Transferability and Generalizability of Regression Models of Ultrafine Particles in Urban Neighborhoods in the Boston Area

Abstract

Land use regression (LUR) models have been used to assess air pollutant exposure, but limited evidence exists on whether location-specific LUR models are applicable to other locations (transferability) or general models are applicable to smaller areas (generalizability). We tested transferability and generalizability of spatial-temporal LUR models of hourly particle number concentration (PNC) for Boston-area (MA, U.S.A.) urban neighborhoods near Interstate 93. Four neighborhood-specific regression models and one Boston-area model were developed from mobile monitoring measurements (34–46 days/neighborhood over one year each). Transferability was tested by applying each neighborhood-specific model to the other neighborhoods; generalizability was tested by applying the Boston-area model to each neighborhood. Both the transferability and generalizability of models were tested with and without neighborhood-specific calibration. Important PNC predictors (adjusted-R2 = 0.24–0.43) included wind speed and direction, temperature, highway traffic volume, and distance from the highway edge. Direct model transferability was poor (R2 < 0.17). Locally-calibrated transferred models (R2 = 0.19–0.40) and the Boston-area model (adjusted-R2 = 0.26, range: 0.13–0.30) performed similarly to neighborhood-specific models; however, some coefficients of locally calibrated transferred models were uninterpretable. Our results show that transferability of neighborhood-specific LUR models of hourly PNC was limited, but that a general model performed acceptably in multiple areas when calibrated with local data.

Development and evaluation of land use regression models for black carbon based on bicycle and pedestrian measurements in the urban environment

“…These two studies also have difficulties to generalize the model to other parts of the same city. This evaluation of the transferability of the models is similar to the spatial cross-validation in the present study, although the neighbourhoods in Patton et al (2015) are not contiguous but around 3 to 12 km apart, and the routes in Hankey and Marshall (2015) also cover a larger area (about 8 x 12 km 2 ).…”

“…A possible reason they gave was that the range spanning the predictor variables within each of the routes was not fully balanced. In the study of Patton et al (2015), measurements were performed in four different neighbourhoods. When models built for one of the neighbourhoods were transferred to the other neighbourhoods, the models performed poorly (R 2 < 0.17, compared to R 2 of 0.23 to 0.42 for the neighbourhood-specific models).…”

Comparisons of traffic-related ultrafine particle number concentrations measured in two urban areas by central, residential, and mobile monitoring

“…Widely-differing monitoring networks have been used to model UFP, and characterize UFP in general, including long-term stationary monitoring (Aalto et al, 2005; Cyrys et al, 2008; Moore et al, 2009), mobile monitoring (Aggarwal et al, 2012; Li et al, 2013; Padró-Martínez et al, 2012; Patton et al, 2015; Steffens et al, 2017; Weichenthal et al., 2016; Zwack et al, 2011), monitoring at central sites and multiple short-term stationary sites (Abernethy et al, 2013; Eeftens et al, 2015; Fuller et al, 2012; Hofman et al, 2016; Klompmaker et al, 2015; Meier et al, 2015; Puustinen et al, 2007; Rivera et al, 2012; Wolf et al, 2017), or a combination of mobile and stationary monitoring (Hankey and Marshall, 2015; Kerckhoffs et al., 2016; Riley et al, 2016; Sabaliauskas et al, 2015) (Table S1). While Kerckhoffs et al (2016) observed modest correlations between on-road and nearby short-term stationary-site PNC, it remains unclear if these results can be generalized to other study areas and other platform comparisons or if use of a particular platform measures systematically different concentrations.…”

A High Resolution Spatiotemporal Model for In-Vehicle Black Carbon Exposure: Quantifying the In-Vehicle Exposure Reduction Due to the Euro 5 Particulate Matter Standard Legislation

“…The most important feature of the µLUR is that the models only use a single traffic covariate and not multiple partially correlating traffic covariates. In standard LUR practice, a large set of possible combinations of short and long distance variants of the traffic data are included, which results in a high risk of overfitting the sparse data points [40][41][42]. The high temporal resolution of the µLUR in combination with the non-linear modeling counteracts this by design.…”