How fast did agriculture spread?

The title question is not about how quickly agricultural skills and knowledge pass between peoples, though that is an intensely interesting subject for archaeologists. It is about the rate of spread from the originating centres which I discussed in my post of 7th November 2014. Modelling the rate of spread depends on understanding the different nature and effects of intensification and extensification.

Intensification is about productivity. It is about getting more yield from a field by tilling, fertilising, irrigating and weeding it. Intensification replaces the nutrients and minerals taken up by the crops harvested in previous years so that fallow periods, during which those nutrients and minerals would be naturally replenished, can be eliminated so that more crops can be planted. The seminal work on these processes is The Conditions of Agricultural Growth: The Economics of Agrarian Change under Population Pressure by Ester Boserup, a Danish economist, though that work has since been challenged and reappraised by many academics. For example, Kathleen Morrison (1994) shows that intensification is not only a response to population increase but may be  driven by trade or food preferences. Agricultural intensification is the reason proposed by Ruddiman (2003, p.264-5) for the anomalous rise in atmospheric methane.

Extensification is about production. It is about getting more yield by expanding the area which may be farmed by clearing forests, creating space for new fields. When those fields lose their fertility, further clearances make space for more fields. Agricultural extensification is the reason proposed by Ruddiman (2003, p.272-3) for the anomalous rise in atmospheric carbon dioxide.

The problem with modelling the different rates of spread of these two processes is that they are not mutually exclusive. Extensification is constrained by the practical limitations of travel to and from fields while some forms of intensification require less labour than does forest clearance. Prehistoric people would have made the choice between intensification, extensification or a combination of the two based on local environmental conditions and local competition for resources.

In their paper Used planet: a global history, Ellis et al. (2013) considered methods for quantifying land use for the period before the availability of historical and instrumental records. Existing attempts typically created models of land use for the historic period and extrapolated them backwards using algorithms based on contemporary land management practices. The two alternatives reviewed used population as a proxy measure for land use.

The HYDE 3.1 model (Klein Goldewijk et al, 2010)  assigned areas of cropland and pasture to a grid of population densities, estimated from a range of sources, for the period 1700-2000. The model runs from 10,000 BCE to date. The initial human population used was 2 million and assumptions are made about regional variations in population growth related to the waxing and waning of empires. A number of uncertainties in the assumptions and results were acknowledged. The prehistoric climate and biome models were derived from those of today, the population start value is subject to great uncertainty and the per capita land-use (assumed to vary little over 12,000 years) may vary greatly depending on the degree of intensification.

The KK10 model (Kaplan et al, 2011) assumed that populations would expand to use all available land before turning to intensification and, importantly, compared a Boserupian scenario, in which per capita land-use reduces over time, with a fixed per capita land-use scenario. The model runs from 6,000 BCE to date. Population estimates were from the same sources used in the HYDE 3.1 study. The model also incorporated climate, soil and CO2 data into a Dynamic Global Vegetation Model.

Results from the HYDE 3.1 model compared with those from the KK10 model (Ellis et al., 2013, p.7980)

The two different models offer contrasting pictures of the potential impact of human agricultural expansion on the environment. The KK10 model clearly impacts earlier and more deeply on both forested areas and savannahs across most regions of the world, consistent with the Early Anthropogenic Hypothesis for the origin of increased greenhouse gases. Neither the researchers who constructed each of the two models, nor the authors of the comparative review, claim that one or other model is correct, indeed they both highlight the uncertainties and weaknesses in the models. What is important is that, with increased empirical evidence and refined models, scientists are improving the understanding of how pre-industrial anthropogenic activity may have contributed to climate change during the Holocene.