What can we know about North Korea’s development from machine learning and satellite imagery (2016 - 2019)?

Keywords: satellite imagery, machine learning, poverty mapping, North Korea
Session Type: Virtual Paper Abstract
Day: Saturday
Session Start / End Time: 2/26/2022 11:20 AM (Eastern Time (US & Canada)) - 2/26/2022 12:40 PM (Eastern Time (US & Canada))
Room: Virtual 68

Abstract

North Korea has long been a black box with few public data for outsiders to assess its economic development. Little is known about changing conditions since a series of economic sanctions that went into effect in 2017. The data deficiency makes it difficult to apply existing traditional methods and recent remote sensing technologies for predicting the country’s economy. Here we present our machine learning model which predicts grid-level economic development by using publicly available satellite imagery. When applied to 10m resolution Sentinel-2 satellite images for the period from 2016 to 2019, our model produces the first-ever predictions of North Korea’s economic development. Using our measure as a proxy of economic development indicates that amid rising pressure from economic sanctions, the centrally planned economy has been directing more resources towards its capital, regions with state-led development projects, and areas with uranium mines. We also tested the applicability of the algorithm on other Asian developing countries. Our method can help monitor development in hard-to-visit and low-resource regions with high granularity thereby guiding sustainable development programs.