definition of the best method of distribution modelling of the avifauna in the park. Maxent proved itself to be capable of generating maps that are mathematically good and ecologically reliable for the birds in studied areas, It was the best result shown shown by the 3 algorithms methods tried out. Specific care is necessary in relation to the species to be modeled so as to assure the feasible layers represent premises of the ecological needs for such species. Further studies on the algorithm responses to different quantities and distributions patterns of occurrence data may elucide its applicability as a tool to help conservation work.

A área de pesquisa científica, denominada computação e biodiversidade, tem por desafio suprir a demanda por tecnologias de apoio à conservação da biodiversidade, fornecendo ferramentas computacionais aplicadas ao estudo da biodiversidade. Os modelos de distribuição geográfica de espécies têm uma importante contribuição para o entendimento da biodiversidade e no apoio para a tomada de decisão em conservação dos recursos de biodiversidade. Os resultados da modelagem de distribuição de espécies, não apresentam com clareza os tratamentos aplicados aos dados na etapa de pré-análise e os critérios de seleção das variáveis preditoras, o que agrava a incerteza quanto aos resultados obtidos e prejudica a reprodutibilidade do experimento. Assim, o objetivo desta pesquisa é apresentar o processo de modelagem de distribuição de espécie com destaque para as atividades de pré-análise e atividades de seleção das variáveis preditoras. O processo de modelagem de distribuição de espécies proposto é avaliado por meio de estudo de caso de Modelagem de Distribuição de Espécies do polinizador Centris hyptidis e das plantas oleaginosas Angelonia campestris e Angelonia cornigera, que considera como fator biótico a especialização da interações entre essas plantas e o polinizador. Neste estudo de caso pode-se verificar que a aplicação de técnicas estatísticas exploratórias de dados na etapa de pré-análise do processo de modelagem distribuição de espécies permite a avaliação da qualidade dos pontos de ocorrência, essenciais para o desempenho preditivo do modelo final. / The area of research called biodiversity informatics, or bioinformatics, has to face the challenge of meeting the demand for technologies to support the conservation of biodiversity, providing computational tools applied to the study of biodiversity. The models of geographic distribution of species have a fundamental implication for understanding the biodiversity and conservation decision making. Researches on modeling species distributions generally do not present clearly the treatments applied to the data in the pre-analysis and criteria for selection of predictor variables, which increases the uncertainty regarding the results and affect the reproducibility of experiment. The objective of this research is to present the process of species distribution modeling with emphasis on the activities of pre-analysis and activities selection of the predictor variables, such as to favor its repeatability and reproducibility by other researchers. The process of modeling species distribution proposed is validated on a case study of modeling distribution of pollinator species Centris hyptidis and oilseed Angelonia campestris and Angelonia cornigera, a biotic factor that considers the specialization of these interactions between plants and pollinators. In this case study we can observe one of the main contributions of this work: the application of statistical techniques for data exploration in the pre-analysis of species distribution modeling process, with improved capacity for evaluation and selection of points of occurrence essential to the performance of the predictive model.

O aumento da importância da cultura da cana-de-açúcar nos últimos anos atraiu de investimentos ao setor sucro-alcooleiro, tornando o planejamento estratégico uma ferramenta essencial na orientação para expansão da cultura em novas áreas e otimização da produção nas áreas tradicionais de cultivo. A modelagem agrícola, por isso, ganhou importância por contribuir neste sentido. Existem diversos modelos de simulação de cana-de-açúcar sendo utilizados em todo o mundo, sendo o DSSAT/CANEGRO um dos mais utilizados. O modelo DSSAT/CANEGRO foi baseado no modelo CERES-MAIZE e desenvolvido na África do Sul. Incluído no conjunto de modelos Decision Support System for Agrotechnology Transfer (DSSAT versão 3.1), e atualizado na versão 4.5 do DSSAT, vem sendo aplicado em diversas regiões do mundo para análise e avanço no conhecimento dos sistemas de produção de cana-deaçúcar. O objetivo do trabalho foi a parametrização e avaliação do modelo calibrado para 5 variedades brasileiras de cana-de-açúcar: CTC 4, CTC 7, CTC 20, RB 86 7515 e RB 83 5486. Foram realizadas medidas em campo experimental em duas localidades para as variedades CTC 4, CTC 7, CTC 20 e obtenção de dados bibliográficos para as variedades RB 86-7515 e RB 83-5486. Uma análise de sensibilidade local, utilizando dados climáticos e pedológicos de Piracicaba foi realizada para determinar a dependência das variáveis simuladas em relação às variações nos parâmetros do modelo. Em seguida, realizou-se a parametrização do modelo utilizando a técnica GLUE (do ingles, generalized likelihood uncertainity estimation). A avaliação das cinco variedades estudadas foi feita utilizando o R²; índice D de Willmott e quadrado médio dos desvios como indicadores estatísticos. Para a variável de índice de área foliar, as variedades CTC 4, CTC 7, CTC 20 apresentaram índice D variando entre 0,870 e 0,944, sendo com indicadores estatísticos similares para as variáveis altura de colmo (D=0,80), perfilhamento (D=0,90) e teor de sacarose (D=0,50). A variedade RB 83-5486 apresentou resultados próximos aos observados para as variáveis de teor de sacarose e massa fresca do colmo. Os piores resultados foram obtidos para a variedade RB 86- 7515, com D=0,743 para massa fresca do colmo, D=0,821 para altura de colmo, D=0,873 para perfilhamento e D=0,665 para massa seca de colmo. O modelo teve melhor desempenho para as condições de Piracicaba, Pradópolis, Guariba e Porto Feliz, no Estado de Sao Paulo, em comparação com as condições de solo e clima de Couripe, Estado de Alagoas. / The importance of sugarcane crop has increasing during the last years, attracting more investments to the sugarcane sector. Thus, the strategic planning has becoming an essential tool to guide the sugarcane expansion to new areas, as well as to help optimize the traditional production systems. To do so, crop modelling has a great usefulness. There are several models for sugarcane crop simulation around the world, and DSSAT/CANEGRO is one of the most used. DSSAT/CANEGRO model was based in the CERES-MAIZE model and developed in South Africa. It was included in Decision Support System for Agrotechnology Transfer (DSSAT 3.1), and updated in DSSAT 4.5 version. It has been applied in many sites around the world to analyze sugarcane productions systems and advance the knowledge regarding the crop. This research aimed to parameterize and evaluation DSSAT/CANEGRO for 5 Brazilian sugarcane varieties: CTC 4, CTC 7, CTC 20, RB 86 7515 e RB 83 5486. CTC 4, CTC 7 and CTC 20 experimental data came from two experimental fields in the State of Sao Paulo and RB 86-7515 and RB 83-5486 varieties data came from bibliographic references. A local sensibility analysis was performed using weather and soil data from Piracicaba to determine the dependency of simulated variables on changes in key parameters. The model was parameterized using the GLUE (generalized likelihood uncertainity estimation) method. The evaluation of five varieties was done using R², Willmott´s index (D) and root mean square error as statistical indicators. D index ranged from 0,870 and 0,944 for green leaf area index for CTC 4, CTC 7 and CTC 20 varieties. The model evaluation was also conducted for the following variables: stalk height (D=0,80), tillering (D=0,90) and sucrose content (D=0,50). The same was done for RB 83-5486 variety using the following variables: sucrose content and stalk fresh mass. The worst results were obtained for variety RB 86-7515, with the following statistics in the validation procedure: (stalk fresh mass D=0,743; stalk height, D=0,821; tillering, D=0,873; stalk dry mass, D=0,665). DSSAT/CANEGRO performed better for Piracicaba, Pradópolis, Guariba and Porto Feliz (State of Sao Paulo) in comparison to the weather and soil conditions of Couripe (State of Alagoas).

O aumento da importância da cultura da cana-de-açúcar nos últimos anos atraiu de investimentos ao setor sucro-alcooleiro, tornando o planejamento estratégico uma ferramenta essencial na orientação para expansão da cultura em novas áreas e otimização da produção nas áreas tradicionais de cultivo. A modelagem agrícola, por isso, ganhou importância por contribuir neste sentido. Existem diversos modelos de simulação de cana-de-açúcar sendo utilizados em todo o mundo, sendo o DSSAT/CANEGRO um dos mais utilizados. O modelo DSSAT/CANEGRO foi baseado no modelo CERES-MAIZE e desenvolvido na África do Sul. Incluído no conjunto de modelos Decision Support System for Agrotechnology Transfer (DSSAT versão 3.1), e atualizado na versão 4.5 do DSSAT, vem sendo aplicado em diversas regiões do mundo para análise e avanço no conhecimento dos sistemas de produção de cana-deaçúcar. O objetivo do trabalho foi a parametrização e avaliação do modelo calibrado para 5 variedades brasileiras de cana-de-açúcar: CTC 4, CTC 7, CTC 20, RB 86 7515 e RB 83 5486. Foram realizadas medidas em campo experimental em duas localidades para as variedades CTC 4, CTC 7, CTC 20 e obtenção de dados bibliográficos para as variedades RB 86-7515 e RB 83-5486. Uma análise de sensibilidade local, utilizando dados climáticos e pedológicos de Piracicaba foi realizada para determinar a dependência das variáveis simuladas em relação às variações nos parâmetros do modelo. Em seguida, realizou-se a parametrização do modelo utilizando a técnica GLUE (do ingles, generalized likelihood uncertainity estimation). A avaliação das cinco variedades estudadas foi feita utilizando o R²; índice D de Willmott e quadrado médio dos desvios como indicadores estatísticos. Para a variável de índice de área foliar, as variedades CTC 4, CTC 7, CTC 20 apresentaram índice D variando entre 0,870 e 0,944, sendo com indicadores estatísticos similares para as variáveis altura de colmo (D=0,80), perfilhamento (D=0,90) e teor de sacarose (D=0,50). A variedade RB 83-5486 apresentou resultados próximos aos observados para as variáveis de teor de sacarose e massa fresca do colmo. Os piores resultados foram obtidos para a variedade RB 86- 7515, com D=0,743 para massa fresca do colmo, D=0,821 para altura de colmo, D=0,873 para perfilhamento e D=0,665 para massa seca de colmo. O modelo teve melhor desempenho para as condições de Piracicaba, Pradópolis, Guariba e Porto Feliz, no Estado de Sao Paulo, em comparação com as condições de solo e clima de Couripe, Estado de Alagoas. / The importance of sugarcane crop has increasing during the last years, attracting more investments to the sugarcane sector. Thus, the strategic planning has becoming an essential tool to guide the sugarcane expansion to new areas, as well as to help optimize the traditional production systems. To do so, crop modelling has a great usefulness. There are several models for sugarcane crop simulation around the world, and DSSAT/CANEGRO is one of the most used. DSSAT/CANEGRO model was based in the CERES-MAIZE model and developed in South Africa. It was included in Decision Support System for Agrotechnology Transfer (DSSAT 3.1), and updated in DSSAT 4.5 version. It has been applied in many sites around the world to analyze sugarcane productions systems and advance the knowledge regarding the crop. This research aimed to parameterize and evaluation DSSAT/CANEGRO for 5 Brazilian sugarcane varieties: CTC 4, CTC 7, CTC 20, RB 86 7515 e RB 83 5486. CTC 4, CTC 7 and CTC 20 experimental data came from two experimental fields in the State of Sao Paulo and RB 86-7515 and RB 83-5486 varieties data came from bibliographic references. A local sensibility analysis was performed using weather and soil data from Piracicaba to determine the dependency of simulated variables on changes in key parameters. The model was parameterized using the GLUE (generalized likelihood uncertainity estimation) method. The evaluation of five varieties was done using R², Willmott´s index (D) and root mean square error as statistical indicators. D index ranged from 0,870 and 0,944 for green leaf area index for CTC 4, CTC 7 and CTC 20 varieties. The model evaluation was also conducted for the following variables: stalk height (D=0,80), tillering (D=0,90) and sucrose content (D=0,50). The same was done for RB 83-5486 variety using the following variables: sucrose content and stalk fresh mass. The worst results were obtained for variety RB 86-7515, with the following statistics in the validation procedure: (stalk fresh mass D=0,743; stalk height, D=0,821; tillering, D=0,873; stalk dry mass, D=0,665). DSSAT/CANEGRO performed better for Piracicaba, Pradópolis, Guariba and Porto Feliz (State of Sao Paulo) in comparison to the weather and soil conditions of Couripe (State of Alagoas).

A energia radiante absorvida pelas copas tem aplicação em estudos de fotossíntese e transpiração de plantas arbóreas. Sua determinação não é simples. Nas últimas duas décadas tem sido realizados trabalhos sobre interceptação de radiação por espécies arbóreas isoladas ou em renques, envolvendo tanto radiação de ondas curtas como de ondas longas (saldo de radiação, Rn). Uma técnica de medida de Rn utiliza radiômetros movendo-se em torno da copa de uma árvore (geometria esférica de medida) ou ao longo de copas de um renque (geometria cilíndrica). Uma alternativa é o uso de modelagem físico-matemática para estimativa de Rn, que também exige medidas para testar modelos. Considerando os poucos trabalhos nessa linha de pesquisa, o presente estudo objetivou: a) avaliar o desempenho de sistemas móveis de integração espaçotemporal de medidas de Rn em renque de um cafezal e de um pomar de limeira ácida localizados no Campus Luiz de Queiroz, USP, Piracicaba, SP; b) estabelecer relações do saldo de radiação das copas de cafeeiros (Rnc) e limeiras ácidas (Rnl) com o saldo de radiação de gramado (Rng) e irradiância solar global (Rg); c) avaliar o desempenho de três modelos físicomatemáticos de estimativa de Rn por comparação com as medidas realizadas com varredura da copa pelos saldo-radiômetros movimentando-se ao longo dos renques (geometria cilíndrica de medidas). Os estudos foram realizados nas quatro estações do ano e em três no pomar, iniciandose respectivamente no outono e no verão de 2008. Os sistemas integradores forneceram valores diários de magnitude coerentes com os de calor latente de vaporização da copa determinados por medidas de transpiração de uma árvore. O desempenho do sistema de grande porte utilizado no pomar exige testes adicionais. O curso diário de Rn em cada posição dos radiômetros em torno da copa mostraram padrões diferenciados conforme orientação do renque e época do ano. No cafezal foram encontradas muito boas relações de Rnc com Rg e Rng nas escalas de 15min, horária e diurna, exceto no verão para 15min e horária e no inverno, na escala diurna, para Rng. Para o pomar, as relações de Rnl com Rg e Rng foram boas nas três escalas temporais, exceto no verão para 15min e horária e no outono na escala diurna. Para o cafezal o modelo de Beer apresentou altos índices de confiabilidade nas épocas do ano, nas escalas horária e diurna; e no pomar se mostrou confiável para inverno, mas não no verão na escala diurna e na escala horária apresentou menor confiabilidade. O modelo de Pilau apresentou boa confiabilidade para o cafezal no outono e menor confiabilidade nas demais estações, enquanto na escala horária verificou-se boa confiabilidade no outono e inverno; para o pomar, o desempenho foi satisfatório para inverno nas duas escalas temporais e insatisfatório para verão na escala diurna. Para o cafezal o modelo de Oyarzun mostrou-se confiável na escala diurna, e menos confiável na escala horária no verão; para o pomar, apresentou boa confiabilidade em ambas escalas temporais no inverno e baixa confiabilidade no verão. / The amount of radiant energy absorbed by the canopy of trees has great application in photosynthesis and transpiration research. Studies of radiation interception by the canopies of an isolated or a sequence of trees in hedgerows have been carried out by a technique which employs moving radiometers around an isolated tree (spherical geometry of measurement) or along a part of a hedgerow (cylindrical geometry). This technique allows spatiotemporal integration of the values. Crop Modeling is an alternative approach to estimate the radiation balance of canopies (Rn). The present study had the following objectives: a) to evaluate the performance of systems for spatiotemporal integration of Rn measurements in hedgerows of a coffee and an acid lime crops at the Luiz de Queiroz Campus of the University of São Paulo, in Piracicaba, SP, Brazil; b) to establish relations of the net radiation of the coffee (Rnc) and of acid lime plants (Rnl) canopies with grass net radiation (Rng) and incoming solar radiation (Rg); c) to evaluate the performance of three models of Radiation Balance (Rn), by comparing modem results with measurements by the moving radiometers along the planting row (cylindrical geometry). The coffee experiment was carried out all year long starting in the fall of 2008 while for the orchard only three seasons were tested starting in the summer 2008. Measurements of Rn by the moving systems were considered reliable, but it is recommended further testing of that of the orchard. The daily course of Rn for each measurement around the canopy showed different patterns for the two crops, in agreement with planting row orientation and season. Concerning to relations of Rnc with Rg and Rng integrated for 15min, hourly and diurnal periods, they were very good for the coffee crop in the three timescales, excepting for the 15min and hourly timescales in the summer and for the Rng for the diurnal period in the winter. For the orchard, the relations of Rnl with Rg and Rng were good, excepting for the summer in the 15min and hourly and in autumn for the diurnal timescales. The Beer´s model showed good reliability for the coffee crop; while for the orchard it showed to be reliable in the winter, but not in the summer for the diurnal period; for the hourly periods it was less reliable. The Pilau´s model showed good reliability for the coffee crop for both the diurnal and hourly periods during the fall, but less reliability in the other seasons; for the orchard, its performance was reliable for the winter in the two timescales and not reliable for the summer in the diurnal scale. The Oyarzun´s model showed to be reliable in the diurnal timescale and less reliable in the hourly timescale for the summer; for the orchard, it showed good reliability for the winter and little reliability for the summer.

A área de pesquisa científica, denominada computação e biodiversidade, tem por desafio suprir a demanda por tecnologias de apoio à conservação da biodiversidade, fornecendo ferramentas computacionais aplicadas ao estudo da biodiversidade. Os modelos de distribuição geográfica de espécies têm uma importante contribuição para o entendimento da biodiversidade e no apoio para a tomada de decisão em conservação dos recursos de biodiversidade. Os resultados da modelagem de distribuição de espécies, não apresentam com clareza os tratamentos aplicados aos dados na etapa de pré-análise e os critérios de seleção das variáveis preditoras, o que agrava a incerteza quanto aos resultados obtidos e prejudica a reprodutibilidade do experimento. Assim, o objetivo desta pesquisa é apresentar o processo de modelagem de distribuição de espécie com destaque para as atividades de pré-análise e atividades de seleção das variáveis preditoras. O processo de modelagem de distribuição de espécies proposto é avaliado por meio de estudo de caso de Modelagem de Distribuição de Espécies do polinizador Centris hyptidis e das plantas oleaginosas Angelonia campestris e Angelonia cornigera, que considera como fator biótico a especialização da interações entre essas plantas e o polinizador. Neste estudo de caso pode-se verificar que a aplicação de técnicas estatísticas exploratórias de dados na etapa de pré-análise do processo de modelagem distribuição de espécies permite a avaliação da qualidade dos pontos de ocorrência, essenciais para o desempenho preditivo do modelo final. / The area of research called biodiversity informatics, or bioinformatics, has to face the challenge of meeting the demand for technologies to support the conservation of biodiversity, providing computational tools applied to the study of biodiversity. The models of geographic distribution of species have a fundamental implication for understanding the biodiversity and conservation decision making. Researches on modeling species distributions generally do not present clearly the treatments applied to the data in the pre-analysis and criteria for selection of predictor variables, which increases the uncertainty regarding the results and affect the reproducibility of experiment. The objective of this research is to present the process of species distribution modeling with emphasis on the activities of pre-analysis and activities selection of the predictor variables, such as to favor its repeatability and reproducibility by other researchers. The process of modeling species distribution proposed is validated on a case study of modeling distribution of pollinator species Centris hyptidis and oilseed Angelonia campestris and Angelonia cornigera, a biotic factor that considers the specialization of these interactions between plants and pollinators. In this case study we can observe one of the main contributions of this work: the application of statistical techniques for data exploration in the pre-analysis of species distribution modeling process, with improved capacity for evaluation and selection of points of occurrence essential to the performance of the predictive model.

Of all of the desktop software available, that for the implementation of a database is some of the most complex. With the increasing number of computer users having access to this sophisticated software, but with no obvious way to learn the rudiments of data modelling for the implementation of a database, there is a need for a simple, convenient method to improve their understanding. The research described in this thesis represents the first steps in the development of a tool to accomplish this improvement. In a preliminary study using empirical research a conceptual model was used to improve novice end users’ understanding of the relational concepts of data organisation and the use of a database software package. The results showed that no conclusions could be drawn about either the artefact used or the method of evaluation. Following the lead of researchers in the fields of both education and information systems, a design research process was developed, consisting of the construction and evaluation of a training artefact. A combination of design research and a design experiment was used in the main study described in this thesis. New to research in information systems, design research is a methodology or set of analytical techniques and perspectives, and this was used to develop a process (development of an artefact) and a product (the artefact itself). The artefact, once developed, needed to be evaluated for its effectiveness, and this was done using a design experiment. The experiment involved exposing the artefact to a small group of end users in a realistic setting and defining a process for the evaluation of the artefact. The artefact was the tool that would facilitate the improvement of the understanding of data modelling, the vital precursor to the development of a database. The research was conducted among a group of novice end users who were exposed to the artefact, facilitated by an independent person. In order to assess whether there was any improvement in the novices’ understanding of relational data modelling and database concepts, they then completed a post-test. Results confirmed that the artefact, trialled through one iteration, was successful in improving the understanding of these novice end users in the area of data modelling. The combination of design research and design experiment as described above gave rise to a new methodology, called experimental design research at this early juncture. The successful outcome of this research will lead to further iterations of the design research methodology, leading in turn to the further development of the artefact which will be both useful and accessible to novice users of personal computers and database software. This research has made the following original contributions. Firstly, the use of the design research methodology for the development of the artefact, which proved successful in improving novice users’ understanding of relational data structures. Secondly, the novel use of a design experiment in an information systems project, which was used to evaluate the success of the artefact. And finally, the combination of the developed artefact followed by its successful evaluation using a design experiment resulted in the hybrid experimental design research methodology. The success of the implementation of the experimental design research methodology in this information systems project shows much promise for its successful application to similar projects.

Managing, handling, exchanging and accessing hydrogeological information depend mainly on the applied hydrogeological data models, which differ between institutions and across countries. Growing interest in hydrogeological information diffusion, combined with a need for information availability, require the convergence of hydrogeological data models. Model convergence makes hydrogeological information accessible to multiple institutions, universities, administration, water suppliers, and research organisations, at different levels: from the local level (on-site measurement teams), to national and international institutions dealing with water resources management. Furthermore, because hydrogeological studies are complex, they require a large variety of high-quality hydrogeological data with appropriate metadata in clearly designed and coherent structures. To respond to the requirement of model convergence, easy information exchange and hydrogeological completeness, new data models have been developed, using two different methodologies. At local-regional level, the HydroCube model has been developed for the Walloon Region in Belgium. This logical data model uses entity-relationship diagrams and it has been implemented in the MS Access environment, further enriched with a fully functional user-interface. The HydroCube model presents an innovative holistic project-based approach, which covers a full set of hydrogeological concepts and features, allowing for effective hydrogeological project management. This approach enables to store data about the project localisation, hydrogeological equipment, related observations and measurements. Furthermore, topological relationships facilitate management of spatially associated data. Finally, the model focuses on specialized hydrogeological field experiments, such as pumping tests and tracer tests. At the international level, a new hydrogeological data model has been developed which guarantees hydrogeological information availability in one standard format in the scope of the FP6 project GABARDINE (Groundwater Artificial recharge Based on Alternative sources of wateR: aDvanced Integrated technologies and management). The model has been implemented in the ArcGIS environment, as a Geospatial Database for a decision support system. The GABARDINE Geospatial Database uses advantages of object-oriented modelling (UML), it follows standards for geoscientific information exchange (ISO/TC211 and OGC), and it is compliant with the recommendations from the European Geospatial Information Working Group. Finally, these two developed models have been tested with hydrogeological field data on different informatics platforms: from MS Access, through a proprietary ArcGIS environment, to the open source, free Web2GIS on-line application. They have also contributed to the development of the GroundWater Markup Language (GWML) Canadian exchange standard, compliant with Geographic Markup Language (GML). GWML has the potential of becoming an international HydroGeology Markup Language (HgML) standard with a strong and continuous support from the hydrogeological community.

“In an ill-structured domain you cannot, by definition, have a pre-compiled schema in your mind for every circumstance and context you may find ... you must be able to flexibly select and arrange knowledge sources to most efficaciously pursue the needs of a given situation.” [57] / In order to interact and collaborate effectively, agents, whether human or software, must be able to communicate through common understandings and compatible conceptualisations. Ontological differences that occur either from pre-existing assumptions or as side-effects of the process of specification are a fundamental obstacle that must be overcome before communication can occur. Similarly, the integration of information from heterogeneous sources is an unsolved problem. Efforts have been made to assist integration, through both methods and mechanisms, but automated integration remains an unachieved goal. Communication and information integration are problems of meaning and interaction, or semantic interoperability. This thesis contributes to the study of semantic interoperability by identifying, developing and evaluating three approaches to the integration of information. These approaches have in common that they are lightweight in nature, pragmatic in philosophy and general in application. / The first work presented is an effort to integrate a massive, formal ontology and knowledge-base with semi-structured, informal heterogeneous information sources via a heuristic-driven, adaptable information agent. The goal of the work was to demonstrate a process by which task-specific knowledge can be identified and incorporated into the massive knowledge-base in such a way that it can be generally re-used. The practical outcome of this effort was a framework that illustrates a feasible approach to providing the massive knowledge-base with an ontologically-sound mechanism for automatically generating task-specific information agents to dynamically retrieve information from semi-structured information sources without requiring machine-readable meta-data. / The second work presented is based on reviving a previously published and neglected algorithm for inferring semantic correspondences between fields of tables from heterogeneous information sources. An adapted form of the algorithm is presented and evaluated on relatively simple and consistent data collected from web services in order to verify the original results, and then on poorly-structured and messy data collected from web sites in order to explore the limits of the algorithm. The results are presented via standard measures and are accompanied by detailed discussions on the nature of the data encountered and an analysis of the strengths and weaknesses of the algorithm and the ways in which it complements other approaches that have been proposed. / Acknowledging the cost and difficulty of integrating semantically incompatible software systems and information sources, the third work presented is a proposal and a working prototype for a web site to facilitate the resolving of semantic incompatibilities between software systems prior to deployment, based on the commonly-accepted software engineering principle that the cost of correcting faults increases exponentially as projects progress from phase to phase, with post-deployment corrections being significantly more costly than those performed earlier in a project’s life. The barriers to collaboration in software development are identified and steps taken to overcome them. The system presented draws on the recent collaborative successes of social and collaborative on-line projects such as SourceForge, Del.icio.us, digg and Wikipedia and a variety of techniques for ontology reconciliation to provide an environment in which data definitions can be shared, browsed and compared, with recommendations automatically presented to encourage developers to adopt data definitions compatible with previously developed systems. / In addition to the experimental works presented, this thesis contributes reflections on the origins of semantic incompatibility with a particular focus on interaction between software systems, and between software systems and their users, as well as detailed analysis of the existing body of research into methods and techniques for overcoming these problems.

Os modelos de distribuição de espécies (MDEs) utilizam dados de ocorrência de campo e variáveis ambientais para indicar locais adequados para a ocorrência de uma espécie. Apesar dos inúmeros trabalhos que avaliam diversos aspectos deste tipo de modelagem, a aplicação desses métodos para predição do distribuição potencial de espécies de aves em escala local para o Parque Estadual da Serra do Mar (PESM) São Paulo, ainda não foi avaliada. Este trabalho analisa o potencial dos MDEs para o referido parque a partir das variáveis ambientais disponíveis para modelagem da área. Para isso, foram desenvolvidos modelos para 23 espécies de aves a partir de dados coletados em levantamento sistemático da avifauna do Núcleo São Sebastião um dos 8 núcleos que integram o PESM. Numa primeira etapa foi utilizado o algoritmo Maxent, por ser um dos mais utilizados segundo a literatura e por sua habilidade em realizar predições a partir de informações incompletas. A seguir a performance do Maxent foi comparada com as performances dos algoritmos GARP e Enviromental Distance. O parâmetro utilizado para analisar o desempenho e definir os melhores modelos preditivos foi a analise do valor de AUC. Os modelos ainda foram avaliados através da matriz de confusão, para estabelecer a taxa de omissão, e realizar o Teste Binomial Duas proporções do BioEstat 5.0 a fim de estabelecer a probabilidade dos acertos destes modelos serem diferentes do acaso. E finalmente foi determinada a área mínima estimada, baseado na premissa que um bom modelo deve prever a menor área possível com o maior numero de acertos ao mesmo tempo. Apesar de existir um número reduzido de camadas disponíveis para modelagem da área em escala local 8 mapas - O Maxent, teve o melhor desempenho entre os algoritmos testados, esse algoritmo se mostrou capaz de gerar mapas estatisticamente bons e biologicamente confiáveis para aves, nas áreas estudadas. São necessários cuidados específicos com relação a escolha das espécies a serem modeladas, assegurando que os camadas disponíveis representem premissas das necessidades ecológicas dessas espécies. Estudos posteriores sobre as respostas do algoritmo a diferentes quantidades e padrões de distribuição dos dados de ocorrência podem elucidar a aplicabilidade do algoritmo como ferramenta para auxiliar nos trabalhos de conservação. / Species Distribution Models (SDM) use Field occurrence data and environmental variables to indicate adequate places for the presence of a species. Despite a number of papers evaluate various aspects on these methodologies, the application of such models in order to predict the potential niche for bird species in Parque Estadual da Serra do Mar (PESM) in Sao Paulo, hasnt been analyzed yet. This paper examines the potential of SDM for the referred park from available variable to the modelling of the area. In order to do so, models were developed for 23 bird species form data collected in a systematic survey in the avifauna in Núcleo Sao Sebastião which is one of the 8 centers that are part of PESM. At first, Maxent algorithm was used as it is one of the most used algorithms according to current literature and due to its ability to accomplish predictions from incomplete information. Maxent performance was then compared to GARP and Environmental Distance algorithms performances. The parameter used to analyze skill and define the best predicting models was AUC value analyses. They were also evaluated though the confusion matrix as to establish the omission rate and take the binomial test of two proportions in order to define the correctness of such models as being different from chance and finally the estimated minimum area was determined based on the premise that a good model must predict the smallest possible area with the biggest number of correctness allowed. Although there are a few number of layers for modelling the area in local scale (just 8 maps), maxent results to be potentially useful for generating maps for potential distribution of the birds from the park. It was the best result shown by the 3 algorithms methods tried out. More studies are necessary as to evaluate the algorithms behavior from a smaller number of records occurrences well as different patterns en these records for the definition of the best method of distribution modelling of the avifauna in the park. Maxent proved itself to be capable of generating maps that are mathematically good and ecologically reliable for the birds in studied areas, It was the best result shown shown by the 3 algorithms methods tried out. Specific care is necessary in relation to the species to be modeled so as to assure the feasible layers represent premises of the ecological needs for such species. Further studies on the algorithm responses to different quantities and distributions patterns of occurrence data may elucide its applicability as a tool to help conservation work.