School of Forestry and Environmental Studies, Yale University, New Haven, CT, USA.
Department of Environmental Health, Boston University School of Public Health, Boston, MA 02118, USA.
Environ Res. 2017 May;155:42-48. doi: 10.1016/j.envres.2017.01.038. Epub 2017 Feb 9.
With growing urbanization, traffic has become one of the main sources of air pollution in Nepal. Understanding the impact of air pollution on health requires estimation of exposure. Land use regression (LUR) modeling is widely used to investigate intraurban variation in air pollution for Western cities, but LUR models are relatively scarce in developing countries. In this study, we developed LUR models to characterize intraurban variation of nitrogen dioxide (NO) in urban areas of Kathmandu Valley, Nepal, one of the fastest urbanizing areas in South Asia.
Over the study area, 135 monitoring sites were selected using stratified random sampling based on building density and road density along with purposeful sampling. In 2014, four sampling campaigns were performed, one per season, for two weeks each. NO was measured using duplicate Palmes tubes at 135 sites, with additional information on nitric oxide (NO), NO, and nitrogen oxide (NOx) concentrations derived from Ogawa badges at 28 sites. Geographical variables (e.g., road network, land use, built area) were used as predictor variables in LUR modeling, considering buffers 25-400m around each monitoring site.
Annual average NO by site ranged from 5.7 to 120ppb for the study area, with higher concentrations in the Village Development Committees (VDCs) of Kathmandu and Lalitpur than in Kirtipur, Thimi, and Bhaktapur, and with variability present within each VDC. In the final LUR model, length of major road, built area, and industrial area were positively associated with NO concentration while normalized difference vegetation index (NDVI) was negatively associated with NO concentration (R=0.51). Cross-validation of the results confirmed the reliability of the model.
The combination of passive NO sampling and LUR modeling techniques allowed for characterization of nitrogen dioxide patterns in a developing country setting, demonstrating spatial variability and high pollution levels.
随着城市化进程的推进，交通已成为尼泊尔空气污染的主要来源之一。了解空气污染对健康的影响需要对暴露情况进行评估。基于地理位置的回归（LUR）模型广泛应用于西方城市的城市内空气污染研究，但在发展中国家，此类模型相对较少。本研究旨在建立 LUR 模型，以描述尼泊尔加德满都谷地（南亚城市化速度最快的地区之一）城市内二氧化氮（NO ）的空间变化。
在研究区域内，采用分层随机抽样方法，根据建筑密度和道路密度，结合有目的的抽样，共选择了 135 个监测点。2014 年，每个季节进行了 4 次采样，每次持续两周。在 135 个监测点使用帕尔默管（Palmes tubes）进行重复测量，在 28 个监测点使用奥加瓦（Ogawa） badges 收集一氧化氮（NO）、二氧化氮（NO ）和氮氧化物（NOx）的浓度数据。LUR 模型的预测变量包括地理变量（如道路网络、土地利用、建筑区域），并考虑了每个监测点周围 25-400m 的缓冲区。
研究区域内各监测点的年平均 NO 浓度范围为 5.7 至 120ppb，其中加德满都和勒利德布尔的村发展委员会（Village Development Committees，VDCs）的浓度高于其他三个区，且同一 VDC 内的浓度也存在差异。在最终的 LUR 模型中，主要道路长度、建筑区域和工业区域与 NO 浓度呈正相关，归一化植被差异指数（Normalized Difference Vegetation Index，NDVI）与 NO 浓度呈负相关（R=0.51）。结果的交叉验证证实了模型的可靠性。
本研究采用被动式 NO 采样和 LUR 模型技术，可在发展中国家环境中对二氧化氮的分布特征进行描述，结果表明该地区的空间变异性较大且污染水平较高。
