The Cities That Drank Their Own Floors

AThe central lowlands of the Yucatan peninsula are an unlikely place to build a city. Rain falls generously there, some 1,500 millimetres a year, but it arrives in a single season and then stops for four or five months, and the limestone beneath drinks whatever reaches it. Across most of the Peten there is no permanent river, and the water table lies more than a hundred metres below the plazas. Yet by the eighth century Tikal and its neighbours were carrying tens of thousands of people through every dry season. They managed it by inverting the usual relationship between a settlement and its supply. Rather than building beside water, the Maya turned the built surface itself into a source, sealing plazas, terraces and causeways with lime plaster so that whatever they shed ran downhill into tanks quarried out of the bedrock. Airborne laser survey flown between 2015 and 2018 traced the resulting network across ground that maps had left blank, and at a scale nobody had budgeted for. Why it stopped working has divided the field ever since.

BDalia Vasquez, a hydrologist with the Central Peten Reservoir Programme in Flores, spent eleven field seasons measuring what that network could hold. Her team surveyed 63 tanks across nine centres between 2011 and 2022, reconstructing original depths from the buried plaster floors rather than from the present ground surface, which centuries of silt had raised by several metres. At Tikal the system came to roughly 900,000 cubic metres, enough on her figures to carry the city through eighteen months without rain. The gain, she stresses, came less from the tanks than from the surfaces feeding them: a plastered plaza shed 68 per cent of the rain that landed on it, against 12 per cent from the forest floor, so a catchment of a few hectares did the work of a small watershed. One tank at Tikal preserved a screen of quartz sand and imported zeolite, a volcanic mineral that strips heavy metals and pathogens from water passing through it; ceramics sealed beneath the screen place its installation before 200 CE.

CCapacity, on Vasquez's account, was never the binding constraint; upkeep was. A tank cut into karst fills with sediment as fast as its catchment erodes, and the cores her team drew record the consequence. At six of the nine centres, sediment accumulating after 750 CE arrives three to four times faster than in the preceding two hundred years, and it carries fragments of plaster with it, the catchment coming apart and washing into the store it fed. Dredging a reservoir is not skilled work, but there is a great deal of it, and it must be finished in the weeks before the rains. Where the cores show dredging continuing, with silt layers cut off and then beginning again, occupation ran on for eighty years beyond the regional median. That extension was confined to centres whose plastered catchment was renewed as well; where the plaster had gone, a dredged tank silted up again within a generation. What failed first, she argues, was the arrangement that put several thousand people on a reservoir floor every spring.

DTomas Iriarte, who reads oxygen isotopes in cave deposits at the Instituto de Paleoclima in Merida, holds that a system emptied by the sky tells you nothing about labour. A stalagmite from the Sotano de Chac, banded finely enough to resolve single years, records eight intervals between 800 and 900 CE in which rainfall ran more than 40 per cent below the long-term mean, four of them lasting three years or longer. Storage sized for one failed season, he argues, says nothing about three in succession, and a city that has drunk its reserves does not care how clean the filter was. He concedes two things. His record comes from a single cave some 300 kilometres north of Tikal, and the age model carries an uncertainty of about twenty-five years, which is coarse against the decade-by-decade abandonments it is asked to explain.

EThe dispute left the seminar room some time ago. Since 2019 a state programme in Peten has been re-lining pre-Columbian tanks to supply villages that currently have their water trucked in, and the engineering it commissions follows whichever account the ministry finds persuasive that year. Vasquez points out, awkwardly for her own case, that no labour force can dredge water that never fell, so the maintenance she treats as decisive could not have saved a city from Iriarte's worst intervals. Iriarte answers that this is precisely why drought explanations keep winning the argument and losing the budget, since a rainfall series specifies nothing a public works department can build. Both object to one habit above all: that the literature goes on reporting reservoir volume, a number a survey can produce, instead of water actually delivered to households, a number nobody has produced at all.

FWhat is emerging is a division by settlement size rather than a verdict. Small centres sat beside aguadas, natural depressions sealed with clay, which recharge from a wider catchment and never depended on plaster; in several districts these were the last places to be given up. The great capitals had traded that independence for capacity, and a manufactured catchment is only as good as the lime and the labour behind it. What nobody can yet test is whether the population figures on which the whole argument rests are sound, since they derive from house counts calibrated on a handful of excavated compounds. What both researchers reject is the framing they inherited, in which a civilisation must fail for one reason, when the evidence describes a system with two independent ways of failing and no obligation to fail by only one.