Development of Land Use Regression (LUR) models and high-resolution spatial mapping of criteria air pollutants: Leveraging Delhi's continuous air monitoring network and remote sensing data

Abstract

High-resolution air pollution maps can help researchers and governments better characterize and mitigate pollution levels in a given region. Land Use Regression (LUR) modeling is a statistical approach capable of predicting pollution levels at a high spatial resolution. In this study, we used pollution data (for the calendar year 2019) from a dense (compared to other Indian regions) regulatory monitoring network in Delhi, India, to develop simple linear and interpretable LUR models for various criteria pollutants. The observed annual mean PM_2.5 and PM₁₀ over Delhi were found to be ∼110 μg m⁻³ and 220 μg m⁻³, respectively. The PM concentration levels were 2.5–4 times higher than the prescribed national ambient standards, while the gaseous criteria pollutants were found to be within the standards (over most of the study area). The performance of the developed LUR models ranged from poor to moderate levels (adjusted-R² values of the models were between 0.14 and 0.63). Land use and road-network related variables were found to be the most common predictors of the observed pollution levels. Moderately performing models (11 out of the developed 20) were then used to predict pollution levels at 50 m spatial intervals and to identify the most polluted districts. The advantages and limitations of using the existing regulatory network data for LUR development, and the other probable potential reasons responsible for the underperformance of the developed models are extensively discussed. To our knowledge, this is one of the few studies carried over Indian region to develop LUR models utilizing regulatory monitoring network data.