Engineering: Theses and Dissertations

Permanent URI for this collection

https://ir.canterbury.ac.nz/handle/10092/840

Browse

Now showing 1 - 20 of 3085

Open Access
Exploring radiata pine breeding strategies using genomic tools.
(2024) McLean, Duncan
Radiata pine (Pinus radiata D. Don) is economically the most important forestry species in New Zealand. The radiata pine breeding program has undergone numerous strategic changes over the past 70 years. Presently, the Radiata Pine Breeding Company (RPBC) is in the process of implementing genomic selection to shorten generation cycles and accelerate genetic progress. We used a stochastic simulation approach to confirm that genomic selection can double the annual genetic gain compared to traditional phenotypic selection cycles. This significant increase is primarily due to the shorter breeding cycle facilitated by GS. The study also found that prediction accuracy and genetic gain increase with larger training populations and clonal testing, but plateau after approximately 2,500 genotypes. In terms of genotype-by-environment interaction (GxE), while it exists, its magnitude does not justify regionalizing breeding zones within New Zealand’s North Island. The thesis further explores the impact of GS on the deployment of genetic gain to production forests, showing that GS significantly increases genetic gain across all scenarios, with clonal deployment via somatic embryogenesis yielding the highest increase. An optimal orchard scenario for control-pollinated (CP) deployment was identified as having low female turnover with high male turnover, balancing genetic gain, seed yield, and inbreeding rates. Lastly, a cost-benefit analysis revealed that open-pollinated (OP) orchards offer a higher return on investment over 20 deployment cycles, primarily due to lower production costs and higher seed yields. This research demonstrates the advantages of genomic selection in accelerating genetic gain and optimising breeding and deployment strategies for Radiata pine in New Zealand.
Open Access
Construction and optimisation of an on-site nitrogen pressure swing adsorption process for Nuenz Ltd.
(2024) Peat, Kendrick
Nitrogen is widely used as an inert gas in the chemical process industry to prevent oxidation. The abundant availability of atmospheric nitrogen has made air the primary feedstock for nearly all nitrogen production processes. The primary objective of this thesis was to construct and optimise an on-site nitrogen generation process for the company Nuenz. A literature review established that stripping reflux pressure swing adsorption using a carbon molecular sieve (CMS) adsorbent and a modified Skarstrom process cycle was the most suitable method for economic nitrogen production at the purity and volume relevant to the company. A pilot-scale stripping reflux pressure swing adsorption rig (pilot PSA rig) that produced high-purity nitrogen (500-5 ppm O2) was successfully constructed using off-the-shelf and in-house fabricated components. The effect of operating parameters and cycle configuration on process performance was experimentally investigated, especially the effect of cycle time, column aspect ratio, equalisation time, cutting time, temperature, productivity and pressure. The performance of the pilot PSA rig was assessed using the performance indicators (PIs): purity, mass efficiency, and productivity. Optimal operating parameters for the pilot PSA rig were established from experimental results. A trade-off between purity and mass efficiency was linked to half-cycle time, with shorter half-cycles enhancing purity, whereas longer half-cycles improved mass efficiency. An increased column aspect ratio enhanced purity without compromising mass efficiency or productivity. A short equalisation step enhanced all PIs of the pilot PSA rig with a 0.5 s equalisation yielding maximum purity improvements. A cut step enhanced mass efficiency and productivity, while a cut time that was 30% of the half-cycle time maximised purity. Low operating temperatures enhanced all PIs at shorter cut step times but caused a purity decline at longer cut step times. Mass efficiency was enhanced at the expense of purity when productivity was increased. However, operating at a short half-cycle time and higher productivity achieved the same purity and efficiency as operating with a longer half-cycle time and lower productivity. Higher operating pressures improved product purity at the expense of mass efficiency. However, a higher operating pressure increased productivity for a given product purity without compromising mass efficiency. An empirical model was established from experimental results, which demonstrated good agreement with purity readings at different cut times, half-cycle times, and productivities and showed good predictive power for predicting purity when a productivity 15% higher than the experimental range was used.
Open Access
MoanaNet : a deep learning framework for global short-to-long-term sea surface temperature anomaly and marine heatwave forecasts.
(2024) Ning, Ding
In the context of escalating climate change, accurate forecasting of sea surface temperatures (SSTs), SST anomalies (SSTAs), and marine heatwaves (MHWs) is crucial due to their profound impacts on marine ecosystems, biodiversity, and human societies. The increasing frequency and intensity of these oceanic temperature oscillations and extreme events necessitate advanced methodologies for reliable predictions to help mitigate their adverse effects. This thesis introduces MoanaNet, an integrated deep learning (DL) framework that enhances forecasting capabilities for SSTs, SSTAs, and MHWs by leveraging spatiotemporal anomaly forecasting methodologies, data science techniques, and domain knowledge. The core innovation of this research lies in its comprehensive exploration of machine learning methods from anomaly, spatial, and temporal perspectives to advance data-driven forecasts of SSTs, SSTAs, and MHWs. By integrating these approaches into a cohesive framework, MoanaNet addresses the complexities inherent in predicting marine climate phenomena. This thesis evaluates the predictive performance of MoanaNet through spatial and temporal analyses, utilizing case studies to demonstrate areas where predictions are more or less successful. Comparisons with existing models and benchmarks further highlight the potential of MoanaNet to significantly improve forecasting of not only ocean temperature oscillations and extreme events but also broader climate phenomena. The findings of this research contribute to ongoing efforts to develop robust data-driven predictive tools for environmental applications, which are substantially different from conventional physics-based numerical approaches. The proposed tools can aid in planning and executing marine conservation strategies and managing the effects of climate anomalies on human activities. The specifi c contributions of the thesis can be summarized as follows: A DL framework that combines three methodologies: imbalanced regression (Chapter 2), graph representation (Chapter 3), and temporal diffusion (Chapter 4), to enhance the forecasting of SSTAs and MHWs from short to long lead times on a global scale is proposed. To the best of my knowledge, this is the rst research to synergistically combine these methodologies to predict SSTAs and MHWs. Furthermore, the framework allows for potential application to other climate extreme forecasts. Each component's primary contribution is detailed below: In Chapters 2 and 4, imbalanced regression and optimization of loss functions tailored for MHW prediction are investigated. Imbalanced regression loss functions, whether chosen from the existing or customized, are found to increase the detection of MHW occurrences, while the standard mean squared error (MSE) loss tends to lower the false alarm rate. Compared to the standard regression loss functions, this approach achieves higher average critical success index (CSI) or symmetric extremal dependence index (SEDI) scores. At 12 selected MHW hotspots, the imbalanced regression loss functions generally outperformed the standard MSE. In Chapters 3 and 4, two new graph construction approaches designed to convert SST grids into SSTA graphs and generate new publicly available SSTA graph datasets are presented. Several classes of graph neural networks, focusing on the use of the GraphSAGE architecture, are investigated. Generally, the graph approach outperformed the persistence model in short-term global SST and SSTA forecasts, provided up-to-two-year-ahead SST forecasts with a recursive model, and improved MHW predictions both on average and at 12 selected MHW hotspot locations. In Chapters 3 and 4, time series forecasting approaches for long-term prediction are integrated. Besides further examining the conventional sliding window approach and recursive method for climate forecasts, and the long short-term memory mechanism, an adaptation of the temporal diffusion process that eliminates the need for sliding windows in climate forecasting is introduced. The diffusion method is found to reduce input requirements and enhance long-term MHW predictive skills. To the best of my knowledge, this is the first attempt to combine temporal diffusion with graph neural networks for climate forecasts. The outcomes of MoanaNet can be split in two categories: spatial and temporal. Spatially, in terms of one-month-ahead forecasts, the SST and SSTA models, evaluated by the root mean squared error (RMSE), performed better than the persistence model across most locations around the world. The MHW models, evaluated by the SEDI, generated forecasts with distinct spatial patterns compared to numerical models on a global scale. Temporally, evaluated by the RMSE, the SST models produced forecasts up to two years ahead, with average RMSEs around 0.8 across all locations. Evaluated by combining the RMSE, precision, recall, CSI, and SEDI, SSTAs and MHWs could be predicted up to six months in advance. In summary, this thesis contributes to machine learning for climate forecasting by introducing MoanaNet, a deep learning framework that improves the prediction of SSTs, SSTAs, and MHWs. By integrating imbalanced regression, graph representation, and temporal diffusion, MoanaNet aims to enhance predictive performance across spatial and temporal scales. The framework not only introduces new tools for climate forecasting but also establishes a foundation for future research in data-driven Earth system modeling, supporting efforts to manage marine ecosystems and mitigate the impacts of climate extremes.
Open Access
Phase selection algorithm design and statistical analysis of Reconfigurable Intelligent Surface (RIS) Systems for 6G communications.
(2024) Inwood, Amy Siobhan
The demand for mobile data is growing exponentially. More devices than ever are connecting to mobile broadband, and new applications are continually being developed. It is expected that technologies such as holographic telepresence, digital twins and selfdriving cars will become available in the coming decades, requiring more higher data rates and lower latency than current mobile generations can provide. 6G is expected to utilise millimetre wave (mmWave) band (30 - 300 GHz) and terahertz band (300 GHz - 3 THz) carrier frequencies to provide these, with initial deployment planned for 2030. However, significant propagation challenges occur at such high frequencies, such as increased path loss and attenuation, and limited reflection, diffraction and random scattering, resulting in shorter possible transmission distances. Traditional solutions to extend coverage range, including the installation of additional base stations (BS) or relays, would be expensive to implement and consume a large amount of power. Reconfigurable intelligent surfaces (RIS) are proposed as a new technology to overcome some of these challenges. A RIS is a mostly passive array of metamaterial elements with the ability to change the phase of an incoming signal. Each RIS element can be tuned independently, allowing for controlled reflections directly influencing the signal path. Therefore, RIS offer a new way to increase channel capacity. A key challenge of RIS design is the selection of reflection coefficients for each element, due to the multidimensionality of the problem. While an optimal design is believed to be intractable in a multi-user (MU) scenario, an optimal design has been derived for the single-user (SU) case. A number of near-optimal methods have been designed for MU situations, however most are iterative, complex and custom, making implementation challenging. Therefore, the first part of this thesis focuses on designing a sub-optimal but very low complexity phase selection method for a MU RIS scenario by adapting the SU optimal design. An equal number of the N RIS elements are designed to serve each of the K users (referred to as a subsurface) on a separate frequency band. This means matched filtering can be used at the receiver. Elements not designed for a given user provide additional random scattering, something that is lacking in high-frequency environments. The analytical mean signal-to-noise ratio (SNR) of a system employing the subsurface design (SD) with a line-of-sight (LoS) RIS-BS channel, and Rayleigh fading user equipment (UE)-RIS and UE-BS channels is derived. An upper bound of the mean data rate is also found. This analysis is verified by simulation and is used to gain insights into system performance. When compared to the lowest complexity MU method known to us, a design that minimises the total mean-squared error (TMSE), the SD method leads to a rate loss due to each UE having only 1 K of the bandwidth of the MU case. However, this difference decreases as correlation increases. Similarly, it is found that spacing RIS elements closer together improves the mean SNR, as this increases correlation. Importantly, the SD leads to a K fold reduction in required channel state information and a N(K2 + K(N + 2)) computational complexity reduction when compared to the TMSE design. In the next section, the analytical mean SNR of a system employing the SD is significantly extended to cover Ricean fading channels, a more realistic and flexible scenario. The SD is also extended to a low-complexity iterative subsurface design (ISD). The ISD sets each subsurface sequentially, using previously set subsurfaces to improve the design. This process can be done once, or repeated until the total SNR increases by less than a specified tolerance (referred to as the converged ISD (CISD)). The analysis is verified and the performances of the ISD and CISD are investigated by simulation. The impact of Ricean K-factor is assessed, and despite the bandwidth restrictions, the SD outperforms the TMSE method at high UE-RIS and RIS-BS link K-factors. The mean rate of the TMSE method falls dramatically when the K-factor nears 1, as the channel cannot support the spatial multiplexing required to support multiple UEs in one frequency band. This is not an issue for the SD design. The impact of clustering UEs at the same location is also considered. Again, the mean rate of the MU method decreases, as channel diversity is reduced, but the SD is unaffected. These are realistic scenarios where the SD offers significant advantages over MU designs. Another key challenge facing RIS is channel state information acquisition. Passive RIS cannot generate pilot signals or process data. Therefore, rapid temporal changes in the channel would be problematic for a RIS system. The final section of this thesis uses second order statistics to investigate the temporal behaviour of a fundamental SU RIS system. Assuming a LoS RIS-BS link, we derive an exact expression for the level crossing rate (LCR) of the RIS link (UE-RIS-BS path) and propose a numerically stable approximation for the LCR of the UE-BS channel. Each LCR expression attained is then utilised to find the corresponding average fade duration (AFD). The mean SNR correlation is also derived for this scenario. Assuming a Ricean RIS-BS link, expressions for the system’s mean spatial correlation matrix and the mean SNR loss due to channel ageing are derived. All analysis is verified by simulation, and the impact of key system parameters is investigated. Crucially, it was found that RIS do not significantly amplify temporal changes in the channel, an important result.
Open Access
How do we balance user privacy and user experience in VR mediated experiences?
(2024) Kumarapeli, Dilshani
Virtual Reality(VR) applications utilise many sensor-tracked user movements to interact with the immersive environment and drive the avatar movements. These innocently seaming movement data are capable of revealing and predicting user behaviours when combined with machine learning algorithms. Even though these findings could be beneficial in areas like education or healthcare, unwanted and unauthorised processing and utilisation of these data could pose significant user privacy risks. To address this issue, this research focuses on introducing and evaluating methodologies to balance user privacy without affecting the user experience in VR-mediated applications. To evaluate the proposed methodologies, we use the most researched usage of these behaviour data, user identity detection, by employing machine learning-based classification algorithms. We begin this PhD research by exploring the nature of machine learning classification models used for user identity detection, analysing their strengths and weaknesses. Based on this analysis, we identify the overall best-performing model and use it to propose and evaluate two types of behaviour filters, assessing their effectiveness in concealing user identity. To validate the effectiveness of the proposed filters, we proposed a new multi-faceted validation model and explored the best ways to evaluate human-based identity recognition in avatars. We then examine the impact of these filters on the overall VR experience. Finally, to gain insights into how actual VR users perceive and prefer to share their behaviour data, we conduct and analyse user interviews. Through these stages, we critically discuss the research implications, suggest future research directions, and provide guidelines for developing and evaluating VR privacy solutions. Overall, this research addresses the question, "How do We Balance User Privacy and User Experience in VR Mediated Experiences?" through a comprehensive research approach that yields promising insights into novel strategies for enhancing VR user privacy. We also identify and highlight several major misconceptions and mistakes common to many existing privacy solutions, offering practical suggestions to overcome these challenges. We hope that the results, observations, and implications presented in this thesis will contribute to the development of more robust, multi-faceted privacy solutions to address the complex behaviour privacy challenges in VR environments.
Open Access
Home fitness equipment design for wheelchair users
(2024) Hui, Zixiong
This research's main aim is to investigate the feasibility of designing home fitness equipment tailored specifically for wheelchair users, with a particular emphasis on upper body training. Utilizing a comprehensive methodology that includes a literature review, user interviews, concept generation, and experimental testing, a prototype was developed and evaluated. The study identifies several key requirements and preferences of wheelchair users regarding home fitness equipment. These include the importance of affordability, space efficiency, flexibility, multifunctionality, and independence. The prototype that was developed redesigns and combines a pull-up bar with a fitness product called the Fitness Pump, effectively meets these needs. It supports a variety of upper body exercises, is easy to use, and can be conveniently mounted and removed from a door frame, making it suitable for users with limited space. Positive feedback from user testing indicates its high usability and effectiveness. However, the research also encountered certain limitations that impacted the design process. These include a limited number of participants, controlled testing environments that may not fully represent real-world conditions, a lack of feedback from the manufacturing process, and insufficient data on the durability of materials under varying environmental conditions. Future research should address these limitations by conducting large-scale testing with diverse samples, long-term usability studies, and exploring the integration of digital technologies such as virtual coaching and online courses to enhance the equipment's functionality and appeal.
Open Access
Phenology and natural enemies of Paropsisterna cloelia (Stål) and Paropsis charybdis Stål (Coleoptera: Chrysomelidae) in New Zealand
(2024) Weser, Carolin
Eucalypts have been grown in New Zealand since the 1860s for several end uses including pulp and paper production, and high-value hardwood. More recently plantations have aimed to provide ground-durable eucalypt timber as a sustainably grown alternative to Pinus radiata D.Don treated with chromated copper arsenate. Six Australian paropsine leaf beetles (Coleoptera: Chrysomelidae) have established in New Zealand so far. The Eucalyptus tortoise beetle, Paropsis charybdis Stål, was the first paropsine that established over 100 years ago, has spread throughout the country, and is currently the most destructive defoliator, causing severe damage to short-fibre plantations of Eucalyptus nitens (H.Deane & Maiden) Maiden and hardwood plantations of Eucalyptus quadrangulata H.Deane & Maiden. Several attempts of classical biological control with egg and larval parasitoids have been undertaken with some success. The Eucalyptus variegated beetle, Paropsisterna cloelia (Stål), was detected in 2016 and defoliation in hardwood plantations of E. bosistoana F.Muell. and E. tricarpa (L.A.S.Johnson) L.A.S.Johnson & K.D.Hill appears to be the most severe. This thesis aimed to quantify the phenology of Pst. cloelia and P. charybdis and the effect of general predators and parasitoids currently present in New Zealand on the two paropsine species. In a comprehensive literature review, chapter one compares life history traits, distribution, host preferences, phenology, natural enemies, and potential pest impacts of Pst. cloelia and P. charybdis in New Zealand. Australian distributions of both species largely overlap, indicating similar climate tolerances. Hence, it can be assumed that Pst. cloelia will also spread throughout New Zealand over time. In New Zealand, host preferences within the Eucalyptus subgenus Symphyomyrtus seem to vary between the two paropsines, with P. charybdis preferentially feeding on eucalypt species from the section Maidenaria, whereas Pst. cloelia seems to prefer the section Adnataria. Paropsisterna cloelia has a higher reproductive output than P. charybdis and advantageous life strategies that may lead to a higher survival rate of immature stages. This gives it the potential for more frequent population outbreaks and subsequently more severe defoliation on preferred hosts. Based on a literature review and personal communications, knowledge gaps were identified and, accordingly, thesis goals determined as follows. Goal 1: Quantify the phenology, meaning the timing of different life stages and the number of generations completed per year, of Pst. cloelia and P. charybdis in Marlborough. Goal 2: Develop a species-specific molecular assay that detects the DNA of Pst. cloelia and P. charybdis within field-collected predatory arthropods. Goal 3: Identify the most abundant predatory arthropod species in the field and quantify the proportion that feed on Pst. cloelia and P. charybdis life stages. Goal 4: Quantify the impact of the two established paropsine egg parasitoids on Pst. cloelia and P. charybdis populations in the field. In chapter two, the timing and abundance of different life stages (phenology) and the number of generations (voltinism) of Pst. cloelia and P. charybdis were quantified on E. bosistoana over two growing seasons at two field sites in Marlborough. Paropsisterna cloelia clearly dominated on the trees, whereas P. charybdis was practically absent. Results show that the availability of new foliage, facilitated primarily by the synchrony of warm temperatures and water availability, determined the number of generations, timing of life stages, and population size of both paropsine species. Timing of P. charybdis and Pst. cloelia life stages coincided at both field sites. As availability of new leaves in spring and mid-summer prompts oviposition, both species had two generations at the site where microclimate facilitated refoliation of trees after a first early summer defoliation event and one generation at the site where summer droughts prohibited refoliation of trees in mid-summer. Additionally, field observations indicate that both species have the potential for a third generation if the growth season is extended by warm spring and autumn temperatures and sufficient rainfall to facilitate production of new foliage earlier in the season as well as after the second summer defoliation event. These conditions may be met in the Northland and coastal Bay of Plenty regions. Consequently, according to current knowledge, Pst. cloelia does not produce more generations than P. charybdis and immature stages are not active for longer during the season. In chapter three, two sensitive quantitative PCR (qPCR) assays with species-specific primers that target the cytochrome oxidase I (COI) gene region were developed. Sequences and/or specimens for five of the six closely related paropsine species present in New Zealand at the time of assay development (i.e., Trachymela sloanei (Blackburn), Dicranosterna semipunctata (Chapius), Trachymela catenata (Chapuis), Paropsisterna beata (Newman), Trachymela sp. Weise) were included as non-target species in the assay development and testing to avoid non-target amplification. Assays reliably detected the DNA of both paropsine species within the bodies of field-collected predatory arthropods, thus identifying paropsine predators. Moreover, a decontamination protocol for field-collected predator individuals was developed to avoid false positives due to potential surface contamination. Natural enemies play an important role in top-down control of pest species. However, in New Zealand, P. charybdis and Pst. cloelia have escaped most of their natural enemies from their native range Australia. In chapter four, field collections quantified richness and abundance (number of individuals) of predator taxa in two E. bosistoana plantations. Additionally, a combination of qPCR analysis and field observations of predation events confirmed that six insect species, two mite species, and three spider families are paropsine predators and quantified their activity as the number of collected specimens that had consumed paropsines. Richness and abundance of predator taxa varied over time and between the two sites and was higher at the site with complete ground cover and neighbouring forestry trials. At the more predator-rich site, the most abundant insect predator, Oechalia schellenbergii (Guérin) (Hemiptera: Pentatomidae), was also the most active paropsine predator, whereas the most abundant spider family, Pisauridae, was disproved as a paropsine predator. The Tasmanian ladybeetle, Cleobora mellyi (Mulsant) (Coleoptera: Coccinellidae), was the second most active predator, but had low abundance in the field. The red mite Anystis baccarum (Linnaeus) (Acari: Anystidae) was abundant at both sites and, at times of high abundance, was frequently seen feeding on eggs and newly hatched larvae. Most predators fed on eggs, few on early-instar larvae, and only O. schellenbergii preyed as both nymphs and adults on all paropsine life stages from egg to adult. Based on abundance, field observations of predation events, and molecular analysis, O. schellenbergii, C. mellyi, and A. baccarum appear to be the most promising candidates to support future paropsine control in New Zealand. In chapter five, parasitism of P. charybdis and Pst. cloelia egg batches collected from E. bosistoana and E. quadrangulata trees at two field sites was quantified to assess the capabilities of the two established egg parasitoids, Enoggera nassaui (Girault) (Hymenoptera: Pteromalidae) and Neopolycystus insectifurax Girault (Hymenoptera: Pteromalidae), to control both P. charybdis and Pst. cloelia populations in the field. Simultaneous to egg collections, egg batch abundance of both paropsines was quantified on the two eucalypt hosts, revealing competitive displacement and host separation between P. charybdis and Pst. cloelia in the field. Paropsisterna cloelia competitively displaced P. charybdis on its most preferred eucalypt hosts, such as E. bosistoana, and consequently, P. charybdis switched to other suitable eucalypt hosts less preferred by Pst. cloelia, such as E. quadrangulata H.Deane & Maiden, if present at the site. If no suitable alternative hosts for P. charybdis were available at the site, P. charybdis was extremely rare. Hatch rates show that the two egg parasitoids controlled the second P. charybdis generation, but did not control Pst. cloelia populations, especially not on host trees most preferred by Pst. cloelia where it was most abundant. Parasitoids increased Pst. cloelia egg mortality through oviposition scars only on less preferred host trees where P. charybdis was also present and where abundance of Pst. cloelia was generally low. Finally, the main findings of all chapters are drawn together in a general conclusion to provide recommendations for integrated pest management of paropsine beetles and future research. Host separation and competitive displacement between P. charybdis and Pst. cloelia suggest that additive defoliation on the same host species is unlikely because P. charybdis appears to avoid direct competition by ovipositing on a different host than Pst. cloelia if available. Paropsisterna cloelia is not affected by the two established egg parasitoids and predators are generally not sufficiently abundant to have an impact on populations. Moreover, it has a higher reproductive output than P. charybdis and advantageous life strategies that may lead to a higher survival rate. The species may therefore have bigger population outbreaks and cause more damage than previously observed by P. charybdis on its most preferred species, such as E. bosistoana. To date, Pst. cloelia has not established in E. nitens or E. globulus plantations in New Zealand and the risk of it replacing P. charybdis there is questionable as the two paropsines prefer hosts belonging to different sections within the eucalypt subgenus Symphyomyrtus. To investigate Pst. cloelia’s future potential impacts, further research should concentrate on phenology in warm and wet regions of New Zealand, where it may have more than two generations. Identifying host preferences and quantifying development rates on different host species could indicate the eucalypt hosts that may be most impacted by Pst. cloelia. To exert effective paropsine population control, populations of paropsine predators need to be enhanced by attracting and retaining more individuals into forest plantations, and this could be achieved by using semio-chemicals, artificial supplementary food, and/or habitat improvement to increase the abundance and diversity of alternative food sources and provide shelter and overwintering sites. Future research could explore how to practically implement these measures in a cost-effective way and whether they can successfully increase predator abundance and pest mortality. Additionally, there would be the option to introduce specialist parasitoids for classical biological control of Pst. cloelia. The present study demonstrated that timing of life stages can vary considerably between sites even in the same geographic region and that monitoring the production of new foliage can provide guidance on how to time pest monitoring. Establishment of development models from laboratory development trials can facilitate predictions on the occurrences of paropsine life stages in the field for more effective timing of control measures. In the future, standardised monitoring procedures to assess pest abundance and defoliation need to be developed and economic thresholds for defoliation and pest populations determined. This is essential for an effective IPM programme and can increase financial benefits and reduce impact on the environment due to more efficient pesticide use.
Open Access
Evaluation of fused particle fabrication additive manufacturing as a recycling method of poly(ethylene terephthalate): Processing-structure-properties relationships
(2024) Hosen, Mohammad Sagor
Fused particle fabrication (FPF), an extrusion-based additive manufacturing (EBAM), is recently introduced as a cutting-edge upcycling technology that provides a fresh avenue for sustainable waste management of semicrystalline thermoplastics and diversifies the applications of recycled thermoplastics. Despite having tremendous potential, FPF application to semicrystalline thermoplastics encounters substantial hurdles rooted in the intricate microstructural transformations during the processing. The dynamic transformations in microstructure lead to warpage through differential shrinkage and consequential decline in the mechanical and thermal stability of the final product. Acknowledging the pivotal role of microstructure in determining the abovementioned properties, the present study investigates on the microstructural evolution and properties change of a semicrystalline thermoplastic, i.e. poly(ethylene terephthalate (PET), following multiple (four) cycles of FPF to establish the processing-microstructure-properties relationship. The study discerns alterations in nucleation, crystal growth, crystal dimensionality, crystallite size, overall crystallinity and microstructural contents (rigid and mobile amorphous components). The findings demonstrate that repeated FPF cycles significantly affect PET microstructure and properties. Initial FPF cycles promoted nucleation, while further cycles enhanced crystal growth with reduced crystallisation rates in the PET microstructure. The microstructural alteration has been identified as an underlying factor and significant contributor to the increased density and volumetric shrinkage, diminished tensile properties and improved thermal stability in the final part following each processing cycle. The altered microstructure decreases specific volume and increases volumetric shrinkage, suggesting a potential warpage of PET-printed parts after the 4th FPF cycle. A reduction in tensile strength, ductility, and brittleness was observed after multiple cycles, alongside an increased thermal-oxidative stability of the recycled PET. The research findings delving into the microstructural insights of semicrystalline thermoplastics will assist in yielding printed parts endowed with tailored mechanical and thermal properties with minimum warpage. The resultant knowledge will facilitate the development of the FPF technology as a sustainable upcycling tool for semicrystalline thermoplastics, marking a stride towards optimised EBAM practices.
Open Access
Hydrogen ironmaking with precipitated iron residues
(2024) McArdle, Joshua
Ironmaking, for the steel industry, contributes ~7% of global CO2 emissions by utilising carbon fuels for process heat and the chemical reducing agent required for reduction of iron ores to metallic iron. Hydrogen (H2) ironmaking can greatly decrease these emissions when H2 is produced from renewable energy sources and used for chemical reduction. Further improvement of ironmaking sustainability and longevity relies on increased process efficiency and the origin and quality of iron ore resources used. Globally, mined iron ores are decreasing in purity, leading to the need for greater ore beneficiation or the potential valorisation of waste iron material, the latter being the focus of this research. Iron’s high natural abundance makes it a common impurity in hydrometallurgical processes, and it requires removal through chemical precipitation, creating a voluminous, unstable, and often toxic precipitated iron residue (PIR) sludge. PIRs are generated on a large scale and often stored in waste ponds and landfills or, in countries with poor environmental regulations, disposed of without suitable environmental protection measures. Both cases create resource waste and potential environmental damage. To mitigate these impacts, there is increased motivation to utilise PIRs. PIR use in ironmaking is appealing due to ironmaking’s sufficiently large scale and growing demand for alternative high purity iron resources. The present research evaluated the utilisation of PIRs in H2-ironmaking, specifically in the New Zealand (NZ) ironmaking context where pellets are assumed to be used in a vertical shaft furnace process. PIR was first synthesised and later pelletized alongside NZ titanomagnetite (TTM) ironsand with PIR blends of 2, 5, 10 and 20 wt.%, and individually (0 and 99 wt.% PIR). A binder was dosed into pellets at 1 wt.%. The respective physical and chemical properties of these materials and pellets where then evaluated before and during the laboratory-based ironmaking experiments, including pelleting, sintering and H2 reduction. Results showed the 99 wt.% PIR pellets had significantly enhanced reduction kinetics compared to NZ TTM (0 wt.% PIR) after sintering at 1200°C. However, the 99 wt.% PIR pellets had decreased compressive strength and comprised considerable sulphur species, creating potential issues during industrial scale ironmaking and the subsequent steelmaking process. Therefore, it was concluded that blending of PIR with TTM was more suitable, with the addition of 5 wt.% PIR negating these issues. Pellets with 5 wt.% PIR demonstrated similar sintered compressive strength (compared with 0 wt.% PIR pellets), no increased sulphur content in DRI beyond practical levels for steelmaking, and an enhancement in H2 reduction kinetics was observed, with up to a 20% decrease in reduction time at higher temperatures. Additionally, the literature review estimated the potential global production of iron precipitates to be on the order of 100 Mtpa, sufficient for ~5 wt.% addition into existing ironmaking technologies or future H2-ironmaking. This potential blending would enhance resource valorisation and minimise harmful waste from hydrometallurgy, while providing the ironmaking industry with a viable alternative feedstock additive.
Open Access
4-dimensional elation Laguerre planes with 9-dimensional automorphism groups.
(2024) Findlay, Rosemary
Topological Laguerre planes are circle geometries with one parallelism relation and a topology on the point set and circle set such that the geometric operations of joining, touching, parallel projection and intersection are continuous. If the point set is finite dimensional, locally compact and connected then it is either of dimension 2 or 4 and the automorphism group of the Laguerre plane is a Lie group. The 4-dimensional elation Laguerre planes with automorphism groups that have dimension at least 10 are all known and have seen signi cant investigation since 1992. The aim of this research is to look at the 4-dimensional elation Laguerre planes with 9-dimensional automorphism groups. We prove that all 4-dimensional elation Laguerre planes with 9- dimensional automorphism groups have a parallel class which is fixed by the automorphism group. With this result we outline a method of classifi cation by drawing on the correspondence between elation Laguerre planes and dual translation planes. This reduces the classi fication to checking if the already classi fied 4-dimensional dual translation planes with group dimension greater than 7 can appear as derived planes at points of the fixed parallel class. We proceed using this method to prove that there is no 4-dimensional elation Laguerre plane with 9-dimensional automorphism group that has, as a derived plane at points of it's fixed parallel class, a dual-translation plane that has an automorphism group of group dimension 8 or more. We additionally construct a first example of a 4-dimensional elation Laguerre plane with 9-dimensional automorphism group thus proving that such planes do exist.
Open Access
Understanding the cascading flood consequences of dam operation and levee breaching in integrated dam-levee catchments.
(2024) Wallace, Thomas
Floods are among the most frequent and damaging natural hazards. The ramifications of flooding extend beyond the immediate physical danger to include damage to homes, infrastructure, agriculture, ecosystems, and public health that can take communities years to recover from. Flood frequency and magnitude will likely worsen in the future due to climate change and urbanisation. Recognising flood risks is increasingly crucial as populations exacerbate the risk by moving to concentrate in vulnerable areas. Understanding and addressing this flood risk is vital for developing effective mitigation strategies, resilient infrastructure, and safeguarding communities from adverse consequences. Flood protection systems play a pivotal role in reducing flood risk by shielding communities from the impacts of floods. These systems employ a range of strategies to manage and control water flow. Among these are defensive mechanisms, which include dams and levees. Levees increase the conveyance capacity of rivers to prevent inundation, ostensibly protecting 10% of the global population (ICOLD, 2022). Dams, in contrast, increase the storage capacity, allowing controlled releases and regulated outflows to mitigate the downstream flood risk. Both dams and levees are integral components of flood defence infrastructure, protecting urban and rural communities alike. However, these systems face increasing risk due to ageing infrastructure, urbanisation, the increasing complexity of management plans and stakeholder relations, and climate change. Understanding these systems and their inherent risk becomes critical for catchment management and selecting adaptation options. This thesis explores the effects of levee breaching and dam operation on flooding and the potential consequences of their interaction. This is done by simulating floods in computational models of four New Zealand dam-levee catchments. The research proposes novel methods for simulating levee breaching and representing dam management in models so impacts can be captured in future risk assessments. Combining these approaches, the research explores cascading consequences. The thesis advances flood risk assessments by acknowledging the interconnectedness of levees and dams. In addressing this gap, these systems may be improved to aid in addressing wider flood risks. If not omitted entirely, flood maps in practice that include levee breaches are often deterministic (van Kalken et al., 2007; Vorogushyn et al., 2010; Wild, 2017). The research developed a method for simulating levee breaches to determine the probability of inundation, given that a breach occurs. Unlike the current probabilistic methods, the method does not rely on computationally intensive Monte Carlo simulations and detailed geotechnical information. Empirical equations were used to set breach development parameters and final dimensions. The method ran one breach per simulation, initiated where the depth exceeded 50% of the levee height. Results from a historical flood in one catchment found that including breaching increased the average inundation by 48%. Including breaching in another catchment reduced the inundation by up to 12% as flows were contained between a levee and floodway. This highlights the possible use of floodways and fuse plug levees to cause preferential flooding. Given that the breaching predominantly flooded cropland and pastures, it may be prudent to develop these before the land is developed and becomes more expensive to purchase and repurpose. Managing dams is critical in not exacerbating flood risk. However, quantitative research on the importance of management in dam operation during floods is still developing. This thesis developed a method for representing stakeholder communication and understanding roles & responsibilities in a computational flood model by varying parameters such as the initial level, release timing, and maximum gate opening. These were simulated at various levels in a range of rainfall patterns and return periods, enabling a more nuanced perspective than previous studies. The hours of gate pre-release were more correlated with outflows in dams designed to provide flood control, as these had larger mechanisms to release water, allowing the flood volume to be evenly distributed throughout the event. However, these mechanisms can cause outflows greater than dams’ inflow without appropriate management. A different catchment showed that legislative regulations can remove sources of misoperation, enabling more consistent outflows, though these need to be created collaboratively to set realistic expectations. The volume five-days before the flood peak was more correlated with outflows in dams with large storages, as these relied on storing water to reduce outflows. However, the outflow reductions in these dams decrease in large events as they act as run-of-river schemes once full. Flow control mechanisms may make these dams more effective in large events and provide a potential climate change adaptation option. Cascading consequences are the chain reaction of events where the consequences of each event amplify or propagate the consequences. Understanding these cascading effects is crucial for comprehensive risk assessment. This thesis investigates the cascading consequences of dam operation and levee breaching. This was achieved by pairing the maximum, median, and minimum dam operation flows with the levee breaching method to create cascading consequence flood maps. Using the maps with land cover and building data, the consequences were calculated in terms of monetary building and land losses. In all four case study locations, at least 20% of the levee length reached the breaching threshold under median dam flows, and two locations reached at least 55%. Dam operations significantly impacted this potential breach length, with the average difference between the minimum and maximum dam flow cases being 25% of the total levee length. Including breaching tripled the potentially exposed area in half the case study catchments. The affected land was predominantly cropland and paddocks. However, this can be converted to higher-value orchards, vineyards, and built environments. Because of this, the potential losses will likely increase under climate change without additional measures to manage flooding. When the full range of scenarios is considered, there is at least a 950% increase in losses between minimum flow dam operation without breaching and maximum flow dam operation with breaching (i.e. best and worst-case scenarios). This range of cascading consequences is often unconsidered in modelling practices. By emphasising the risks, this research contributes to addressing the levee paradox, advocating against building developments in areas presumed to be flood-safe due to levees’ perceived absolute protection. This approach promotes greater resilience in flood-prone areas. As global flood losses increase, understanding our flood protection systems is vital for making informed adaptation decisions. As such, acknowledging risks like levee breaching, dam operation, and their cascading consequences is crucial. This thesis explores these topics, finding they significantly impact inundation and the resultant losses. Implementing this research in practice could transform the understanding of flood risk as by improving our knowledge of these systems, aiding in addressing broader flood risks and improving our communities’ resilience.
Open Access
Enhancing non-expert rescuers’ emotional experience in cardiac arrest emergencies : a demonstrative design of an emergency care device.
(2024) Chen, Zhen
Global ageing is leading to an increase in cardiac arrest incidents among senior citizens, posing a significant societal challenge. Most out-of-hospital cardiac arrest (OHCA) incidents occur at patient’s domestic residences, limiting their immediate access to professional help and Automated External Defibrillators (AEDs). In such instances, Cardiopulmonary Resuscitation (CPR) performed by caregivers, often the patient's family with no expert knowledge of emergency care, becomes crucial. Compared to professional rescuers, non-expert rescuers such as caregivers could face challenges in executing effective CPR due to skill gaps and potential emotional barriers in performing such procedures on family members. This study investigates the experience of caregivers in cardiac arrest scenarios, a critical but often overlooked aspect in the design of emergency care devices. We aim to understand 1) how emotions affect caregivers' performance and experience during domestic cardiac arrest incidents and 2) how design can support their practical and emotional needs, enhancing their performance. Semi-structured interviews with professional rescuers and caregivers reveal both emotional and skill-related challenges that caregivers might confront before, during, and after cardiac arrest incidents, such as unfamiliarity with rescue procedures, unskilled in CPR, physical strain, fear of approaching a collapsed person, anxiety about causing harm, lack of confidence, and moral pressure from social ties. These challenges can lead to adverse reactions that further hinder their CPR performance. The study highlights the importance of including emotional support for non-expert rescuers in OHCA incidents. By incorporating human-centred design principles, we propose an inclusive design guideline for emergency care devices and practical design recommendations to mitigate emotional barriers and assist operational performance for non-expert rescuers. Under these design guidelines, we conducted two follow-up interviews to engage non-expert rescuers in co-design sessions to propose an emergency care device concept. With the co-design results, this study established the final design concept for this device and created a digital prototype demonstration of our final design concept. Finally, we invited two non-expert participants to evaluate the demonstration through an experiential evaluation. This study contributes to the field by presenting a design demonstration of an emergency care device, emphasising the emotional and practical support to enhance the experience of non-expert rescuers in domestic OHCA incidents.
Open Access
Inference of effective rheological laws for shallow lava flow models from free-surface measurements.
(2024) Muchiri, David Kibe
To understand, describe, simulate, and predict the dynamics of geophysical lava flows for hazard mitigation, this research work presents mathematical modelling and simulations of three-dimensional shallow free-surface flows of Herschel–Bulkley viscoplastic fluids down an inclined non-flat topography with varying basal conditions. Starting from a systematic perturbation analysis, lower-order models of the Navier–Stokes equations are derived by employing the multi-regime approach, which allows to model different flow regimes originating from the variation of the mean-slope and/or the basal conditions. In particular, the lubrication model, the shallow water equations, and the depth-averaged heat equation are derived for multi-regime viscoplastic flows applicable to a natural setting consisting of an irregular topography with perturbations in basal slipperiness. Two flow regimes (Regime A, the basic/classical one, and Regime B, the enriched one with a corrective pressure term) corresponding to different balances between shear and pressure forces are defined and investigated. Flow models corresponding to these regimes are calculated as perturbations of the zeroth-order solutions. The classical reference models in the literature are recovered by considering their respective cases on a flat-inclined surface. The flow solutions of the two regimes are compared; the difference appears, in particular, in the vicinity of sharp changes of slopes. Nonetheless, both regime models are compared with experiments and are found to be in good agreement. Moreover, the derived shallow water equations and the lubrication approximation are compared to test their validity limits in different flow regimes in terms of the inclination angle, aspect ratio, basal perturbations, and Reynolds numbers. In the limit of low Reynolds numbers, both models are observed to agree well with viscoplastic experiments on a flat topography. The significant difference is observed at early times of dam-break, where the lubrication approximation overestimates the speed of the front position compared to the experiments. Also, it is shown that the flow dynamics over a perturbed topography are well predicted by the shallow water equations compared to the lubrication model. Generally, the shallow water equations are observed to have a wider predictive limit (e.g., in terms of the slope angle and Reynolds numbers) than the lubrication approximation. Furthermore, to investigate the effects of rheological parameters: power-law index (n), consistency index (K), and yield stress (τc), on flow height and velocity of Herschel– Bulkley fluids over different topographies, three practical examples of dam-break flow cases are studied: a dam-break on an inclined flat surface, a dam-break over a nonflat topography, and a dam-break over a wet bed. The effects of bed slopes and depth ratios between upstream and downstream fluid levels on flow dynamics are also analyzed. The numerical results are compared with experimental data from the literature and are found to be in good agreement. Results show that increasing any of the three rheological parameters decreases the fluid front position, peak height, and mean velocity, for both dry and wet bed conditions. Lastly, by formulating an inverse problem, we can infer the rheological parameters (K, τc, n) from free-surface measurements based on experimental data of flows around an occlusion. The rheological identification problem is formulated to minimize an objective function that measures the discrepancy between the elevation hydrographs from the model output and experimental data. The inverse solver is tested on both synthetic and laboratory data. The set of rheological parameters inferred is compared with the values measured on a rheometer for the fluid used in the experiments. The results have shown that hydrograph measurements at the wetted solid-fluid interface contain information of the fluid flow, which can be used to retrieve the unknown rheology for hazard mitigation and/or aid prediction of the wetting dynamics. Novel results include the derivation and validation of the shallow water models for 3D Herschel–Bulkley flows that account for basal elevation and basal slipperiness, a formal comparison of the lubrication and shallow water equations with experiments, numerical simulations of viscoplastic dam-break flows over a dry non-flat bed and over a wet bed, and the inference of rheological parameters for viscoplastic flows using elevation hydrographs around an occlusion.
Open Access
Characterization of fluid and thermal flows through 3D printed triply periodic minimal surfaces.
(2024) Hawken, Mathew Burnett
Porous materials are prevalent in both nature and industry. In industry, heat and mass transfer processes often require porous materials. Characterizing the flow and heat transfer facilitates the design of optimal geometries to meet the required performance, such as pressure drop, heat transfer or mass transfer rates. The often-random nature of porous materials hinders control of the geometry. Experiments or simulations can be employed to understand the fluid flow through porous structures. Magnetic resonance imaging (MRI) offers a range of techniques to determine flow and heat transfer and provide validation of simulations. This thesis aims to develop methods to characterize flow and heat transfer through an ordered porous material based on triply periodic minimal surfaces (TPMS). Additive manufacturing was used to generate a range of different TPMS structures. Initial investigations focused on determining the ability to produce the desired structure accurately and the effects of changing structure on the pressure drop. The pressure drop showed that for aspect ratios as small as 2.1, no wall effects were present, and the transition between laminar and turbulent flow occurred between Reynolds numbers of 60 and 600. Then, the validation of MRI sequences to investigate both velocity and temperature was conducted. First, a sequence to measure spatially resolved velocities was implemented on a commercially available low-field MRI machine. Furthermore, the flow field mapped by the sequence was validated inside a straight tube and later TPMS with volumetric flow rate discrepancies of 6% in both. Second, temperature measurements based on the proton resonance frequency method (PRF) were validated for a high-field MRI machine. The implementation involved a flow compensated spin echo sequence, which sampled both positive and negative times relative to the spin echo centre and detected temperature changes of less than 1°C. Finally, by combining velocity and temperature maps on a high-field MRI machine, the heat transfer and fluid flow were measured for the first time inside a TPMS structure. From the velocity and temperature maps, the local geometry was shown to cause variations in local temperatures and velocity. Comparison with an available numerical simulation highlighted the consistency between flow fields while isolating a potential discrepancy in thermal flow. These maps demonstrated the potential for MRI to characterize both the heat and flow inside a TPMS structure and the potential to discover previously unknown features.
Open Access
Triplet Matroids and Closure in Phylogenetics
(2024) Levy, Maayan
This thesis examines a meeting point of two combinatorial structures: matroids and phylogenetic trees. Matroids are a tool for generalising the notion of independence, with areas of application including graph theory, linear algebra, and coding theory among others. Meanwhile, phylogenetic trees are representations of ancestral relationships, with common applications in computational biology including hypothesising evolutionary histories, modelling disease transmission, and classifying species. These trees can be broken down into sets of rooted triples, which serve as a list of lineage constraints that encode a phylogenetic tree. Rooted triples are useful for determining tree compatibility and constructing supertrees from overlapping datasets. In this thesis, we investigate a class of matroids arising from sets of rooted triples that minimally encode trees, which we term triplet matroids. The matroid structure is useful for revealing which rooted triples in a set are essential and which are redundant. This ties in closely with the notion of closure, which concerns the lineage constraints logically implied by a set of rooted triples or phylogenetic trees. Our main result is that all triplet matroids are graphic. We provide a polynomial-time algorithm for constructing a graphic representation of a given triplet matroid, which serves as a helpful visualisation of the closure and dependencies of a set of rooted triples. In addition to this, we explore the unusual behaviour of the class of triplet matroids under deletion and contraction. We also demonstrate the effects of modifying a phylogenetic tree on its associated triplet matroid. Finally, we characterise the class of graphs that arise as graphic representations of triplet matroids. An additional chapter departs from triplet matroids, focusing instead on network greedoids. Greedoids and phylogenetic networks are generalisations of matroids and phylogetic trees respectively. We prove that a greedoid structure arises from exactly the networks that are tree-child, thus providing an additional characterisation of the class of tree-child networks. For future investigation, we introduce the greedoid polynomial and pose a question about its ability to distinguish network greedoids.
Open Access
Numerical simulation of two-phase flows with the conservative level set method.
(2024) Rate, Keetley
In this thesis, we will develop an implementation of Olsson and Kreiss’s conservative level set method. Do achieve this, we must cover key aspects of fluid dynamics, numerical methods for PDEs, including the finite element method. Prior to discussing aspects of two-phase flow, we first overview the foundational principles of fluid dynamics, as detailed in chapter 2. This chapter entails derivations of the governing equations of fluid flow which follow from conservation of mass and momentum, as well as some fundamental solutions to them. In chapter 3, we discuss how approximate solutions the Naiver-Stokes equation may be obtained via a splitting integration method and a projection step. To complete this, we discuss how the Finite Element Method is used to find the numerical solution a general boundary value problem, allowing us to solve the equations arising from projection methods for the Naiver-Stokes equations. This will give us the necessary tools for simulating two-phase flows in chapter 4, where we will see how the level set method can be used to solve the free-surface problem inherent in multiphase flows. This will leave us with a numerical method capable of tackling various problems with practical applications, some of which are detailed in chapter 5. In chapter 6, we examine one aspect of non-Newtonian fluid behavior within the context of viscoplastic or yield stress fluids. Here, we investigate the fundamentals of viscoplastic dynamics through the idealized Bingham fluids, which provide a simple model of viscoplasticity. We will also see how regularizing viscosity can be used to find the numerical solution to viscoplastic flows, allowing the modeling two-phase viscoplastic flows with a conservative level set method.
Open Access
A Kaupapa Māori approach to the Storage and Collection of Taonga Seeds
(2024) Shadbolt, Marcus-Rongowhitiao Te Puni
Due to the eﬀects of climate change and widespread ecological destruction, we are seeing global species loss on an unprecedented scale. In response to this, seed banking has become one method of storing at-risk species safely, while simultaneously supporting ecological restoration. Seed banking has therefore become a vital practice globally for ensuring the continual supply of seeds, in both agricultural and conservation projects. In Aotearoa, knowledge of how to store native seeds is limited, as the local science system has yet to truly utilise it as a method of conservation. This thesis therefore aims to look at both the technical aspects of how to store seeds native to Aotearoa, and what this may look like ethically, legally, and appropriately from an Indigenous Māori perspective. The technical part of this thesis focused on ﬁve species of the Coprosma genus and aimed to ﬁnd the optimal germination method for each one, as well as whether these species show signs of desiccation or freezing sensitivity. Of my study species, C. robusta was identiﬁed as orthodox, while C. propinqua, C. rugosa, C. rhamnoides, and C. autumnalis are all varying degrees of non-orthodox. Among them, C. propinqua is intermediate with decreasing viability as temperatures decreased, and C. autumnalis was completely recalcitrant with no germination after drying. Coprosma rugosa and C. rhamnoides are both intermediate but with a signiﬁcantly lower number of germinations than in C. propinqua. More research is needed on these species, speciﬁcally into how long in storage these species can last, in the case of those which can be stored safely. The cultural aspect of this thesis, however, focused on addressing the past injustices faced by Indigenous peoples, speciﬁcally Māori, in science and conservation, while discussing how to build an appropriate and ethical seed banking system from the outset in Aotearoa. This chapter aimed to bring together both international policy and legal precedents from Aotearoa related to seed ownership. Based on these, I propose a set of best-practice guidelines for working with Māori in relation to seed banking. These protocols bring together the current literature on appropriate engagement, and personal experiences of myself and colleagues as Māori people working in the environmental space. Ultimately, between these two seemingly separate aims, the overall goal of this thesis is to support the growth of the relatively new seed banking sector in Aotearoa, so that as the nation progresses, we do it from an ethical and appropriate position.
Open Access
Computational bioacoustics for the detection of rare acoustic events
(2024) McEwen, Ben
The field of computational bioacoustics is a rapidly developing area of research. Tradi- tional bioacoustic monitoring has a long history of use particularly for monitoring of avian species. Species that were previously infeasible to monitor due to resource and time constraints are now a possibility. In New Zealand, bioacoustic monitoring has never been applied to the detection and surveillance of invasive terrestrial mammals. This dissertation presents the first research of its kind investigating the use of computational tools for the detection of invasive species. We evaluate the use of computational methods for the detection of common brushtail possums (Trichosurus vulpecula) a species that is invasive to New Zealand we also investigate the potential of acoustic monitoring of other challenging target species such as mustelids and rats. Low population densities (such as post-eradication efforts) encounter extremely sparse detections where it becomes challenging to reliably differentiate between species absence and non-detection. This issue of non-detection is accentuated by the use of passive acoustic monitoring technology at a landscape-scale. Thousands of hours of raw acoustic data can be collected weekly. Most of this data is empty of events of interest. It is infeasible for a human to manually analyse this data. The post-processing of acoustic data is a common challenge for passive acoustic monitoring applications. Rare feature detection (especially for challenging applications such as incursion detection and probability of absence testing) is currently at the border of feasibility using state- of-the-art computational bioacoustics tools. We present an active few-shot learning methodology that combines semi-supervised prototypical learning methods for efficient analysis of acoustic data with limited existing samples. We evaluate this methodology on an invasive species detection dataset demonstrating high performance at a range of data availability contexts. This methodology achieves a test accuracy of 98.4% (with fine- tuning) as well as 81.2% test accuracy using 2-shot, 2-way prototypical learning without fine-tuning, demonstrating high performance at varying data availability contexts. The development of improved methodologies capable of detecting rare acoustic features has clear benefits for other bioacoustic monitoring applications. This technology can be applied to other challenging applications such as monitoring of rare and at- risk species. These methods can be applied to scale the efficiency of data analysis of existing bioacoustic monitoring applications to achieve landscape-scale monitoring. Invasive species detection using bioacoustics represents a challenging but valuable area of research. We present the development of publicly available methodologies, tools and the first invasive species dataset containing 3500 labelled samples and over 1300 samples of brushtail possum vocalisations. The results of this work indicate the potential of computational bioacoustic methods for the detection of invasive species as well as rare and at-risk species that encounter similar monitoring challenges. In addition to improved bioacoustic detection methods, we also investigate the use of bioacoustic noise reduction methods evaluating both signal processing-based methods and deep audio enhancement methods. We investigate the efficacy of noise reduction for the performance of downstream segmentation and classification tasks identifying the limitations of common perceptual metric-based approaches. We find that noise reduction results in no improvement in segmentation precision and recall with an average AUC performance decrease of 19.2%. We also demonstrate no benefit in classification accuracy when tested using state-of-the-art time-domain and time-frequency-domain audio classification models with a marginal decrease in average validation accuracy of 0.41%. We contrast these findings with common perceptual metrics which demonstrate consistent increases in perceptual quality when noise reduction is applied with SnNR increases ranging from 14.0% to 41.3%. We also present initial work that investigates the use of visual detection methods. We develop predictive tracking methods used to improve population estimates of low- resolution and low-frame rate thermal cameras. We discuss the use of these tools within the context of predator-free New Zealand and the current pest management paradigm.
Open Access
Robotic arm path planning for autonomous grape vine pruning.
(2024) de Wet, Andrew
The autonomous grape vine pruning robot being developed by the Computer Vision Research Group at the University of Canterbury will be capable of imaging vines to create a 3D model of them, navigating around orchards, and pruning vines with a high degree of freedom robotic arm. This thesis develops the system to plan and execute motions for the robotic arm. Path planning for robotic arms involves finding a sequence of joint positions in the robot’s configuration space that can be moved through to create a motion in Cartesian space that avoids any collisions and gets the end effector of the robot, in this case a pruning tool, to a goal position and orientation. Because of the infeasibility of representing the grape vine as an obstacle in the robot’s configuration space, sampling based path planning algorithms that only collision check specific randomly sampled joint positions offer the best approach. Robotics Operating System (ROS) is used as a framework that enables development with a simulated robot and operation of the real robot. The six degree of freedom UF850 by Ufactory is selected with its suitability to the application verified through simulating path planning on 30 synthetic vines, where for all vines a valid pruning strategy given the points the arm could reach can be determined. An implementation method where single paths are found between cut points and a safe plane offset from the vines is proposed to minimise the distance traversed by the planning algorithm and to configure the problem in a way that leads to high quality solutions. The performance of 10 different path planning algorithms with different features from the literature is tested. Following a successful extensive evaluation, Informed RRT* is the selected algorithm with an average length of individual paths of 1.88 seconds and 81.9% of possible cut points reachable. The proposed implementation to prune whole vines maximises the quality of the paths found by using all of the time spent executing one cut to plan the next. While operating the real UF850 it is able to successfully execute a realistic pruning strategy on all of the 20 vines tested. It does so with an average time per cut of 10.85 seconds and an average time per vine of 139.3 seconds. These results demonstrate that once integrated with the full rover the system developed provides an effective approach to autonomous grape vine pruning.
Open Access
Practical pavement distress detection via photogrammetry.
(2024) Zhang, Meng
Overall, the proposed approach in this thesis offers a cost-effective, accessible, and accurate system for pavement distress detection, with the integration of fast perspective transformation and spatial data analysis playing a pivotal role in improving the assessment and management of road infrastructure. The structure of this thesis is organized as follows: • Introduction: This section sets the stage for the research by highlighting the importance of pavement distress detection, the challenges of traditional methods, and the contributions of the thesis. • Literature Review: This section provides a comprehensive overview of the existing literature, covering both background of general models in the field of computer vision and specific techniques in pavement distress detection, or anything directly related to the thesis: ‣ Background – Traditional Automated Techniques: Discussion on sensor-based solutions, stereo vision methods, and detection with structured light. – Deep Learning in Computer Vision: Examination of the evolution of deep learning models in computer vision, as well as important architectures such as the U-Net, the R-CNN, the YOLO model family, and latest transformer-based networks. – Depth Estimation Methods: Exploration of stereo vision and disparity maps, monocular depth estimation, and the integration of depth perception with an example. – Geographic Information Systems (GIS) Integration: Analysis of how GIS can be integrated into pavement distress detection. ‣ Related Work: Review of recent studies and research papers in the field of pavement distress detection, focusing on the methodologies, techniques, and findings. – Deep Learning based Distress Detection: include transfer learning and previous work on neural network based pavement distress detection models. • Data Collection, Cleansing, Labeling, and Transformation: This section describes the methodology and equipment selection, transportation selection, data collection strategy, data cleansing, data labeling, and data transformation processes. • Model Re-implementation and Tweaking: It focuses on model selection and re-implementation (Mask R-CNN), as well as model tweaking and hyperparameter optimization. • GIS Analysis for Pavement Distress Detection: This section discusses the integration of spatial data analysis using Geographic Information Systems (GIS) to provide geographic context to the pavement distress data. • Conclusion and Future Work: Providing the key findings of the thesis and outlines potential directions for future research.