A team of demographers and mathematicians released the most granular global migration maps ever compiled, covering 230 nations from 1990 to 2023. Published in Nature on June 10, the data show annual flows rising from 13 million in 2000 to about 35 million in 2023, including migrants and refugees. Researchers trained an AI model on UN, national statistics and Facebook inputs to generate yearly estimates.

Combining classic migration equations with a deep‑learning network, the model ingested dozens of predictors—economic output, trade links, religious affinity, conflict incidence, colonial history, and language overlap. This hybrid approach revealed that the Middle East attracted the largest inbound flow, with roughly 19 million people moving from South Asia to Gulf states between 2010 and 2023, while internal European moves topped 20 million since 1990.

Demographers say the yearly resolution uncovers short‑term spikes missed by five‑year UN or ten‑year World Bank tables, such as the 1994 exodus of nearly 950,000 Rwandans to the Democratic Republic of Congo during the civil war. All code, data and trained models are publicly released, giving policymakers immediate tools for school capacity planning, social‑benefit allocation, and labor‑market forecasting.