The disappearance of the small-bodied hominin species Homo floresiensis from the Liang Bua cave on the Indonesian island of Flores has been debated for two decades. The discovery of these individuals in 2003 triggered a wave of speculation that ranged from contact with modern humans to disease to volcanic catastrophe. The revised stratigraphy released in 2016 closed the door on many of those theories, showing that the species vanished well before Homo sapiens reached the site. What remained was a mystery defined by absence. There was no evidence of violence. No sign of mass death. No abrupt cultural layer marking a final moment. The new study from Communications Earth & Environment changes everything by introducing a complete, high resolution, season specific rainfall reconstruction from a stalagmite growing less than two kilometres from Liang Bua. This single record captures the final 45,000 years of climate conditions that shaped the fate of the species. It shows that the disappearance of Homo floresiensis and its primary prey animal Stegodon florensis insularis coincided with a long decline in water availability that reached its most severe point between 61,000 and 55,000 years ago. The evidence is precise, dated, and directly connected to the freshwater supply that sustained the small community at Liang Bua.
The stalagmite in question, labelled LR09 K2, formed inside Liang Luar cave. Its chemistry changes in step with rainfall. Magnesium to calcium ratios rise when the region dries, because lower water infiltration allows more prior calcite precipitation in the karst above, concentrating magnesium. Oxygen isotopes track a mixture of seasonal rainfall sources. Together they provide a high fidelity measurement of annual, summer, and winter rainfall going back 91,000 years. The data reveal a climate that shifted from wet and stable to variable and highly seasonal, before finally collapsing into a state marked by minimal summer rainfall and persistent dryness across the year.
Between 76,000 and 61,000 years ago mean annual rainfall dropped from roughly 1560 millimetres to around 990 millimetres. This was not a short lived event. This was a 15,000 year slide driven by changes in monsoon behaviour, sea level, and the position of the Intertropical Convergence Zone. The most critical detail is that summer rainfall, which accounts for most of the water that recharges rivers and groundwater, declined by more than half. In the best years of Marine Isotope Stage 5 the uplands above Liang Bua received over a metre of rain during the monsoon months. By 61,000 years ago that value had fallen to about 450 millimetres, the lowest point in the entire 91,000 year climate sequence. Winter rainfall also decreased, further stressing river flow during the driest months. This combination pushed the region toward a condition in which perennial streams shifted to seasonal flow and local aquifers became increasingly sensitive to evaporation and seepage loss.
The hydrological collapse is recorded not only in the stalagmite chemistry but also in the uranium activity ratios that rise when groundwater residence time increases, indicating slow infiltration through dry karst. The rise begins around 55,000 years ago, precisely when the region reached maximum aridity. The chemical evidence shows that the karst system above Liang Luar and Liang Bua was drying out, meaning that the river feeding the Wae Racang valley would have become unreliable during the later dry seasons.
This record aligns tightly with fossil changes inside Liang Bua. Stegodon florensis insularis dominated the assemblage during the entire occupation sequence of Homo floresiensis. These pygmy proboscideans had the same strict requirements for freshwater that modern elephants possess. Tooth enamel preserves the oxygen isotope signature of the water these animals drank. The isotopes track the same swings seen in the stalagmite, confirming that these animals relied on the same source of groundwater fed river water. The match is tight across multiple excavation sectors, indicating that the isotopic signal is not an artefact of sampling. Only three individuals stand out as outliers, and these likely drank from small localised pools or runoff sources before deposition. The rest point to a herd reliant on a single drying catchment.
The fossil record shows that Stegodon abundance declines in the same interval when rainfall seasonality intensifies. Ninety two percent of all Stegodon remains from the key sectors fall between 76,000 and 62,000 years ago. After 62,000 years ago there are almost none. The youngest Stegodon in the entire deposit is dated to 57,000 years ago. This nearly matches the predicted depth based on the stalagmite curve and precedes the arrival of modern humans by nearly ten thousand years. The decline was not sudden. It begins as the climate enters its second phase around 76,000 years ago, a period marked by wetter summers during abrupt Greenland stadial events but increasing dry season stress. The presence of young Stegodons in the assemblage shows that Homo floresiensis hunted or scavenged subadults, which are the most vulnerable individuals during water stress. As water availability decreased these animals would have been forced to move more frequently or travel farther to find drinkable sources. The youngest individuals would have struggled to survive long dry intervals, creating a population pressure independent of predation.
Homo floresiensis tracked closely with Stegodon distribution. Their skeletal remains follow the same pattern, with the last appearing at 61,000 years ago. The species persists through the onset of high seasonality but disappears just as summer rainfall reaches the lowest point in the entire climate record. The river would have dwindled during repeated dry seasons. Groundwater recharge would have failed. Water stress becomes visible in the stalagmite through the rising uranium activity ratios that mark long residence time in a drying karst. This is the same interval when Stegodon becomes nearly absent in all excavation sectors. The site does not show signs of catastrophic death or sudden replacement. Instead it shows a weakening presence of both predator and prey, followed by abandonment.
The authors of the study outline three climate phases. The first, from 91,000 to 76,000 years ago, was wet and stable. Rainfall was higher than today and seasonality was low. River baseflow would have been reliable year round. The second phase, from 76,000 to 61,000 years ago, saw higher seasonality, meaning strong wet seasons punctuated by harsh dry seasons. This created open grazing environments and concentrated animals near remaining water. The fossil record shows increased interaction between Homo floresiensis and Stegodon in this period, matching the ecological expectation that water restricted landscapes force predators and prey into close proximity. The third phase begins at 61,000 years ago and marks the true collapse. Mean annual rainfall drops below modern levels. Summer rainfall reaches its minimum. Winter rainfall remains weak. Groundwater signals show acute drying inside the karst. From this point onward the support capacity of the landscape falls below the threshold needed to sustain Stegodon herds and the hominin group that depended on them.
This is the first time a precise, season specific rainfall reconstruction has been placed directly against the archaeological and palaeontological record at Liang Bua. The alignment is exact enough to resolve decade scale behaviours in a sequence tens of thousands of years long. The evidence does not require sudden extinction. It instead implies migration away from the Wae Racang valley into more favourable catchments. Stegodon may have shifted to better watered parts of western Flores. Homo floresiensis may have followed. If Stegodon populations were already small, even low intensity predation on juveniles would have accelerated decline, but the climate conditions already placed them under extreme stress. The absence of both species in the upper layers of Unit 1B and above suggests that neither returned to Liang Bua after the dry phase set in.
This scenario explains the long quiet in the archaeological record before the arrival of Homo sapiens around 46,000 years ago. It also resolves the puzzle of why Homo floresiensis persisted on Flores for so long, only to vanish without evidence of a violent end. The species endured repeated glacial and interglacial cycles but met its limits when a rare convergence of lower sea levels, cooling sea surface temperatures, and weakened monsoon circulation altered the balance of summer and winter rainfall. This created an environment in which the primary freshwater source became unreliable. Once the dry phase began it lasted long enough that recovery was unlikely within the lifetime of a small isolated population.
The broader implication is that island dwarf species are extremely sensitive to freshwater thresholds. The collapse described in the stalagmite record is not a global climate anomaly. It is a regional reorganisation of monsoon systems that pushed one valley on one island into a state where water availability could no longer sustain the animals that shaped its ecological web. The precision of the record makes it clear that the disappearance of Homo floresiensis was not a rapid or mysterious event. It was a gradual withdrawal triggered by the simplest environmental pressure possible, the loss of dependable water.
The study provides a foundation for reinterpreting the last known population structure of Homo floresiensis. It suggests that future excavations on Flores should target areas with more reliable aquifers or larger catchments. If the species survived beyond 61,000 years ago its presence would likely be found in zones where summer rainfall remained strong enough to maintain perennial streams during the driest years. The stalagmite record gives exact rainfall profiles for those periods, offering a framework for locating remaining refugia.
The decline of Homo floresiensis and Stegodon florensis insularis at Liang Bua is no longer an enigma. The decisive factor was progressive aridification that peaked between 61,000 and 55,000 years ago. The record shows that the species did not disappear in isolation but in parallel, bound together by a shrinking river in a valley that could no longer sustain them. This reconstruction replaces speculation with measurable evidence and reframes one of the most debated chapters in human evolution as a story defined by climate pressure rather than contact or catastrophe.
Source:Â
Based on the 2025 study in Communications Earth & Environment: https://doi.org/10.1038/s43247-025-02961-3
