This dissertation presents the results of a multi-year interdisciplinary study of a dynamic coastal dune ecosystem restoration effort in Pacific Rim National Park Reserve in British Columbia, Canada. The research is the result of a collaboration with Parks Canada Agency (PCA) who, under the Species at Risk Act (SARA), are mandated to restore habitat for SARA listed species within the dune complex. In response, PCA committed to, and implemented, a dynamic dune ecosystem restoration program that involved widespread removal of invasive vegetation (Ammophila spp.), transplanting of native vegetation, introduction of an endangered species, and volunteer programs to prevent re-growth of Ammophila. A comprehensive monitoring program was developed with PCA and undertaken by the author and PCA collaborators from start of the project in Summer 2008 to Fall of 2012. This dissertation is the product of independent research by the author carried out under the supervision of the advisory committee and does not reproduce written materials prepared for, or by, PCA. The dissertation consists of three separate journal manuscripts (the first two published by completion of the dissertation) that stand alone as independent investigations but are structured here to provide a natural progression of research findings and allow for an overall synthesis of ideas and broader contributions of the research. The dune restoration program afforded an opportunity to review restoration trends and methods and implement a strategy and monitoring protocols based on leading edge science. Accordingly, the first manuscript, Chapter 2, summarises recent trends in coastal dune restoration, discusses relevant research surrounding beach-dune morphodynamics and coastal dune activity, and reviews preliminary data from the project. The study identifies usable control data for the project and builds the criteria for assessing the project as a whole. The second manuscript, Chapter 3, presents and analyses the core data obtained for the dissertation - 5 years of geomorphic monitoring from detailed land surveys with 3 years of analysis of beach-foredune-transgressive dune sediment budget responses derived from aerial LiDAR surveys. This chapter identifies several trends in the dune systems’ response to restoration that, with reference to the indicators developed in Chapter 2, suggest improved levels of dynamism in the landscape. Finally, Chapter 4 (manuscript 3), extends the findings of the restoration study and utilises the rich data set obtained from the restoration program to develop a dynamic mapping technique that better conveys the spatial-temporal morphodynamic behaviour of dune ecosystems. The study comments broadly on the potential to apply these data and techniques to the study of disturbance events in beach-dune systems. The dissertation is concluded (Chapter 5) with an overall summary of key research objectives and contributions, and presents recommendations for future research. / Graduate

Buffel grass (Cenchrus ciliaris ) has been established in Uluṟu‐Kata Tjuta National Park since 1968. To date, the influence of buffel grass on the Park's flora and fauna has been largely unassessed. The objectives of this study were to determine if buffel grass dominates vegetation communities at the base of Uluṟu and if buffel grass habitats are associated with lower reptile and amphibian species richness than endemic vegetation communities. We used vegetation transects to measure the amount of buffel grass and genera of endemic vegetation at 26 sampling locations around the base of Uluṟu. The vegetation survey data were paired with pitfall trap data from reptile and amphibian captures at the same sampling locations. Indicator species analysis and non‐metric multidimensional scaling were used to analyse the vegetation and herpetofaunal community data. Our analyses determined five distinct vegetation communities around Uluṟu. At the base of Uluṟu, buffel grass dominated half of sampled areas and the rest of the inselberg's base was dominated by Themeda grasses. Buffel grass habitats had significantly higher herpetofaunal species richness than the Themeda habitats that dominated other areas at Uluṟu's base. Herpetofauna species richness in buffel grass‐dominated habitats was also significantly higher than all vegetation communities except for Triodia‐dominated habitats. These observations do not directly indicate that buffel grass presence promotes higher species richness of reptiles and amphibians since the observed patterns may be driven by factors such as proximity to breeding sites and abiotic variables not directly related to the grass itself.

The braided rivers and floodplains in the Upper Waitaki Basin (UWB) of the South Island of New Zealand are critical habitats for endangered and threatened fauna such as the black stilt. However, this habitat has degraded due to introduced predators, hydropower operations, and invasive weeds including Russell lupins. While conservation efforts have been made to restore these habitats, flood events may provide a natural mechanism for removal of invasive vegetation and re-creation of natural floodplain habitats. However, little is understood about the hydraulic effects of floods on vegetation and potential mortality in these dynamic systems. Therefore, this thesis analyzed the flood-induced processes that cause lupin mortality in a reach of the Ahuriri River in the UWB, and simulated various sized flood events to assess how and where these processes occurred. To determine the processes that cause lupin mortality, post-flood observations were utilized to develop the hypothesis that flood-induced drag, erosion, sediment deposition, inundation, and trauma were responsible. Field and laboratory experiments were conducted to evaluate and quantify these individual processes, and results showed that drag, erosion, sediment deposition and inundation could cause lupin mortality. Utilizing these mortality processes, mortality thresholds of velocity, water depth, inundation duration, and morphologic changes were estimated through data analysis and evaluation of various empirical relationships. Delft3D was the numerical model used to simulate 2-dimensional flood hydraulics in the study-reach and was calibrated in three stages for hydraulics, vegetation, and morphology. Hydraulic calibration was achieved using the study-reach topography captured by Structure-from-Motion (SfM) and various hydraulic data (depth, velocity, and water extent from aerial photographs). Vegetation inclusion in Delft3D was possible utilizing a function called ‘trachytopes’, which represented vegetation roughness and flow resistance and was calibrated utilizing data from a lupin-altered flow conveyance experiment. Morphologic calibration was achieved by simulating an observed near-mean annual flood event (209 m3 s-1) and adjusting the model parameters until the simulated morphologic changes best represented the observed morphologic changes captured by pre- and post-flood SfM digital elevation models. Calibration results showed that hydraulics were well represented, vegetation inclusion often improved the simulated water inundation extent accuracy at high flows, but that local erosion and sediment deposition were difficult to replicate. Simulation of morphological change was expected to be limited due to simplistic bank erosion prediction methods. Nevertheless, the model was considered adequate since simulated total bank erosion was comparable to that observed and realistic river characteristics (riffles, pools, and channel width) were produced. Flood events ranging from the 2- to 500-year flood were simulated with the calibrated model, and lupin mortality was estimated using simulation results with the lupin mortality thresholds. Results showed that various degrees of lupin mortality occurred for the different flood events, but that the dominant mortality processes fluctuated between erosion, drag, and inundation. Sediment deposition-induced mortality was minimal, but was likely under-represented in the modeling due to poor model sediment deposition replication and possibly over-restrictive deposition mortality thresholds. The research presented in this thesis provided greater understanding of how natural flood events restore and preserve the floodplain habitats of the UWB and can be used to aid current and future braided river conservation and restoration efforts.

Thesis (MScConEcol (Conservation Ecology and Entomology))--University of Stellenbosch, 2011. / ENGLISH ABSTRACT: Invasive alien plants (IAPs) reduce streamflow and threaten the biodiversity of South Africa’s Cape Floristic Region. Up-to-date information on invasive vegetation is required for land management agencies to formulate policies and make appropriate resource management decisions. Invasion maps are typically not updated often enough because of the time and expenses required to do so. As a result, invasion maps for South Africa are limited to coarse resolution data or isolated small scale studies. Invasive alien plants change the landscape by destabilizing catchments and thereby increasing soil erosion, altering fire regimes and hydrology, as well as changing the physical and chemical composition of the soil. Information on IAPs is needed at a landscape scale. Remote sensing is a powerful tool that can be used to characterise landscapes in a biologically meaningful manner. The Normalised Difference Vegetation Index (NDVI) was used to create an up-to-date invasion map of the Agulhas Plain, lying at the heart of the species rich Cape Floristic Region. This information was combined with actual evapotranspiration data from the Surface Energy Balance Algorithm for Land (SEBAL) study done by Water Watch and the Council for Scientific and Industrial Research. The results showed that invasive vegetation uses more water than natural fynbos vegetation and that the greatest amount of water would be made available by clearing the invaded deep sands on the Agulhas Plain. These deep sand areas conflict with the priority areas of the Working for Water programme. This IAP eradication programme targets sparsely invaded upland areas for long-term sustainability. The recommendation of this study is to clear invaded wetland and riparian areas as these zones yield the greatest hydrological benefit per hectare and meet the priorities of Working for Water. Overall, 36 million cubic meters of water would be made available by clearing the Agulhas Plain. It can be concluded that there is a significant hydrological benefit to clearing invasive alien vegetation on the Agulhas Plain. / AFRIKAANSE OPSOMMING: Indringerplante (IP) verminder stroomvloei en bedreig die biodiversiteit van Suid-Afrika se Kaapse Floristiese Streek. Die nuutste inligting oor uitheemse plantegroei is nodig vir grondbestuuragentskappe om beleide te formuleer vir die neem van toepaslike hulpbronbestuur besluite. As gevolg van die tyd en uitgawes wat nodig is om indringingskaarte op te dateer, word dit gewoonlik nie dikwels genoeg gedoen nie. Dus is indringingskaarte vir Suid-Afrika beperk tot growwe resolusie data of geïsoleerde kleinskaal studies. Indringerplante verander die landskap deur opvangsgebiede te destabiliseer en sodoende te lei tot gronderosie, verandering van vuurregimes en hidrologie, sowel as die verandering in die fisiese en chemiese samestelling van die grond. Inligting oor IP is nodig op 'n landskapskaal. Afstandswaarneming is 'n kragtige tegniek wat gebruik kan word om landskappe op 'n biologies betekenisvolle manier te karakteriseer. Die Normalised Difference plantegroei-indeks (NDVI) is gebruik om 'n opgedateerde indringingskaart van die Agulhas-vlakte, wat in die hart van die spesiesryke Kaapse Floristiese Streek lê, te skep. Hierdie inligting is gekombineer met die werklike evapotranspirasie data vanaf die Surface Energy Balance Algorithm for Land (SEBAL) studie gedoen deur Water Watch en die Raad vir Wetenskaplike en Nywerheidnavorsing. Die resultate het getoon dat uitheemse plantegroei meer water gebruik as natuurlike fynbosplantegroei en dat die grootste hoeveelheid van hierdie water beskikbaar gestel sal word deur IP op diepsand op die Agulhas-vlakte skoonte maak. Hierdie diepsand areas is in konflik met die prioriteitsgebiede van die Werk vir Water-program. Hierdie IP uitroeiingsprogram fokus op yl ingedringde berggebiede vir langtermyn volhoubaarheid. Die aanbeveling van hierdie studie is om duidelik ingedringde vleilande en oewergebiede skoon te maak, siende dat hierdie sones die hoogste opbrengs en die grootste hidrologiese voordeel per hektaar bied, en voldoen aan die prioriteite van Werk vir Water. In totaal sou 36 miljoen kubieke meter water beskikbaar gestel word deur die skoonmaak van die Agulhas-vlakte. Dus kan dit afgelei word dat die verwydering van hidrologiese indringerplante op die Agulhas-vlakte 'n beduidende voordeel sal inhou.

Les images hyperspectrales suscitent un intérêt croissant depuis une quinzaine d'années. Elles fournissent une information plus détaillée d'une scène et permettent une discrimination plus précise des objets que les images couleur RVB ou multi-spectrales. Bien que les potentialités de la technologie hyperspectrale apparaissent relativement grandes, l'analyse et l'exploitation de ces données restent une tâche difficile et présentent aujourd'hui un défi. Les travaux de cette thèse s'inscrivent dans le cadre de la réduction et de partitionnement des images hyperspectrales de grande dimension spatiale. L'approche proposée se compose de deux étapes : calcul d'attributs et classification des pixels. Une nouvelle approche d'extraction d'attributs à partir des matrices de tri-occurrences définies sur des voisinages cubiques est proposée en tenant compte de l'information spatiale et spectrale. Une étude comparative a été menée afin de tester le pouvoir discriminant de ces nouveaux attributs par rapport aux attributs classiques. Les attributs proposés montrent un large écart discriminant par rapport à ces derniers et par rapport aux signatures spectrales. Concernant la classification, nous nous intéressons ici au partitionnement des images par une approche de classification non supervisée et non paramétrique car elle présente plusieurs avantages: aucune connaissance a priori, partitionnement des images quel que soit le domaine applicatif, adaptabilité au contenu informationnel des images. Une étude comparative des principaux classifieurs semi-supervisés (connaissance du nombre de classes) et non supervisés (C-moyennes, FCM, ISODATA, AP) a montré la supériorité de la méthode de propagation d'affinité (AP). Mais malgré un meilleur taux de classification, cette méthode présente deux inconvénients majeurs: une surestimation du nombre de classes dans sa version non supervisée, et l'impossibilité de l'appliquer sur des images de grande taille (complexité de calcul quadratique). Nous avons proposé une approche qui apporte des solutions à ces deux problèmes. Elle consiste tout d'abord à réduire le nombre d'individus à classer avant l'application de l'AP en agrégeant les pixels à très forte similarité. Pour estimer le nombre de classes, la méthode AP utilise de manière implicite un paramètre de préférence p dont la valeur initiale correspond à la médiane des valeurs de la matrice de similarité. Cette valeur conduisant souvent à une sur-segmentation des images, nous avons introduit une étape permettant d'optimiser ce paramètre en maximisant un critère lié à la variance interclasse. L'approche proposée a été testée avec succès sur des images synthétiques, mono et multi-composantes. Elle a été également appliquée et comparée sur des images hyperspectrales de grande taille spatiale (1000 × 1000 pixels × 62 bandes) avec succès dans le cadre d'une application réelle pour la détection des plantes invasives. / The interest in hyperspectral image data has been constantly increasing during the last years. Indeed, hyperspectral images provide more detailed information about the spectral properties of a scene and allow a more precise discrimination of objects than traditional color images or even multispectral images. High spatial and spectral resolutions of hyperspectral images enable to precisely characterize the information pixel content. Though the potentialities of hyperspectral technology appear to be relatively wide, the analysis and the treatment of these data remain complex. In fact, exploiting such large data sets presents a great challenge. In this thesis, we are mainly interested in the reduction and partitioning of hyperspectral images of high spatial dimension. The proposed approach consists essentially of two steps: features extraction and classification of pixels of an image. A new approach for features extraction based on spatial and spectral tri-occurrences matrices defined on cubic neighborhoods is proposed. A comparative study shows the discrimination power of these new features over conventional ones as well as spectral signatures. Concerning the classification step, we are mainly interested in this thesis to the unsupervised and non-parametric classification approach because it has several advantages: no a priori knowledge, image partitioning for any application domain, and adaptability to the image information content. A comparative study of the most well-known semi-supervised (knowledge of number of classes) and unsupervised non-parametric methods (K-means, FCM, ISODATA, AP) showed the superiority of affinity propagation (AP). Despite its high correct classification rate, affinity propagation has two major drawbacks. Firstly, the number of classes is over-estimated when the preference parameter p value is initialized as the median value of the similarity matrix. Secondly, the partitioning of large size hyperspectral images is hampered by its quadratic computational complexity. Therefore, its application to this data type remains impossible. To overcome these two drawbacks, we propose an approach which consists of reducing the number of pixels to be classified before the application of AP by automatically grouping data points with high similarity. We also introduce a step to optimize the preference parameter value by maximizing a criterion related to the interclass variance, in order to correctly estimate the number of classes. The proposed approach was successfully applied on synthetic images, mono-component and multi-component and showed a consistent discrimination of obtained classes. It was also successfully applied and compared on hyperspectral images of high spatial dimension (1000 × 1000 pixels × 62 bands) in the context of a real application for the detection of invasive and non-invasive vegetation species.

Les images hyperspectrales sont des images complexes qui ne peuvent être partitionnées avec succès en utilisant une seule méthode de classification. Les méthodes de classification non coopératives, paramétriques ou non paramétriques peuvent être classées en trois catégories : supervisée, semi-supervisée et non supervisée. Les méthodes paramétriques supervisées nécessitent des connaissances a priori et des hypothèses sur les distributions des données à partitionner. Les méthodes semi-supervisées nécessitent des connaissances a priori limitées (nombre de classes, nombre d'itérations), alors que les méthodes de la dernière catégorie ne nécessitent aucune connaissance. Dans le cadre de cette thèse un nouveau système coopératif et non supervisé est développé pour le partitionnement des images hyperspectrales. Son originalité repose sur i) la caractérisation des pixels en fonction de la nature des régions texturées et non-texturées, ii) l'introduction de plusieurs niveaux d'évaluation et de validation des résultats intermédiaires, iii) la non nécessité d'information a priori. Le système mis en ouvre est composé de quatre modules: Le premier module, partitionne l'image en deux types de régions texturées et non texturées. Puis, les pixels sont caractérisés en fonction de leur appartenance à ces régions. Les attributs de texture pour les pixels appartenant aux régions texturées, et la moyenne locale pour les pixels appartenant aux régions non texturées. Le deuxième module fait coopérer parallèlement deux classifieurs (C-Moyen floue : FCM et l'algorithme Adaptatif Incrémental Linde-Buzo-Gray : AILBG) pour partitionner chaque composante. Pour rendre ces algorithmes non supervisés, le nombre de classes est estimé suivant un critère basé sur la dispersion moyenne pondérée des classes. Le troisième module évalue et gère suivant deux niveaux les conflits des résultats de classification obtenus par les algorithmes FCM et AILBG optimisés. Le premier identifie les pixels classés dans la même classe par les deux algorithmes et les reportent directement dans le résultat final d'une composante. Le second niveau utilise un algorithme génétique (GA), pour gérer les conflits entre les pixels restant. Le quatrième module est dédié aux cas des images multi-composantes. Les trois premiers modules sont appliqués tout d'abord sur chaque composante indépendamment. Les composantes adjacentes ayant des résultats de classification fortement similaires sont regroupées dans un même sous-ensemble et les résultats des composantes de chaque sous-ensemble sont fusionnés en utilisant le même GA. Le résultat de partitionnement final est obtenu après évaluation et fusion par le même GA des différents résultats de chaque sous-ensemble. Le système développé est testé avec succès sur une grande base de données d'images synthétiques (mono et multi-composantes) et également sur deux applications réelles: la classification des plantes invasives et la détection des pins. / Hyperspectral and more generally multi-component images are complex images which cannot be successfully partitioned using a single classification method. The existing non-cooperative classification methods, parametric or nonparametric can be categorized into three types: supervised, semi-supervised and unsupervised. Supervised parametric methods require a priori information and also require making hypothesis on the data distribution model. Semi-supervised methods require some a priori knowledge (e.g. number of classes and/or iterations), while unsupervised nonparametric methods do not require any a priori knowledge. In this thesis an unsupervised cooperative and adaptive partitioning system for hyperspectral images is developed, where its originality relies i) on the adaptive nature of the feature extraction ii) on the two-level evaluation and validation process to fuse the results, iii) on not requiring neither training samples nor the number of classes. This system is composed of four modules: The first module, classifies automatically the image pixels into textured and non-textured regions, and then different features of pixels are extracted according to the region types. Texture features are extracted for the pixels belonging to textured regions, and the local mean feature for pixels of non-textured regions. The second module consists of an unsupervised cooperative partitioning of each component, in which pixels of the different region types are classified in parallel via the features extracted previously using optimized versions of Fuzzy C-Means (FCM) and Adaptive Incremental Linde-Buzo-Gray algorithm (AILBG). For each algorithm the number of classes is estimated according to the weighted average dispersion of classes. The third module is the evaluation and conflict management of the intermediate classification results for the same component obtained by the two classifiers. To obtain a final reliable result, a two-level evaluation is used, the first one identifies the pixels classified into the same class by both classifiers and report them directly to the final classification result of one component. In the second level, a genetic algorithm (GA) is used to remove the conflicts between the invalidated remaining pixels. The fourth module is the evaluation and conflict management in the case of a multi-component image. The system handles all the components in parallel; where the above modules are applied on each component independently. The results of the different components are compared, and the adjacent components with highly similar results are grouped within a subset and fused using a GA also. To get the final partitioning result of the multi-component image, the intermediate results of the subsets are evaluated and fused by GA. The system is successfully tested on a large database of synthetic images (mono and multi-component) and also tested on two real applications: classification of invasive plants and pine trees detection.