Using satellite data and machine learning to study conflict-induced environmental and socioeconomic destruction in data-poor conflict areas: The case of the Rakhine conflict

This paper studies socioeconomic and environmental changes in the neighboring areas Bangladesh-Myanmar border from 2012 to 2019, thus covering the period before and after the 2017 Rakhineconflict in Myanmar and outflux of refugees across the border to Bangladesh. Given the scarcity andcostliness of traditional data collection methods in such conflict areas, the paper uses a novelmethodological model based on very-high-resolution satellite imagery, nighttime satellite imagery,and machine-learning algorithms to generate reliable and reusable data for comparative assessment ofthe impacts of the Rakhine conflict. Assessments of welfare and environmental risks using thisapproach can be accurate and scalable across different regions and times when other data areunavailable. Keyfindings are: the general livelihood situation has worsened and income sourcesshrunk in Rakhine; forced migration damaged the ecologically fragile regions in the two countries; thedestruction of aquaculture wetland ecosystems is observed in Rakhine; the deforestation rate reached20% in Rakhine and 13% on the Bangladeshi side of the border. The results can provide guidance topolicymakers and international actors as they work to repatriate the victims of the conflict in Rakhineand minimize the conflict’s security and environmental consequences. The methodology can beapplied to other data-poor conflict and refugee areas in the world.