SLEUTH-3r Modeling of Urban Expansion in a Challenging Case: Insights from Ottawa’s Dispersed Growth Landscape

Abstract

This study examines the use of the slope, land use, exclusion, urban, transportation, and hillshade (SLEUTH-3r) urban modeling tool for predicting urban growth in large areas with dispersed urban development. The SLEUTH-3r, an enhanced version of the original SLEUTH model, was developed to improve urban growth predictions across expansive regions by optimizing the calibration process and balancing different growth patterns such as edge and spontaneous growth. This was achieved through the use of multipliers, such as the diffusion multiplier ( D _M ), which regulates the initial urbanization pixels and adjusts the model’s tendency toward edge growth. We applied SLEUTH-3r to predict urban growth in the National Capital Region of Canada, characterized by multiple urban clusters. D _M values were tested for both original (30 × 30 m) and reduced-size (90 × 90 m) images, using two versions of the exclusion layer with low and high Monte Carlo iterations for model calibration. We assessed urbanization likelihood images, fit metrics, landscape metrics for predicted land use/cover (LULC), and the time required for modeling. Surprisingly, the use of reduced-size images (90 × 90 m) yielded more accurate predictions of urban growth, providing sufficient urban pixels for future projections. Our results showed that the coefficients and fit metrics of SLEUTH-3r should be interpreted relative to the model settings and input specifications. Thus, we recommend integrating assessments beyond basic fit metrics into final predictions, including insights from calibration, validation, and prediction phases. The reduced-size images outperformed the original-sized images in urbanization likelihood, fit metrics, landscape metrics of the target year’s LULC image, and calibration time. Given SLEUTH-3r’s advanced capabilities and the availability of high computational speed and memory, researchers may opt for highly detailed data. However, for large-scale urban modeling with dispersed growth, we recommend testing reduced-size images alongside original-resolution images during coarse calibration to determine the optimal resolution, D _M , and the exclusion layer for obtaining reliable results.