Skip to content

Home

rsd - Reference-based standardization framework for drought indices under distribution shift rsd - Reference-based standardization framework for drought indices under distribution shift

A Python package for standardized hydroclimate drought indices (SPI, SSI, SDI, SPEI) with a hybrid empirical-parametric, reference-based estimator: empirical CDF core, generalized Pareto tails for extrapolation.

The fixed-reference design preserves the real differences in magnitude between, say, a factual climate run and a counterfactual, between historical and future projections, or between two observational products. Without a fixed reference, every series is mapped to its own local distribution and these differences vanish.

What you get

  • rsd.standardize for 1-D series and rsd.standardize_xr for N-D xarray grids (with optional dask parallelism).
  • Three methods sharing a common interface: the hybrid empirical-core / GPD-tail rsd method, plus monthwise ECDF and monthwise parametric baselines for direct comparison on the same input.
  • rsd.diagnose for sanity-checking a reference series before standardizing (optional, pip install rsd[diagnostics]).

The Method overview covers the advanced features (bounded variables via a logit pre-transform, custom warning categories, plotting-position choices) and the scientific rationale.

Where to go from here

  • Installation - install profiles for NumPy-only, xarray, and diagnostics.
  • Quick start - 1-D and N-D worked examples.
  • Method overview - the three-problem chain, the three-method solution, and when to prefer one method over another.
  • Examples - runnable notebooks.
  • API reference - every public function with signatures and parameters auto-generated from the source.
  • How to cite - the WRR methodology paper and the software DOI.

Companion paper

This package implements the methodology described in:

Tsilimigkras, A., Grillakis, M., & Koutroulis, A. (2026). A reference-based standardization framework for hydroclimate drought indices under distribution shift. Manuscript submitted to Water Resources Research. DOI pending acceptance.

If you use rsd in your research, please cite the paper as the primary reference; the software DOI is supplementary. See How to cite.