Abstract:
Road traffic accidents are a significant public safety concern, especially in urban areas where congestion and infrastructural limitations increase the likelihood of collisions. This study investigates road traffic accidents within the Kandy Police Division in Sri Lanka by applying geostatistical interpolation techniques to model the spatial distribution of 2,099 RTAs reported from January 2022 to March 2024.
The analysis evaluates the suitability of two interpolation methods: Inverse Distance Weighting (IDW), a deterministic approach, and Ordinary Kriging, a model-based geostatistical method that incorporates spatial autocorrelation. For the Kriging analysis, an empirical semivariogram was developed to quantify the spatial dependence structure of the accident data, and accident counts were log- transformed to approximate normality prior to spatial prediction with Ordinary Kriging. Four theoretical models, Spherical, Exponential, Gaussian, and Matérn, were fitted to the empirical semivariogram. The Spherical model outperformed the others, followed the empirical observed points closely, and reached the sill, with a sum of squared errors of 2.59 between empirical and fitted semivariograms. It was therefore selected as the best fit for spatial prediction using Kriging. Both interpolation methods were applied on a regular 10 × 10 grid across the study area to estimate accident frequencies. The performance of both methods was assessed using cross-validation, and predictive accuracy was evaluated through Mean Error (ME), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE).
Results showed that Ordinary Kriging provided slightly better predictive accuracy than IDW, with lower values for MAE (24.26 and 35.13), RMSE (40.04 and 52.87), and ME (-4.94 and -6.48), respectively. These findings reveal the effectiveness of geostatistical modeling, particularly Kriging, in identifying high-risk areas and supporting data-driven decision making. The findings of this study will support urban planners, traffic engineers, and policymakers in guiding targeted road safety measures, prioritizing infrastructure improvements, and effectively allocating resources in accident-prone zones.
