Reconstructing hunter-gatherer planet Earthusing machine-learning
Hamilton, M. J.; Walker, R. S.; Buchanan, B.; Blasi, D. E.; Bowern, C.
Show abstract
Estimating the total human population size (i.e., abundance) of the preagricultural planet is important for setting the baseline expectations for human-environment interactions if all energy and material requirements to support growth, maintenance, and well-being were foraged from local environments. However, demographic parameters and biogeographic distributions do not preserve directly in the archaeological record. Rather than attempting to estimate human abundance at some specific time in the past, a principled approach to making inferences at this scale is to ask what the human demography and biogeography of a hypothetical planet Earth would look like if populated by ethnographic hunter-gatherer societies. Given ethnographic hunter-gatherer societies likely include the largest, densest, and most complex foraging societies to have existed, we suggest population inferences drawn from this sample provide an upper bound to demographic estimates in prehistory. Our goal in this paper is to produce principled estimates of hunter-gatherer abundance, diversity, and biogeography. To do this we trained an extreme gradient boosting algorithm (XGBoost) to learn ethnographic hunter-gatherer population densities from a large matrix of climatic, environmental, and geographic data. We used the predictions generated by this model to reconstruct the hunter-gatherer biogeography of the rest of the planet. We find the human abundance of this world to be 6.1{+/-}2 million with an ethnolinguistic diversity of 8,330{+/-}2,770 populations, most of whom would have lived near coasts and in the tropics. Significance StatementUnderstanding the abundance of humans on planet Earth prior to the development of agriculture and the industrialized world is essential to understanding human population growth. However, the problem is that these features of human populations in the past are unknown and so must be estimated from data. We developed a machine learning approach that uses ethnographic and environmental data to reconstruct the demography and biogeography of planet Earth if populated by hunter-gatherers. Such a world would house about 6 million people divided into about 8,330 populations with a particular concentration in the tropics and along coasts.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Local increases in admixture with hunter-gatherers followed the initial expansion of Neolithic farmers across continental Europe 95%
- Impact of Holocene environmental change on the evolutionary ecology of an Arctic top predator 94%
- Anthropogenic climate change will likely outpace coral range expansion 94%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.