This research focuses on finding patterns in oceanic Planetary Boundary Layer (PBL) by analyzing satellite imagery and the outputs of machine learning (ML) algorithms. The PBL, located in the lowest part of the atmosphere (~1000m) is nearly always turbulent while the flow above the PBL is comparatively smooth. The downward transfer of momentum from the atmosphere above the PBL into the ocean and the exchanges of heat and water vapor between the ocean and atmosphere occur in the PBL. Understanding and modelling these exchanges is an important aspect of climate science. Even though the PBL is turbulent, its flow spontaneously generates organized coherent secondary circulations in the form of small convective honeycomb-like cells (MC) or long wind-aligned overturning rolls (WS). These flow patterns modulate wind-generated cm-scale ocean surface waves. The Sentinel-1 satellite constellation carries microwave (5 cm wavelength) radars that capture very high-resolution images of the ocean surface. The images are 20 x 20 km and are spaced by ~100 km, but sample nearly all the global oceans with each satellite acquiring ~65,000 images per month. The images are analyzed to find patterns indicative of WS or MC structures in the PBL. Several machine learning (ML) algorithms have been developed to analyze these images and predict whether the PBL above the image site contains WS or MC structures. I focus on a subset of 2100 images acquired in a small region of the tropical Atlantic Ocean; each having been hand-classified by a panel of five experts. My goal is to assess the ML models and calibrate a new ML model according to analysis of their outputs. I anticipate analyzing multiple patterns, including variance throughout the day-night cycle, seasonal changes, and geographical trends.

Dehydration is a silent but pervasive health risk, particularly for older adults in assisted living home settings, where prevalence rates can reach up to 60%. Medications that increase fluid loss place seniors at a heightened risk, leading to severe complications including urinary tract infections, falls, cognitive decline and hospitalisations. Caregivers continue to struggle to monitor fluid intake effectively, with less than 10% maintaining consistent hydration logs. Existing hydration monitoring solutions are often invasive, expensive and poorly suited for non-medical care settings. To address this critical issue, we developed a novel, non-invasive hydration monitoring system designed for elderly care environments. Unlike existing methods that rely on highly variable sweat salt concentrations, our approach leverages ultrasound-based elasticity measurements to assess hydration status. Changes in hydration levels alter the biomechanical properties of skin and muscle, affecting the speed at which ultrasound waves travel through tissue. By using a dual-transducer system to induce and measure shear wave propagation, we can quantify hydration status in real-time. The device provides both quantitative readouts for longitudinal tracking and intuitive qualitative feedback, similar to a blood pressure monitor's high-normal-low classification, ensuring ease of use without specialised training. Initial testing demonstrates promising accuracy and usability, positioning our solution as a practical solution to improve hydration management, prevent dehydration-related complications, and enhance quality of life for elderly residents. By empowering caregivers with a reliable, accessible hydration monitoring tool, our solution has the potential to significantly reduce healthcare costs, improve patient outcomes, and transform hydration care in aging populations.