Research
My research develops statistically principled and computationally scalable methods for spatial and spatio-temporal data, especially when relationships vary locally across space.
I develop localized variable-selection and multiresolution frameworks designed to recover spatially heterogeneous associations. Current work includes local LASSO/SCAD-type procedures, wavelet representations, and spatial intensity models for event data.
My work in environmental statistics studies extreme and threshold-exceedance behavior in complex spatial systems. Applications include climate extremes, streamflow, air pollution, and risk mapping.
I am interested in scalable approximations for spatial mixed models and latent Gaussian structures, including restricted likelihood, ICAR-type effects, and structured variational approximations.
I investigate graph neural networks, deep spatial learning, and network-aware prediction where geometry and spatial dependence should be integrated directly into model architecture and loss construction.
Recent work considers intensity and pair-correlation estimation when events live on curved domains or Riemannian manifolds, including measurement-error-aware local estimation.
I am interested in feature selection and learning for irregular spatial transcriptomics, fMRI, and other high-dimensional spatial biomedical datasets.