Populations Trends Across the US Southwest

Kenneth "Blake" Vernon

Explore humanity's evolving presence in the US Southwest from 750 to 1600 CE. Hovering over a shape will give you more details about that time and place.

1. Population estimates are aggregated across HUC10 watersheds and 25-year intervals. Values are based on estimates of the number of occupied rooms per time interval. For visualization purposes, values are log-transformed (zero counts are calculated as log(1e-5)).

How was this map made?

This map shows population estimates aggregated across HUC10 watersheds watersheds (a level of definition of watersheds used by the US Geological Survey) and 25-year intervals. To arrive at those estimates, we counted up the rooms in each residential site in each watershed, under the assumption that more rooms means more people. This is a common strategy in archaeology, and it is also the strategy used by demographers today to estimate population in areas where governments lack the resources to conduct a census (for example, in the European Union's Global Human Settlement Layer).

Unfortunately, it is not reasonable to infer population size directly from room counts. That would be like arguing that there must be 100 people staying at a hotel because the hotel has 100 rooms. In reality, some number of rooms will typically be vacant on a given day. The situation gets more complicated if we imagine the hotel adding or removing rooms over time, which is what happened at residential sites in the US Southwest during the time period being modeled.

To account for vacancy and changes in the sizes of settlements over time, we simply assumed that each room represents one person and that persons were evenly (or uniformly) distributed over the years each site was occupied, based on the associated pottery. For instance, if people lived at a residential site for 100 years, that's four 25-year intervals, and there are four rooms at that residence, then it is assumed that one person lived at the site in each 25-year interval. This was done for each residential site in the cyberSW database, then those estimates were added up for each watershed and time interval.

There's a lot of variation in these estimates, with some areas having huge populations, and others almost no human presence at all. It's hard to visualize spatial and temporal trends with information like this, so the map actually displays the logarithm of occupied rooms (zero counts are listed as log(1e-5)).