LUR Models Research Articles

Development of Land Use Regression models for predicting exposure to NO2 and NOx in Metropolitan Perth, Western Australia

This study developed LUR models for predicting exposure to NO2 and NOx among of 12,203 elderly men in Perth. NOx and NO2 concentrations were determined for 2-week periods in summer, autumn, and winter, from January to September 2012, at 43 sites. The LUR models were developed to predict annual average concentrations of nitric oxides based upon land use, population/household density, and traffic variables within different buffer sizes, following the procedures of the European Study of Cohort for Air Pollution Effects program. The sample mean and standard deviation of the annual average concentrations of NO2 and NOx were 10.1 ± 5.3 μg/m3 and 18.7 ± 11.7 μg/m3 respectively, lower than those of ESCAPE study areas. The LUR models explained 69% of the variance in NO2 and 75% variance of NOx. Both the NO2 and NOx models had similar predictors, including traffic intensity on the nearest roads, household density within-1000 m industrial activities within-5000 m, and road length within-50 m.

Intra-urban variation of ultrafine particles as evaluated by process related land use and pollutant driven regression modelling

The microscale intra-urban variation of ultrafine particle concentrations (UFP, diameter Dp<100nm) and particle number size distributions was studied by two statistical regression approaches. The models were applied to a 1km2 study area in Braunschweig, Germany. A land use regression model (LUR) using different urban morphology parameters as input is compared to a multiple regression type model driven by pollutant and meteorological parameters (PDR). While the LUR model was trained with UFP concentration the PDR model was trained with measured particle number size distribution data. The UFP concentration was then calculated from the modelled size distributions. Both statistical approaches include explanatory variables that try to address the ‘process chain’ of particle emission, dilution and deposition.LUR explained 74% and 85% of the variance of UFP for the full data set with a root mean square error (RMSE) of 668cm−3 and 1639cm−3 in summer and winter, respectively. PDR explained 56% and 74% of the variance with RMSE of 4066cm−3 and 6030cm−3 in summer and winter, respectively. Both models are capable to depict the spatial variation of UFP across the study area and in different outdoor microenvironments. The deviation from measured UFP concentrations is smaller in the LUR model than in PDR.The PDR model is well suited to predict urban particle number size distributions from the explanatory variables (total particle number concentration, black carbon and wind speed). The urban morphology parameters in the LUR model are able to resolve size dependent concentration variations but not as adequately as PDR.

Investigating the role of transportation models in epidemiologic studies of traffic related air pollution and health effects

In two earlier case–control studies conducted in Montreal, nitrogen dioxide (NO2), a marker for traffic-related air pollution was found to be associated with the incidence of postmenopausal breast cancer and prostate cancer. These studies relied on a land use regression model (LUR) for NO2 that is commonly used in epidemiologic studies for deriving estimates of traffic-related air pollution. Here, we investigate the use of a transportation model developed during the summer season to generate a measure of traffic emissions as an alternative to the LUR model. Our traffic model provides estimates of emissions of nitrogen oxides (NOx) at the level of individual roads, as does the LUR model. Our main objective was to compare the distribution of the spatial estimates of NOx computed from our transportation model to the distribution obtained from the LUR model. A secondary objective was to compare estimates of risk using these two exposure estimates. We observed that the correlation (spearman) between our two measures of exposure (NO2 and NOx) ranged from less than 0.3 to more than 0.9 across Montreal neighborhoods. The most important factor affecting the “agreement” between the two measures in a specific area was found to be the length of roads. Areas affected by a high level of traffic-related air pollution had a far better agreement between the two exposure measures. A comparison of odds ratios (ORs) obtained from NO2 and NOx used in two case–control studies of breast and prostate cancer, showed that the differences between the ORs associated with NO2 exposure vs NOx exposure differed by 5.2–8.8%.

Comparing land use regression and dispersion modelling to assess residential exposure to ambient air pollution for epidemiological studies

BackgroundLand-use regression (LUR) and dispersion models (DM) are commonly used for estimating individual air pollution exposure in population studies. Few comparisons have however been made of the performance of these methods. ObjectivesWithin the European Study of Cohorts for Air Pollution Effects (ESCAPE) we explored the differences between LUR and DM estimates for NO2, PM10 and PM2.5. MethodsThe ESCAPE study developed LUR models for outdoor air pollution levels based on a harmonised monitoring campaign. In thirteen ESCAPE study areas we further applied dispersion models. We compared LUR and DM estimates at the residential addresses of participants in 13 cohorts for NO2; 7 for PM10 and 4 for PM2.5. Additionally, we compared the DM estimates with measured concentrations at the 20–40 ESCAPE monitoring sites in each area. ResultsThe median Pearson R (range) correlation coefficients between LUR and DM estimates for the annual average concentrations of NO2, PM10 and PM2.5 were 0.75 (0.19–0.89), 0.39 (0.23–0.66) and 0.29 (0.22–0.81) for 112,971 (13 study areas), 69,591 (7) and 28,519 (4) addresses respectively. The median Pearson R correlation coefficients (range) between DM estimates and ESCAPE measurements were of 0.74 (0.09–0.86) for NO2; 0.58 (0.36–0.88) for PM10 and 0.58 (0.39–0.66) for PM2.5. ConclusionsLUR and dispersion model estimates correlated on average well for NO2 but only moderately for PM10 and PM2.5, with large variability across areas. DM predicted a moderate to large proportion of the measured variation for NO2 but less for PM10 and PM2.5.

The association of LUR modeled PM2.5 elemental composition with personal exposure

Background and aimsLand use regression (LUR) models predict spatial variation of ambient concentrations, but little is known about the validity in predicting personal exposures. In this study, the association of LUR modeled concentrations of PM2.5 components with measured personal concentrations was determined. The elements of interest were copper (Cu), iron (Fe), potassium (K), nickel (Ni), sulfur (S), silicon (Si), vanadium (V) and zinc (Zn). MethodsIn Helsinki (Finland), Utrecht (the Netherlands) and Barcelona (Spain) five participants from urban background, five from suburban background and five from busy street sites were selected in each city (15 participants per city). Outdoor, indoor and personal 96-hour PM2.5 samples were collected by the participants over periods of two weeks in three different seasons (winter, summer and spring/autumn) and the overall average was calculated. Elemental composition was measured by ED-XRF spectrometry. The LUR models for the average ambient concentrations of each element were developed by the ESCAPE project. ResultsLUR models predicted the within-city variation of average outdoor Cu and Fe concentrations moderately well (range in R2 27–67% for Cu and 24–54% for Fe). The outdoor concentrations of the other elements were not well predicted. The LUR modeled concentration only significantly correlated with measured personal Fe exposure in Utrecht and Ni and V in Helsinki. The LUR model predictions did not correlate with measured personal Cu exposure. After excluding observations with an indoor/outdoor ratio of >1.5, modeled Cu outdoor concentrations correlated with indoor concentrations in Helsinki and Utrecht and personal concentrations in Utrecht. The LUR model predictions were associated with measured outdoor, indoor and personal concentrations for all elements when the data for the three cities was pooled. ConclusionsWithin-city modeled variation of elemental composition of PM2.5 did not predict measured variation in personal exposure well.

Assessing ozone exposure for epidemiological studies in Malmö and Umeå, Sweden

Ground level ozone [ozone] is considered a harmful air pollutant but there is a knowledge gap regarding its long term health effects. The main aim of this study is to develop local Land Use Regression [LUR] models that can be used to study long term health effects of ozone. The specific aim is to develop spatial LUR models for two Swedish cities, Umeå and Malmö, as well as a temporal model for Malmö in order to assess ozone exposure for long term epidemiological studies. For the spatial model we measured ozone, using Ogawa passive samplers, as weekly averages at 40 sites in each study area, during three seasons. This data was then inserted in the LUR-model with data on traffic, land use, population density and altitude to develop explanatory models of ozone variation. To develop the temporal model for Malmö, hourly ozone data was aggregated into daily means for two measurement stations in Malmö and one in a rural area outside Malmö. Using regression analyses we inserted meteorological variables into different temporal models and the one that performed best for all three stations was chosen. For Malmö the LUR-model had an adjusted model R2 of 0.40 and cross validation R2 of 0.17. For Umeå the model had an adjusted model R2 of 0.67 and cross validation adjusted R2 of 0.48. When restricting the model to only including measuring sites from urban areas, the Malmö model had adjusted model R2 of 0.51 (cross validation adjusted R2 0.33) and the Umeå model had adjusted model R2 of 0.81 (validation adjusted R2 of 0.73). The temporal model had adjusted model R2 0.54 and 0.61 for the two Malmö sites, the cross validation adjusted R2 was 0.42. In conclusion, we can with moderate accuracy, at least for Umeå, predict the spatial variability, and in Malmö the temporal variability in ozone variation.

OXIDATIVE POTENTIAL IN 10 EUROPEAN AREAS OF THE ESCAPE STUDY: SPATIAL VARIATION AND LAND USE REGRESSION MODELS

Background:TRANSPHORM and ESCAPE are EU FP7 funded projects providing knowledge on the impact of air pollutionon human health in Europe by adding air pollution exposure assessment to health data available from European cohort studies. Oxidative potential is increasingly used as metric related to PM adverse health effects considering PM chemical complexity. Aim: The goal of the present study was assessment of PM2.5 oxidative potential levels in 10 European study areas and development of land use regression models (LUR) in order to assess exposure of cohort studies participants to oxidative potential. Methods: Oxidative potential (OP) was measured with the DTT assay in: The Netherlands, Oslo, Barcelona, Munich, Copenhagen, Helsinki, Athens, Paris, Rome, in addition to NOx, PM2.5 and PM10 (mass, absorbance), PAH’s, EC/OC and metals. Three two-week samples were taken at 20–40 street , urban and regional background sites in each study areas. LUR models were developed based on annual average concentrations and various traffic and land-use related GIS-predictor variables. Results: OP level were highest in Athens (0.28nmolDTT/min*m3) and Rome, Paris, Barcelona (0.23). The lowest OP was in Oslo, London, Helsinki – 0.13, 0.14, 0.15nmolDTT/min*m3). The median ratio of at street locations compared to urban background was 1.11. The median regional to urban background OP ratio was 1.04. OP levels during different seasons differed significantly in 9 study areas. The median cold to warm season OP ratio was 1.51. Spatial OP correlations with other components varied across areas, the most consistent were found in Athens (PM2.5, PM2.5abs, OC, Zn, Ni, S). In 6 locations LUR models could be developed with the highest model explained variance (R2) in The Netherlands – 66%. The median R2 was 25%. Cross validation R2 was on average 10% lower. Conclusions: Oxidative potential levels differed across Europe. Relatively low R2 of LUR models can be explained by lack of identified sources of oxidative potential among variables used for models development.

A Hybrid Approach to Estimating National Scale Spatiotemporal Variability of PM2.5in the Contiguous United States

Airborne fine particulate matter exhibits spatiotemporal variability at multiple scales, which presents challenges to estimating exposures for health effects assessment. Here we created a model to predict ambient particulate matter less than 2.5 μm in aerodynamic diameter (PM2.5) across the contiguous United States to be applied to health effects modeling. We developed a hybrid approach combining a land use regression model (LUR) selected with a machine learning method, and Bayesian Maximum Entropy (BME) interpolation of the LUR space-time residuals. The PM2.5 data set included 104,172 monthly observations at 1464 monitoring locations with approximately 10% of locations reserved for cross-validation. LUR models were based on remote sensing estimates of PM2.5, land use and traffic indicators. Normalized cross-validated R(2) values for LUR were 0.63 and 0.11 with and without remote sensing, respectively, suggesting remote sensing is a strong predictor of ground-level concentrations. In the models including the BME interpolation of the residuals, cross-validated R(2) were 0.79 for both configurations; the model without remotely sensed data described more fine-scale variation than the model including remote sensing. Our results suggest that our modeling framework can predict ground-level concentrations of PM2.5 at multiple scales over the contiguous U.S.

Evaluation of Land Use Regression Models for NO2 and Particulate Matter in 20 European Study Areas: The ESCAPE Project

Land use regression models (LUR) frequently use leave-one-out-cross-validation (LOOCV) to assess model fit, but recent studies suggested that this may overestimate predictive ability in independent data sets. Our aim was to evaluate LUR models for nitrogen dioxide (NO2) and particulate matter (PM) components exploiting the high correlation between concentrations of PM metrics and NO2. LUR models have been developed for NO2, PM2.5 absorbance, and copper (Cu) in PM10 based on 20 sites in each of the 20 study areas of the ESCAPE project. Models were evaluated with LOOCV and "hold-out evaluation (HEV)" using the correlation of predicted NO2 or PM concentrations with measured NO2 concentrations at the 20 additional NO2 sites in each area. For NO2, PM2.5 absorbance and PM10 Cu, the median LOOCV R(2)s were 0.83, 0.81, and 0.76 whereas the median HEV R(2) were 0.52, 0.44, and 0.40. There was a positive association between the LOOCV R(2) and HEV R(2) for PM2.5 absorbance and PM10 Cu. Our results confirm that the predictive ability of LUR models based on relatively small training sets is overestimated by the LOOCV R(2)s. Nevertheless, in most areas LUR models still explained a substantial fraction of the variation of concentrations measured at independent sites.

Temporal stability of land use regression models for traffic-related air pollution

BackgroundLand-use regression (LUR) is a cost-effective approach for predicting spatial variability in ambient air pollutant concentrations with high resolution. Models have been widely used in epidemiological studies and are often applied to time periods before or after the period of air quality monitoring used in model development. However, it is unclear how well such models perform when extrapolated over time. ObjectiveThe objective of this study was to assess the temporal stability of LUR models over a period of 7 years in Metro Vancouver, Canada. MethodsA set of NO and NO2 LUR models based on 116 measurements were developed in 2003. In 2010, we made 116 measurements again, of which 73 were made at the exact same location as in 2003. We then developed 2010 models using updated data for the same predictor variables used in 2003, and also explored additional variables. Four methods were used to derive model predictions over 7 years, and predictions were compared with measurements to assess the temporal stability of LUR models. ResultsThe correlation between 2003 NO and 2010 NO measurements was 0.87 with a mean (sd) decrease of 11.3 (9.9) ppb. For NO2, the correlation was 0.74, with a mean (sd) decrease of 2.4 (3.2) ppb. 2003 and 2010 LUR models explained similar amounts of spatial variation (R2 = 0.59 and R2 = 0.58 for NO; R2 = 0.52 and R2 = 0.63 for NO2, in 2003 and in 2010 respectively). The 2003 models explained more variability in the 2010 measurements (R2 = 0.58–0.60 for NO; R2 = 0.52–0.61 for NO2) than the 2010 models explained in the 2003 measurements (R2 = 0.50–0.55 for NO; R2 = 0.44–0.49 for NO2), and the 2003 models explained as much variability in the 2010 measurements as they did in the 2003 measurements. ConclusionLUR models are able to provide reliable estimates over a period of 7 years in Metro Vancouver. When concentrations and their variability are decreasing over time, the predictive power of LUR models is likely to remain the same or to improve in forecasting scenarios, but to decrease in hind-casting scenarios.

Nitrogen dioxide levels estimated from land use regression models several years apart and association with mortality in a large cohort study

BackgroundLand Use Regression models (LUR) are useful to estimate the spatial variability of air pollution in urban areas. Few studies have evaluated the stability of spatial contrasts in outdoor nitrogen dioxide (NO2) concentration over several years. We aimed to compare measured and estimated NO2 levels 12 years apart, the stability of the exposure estimates for members of a large cohort study, and the association of the exposure estimates with natural mortality within the cohort.MethodsWe measured NO2 at 67 locations in Rome in 1995/96 and 78 sites in 2007, over three one-week-long periods. To develop LUR models, several land-use and traffic variables were used. NO2 concentration at each residential address was estimated for a cohort of 684,000 adults. We used Cox regression to analyze the association between the two estimated exposures and mortality.ResultsThe mean NO2 measured concentrations were 45.4 μg/m3 (SD 6.9) in 1995/96 and 44.6 μg/m3 (SD 11.0) in 2007, respectively. The correlation of the two measurements was 0.79. The LUR models resulted in adjusted R2 of 0.737 and 0.704, respectively. The correlation of the predicted exposure values for cohort members was 0.96. The association of each 10 μg/m3 increase in NO2 with mortality was 6 % for 1995/96 and 4 % for 2007 LUR models. The increased risk per an inter-quartile range change was identical (4 %, 95 % CI:3–6 %) for both estimates of NO2.ConclusionsMeasured and predicted NO2 values from LUR models, from samples collected 12 years apart, had good agreement, and the exposure estimates were similarly associated with mortality in a large cohort study.

Redistribution of Traffic Related Air Pollution Associated with a New Road Tunnel

The aim of this study was to assess the effect of a new road tunnel on the concentration and distribution of traffic-related air pollution (TRAP), specifically nitrogen dioxide (NO(2)) and particulate matter (PM), and to determine its relationship to change in traffic flow. We used continuously recorded data from four monitoring stations at nonroadside locations within the study area and three regional monitors outside the area. The four monitors in the study area were in background locations where smaller pollutant changes were expected compared with changes near the bypassed main road. We also deployed passive samplers to assess finer spatial variability in NO(2) including application of a land use regression model (LUR). The study was conducted from 2006 to 2008. Analysis of the continuously recorded data showed that the tunnel intervention did not lead to consistent reductions in NO(2) or PM over the wider study area. However, there were significant decreases in NO(2), NO(x), and PM(10) in the eastern section of the study area. Analysis of passive sampler data indicated that the greatest reductions in NO(2) concentrations occurred within 100 m of the bypassed main road. The LUR model also demonstrated that changes in NO(2) were most marked adjacent to the bypassed main road. These findings support the use of methods that highlight fine spatial variability in TRAP and demonstrate the utility of traffic interventions in reducing air pollution exposures for populations living close to main roads.

Long-term exposure models for traffic related NO2 across geographically diverse areas over separate years

Although recent air pollution epidemiologic studies have embraced land-use regression models for estimating outdoor traffic exposure, few have examined the spatio-temporal variability of traffic related pollution over a long term period and the optimal methods to take these factors into account for exposure estimates. We used home outdoor NO 2 measurements taken from eight geographically diverse areas to examine spatio-temporal variations, construct, and evaluate models that could best predict the within-city contrasts in observations. Passive NO 2 measurements were taken outside of up to 100 residences per area over three seasons in 1993 and 2003 as part of the Swiss cohort study on air pollution and lung and heart disease in adults (SAPALDIA). The spatio-temporal variation of NO 2 differed by area and year. Regression models constructed using the annual NO 2 means from central monitoring stations and geographic parameters predicted home outdoor NO 2 levels better than a dispersion model. However, both the regression and dispersion models underestimated the within-city contrasts of NO 2 levels. Our results indicated that the best models should be constructed for individual areas and years, and would use the dispersion estimates as the urban background, geographic information system (GIS) parameters to enhance local characteristics, and temporal and meteorological variables to capture changing local dynamics. Such models would be powerful tools for assessing health effects from long-term exposure to air pollution in a large cohort. ► Long-term exposure models developed for 8 different areas in Switzerland. ► Unique comparison of LUR models using seasonal NO 2 measurements made a decade apart. ► Best LUR models used land-use, dispersion, meteorological, and temporal predictors. ► Individual local area models by year best captured NO 2 spatial distribution. ► Implications on the generalizability of LUR models for air pollution health studies.

LAND USE REGRESSION MODELS OF PM 10 AND NO 2 CONCENTRATIONS IN TAIPEI

Background and Aims: This study developed land use regression models to assess long-term exposure to ambient particulate matters of PM10 and NO2, in order to minimize the exposure misclassification and achieve better individual exposure assessment for the Taiwan cohorts of the ESCAPE study. Methods: We used a population density based sampling strategy to select 40 locations as our NO2 measurement sampling sites in Taipei, while PM10 measurements were conducted at 20 sites among them, and one additional background site for annual average adjustment. PM10 and NO2 concentrations were monitored simultaneously by using Harvard Impactors and Ogawa passive samplers respectively for 2-week period in three seasons during 2009 and 2010. With collected land-use datasets, containing land-cover and traffic related data, we conducted the GIS analyses to obtain a series of potential predictor variables in concentric circle buffers with several radii ranging from 25 to 5000m of each monitoring site. Based on stochastic modeling techniques, we combined the average concentrations and values of predictor variables in stepwise procedure to develop the final LUR models for PM10 and NO2. Results: The adjusted annual concentrations were 48.6±5.9 •g/m3 for PM10 and 26.0±6.5 ppb for NO2. There were 3 predictors for the final model of PM10 (adjusted R2=0.78), including major road length within 25m, industry areas within 1000m, and high density residential areas within 100m. And there were 4 predictors for the NO2 model (adjusted R2=0.66), including low density residential areas within 500m, major road length within 25m, urban green areas within 300m, semi-natural and forested areas within 500m. Conclusions: Our study indicates that exposure to PM10 and NO2 for subjects in Taiwan cohorts of the ESCAPE study can be well estimated by LUR models with 3-4 predictors.

AIR QUALITY CHANGES AFTER COMMISSIONING OF AN URBAN ROAD TUNNEL

Background &amp; Aims: The commissioning of a new road tunnel in Sydney, Australia, in March 2007 provided an opportunity to assess the effect of this intervention on air pollutant concentrations in the affected area. Dispersion model-based predictions were that pollutant concentrations would decrease around the main bypassed road where traffic flow was halved and would increase along feeder roads to the tunnel. Methods: PM10, PM2.5 and NO2 data from six continuous monitoring stations erected in the study area were used and compared with regional changes in air quality. We also collected data on NO2, as a marker of traffic related air pollution, using a network of passive samplers. Spatial variability was assessed by land use regression. The study period was Jan 2006 to Dec 2008. Results: Data from the continuous monitors showed that, after adjusting for regional changes in air quality, the opening of the tunnel did not lead to consistent reductions in either PM or NO2 in the study area as a whole. However, the passive sampler data showed that, while NO2 levels were not significantly reduced in the first post-tunnel year (-2.1%, 95% CI -5.9% to +1.8%), they were significantly reduced in the second post-tunnel year (-16.1%, 95% CI -19.2% to -12.7%). Further, passive sampler NO2 concentrations were reduced in both post-tunnel years along the major by-passed road. The LUR model also demonstrated reductions of NO2 (up to 10 ppb) along the by-passed road and an increase in NO2 of up to 5 ppb along the feeder roads. Conclusion: This study highlights the need to study fine scale spatial variability when assessing the impact of traffic interventions and supports the effectiveness of traffic interventions in reducing exposures for people living close to main traffic routes.

Stability of measured and modelled spatial contrasts in NO2 over time

ObjectivesLand use regression (LUR) modelling is a popular method to estimate outdoor air pollution concentrations at the home and/or work addresses of individual subjects in epidemiological studies. Typically, such models...

Nitrogen Dioxide Spatial Variability in Rome (Italy): An Application of the LUR Model Over a Decade

Nitrogen Dioxide Spatial Variability in Rome (Italy): An Application of the LUR Model Over a Decade