Modelling and Optimising Flows Using Parallel Spatial Interaction Models

The paper demonstrates some of the benefits that high performance computing has to offer geographers. It reports the results of porting and then using very large spatial interaction models on the Cray T3D parallel supercomputer in modelling and spatial optimisation applications that would otherwise have been judged computationally infeasible. Keywords optimisation, simulated annealing, parallel design, data parallel, retail networks, modelling Background The entropy-maximising spatial interaction model is used for modelling many different types of flow data in both research and applied contexts. Examples of flow data are: journey to work trips, airline traffic, retail behaviour, world trade and telephone traffic in which flows of people, money, information etc., connect origin zones to destination zones. These zones have a geographical expression on maps whilst the flow data provides a useful summary of the very complex processes and behaviour patterns that generated them; for example..