evidoria

← Back to browse

Good practice Imported

HRSI Poverty Maps — Satellite-and-Machine-Learning Targeting for São Tomé and Príncipe's Safety Net

Sao Tome and Principe · São Tomé · See the Sao Tome and Principe profile · See the São Tomé profile

Evidence: Descriptive / self-reported Top 24% 73/100 · Ask Evidence Copilot about this practice

A World Bank team built 110m-resolution poverty maps for São Tomé and Príncipe by combining satellite imagery with machine learning, giving planners a low-cost, frequently updatable tool to geographically target social safety-net programmes where survey data is scarce.

110m x 110m
Map resolution
HRSI Poverty Maps — Satellite-and-Machine-Learning Targeting for São Tomé and Príncipe's Safety Net

Details

Promoter
World Bank Poverty and Equity Global Practice, in support of São Tomé and Príncipe's social protection programme
Period
2022–
Keywords
poverty mapping, machine learning, satellite imagery, social protection targeting

Context

Social protection programmes must target limited resources geographically, but household-survey poverty data is expensive to collect and quickly goes stale, a particular constraint for a small island nation with limited statistical capacity.

Activities

World Bank researchers (Fisker, Gallego-Ayala, Malmgren Hansen, Sohnesen, Murrugarra) built High-Resolution Satellite Imagery poverty maps at 110m x 110m resolution, combining a machine-learning poverty-incidence model with population-density data to rank geographic cells by estimated poverty.

Results

The methodology has since been referenced as a case study in the Asian Development Bank's guidebook on mapping poverty through data integration and AI, reflecting a wider World Bank research programme applying comparable methods elsewhere.

Conclusions

Public documentation covers methodology, resolution and design intent, but there is no published evaluation of how much targeting accuracy improved once the maps were used operationally, nor confirmed adoption figures from São Tomé and Príncipe's safety-net administration.

Implementation
Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)
Staffing & skills
World Bank Poverty and Equity Global Practice

Conditions for success

  • available satellite imagery and population-density data
  • machine-learning poverty-incidence model

Common failure modes

  • no published evaluation of operational targeting-accuracy improvement
  • no confirmed in-country institutional adoption

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful