A Smart Framework for Zoning the Spatial Capacity of Urban Physical Development by Integrating GIS and Machine Learning: A Case Study of Tabriz
A Smart Framework for Zoning the Spatial Capacity of Urban Physical Development by Integrating GIS and Machine Learning: A Case Study of Tabriz
sajad amanpour1 yaghoub kohansal2
1) Master s Student in Urban Planning, University of Guilan, Rasht, Iran -
2) Master s Student in Urban Planning, University of Guilan, Rasht, Iran -
IOI Code
:
ABKD-ACAFDB
Publication :
Fourth International Conference on applied researches in civil engineering, architecture and urban planning (icarcau.com)
Abstract :
Any determination of the spatial capacity for physical urban development is a multi-criteria problem in which the suitability of an area for development depends on a set of natural, environmental, accessibility, infrastructural, and spatial-structure conditions. In conventional approaches, geographic information systems and multi-criteria evaluation have been used to convert these factors into spatial layers, standardize them, and produce suitability maps. Methods such as the analytic hierarchy process have also been used to determine criterion weights; however, their dependence on expert judgment and the sensitivity of results to the selection of criteria and weights are among the issues discussed in the literature (Malczewski, 1999; Zhao et al., 2025). In recent years, machine learning has gradually entered this field and has been used to extract criterion importance, model nonlinear relationships, and generate spatial suitability levels (Zhao et al., 2024).
The present study, using a review–analytical approach, examines the evolution of methods for assessing urban development suitability and spatial capacity, from GIS-MCE and traditional weighting methods to machine learning and explainable artificial intelligence. Domestic and international studies are compared in terms of research problem, data type, spatial criteria, algorithm used, validation method, and interpretability of results. At the international level, recent studies have pursued the use of Random Forest, XGBoost, and other machine learning algorithms to assess development suitability and extract variable importance. In addition, spatial validation and Explainable AI methods have been considered in more recent studies as responses to two issues: overestimation of model performance and the black-box nature of some algorithms (Nguyen et al., 2026; Wang et al., 2025). In Iran, studies related to Tehran, Tabriz, Urmia, and other cities show that machine learning and remote sensing have been used more in modeling and predicting urban growth and land-use change. Regarding Tabriz, previous studies have used Cellular Automata, regression, neural networks, and machine learning algorithms to analyze urban growth (Ashkriz et al., 2021; Mahmoudzadeh et al., 2022; Azizi et al., 2022).
The review findings indicate that the research gap cannot be sought merely in the lack of using artificial intelligence. The more important issue is how to combine definitive development constraints, influential spatial factors, empirical evidence derived from past development patterns, spatial validation, and model interpretability. Accordingly, the article proposes a four-layer framework for zoning the spatial capacity for development, in which definitive constraints are first separated from influential factors; then spatial factors are organized in a GIS environment and analyzed using machine learning; and the model output is also examined using interpretability methods. The city of Tabriz, as an example with a considerable research background in urban growth modeling, provides a suitable context for explaining the applicability of this framework. The main conclusion of the study is that the use of machine learning can assist spatial decision-making only when it is accompanied by a clear methodological structure, precise problem definition, control of spatial dependence, separation of constraints and factors, and interpretation of results.
Keywords : spatial development capacity physical development geographic information system (gis) multi criteria evaluation machine learning tabriz