Ecological land classification refers to an integrated survey in which areas of land, as ecosystems, are classified according to their ecological unity. In Canada, the approach was first advanced, nationally, in 1969 and was termed 'Bio-physical Land Classification'. This approach, which was derived from several foreign and domestic precedents, has been employed by various independent survey organizations throughout Canada to secure an ecological data bases for resource planning and management consideration. Because coordination was lacking between these organizations, modifications of this approach have taken place independently and often have been weighted according to the investigator's personal interests or capabilities. As such, the approach currently possesses a disparate character which is difficult to define singularly. To identify the current status in methodology, Canadian works in this field were comparatively analyzed. One result which stands out prominently from the analysis is that there are multifarious forms of ecological land classification. While they tend to achieve the same results, and demonstrate numerous commonalities land ecosystems have been manifested by combinations of criteria which are not always the same. Considerable confusion surrounds the nomenclature, the criteria for definitions and the criteria for recognition. Based on the analysis, hierarchical categories eco- province, ecoregion, ecodistrict, ecosection and ecotype are proposed. These are land ecosystems which possess a common recognized identity based on a unified pattern of biological and physical land characteristics. Each category coincides with a different order of generalization. Based largely on material extracted from past studies, criteria for recognition are stated. / Land and Food Systems, Faculty of / Graduate

Cette thèse traite deux problèmes importants dans les domaine de l'apprentissage automatique et du raisonnement dans l'incertain : comment évaluer une relation de similarité entre deux informations incertaines et comment assurer la classification \`a partir de données incertaines. Notre première principale contribution est de proposer une approche, appelée arbre de décision possibiliste, qui permet de construire des arbres de décision à partir de données d'apprentissage imparfaites. Plus précisément, elle traite des données caractérisées par des classes incertaines o\`u l'incertitude est modélisée avec la théorie des possibilités quantitative. Nous avons développé trois approches d'arbres de décision possibilistes. Pour chacune des approches, nous avons été confrontés à résoudre plusieurs problèmes pour pouvoir construire des arbres de décision possibilistes, tels que, comment définir une mesure de sélection d'attributs quand les classes sont représentes par des distributions de possibilité, comment trouver les critères d'arrêt et comment les feuilles vont être étiquetées dans ce contexte incertain. La première approche, appelée arbre de décision possibiliste basée sur la non- spécificité, utilise le concept de non-spécificité relatif à la théorie des possibilités dans la définition de sa mesure de sélection d'attributs. Cette approche maintient les distributions de possibilité durant toutes les étapes de la procédure de construction et ce particulièrement, au moment de l'évaluation de la quantité d'information apportée par chaque attribut. En revanche, la deuxième et la troisième approches, appelées arbre de décision possibiliste basé sur la similarité et arbre de décision possibiliste basé sur le clustering, éliminent automatiquement les distributions de possibilité dans leurs mesures de sélection d'attributs. Cette stratégie a permis d'étendre le ratio de gain et, par conséquent, d'étendre l'algorithme C4.5 pour qu'il puisse traiter des données libellées par des classes possibilistes. Ces deux dernières approches sont principalement basées sur le concept de similarité entre les distributions de possibilité étudié dans la thèse.La deuxième principale contribution de cette thèse concerne l'analyse des mesures de similarité en théorie des possibilités. En effet, un challenge important était de fournir une analyse des mesures de similarité possibiliste conduite par les propriétés qu'elles doivent satisfaire. Après avoir montré le rôle important de la notion d'incohérence dans l'évaluation de la similarité en théorie des possibilités, une nouvelle mesure, appelée affinité de l'information a été proposée. Cette mesure satisfait plusieurs propriétés que nous avons établies. A la fin de cette thèse, nous avons proposé des expérimentations pour comparer et montrer la faisabilité des approches d'arbres de décision possibilistes que nous avons développées. / This thesis concerns two important issues in machine learning and reasoning under uncertainty: how to evaluate a similarity relation between two uncertain pieces of information, and how to perform classification from uncertain data. Our first main contribution is to propose a so-called possibilistic decision tree which allows to induce decision trees from training data afflicted with imperfection. More precisely, it handles training data characterized by uncertain class labels where uncertainty is modeled within the quantitative possibility theory framework. We have developed three possibilistic decision tree approaches. For each approach, we were faced and solved typical questions for inducing possibilistic decision trees such as how to define an attribute selection measure when classes are represented by possibility distributions, how to find the stopping criteria and how leaves should be labeled in such uncertain context. The first approach, so-called, non-specificity-based possibilistic decision tree uses the concept of non-specificity relative to possibility theory in its attribute selection measure component. This approach keeps up the possibility distributions within all the stages of the building procedure and especially when evaluating the informativeness of the attributes in the attribute selection step. Conversely, the second and the third approaches, so-called similarity-based possibilistic decision tree and clustering-based possibilistic decision tree, automatically, get rid of possibility distributions in their attribute selection measure. This strategy has allowed them to extend the gain ratio criterion and hence to extend the C4.5 algorithm to handle possibilistic labeled data. These two possibilistic decision tree approaches are mainly based on the concept of similarity between possibility distributions.This latter issue constitutes our second main contribution in this thesis. In fact, an important challenge was to provide a property-based analysis of possibilistic similarity measures. After showing the important role that inconsistency could play in assessing possibilistic similarity, a new inconsistency-based possibilistic similarity measure, so-called information affinity has been proposed. This measure satisfies a set of natural properties that we have established. Finally, we have conducted experiments to show the feasibility and to compare the different possibilistic decision tree approaches developed in this thesis.

Notre étude a commencé par une brève présentation de la langue persane et des parties du discours suivi par la démonstration de l'adjectif persan dans la grammaire contemporaine, notion floue dans les études antérieures. Étant donné que l'adjectif persan n'a pas de marquage flexionnel, cette étude exploratoire a pour but d’identifier des critères pertinents pour distinguer la catégorie adjectivale des autres catégories existantes (notamment nominale et verbale). Le résultat de ce travail nous a conduits à une étude morphosyntaxique et sémantique nous permettant de conclure au fait que l'adjectif persan est une sous-catégorie de la catégorie nominale. / Our study begins with a brief presentation of the Persian language and the parts of speech followed by the demonstration of the Persian adjective in the contemporary grammar and to look for this vague notion in the previous studies. Given that the Persian adjective has no inflected marking, this search aims to identifying distinctive criteria to distinguish the adjectival category from the other existing categories (in particular nominal and verbal). The results of this work conduct us towards a morphosyntactic and semantic study which allowing us to conclude that the Persian adjective is a subcategory of the nominal category.

Gem garnet chemistry is described using the end-members pyrope, almandine, spessartine, andradite, grossular, and uvarovite. The large variation in garnets makes classification difficult. Garnet gemstones are typically classified by means of their refractive indices, together with specific gravity. However, the range of refractive indices for gem garnets can be restrictive and unreliable in gemstone classification. Chemical classification is generally the most accurate means of classification and was used in this thesis by means of a portable XRF for the non-destructive chemical analyses of 1513 garnet gemstones. Colour, refractive index and magnetic susceptibility were also determined. The garnets were divided into two species: ugrandite (uvarovite, grossular and andradite) and pyralspite (pyrope, spessartine and almandine). The chemistry of the pyralspite species was very diverse with a large range in end-member proportions producing extensive solid solutions between end-members. A diverse range in colour, refractive index, and magnetic susceptibility was also observed in the pyralspite species. However, no distinction based on refractive index and magnetic susceptibility could be made in the pyralspite species. Distinctive chemistries were observed in the ugrandite species, which correlates with the magnetic susceptibility and refractive index. Colour was the exception as no relationship between colour and chemistry was observed. The samples with unusual compositions were reported such as pyrope-andradite, spessartine-grossular, and almandine-grossular. This study concluded that refractive index, magnetic susceptibility and colour should not be used in isolation because this can lead to misinterpretation. Rather, analytical techniques, if available, should be used. / Dissertation (MSc)--University of Pretoria, 2016. / Geology / Unrestricted

In the forestry industry tree trunks are currently classified manually. The object of this thesis is to answer whether it is possible to automate this using modern computer hardware and image-classification of tree-trunks using machine learning algorithms. The report concludes, based on results from controlled experiments that it is possible to achieve an accuracy above 90% across the genuses Birch, Pine and Spruce with a classification-time per tree shorter than 500 milli seconds. The report further compares these results against previous research and concludes that better results are probable.

This study examines the evolutionary history of the cosmopolitan beaksedge tribe Rhynchosporeae (ca. 386 spp.; Cyperaceae) using phylogenetics. Taxon sampling covers 25 of the 28 taxonomic sections proposed for the tribe. I compare a history inferred for Rhynchosporeae using a single plastid gene (Chapter 2) with one inferred using hundreds of loci (Chapter 4). The latter involves a sequencing methodology I develop with collaborators that can be applied across angiosperms (Chapter 3). Chapter 2 recognizes that Rhynchosporeae has high levels of endemicity (≥ 44%) in tropical and subtropical American savannas and can provide insights into the diversification of their biotas. Wind pollination, occupation of a savanna habitat, and a C3 photosynthetic pathway are common in the tribe, but showy (presumably insect-pollinated) inflorescences, occupation of forest habitat, and a C4 pathway also occur. I reconstructed a dated phylogenetic hypothesis for 79 taxa, using the trnL/F plastid region, inferring a mean crown-group age of 56 million years. Fitch parsimony infers the most recent common ancestor (MRCA) to have occupied a savanna habitat with eight or more shifts to forest. Features associated with insect pollination—white bracts and spikelets—were shown to evolve six or more times but were not correlated with the shifts to forest habitat where wind pollination is likely to be less effective. I found evolutionary correlations in the pairwise comparisons of bract color versus spikelet color and bract positioning versus bract color. Members with anatomies associated with C4, though anatomically variable, form a clade with a crown age of 19 million years. In Chapter 3, with collaborators I develop a robust probe design process to identify 499 low-copy nuclear regions and 18 high-copy functional genes for hybrid enrichment. We obtained >90% enrichment success for target regions. Between 159 and 488 orthologs were retained in alignments used for phylogenetic inference at deep and shallow levels across the angiosperms. A sampling strategy focusing on incremental removal of incongruent loci in combination with removal of sites with high rates of change produced 196 alignments for phylogenetic inference. The phylogenetic hypotheses at each sample level represent outcomes under different regions of parameter space. These outcomes were presented using heatmaps that depict bootstrap support at all nodes for those 196 levels of parameter space. This provided a new approach for sensitivity analyses and for testing the robustness of any hypotheses. A randomization methodology for hypotheses testing at specific nodes takes advantage of the heatmap approach. Focusing on the difficult-to-resolve eudicot, monocot, Magnoliid nodes the analysis revealed that the supermatrix approach produced a spuriously confident yet conflicting result in some regions of parameter space. With >97% of the data, supermatrix analyses supported eudicot and Magnoliids as sister. Support switched to strongly support the eudicot and monocot sister relationship at higher levels of data removal. In contrast the coalescent model consistently supported the latter relationship across most of the parameter space. Overall the eudicot and monocot sister relationship is robustly favored. In Chapter 4, I reexamine beaksedge (tribe Rhynchosporeae ) diversification but employ the anchored hybrid enrichment protocols developed in Chapter 3. A dated phylogenetic hypothesis for 115 taxa in the tribe and 11 outgroup taxa inferred a mean crown-group age for the tribe of 43.2 million years. Ancestral state reconstruction using stochastic mapping infers an open (savanna) habitat for the MRCA. This was the common state along 77% of the total branch lengths. However, there was an average of 22 independent shifts from open habitats into forest understory or edges in its descendants. The common state was the typical seasonally wet savanna soils. The state associated with their occurrence in dry, well drained soils was reconstructed for 4% of the total branch lengths, but there was an average of 11.2 transitions to that state. There were 3.7 transitions to the state where plants typically occur in standing or flowing water. An average of 5.9 transitions from nondescript brown or green inflorescences associated with wind pollination to those associated with insect pollination (white spikelets and/or bracts) were inferred but these were not correlated with the shifts to forest habitat. Members with C4 anatomy formed a clade that diverged from a sister clade containing taxa with C3 photosynthetic anatomies 26 MYA; this is earlier than previously thought. Most of the taxonomic sections described by Shirley Gale and Georg Kükenthal for Rhynchospora and Pleurostachys were not monophyletic. I also briefly discuss the possible significance of detecting a recently described repetitive satellite DNA element known to be associated with the centromeric protein CENH3. / A Dissertation submitted to the Department of Biological Science in partial fulfillment of the Doctor of Philosophy. / Fall Semester 2016. / November 14, 2016. / Anchored Phylogenomics, Angiosperms, evolution, photosynthesis, pollination, Rhynchospora / Includes bibliographical references. / Austin Mast, Professor Directing Dissertation; William Parker, University Representative; Alice Winn, Committee Member; Scott Steppan, Committee Member; Brian Inouye, Committee Member.

The original intent of this study was to reinvestigate and confirm the validity of thirteen species of Podocotyle reported from Dillon Beach, California. Eight species were described by Park (1937) and five described by Johnson (1949-unpublished). Among these species basic similarities were evident in such features as: cirrus sac, seminal vesicle, body shape, extent and distribution of vitellaria, which indicated the possibility of conspecificity. Because four of these species (Podocotly blennicottusi, P. california, P. pedunculata, and P. pacifica) were also reported from fishes of the Pacific Northwest coast by Ching (1960), Pratt and McCauley (1961), Arai (1969), and Hall and Pratt (1969), it was decided to expand the investigation to include these forms and other Podocotyle spp, from coastal areas of the world in an attempt to determine the number of valid species and introduce and improved method of species determination within this group. Some 489 tide pool fish, representing 23 species, were collected and examined at the Pacific Marine Station, Dillon Beach, California, between July 1969 and August 1970. Specimens of Pococotyle, obtained by the investigator, were augmented by material from previous collections in that area.

The field of customs is commonly referred to as that of imports and exports. It is perceived as a maze of processes, procedures, and forms required to enable a customs administration to perform their wide range of responsibilities. One of the responsibilities of a customs administration is the collection of duties, which necessitates classification of the goods in question. This study sets out to determine the extent of customs control in relation to tariff classification in South Africa. The starting point is the establishment of the foundations of customs, both internationally and in South Africa. After origin and valuation, tariff classification is the third technical customs-related focus area. An analysis of the responsibilities of the customs administration in South Africa confirms the importance of revenue collection and, subsequently, tariff classification. As a result of South Africa’s membership of the World Customs Organization, specific obligations in relation to tariff classification are incurred. The implementation and application of the international provisions are considered and compared in South Africa, Australia, and Canada. Not only is South Africa’s existing legislation considered, but also two new Acts. It is found that despite similarities in the implementation of the Harmonized System Convention into the legislation of the three countries, South Africa’s existing legislation makes the most detailed provision for the Harmonized System and its aids. This is based on the finding that the legislation in Australia and Canada, as well as the two new Acts in South Africa, do not have the same comprehensive provisions. A critical review of the varying processes of classification in the three countries suggests that more suitable and effective processes could be implemented in South Africa. In addition, a synopsis of some of the principles developed in case law is provided and compared. In relation to facilitation, the access to relevant information and the adequacy thereof, as well as the availability of rulings, are considered. Differences in the approach to dispute resolution in the three countries are furthermore provided. Proposals are made to address the discrepancies in the implementation and application of the legislation, the process of classification, the principles developed in case law, the enhancement of related guides, the publication of tariff classification rulings, and the extent of facilitation and dispute resolution. Finally it is recommended that an independent and expert tribunal is established to adjudicate technical customs matters. / Thesis (LLD)--University of Pretoria, 2015. / Mercantile Law / Unrestricted

No description available.

No description available.