Limescale in the Pont du Gard aqueduct traces 300 years of water crises and one deliberate diversion

The record depends on hard water leaving a layer at a time. Inflow and temperature vary through the year, so the layers differ from one another, and the Mainz geologists read them the way a dendrochronologist reads rings ^[10]. Ten sample sites along 50 kilometers of channel works out to one about every five kilometers ^[9][2]. That spacing is what lets the rural reaches be compared against the stretch approaching the city.

The aqueduct's first decade sits at the bottom of every sample. "At the base of the samples, there is a carbonate layer about two centimeters thick, covered by a thin layer of clay," said Cees Passchier ^[11]. He attributes that clay to a drought in the western Mediterranean, an episode the Roman historian Tacitus also recorded ^[12]. Above it the chemistry changes. "The nature of the carbonate changes in the subsequent layers: after 11 years, a different type of water was introduced. An additional source was tapped to ensure the aqueduct had a sufficient water supply," Passchier said ^[13]. Counting from a construction date around 50 A.D., that lands at about 61 A.D. ^[1]. The stonework agrees: sidewalls on the Pont du Gard were raised, reinforced and partly buttressed to hold a higher water level ^[14].

Around 150 A.D. the layers thin sharply, meaning the level in the channel fell ^[15]. "About half of the water supplying the city was diverted to irrigate fields," said Gul Surmelihindi ^[16], who takes the size of the withdrawal as evidence that it was authorized ^[17]. Only highly influential individuals could have arranged it, she argues, and the city does not appear to have needed the water ^[18]. The half comes from the drop in deposition; the report does not include the conversion from layer thickness to flow ^[27]. The channel then ran for roughly another century at the lower volume, working without fault ^[19].

From the mid-third century the carbonate turns porous and picks up clay and sand, which is what an unmaintained channel deposits ^[20]. "Great economic and political turmoil prevailed across many parts of the Roman Empire. The carbonate deposits reveal that the city's condition was deteriorating," said Surmelihindi ^[21]. Then, in the samples near the city, deposition stops around 270 A.D. ^[22]. In the rural sections it continues, which the team attributes to owners of large estates taking water upstream ^[23]. Water was still moving in the upper channel, so a dry spring or a collapsed section cannot explain the halt near Nimes ^[23].

That 270 date is a layer count anchored to a construction date around 50 A.D., so it is only as good as the counting across some 220 layers ^[2][10][3]. The span is also shorter than the roughly 300 years of intervention Surmelihindi describes ^[7]; the extra time sits in the rural reaches, where carbonate kept forming after the city was cut off ^[23]. The samples record water chemistry, water level and the state of the channel, so depopulation is an inference drawn from them ^[6]. The plague reading comes from co-author Andrew Wilson of the University of Oxford ^[25]. "Not much is known about this plague in Western Europe," Surmelihindi said ^[26].