← Back home

thomasholland.uk/cv

Thomas Holland

Software and science, mostly pointed at the Earth.

Thomas Holland

Projects

University of Cambridge | 2025-2026

Improving Methods for Estimating Modern Ice-Driven Sea Level Change using Bayesian Inference

  • Developed infinite-dimensional Bayesian inversion for recovering ice-thickness change directly from satellite and ice altimetry.
  • Built Gaussian-prior structure with heat-kernel covariance and based on physically motivated priors to separate ice and firn contributions.
  • Extended framework to joint inversion over ice thickness, firn compaction, and ocean dynamic topography, using a derived joint operator matrix factorisation to significantly reduce computational overhead in numerical modeling.
  • Implemented in Python using PySLFP and pygeoinf, validating robustness under synthetic twin experiments and prior misspecification.
GitHub Report

University of Cambridge | 2025-2026

Bayesian Data Assimilation and Forecasting of Chaotic Fault Slip

  • Implemented a custom grid-based exact Bayesian filtering for nonlinear state estimation of a chaotic frictional fault model in Julia, with Rauch-Tung-Striebel smoothing for retrospective posterior refinement.
  • Used the near-2D chaotic attractor of a two-state-variable spring-slider to propagate full posteriors without linearisation or ensemble approximation.
  • Evaluated calibration with z-score diagnostics and tested forecast skill against a persistence baseline across deterministic and stochastic truth scenarios.
  • Identified noise on unobserved fault-state variables as the main limit on probabilistic forecast skill.
GitHub Report

Rust simulation tooling | 2025

Seismic Wave Forward Modeller

  • Built 2D elastic wave simulator with staggered-grid finite differences, multi-source injection, and CFL-safe timestep selection.
  • Parallelised time-stepping with Rayon and added energy, divergence, and curl diagnostics for debugging and analysis.
  • Structured experiments through TOML configuration for repeatable runs without code changes.
GitHub

Julia simulation | 2026

Dual Limit Cycle SDE

  • Built a 2D stochastic dynamical system with two stable limit cycles at arbitrary positions and radii, where noise drives rare Kramers-type transitions between them.
  • Implemented Itô SDE integrator with a gradient potential encoding both cycles exactly — the barrier height ΔV maps cleanly onto the Kramers escape rate for tuning transition timescales.
  • Added parallel Fokker-Planck PDE solver (conservative upwind advection, central-difference diffusion, RK4) that evolves the probability density ρ(x,y,t) on a grid with async snapshot rendering.
  • Included event-driven transition detection via a ContinuousCallback that fires at the equidistant surface separating the two basins.
  • Acted as a playground and precursor to my Bayesian Data Assimilation and Forecasting of Chaotic Fault Slip project.
GitHub

Education

University of Cambridge | 2025-Current

Master of Science, Geological and Earth Sciences/Geosciences

Dissertation: Bayesian Inversion for Sea Level and Ice Sheet Signals. Focused on geophysics and computational methods, with additional study in dynamical systems and chaos, seismology, paleontology, major paleo-environmental change, and critical mineral resources.

University of Cambridge | 2022-2025

BA Natural Sciences, Geological and Earth Sciences/Geosciences

Part II: Earth Sciences, including field project, petrology, geophysics, and ancient life and environments. Earlier years combined Earth Sciences with plant sciences, biology, evolution, and mathematical biology.

The King's School, Worcester | 2014-2021

A*A*A* in Biology, Chemistry and Maths with A* in Extended Project Qualification

← Back home