Data sets of increasing size and complexity require new approaches for prediction as the sheer volume of data from disparate sources inhibits joint processing and modeling. Rather modular segmentation is required, in which a set of models process (potentially overlapping) partitions of the data to independently construct predictions. This framework enables individuals models to be tailored for specific selective superiorities without concern for existing models, which provides utility in cases of segmented expertise. However, a method for fusing predictions from the collection of models is required as models may be correlated. This work details optimal principles for fusing binary predictions from a collection of models to issue a joint prediction. An efficient algorithm is introduced and compared with off the shelf methods for binary prediction. This framework is then implemented in an applied setting to predict instances of civil unrest in Central and South America. Finally, model fusion principles of a spatiotemporal nature are developed to predict civil unrest. A novel multiscale modeling is used for efficient, scalable computation for combining a set of spatiotemporal predictions. / Ph. D.

[pt] Muitos fatores podem in uenciar a velocidade de um veículo numa rodovia ou estrada, mas dois deles são observados diariamente pelos motoristas: sua localização e o momento do dia. Obter modelos que retornem a velocidade média como uma função do espaço e do tempo é ainda uma tarefa desafiadora. São muitas as aplicações para esses tipos de modelos, como por exemplo: tempo estimado de chegada, caminho mais curto e previsão de tráfico, deteccção de acidente, entre outros. Este estudo propõe um modelo de previsão baseado em uma média espaço-temporal da velocidade média/instantânea coletada de dados históricos de GPS. A grande vantagem do modelo proposto é a sua simplicidade. Além disso, os resultados experimentais obtidos de caminhões de entrega de combustíveis, por todo o ano de 2013 no Brasil, indicaram que a maioria das observações podem ser preditas usando esse modelo dentro de uma tolerância de erro aceitável. / [en] Many factors may inuence a vehicle speed in a road, but two of them are usually observed by many drivers: its location and the time of the day. To obtain a model that returns the average speed as a function of position and time is still a challenging task. The application of such models can be in different scenarios, such as: estimated time of arrival, shortest route paths, traffic prediction, and accident detection, just to cite a few. This study proposes a prediction model based on a spatio-temporal partition and mean/instantaneous speeds collected from historic GPS data. The main advantage of the proposed model is that it is very simple to compute. Moreover, experimental results obtained from fuel delivery trucks, along the whole year of 2013 in Brazil, indicate that most of the observations can be predicted using this model within an acceptable error tolerance.

Doctor of Philosophy / Department of Biological & Agricultural Engineering / Stacy L. Hutchinson / The Soil Conservation Service-Curve Number (SCS-CN) method is widely used to estimate direct runoff from rainfall events; however, the method does not account for the dynamic rainfall-runoff relationship. This study used back-calculated curve numbers (CNs) and Normalized Difference Vegetation Index (NDVI) to develop NDVI-based CNs (CN[subscript]NDV) using four small northeastern Kansas grassland watersheds with average areas of 1 km² and twelve years (2001–2012) of daily precipitation and runoff data. Analysis indicated that the CN[subscript]NDVI model improved runoff predictions compared to the SCS-CN method. The CN[subscript]NDVI also showed greater variability in CNs, especially during growing season, thereby increasing the model’s ability to estimate relatively accurate runoff from rainfall events since most rainfall occurs during the growing season. The CN[subscript]NDVI model was applied to small, disturbed grassland watersheds to assess the model’s ability to detect land cover change impact for military maneuver damage and large, diverse land use/cover watersheds to assess the impact of scaling up the model. CN[subscript]NDVI application was assessed using a paired watershed study at Fort Riley, Kansas. Paired watersheds were identified through k-means and hierarchical-agglomerative clustering techniques. At the large watershed scale, Daymet precipitation was used to estimate runoff, which was compared to direct runoff extracted from stream flow at gauging points for Chapman (grassland dominated) and Upper Delaware (agriculture dominated) watersheds. In large, diverse watersheds, CN[subscript]NDVI performed better in moderate and overall flow years. Overall, CN[subscript]NDVI more accurately simulated runoff compared to SCS-CN results: The calibrated model increased by 0.91 for every unit increase in observed flow (r = 0.83), while standard CN-based flow increased by 0.506 for every unit increase in observed flow (r = 0.404). Therefore, CN[subscript]NDVI could help identify land use/cover changes and disturbances and spatiotemporal changes in runoff at various scales. CN[subscript]NDVI could also be used to accurately estimate runoff from precipitation events in order to instigate more timely land management decisions.

No description available.

Les zones humides sont écosystèmes reconnus de vitale importance pour la conservation de la biodiversité et pour un développement soutenable. En Colombie, 26 % du territoire continental national est couvert de ces écosystèmes. Le complexe de zones humides urbaines de Bogota, en fait partie, avec 15 écosystèmes, dont la Convention Ramsar reconnaît 11. Ils sont uniques et jouent un rôle important dans l’approvisionnement des services écosystèmes à la zone urbaine. Cependant, ces écosystèmes urbains font face à de nombreux défis en raison de leur emplacement. Les causes et les conséquences de leur transformation sont très complexes. En appliquant des approches des systèmes complexes, sa dynamique de changement peut être étudiée. Les automates cellulaires sont l’une des techniques largement utilisées dans la modélisation de la dynamique spatiotemporelle des changements de l’usage et de l’occupation des sols. Cette étude propose l’analyse et la simulation des zones humides urbaines en appliquant une approche hybride par un modèle couplé de chaîne de Markov, de réseaux de neurones artificiels et d’automates cellulaires, afin d’estimer leurs changements d’étendue pour les années 2016, 2022, 2028 et 2034 dans la ville de Bogota, en Colombie. Pour extraire le changement d’occupation et d’utilisation du sol, trois images analogues des années 1998, 2004 et 2010 ont été a utilisées. Les résultats ont montré une diminution de 0,30 % de la couverture des zones humides en douze ans. De plus, les résultats suggèrent que la couverture des zones humides représentera 1,97 % de la zone d’étude totale en 2034, représentant une probabilité de diminution de 14 % en 24 ans. D’ailleurs, en appliquant l’analyse d’intensité, il a été constaté que le gain de cultures et de pâturages cible la perte de zones humides. Bien dont ces écosystèmes soient protégés et d’utilisation restreinte, leur patron de réduction se poursuivra en 2034. La pertinence de ce projet réside dans sa contribution potentielle au processus décisionnel au sein de la ville et en tant qu’instrument de gestion des ressources naturelles. En outre, les résultats de cette étude pourraient aider à atteindre l’objectif de développement durable 6 « Eau propre et assainissement » et l’atténuation du changement climatique. / Wetlands are ecosystems recognized as being of vital importance for the conservation of biodiversity and for sustainable development. In Colombia, 26% of the national continental territory is covered by these ecosystems. The complex of urban wetlands of Bogota is one of them, with 15 ecosystems, of which the Ramsar Convention recognizes 11. They are unique and play an important role in providing ecosystem services to the urban area. However, these urban ecosystems face many challenges due to their location. The causes and consequences of their transformation are very complex. By applying complex systems approaches, the dynamics of change can be studied. Cellular automata is one of the widely used techniques in modeling the spatiotemporal dynamics of land use and land cover changes. This study proposes the analysis and simulation of urban wetlands by applying a hybrid approach through a coupled model of the Markov chain, artificial neural networks, and cellular automata, in order to estimate the extent of changes for the years 2016, 2022, 2028, and 2034 in the city of Bogota, Colombia. To extract the change in land cover and land use, three analogous images from the years 1998, 2004, and 2010 were used. The results showed a 0.30% decrease in wetland coverage in twelve years. Furthermore, the results suggest that wetland cover will be 1.97% of the total study area in 2034, representing a 14% probability of a decrease in 24 years. Moreover, by applying the intensity analysis, it was found that the gain of crop and pastureland targets the loss of wetlands. Although these ecosystems are protected and of limited use, their pattern of reduction will continue in 2034. The relevance of this project lies in its potential contribution to decision-making within the city and as a natural resource management tool. In addition, the results of this study could help achieve Sustainable Development Goal 6 “Clean Water and Sanitation” and climate change mitigation.