A sub-neighborhood scale land use regression model for predicting NO(2)

Analysis of spatial distribution characteristics and main influencing factors of heavy metals in road dust of Tianjin based on land use regression models

Land use regression (LUR) models are mainly used for the simulation and prediction of conventional atmospheric pollutants. Whether the LUR models can be expanded to study more toxic and hazardous pollutants (such as heavy metals) remains to be verified. Combined with the factors of road, land use type, population, pollution enterprise, meteorology, and terrain, the LUR models were used to simulate the spatial distribution characteristics of heavy metals in road dust and determine the main influencing factors. Samples of road surface dust were collected from 144 evenly distributed points in Tianjin, China, with 108 modelling points and 36 verification points. The R² values of the LUR models of Cd, Cr, Cu, Ni, and Pb contents were 0.301, 0.412, 0.399, 0.496, and 0.377, and their error rates were 2.72%, 4.96%, 4.64%, 8.91%, and 4.94%, respectively. The error rates of the kriging interpolation models were 3.33%, 6.50%, 5.14%, 18.30%, and 22.87%, which were all greater than those of the LUR models. The estimation effect of the LUR models was more refined than that of the kriging interpolation models. The contents of most heavy metals (except Ni) in road dust of the central area in Tianjin were generally higher than those of the surrounding areas. The heavy metal contents in road dust of Tianjin were mainly affected by road variables and meteorological variables. The LUR models were suitable for small-scale spatial prediction of heavy metals in urban road dust within urban areas.

Predicting intraurban airborne PM1.0-trace elements in a port city: Land use regression by ordinary least squares and a machine learning algorithm

Airborne particulate matter (PM) has been associated with cardiovascular and respiratory morbidity and mortality, and there is some evidence that spatially varying metals found in PM may contribute to adverse health effects. We developed spatially refined models for PM trace elements using ordinary least squares land use regression (OLS-LUR) and machine leaning random forest land-use regression (RF-LUR). Two-week integrated measurements of PM_1.0 (median aerodiameter < 1.0 μm) were collected at 50 sampling sites during fall (2010), winter (2011), and summer (2011) in the Halifax Regional Municipality, Nova Scotia, Canada. PM_1.0 filters were analyzed for metals and trace elements using inductively coupled plasma-mass spectrometry. OLS- and RF-LUR models were developed for approximately 30 PM_1.0 trace elements in each season. Model predictors included industrial, commercial, and institutional/ government/ military land use, roadways, shipping, other transportation sources, and wind rose information. RF generated more accurate models than OLS for most trace elements based on 5-fold cross validation. On average, summer models had the highest cross validation R² (OLS-LUR = 0.40, RF-LUR = 0.46), while fall had the lowest (OLS-LUR = 0.27, RF-LUR = 0.31). Many OLS-LUR models displayed overprediction in the final exposure surface. In contrast, RF-LUR models did not exhibit overpredictions. Taking overpredictions and cross validation performances into account, OLS-LUR performed better than RF-LUR in roughly 20% of the seasonal trace element models. RF-LUR models provided more interpretable predictors in most cases. Seasonal predictors varied, likely due to differences in seasonal distribution of trace elements related to source activity, and meteorology.

Statistical Approaches for Forecasting Primary Air Pollutants: A Review

Air pollutant forecasting can be used to quantitatively estimate pollutant reduction trends. Combining bibliometrics with the evolutionary tree and Markov chain methods can achieve a superior quantitative analysis of research hotspots and trends. In this work, we adopted a bibliometric method to review the research status of statistical prediction methods for air pollution, used evolutionary trees to analyze the development trend of such research, and applied the Markov chain to predict future research trends for major air pollutants. The results indicate that papers mainly focused on the effects of air pollution on human diseases, urban pollution exposure models, and land use regression (LUR) methods. Particulate matter (PM), nitrogen oxides (NOx), and ozone (O3) were the most investigated pollutants. Artificial neural network (ANN) methods were preferred in studies of PM and O3, while LUR were more widely used in studies of NOx. Additionally, multi-method hybrid techniques gradually became the most widely used approach between 2010 and 2018. In the future, the statistical prediction of air pollution is expected to be based on a mixed method to simultaneously predict multiple pollutants, and the interaction between pollutants will be the most challenging aspect of research on air pollution prediction. The research results summarized in this paper provide technical support for the accurate prediction of atmospheric pollution and the emergency management of regional air quality.

Improving spatial prediction of health risk assessment for Hg, As, Cu, and Pb in soil based on land-use regression

Heavy-metal pollution is a significant health and environmental concern in areas of rapid industrialization in China. The accuracy of spatial mapping of pollutant is the main constraint on spatial prediction of health risks. Our study addressed the possibility of improving spatial prediction accuracy of risk assessment. We developed land-use regression (LUR) models for Hg, As, Cu, and Pb based on surface soil sampling, land-use data, and soil properties. The regression results suggested that LUR was more accurate than ordinary kriging method. Spatial prediction accuracy of Hg, As, Cu, and Pb were improved by 15%, 59%, 36%, and 20%, respectively. Then, spatial distribution of health risk was assessed by using distributions of heavy metal and exposure parameters. Chronic risk of children was controlled by distribution of Pb and carcinogenic controlled by As. The result indicated that Pb and As were the main sources of health risk for children in Kunshan. Chronic and carcinogenic risk maps could clearly show where we should pay attention to and control the risk. This study provided a simple approach to draw spatially explicit maps of health risk which were useful for pollution control and public health risk management.

Assessing NO2 Concentration and Model Uncertainty with High Spatiotemporal Resolution across the Contiguous United States Using Ensemble Model Averaging

NO₂ is a combustion byproduct that has been associated with multiple adverse health outcomes. To assess NO₂ levels with high accuracy, we propose the use of an ensemble model to integrate multiple machine learning algorithms, including neural network, random forest, and gradient boosting, with a variety of predictor variables, including chemical transport models. This NO₂ model covers the entire contiguous U.S. with daily predictions on 1-km-level grid cells from 2000 to 2016. The ensemble produced a cross-validated R² of 0.788 overall, a spatial R² of 0.844, and a temporal R² of 0.729. The relationship between daily monitored and predicted NO₂ is almost linear. We also estimated the associated monthly uncertainty level for the predictions and address-specific NO₂ levels. This NO₂ estimation has a very high spatiotemporal resolution and allows the examination of the health effects of NO₂ in unmonitored areas. We found the highest NO₂ levels along highways and in cities. We also observed that nationwide NO₂ levels declined in early years and stagnated after 2007, in contrast to the trend at monitoring sites in urban areas, where the decline continued. Our research indicates that the integration of different predictor variables and fitting algorithms can achieve an improved air pollution modeling framework.

Can dispersion modeling of air pollution be improved by land-use regression? An example from Stockholm, Sweden

Both dispersion modeling (DM) and land-use regression modeling (LUR) are often used for assessment of long-term air pollution exposure in epidemiological studies, but seldom in combination. We developed a hybrid DM-LUR model using 93 biweekly observations of NO_x at 31 sites in greater Stockholm (Sweden). The DM was based on spatially resolved topographic, physiographic and emission data, and hourly meteorological data from a diagnostic wind model. Other data were from land use, meteorology and routine monitoring of NO_x. We built a linear regression model for NO_x, using a stepwise forward selection of covariates. The resulting model predicted observed NO_x (R²=0.89) better than the DM without covariates (R²=0.68, P-interaction <0.001) and with minimal apparent bias. The model included (in descending order of importance) DM, traffic intensity on the nearest street, population (number of inhabitants) within 100 m radius, global radiation (direct sunlight plus diffuse or scattered light) and urban contribution to NO_x levels (routine urban NO_x, less routine rural NO_x). Our results indicate that there is a potential for improving estimates of air pollutant concentrations based on DM, by incorporating further spatial characteristics of the immediate surroundings, possibly accounting for imperfections in the emission data.

Assessing the Potential of Land Use Modification to Mitigate Ambient NO₂ and Its Consequences for Respiratory Health

Understanding how local land use and land cover (LULC) shapes intra-urban concentrations of atmospheric pollutants-and thus human health-is a key component in designing healthier cities. Here, NO₂ is modeled based on spatially dense summer and winter NO₂ observations in Portland-Hillsboro-Vancouver (USA), and the spatial variation of NO₂ with LULC investigated using random forest, an ensemble data learning technique. The NO2 random forest model, together with BenMAP, is further used to develop a better understanding of the relationship among LULC, ambient NO₂ and respiratory health. The impact of land use modifications on ambient NO₂, and consequently on respiratory health, is also investigated using a sensitivity analysis. We find that NO₂ associated with roadways and tree-canopied areas may be affecting annual incidence rates of asthma exacerbation in 4-12 year olds by +3000 per 100,000 and -1400 per 100,000, respectively. Our model shows that increasing local tree canopy by 5% may reduce local incidences rates of asthma exacerbation by 6%, indicating that targeted local tree-planting efforts may have a substantial impact on reducing city-wide incidence of respiratory distress. Our findings demonstrate the utility of random forest modeling in evaluating LULC modifications for enhanced respiratory health.

Development, Evaluation, and Comparison of Land Use Regression Modeling Methods to Estimate Residential Exposure to Nitrogen Dioxide in a Cohort Study

We used a network of 135 NO₂ passive diffusion tube sites to develop land use regression (LUR) models in a UK conurbation. Network sites were divided into four groups (32-35 sites per group) and models developed using combinations of 1-3 groups of "training" sites to evaluate how the number of training sites influenced model performance and residential NO₂ exposure estimates for a cohort of 13 679 participants. All models explained moderate to high variance in training and independent "hold-out" data (Training adj. R²: 62-89%; Hold-out R²: 44-85%). Average hold-out R² increased by 9.5%, while average training adj. R² decreased by 7.2% when the number of training groups was increased from 1 to 3. Exposure estimate precision improved with increasing number of training sites (median intralocation relative standard deviations of 19.2, 10.3, and 7.7% for 1-group, 2-group and 3-group models respectively). Independent 1-group models gave highly variable exposure estimates suggesting that variations in LUR sampling networks with relatively low numbers of sites (≤35) may substantially alter exposure estimates. Collectively, our analyses suggest that use of more than 60 training sites has quantifiable benefits in epidemiological application of LUR models.

A land use regression application into assessing spatial variation of intra-urban fine particulate matter (PM2.5) and nitrogen dioxide (NO2) concentrations in City of Shanghai, China

Intra-urban assessment of air pollution exposure has become a priority study while international attention was attracted to PM2.5 pollution in China in recent years. Land Use Regression (LUR), which has previously been proved to be a feasible way to describe the relationship between land use and air pollution level in European and American cities, was employed in this paper to explain the correlations and spatial variations in Shanghai, China. PM2.5 and NO2 concentrations at 35-45 monitoring locations were selected as dependent variables, and a total of 44 built environmental factors were extracted as independent variables. Only five factors showed significant explanatory value for both PM2.5 and NO2 models: longitude, distance from monitors to the ocean, highway intensity, waterbody area, and industrial land area for PM2.5 model; residential area, distance to the coast, industrial area, urban district, and highway intensity for NO2 model. Respectively, both PM2.5 and NO2 showed anti-correlation with coastal proximity (an indicator of clean air dilution) and correlation with highway and industrial intensity (source indicators). NO2 also showed significant correlation with local indicators of population density (residential intensity and urban classification), while PM2.5 showed significant correlation with regional dilution (longitude as a indicator of distance from polluted neighbors and local water features). Both adjusted R squared values were strong with PM2.5 (0.88) being higher than NO2 (0.62). The LUR was then used to produce continuous concentration fields for NO2 and PM2.5 to illustrate the features and, potentially, for use by future studies. Comparison to PM2.5 studies in New York and Beijing show that Shanghai PM2.5 pollutant distribution was more sensitive to geographic location and proximity to neighboring regions.

Integrating High-Resolution Datasets to Target Mitigation Efforts for Improving Air Quality and Public Health in Urban Neighborhoods

Reducing exposure to degraded air quality is essential for building healthy cities. Although air quality and population vary at fine spatial scales, current regulatory and public health frameworks assess human exposures using county- or city-scales. We build on a spatial analysis technique, dasymetric mapping, for allocating urban populations that, together with emerging fine-scale measurements of air pollution, addresses three objectives: (1) evaluate the role of spatial scale in estimating exposure; (2) identify urban communities that are disproportionately burdened by poor air quality; and (3) estimate reduction in mobile sources of pollutants due to local tree-planting efforts using nitrogen dioxide. Our results show a maximum value of 197% difference between cadastrally-informed dasymetric system (CIDS) and standard estimations of population exposure to degraded air quality for small spatial extent analyses, and a lack of substantial difference for large spatial extent analyses. These results provide the foundation for improving policies for managing air quality, and targeting mitigation efforts to address challenges of environmental justice.

Neighborhood-Scale Spatial Models of Diesel Exhaust Concentration Profile Using 1-Nitropyrene and Other Nitroarenes

With emerging evidence that diesel exhaust exposure poses distinct risks to human health, the need for fine-scale models of diesel exhaust pollutants is growing. We modeled the spatial distribution of several nitrated polycyclic aromatic hydrocarbons (NPAHs) to identify fine-scale gradients in diesel exhaust pollution in two Seattle, WA neighborhoods. Our modeling approach fused land-use regression, meteorological dispersion modeling, and pollutant monitoring from both fixed and mobile platforms. We applied these modeling techniques to concentrations of 1-nitropyrene (1-NP), a highly specific diesel exhaust marker, at the neighborhood scale. We developed models of two additional nitroarenes present in secondary organic aerosol: 2-nitropyrene and 2-nitrofluoranthene. Summer predictors of 1-NP, including distance to railroad, truck emissions, and mobile black carbon measurements, showed a greater specificity to diesel sources than predictors of other NPAHs. Winter sampling results did not yield stable models, likely due to regional mixing of pollutants in turbulent weather conditions. The model of summer 1-NP had an R(2) of 0.87 and cross-validated R(2) of 0.73. The synthesis of high-density sampling and hybrid modeling was successful in predicting diesel exhaust pollution at a very fine scale and identifying clear gradients in NPAH concentrations within urban neighborhoods.

Accounting for spatial effects in land use regression for urban air pollution modeling

In order to accurately assess air pollution risks, health studies require spatially resolved pollution concentrations. Land-use regression (LUR) models estimate ambient concentrations at a fine spatial scale. However, spatial effects such as spatial non-stationarity and spatial autocorrelation can reduce the accuracy of LUR estimates by increasing regression errors and uncertainty; and statistical methods for resolving these effects--e.g., spatially autoregressive (SAR) and geographically weighted regression (GWR) models--may be difficult to apply simultaneously. We used an alternate approach to address spatial non-stationarity and spatial autocorrelation in LUR models for nitrogen dioxide. Traditional models were re-specified to include a variable capturing wind speed and direction, and re-fit as GWR models. Mean R(2) values for the resulting GWR-wind models (summer: 0.86, winter: 0.73) showed a 10-20% improvement over traditional LUR models. GWR-wind models effectively addressed both spatial effects and produced meaningful predictive models. These results suggest a useful method for improving spatially explicit models.

Assessing the relationship among urban trees, nitrogen dioxide, and respiratory health

Modeled atmospheric pollution removal by trees based on eddy flux, leaf, and chamber studies of relatively few species may not scale up to adequately assess landscape-level air pollution effects of the urban forest. A land use regression (LUR) model (R(2) = 0.70) based on NO2 measured at 144 sites in Portland, Oregon (USA), after controlling for roads, railroads, and elevation, estimated every 10 ha (20%) of tree canopy within 400 m of a site was associated with a 0.57 ppb decrease in NO2. Using BenMAP and a 200 m resolution NO2 model, we estimated that the NO2 reduction associated with trees in Portland could result in significantly fewer incidences of respiratory problems, providing a $7 million USD benefit annually. These in-situ urban measurements predict a significantly higher reduction of NO2 by urban trees than do existing models. Further studies are needed to maximize the potential of urban trees in improving air quality.

An LUR/BME framework to estimate PM2.5 explained by on road mobile and stationary sources

Knowledge of particulate matter concentrations <2.5 μm in diameter (PM2.5) across the United States is limited due to sparse monitoring across space and time. Epidemiological studies need accurate exposure estimates in order to properly investigate potential morbidity and mortality. Previous works have used geostatistics and land use regression (LUR) separately to quantify exposure. This work combines both methods by incorporating a large area variability LUR model that accounts for on road mobile emissions and stationary source emissions along with data that take into account incompleteness of PM2.5 monitors into the modern geostatistical Bayesian Maximum Entropy (BME) framework to estimate PM2.5 across the United States from 1999 to 2009. A cross-validation was done to determine the improvement of the estimate due to the LUR incorporation into BME. These results were applied to known diseases to determine predicted mortality coming from total PM2.5 as well as PM2.5 explained by major contributing sources. This method showed a mean squared error reduction of over 21.89% oversimple kriging. PM2.5 explained by on road mobile emissions and stationary emissions contributed to nearly 568,090 and 306,316 deaths, respectively, across the United States from 1999 to 2007.

Role of highway traffic on spatial and temporal distributions of air pollutants in a Swiss Alpine valley

Traffic-related air pollutants show high spatial variability near roads, posing a challenge to adequately assess exposures. Recent modeling approaches (e.g. dispersion models, land-use regression (LUR) models) have addressed this but mostly in urban areas where traffic is abundant. In contrast, our study area was located in a rural Swiss Alpine valley crossed by the main North-south transit highway of Switzerland. We conducted an extensive measurement campaign collecting continuous nitrogen dioxide (NO₂), particulate number concentrations (PN), daily respirable particulate matter (PM10), elemental carbon (EC) and organic carbon (OC) at one background, one highway and seven mobile stations from November 2007 to June 2009. Using these measurements, we built a hybrid model to predict daily outdoor NO₂ concentrations at residences of children participating in an asthma panel study. With the exception of OC, daily variations of the pollutants followed the temporal trends of heavy-duty traffic counts on the highway. In contrast, variations of weekly/seasonal means were strongly determined by meteorological conditions, e.g., winter inversion episodes. For pollutants related to primary exhaust emissions (i.e. NO₂, EC and PN) local spatial variation strongly depended on proximity to the highway. Pollutant concentrations decayed to background levels within 150 to 200 m from the highway. Two separate daily NO₂ prediction models were built using LUR approaches with (a) short-term traffic and weather data (model 1) and (b) subsequent addition of daily background NO₂ to previous model (model 2). Models 1 and 2 explained 70% and 91% of the variability in outdoor NO₂ concentrations, respectively. The biweekly averaged predictions from the final model 2 agreed very well with the independent biweekly integrated passive measurements taken at thirteen homes and nine community sites (validation R(2)=0.74). The excellent spatio-temporal performance of our model provides a very promising basis for the health effect assessment of the panel study.

A land use regression model for ultrafine particles in Vancouver, Canada

Methods to characterize chronic exposure to ultrafine particles (UFP) can help to clarify potential health effects. Since UFP are not routinely monitored in North America, spatiotemporal models are one potential exposure assessment methodology. Portable condensation particle counters were used to measure particle number concentrations (PNC) to develop a land use regression (LUR) model. PNC, wind speed and direction were measured for sixty minutes at eighty locations during a two-week sampling campaign. We conducted continuous monitoring at four additional locations to assess temporal variation. LUR modeling utilized 135 potential geographic predictors including: road length, vehicle density, restaurant density, population density, land use and others. A novel approach incorporated meteorological data through wind roses as alternates to traditional circular buffers. The range of measured (sixty-minute median) PNC across locations varied seventy-fold (1500-105000 particles/cm(3), mean [SD] = 18200 [15900] particles/cm(3)). Correlations between PNC and concurrently measured two-week average NOX concentrations were 0.6-0.7. A PNC LUR model (R(2) = 0.48, leave-one-out cross validation R(2) = 0.32) including truck route length within 50 m, restaurant density within 200 m, and ln-distance to the port represents the first UFP LUR model in North America. Models incorporating wind roses did not explain more variability in measured PNC.

Assessing the spatial distribution of nitrogen dioxide in London, Ontario

Land use regression (LUR) models have been widely used to characterize the spatial distribution of urban air pollution and estimate exposure in epidemiologic studies. However, spatial patterns of air pollution vary greatly between cities due to local source type and distribution. London, Ontario, Canada, is a medium-sized city with relatively few and isolated industrial point sources, which allowed the study to focus on the contribution of different transportation sectors to urban air pollution. This study used LUR models to estimate the spatial distribution of nitrogen dioxide (NO2) and to identify local sources influencing NO2 concentrations in London, ON. Passive air sampling was conducted at 50 locations throughout London over a 2-week period in May-June 2010. NO2 concentrations at the monitored locations ranged from 2.8 to 8.9 ppb, with a median of 5.2 ppb. Industrial land use, dwelling density, distance to highway, traffic density, and length of railways were significant predictors of NO2 concentrations in the final LUR model, which explained 78% of NO2 variability in London. Traffic and dwelling density explained most of the variation in NO2 concentrations, which is consistent with LUR models developed in other Canadian cities. We also observed the importance of local characteristics. Specifically, 17% of the variation was explained by distance to highways, which included the impacts of heavily traveled corridors transecting the southern periphery of the city. Two large railway yards and railway lines throughout central areas of the city explained 9% of NO2 variability. These results confirm the importance of traditional LUR variables and highlight the importance of including a broader array of local sources in LUR modeling. Finally, future analyses will use the model developed in this study to investigate the association between ambient air pollution and cardiovascular disease outcomes, including plaque burden, cholesterol, and hypertension.

A land use regression model incorporating data on industrial point source pollution

Advancing the understanding of the spatial aspects of air pollution in the city regional environment is an area where improved methods can be of great benefit to exposure assessment and policy support. We created land use regression (LUR) models for SO2, NO2 and PM10 for Tianjin, China. Traffic volumes, road networks, land use data, population density, meteorological conditions, physical conditions and satellite-derived greenness, brightness and wetness were used for predicting SO2, NO2 and PM10 concentrations. We incorporated data on industrial point sources to improve LUR model performance. In order to consider the impact of different sources, we calculated the PSIndex, LSIndex and area of different land use types (agricultural land, industrial land, commercial land, residential land, green space and water area) within different buffer radii (1 to 20 km). This method makes up for the lack of consideration of source impact based on the LUR model. Remote sensing-derived variables were significantly correlated with gaseous pollutant concentrations such as SO2 and NO2. R2 values of the multiple linear regression equations for SO2, NO2 and PM10 were 0.78, 0.89 and 0.84, respectively, and the RMSE values were 0.32, 0.18 and 0.21, respectively. Model predictions at validation monitoring sites went well with predictions generally within 15% of measured values. Compared to the relationship between dependent variables and simple variables (such as traffic variables or meteorological condition variables), the relationship between dependent variables and integrated variables was more consistent with a linear relationship. Such integration has a discernable influence on both the overall model prediction and health effects assessment on the spatial distribution of air pollution in the city region.

Improving spatial concentration estimates for nitrogen oxides using a hybrid meteorological dispersion/land use regression model in Los Angeles, CA and Seattle, WA

Predictions from a simple line source dispersion model, Caline3, were included as a covariate in a land use regression (LUR) model for NO(X)/NO(2) in Los Angeles, CA and Seattle, WA. The Caline3 model prediction assumed a unit emission factor for all roadway segments (1.0g/vehicle-mile). The NO(X) and/or NO(2) measurements for LA and Seattle were obtained from a comprehensive measurement campaign that is part of the Multi-Ethnic Study of Atherosclerosis Air Pollution Study (MESA Air). The measurement campaigns in both cities were approximately 2weeks in duration employing approximately 145 measurement sites in Greater LA and 26 sites in Seattle. The best "standard" LUR model (obtained without the inclusion of the Caline3 predictions) in LA had R(2) values of 0.53 for NO(X) and 0.74 for NO(2). The leave-one-out cross-validated R(2) values for NO(X) and NO(2) were 0.45 and 0.71, respectively. The equivalent "standard" NO(2) model for Seattle had an R(2) of 0.72 and a leave-one-out cross-validated R(2) of 0.63. When the Caline3 variable was included in the LA hybrid model, the R(2) values were 0.71 and 0.79 for NO(X) and NO(2), respectively. The corresponding cross-validated R(2) values were 0.66 and 0.77, for NOX and NO2, respectively. In Seattle, the inclusion of the Caline3 variable resulted in a NO(2) model with an R(2) of 0.81 and a corresponding cross-validated R(2) of 0.67. In LA, hybrid model performance was not affected by excluding roadways with annual average daily traffic volumes (AADT)<100,000. When the Caline3 predictions for heavy-duty trucks and lighter-duty vehicles were modelled as separate terms, the estimated fleet average NO(X) emission factors were 8.9 (SE=0.7) and 0.16 (SE=0.12) grams NO(X)/vehicle mile for heavy-duty and lighter-duty vehicles, respectively. These values are consistent with fleet average emission factors computed for LA with EMFAC 2007.

A land use regression for predicting NO2 and PM10 concentrations in different seasons in Tianjin region, China

Land use regression (LUR) model was employed to predict the spatial concentration distribution of NO2 and PM10 in the Tianjin region based on the environmental air quality monitoring data. Four multiple linear regression (MLR) equations were established based on the most significant variables for NO2 in heating season (R2 = 0.74), and non-heating season (R2 = 0.61) in the whole study area; and PM10 in heating season (R2 = 0.72), and non-heating season (R2 = 0.49). Maps of spatial concentration distribution for NO2 and PM10 were obtained based on the MLR equations (resolution is 10 km). Intercepts of MLR equations were 0.050 (NO2, heating season), 0.035 (NO2, non-heating season), 0.068 (PM10, heating season), and 0.092 (PM10, non-heating season) in the whole study area. In the central area of Tianjin region, the intercepts were 0.042 (NO2, heating season), 0.043 (NO2, non-heating season), 0.087 (PM10, heating season), and 0.096 (PM10, non-heating season). These intercept values might imply an area's background concentrations. Predicted result derived from LUR model in the central area was better than that in the whole study area. R2 values increased 0.09 (heating season) and 0.18 (non-heating season) for NO2, and 0.08 (heating season) and 0.04 (non-heating season) for PM10. In terms of R2, LUR model performed more effectively in heating season than non-heating season in the study area and gave a better result for NO2 compared with PM10.

Mobile monitoring of particle light absorption coefficient in an urban area as a basis for land use regression

Land use regression (LUR) is used to map spatial variability in air pollutant concentrations for risk assessment epidemiology, and air quality management. Conventional LUR requires long-term measurements at multiple locations, so application to particulate matter has been limited. Here we use mobile monitoring to characterize spatial variability in black carbon concentrations for LUR modeling. A particle soot absorption photometer in a moving vehicle was used to measure the absorption coefficient (sigma(ap)) during summertime periods of peak afternoon traffic at 39 locations. LUR was used to model the mean and 25th, 50th, 75th, and 90th percentile values of the distribution of 10 s measurements at each location. Model performance (measured by R2) was higher for the 25th and 50th percentiles (0.72 and 0.68, respectively) than for the mean, 75th and 90th percentiles (0.51, 0.55, and 0.54, respectively). Performance was similar to that reported for conventional LUR models of NO2 and NO in this region (116 sites) and better than that for mean sigma(ap) from fixed-location samplers (25 sites). Models of the mean, 75th, and 90th percentiles favored predictors describing truck, rather than total, traffic. This approach is applicable to other urban areas to facilitate the development of LUR models for particulate matter.

A distance-decay variable selection strategy for land use regression modeling of ambient air pollution exposures

Land use regression (LUR) has emerged as an effective and economical means of estimating air pollution exposures for epidemiological studies. To date, no systematic method has been developed for optimizing the variable selection process. Traditionally, a limited number of buffer distances assumed having the highest correlations with measured pollutant concentrations are used in the manual stepwise selection process or a model transferred from another urban area. In this paper we propose a novel and systematic way of modeling long-term average air pollutant concentrations through "A Distance Decay REgression Selection Strategy" (ADDRESS). The selection process includes multiple steps and, at each step, a full spectrum of correlation coefficients and buffer distance decay curves are used to select a spatial covariate of the highest correlation (compared to other variables) at its optimized buffer distance. At the first step, the series of distance decay curves is constructed using the measured concentrations against the chosen spatial covariates. A variable with the highest correlation to pollutant levels at its optimized buffer distance is chosen as the first predictor of the LUR model from all the distance decay curves. Starting from the second step, the prediction residuals are used to construct new series of distance decay curves and the variable of the highest correlation at its optimized buffer distance is chosen to be added to the model. This process continues until a variable being added does not contribute significantly (p>0.10) to the model performance. The distance decay curve yields a visualization of change and trend of correlation between the spatial covariates and air pollution concentrations or their prediction residuals, providing a transparent and efficient means of selecting optimized buffer distances. Empirical comparisons suggested that the ADDRESS method produced better results than a manual stepwise selection process of limited buffer distances. The method also enables researchers to understand the likely scale of variables that influence pollution levels, which has potentially important ramifications for planning and epidemiological studies.