As sensor data becomes more and more available, there is an increasing interest in assimilating real time sensor data into spatial temporal simulations to achieve more accurate simulation or prediction results. Particle Filters (PFs), also known as Sequential Monte Carlo methods, hold great promise in this area as they use Bayesian inference and stochastic sampling techniques to recursively estimate the states of dynamic systems from some given observations. However, PFs face major challenges to work effectively for complex spatial temporal simulations due to the high dimensional state space of the simulation models, which typically cover large areas and have a large number of spatially dependent state variables. As the state space dimension increases, the number of particles must increase exponentially in order to converge to the true system state. The purpose of this dissertation work is to develop localized particle filtering to support PFs-based data assimilation for large-scale spatial temporal simulations. We develop a spatially dependent particle-filtering framework that breaks the system state and observation data into sub-regions and then carries out localized particle filtering based on these spatial regions. The developed framework exploits the spatial locality property of system state and observation data, and employs the divide-and-conquer principle to reduce state dimension and data complexity. Within this framework, we propose a two-level automated spatial partitioning method to provide optimized and balanced spatial partitions with less boundary sensors. We also consider different types of data to effectively support data assimilation for spatial temporal simulations. These data include both hard data, which are measurements from physical devices, and soft data, which are information from messages, reports, and social network. The developed framework and methods are applied to large-scale wildfire spread simulations and achieved improved results. Furthermore, we compare the proposed framework to existing particle filtering based data assimilation frameworks and evaluate the performance for each of them.

Simulation software is being increasingly used in various domains for system analysis and/or behavior prediction. Traditionally, researchers and field experts need to have access to the computers that host the simulation software to do simulation experiments. With recent advances in cloud computing and Software as a Service (SaaS), a new paradigm is emerging where simulation software is used as services that are composed with others and dynamically influence each other for service-oriented simulation experiment on the Internet. The new service-oriented paradigm brings new research challenges in composing multiple simulation services in a meaningful and correct way for simulation experiments. To systematically support simulation software as a service (SimSaaS) and service-oriented simulation experiment, we propose a layered framework that includes five layers: an infrastructure layer, a simulation execution engine layer, a simulation service layer, a simulation experiment layer and finally a graphical user interface layer. Within this layered framework, we provide a specification for both simulation experiment and the involved individual simulation services. Such a formal specification is useful in order to support systematic compositions of simulation services as well as automatic deployment of composed services for carrying out simulation experiments. Built on this specification, we identify the issue of mismatch of time granularity and event granularity in composing simulation services at the pragmatic level, and develop four types of granularity handling agents to be associated with the couplings between services. The ultimate goal is to achieve standard and automated approaches for simulation service composition in the emerging service-oriented computing environment. Finally, to achieve more efficient service-oriented simulation, we develop a profile-based partitioning method that exploits a system’s dynamic behavior and uses it as a profile to guide the spatial partitioning for more efficient parallel simulation. We develop the work in this dissertation within the application context of wildfire spread simulation, and demonstrate the effectiveness of our work based on this application.

Simulation software is being increasingly used in various domains for system analysis and/or behavior prediction. Traditionally, researchers and field experts need to have access to the computers that host the simulation software to do simulation experiments. With recent advances in cloud computing and Software as a Service (SaaS), a new paradigm is emerging where simulation software is used as services that are composed with others and dynamically influence each other for service-oriented simulation experiment on the Internet. The new service-oriented paradigm brings new research challenges in composing multiple simulation services in a meaningful and correct way for simulation experiments. To systematically support simulation software as a service (SimSaaS) and service-oriented simulation experiment, we propose a layered framework that includes five layers: an infrastructure layer, a simulation execution engine layer, a simulation service layer, a simulation experiment layer and finally a graphical user interface layer. Within this layered framework, we provide a specification for both simulation experiment and the involved individual simulation services. Such a formal specification is useful in order to support systematic compositions of simulation services as well as automatic deployment of composed services for carrying out simulation experiments. Built on this specification, we identify the issue of mismatch of time granularity and event granularity in composing simulation services at the pragmatic level, and develop four types of granularity handling agents to be associated with the couplings between services. The ultimate goal is to achieve standard and automated approaches for simulation service composition in the emerging service-oriented computing environment. Finally, to achieve more efficient service-oriented simulation, we develop a profile-based partitioning method that exploits a system’s dynamic behavior and uses it as a profile to guide the spatial partitioning for more efficient parallel simulation. We develop the work in this dissertation within the application context of wildfire spread simulation, and demonstrate the effectiveness of our work based on this application.

Les forêts sont une ressource naturelle importante sur le plan écologique, culturel et économique, et sont confrontées à des défis croissants en raison des changements climatiques. Ces défis sont difficiles à prédire en raison de la nature complexe des interactions entre le climat et la végétation, dont une le feu. Compte tenu de l’importance des écosystèmes forestiers, des dangers potentiels des feux de forêt et de la complexité de leurs interactions, il est primordial d'acquérir une compréhension de ces systèmes à travers le prisme de la science des systèmes complexes. La science des systèmes complexes et ses techniques de modélisation associées peuvent fournir des informations sur de tels systèmes que les techniques de modélisation traditionnelles ne peuvent pas. Là où les techniques statistiques et basées sur équations cherchent à contourner la dynamique non-linéaire, auto-organisée et émergente des systèmes complexes, les approches de modélisation telles que les automates cellulaires et les modèles à base d'agents (MBA) embrassent cette complexité en cherchant à reproduire les interactions clés de ces systèmes. Bien qu'il existe de nombreux modèles de comportement du feu qui tiennent compte de la complexité, les MBA offrent un terrain d'entente entre les modèles de simulation empiriques et physiques qui peut fournir de nouvelles informations sur le comportement et la simulation du feu. Cette étude vise à améliorer notre compréhension du feu dans le contexte de la science des systèmes complexes en développant un tel MBA de propagation du feu. Le modèle utilise des données de type de carburant, de terrain et de météo pour créer l'environnement des agents. Le modèle est évalué à l'aide d’une étude de cas d'un incendie naturel qui s'est produit en 2001 dans le sud-ouest de l'Alberta, au Canada. Les résultats de cette étude confirment la valeur de la prise en compte de la complexité lors de la simulation d'incendies de forêt et démontrent l'utilité de la modélisation à base d'agents pour une telle tâche. / Forests are an ecologically, culturally, and economically important natural resource that face growing challenges due to climate change. These challenges are difficult to predict due to the complex nature of the interactions between climate and vegetation. Furthermore, fire is intrinsically linked to both climate and vegetation and is, itself, complex. Given the importance of forest ecosystems, the potential dangers of forest fires, and the complexity of their interactions, it is paramount to gain an understanding of these systems through the lens of complex systems science. Complex systems science and its attendant modeling techniques can provide insights on such systems that traditional modelling techniques cannot. Where statistical and equation-based techniques seek to work around the non-linear, self-organized, and emergent dynamics of complex systems, modelling approaches such as Cellular Automata and Agent-Based Models (ABM) embrace this complexity by seeking to reproduce the key interactions of these systems. While there exist numerous models of fire behaviour that account for complexity, ABM offers a middle ground between empirical and physical simulation models that may provide new insights into fire behaviour and simulation. This study seeks to add to our understanding of fire within the context of complex systems science by developing such an ABM of fire spread. The model uses fuel-type, terrain, and weather data to create the agent environment. The model is evaluated with a case study of a natural fire that occurred in 2001 in southwestern Alberta, Canada. Results of this study support the value of considering complexity when simulating forest fires and demonstrate the utility of ABM for such a task.