Getting Started with Spatial Analysis

The spatial workflows run on Python with GeoPandas for vector data and Matplotlib for rendering. This page gets you from a clean environment to a first rendered boundary map.

Create the environment

The notebooks were authored on Python 3.12. Create an isolated environment with the geospatial stack:


conda create -n geo-spatial -c conda-forge python=3.12 \
  geopandas matplotlib pyreadstat xlrd openpyxl pandas numpy
conda activate geo-spatial

geopandas pulls native libraries (GDAL, PROJ, GEOS). Installing from conda-forge is the reliable path — pip install geopandas often fails to resolve the system GDAL.

Get the source

The notebooks are versioned in the repository; their input layers live on the Hugging Face dataset nhic/rwanda-spatial, not in git. Download and unzip them into the layout the notebooks expect:


pip install -U huggingface_hub
cd analysis/spatial-analysis
hf download nhic/rwanda-spatial --repo-type dataset --local-dir hf
unzip -o hf/rwanda-admin-boundaries.zip -d rwanda
unzip -o hf/cells-admin4.zip           -d Cell_Shapefile
unzip -o hf/villages-admin5.zip        -d Village_shapefile
unzip -o hf/wetlands.zip               -d Wetland
unzip -o hf/environmental-layers.zip   -d shapefiles
cp hf/climate-data.xls "Climate data.xls"

Load a boundary shapefile

Each administrative level is a single shapefile. Reading one returns a GeoDataFrame:


import geopandas as gpd
 
sectors = gpd.read_file("rwanda/rwa_adm3_2006_NISR_WGS1984_20181002.shp")
print(sectors.crs)          # EPSG:4326 (WGS84)
print(len(sectors))         # ~416 sectors

Plot it


import matplotlib.pyplot as plt
 
fig, ax = plt.subplots(figsize=(9, 8))
sectors.boundary.plot(ax=ax, linewidth=0.3)
ax.set_axis_off()
plt.show()

This renders the sector geometries — the base layer every choropleth builds on:

Where things live

Path	Contents
`analysis/spatial-analysis/*.ipynb`	The `Climate_maps` and `Bar-Plots` notebooks
`analysis/spatial-analysis/rwanda/`	NISR admin boundaries (`admin1`–`admin4`)
`analysis/spatial-analysis/Village_shapefile/`	Village boundaries (`admin5`)
`analysis/spatial-analysis/shapefiles/`	Water, wetland, and land-cover context layers
`analysis/spatial-analysis/Climate data.xls`	ERA5-Land climate variables by health facility

To regenerate the map images, run the generator — it executes both notebooks and writes the PNGs to public/analytics/spatial-analysis/ (git-ignored), then publish them to Hugging Face:


python analysis/spatial-analysis/generate_figures.py
bash scripts/upload-spatial-to-hf.sh

The site build itself stays Node-only — Python is only needed to regenerate figures, never to serve the docs.

Next steps

Climate MapsFollow the full pipeline from raw ERA5 data to a two-panel choropleth.Data ReferenceEvery layer, its source, and its coordinate system.