Observations and modelling of intensity timeseries for biomedical and astrophysical applications

Abstract

Statistical benchmarking was successfully applied to quantify and resolve below-noise-floor observations of intensity timeseries in both biomedical and astrophysical contexts. Noise models, and fully synthetic camera images were generated from Randox biochip tests, using statistical Z-score analysis and Monte-Carlo modelling techniques. These models were used to benchmark and develop a noise suppression code downstream. Additionally, image analysis software was created to automatically co-align and pre-process successive frames of observation. This resulted in greatly improved identification and extraction of below-noise-floor signatures, increasing the sensitivity and speed of important interventional biomedical testing, and directly leading to improvement in patient outcomes. These same techniques were then adapted for use for astrophysical analysis, in the investigation of stellar nanoflares. These small-scale flare events are below the noise floor for observation, but through combined statistical and Fourier analysis, alongside observation driven modelling, I extracted the underlying nanoflare characteristics in fully convective M dwarf stellar lightcurves. This analysis uncovered the hidden science in these stars, which are dominated by nanoflare energy hidden below seemingly quiescent and noisy lightcurves. There are two key parameters which govern the frequency and lifetime of flare signatures: α and τ. Flare events are believed to be governed by a power-law, equating the frequency of events to their rate of energy release. The index of this power-law is α, and is a key indication of the rate of energy release. Higher values of α correspond to more frequent flaring at lower energies. The other key nanoflare characteristic is the e-folding time or ‘τ’. This is the time taken by a flare to lose approximately 1/e of its original energy through radiative cooling. In an initial study, I found the first observational evidence for stellar nanoflare activity in 3 fully convective M dwarf stars. These exhibited power-law indices in excess of α ≥ 3, a value much enhanced from power-law indices seen in other stars at higher energies. I then completed a follow up study, linking the enhanced rate of nanoflare activity to the convective nature of the stars finding that partially conductive MV stars exhibited little to no nanoflare activity signature. In contrast, fully convective MV stars exhibited α and τ values consistent with frequent nanoflare activity. This was likely due to altered plasma conditions, resulting in increased plasma resistivity and consequently enhanced Sweet-Parker reconnection at the nanoflare energy scale. This link to plasma resistivity suggests that the altered stellar dynamo is not involved in nanoflare enhancement, lending support to emerging theories of solar-like dynamo activity in fully convective stars. This holds implications for the nature of the dynamo in our own Sun. The role of stellar nanoflares in answering these–and other–key questions have only just begun. This synergistic statistical analysis has led to improved healthcare diagnostics and broadened our knowledge of poorly understood stellar nanoflares. Going forward, this style of cross-disciplinary research holds the potential to advance knowledge, while providing a key financial incentive to carry it out. Note that a redacted version of the thesis is available until 31 December 2027. After this point, the full, unredacted version will be visible.