BISMA
Not a member yet
5916 research outputs found
Sort by
Juvenile survival and adult phenology (breeding and moult) in kororā / little penguins (Eudyptula minor albosignata) at Pōhatu / Flea Bay, New Zealand.
No full textThe kororā/little penguin (Eudyptula minor) is the smallest of the extant penguin species (~1 kg) and is found across New Zealand and Australia. Little is known about juvenile survival of kororā except for colonies at Phillip Island (Australia) and Oamaru (New Zealand) that have had ongoing monitoring programmes that span over 30 years. At the Pōhatu/Flea Bay colony on Banks Peninsula, New Zealand a monitoring programme was established in 2022 to provide scientific support to over 40 years of conservation work at Pōhatu/Flea Bay. The monitoring programme has provided data for a baseline study into juvenile survival and adult phenology of kororā at this location. This thesis has two distinct studies presented in two data chapters. Both recognise that due to the restricted range of kororā at Pōhatu, they must be researched as an independent unit. It is also established that the survival of juveniles and the phenological timing of adults is tied directly to the health of the local marine environment. In the study of juvenile survival (chapter 2), weekly monitoring data that records the presence of kororā in the nest and nesting activities (e.g. presence of eggs, chicks) and records of chicks marked (n=358) over 3 seasons (2021, 2022, 2023) was used to carry out a capture mark recapture analysis to determine how many kororā have been resighted in nest boxes in the colony (n=44). I also used the capture mark recapture data to model apparent survival using Cormack-Jolly-Seber and generalised linear modelling to determine what life history traits could influence survival outcomes (weight at fledging, whether the bird had been rehabilitated, and the time the chick was in the nest). I found that 12.1% of marked fledged chicks have been resighted in Pōhatu/Flea Bay including rehabilitated kororā (n=15). The apparent survival rate for juvenile kororā over three seasons was 13% which is comparable to other colonies even though it does not cover a full recruitment period (approximately three – four years). The model that best explained apparent survival of kororā included increased weight of the chick at fledging and whether the chick had been in rehabilitation.
The adult phenology study (chapter 3) also used weekly monitoring data combined with publicly available environmental data to determine if environmental variables, e.g. sea surface temperature, chlorophyll-a concentrations, air temperature and wind speed, affected phenological events across breeding and moult for marked adult kororā (n=163) at Pōhatu/Flea Bay. I found that over the 2022-23 season, chicks fledged earlier (on average 6 days) and adults moulted later (on average 21 days) which was related to a marine heatwave event which peaked at nearly 20oC in January 2023. During the 2022-23 breeding season there was also 84 failed eggs, 74 dead chicks and 71 uplifted chicks indicating mass starvation caused by this event. I suggest that the shift in fledging and moult timings over this stage were in response to this one-off event. Together these findings provide baseline information regarding juvenile survival and adult phenology of kororā at Pōhatu/Flea Bay. This research also indicates support for conservation measures including rehabilitation of underweight kororā as well as giving support to measures to protect the local marine environment. This is especially important as ocean temperatures are predicted to increase, adding additional pressure to this important seabird species
Tree Canopy Cover in Te Awamutu 2021
No full textThe aim of this report is to provide local authorities in New Zealand with a basic understanding of the urban tree canopy cover within their cities and towns
Science Communication: Southshore Spit as a Dynamic Interface. Report prepared part of the GEOG309 Research for Resilient Communities and Environments course, University of Canterbury, 2025.
No full textThis project, working with Justin Cope from Environment Canterbury and Deidre Hart from the University of Canterbury, researches some of the processes and systems which make up the Southshore Spit, a dynamic coastal interface sitting on the South Islands east coast in Christchurch. The research investigates both the natural influences on the system and the anthropogenic and using science communication to provide the community with information on these, improving awareness and resilience. This project aims to use short form content to positively and effectively communicate coastal processes to residents and the wider Christchurch community.We were tasked by our community partners to make our videos digestible to the public so, as a group we brainstormed a series of video ideas and produced the final four of: dune processes, vegetation and ecology, sediment budget and management and the historical and social significance of the spit. From these ideas we storyboarded the scripts and camera shots, ensuring simple and effective language was used, acquired DSLR cameras and drone footage, and filmed. Then, using Final Cut Pro and feedback from our peers and community partners, we edited and changed the videos to create the final products.We found that short-form content was effective at communicating science in both an accessible and engaging way. This was discovered from the reviews of our peers following our presentation and from sharing them to high school students at the Youth Innovation Hub at the Adaptation Futures 2025 conference. Simple language and captivating imagery kept the attention of viewers whilst conveying important information.We reviewed that dune formation and stability are closely associated with the presence of vegetation, Sloss et al. (2012) & Silva et al. (2016), and that the Southshore Spit is supplied sediment from the Waimakariri River from a process called longshore drift, Hicks et al. (2018). Further review found that vegetation also plays a key role in coastal protection, Feagin et al. (2010) & van Puijenbroek et al. (2017) and highlights the cultural and historical significance of the spit to Māori, Ngāi Tahu & Ngā Rūnanga o Waitaha (2013). Carass (2021), Rugrien (2023) & Prindle et al. (2024) found that short-form content is both preferred and helpful for communicating information and engaging the viewers.Limitations of this research included the inability to communicate more in-depth information due to the lack of time in videos and the difficulty of getting our content to reach older demographics who tend to be less online. Further research should focus on sound design improvements, longer-form content development to further develop ideas and additional short form videos to better portray the various aspects that make up the spit
Prediction of anomalous and extreme weather events in Antarctica : an application-focused study on the use of deep learning in environmental data science.
No full textIn this thesis, we present an application-based study on the use of deep-learning methodologies
for detecting and predicting extreme and anomalous weather events in Antarctica.
Motivated by the increasing frequency of climate extremes and the underrepresentation
of polar regions in machine learning (ML)-based environmental studies, we explored
three core research areas:
(1) In the first project, we present the first-ever study to analyse the use of unsupervised
and deep learning methods for detecting and predicting foehn wind events. The results
demonstrate that unsupervised models can successfully identify foehn days with competitive
accuracy, providing a promising foundation for location-independent classification
methods. Furthermore, the best-performing deep learning models demonstrate strong
performance on both observational (macro F1-score of 0.95) and simulated data (macro
F1-score of 0.79). (2) The second project focuses on forecasting weekly Antarctic sea
ice concentration anomalies at the subseasonal time scale using spatiotemporal deep
learning models. Multiple graph-based models alongside a conditional diffusion model
(CoDiCast) are tested on a novel superpixel-aggregated graph representation of Antarctic
sea ice. The results show that the best-performing graph model reduces forecast error
by up to 33% over the persistence baseline. Additionally, we provide new insights into
the physical drivers of sea ice variability using explainable AI techniques. (3) In the final
project, we introduce the first ML-based study for predicting average daily temperatures
and heatwaves for the Antarctic continent. To achieve this goal, we created a novel
graph-based dataset by integrating satellite and automatic weather station (AWS) data
and designing a diffusion convolutional recurrent neural network (DCRNN). Compared
to the Antarctic Mesoscale Prediction System (AMPS) and DeepMind’s GraphCast, the
DCRNN demonstrates superior performance in both accuracy and heatwave classification.
The results highlight the potential of deep spatiotemporal models to enhance local
weather forecasting in remote and data-scarce regions
Chasing storms : tracking event flow in a mountain catchment.
No full textSnow-dominated mountain catchments are important sources of freshwater for local downstream users. However, in a warming world, the hydrological processes that take place in these mountainous regions are changing. For this reason, the continuous monitoring of mountain catchments is crucial for up-to-date and effective management of water resources downstream. This thesis attempted to quantify source contributions to streamflow through the use of mixing model analysis to better understand these hydrological processes. A limitation of this model is the requirement for each hydrological component (precipitation, snow, snowmelt, and groundwater) to have minimal isotopic variation while being isotopically dissimilar from one another. Therefore, the primary aim of this thesis is to evaluate the mechanisms and processes that control the isotopic composition of source waters to a Canterbury mountain stream.
This thesis characterised the natural variation of meteoric waters, shallow groundwater, and streamflow in the Camp Stream catchment, Southern Alps, New Zealand from April to December 2024. This study incorporated the use of field-based techniques, laboratory work, and isotope analysis, to better understand the hydrological processes in the Camp Stream catchment.
This study has produced the following findings:
1. The major mechanisms controlling the variability of the isotopic composition of waters within the Camp Stream catchment are seasonality and the origin of the moisture source.
2. Isotopic deviations of Camp Stream occur simultaneously with large influxes of new water from precipitation events or snowmelt.
3. The preferential flow pathway within the Camp Stream catchment is via sub-surface flow, as groundwater is the predominant contributor to the streamflow of Camp Stream.
4. The natural storage of water as snow is likely changing in the Camp Stream catchment, which in turn has possibly changed the timing and extent of peak streamflow in Camp Stream
Boosting in reflected feature spaces.
No full textIn this thesis, we investigate reflecting the training data of binary classification problems
with quantitative predictors when training a model using Adaboost. We use depth-one
decision trees (stumps) as weak learners. Three methods of reflection were investigated:
(1) a random reflection, (2) a reflection which aligns the direction of maximum variance in
one of the classes to the first coordinate axis and (3) a combination of the two approaches.
We benchmarked the three reflection approaches and compared the classification error,
F1 and AUCdiscrete to the standard Adaboost algorithm with no reflections. We found
that in all but one dataset (11 out of 12) the second reflection method achieved lower
classification error than the standard Adaboost algorithm
Enriching LiDAR forest inventory with height-based competition indices.
No full textAt a trial forest in Canterbury, New Zealand, the efficacy of distance-dependent, distance-independent, and crown area inclusive competition indices in improvement of tree-height based Diameter at Breast Height (DBH) prediction, were compared at the plot and individual tree-level, for the purpose of improving LiDAR based individual tree DBH prediction, and thus its implementation in the forestry industry.
Distance independent competition indices were found to be the most useful, and improved tested individual tree models’ DBH prediction residuals by up to ±0.6 cm. The most useful indices were Stage’s (1973) (DII6) and Tomé and Burkhart’s (1989) (DII10) indices, modified to consider height instead of DBH. Both considered the target tree’s height in relation to only taller trees. Where models included indices which were not DII6 & DII10, the number of neighbours included in the index’s calculation was statistically significant. With the Individual Tree Approach (ITA), each number of neighbours, two, four, six, and eight, was statistically significant from every other number, with eight neighbours improving residuals the most. The best ITA model had an RSE of 2.88 cm, and a R² 88.6%. The best Area Based Approach (ABA) model , where tree variables were summarised at the plot level, had an RSE of 1.86 cm, and a R² of 94.5%. Both best-performing ABA and ITA models included Box-Cox transformations for distribution normalisation
Oceanic diplomacy: Reasserting indigenous pathways through the contemporary Pacific - Foreword
No full textBUSINESS TENURE, OPERATING HOURS, AND COMPETITORS: EFFECTS ON CUSTOMERS AND FINANCIAL IMPLICATIONS
No full textThis study aims to examine the effect of business tenure, operating hours, and the number of competitors on the number of customers for street vendors in Medan, along with the financial implications of customer traffic. Using a quantitative approach with a case study method, data were collected through structured questionnaires, involving direct numerical input provided by street vendors. The research was conducted in selected districts of Medan using purposive sampling, with a total sample of 30 respondents. The data were analyzed using multiple linear regression. The findings indicate that, collectively, business tenure, operating hours, and the number of competitors significantly affect customer volume. However, on an individual basis, only operating hours had a statistically significant impact, suggesting that longer operating hours are positively associated with attracting more customers. In contrast, business tenure and competition did not show a meaningful influence. The regression model met the assumptions of normality and showed no signs of heteroscedasticity. From a financial perspective, the results imply that optimizing daily operating hours may enhance customer reach and, subsequently, improve income potential, whereas simply having longer business experience or facing more competitors does not necessarily guarantee higher earnings
Waste-derived CaO from green mussel shells as a highly stabilized and superior sorbent for cyclic CO2 capture
No full textThis study aims to mitigate CO2 emissions by utilizing and developing diverse waste
natural resources as calcium oxide precursors, prepared through the citric acid-assisted sol-gel
method, to improve capturing capacity in cyclic carbonation/calcination reactions. The study
evaluates the influence of carbonation temperature ranging from 550-750 °C, and CO2 flow rates
(comprising 15 vol% of CO2 gas feed) ranging from 9-15 mL/min and assesses the stability of
CaO derived from waste natural resources concerning cyclic CO2 capture performance.
Furthermore, characterization of CaO derived from waste natural resources and the spent sorbent
was conducted using nitrogen adsorption-desorption analysis, X-ray diffraction (XRD), X-ray
fluorescence (XRF), carbon dioxide temperature-programmed desorption (CO2-TPD),
thermogravimetric analysis (TG/DTG), and field-emission scanning electron microscopy (FE-87 SEM)