Prof. Mohammad Belal
I am an atomic, molecular, and optical (AMO) physicist with broad research experience in experimental soft condensed matter, optoelectronics, high-power fiber lasers, optical-atomic-molecular spectroscopy, coherent atomic control, and quantum technologies. At NOC, I apply this expertise to develop next-generation intelligent monitoring systems for complex environments, including oceans, glaciers, cities, and terrestrial landscapes. Ocean Fibre Sensing (OFS), is one such initiative that I lead. OFS exploits optical-fibres within legacy submarine cables of energy and telecommunications, and transforms them into dense arrays of sensors with unprecedented (dense space-time) ambient environmental observations, alongside enabling cirtical insights into cable-health. Advanced signal processing/analytics combined with bespoke machine learning and artifical intelligence algorithms, not only bakes intelligence into OFS but also preserves efficient analysis of large, multi-dimensional datasets, on high-performance compute architectures. On-the-fly analytics to data-streams (>10 GiB/s) has enabled exploration of alternative compute architectures (e.g., Quantum-based) to overcome classical compute bottlenecks.
My work combines physics, engineering, advanced signals-analysis and data science to provide high-resolution, wide-area insights, transforming complex measurements into clear visualizations and actionable information. Ultimately, these efforts aim to understand, manage, and respond to changes in marine, maritime, glacial, and urban environments, helping to quantify and protect our planet’s critical natural assets.
Projects
Ongoing PhD projects:
1. Coherent processing of ambient marine noise (Mr. Konstantinos Drylerakis; 2022-2026)
Brief description: Leveraging the spatial organisation of underwater DAS measurements across multiple stages of analysis, the project demonstrates how data-driven processing and physical modelling can complement one another to enable richer large-scale analysis of ambient marine noise. Motivated by the potential for DAS maritime applications, physically informed criteria are introduced to identify local vessel-related components within a data-driven spatiotemporal singular value decomposition framework. The resulting subspace isolates relevant structure and supports marine source detection and tracking tasks. The analysis is then extended to the spatial correlation structure of the measurements in order to enhance vessel-related signatures embedded in strong ambient marine noise. A physically derived feasible set of spatial correlation matrices is constructed as an ideal reference against which local empirical correlations can be assessed and processed, enabling interpretable feature extraction and model-conditioned denoising. Finally, the same correlation framework is incorporated as a stochastic augmentation within self-supervised representation learning to support physics-informed deep feature extraction. The resulting representations exhibit improved downstream utility and class-wise organisation relative to generic self-supervised pretraining. Thereby, highlighting their potential for DAS feature extraction pipelines
2 Learning Speech Representations Based on Koopman Constrained Temporal Dynamics (Mr Nikos Chazaridis; 2022-2027)
Brief description: Speech signals contain multiple overlapping factors of variation, including speaker identity, linguistic content, and recording conditions. Learning representations that separate these factors is challenging, particularly without explicit supervision. This thesis investigates how temporal variation in speech can guide disentangled representation learning using Koopman operator theory. It addresses the challenge of separating speech attributes according to their temporal dynamics. The thesis makes three main contributions. First, it explores the requirements and limitations of learning temporally disentangled representations with a Koopman inspired architecture for separating persistent and faster-varying components on a several synthetic scenarios. Second, it introduces a dual-encoder speech autoencoder that separates speaker characteristics from linguistic content, incorporating a novel multistep Koopman prediction objective and evaluating the representations in speaker verification and phoneme recognition. Third, it proposes a temporal decomposition strategy for self-supervised speaker encoders that distinguishes persistent recording conditions from speaker information, aiming to improve speaker recognition under noisy, real-world conditions.
3. Physical sciences PhD projects | The University of Edinburgh
4. (CURIOSITY)Cryosphere and Underwater Remote Inspection and Observation using Optic-fibre based Sustainable noise-InTerferometrY (Mr. Gareth Dowsett, 2026-2030)
Brief description: Technological advances in ocean and ice observation have made it possible to measure phenomena across a wide range of scales. However, a crucial blind spot remains in our observing capabilities: no established approach exists to robustly measure the interior at scales O(10-100 m) over a significant spatio-temporal extent. Important phenomena lying within the above observational blind spot permeate almost every topical problem in physical oceanography and ocean climate change. Furthermore, observations of oceanic processes under ice are very challenging, typically involving high risk autonomous missions. Unlocking progress in tackling these problems thus requires a new way of observing the ocean. This project will demonstrate a novel, low-cost, wide-area approach combining marine cables with natural sound to characterize the ocean and cryosphere. This has been demonstrated previously for the ocean using underwater microphones. Within this project there will be two major advances on previous work (1) the extension of existing techniques to utilise the very large data sets provided by optical fibre sensing and (2) develop the application of the methodology in ice covered oceans for both seafloor and on ice cables. These will require a range of mathematical approaches to: (1) the calculation of background noise cross-correlation on short time scales using empirical greens functions or kernel (gaussian and polynomial) methods; and (2) numerical modelling using a combination of underlying theory across hydrodynamics, ice dynamics and acoustic propagation. A range of data sets suitable for this approach are already available and will be supplemented with upcoming fieldwork.
Funded (ongoing and upcoming) projects:
- SBRI-2020 (PI): Submarine High-fidelity Active-monitoring of Renewable energy Cables (SHARC)
- Trans National Access' 2021-A (PI): Fibre-optic Intelligent Submarine High-fidelity Environmental Sensing (FISHES)
- Trans National Access' 2021-B(PI): Sea-floor High-fidelity Optic-fibre-based Renewables-infrastructure Sensing and Evaluation (SeaHORSE)
- Progeny-Phase1 2022 (PI): Next Generation Underwater Sensing
- PycnoGen 2022-2027 (EP/X025136/1; Co-I): Generation of the global ocean internal pycnocline in the ice-covered Southern Ocean
- Progeny-Phase2 2023 (Co-I): Next Generation Underwater Communications
- NSF-NERC 2024-2027 (NE/Z000408/1; UK-PI): Collaborative Research: Direct comparison of ocean temperature and velocity structure from in-situ measurements and distributed optical fiber sensing
- GSRF 2024 - 2026 (NE/Y003365/1; PI)& Wide-area low-cost sustainable ocean temperature and velocity structure extraction using distributed fibre optic sensing within legacy seafloor cables
- Emso-Eric 2024 - 2026 (PI): Submarine noise-Evaluation and Analytics using Low-cost Sustainable -sensing (SEALS)
- MoD/QQ 2024 (4 months; PI): Geospatial Acoustic Underwater Smart Sensing (GAUSS)
- Xlinks 2024 (PI): Monitoring system evaluation
- SOUNDSCALE 2025 - 2027 (MR/Z505845/1; Co-I): Sensing On Urban Noise: Distributed Sensing For Collaborative And Sustainable Cityscapes And Living Environments
- FULL-OCEAN-FIBRE 2025 - 2028 (ARIA: Forecasting Tipping Points)
- AtlantiS 2024-2029 (NE/Y005589/1): Novel data-processing & interpretation routine development for submarine fibre-optic cables data
Roles Provided
I lead Ocean Fibre Sensing (OFS) research at NOC. It comprises optoelectronic interrogation of optical fibres within submarine cables of energy and telecommmunications combined with advanced signal processing/analytics to enable intelligent, sustainable, wide-area, real-time monitoring of diverse and complex environments, e.g., oceans, glaciers, cities, and terrestrial landscapes whilst also monitoring cable-health. The intelligence layer is anchored in custom signal processing pipelines built in combination with machine learning and AI algorithms, allowing efficient analysis of large, complex, multi-parameter datasets. Compute enabling these advancements comprises advanced GPU-CPU architectures besides quantum compute exploration which is also underway to overcome classical-compute bottlenecks.
