This repository contains research code supporting a manuscript on monthly rainfall interpolation in Singapore. The method uses climate-model simulations to define a Gaussian-process prior, calibrates station-specific noise from observations, and applies ordinary kriging to the remaining spatial residuals.
The manuscript is under revision and this repository is still being organized. This README is a temporary overview; a publication-ready release will add a clean environment specification, data provenance, an end-to-end reproduction workflow, and detailed documentation.
utils.py: core Gaussian-process interpolation and evaluation functions.two_stage_validation.ipynb: leave-one-station-out validation of the two-stage method.kge_three_row_composite.ipynb: consolidated KGE comparison and diagnostic figure.sn_initialization_sweep.ipynb: station-noise initialization sensitivity experiment.gaussian_check.ipynb: station–month rainfall distribution diagnostics.data/: observational and simulation inputs used by the notebooks.figures/: generated manuscript and revision figures.
Run the notebooks from the repository root. The current workflows use the standard Python scientific stack together with scikit-learn, PyKrige, GeoPandas, and Cartopy. Exact dependency versions and reproducibility instructions will be provided during the planned repository cleanup.
Citation information will be added when the associated manuscript is published.