Session
Session 1: Big Data From Small Satellites 1
Abstract
Understanding of economic, social, and cultural characteristics of a society is critical to effective government policy and successful commercial undertakings. Obtaining this information, however, often requires direct observations and interactions with the local populace, entailing significant costs and potentially exposing data collectors to heightened risks. We address this challenge by combining automated processing of satellite imagery with advanced modeling techniques. We have developed methods for inferring measures of well-being, governance, and related socio-cultural attributes from satellite imagery. This research represents a fundamental innovation in the study of human geography by explicitly analyzing the relationship between the observable physical attributes and the societal characteristics and institutions of the region.
Through analysis of commercial satellite imagery and coincident survey data, we have developed and tested models for rural Afghanistan and selected countries in sub-Saharan Africa (Botswana, Kenya, Zimbabwe). The findings show the potential for predicting peoples’ attitudes about the economy, security, leadership, social involvement, and related questions, based only on the imagery-derived information. When tested on sequestered data in Afghanistan, the image-based model predict with 79% accuracy whether villagers will volunteer their time to support a community project (an indicator of social capital), and with 78% accuracy whether the village will look to the government or local resources for protection. These models also predict the likelihood of a respondent supporting the village council, growing opium poppies, or going on Hajj. Similar models for the African region also provide useful indicators. Models for predicting economic attributes (presence of key infrastructure, attitudes about the economy, perceptions of crime, and outlook towards the future) all exhibit statistically significant performance.
These methods show significant promise for assessing key social indicators. However, the models only capture a snapshot in time. With the emergence of new small imaging satellites, the potential for temporal analysis offers a substantial improvement over previous work. By monitoring changes in physical structures and patterns of commercial activities, a far richer understanding of societies will be possible through image-based methods. Using a recent airborne imagery collection campaign, which provides a surrogate for the anticipated frequent coverage achievable with small satellites, we demonstrate a proof-of-concept analysis of traffic patterns and temporal analysis to understand local economic activities. The paper concludes with recommendations for future exploration.
Presentation
Dynamic Socio-Economic Analysis: The Power of Persistence from Small Satellites
Understanding of economic, social, and cultural characteristics of a society is critical to effective government policy and successful commercial undertakings. Obtaining this information, however, often requires direct observations and interactions with the local populace, entailing significant costs and potentially exposing data collectors to heightened risks. We address this challenge by combining automated processing of satellite imagery with advanced modeling techniques. We have developed methods for inferring measures of well-being, governance, and related socio-cultural attributes from satellite imagery. This research represents a fundamental innovation in the study of human geography by explicitly analyzing the relationship between the observable physical attributes and the societal characteristics and institutions of the region.
Through analysis of commercial satellite imagery and coincident survey data, we have developed and tested models for rural Afghanistan and selected countries in sub-Saharan Africa (Botswana, Kenya, Zimbabwe). The findings show the potential for predicting peoples’ attitudes about the economy, security, leadership, social involvement, and related questions, based only on the imagery-derived information. When tested on sequestered data in Afghanistan, the image-based model predict with 79% accuracy whether villagers will volunteer their time to support a community project (an indicator of social capital), and with 78% accuracy whether the village will look to the government or local resources for protection. These models also predict the likelihood of a respondent supporting the village council, growing opium poppies, or going on Hajj. Similar models for the African region also provide useful indicators. Models for predicting economic attributes (presence of key infrastructure, attitudes about the economy, perceptions of crime, and outlook towards the future) all exhibit statistically significant performance.
These methods show significant promise for assessing key social indicators. However, the models only capture a snapshot in time. With the emergence of new small imaging satellites, the potential for temporal analysis offers a substantial improvement over previous work. By monitoring changes in physical structures and patterns of commercial activities, a far richer understanding of societies will be possible through image-based methods. Using a recent airborne imagery collection campaign, which provides a surrogate for the anticipated frequent coverage achievable with small satellites, we demonstrate a proof-of-concept analysis of traffic patterns and temporal analysis to understand local economic activities. The paper concludes with recommendations for future exploration.
