Spatiotemporal poverty mapping using deep learning and statistical models

See also: https://eo-cdt.org/projects/spatiotemporal-poverty-mapping-using-deep-learning-and-statistical-models/

Scientific background and motivation

Reliable techniques to accurately predict socioeconomic indicators such as poverty at national, regional, and global scales are critical factors for the current and future prosperity of people and the planet. The United Nations in 2015 defined the Sustainable Development Goals (SGDs), a set of 17 actions based on socioeconomic indicators to improve health and education, reduce inequality, and spur economic growth. However, the scarce availability of accurate and up-to-date well-being/poverty survey data remains a significant challenge for SDG tracking of socioeconomic indicators. An alternative is to use Earth Observation (EO) satellite data sources, which contain fine-grained information on most of the planet over frequent periods and have clear comparability on a global scale, but show great variations for different regions and geographic settings, e.g., rural areas are usually under-represented. An example of the contribution that geospatial data can make for geographic targeting of resources can be found in Smythe and Blumenstock (2022), where poverty maps derived from EO data and other data products such as road networks, elevation and mobile phone usage were used for assisting in geographic targeting of anti-poverty payments in Nigeria.

Previous studies based on EO data are usually based on deep-learning models, which provide insights into global poverty progress and assess comparisons between countries. However, it is at the sub-national level where decisions and actions need to be taken if the SDGs are to be achieved, and generalized global models based solely on EO data do not provide the level of detail that practitioners and decision-makers need for high-resolution geographic targeting of resources. The local studies on the other hand are restricted to a few locations and require a lot of user input, specific information on social and ecological systems. Therefore, a method to develop actionable policy at a sub-national level that offers intuitive outputs of SDG indicators and can be used by decision-makers for geographic targeting of resources is still lacking.

Aims and objectives

 This project will build spatiotemporal maps of poverty on the sub-national scale to support the implementation of the SDGs and enable evidence-based decision-making by combining EO data with local fine-resolution assessments. This task will involve understanding associations between EO metrics and socioeconomic conditions as well as the relationships between poverty and geospatial proxies in different countries, counties, and wards. Moreover, the high temporal resolution of the EO data will be used to track changes in SDGs metrics and identify spatial locations with unusual changes in patterns in the signal so that new surveys targeting those regions can be commissioned.

Methodology 

Deep learning techniques, such as Convolutional Neural Networks (CNNs), are increasingly used for predictive analytics with remote sensing images and tasks such as ground object detection, population, land mapping, etc. This project will investigate deep learning techniques to fill spatial gaps in earth observation-based (EO) products. However, a drawback of using solely deep learning models to derive data-driven policy and geographic targeting across time and space is their lack of interpretability. Indeed, these models are well known to be black boxes, making the results not easily explained, justified or intuitive, therefore reducing their practicability for policy-making purposes. Statistical models, on the other hand, are designed such that the parameters reflect the relationship between different features of the data and therefore are interpretable and transferable. Although this level of interpretability is not possible in a black-box deep learning model, they are remarkably accurate for prediction purposes. To address this dichotomy, this project will develop a novel workflow that accurately reproduces SGD indicators while retaining the interpretability of statistical models.

Previous studies established relationships between household poverty from household survey data and geospatial data for the surrounding area, but household data is available only partially for a specific ward. The assumption of homogeneity between wards is not valid in general, making the transferability an issue for wards with large variations in socioecological systems. Geostatistical models based on Gaussian processes will be investigated to address the problem of transferability and ultimately predict poverty even at locations where no data is available by borrowing information from neighboring regions. The approach will incorporate multiple EO satellite data and local fine-resolution assessments via spatiotemporal modeling. It is likely that the study will have a focus in East Africa.

SENSE CDT

This PhD is part of the NERC and UK Space Agency funded Centre for Doctoral Training "SENSE": the Centre for Satellite Data in Environmental Science. SENSE will train 50 PhD students to tackle cross-disciplinary environmental problems by applying the latest data science techniques to satellite data. All our students will receive extensive training on satellite data and AI/Machine Learning, as well as attending a field course on drones, and residential courses hosted by ESA (Rome). All students will experience extensive training on professional skills, including spending 3 months on an industry placement. See http://www.eo-cdt.org