A Knowledge-Guided Approach For Urban Growth Mapping Using Deep Learning

Abstract

The increase in urbanization has emerged as one of the critical issues of sustainable development, especially in the fast growing metropolis of the developing world. This paper demonstrates a knowledge-based deep learning method to map and study urban growth in Hillah city, Babil Governorate, Iraq, in a 20-year span (20042024). Integration of multisource geospatial data was performed, namely Landsat imagery (2004, 2014, 2024), Land Surface Temperature (LST) created by the MODIS, Digital Elevation Model (SRTM) and spectral indexes such as Normalized Difference Vegetation Index (NDVI) and Normalized Difference Built-up Index (NDBI). The Support Vector Machine (SVM) and deep learning classifiers were used to detect the land use and land cover (LULC) change, and the accuracy was over 0.8 according to the Kappa coefficient of all years. According to the findings, the built-up area will increase drastically (16.59 km2 in 2004 to 45.88 km2 in 2024) at the cost of vegetation cover and an increase of surface temperatures, a phenomenon that indicates the intensification of urban heat island effect. The findings show that knowledge-based indexation is efficient in deep learning systems of urban monitoring. This research methodology can deliver practical information to the urban planners and policy makers in order to render a sustainable development and preservation strategy feasible to other highly populated areas.