The area that was actually measured

The survey measured a total of about 4,500 km² by airborne laser scanning, mapping 1 km wide strips every 10 km along straight transects roughly 450 km long, and documented more than 400 well-preserved monumental ceremonial sites. More than half of the scanned area was systematically assessed.[1]

From that count, more than 20,000 ceremonial centres are projected across the whole region. The area the projection covers is about 183,000 km², more than forty times the area scanned, and those 183,000 km² themselves amount to less than three percent of Greater Amazonia. When a 1 km wide strip is scanned every 10 km, one tenth of the corridor is seen. So the number reaches beyond ten times what was seen, out to more than forty times it.[1]

From count to population

The second extension is on the population side. The authors give a range of 1.25 to 3 million for the start of the first millennium and call that range cautious; set against the 1 to 10 million proposed earlier for the whole Amazon, the claim to caution becomes intelligible. Moving from a count of structures to a count of people requires an assumption about how many people each structure served, and that assumption does not come from the scanning.[1]

The strength of the extension depends on whether the transects sample earthwork density without bias. If earthworks cluster along watercourses or on flood-free high ground, a transect cutting across those features and one running parallel to them read entirely different densities, and the orientation of the transects enters the result. Alternatively the structures may genuinely be distributed evenly across the region, in which case the figure of 20,000 stands. But separating the two cases takes a second survey.[1]

What would settle it

What would move the argument forward is a second sampling design rather than a larger number: transects flown at a different spacing or in a different orientation over the same 183,000 km². If two separate geometries give densities close to one another, the extension gains support; if they diverge, the difference comes from where the transects ran rather than from the structures. The finding itself is solid: more than 400 monumental sites were found, and they exist. What needs testing is the step that runs from those four hundred to twenty thousand.[1]