Land Suitability Zoning of Natural Lands in Ahvaz for Infill Development Using a Fuzzy–Fuzzy Inference System and GIS

Introduction                                  
Rapid urbanization, particularly in developing countries, has placed immense strain on finite urban resources, manifesting in challenges such as agricultural land degradation, heightened urban heat island effects, and escalating infrastructure costs. As projections indicate that nearly 70% of the global population will reside in cities by 2050, contemporary urban planning has pivoted towards strategies that promote sustainable intensification. Infill development, which focuses on the optimal utilization of vacant, underused, or deteriorated parcels within existing urban fabrics, has emerged as a critical paradigm. This approach not only curbs urban sprawl and preserves natural hinterlands but also revitalizes distressed areas, potentially enhancing urban quality of life and mitigating spatial inequalities.
However, identifying land suitable for infill development is inherently complex, characterized by multi-criteria and multi-layered decision-making processes. Influential criteria span physical (e.g., building density, structural age, land-use compatibility), accessibility (e.g., proximity to transport networks, service centers), socio-economic (e.g., social cohesion, population density, deprivation levels), and urban regulatory dimensions (e.g., legal constraints, height restrictions). Compounding this complexity is the pervasive uncertainty embedded within urban data systems, arising from spatial inaccuracies, heterogeneity in stakeholder opinions, and ambiguity in defining qualitative concepts such as "social cohesion." Traditional Type-1 Fuzzy Inference Systems (T1FIS), while valuable, often struggle with these compounded, multi-layered uncertainties, especially when precise membership functions for linguistic variables are difficult to define from inconsistent or scarce data.
To address these limitations, advanced computational models like Type-2 Fuzzy Inference Systems (T2FIS) have been developed. Specifically, Interval Type-2 Fuzzy Sets (IT2FS) introduce an additional dimension of flexibility by using footprint-of-uncertainty (FOU) in membership functions, enabling the simultaneous modeling of ambiguity in data, model parameters, and interpretive rules. This capability is particularly transformative for urban planning in contexts like Iran, where data scarcity and inconsistency are common. Therefore, this research presents a novel, integrated framework combining Geographic Information Systems (GIS) and an Interval Type-2 Fuzzy Inference System to zone land suitability for infill development in Ahvaz, Iran. The primary objective is to enhance zoning accuracy under ambiguous conditions by systematically managing data, model, and interpretative uncertainties.
 
 
Mothodology
This study employed a robust hybrid methodology integrating spatial analysis within ArcGIS 10.8 and fuzzy modeling in MATLAB R2018b. The study area encompassed the legal boundary of Ahvaz metropolis, covering approximately 18,006 hectares across eight urban districts.
Data Collection and Preparation: Four primary criteria clusters were established: 1) Physical (building density, structural age, land use), 2) Accessibility (distance from service centers, transport network density), 3) Social (social cohesion, population density), and 4) Urban Regulatory (legal constraints, height regulations). Data were sourced from municipal organizations, the Ahvaz comprehensive plan, and the Iranian Statistical Center. Qualitative data were transformed into quantitative measures using fuzzy scales. All spatial layers were integrated, normalized to a [0,1] interval, and rasterized to ensure compatibility with the fuzzy model.
Fuzzy System Design: An Interval Type-2 Fuzzy Inference System was designed to manage multi-layered uncertainties. Input variables were modeled using Gaussian and Triangular Interval Type-2 Membership Functions. For each variable, five linguistic values—Very Low, Low, Medium, High, Very High—were defined. The parameters of these membership functions were optimized using a fuzzy clustering algorithm. A knowledge-based rule repository of 81 IF-THEN rules was constructed, synthesizing expert opinion and prior literature. The Mamdani fuzzy inference mechanism was employed.
Inference and Defuzzification: The system processed the integrated criteria through its rule base. To generate a crisp output, a two-step process was implemented: first, type-reduction was performed using the Karnik-Mendel (KM) algorithm to convert the Interval Type-2 fuzzy output sets into Type-1 fuzzy sets; second, defuzzification was applied using the centroid method to produce a final suitability score between 0 and 100 for each raster pixel (10m x 10m).
Validation: The model's validity was assessed through a tripartite strategy: 1) Sensitivity Analysis to evaluate output stability against input parameter variations, 2) Field Validation to compare results with identified vacant/underused lands in approved urban plans, and 3) Comparative Analysis against classical methods (like standard AHP and Type-1 FIS) using quantitative error metrics such as Root Mean Square Error (RMSE) and the Kappa coefficient.
 
Findings
The application of the T2FIS model yielded a detailed suitability map for infill development across Ahvaz, classified into five potential classes.
Spatial Distribution of Potential: The analysis revealed a nuanced distribution of suitability. Only a combined 20% of Ahvaz's total area (3,639 hectares) exhibited Moderate potential for infill development. Higher suitability classes were more limited, with 13.7% (2,433 ha) classified as High and merely 6.3% (1,099 ha) as Very High potential. Consequently, the majority of the city's land (59% or 10,796 hectares) fell into the Low and Very Low suitability categories, indicating significant constraints for inward development under the defined criteria.
Identification of Priority Zones: The spatial analysis identified distinct geographic patterns. Districts 4, 6, and 8 emerged as the most suitable zones, concentrating the largest contiguous areas of High and Very High potential. This was attributed to a confluence of favorable factors: better access to public transportation networks (particularly in District 4), balanced population densities conducive to redevelopment (District 6), and the availability of legally developable land (District 8). In contrast, the northern and northeastern sectors of the city, along with the historical core (District 1), showed predominantly Low to Very Low potential, constrained by factors like regulatory limitations, physical deterioration, or poor accessibility.
Model Performance: A key finding was the enhanced performance of the proposed T2FIS model. Comparative evaluation demonstrated that the integrated framework, by managing compound uncertainties, improved zoning accuracy by approximately 40% compared to classical Type-1 FIS and Analytic Hierarchy Process (AHP) methods under the same ambiguous data conditions. The model successfully handled heterogeneous data inputs (raster, vector, and qualitative survey data) and inconsistencies within the expert-derived rule base.
 
Discussion and Conclusion
This study demonstrates the significant efficacy of integrating Interval Type-2 Fuzzy Logic with GIS for complex urban land suitability analysis. The primary innovation lies not merely in the application of an advanced fuzzy system, but in its deliberate design to address the multi-layered uncertainties endemic to urban planning in data-scarce environments. By employing IT2FS, the model accommodated ambiguity in input data (e.g., measuring "social cohesion"), model parameters (e.g., defining membership function bounds), and rule interpretation, moving beyond the limitations of simpler deterministic or Type-1 fuzzy models.