Geoenvironment factors guided coastal urban growth prospect (UGP) delineation using heuristic and machine learning models

Abstract

Unplanned urban growth along the coastal cities may disturb the balance between the ecosystem and urbanisation; thus, a development strategy guided by geoenvironmental factors is necessary. We have attempted to delineate suitable Urban Growth Prospect (UGP) zones within the coastal region of the Vishakhapatnam district of India, based on geoenvironmental factors. We conduct an initial assessment of the temporal urban growth pattern in last 15 years using Landsat data and published landuse maps, demonstrating that the urban growth has percolated to unstable geomorphic units, making it vulnerable to flooding. Geoenvironmental factors (geomorphology, lithology, structure, land use, soil, slope and groundwater) based multi-thematic data integration approach is adopted to delineate UGP zones using heuristic and machine learning techniques like Analytical Hierarchy Process (AHP), Artificial Neural Network (ANN), Random Forest (RF), Logistic Regression (LR). We used present-day field observations to train the predictive models like ANN, LR & RF and test the outcomes with accuracy assessment ratios (detection percentage, miss factor etc.). The ANN-derived result presents the best accuracy with a detection percentage of 85% and predicts a 55% spatial scope of urban expansion, which was further refined using the mask of the strategic and ecologically sensitive location with appropriate buffers. The final map represents the UGP zones within the city and surroundings, encompassing stable landform units. The study provides insight into the usefulness of the heuristic and machine learning models in predictive analysis, using geoenvironmental factors for urban expansion in fast-growing coastal cities.