In this thesis, we study a class of multi-stable reaction-diffusion systems used to model competing species. Systems in this class possess uniform stable steady states representing semi-trivial solutions. We start by considering a bistable, interaction, where the interactions are of classic “Lotka-Volterra” type and we consider a particular problem with relevance to applications in population dynamics: essentially, we study under what conditions the interplay of relative motility (diffusion) and competitive strength can cause waves of invasion to be halted and reversed. By establishing rigorous results concerning related degenerate and near-degenerate systems,we build a picture of the dependence of the wave speed on system parameters. Our results lead us to conjecture that this class of competition model has three “zones of response” in which the wave direction is left-moving, reversible and right-moving, respectively and indeed that in all three zones, the wave speed is an increasing function of the relative motility. Moreover, we study the effects of domain size on planar and non-planar interfaces and show that curvature plays an important role in determining competitive outcomes. Finally, we study a 3-species Lotka-Volterra model, where the third species is treated as a bio-control agent or a bio-buffer and investigate under what conditions the third species can alter the existing competition interaction.

Throughout history species has gone extinct due to anthropogenic activities. During the last century efforts have been done to reintroduce species back into the wild. Zoos that originally were created as amusement parks for people have today a new purpose; to keep and breed species in captivity for later reintroductions in the wild. However a relaxed environment such as a zoo leads to a general fitness decline of up to 40% per generation in captivity. The probability of a successful reintroduction of a species that has been bred in a zoo will be lower the longer time it has been kept in captivity. The reintroduction of a captive bred species can also cause secondary extinctions and other negative effects on the food-web. Both changes in the community caused by the loss of a species and changes in the species itself caused by captivity can be expected to affect the outcome of a reintroduction attempt. Using a modeling approach I here investigate how the reintroduction of a captive bred species (at three different trophic levels; basal, intermediate and top predator species) affects a food-web and what risks there are in reintroducing it. A Lokta-Volterra model with type II functional response is used. I investigate three scenarios: the reintroduction of a species with 0% change in its attributes, 40% change in its attributes and 75% change in its attributes. It was found that the most important factor for reintroduction success when reintroducing a species is whether it is a producer species (basal species) or a consumer species (intermediate and top predator species). The producer species were most sensitive to the changes in their attributes, whilst consumer species were more sensitive to change in the food-web (Euclidian distance). The producer species were found to cause most secondary extinctions in all scenarios, hence indicating that it is a bottom-up controlled food-web. The present study suggests that the success of a reintroduction attempt is affected both by the degree of changes in the food web caused by the initial loss of the species and the degree of change in the species itself caused by captivity.

Submitted by SIDNEY BRUCE SHIKI null (sbshiki@gmail.com) on 2016-04-02T14:43:01Z No. of bitstreams: 1 Thesis_PhD_SBShiki.pdf: 10090180 bytes, checksum: c44d2ebecbed6d011cf61ceabdfd3494 (MD5) / Approved for entry into archive by Felipe Augusto Arakaki (arakaki@reitoria.unesp.br) on 2016-04-05T14:42:00Z (GMT) No. of bitstreams: 1 shiki_sd_dr_ilha.pdf: 10090180 bytes, checksum: c44d2ebecbed6d011cf61ceabdfd3494 (MD5) / Made available in DSpace on 2016-04-05T14:42:00Z (GMT). No. of bitstreams: 1 shiki_sd_dr_ilha.pdf: 10090180 bytes, checksum: c44d2ebecbed6d011cf61ceabdfd3494 (MD5) Previous issue date: 2016-03-04 / Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) / Estruturas com comportamento não-linear são frequentes em dinâmica estrutural, principalmente considerando componentes parafusados, com juntas, folgas ou estruturas flexíveis sujeitas à grandes deslocamentos. Desse modo, o monitoramento de estruturas com métodos lineares clássicos, como os baseados em parâmetros modais, podem falhar drasticamente em caracterizar efeitos não-lineares. Neste trabalho foi proposta a utilização de séries de Volterra para identificação de sistemas mecânicos não-lineares em aplicações de detecção de danos e quantificação de parâmetros. A propriedade deste modelo de representar separadamente os componentes de resposta linear e não-linear do sistema foi aplicada para se construir índices de dano que evidenciam a necessidade de modelagem não-linear. Além disso métricas de resíduo linear e não-linear dos termos do modelo de Volterra são empregadas para identificar modelos paramétricos da estrutura. As metodologias propostas são ilustradas em bancadas experimentais de modo a evidenciar a importância de fenômenos não-lineares para o monitoramento de estruturas. / Nonlinear structures are frequent in structural dynamics, specially considering screwed components, with joints, clearance or flexible components presenting large displacements. In this sense the monitoring of systems based on classical linear methods, as the ones based on modal parameters, can drastically fail to characterize nonlinear effects. This thesis proposed the use of Volterra series for nonlinear system identification aiming applications in damage detection and parameter quantification. The property of this model of representing the linear and nonlinear components of the response of a system was used to formulate damage features to make clear the need of nonlinear modeling. Also metrics based on the linear and nonlinear residues of the terms of the Volterra model were employed to identify parametric models of the structure. The proposed methodologies are illustrated in experimental setups to show the relevance of nonlinear phenomena in the structural health monitoring. / FAPESP: 2012/04757-6 / FAPESP: 2013/25148-0 / FAPESP: 2012/21195-1 / FAPESP: 2015/03560-2

Orientador: Samuel da Silva / Abstract: Nonlinear structures are frequent in structural dynamics, specially considering screwed components, with joints, clearance or flexible components presenting large displacements. In this sense the monitoring of systems based on classical linear methods, as the ones based on modal parameters, can drastically fail to characterize nonlinear effects. This thesis proposed the use of Volterra series for nonlinear system identification aiming applications in damage detection and parameter quantification. The property of this model of representing the linear and nonlinear components of the response of a system was used to formulate damage features to make clear the need of nonlinear modeling. Also metrics based on the linear and nonlinear residues of the terms of the Volterra model were employed to identify parametric models of the structure. The proposed methodologies are illustrated in experimental setups to show the relevance of nonlinear phenomena in the structural health monitoring. / Resumo: Estruturas com comportamento não-linear são frequentes em dinâmica estrutural, principalmente considerando componentes parafusados, com juntas, folgas ou estruturas flexíveis sujeitas à grandes deslocamentos. Desse modo, o monitoramento de estruturas com métodos lineares clássicos, como os baseados em parâmetros modais, podem falhar drasticamente em caracterizar efeitos não-lineares. Neste trabalho foi proposta a utilização de séries de Volterra para identificação de sistemas mecânicos não-lineares em aplicações de detecção de danos e quantificação de parâmetros. A propriedade deste modelo de representar separadamente os componentes de resposta linear e não-linear do sistema foi aplicada para se construir índices de dano que evidenciam a necessidade de modelagem não-linear. Além disso métricas de resíduo linear e não-linear dos termos do modelo de Volterra são empregadas para identificar modelos paramétricos da estrutura. As metodologias propostas são ilustradas em bancadas experimentais de modo a evidenciar a importância de fenômenos não-lineares para o monitoramento de estruturas. / Doutor