Employment locations are modelled using census meshblock data. First, the meshblock polygon layers are converted to a point layer of meshblock centroids. Then total employment statistics are attached to these points. This yields a series of points across the study area representing sites of employment, which have a weighting for their significance (ie number of jobs in each meshblock).
Randomly distributing all jobs as points within meshblocks, while excluding roads and other non-employment related land uses, would yield a more realistic distribution of employment opportunities. However in a city the size of Christchurch, this would result in a dataset of ~145,000 points as opposed to ~3000 points for meshblock centroids. This would then generate up to ~142,000 extra potential paths which would need to be solved for each site for which the accessibility score is to be calculated. This would result in a large increase in processing time and for the purposes of this research would not assist the development of the accessibility model methodology. Nevertheless, for real world examples this approach may be worthwhile if combined with a Monte Carlo statistical simulation.