City-level vulnerability to COVID pandemic: Spatio-temporal analysis of land use and land cover type effect on COVID spread

Abstract

In late 2019, Wuhan reported the initial cases of COVID-19, and within a short period, it was declared a global pandemic. The pandemic had a profound impact globally, particularly affecting developing countries. This research employs spatio-temporal techniques to analyze the impact of land use factors on COVID-19 spread in Islamabad and to predict hotspots based on these factors. The land use factors considered include hospitals, pharmacies, local bus stops, metro stations, and supermarkets. Spatial autocorrelation and Hotspot Analysis were used to identify disease clusters, followed by Pearson’s correlation analysis to determine the influence of selected factors. Subsequently, machine learning techniques were applied to predict hotspots. Local bus stops emerged as the most significant factor contributing to the virus’s spread. Among the five classification models, K-Nearest Neighbour (KNN) demonstrated the best performance for hotspot prediction, with an accuracy of 92.9%, precision of 86.5%, recall of 98.3%, F1-score of 92.0%, and an AUC value of 0.989. This research provides valuable insights for policymakers, aiding in the identification of problem areas and optimizing resource allocation to address similar viral outbreaks in the future.