GeoDevelopment Observatory of Bolivia

Name: Satellite embeddings
License: https://creativecommons.org/licenses/by/4.0/

GDO-Bolivia is a public-access platform of open subnational indicators, notebooks, web apps, and research insights for monitoring sustainable regional development in Bolivia with satellite data and AI.

Indicators

SDG 1: No poverty: The SDG 1 (No poverty) index of the Municipal Atlas of the SDGs and its four indicators, multidimensional poverty, unsatisfied basic needs, extreme energy poverty and access to basic services, for every municipality.
SDG 2: Zero hunger: The Zero hunger index of the Municipal Atlas of the SDGs and its four indicators of child malnutrition, obesity in women, farm size and tractors, for every municipality.
SDG 3: Good health and well-being: The SDG 3 index of the Municipal Atlas of the SDGs and its nine indicators of births, child mortality, infectious diseases and adolescent fertility.
SDG 4: Quality education: The Municipal Atlas’s SDG 4 index of quality education and its six indicators of school dropout, adult education, literacy and qualified teachers, each stored at the year it was measured.
SDG 5: Gender equality: The SDG 5 gender equality index of the Municipal Atlas of the SDGs and its four gender parity indicators in schooling, poverty and work, for every municipality.
SDG 6: Clean water and sanitation: The SDG 6 index of the Municipal Atlas of the SDGs and its three indicators of drinking water, sanitation and wastewater treatment in 2017, for every Bolivian municipality.
SDG 7: Affordable and clean energy: The SDG 7 clean energy index and its four indicators of electricity coverage, residential electricity use, clean cooking energy and energy CO2 emissions per person.
SDG 8: Decent work and economic growth: The SDG 8 index of decent work and economic growth for the 339 municipalities, with its seven indicators of labour participation, youth inactivity, idle electricity meters, bank branches and migration intermediation.
SDG 9: Industry, innovation and infrastructure: The SDG 9 index of the Municipal Atlas of the SDGs and its five indicators of road and rail links, phone access, mobile antennas and school technology, for every municipality.
SDG 10: Reduced inequalities: The Atlas’s SDG 10 index for every municipality, built from inequality in schooling, inequality in household electricity use and the share of people who do not speak Spanish.
SDG 11: Sustainable cities and communities: The SDG 11 index of the Municipal Atlas of the SDGs for every municipality, with its three indicators of housing, sanitation and public transport.
SDG 13: Climate action: The SDG 13 (Climate action) index of the Municipal Atlas of the SDGs 2020 and its three indicators, climate change vulnerability, CO₂ emissions per person and deforestation, for every municipality.
SDG 15: Life on land: The Life on land index of the Municipal Atlas of the SDGs for 339 municipalities, built from the share of protected area and biodiversity loss from deforestation.
SDG 16: Peace, justice and strong institutions: The SDG 16 index of the 339 municipalities, built from homicide rates, municipal budget execution and birth registration of children under 5.
SDG 17: Partnerships for the goals: The SDG 17 index of the Municipal Atlas of the SDGs and its two indicators of municipal finance, local tax revenue and public investment per capita, in 2017.
Night-time lights per capita: Nine years of VIIRS night-time lights per person for every municipality, with a long-run trend: a satellite proxy for local economic activity.
Population: Two decades of gridded population estimates and population density for every municipality: the denominator for per-capita indicators and population-weighted averages.
Satellite embeddings: What every municipality looks like from space, compressed by a deep-learning model into 64 numbers: features for machine learning.

Tools

geo4bolivia: load, link and build subnational data: The observatory’s Python package. Load every indicator as a tidy panel with boundaries, link your own data to the same units, and build new indicators across municipalities, provinces and departments.
geometrics: regional growth, convergence and inequality: A companion package for spatial analysis: maps, spatial dependence, β-, σ- and club convergence, spatial regressions, distribution dynamics and inequality, each with a plain-language reading.
expdpy: explore and model panel data: A companion package for panel data: within/between variation, trajectories and transitions, then pooled, fixed and random effects, Hausman tests and clustered standard errors, each with a plain-language reading.
Load, link and build Bolivian indicators with geo4bolivia: A hands-on tour of the observatory’s data: load it with and without the package, move it across scales, link your own table by name or INE code, and build an indicator from a raster.
Exploratory data analysis of multidimensional poverty: How poverty varies across 339 municipalities and 9 departments, and how it relates to population and night-time lights.
Exploratory spatial data analysis of poverty: Is a municipality as poor as its neighbours? Spatial weights, Moran’s I and LISA clusters of multidimensional poverty.
Spatial distribution and dependence of poverty: Two ways to classify a poverty map, then spatial weights, the spatial lag, Moran’s I and local clusters.
Spatial inequality of poverty and night-time lights: Gini, Theil between and within departments, and the spatial Gini: how much of Bolivia’s inequality is regional?
Where do lights predict poverty? GWR and MGWR: Geographically weighted regressions let the link between night-time lights and poverty vary across Bolivia.
Poverty in Bolivia: an extended EDA and ESDA: One workflow from descriptive statistics to infrastructure correlates, spatial clusters, bivariate LISA, inequality and GWR.
Predicting poverty from satellite embeddings: Random forest and XGBoost predict multidimensional poverty from 64 satellite embedding dimensions, with random and spatial cross-validation.
GeoExplorer v1.0: Bolivia from outer space: Swipe between 2013 and 2019 to see how population, night-time lights, land cover and GDP changed across Bolivia’s municipalities.

Insights

Human capital constraints and the regions of Bolivia: Where malnutrition, language barriers, school dropout and unequal education cluster across 339 municipalities, and how they divide Bolivia into seven or eight regions.

Data, code and documentation: GitHub repository.