K'gari, the world's largest sand island, hosts more than half the world's perched lakes — freshwater bodies held above sea level by impermeable layers of cemented sand, decomposed organic matter, aluminum, and iron. These lakes were assumed to have contained water continuously since the end of the last ice age, 11,500 years ago. They are ancient. They look permanent.
Tibby, Burns, and Hofmann (Journal of Quaternary Science, 2025) found that several of K'gari's deepest and oldest lakes — including Lake Boorangoora, Lake Allom, and Basin Lake — dried out between 7,500 and 5,500 years ago. The evidence is a gap: sediment cores show missing layers where continuous deposition should appear. The absence of mud means the absence of water. The lakes were not there.
This would be unremarkable if it coincided with a drought. It didn't. The mid-Holocene in subtropical eastern Australia was a period of higher rainfall. Minjerribah, a sand island 250 kilometers to the south, maintained its lakes through the same period. The region was wetter. K'gari was dry.
The mechanism is directional. Southeasterly trade winds supplied rainfall to Minjerribah but not K'gari. The rain was real. It fell in the wrong place. The regional climate descriptor — “wetter period” — is true at one spatial scale and catastrophically false at another. K'gari's lakes didn't know about the regional average. They only knew whether rain fell on them.
Perched lakes are particularly sensitive to this spatial specificity because they are entirely rain-fed. They have no connection to groundwater or rivers. Their only water source is precipitation that falls directly on the catchment above their impermeable floor. A directional shift in prevailing winds can redirect rainfall across a distance smaller than the spatial resolution of a climate reconstruction while leaving the regional average unchanged. The lake dries. The proxy record — the one that doesn't include K'gari's own sediment — says it was wet.
The structural observation is about the resolution of climate descriptors. “Wet period” and “dry period” are spatial averages presented as temporal categories. They describe when, but they hide where. A period is wet because most of the proxies in the reconstruction are wet — but the proxies are not uniformly distributed, and the rainfall is not uniformly distributed either. The average is a statement about the collection of measurement sites, not about every location within the region. Any specific location can diverge from the average without the average knowing.
This is not the discovery that local climate can differ from regional trends — that's obvious. The discovery is that perched lakes on sand islands serve as spatial detectors of this divergence precisely because they have no hydraulic buffer. Lakes connected to aquifers or rivers integrate water over large areas, smoothing spatial variability. Perched lakes integrate over nothing but their own patch of sky. When the wind shifts, they respond immediately and completely. The regional average keeps rising. The lake disappears.
The Butchulla people, whose traditional country K'gari is, call the lakes “the eyes of K'gari.” The eyes closed for two thousand years during a period that climate science describes as wet.