Website Β· Python package Β· Package docs Β· PyPI Β· How the database works Β· Quickstart notebook Β· Cite
The GeoDevelopment Observatory of Bolivia is a public-access platform for analyzing, monitoring and evaluating sustainable regional development in Bolivia, at every spatial scale: departments, provinces, municipalities, cities and grids. It has three parts:
- GeoDevelopment Indicators & Tools: social, economic and environmental indicators from satellite imagery, censuses and administrative records, at every spatial scale, with the Python packages, notebooks and web apps to load, build and analyze them.
- GeoDevelopment Projects: research projects built on the observatory: the subnational SDG Atlas, regional shocks and spillover effects and monitoring the space-time dynamics of local pollution.
- GeoDevelopment Insights: accessible research summaries.
It serves:
- Researchers, who need ready-made subnational panels with reproducible code
- Decision-makers, who design, target and evaluate regional policies
- Citizens, who want to understand development in their own municipality or department
Its indicators build on the Atlas de los ODS en Bolivia 2026 of SDSN Bolivia (the Sustainable Development Index, its 15 goal indices and 81 indicators near 2012 and near 2024) and on DS4Bolivia (population, night-time lights and satellite embeddings).
| This repository: quarcs-lab/gdo4bolivia | The package: quarcs-lab/geo4bolivia | |
|---|---|---|
| Holds | The data (data/, geo/, metadata/), the originals and the builders that regenerate them (raw/, builders/), the notebooks and the website |
The Python package geo4bolivia, its tests and its documentation |
| Website | quarcs-lab.github.io/gdo4bolivia: indicators, tools, projects and insights | quarcs-lab.github.io/geo4bolivia: get started, articles and the function reference |
| Use it | Every file is a plain URL; nothing to install | pip install geo4bolivia (PyPI) |
The package reads the files of this repository, so the data can be updated without a new release of the package. Both repositories carry the same version tags: ref="v0.7.0" in the package loads the data as they were at that release.
!pip install -q geo4bolivia # every snippet runs in Colab
import geo4bolivia as geo
import plotly.express as px # preinstalled in Colab
geo.datasets() # what is available: scales, years
gdf = geo.load("sdg", years=2024, geometry=True) # 343 municipalities, boundaries
px.choropleth_map(gdf, geojson=gdf.geometry, locations=gdf.index, color="index_sdg",
hover_name="name", center={"lat": -16.3, "lon": -64}, zoom=4)The package also runs the standard analyses, each in one call that returns a figure, a table and a reading of the numbers in words:
!pip install -q geo4bolivia
import geo4bolivia as geo
mpi = geo.load("sdg01", variables="sdg1_2_mpi", groups=["department"])
geo.map(mpi, "sdg1_2_mpi", year=[2012, 2024],
scheme="fisher_jenks").show() # both periods, common classes
res = geo.lisa(mpi, "sdg1_2_mpi", year=2024) # where does poverty cluster?
print(res.interpret()) # the result in words
res.fig # the interactive cluster mapAll data share one spatial backbone. Every unit at every scale has a geo ID that nests by prefix (numbered as in GADM 4.1): municipality BO-04-17-04 (Nuestra SeΓ±ora de La Paz) lies in province BO-04-17 (Murillo), which lies in department BO-04 (La Paz), which lies in Bolivia BO. INE codes, DS4Bolivia's asdf_id and GADM codes are kept as columns of geo/units.csv. The municipalities are the 343 municipalities and indigenous autonomies (GAIOC) of SDSN Bolivia's municipal map of 2025, the units of the SDG Atlas 2026; provinces, departments and the country are their exact unions. Each dataset is a panel with one row per unit (geo_id) and year, published as CSV and Parquet.
| Dataset | Title | Scale | Period | Files |
|---|---|---|---|---|
sdg |
SDG Atlas 2026: the Sustainable Development Index (IDS) and its ranking, with census population | 343 municipalities | ~2012, ~2024 | data/municipality/ |
sdg01 β¦ sdg17 |
SDG Atlas 2026, one dataset per goal: the goal index and its 81 indicators in all | 343 municipalities | ~2012, ~2024 | data/municipality/ |
pop |
Population (gridded estimates) | 333 municipalities | 2001β2020 | data/municipality/ |
ntl |
Night-time lights per capita (log) and its long-run trend | 333 municipalities | 2012β2020 | data/municipality/ |
emb |
Satellite embeddings, 64 dimensions, area-weighted | 333 municipalities | 2017 | data/municipality/ |
embpw |
Satellite embeddings, 64 dimensions, population-weighted | 333 municipalities | 2017 | data/municipality/ |
geo |
Units, names, codes, boundaries and crosswalks for every scale | 466 units | β | geo/ |
Each dataset has three descriptions:
- a Frictionless datapackage
- a generated page in
docs/datasets/, with its variable dictionary, coverage, sources and citation - its rows in
definitions_gdo-bolivia.csv, which lists every variable of every dataset
The SDG Atlas 2026 measures each indicator twice, near the 2012 census and near the 2024 census; the panels store the two periods as 2012 and 2024, and each variable's actual reference years are in its metadata. New municipalities keep being created in Bolivia: since version 0.4 the municipalities are SDSN Bolivia's 343-unit map of 2025. The 339 municipalities of GeoBolivia's 2015 map kept their geo IDs (numbered as in GADM 4.1) and the four units created since were numbered after them. Population, night-time lights and satellite embeddings were computed on the 2015 polygons, so they cover the 333 municipalities whose territory did not change (how maps change).
1. With the package, geo4bolivia (get started, function reference). It finds the variables and their documentation, and loads them as labelled panels with names and boundaries:
!pip install -q geo4bolivia
import geo4bolivia as geo
geo.search("poverty") # variables, datasets, years
geo.info("sdg01") # coverage, sources, licenses
mpi = geo.load("sdg01", variables="sdg1_2_mpi",
groups=["department"]) # ~2012 and ~2024, with names
geo.describe(mpi) # labelled summary per period
geo.change(mpi, "sdg1_2_mpi") # change and rank change per unit
geo.aggregate(mpi, to=["department", "country"],
dataset="sdg01") # census-population weighted
geo.combine(["ntl", "pop"], scale="municipality") # two datasets, one panel
geo.load("ntl", ref="v0.7.0") # a frozen release of the data
geo.map(mpi, "sdg1_2_mpi", year=2024) # interactive map (Plotly)2. Without the package. Every file is a plain URL that pandas and geopandas (preinstalled in Colab, like plotly) read directly:
import pandas as pd
import geopandas as gpd
import plotly.express as px
BASE = "https://raw.githubusercontent.com/quarcs-lab/gdo4bolivia/main/"
df = pd.read_csv(BASE + "data/municipality/sdg.csv") # geo_id, year, variables
units = pd.read_csv(BASE + "geo/units.csv") # names, parents, codes
gdf = gpd.read_file(BASE + "geo/boundaries/municipality.geojson") # polygons
gdf = gdf.merge(df[df.year == 2024], on="geo_id")
px.choropleth_map(gdf, geojson=gdf.geometry, locations=gdf.index, color="index_sdg",
hover_name="name", center={"lat": -16.3, "lon": -64}, zoom=4)Every analysis function takes a dataset id or a table with a geo_id column, finds the boundaries, the spatial weights and the labels itself, and returns a result with .fig (an interactive Plotly figure), .df, .gt (a formatted table, where one helps) and .interpret() (the numbers read in words):
!pip install -q geo4bolivia
import geo4bolivia as geo
res = geo.beta_convergence("sdg", "index_sdg") # did lagging municipalities catch
print(res.interpret()) # up between ~2012 and ~2024?
res.fig| You want to know | Functions |
|---|---|
| What does a variable look like? | describe, distribution, extremes, correlation, scatter, trend |
| Is it clustered in space, and where? | map, weights, connectivity, moran, lisa |
| Are the units that started lower catching up? | beta_convergence, sigma_convergence, convergence_clubs |
| Who moved, and do neighbours matter? | rank_mobility, lisa_transitions, transitions, spatial_transitions |
| How unequal are the units, between or within which groups? | inequality, theil_decomposition |
| What goes with what, once space is accounted for? | regression, panel_model, spatial_diagnostics, spatial_model |
| Does an association change from place to place? | mgwr (multiscale GWR) |
The articles of the package's documentation work one question each on these data, and the function reference has an example for every function.
The package can also:
- link your own tables to the same units, by names in any spelling or by INE and DS4Bolivia codes
- extract indicators from rasters
- move data onto grids or other zones
- save the results in the same documented format
- map and analyse them with the same functions
!pip install -q "geo4bolivia[raster]" # [raster] adds rasterio for geo.zonal()
import pandas as pd
import geo4bolivia as geo
mine = pd.DataFrame({"mun_id": [20101, 70101], "year": 2020,
"teachers": [9120, 11034]})
mine = geo.link(mine, scale="municipality", on="mun_id",
key="mun_id") # INE code 20101 -> BO-04-17-04
ntl = geo.zonal("viirs_2020.tif", scale="municipality",
name="ntl", year=2020) # your raster: ntl_sum, ntl_mean
panel = geo.combine(["pop"], scale="municipality", years=2020, extra=[ntl, mine])
geo.save(panel, "my_indicator", scale="municipality", directory="my_data")
geo.map(panel, "ntl_mean", year=2020) # maps and analyses take your panel tooThe full guide is docs/data-infrastructure.md. It covers the scales and IDs, the metadata and aggregation rules, and recipes for building indicators. To add a dataset to the observatory, see CONTRIBUTING_DATA.md.
Three pairs of municipalities share a name across departments: San Javier (Beni, Santa Cruz), El Puente (Tarija, Santa Cruz) and Santa Rosa (Beni, Pando). In DS4Bolivia, the attribute rows of each pair (names, INE codes and the SDG Atlas 2020 indicators) are attached to the other polygon of the pair. Population, lights and embeddings, which DS4Bolivia computed from the polygons, follow the polygons; the SDG Atlas 2026 comes from SDSN Bolivia and is linked by INE code. Details and evidence: raw/README.md.
| Package | What it does | Install (Google Colab) |
|---|---|---|
| geo4bolivia (the observatory's package; repository, reference) | Find variables and their sources, licenses and actual years; load every indicator as a labelled panel with names and boundaries; compare periods; aggregate across scales with coverage; link your own data and extract indicators from rasters; and run the fundamental analyses in one call each: maps, spatial dependence, convergence, transitions, inequality, regression, panel and spatial models, multiscale GWR | !pip install -q geo4bolivia |
| expdpy | A companion for specialised panel methods: within and between variation, correlated random effects, instrumental variables, event studies; geo.for_expdpy() prepares the observatory's data for it in one line |
!pip install -q geo4bolivia expdpy |
| geometrics | A companion for specialised spatial methods, such as a spatial model under several weights matrices, plain GWR and teaching sandboxes; geo.for_geometrics() prepares the observatory's data for it in one line |
!pip install -q geo4bolivia geometrics |
geo4bolivia covers the fundamentals on its own; the two companions take over where a study needs a specialised method. Its default install runs loading, maps and the core analyses. Extras add the libraries of a few methods: "geo4bolivia[dynamics]" (Markov transitions), [mgwr] (multiscale GWR), [panel] (panel models), [raster] (zonal statistics), or [all].
| Notebook | Description |
|---|---|
| Find, load and understand the data with geo4bolivia | The package tutorial: discover datasets and their documentation, load with and without the package, describe and map, compare ~2012 with ~2024, move across scales, link your own table, and build and save an indicator from a raster |
| One question, three packages | Did municipalities converge in sustainable development, ~2012 to ~2024? geo4bolivia answers it (maps, Ξ²-convergence, inequality, spatial clusters); expdpy and geometrics each take it one specialised step further |
| Exploratory data analysis | Multidimensional poverty in 2012 and 2024, its change, census population and night-time lights: summaries, departments, treemaps and scatter plots |
| Exploratory spatial data analysis | Maps, spatial weights, Moran's I and LISA clusters of multidimensional poverty in 2012 and 2024 |
| Spatial distribution and dependence | Box-plot and Fisher-Jenks maps, KNN weights, the spatial lag, Moran's I and LISA clusters at two significance levels |
| Spatial inequality | Gini, Theil decomposition between and within departments, spatial Gini of poverty (2012, 2024) and lights |
| Spatial heterogeneity (MGWR) | Where night-time lights predict poverty better or worse: a multiscale geographically weighted regression |
| Extended EDA and ESDA | A full workflow on poverty, infrastructure and lights: statistics, correlations, clusters and bivariate maps |
| Predicting poverty from satellite embeddings | Random forest and XGBoost predict multidimensional poverty from 64 embedding dimensions, with random and spatial (leave-one-department-out) cross-validation and residual maps |
| Dynamics of sustainable development, 2012β2024 | Change in the Sustainable Development Index and its 15 goals, Ξ²-convergence, inequality, rank mobility and spatial clusters of progress |
- GeoExplorer v1.0 β Bolivia: space-time dynamics of population, night-time lights, land cover and GDP, 2013β2019 (source in
apps/)
| Project | What it does |
|---|---|
| The subnational SDG Atlas of Bolivia | SDSN Bolivia's municipal indices of the Sustainable Development Goals, ~2012 and ~2024; its 2026 edition is the sdg and sdg01β¦sdg17 datasets |
| Regional shocks and spillover effects | scspill: synthetic control when a shock spills over to neighbouring regions |
| Monitoring space-time dynamics of local pollution | satellite monitoring of local air pollution across municipalities (website coming soon) |
Research that uses the observatory's data, summarised in plain language on the website:
| Insight | Paper |
|---|---|
| Human capital constraints and the regions of Bolivia | Mendez, C., & Gonzales, E. (2021). Human capital constraints, spatial dependence, and regionalization in Bolivia: A spatial clustering approach. EconomΓa, 44(87), 115β145. https://doi.org/10.18800/economia.202101.007 |
gdo4bolivia/
βββ site/ # The website: React + Tailwind sources, build.mjs, map boundaries (data/)
βββ robots.txt, site.webmanifest
βββ project.yaml # Project identity: name, package, repository, country codes, map settings
βββ raw/ # Original inputs, never edited (see raw/README.md)
βββ builders/ # geo.py + one builder per dataset: raw/ -> geo/, data/, docs/datasets/
βββ geo/ # Spatial backbone: scales.yaml, units, boundaries/, crosswalks/
βββ data/<scale>/ # Published panels: <dataset>.csv + .parquet (generated)
βββ metadata/ # One Frictionless datapackage per dataset (source of truth)
βββ docs/ # data-infrastructure.md (the guide) + datasets/ (generated)
βββ definitions_gdo-bolivia.csv # All variables of all datasets (generated)
βββ tests/ # test_data.py (validation, builders) + test_project.py (Bolivia facts)
βββ notebooks/ # Colab notebooks (.ipynb) paired with MyST markdown (.md)
βββ apps/ # Earth Engine app source code
βββ code/ # Pointers to data-preparation code
βββ scripts/ # web_boundaries.py, indicator maps, site images, new_project.py (template)
βββ images/ # Site assets
βββ .github/workflows/ # pages.yml (builds and deploys the website)
βββ CONTRIBUTING_DATA.md # How to add a dataset, a scale or a municipality map of another vintage
βββ TEMPLATE.md # The two-repository layout and the geo4<country> template
βββ pyproject.toml, uv.lock # Environment (uv): the geo4bolivia package and the notebooks' libraries
βββ CITATION.cff, LICENSE (code), LICENSE-DATA (data), CLAUDE.md
The Python package is not in this repository: its code, tests and documentation are in quarcs-lab/geo4bolivia. This repository installs it as a dependency.
uv sync # environment: the geo4bolivia package (every extra) + the notebooks' libraries
uv run geo4bolivia build # regenerate geo/, data/, docs/ from raw/ with builders/
uv run geo4bolivia validate # check IDs, schemas, nesting and file consistency
uv run pytest # the data validate, every dataset has a builder, Bolivia facts
uv run jupyter lab notebooks/ # open the notebooks
npm ci --prefix site # website build tools (Node 20+), once
npm run --prefix site preview # build the website and serve it at http://localhost:8765/The geo4bolivia command comes from the package, which finds this checkout and runs the builders in builders/. While a new version of the package is being developed, [tool.uv.sources] in pyproject.toml installs it from a checkout next to this repository instead of PyPI.
The package is released on PyPI from its own repository (RELEASING.md). At each release this repository gets the same vX.Y.Z tag, so geo.load("ntl", ref="v0.7.0"), or a raw URL with v0.7.0 in place of main, returns the data as they were at that release.
GDO-Bolivia was created from geo4cambodia, the template of a family of geo4<country> projects. They share one design: geo IDs, long panels, Frictionless metadata and the same package API. Every project-specific setting lives in project.yaml. Since version 0.6 the package and the data live in two repositories. See TEMPLATE.md.
The site's sources are in site/: a React app (site/src/app.jsx) whose items (indicators, packages, notebooks, apps, projects, insights, events) are entries of the CONTENT_REGISTRY array in site/src/content.js. site/build.mjs bundles it with esbuild and Tailwind and pre-renders a page for every item at <section>/<id>/, with its own preview card and structured data, plus the sitemap, the search index and the data of the interactive maps. The website workflow builds and deploys it to GitHub Pages on every push to main. The maps draw SDSN Bolivia's municipal boundaries of 2025, simplified for the web by scripts/web_boundaries.py.
- Carlos Mendez (Nagoya University, Japan)
Contact: carlos.mendez.n9 [at] f.mail.nagoya-u.ac.jp
Community: questions and ideas are welcome in GitHub Discussions.
Mendez, C. (2026). GeoDevelopment Observatory of Bolivia. https://quarcs-lab.github.io/gdo4bolivia/
@misc{gdobolivia2026,
author = {Mendez, Carlos},
title = {{GeoDevelopment Observatory of Bolivia}},
year = {2026},
howpublished = {\url{https://quarcs-lab.github.io/gdo4bolivia/}},
note = {Data and notebooks: https://github.com/quarcs-lab/gdo4bolivia}
}To cite the Python package, see its README.
Please also cite the original source of each dataset you use. The citations are listed on each dataset's page in docs/datasets/.
Disclaimer. The indicators, tools, projects and insights are the responsibility of the individual researchers and do not necessarily represent the views of their affiliated institutions or related organizations. Maps are for geo-statistical use only. They have a referential, informational character and do not represent an official definition of Bolivia's territorial boundaries.
License. Code: MIT. Data: CC BY 4.0, except where an original source requires otherwise. The boundaries (geo/boundaries/) are SDSN Bolivia's municipal map of 2025, used and shared with SDSN Bolivia's permission (cite βSDSN Bolivia (2025). LΓmites Municipales Bolivia 2025 [Archivo shapefile]. https://sdsnbolivia.org/datos-espaciales/β). Each dataset's sources, their licenses and the credits they ask for are listed in its metadata, its page in docs/datasets/ and its page on the website. The indicator maps use Esri's World Dark Gray Canvas basemap (Esri, HERE, Garmin, Β© OpenStreetMap contributors). Section illustrations come from the geo4cambodia template (generated with OpenAI image generation via Codex); satellite imagery in the hero and the section backgrounds is Sentinel-2 cloudless 2016 by EOX IT Services GmbH (CC BY 4.0), except the Laguna Verde background from Sentinel-2 cloudless 2024 (CC BY-NC-SA 4.0), and NASA Blue Marble (public domain).
