Emulating human expertise in rezoning: a case study of the Australian Statistical Geography Standard

Large geographical areas must be rezoned into smaller areas to support spatial statistical analyses providing community-level insights. The rezoning process is influenced by qualitative and subjective agency-specific requirements, leading to challenges such as information loss, and limited results comparability. This study aims to quantify and compare common rezoning criteria - compactness, homogeneity and equality, and explainspatial scientists' decisions when rezoning. This will allow us to assess their importance in rezoning, facilitating greater statistical framework comparability and assist in process automation. The methodology developed in this study relies on the ordinary and Tikhonov regularised least squares. This approach provides a weighted measure of each criterion’s contribution, ensuring stability, accuracy and consistency of a geographic framework. This study’s motivating example is the Australian Statistical Geography Standard (ASGS), and its basic spatial unit - Mesh Block. Further testing was conducted on simulated data. We combined our methodology with an existing aggregation algorithm, HeLP. Using the rezoning criteria with the addition of context-specific criteria, such as land use adherence, and natural and urban boundaries, we were able to recreate a rezoning schema that attains the same underlying properties as the ASGS. We provide a heuristic approach for decision-making in rezoning parcel-based systems such as the ASGS.