Satellite-Based NO₂ and Model Validation in a National Prediction Model Based on Universal Kriging and Land-Use Regression

Abstract

Epidemiological studies increasingly rely on exposure prediction models. Predictive performance of satellite data has not been evaluated in a combined land-use regression/spatial smoothing context. We performed regionalized national land-use regression with and without universal kriging on annual average NO2 measurements (1990–2012, contiguous U.S., EPA sites). Regression covariates were dimension-reduced components of 418 geographic variables including distance to roadway. We estimated model performance with two cross-validation approaches: using randomly selected groups and, in order to assess predictions to unmonitored areas, spatially clustered cross-validation groups. Ground-level NO2 was estimated from satellite-derived NO2 and was assessed as an additional regression covariate. Kriging models performed consistently better than non-kriging models. Among kriging models, conventional cross-validated R2 (R2cv) averaged over all years was 0.85 for the satellite data models and 0.84 for the models without satellite data. Average spatially clustered R2cvwas 0.74 for the satellite data models and 0.64 for the models without satellite data. The addition of either kriging or satellite data to a well-specified NO2 land-use regression model each improves prediction. Adding the satellite variable to a kriging model only marginally improves predictions in well-sampled areas (conventional cross-validation) but substantially improves predictions for points far from monitoring locations (clustered cross-validation).

The Ozone Monitoring Instrument: overview of 14 years in space

“…OMI NO 2 data have been used in a number of recent health studies (e.g., Hystad et al, 2011Hystad et al, , 2012Novotny et al, 2011;Prud'homme et al, 2013;Vienneau et al, 2013;Knibbs et al, 2014;Hoek et al, 2015;Belche et al, 2015;Crouse et al, 2015;de Hoogh et al, 2016;Young et al, 2016). For example, Belche et al (2013) found that annual OMI NO 2 column density data correlate well (r = 0.93) with surface data in southern California and provide a reliable measure of spatial variability for NO 2 exposure assessment.…”

High-resolution spatial distribution and associated uncertainties of greenhouse gas emissions from the agricultural sector

“…Kim and Dall'erba (2014) found a high spatial correlation of fossil fuel CO 2 emissions from crop production in the US; this might also apply to other GHG emissions in the agricultural sector. So, in advanced analysis, it is worth considering the correlation between some proxy data, for example, using tools of geostatistical modelling such as universal kriging (Young et al 2016) or autoregressive methods, and among them conditional (Horabik and Nahorski 2010) or spatial (Kim and Dall'erba 2014) autoregression models.…”

A high-resolution and observationally constrained OMI NO<sub>2</sub> satellite retrieval

“…In another effort to merge model and satellite data, Lamsal et al (2008) was able to infer surface-level NO 2 concentrations from OMI NO 2 by applying local scaling factors from a global model. There has also been an emergence of a technique that combines land-use regression techniques with satellite information to infer ground-level NO 2 concen-trations (Novotny et al, 2011;Vienneau et al, 2013;Lee et al, 2014;Bechle et al, 2015;Young et al, 2016). While each individual technique is useful, all of the aforementioned techniques use model data to adjust existing satellite data but do not address issues inherent with the satellite retrieval methodology.…”