Spatial distribution of ultrafine particles in urban settings: A land use regression model

详细信息	查看全文 | 推荐本文 |

• 作者：Marcela Rivera^a ; ^b ; ^c ; ^{arivera@creal.cat} ; ^{marcelariverav@gmail.com} ; Xavier Basaga&ntilde ; a^a ; ^b ; Inmaculada Aguilera^a ; ^b ; ^d ; David Agis^a ; ^b ; Laura Bouso^a ; ^b ; Maria Foraster^a ; ^b ; ^c ; ^d ; Mercedes Medina-Ramó ; n^a ; Jorge Pey^e ; Nino Kü ; nzli^f ; ^g ; Gerard Hoek^h
• 关键词：Particle number concentration ; PM0.1 ; LUR ; Exposure assessment ; Monitoring ; Outdoor home ; Traffic ; Girona ; Spain
• 刊名：Atmospheric Environment
• 出版年：2012
• 期刊代码：11_13522310
• 类别：env
• 出版时间：July, 2012
• 卷：54
• 期：Complete
• 页码：657-666
• 文件大小：644 K

摘要

Background

The toxic effects of ultrafine particles (UFP) are a public health concern. However, epidemiological studies on the long term effects of UFP are limited due to lacking exposure models. Given the high spatial variation of UFP, the assignment of exposure levels in epidemiological studies requires a fine spatial scale. The aim of this study was to assess the performance of a short-term measurement protocol used at a large number of locations to derive a land use regression (LUR) model of the spatial variation of UFP in Girona, Spain.

Methods

We measured UFP for 15 min on the sidewalk of 644 participants鈥?homes in 12 towns of Girona province (Spain). The measurements were done during non-rush traffic hours 9:15-12:45 and 15:15-16:45 during 32 days between June 15 and July 31, 2009. In parallel, we counted the number of vehicles driving in both directions. Measurements were repeated on a different day for a subset of 25 sites in Girona city. Potential predictor variables such as building density, distance to bus lines and land cover were derived using geographic information systems. We adjusted for temporal variation using daily mean NO_x concentrations at a central monitor. Land use regression models for the entire area (Core model) and for individual towns were derived using a supervised forward selection algorithm.

Results

The best predictors of UFP were traffic intensity, distance to nearest major crossroad, area of high density residential land and household density. The LUR Core model explained 36%of UFP total variation. Adding sampling date and hour of the day to the Core model increased the R² to 51%without changing the regression slopes. Local models included predictor variables similar to those in the Core model, but performed better with an R² of 50%in Girona city.

Independent LUR models for the first and second measurements at the subset of sites with repetitions had R²鈥檚 of about 47%. When the mean of the two measurements was used R² improved to 72%.

Conclusions

LUR models for UFP were developed, based on a highly cost-effective short-term monitoring campaign at a large number of sites, with fair performance. Complementing the approach with further strategies to address sources of temporal variation of UFP is likely to result in improved models as indicated by the good performance of the model based on the subset of sites with one repeated measurement. Our approach is promising for UFP and possibly for other PM components requiring active sampling.