Abstract
This thesis presents a spatial analysis of pre-European Pa sites across New Zealand with a focus on the North Island. Few spatial analyses of Pa have utilised sophisticated statistical methods or been carried out at a large scale. Those that have largely do not account for potential confounders or investigate the implications of their results in terms of social organisation. These past studies therefore motivate our three research questions: What is the density distribution and clustering pattern of Pa sites in New Zealand? What are the dominant variables governing the density and clustering of Pa sites across New Zealand? What are the implications of the identified patterns in terms of social organisation?
Kernel density estimates are calculated for the Pa of the North and South Island. These estimates support previous statements by researchers about the distribution of Pa: Pa are most dense in the north of the North Island, decreasing southwards. The density of Pa is then compared to the density of non-Pa via the relative risk function in order to curb the confounding effects of population and bias introduced by unsystematic surveying. This shows that Pa density is relatively high through the central North Island and relatively low through Auckland and Northland.
Clustering is explored using the pair-correlation function (PCF). This reveals that Pa show strong evidence of large clusters in the absence of deterministic heterogeneity. Heterogeneity in the Pa point pattern from sources such as geography and the human population distribution is then accounted for by utilising the inhomogeneous PCF. This suggests that Pa in the North Island form clusters that then repel each other. The central and northern regions identified by the relative risk are compared: the central North Island shows the same pattern as the North Island in general but the northern region demonstrates a second level of clustering at larger distances indicating clusters of clusters."
The spatial relative risk of Pa is modelled against temperature, solar radiation, land wetness, distance from the coast, soil drainage class, and slope. In all the models fitted only the distance to coast parameter is significant. The parameter estimate indicates that the relative risk of observing a Pa increases as the distance from the coast increases.
The spatial patterns identified are interpreted as indicating that society and by extension Pa were centred around the hapu, which were largely independent of their respective iwi, except in the northern North Island where there was more political centralisation. This resulted in a proportionately small number of large Pa in the north and large number of small Pa through the central North Island. More neighbours through the central region may have also promoted more Pa building.
We conclude that Pa are driven more by social factors than environmental ones.