There are a lot of different ways to get at credible estimates of a population and sub-population size, especially in a war zone.

A census is not really a good one for strategic/tactical purposes. Example: The US census is done every ten years, with an annual and mid-point revised estimate.

Usually, the census preliminary figures follow the census by about six months, with the final "agreed" population data out in about three years. Those are the figures used for political reapportionments and per head fund allocations. They result from lots of arm wrestling about how many homeless people and categories of undercounts. If somebody is using decennial census data for, or relying on it for future use for strategic and tactical decisions, they are uninformed, and probably wrong.

Population figures, especially at the district, subdistrict level are real time dynamic figures, like political boundaries. They need to incorporate real-time refugee movements, temporary displacements, and any number of factors.

Like political boundaries in a conflict zone, they are highly-dynamic, and, in part, evidence of conflict patterns.

To break that down even further to subterranean folks (bad guys) requires real time analysis---aggregation of field data---in a way that current battlefield intel does not routinely provide.

I was, at first, frustrated by military folks having little faith in data processes and planning systems, but I always remind myself that there are two different things: (1) high quality information created and managed by experienced professionals, and (2) the GIGO that floats out to the field.

They are getting old, and often out-dated information instead of real-time dynamic stuff that they need.

Every time a group moves from a village to a tent outside of Kabul, they register as refugees to get food (and the tent). Refugee folks know exactly where people come from, and why they left. Mil needs a real-time civilian interface to that data.

From experience, if you want to know how many people are in a town, ask an SF about the town he watches, or use a thermal scan. Or, count the number of occupied dwelling units in a village (that's air photo count minus destroyed/uninhabitable), and randomly sample the population per household in that village (How many families, people? Most of this is routinely done anyway but not systematized).

Want to get down to bad guys. My understanding is, sometimes, that is a count best done at night. Who is out wandering around?

Soldiers, in their normal course of business, are doing census work---it is just not being compiled. How many people live in that settlement? How many high-top sneakers per male population? Is that recorded settlement formally occupied, and by who?

Again, this stuff needs to be routinely updated (by standing FRAGO if necessary) if it is to be accurate and meaningful.

As much as anything, it is the changes in patterns that tell the stories and contain predictive and measurable capabilities, but, for that, you have to have a base line.

There is just no system for boring and mundane stuff like this. Yet....

Steve