The collective memory of floods and droughts influences where people settle, how they build and how they prepare for the next climatic extreme. Over decades, that memory can fade, and communities may return to places that once proved dangerous. How such memory decayed over centuries is hard to observe because the climate past societies experienced is usually known only from their own records.
A new study published in Nature Communications tackles this problem by comparing two independent kinds of evidence.
The study, led by Dr. Shuang Song of the Max Planck Institute of Geoanthropology, with colleagues from Nanjing University, Beijing Normal University and Fudan University, examined official histories, local gazetteers, diaries and inscriptions from 32 sites in the Yellow River Basin of northern China. They compared these with tree-ring reconstructions of summer rainfall and river flow – records which do not pass through human perception.
Researchers found that the two records broadly agree, but where they do differ, they do not do so randomly. Archival writers tended to overstate a year's severity when it was only slightly drier or wetter than similar events a few years earlier, and to understate it when it was the opposite. Many records make such comparisons explicitly.
"Historical writers were not simply recording the weather. They were comparing each year with what they and their contemporaries remembered," Song says.
To measure how far back that memory reached, the team applied artificial time frames to the tree-ring record, so that each year was judged only against the preceding decades. Agreement between the written archives and tree-ring records improved most for windows of 20 to 40 years, with an average optimum of about 33 years. This range closely matches estimates of life expectancy in historical northern China, suggesting that each generation's sense of "normal" was set by the climate it had lived through. Psychologists and ecologists call this a shifting baseline.
To test these results, the researchers built an agent-based model in which simulated observers are born, age, record extreme years and die. Only when observers judged climate against their own lifetime did the model reproduce the generational pattern seen in the archives, with length of lifetime as the main factor determining where the pattern peaked. This is consistent with the social phenomenon known as generational amnesia, in which a generation “forgets” things that they did not directly experience.
"As climate change accelerates, understanding how each generation resets its sense of what is normal becomes important for how societies prepare for extremes," says Dr. Patrick Roberts, Director of the Department of Co-evolution of Land Use and Urbanisation at the Max Planck Institute of Geoanthropology and senior author of the study.
The authors argue that although the model is simple, the results suggest that documentary records of past climate should be read as climate signals filtered through human memory. Correcting for that filter could improve reconstructions of past climate and help explain why risk awareness fades between disasters.
Nature Communications
Archival and palaeoenvironmental documentation of extreme climate events reveals generational amnesia in collective memory
9-Nov-2026