Poverty is one of the chronic problems of the XXI century and, despite the recent decrease of global economic inequalities between and within countries, in 2016 about 800 million people still lived in extreme poverty condition, with many of them located in sub-Saharan Africa and Southern Asia. In this context, poverty alleviation programmes generally rely on data about local economic livelihood for identifying places with highest need for aid. Nevertheless, this information traditionally comes from patchy and logistically challenging household surveys which normally happen to be extremely expensive. As a result, policymakers and public sector stakeholders lack key data necessary for targeting anti-poverty programs or properly measuring their effectiveness. Given the challenges of scaling up traditional data collection efforts, in the past few years alternative strategies have been proposed for assessing the degree of poverty based on satellite data.
The main objective of EO4Poverty is to implement a novel system based on advanced machine- and deep-learning techniques for generating national spatial poverty maps by jointly exploiting EO-based products (in particular derived from Copernicus Sentinel data) and non-EO based products (e.g., roads and transportation networks, social media) coupled with in-situ reference information gathered from publicly available household surveys. The project aims to improve existing approaches and to provide an easily transferable service for creating maps of actual support to the end-users.