The physical conditions in informal settlements are not appropriate due to uncontrolled and unplanned growth in these areas. Simulation of informal settlements growth leading to better understanding of the complexity of its spatial process and how its formation. The phenomenon of informal settlements growth as a complex system has a macro pattern that results from the interaction of downstream agents at the micro level, and accordingly the feedback of the macro patterns affects the interactions at the micro level. In this study, it is attempted to consider this complex process of informal settlements growth. Agent-based models have the innate ability to model the behavior and interaction of agents and provide an ideal framework to study the informal settlements growth.
Since the informal settlements are often formed organically the geometric shape of the land parcels is highly irregular, therefore choose a suitable spatial unit is among the challenges that local scale of simulation is confronted. In this study, it is attempted to address this challenge. Decision maker agents were combined with land parcel agents and simulation was performed on a local scale. In this model, the suitability of the land parcel agents was calculated based on factors of neighborhood, accessibility, physical suitability and spatial constraints using the spatial analyses in the GIS environment. Each parcel calculates its own general suitability value using these factors. Decision making for selecting the parcels was carried out by household agents who worked in downtown and those who worked in industrial areas around the city. For urban worker agents, a land parcel is desirable which has high suitability, small area, close distance to the city center and same agent type in the majority of the inhabitants around the land parcel. Besides, for the industrial worker agents a land parcel is desirable that has all conditions mentioned above and is also close to the industrial areas. The sensitivity of developed model was investigated on the basis of shape and size of land parcels more specifically. In this study, the rectangle and square shapes were used as the common geometric shapes of the land parcels in 20 different scenarios. The area of the residential parcels is always considered among the most important physical indicators for investigation of informal settlements at a local scale. So, in this study, it is attempted to investigate the land parcels in various areas. Besides, in this study, a fishnet grid of cells was created in two cell sizes of 2 and 5 m and the study area was also investigated with this configuration. In other words, by creating a raster-like space and comparing its results with the vector status it is attempted that the role of selecting the spatial unit in simulation process of informal settlements local growth also be analyzed from the perspective of the data model. All of these 22 scenarios was implemented using the developed agent based model on one of the informal settlements neighborhoods in the Kashan city of Isfahan province of Iran. Evaluation the results by real data showed that the accuracy of the model developed by rectangular land parcels with an area of 50 to 150 m² is better than other vector and raster-like spatial units and is equal to 64%. Besides, the results showed that the overall accuracy of the raster-like models is significantly decreased compared to the vector structures.