Browsing by Author "Groot, Paul"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Open Access
    Capturing transients: an application of biostatistics to astronomy
    (2022) van Dyk, Anke; Mcbride, Vanessa; Groot, Paul
    Capture-recapture has been identified as a possible use case for estimating the underlying size of astrophysical transient populations. In this work, we present a series of exploratory analyses using capture-recapture methods from biostatistics. In the first of three separate analyses, we reproduce results of Laycock (2017). Strategically sampled X-ray lightcurves of simulated populations of high mass X-ray binaries (HMXBs) are used to probe estimator behaviour and efficiency. Overall, these statistically closed population estimators converge to the input population with increasing number of observations, yet estimator efficiency is shown to be significantly be affected by sampling strategy. I then employ nonstandard estimator models to account for variations in capture probability of individuals within the population, categorised into ‘behavioural', ‘temporal', and ‘heterogeneous' effects. In the second analysis, we present a methodology for closed population capture-recapture analysis using real data from the OGLE-IV XROM survey. The data samples consisted of observations that were grouped into epochs. The large variation in quiescent magnitude of the population creates heterogeneity in the capture probability of sources which requires non-standard modelling. Estimation of population size is therefore limited by the choice of observational magnitude threshold. Bias corrected estimation proves to be potentially useful in this context. In the third and final investigation, we present a ‘robust design' approach with a population of Dwarf Nova located towards and in the Galactic Bulge identified from the OGLE-II, -III, and -IV phases. This approach combines closed and open population practices that allows new individuals identified between the survey phases to be added to the study sample for dynamical estimation. These investigations provide a future course for population size estimation of transients and variable stellar population alongside population synthesis simulations. The generation of capture histories remain non-trivial through the choice of observation grouping, brightness scale, and imposed flux threshold. Further, there remain several unexplored avenues of inquiry and refinement for this methodology pertaining to astronomy using explanatory variables in the modelling. Recommendations are made for further exploration of the topic.
  • No Thumbnail Available
    Item
    Open Access
    Developing instrumentation and software for rapid follow-up and characterisation of near-Earth Asteroids
    (2025) Ngwane, Thobekile sandra; Erasmus, Nicolas; Groot, Paul
    Near-Earth Asteroids (NEAs), a subset of minor bodies in the Solar System, result from resonant interactions with major planets, particularly Jupiter, leading to their escape from the main asteroid belt. The International Astronomical Union's Minor Planet Center (MPC) database, as of December 2024, lists approximately 37,000 discovered NEAs, with an average daily discovery rate of 10 from dedicated survey programs like Catalina Sky Survey (CSS), the Panoramic Survey Telescope and Rapid Response System (PanSTARRS), and the Asteroid Terrestrial-impact Last Alert System (ATLAS). This project uses the robotic observing capabilities of the South African Astronomical Observatory's 1-meter telescope, Lesedi, equipped with the Mookodi instrument. Observations are scheduled in robotic mode using automated Python scripts, enabling rapid follow-up of newly discovered NEAs, often within the same night of detection. This rapid response is essential, as smaller asteroids (< 300 metres)—a significantly understudied group— quickly dim as they move away from Earth, making precise measurements challenging. Since the start of this project in February 2023, approximately 230 NEAs have been successfully observed in robotic mode, with an average absolute magnitude (H-magnitude) of 24.4. This magnitude corresponds to asteroid sizes ranging from 32 to 78 metres, depending on an assumed albedo of 0.05 to 0.30. Approximately 75% of these asteroids have a diameter (D) of less than 100 metres. Among the observed NEAs, 15 have been classified as potentially hazardous asteroids (PHAs). The findings presented in this study are based on multi-filter photometry and astrometric measurements collected as part of the program. The astrometric data significantly contributes to the MPC's orbital refinement and the observed NEAs designation. Photometric observations using g, r, and i filters enable the extraction of g - r and r - i colours, which approximate the spectral slope. These colours aid in determining the most likely taxonomic type (S, C, X, D, Q, or V-types in this project) of the observed NEAs, as defined by the Bus-DeMeo Classification Scheme. This provides insight into their composition. Using the collected data, the compositional distribution of the small NEA population was determined and compared with previous studies investigating a larger size population.
  • Thumbnail Image
    Item
    Open Access
    Machine Learning techniques to discover and understand the population of flare stars in MeerLICHT data
    (2022) Bangiso, Aphiwe; Groot, Paul; Buckley, David; Johnston, Cole
    In this era of information overload, machine learning and artificial intelligence have been increasingly popular in various fields, including the field of astronomy. These approaches attempt to extract meaningful information from the data through automated means. In this work, we develop generic machine learning models that classify a given transient object from the observed light curve. We train random forest (sect 4.1.1) and multilayer perceptron neural network (sect 4.1.3) models on simulated LSST PLAsTiCC data and real data from the MeerLICHT survey. We found that the random forest model outperforms the neural network model in both data sets, achieving test accuracy of 66.0% and 98.0% in the PLAsTiCC and MeerLICHT data respectively. On the other hand, the neural network model achieved test accuracy of 65.7% and 86.6 % in the PLAsTiCC and MeerLICHT data respectively. For PLAsTiCC simulated data, we also show that grouping all types of supernovae into one aggregate class and discarding distance information improves the performance of both models to 96.5% and 96.0% for random forest and neural networks respectively. As additional work, we attempt to find sub-classes within the M-type class in MeerLiCHT data using k-means and hierarchical clustering algorithms. We find two distinct sub-classes in this data. Namely variable and non-variable M-type stars.