eofigs

Analysis-aware, publication-ready figures for Earth observation research.

Study area map	Trend map	Event composite
DEM accuracy (ICESat-2)	False discovery rate	House style

More examples in the gallery.

eofigs turns Earth observation analysis outputs (rasters, time series, validation pairs) into journal-ready figures that know the analysis, the variable and the printed size:

• Recipe figures for common analyses: trend with significance, event composites, and later recovery, burn severity, disturbance and divergence.
• A variable library with fixed colour maps, symbols and units, so NDVI, LST or GPP look the same in every figure, and their anomalies and trends get zero-centred diverging colours.
• Print-size rendering: rasters are read and reduced to the pixels the printed figure needs, from its width in mm and its dpi, keeping memory use and PDF size small.
• Journal presets and one-call export to vector PDF plus 600 dpi PNG or TIFF, with a check of size, resolution, smallest font, font embedding and colour-blind safety.

Installation

pip install eofigs

eofigs needs Python 3.11 or newer. On Windows and macOS, conda-forge is the simplest source of its geospatial libraries; see the installation guide.

Quick start

import numpy as np
import eofigs

eofigs.set_style("elsevier")  # fonts, sizes and widths of the journal
rng = np.random.default_rng(1)
field = rng.gamma(5, 30, 300)  # observed aboveground biomass (Mg ha⁻¹)
model = 0.9 * field + rng.normal(10, 20, 300)  # predicted
fig = eofigs.accuracy_scatter(field, model, variable="agb")
print(fig.eofigs_layout["caption"])  # a ready caption sentence
print(eofigs.check_figure(fig))  # size, dpi, smallest font, fonts, colour-blind safety
eofigs.save_figure(fig, "fig_agb")  # vector PDF and a 600 dpi raster

More in the quickstart and the gallery.

Run with Docker

A container image with eofigs and its geospatial stack (cartopy, rioxarray, geopandas, fonts) is published on GitHub Container Registry for each release (0.2.0, latest; dev follows the main branch):

docker run --rm -v "$PWD":/work ghcr.io/kmkamilkhel/eofigs:0.2.0 python script.py

Roadmap

Available now: set_style, style, new_figure, label_panels, save_figure, check_figure, cvd_preview, tidy_axes, palettes, variable, register_variable, fit_raster, trend, stats.fdr_mask, figure_set, study_area_map, add_north_arrow, add_scale_bar, add_inset, trend_map, event_composite and accuracy_scatter.

Released in 0.2: false discovery rate control of significance marks and the seasonal Mann–Kendall test.

Planned: burn severity, recovery, disturbance and divergence recipes, SHAP and partial dependence styling, confusion matrix, Google Earth Engine input through Xee at print scale; an R package with matching function names, after the Python API is stable.

Contributing

Questions and ideas are welcome in Discussions; bugs and feature requests in Issues. See CONTRIBUTING.md and the Code of Conduct.

Citation

If you use eofigs, please cite it. The software is archived on Zenodo: doi:10.5281/zenodo.24053760 covers all versions, and each release has its own DOI (v0.1.1: doi:10.5281/zenodo.24084780). The metadata are in CITATION.cff; the citation page has APA and BibTeX entries.

License

MIT. See LICENSE. Data files have their own licences, listed in DATA_LICENSES.md.