Fig 1.
(Top left) Rapa Nui in East Polynesia, (top right) locations of image-ahu on Rapa Nui, and (bottom) Ahu Tongariki with moai (Photo by R.J. DiNapoli).
Fig 2.
Archaeological and environmental data used in our analyses.
(A) locations of image-ahu, (B) locations of freshwater sources, (C) marine resource locations, (D) minimal rock mulch classification, (E) medial rock mulch classification, (F) maximal rock mulch classification.
Fig 3.
(A): L-function of ahu compared to 39 simulated realizations of CSR; (B): inhomogeneous L-function of ahu compared to 39 simulated realizations of CSR. Y-axes are the values of the L-functions at separation distances (r) in meters (x-axes). Results indicate that image-ahu have an inhomogeneous intensity lacking second-order properties, though there is some evidence for dispersion at distances >1500 m. Black lines are the empirical L-functions, red dashed lines are the theoretical expectations under the null model, and the grey-shaded region is the envelope of 39 Monte Carlo simulations of the null model.
Fig 4.
Spatial Kolmogorov-Smirnov (SKS) tests.
SKS tests for the relationship between image-ahu (black squares) and distance (m) from subsistence resource locations (choropleth maps). Observed distribution (black lines) is compared to the expected distribution under CSR (dashed red lines) with the alternative hypothesis being that ahu are nearer to these resources than random. Results suggest ahu are significantly clustered near freshwater sources (D+ = 0.59, p<0.0001), marine resource locations (D+ = 0.65, p<0.0001), and the maximal rock mulch garden classification (D+ = 0.24, p<0.0001). Results for minimal and medial mulch classifications can be found in S8 File.
Table 1.
Point-process model selection for the relationship between ahu and subsistence resources.
Smaller change in information criteria score (ΔBIC and ΔAIC) and higher weight indicate best-fitting model.
Fig 5.
Effect of distance from freshwater sources on the spatial intensity of image-ahu.
Ahu spatial intensity declines with distance from freshwater sources. Grey-shaded region represents the 95% confidence interval.
Table 2.
Covariates for best-fitting model 5.
Negative covariate estimates indicate that ahu intensity decreases with distance from the coast and freshwater sources.
Fig 6.
Residual L-function for best-fitting model.
Red dashed line is the theoretical L-function of the model, grey shaded region represents the upper and lower bounds of 39 Monte Carlo simulated realizations of the model (p = 0.05), and the black line is the L-function for ahu. Y-axis is the value of the L-function at distances (r) in meters (x-axis) Results indicate no significant deviation between model 5 and the data.
Fig 7.
Simulated realizations of the best-fitting model.
20 simulated realizations of the best-fitting model 5 incorporating distance from the coastline and distance from freshwater sources.