The Maya Collapse: The Cities That Went Silent, and the People Who Did Not Vanish

At its height, roughly 600 to 800 CE, the Classic Maya civilization spanned hundreds of polities and raised monumental architecture of extraordinary sophistication at Tikal, Calakmul, Copan, Palenque, Caracol and Naranjo. It developed the most complete writing system in the pre-Columbian Americas, the Maya hieroglyphs; an elaborate Long Count calendar; and sophisticated knowledge of astronomy, mathematics including the concept of zero, and agriculture.

The collapse was not an event. It was a gradual, diachronous process running roughly 150 to 200 years, from about 800 to about 1000 CE, and our own research file's key finding calls it a rolling wave of abandonment. Any telling that puts it in a year, or in a single failed harvest, is telling something the file itself refuses in its opening summary.

What archaeologists actually count, when they say a Maya city collapsed, is four things: the cessation of dated stelae, meaning the last Long Count dates carved in stone; the end of elite construction; the abandonment of palace and ceremonial compounds; and dramatic population decline. Notice that three of the four are measures of what the ruling order stopped doing. That distinction runs through this whole article.

The wave opens in the west. Palenque's last recorded ruler is placed at about 799 CE, and the city was abandoned by roughly 850 CE. One caution on that first number, since this article would rather be exact than tidy: our own file says the last recorded ruler died in 799, while the wider literature more usually records 799 as an accession date for Palenque's last known ruler. This pass did not resolve which, so the year is carried here as attached to that ruler generally, not to a specific event.

Then the sequence tightens. Quirigua's last dated monument is 810 CE. Naranjo's last stela is 820 CE. Within a single decade, two established capitals stop recording themselves.

Copan's last carved monument is dated 822 CE, and it is the most eloquent object in this entire file, because the carvers did not finish it. It is Altar L, an altar rather than a stela. One face carries the enthronement scene of the ruler Ukit Took', a second face was begun, and two faces were left entirely blank. Ukit Took' acceded on 6 February 822 and, on the surviving evidence, was never fully installed as king. A correction to our own file belongs here in the open: E_5_04 calls this monument a stela in two places. It is an altar, and this article uses the externally verified description.

Estimates of peak Classic Maya population in the southern lowlands range from 3 to 10 million, with some estimates as high as 15 million depending on the geographic scope. Our own file states this range in its body and separately states 10 to 15 million in its summary, without declaring the wider scope that would justify the higher figure. Those two statements do not sit together, and this article carries the body range as the sourced one.

There is a published number that grounds the high end, and our file does not cite it. In 2018 Canuto, Estrada-Belli and colleagues published in Science the results of a consortium of 18 scholars, enabled by Fundacion PACUNAM, analysing airborne laser scanning data across more than 2,100 square kilometers of the Maya Biosphere Reserve in northern Guatemala. It was the largest single lidar survey in the history of Mesoamerican archaeology. The team identified 61,480 ancient structures, derived an average population density of about 120 people per square kilometer, and estimated 7 to 11 million people at the height of the Late Classic period, 650 to 800 CE.

Whatever the peak was, the fall from it was severe. Populations in the southern lowlands declined by an estimated 90 to 95 percent between 800 and 1000 CE, on the evidence of lost settlement density in archaeological surveys, reduced ceramic production, and the absence of monumental construction. Read that number for exactly what it is: a regional depopulation figure for the southern lowlands. It is not a figure about Maya people, and the next section is where it gets its context rather than its footnote.

Some areas of the southern lowlands were essentially depopulated and never significantly reoccupied before the Spanish arrival. That is the part of the story the jungle-swallowed-city photographs actually earn, and it applies to particular districts of one region, not to the Maya world.

The Maya civilization did not completely disappear. While the southern cities were being abandoned, the northern Yucatan flourished: the Puuc region, meaning Uxmal, Kabah and Labna, and above all Chichen Itza. The same centuries that read as an ending in the Peten read as an ascent in the north.

Chichen Itza was a major power from roughly 900 to 1100 CE, and Mayapan followed it from roughly 1200 to 1450 CE. Set those dates against the table above. Chichen Itza's rise begins in the same generation as Tonina's last inscription.

The north's exemption was real but partial and temporary, and a 2016 study says so precisely. Hoggarth, Kennett and colleagues, publishing in Global and Planetary Change, combined Bayesian radiocarbon models, calendar dates on carved monuments and climate proxies, and identified three major intervals of drought between 820 and 1100 CE and two separate waves of political collapse: the asynchronous collapse of multiple southern polities between about 850 and 925 CE, and then the political collapse of Chichen Itza itself between about 1000 and 1050 CE, coinciding with the longest and most severe drought in the regional record, 1000 to 1100 CE. Political developments through the Postclassic, 1000 to 1517 CE, correspond with a return to higher rainfall. The northern Maya did not escape drought. They met a different drought on a different timetable.

At the time of Spanish contact, a period running from 1524 to 1697 CE, millions of Maya-speaking people inhabited Yucatan, Guatemala and adjacent regions. That end date is worth its own sentence. The Itza Maya at Nojpeten, on Lake Peten Itza, did not fall to the Spanish until 1697: more than eight centuries after Tikal raised its last stela, an independent Maya kingdom was still holding out, in the same lowlands.

So, flatly: no, the Maya did not disappear. Our own file files that claim in Tier 4 as debunked, and puts a living-population figure beside it. Approximately 6 to 7 million Maya-speaking people live in Mexico, Guatemala, Belize, Honduras and El Salvador today, and that is the floor rather than the ceiling: the Smithsonian's National Museum of the American Indian states that more than seven million Maya live in their original homelands in Mesoamerica and in countries around the world. Britannica gives roughly 30 Mayan languages spoken by more than five million people in the early 21st century, most of those speakers bilingual in Spanish, and in Guatemala alone figures in the 6.5 million range are reported across the country's 22 Maya peoples. The totals differ because they count different things, people or language speakers, so take the range and note which is which. The Classic Maya collapse was a regional phenomenon confined to the southern lowlands. It was the end of a particular political and cultural system, not the extinction of a people.

The first record came out of the sea. Gerald Haug and colleagues, publishing in Science in 2003, measured titanium concentrations in a marine sediment core from the Cariaco Basin off Venezuela, where titanium serves as a proxy for terrestrial runoff and therefore for rainfall on land. The record shows three intense multi-year droughts centred at approximately 810, 860 and 910 CE, each lasting several years, with the wet periods between them briefer than the climatological mean. Compare those three dates with the last-monument table and the correlation looks uncanny.

Which is exactly why the brake belongs immediately after, and our own file supplies it in its counter-arguments. The exact correlation between specific drought events in paleoclimate records and specific polity collapse dates from epigraphy has been questioned. Sediment core dating often carries uncertainties of decades, which makes precise cause-and-effect attribution difficult. A correlation that looks tight to the year may be a correlation between a carved date good to the day and a sediment date good to a human lifetime.

The second record came out of a cave, and it says more than it is usually quoted as saying. Douglas Kennett and colleagues, publishing in Science in 2012, ran oxygen isotope analysis on a precisely dated speleothem from Yok Balum Cave in Belize. The record describes a sequence, not a single peak: anomalously high rainfall between 440 and 660 CE, which favoured unprecedented population expansion and the proliferation of political centres; a drying trend from 660 to 1000 CE, which the authors tie to the balkanization of polities, increased warfare and the asynchronous disintegration of polities, with multidecadal droughts particularly between 820 and 870 CE and roughly a 40 percent reduction in summer rainfall across the Terminal Classic; and then population collapse in the context of an extended drought from 1020 to 1100 CE. That last window sits later than the 800 to 1000 CE range section 03 gives for the regional population fall, and the two are not measuring the same thing: section 03's figure is a survey-based estimate of how much of the southern lowland population was lost across those two centuries, while this one is a single cave record placing population collapse against its own rainfall sequence. Both are carried here as their sources state them. Note what that first interval means: on this record, climate helped build the Classic Maya world before it helped unmake it. One correction to our own file, because it compresses this paper: the superlative driest period in the past 2,000 years belongs to the 1020 to 1100 drought, after the Terminal Classic, and should not be attached to the collapse window.

The third record put a number on it, and it is the best-sourced figure in this article. Nicholas Evans and colleagues, publishing in Science in 2018, applied gypsum isotope hydrology to sediment cores from Lake Chichancanab in Yucatan, analysing the water structurally bound inside sedimentary gypsum that was laid down under drought conditions. They estimated that annual precipitation fell by 41 to 54 percent during the most severe Terminal Classic droughts, with intervals of up to 70 percent reduction during the driest episodes. The same study found relative humidity down by 2 to 7 percent against present-day levels. This is what a halved rainfall looks like when somebody finally measures it instead of inferring it.

The fourth record is new, and it is the one that makes the drought human-scaled. James and colleagues, publishing in Science Advances in August 2025, ran oxygen-isotope analysis on a stalagmite from a cave in northwest Yucatan and resolved rainfall for individual wet and dry seasons between 871 and 1021 CE, the first time seasonal resolution has been achieved for the Terminal Classic. The record contains droughts lasting from 1 to 13 years, including one that ran the full 13. And its own headline finding points away from a single story: political activity at major northern sites including Chichen Itza and Uxmal declined at different times relative to those droughts, implying differential cultural responses to the same climate stress.

Drought alone does not explain the collapse. The contemporary northern Yucatecan Maya, at Uxmal and Chichen Itza, experienced the same droughts and survived them, or flourished through them, during the Terminal Classic. If the same rain fell on the north and the north did not empty, then rainfall by itself is not the explanation for the south.

Our file goes further and names the failure mode by its scholarly name. Researchers caution against environmental determinism, meaning over-attributing the collapse to drought while underweighting human agency, political dynamics and social factors. That northern survival suggests institutional resilience, geographic factors such as access to cenotes and coastal resources, and different political structures mattered as much as climate did. Cenotes are the point worth holding on to, because it turns an abstraction into a landscape: in the north, permanent water sat in open sinkholes below the surface even in years when the rain did not come.

Our file's stated emerging consensus, synthesized from David Webster's The Fall of the Ancient Maya (2002), Richardson Gill's The Great Maya Droughts (2000) and the work of Takeshi Inomata and colleagues, is multi-causal: prolonged drought acted as the ultimate stressor on a system already weakened by overpopulation relative to agricultural capacity, by deforestation and soil erosion in the thin karstic soils of the Peten, by intensified warfare between rival polities, and by the erosion of political legitimacy as divine kings proved unable to deliver rainfall and prosperity. The collapse was a combined ecological, political and social failure. Read the operative phrase carefully. It says drought was the ultimate stressor on an already weakened system, which is a different sentence from drought caused the collapse.

Family one, the forest. Pollen core studies from multiple lowland lakes, including Lake Salpeten, Lago Petexbatun and Lake Peten Itza, show evidence of massive deforestation during the Late Classic, with forest pollen declining dramatically and disturbance indicators such as grasses and weeds increasing.

David Lentz and colleagues documented that clearance around Tikal in 2014, estimating that the Tikal polity cleared roughly 80 percent of its surrounding forest for agriculture, construction timber and lime plaster production, burning enormous quantities of wood to make the white plaster that covered Maya buildings. That last detail is the one to sit with: the brilliant white cities were, quite literally, built by burning the forest they stood in.

Why that would matter is a mechanism argument rather than a measurement, and our file states it in the conditional, which this article keeps. Deforestation on the thin karstic soils of the Peten would have exacerbated drought impacts by reducing soil moisture retention, increasing surface runoff and erosion, and potentially altering local rainfall patterns. Would have. Potentially. Those words are load-bearing.

Family two, war. Hieroglyphic texts and archaeological evidence document escalating warfare among lowland Maya polities during the 8th and 9th centuries CE, and defensive walls, hastily built fortifications and evidence of violent destruction increase dramatically in the 9th century at multiple sites. The word hastily is doing real work there: these are not the works of a state planning ahead.

The best-excavated instance is the Petexbatun War of the late 8th century, documented by Arthur Demarest of Vanderbilt University through excavations at Dos Pilas, Aguateca and surrounding sites. It shows a wave of destruction, fortification building, and eventual abandonment directly caused by interstate warfare.

Whether warfare was a cause or a symptom is genuinely unresolved, and this article does not resolve it. On one reading, drought-induced food scarcity likely intensified conflict over diminishing resources, which makes war a downstream effect. On the other, Demarest emphasizes that in the Petexbatun region warfare clearly preceded the major drought episodes, which would make political fragmentation and militarism independent causes rather than drought consequences. That second position comes from the scholar who dug the relevant sites, and it is the strongest single argument against a drought-first narrative. Both readings are carried here at strength.

Family three, the throne. Classic Maya political legitimacy rested on the k'uhul ajaw, the divine king, as intermediary between the human and supernatural worlds, responsible for propitiation rituals including bloodletting ceremonies and invocations to bring rain. That job description is the vulnerability: a ruler whose authority is founded on delivering rain has staked it on the weather.

Lisa Lucero argued in 2002 that when drought persisted despite royal rituals, the ideological foundation of divine kingship eroded and populations lost faith in the ruling elite's ability to manage the cosmic order. Keep the attribution visible, because this is the most narratively attractive claim in the whole file and among the least directly evidenced. It is one named scholar's argument, at Tier 2, and it is offered here as that.

There is a physical trace that fits it, and it cuts against the emptying story hard. At several sites, late occupation phases show non-elite squatters occupying former palace and ceremonial spaces, which suggests the breakdown of social hierarchy preceded final abandonment. People were still there. They were living in the throne rooms. The cities did not empty first: the ruling order failed first.

Family four, the arithmetic of land. Richardson Gill, in The Great Maya Droughts (2000), emphasized that Late Classic Maya populations had likely exceeded the sustainable carrying capacity of their landscape, a conclusion supported by soil exhaustion, widespread terracing that indicates pressure on marginal land, and agriculture expanding into the bajos, the seasonal swamplands. Under normal rainfall the overshoot was manageable; under drought the food production shortfall became catastrophic. That is the cleanest one-sentence version of the multi-causal thesis anywhere in our file, and it is worth noticing that it sits at Tier 2, not Tier 1.

An eruption of an unidentified tropical volcano in 818 to 822 CE, detected as a sulfate spike in ice cores, may have contributed short-term cooling or climate disruption coinciding with the first pulse of collapse. Direct evidence linking this eruption to Maya drought is lacking, and note that the volcano itself has no name because nobody has identified it.

Researchers have speculated that epidemic diseases, unrelated to the post-contact Old World diseases that came later, may have contributed to the population decline. No specific pathogen has been identified in the pre-Columbian Maya record for the Terminal Classic, which leaves this as a shape in the data with nothing in it.

Takeshi Inomata and colleagues have suggested the collapse may have involved cascading refugee movements, with populations fleeing drought-stricken or warfare-torn polities overwhelming neighbouring polities' capacity and triggering sequential collapses. Our file rates this plausible but difficult to demonstrate archaeologically, and the difficulty is real: the pattern such a cascade would leave looks a great deal like the pattern several other explanations would leave.

In 2025, McCool, Codding, Kennett and colleagues published in the Proceedings of the National Academy of Sciences a study applying population-ecology models and piecewise structural equation models to paleoclimatological and demographic data for the Classic Period Maya lowlands, 250 to 1000 CE. They find that the rise and expansion of Classic Maya cities resulted from the interaction of climate downturns, intergroup conflict and strong economies of scale realized through capital investment in agricultural infrastructure. Their reported counterintuitive result is that the abandonment of cities occurred under improving climatic conditions, at the point when conditions in rural areas got better.

No, the collapse has no supernatural explanation, and it has nothing to do with the supposed end of the Maya calendar in 2012 CE. The Long Count does not predict the end of the world. The 2012 date, 13.0.0.0.0, simply marks the end of one great cycle, a b'ak'tun, and the beginning of another. Our library's file on Maya epigraphy and calendar science makes the case in more detail: only one ancient Maya inscription, Tortuguero Monument 6, references the 13.0.0.0.0 date at all, and its damaged text appears to describe a ritual; inscriptions at Palenque and Coba reference dates millions of years in the future, which is not the behaviour of a calendar that expects to stop; and the supposed galactic alignment occurs over a roughly 36-year window and is a modern construct absent from Maya texts. The 2012 phenomenon was a product of New Age appropriation, from Jose Arguelles and John Major Jenkins, rather than of Maya scholarship.