Analysis of spatial and time differences in real estatevalues in the context of earthquake risk: A study based on the GTWR model

Accurate and reliable valuation of real estate is essential not only for individual investment decisions but also for the stability and sustainability of national economies. The value of a property is shaped by a multidimensional interaction that includes its structural characteristics, locational features, environmental attributes, and various economic indicators. In addition, extraordinary events such as natural disasters, pandemics, or socio-political crises can cause abrupt shifts in real estate markets. In this context, the present study focuses on Onikişubat District of Kahramanmaraş, one of the areas most severely affected by the earthquake that struck the region on February 6, 2023.To examine value changes before and after the disaster, real estate sales data from 2022, 2023, and 2025 were collected and integrated from various property listing platforms, resulting in a dataset consisting of approximately 2,704 residential units. These data were transferred into a Geographic Information System environment, where coordinate information was processed and distance based spatial variables such as proximity to transportation networks, public facilities, social amenities, and geological risk zones were calculated. Additionally, approximately 31,000 cadastral parcels obtained from open-source platforms such as Parsel Sorgu and OpenStreetMap were incorporated to construct a comprehensive spatial base for the study area. To model and predict property values, three analytical approaches were applied: Ordinary Least Squares (OLS), Geographically Weighted Regression (GWR), and Geographically and Temporally Weighted Regression (GTWR). A comparative assessment of these models revealed that classical regression is insufficient in regions with pronounced spatial and temporal heterogeneity. While the GWR model better captured spatial variability, the GTWR model demonstrated superior predictive performance by integrating both spatial and temporal dimensions. The models were evaluated using performance metrics such as R², RMSE, MAE, and MAPE, and their strengths and limitations were discussed in detail.This study contributes to the literature by revealing the spatial patterns of value changes following the earthquake, their relationship with risk zones, and the regional differentiation of factors influencing property values. The findings also provide valuable insights for local authorities, planners, and investors by supporting more informed and spatially sensitive decision-making processes.