Comparison of Interpolation Techniques for Prolonged Exposure Estimation: A Case Study on Seven Years of Daily Nitrogen Oxide in Greater Sydney

Continuous exposure to air pollutants over a long period of time adversely affects population health. Addressing this issue may help in reducing the disease burden. Thus, it is crucial to understand the spatial and spatiotemporal variation in this prolonged exposure to ambient air pollutants to make informed decisions. The objective of this study is to evaluate the performance of most commonly used spatial interpolation techniques in sparsely located real-world sensor data for the purpose of estimating the prolonged exposure to air pollutants. The secondary data obtained from NSW Air Quality Monitoring Network (AQMN) sites within Greater Sydney during 1st January 2011 - 31st December 2017 by considering the daily concentrations of Nitrogen Oxide (NO) were used for this study. Nearest Neighbour (NN) interpolation, Inverse Distance Weighted (IDW) interpolation without search radius and with search radius (10 km, 15 km, 20 km, 25 km, 30 km, 35 km, 40 km, 45 km and 50 km) were used to estimate the daily concentrations at unknown locations. The performance of these interpolation techniques was assessed based on leave location-out cross-validation (LLO-CV) using Root Mean Square Error (RMSE), Index of Agreement (d) and Coefficient of Determination ( \(R^2\) ). Results revealed that, IDW with search radius of 25 km and power value of one performed better for the given dataset. IDW outperformed NN interpolation technique. These findings may help policy makers to come up with strategies for disease management, control and mitigation.