Redirecting to original paper in 30 seconds...
Click below to go immediately or wait for automatic redirect
📄 Abstract
Abstract: Accurate, fine-grained poverty maps remain scarce across much of the Global
South. While Demographic and Health Surveys (DHS) provide high-quality
socioeconomic data, their spatial coverage is limited and reported coordinates
are randomly displaced for privacy, further reducing their quality. We propose
a graph-based approach leveraging low-dimensional AlphaEarth satellite
embeddings to predict cluster-level wealth indices across Sub-Saharan Africa.
By modeling spatial relations between surveyed and unlabeled locations, and by
introducing a probabilistic "fuzzy label" loss to account for coordinate
displacement, we improve the generalization of wealth predictions beyond
existing surveys. Our experiments on 37 DHS datasets (2017-2023) show that
incorporating graph structure slightly improves accuracy compared to
"image-only" baselines, demonstrating the potential of compact EO embeddings
for large-scale socioeconomic mapping.
Authors (2)
Markus B. Pettersson
Adel Daoud
Submitted
November 3, 2025
Key Contributions
This paper proposes a novel graph-based approach using low-dimensional satellite embeddings and a probabilistic loss function to improve the accuracy and generalization of poverty mapping, especially in regions with limited survey data and privacy-protected coordinates. It demonstrates the effectiveness of incorporating spatial relationships and compact Earth Observation embeddings for large-scale socioeconomic prediction.
Business Value
Enables more accurate and granular poverty mapping, which can inform targeted development interventions, resource allocation, and policy-making for NGOs, governments, and international organizations.