Subjectivity versus Objectivity: Comparative Study between Brute Force Method and Genetic Algorithm for Calibrating the SLEUTH Urban Growth Model

Urban growth models (UGM) as regional planning tools are of great interest for quantitative analysis of urban complex systems. As a crucial step, model calibration is one of the most important and challenging steps when trying to simulate a spatial phenomenon. The current paper adopts two different approaches to calibrate a popular geospatial simulation model, the SLEUTH UGM. The conventional Brute Force as a subjective method and the genetic algorithm (GA) as an objective approach were implemented to calibrate the model for three study locations of Azadshahr, Gonbadekavoos, and Gorgan Cities, Golestan Province, Iran. Model simulation success was measured and compared for three modeling efforts using multiple methods [optimized SLEUTH metric (OSM), Kappa coefficient, receiving operator characteristic (ROC) statistic and landscape metrics]. Results indicated that GA-based model calibration out-performed the Brute Force method in terms of landscape metrics, Kappa coefficient (Khisto) and the final OSM values. On the other hand, the Brute Force model yielded better results for Klocation. Both models depicted an approximately equal performance in terms of the ROC statistic. The majority of the resultant growth coefficients derived from both methods were relatively close, while GA-base model calibration out-paced the Brute Force with a noticeable less time-consuming process to calibrate the model.