Abstract In Low- and Middle-Income Countries, rapid urbanization has led to poorer air quality, yet pollution monitoring networks are often sparse or non-existent. Few previous studies have sought to understand the unique predictors of air pollution exposure in Indian urban environments. Our study monitored and modeled nitrogen dioxide (NO2) in Mysore, a rapidly urbanizing city in India. NO2 sampling was conducted in four seasonal campaigns (each lasting 2 weeks) in 2016–2017, at 150 sites throughout Mysore. Seasonal spatial interpolation of NO2 levels was conducted using 2 distinct models, the first utilizing a land use regression (LUR) approach and the second using universal kriging methods. Model performance was determined using adjusted R2, and validated using leave-one-out cross validation. Measured NO2 concentrations ranged from 0.3 to 51.9 ppb across the four seasons of the study period, with higher concentrations in the center of the city. In the LUR model (R2 = 0.535), proximity to major roads, point sources of pollution such as industrial sites and religious points of interest (PoI), land uses with high human activity, and high population density were associated with higher levels of NO2. Proximity to minor roads and coverage of land uses characterized by low human activity were inversely associated with air pollution. Cross-validation of results confirmed the reliability of each model. Few studies have applied spatially heterogeneous sampling to assess ambient air pollution levels in India. The combination of passive NO2 sampling and LUR/kriging modeling methods allowed for characterization of NO2 patterns in Mysore. While previous work indicates traffic pollution as a major contributor to ambient air pollution levels in urbanizing centers in Asia, our results indicate the influence of other pollution factors (e.g., point sources), as well as highly localized characteristics of the urban environment (e.g., proximity to religious points of interest) in urban India. Areas of Mysore consistently experienced pollution in excess of World Health Organization (WHO) health-protective guidelines for NO2.

中文翻译：

---

在印度迈索尔使用通用克里金法和土地利用回归模型对城市暴露进行低成本 NO2 监测和预测

摘要 在中低收入国家，快速的城市化导致空气质量下降，但污染监测网络往往稀少或不存在。以前的研究很少试图了解印度城市环境中空气污染暴露的独特预测因素。我们的研究对印度快速城市化城市迈索尔的二氧化氮 (NO2) 进行了监测和建模。NO2 采样是在 2016-2017 年的四个季节性活动（每个持续 2 周）中在整个迈索尔的 150 个地点进行的。NO2 水平的季节性空间插值使用 2 个不同的模型进行，第一个使用土地利用回归 (LUR) 方法，第二个使用通用克里金方法。使用调整后的 R2 确定模型性能，并使用留一法交叉验证进行验证。测得的 NO2 浓度范围为 0.3 到 51。整个研究期间的四个季节为 9 ppb，城市中心的浓度更高。在 LUR 模型 (R2 = 0.535) 中，靠近主要道路、工业场所和宗教景点 (PoI) 等污染点源、人类活动频繁的土地用途和高人口密度与较高水平的 NO2 相关. 靠近次要道路和以人类活动低为特征的土地利用覆盖率与空气污染呈负相关。结果的交叉验证证实了每个模型的可靠性。很少有研究应用空间异质采样来评估印度的环境空气污染水平。被动二氧化氮采样和 LUR/克里金建模方法相结合，可以表征迈索尔的二氧化氮模式。虽然之前的工作表明交通污染是亚洲城市化中心环境空气污染水平的主要贡献者，但我们的结果表明其他污染因素（例如点源）以及城市环境的高度局部特征（例如，靠近宗教景点）在印度城市。迈索尔地区的污染持续超过世界卫生组织 (WHO) 的 NO2 健康保护指南。