Independent Validation of National Satellite-Based Land-Use Regression Models for Nitrogen Dioxide Using Passive Samplers

Abstract

Including satellite observations of nitrogen dioxide (NO) in land-use regression (LUR) models can improve their predictive ability, but requires rigorous evaluation. We used 123 passive NO samplers sited to capture within-city and near-road variability in two Australian cities (Sydney and Perth) to assess the validity of annual mean NO estimates from existing national satellite-based LUR models (developed with 68 regulatory monitors). The samplers spanned roadside, urban near traffic (≤100 m to a major road), and urban background (>100 m to a major road) locations. We evaluated model performance using R (predicted NO regressed on independent measurements of NO), mean-square-error R (MSE-R), RMSE, and bias. Our models captured up to 69% of spatial variability in NO at urban near-traffic and urban background locations, and up to 58% of variability at all validation sites, including roadside locations. The absolute agreement of measurements and predictions (measured by MSE-R) was similar to their correlation (measured by R). Few previous studies have performed independent evaluations of national satellite-based LUR models, and there is little information on the performance of models developed with a small number of NO monitors. We have demonstrated that such models are a valid approach for estimating NO exposures in Australian cities.

The Urban Liveability Index: developing a policy-relevant urban liveability composite measure and evaluating associations with transport mode choice

“…The data that support the findings of this study are or were available from ABS [72–75, 78], PSMA [42], DSE [76, 82], VEAC [77], Knibbs [80], Acxiom [81], and ACECQA [83]. However, restrictions apply to the availability of some of these data, which were used under license for the current study [42, 81] or are historical and no longer hosted in the same form online by the source institutions [76, 77, 82, 83].…”

“…Hence, a sensitivity analysis was conducted to examine the impact including an air quality indicator in the ULI. The 2011 Melbourne segment of a national, longitudinal ambient air pollution dataset derived from a satellite-based land-use regression model was acquired featuring Mesh Block-level annual average prediction of nitrogen dioxide (NO 2 ) [79]; the surface model was validated to predict 69% of spatial variation in annual NO 2 (root mean square error 25%) [80]. Mesh Block-level modelled annual average NO 2 serves as a general proxy for traffic-related and other combustion-derived air pollutants.…”

Comparison of model estimates from an intra-city land use regression model with a national satellite-LUR and a regional Bayesian Maximum Entropy model, in estimating NO2 for a birth cohort in Sydney, Australia

“…Generalised estimating equation (GEE) models were used to develop annual models for 2006-2011, but for this study were updated with 2013 data (satellite and fixed monitor data). The Sat-LUR model was validated using passive sampler NO 2 data (including data from this study) and was found to predict 58% of variability across all sites and 69% variability at the urban near-traffic and background sites in two capital cities (Sydney and Perth) (Knibbs et al 2016).…”

Residential NO2 exposure is associated with urgent healthcare use in a thunderstorm asthma cohort

“…The LUR model estimates mean annual NO 2 levels from a combination of predictors including tropospheric NO 2 columns observed by satellite, land use and roads. This model has been both internally and externally validated across 68 and 123 different NO 2 monitoring sites, respectively, across all Australian states [ 24 25 ]. The model captures 66%–81% of spatial variability in annual NO 2 in urban locations, with a prediction error of 19%–25% [ 24 25 ].…”