Weiyue Zheng

Research title: Development of a spatio-temporal data fusion framework for integrating point and gridded soil moisture observations

Research Summary

I'm a final-year PhD candidate in Statistics at the University of Glasgow, supervised by Professors Marian Scott, Claire Miller, and Andrew Elliott. I develop methodology for spatio-temporal data fusion under spatial misalignment, integrating point-sensor data with gridded satellite data. Methodologically, I work on Bayesian hierarchical models (INLA–SPDE), graph-regularised machine learning (Geo-XGBoost with Laplacian penalties), and conformal prediction for uncertainty quantification. My applications focus on soil-moisture mapping and short-term forecasting.

Research interests

  • Spatial & Spatio-temporal Statistics: Gaussian random fields (Matérn/SPDE), separable vs non-separable covariance models, change-of-support and misaligned covariates, hierarchical/latent-process modelling.
  • Data Fusion: Probabilistic fusion of point and areal data, multiresolution modelling, handling missingness.
  • Geospatial Machine Learning: Geo-XGBoost with graph-Laplacian regularisation and custom objectives. Uncertainty Quantification: Conformal prediction (local/spatio-temporal).
  • Applications: Soil-moisture mapping, short-term prediction; integration of in-situ sensors with Copernicus satellite data and related products.
  • Computing & Data: Reproducible R/Python pipelines (INLA, XGBoost, PyTorch), HPC workflows.

Supervisors

Teaching

  • Time Series
  • Spatial Statistics
  • Introduction to R and Python
  • Flexible regression
  • Generalised linear model