South Arabia’s prehistoric monument landscape shows social resilience to climate change

PLOS One | https://doi.org/10.1371/journal.pone.0323544 May 28, 2025 13 / 24 technological requirements (X2 (1, N = 298) = 36.994, p < 0.001). Single-episode constructions are significantly correlated with larger volumes (Fig 7, left), which is consistent with a large assembly of people. Conversely, accretive constructions are correlated with smaller monument volume, which is consistent with small groups and sequential labor. The greater labor invested in large platforms involved simultaneous labor—more adults to lift the large stones. We found chronological patterns. Our survey found fewer platforms than other monument types, a fraction of the HCT or triliths counted, despite the longer timespan during which platforms were built. Within our sample, roughly one plat- form per 100 years contrasts with one HCT every 12 years and one trilith (full line of elements) every four (the frequency of triliths increases if each plinth is counted as an accretive element). Furthermore, an ANOVA with Tukey post-hoc test showed that platforms have significantly higher (F(4,291) = 6.034, p = 0.00011) total stone volumes than the later monu- ment types (Fig 7, right) and thereby represent the greatest cumulative labor investment. With HCT, a subsequent pulse of monument construction began around 5200 cal BP. Because HCT would not hold up if built in accretive phases, their construction also indicates single-episode construction (which makes sense when burying a corpse in the heat). There are exceptions [20]. Nonetheless, HCT have significantly smaller stone volumes than platforms.

The next chronological-cultural pulse—accretive trilith monuments--follows increasingly arid conditions beginning around 2700 cal BP [43,44]. Monuments with accretive construction (triliths, boat graves) have significantly lower (F(1,294) = 9.776, p = 0.00194) total stone volumes, compared to those with episodic construction. Fig 7 shows that accre- tive monuments, such as the numerous triliths, have few large stones (> 0.75 m). Fig 5. The regularized multinomial logistic regression model coefficients show the relationship between the environmental variables and the monument types. Standard errors cannot be accurately estimated in regularized logistic regression. A positive coefficient indicates that the monument type is associated with higher values of that variable, whilst a negative coefficient shows that the monument type is associated with lower values of that variable. https://doi.org/10.1371/journal.pone.0323544.g005