Spatial Clustering, Detection and Analysis of

https://doi.org/10.1016/B978-008044910-4.00523-XGet rights and content

The distribution of geographical phenomena through space is autocorrelated if the measurement at one location is related to that at other locations. Positive spatial autocorrelation refers to the patterns where nearby or neighboring values are similar, while negative spatial autocorrelation refers to the patterns where nearby or neighboring values are dissimilar. Spatial cluster refers to positive spatial autocorrelation. Cluster detection is important for spatial pattern analysis since the presence of spatial clustering violates the common assumption of most statistics on the independence among the samples.

Cluster detection and analysis fulfill two tasks: to test if the distribution is clustered and to detect the location and extent of clusters. These two tasks correspond to 1st-order and 2nd-order pattern analysis, which are carried out through global and local scale analyses, respectively. To evaluate the clustering of a distribution, the observed measurements are compared to the expected measurements that are derived from a random distribution or Monte Carlo simulation. Edge effects and scale issues are two major concerns for cluster detection.

Clustering detection can be conducted for point, linear/flow, and areal data. The dynamics of spatial clustering patterns through time are revealed by space–time cluster analysis. When the correlation among multiple variables or different types of geographical events is concerned, multivariate spatial pattern analysis is conducted. Cross-scale spatial clustering analysis reveals the dynamics of clustering pattern of the same sample across different analysis scales.

References (0)

Cited by (0)

View full text