Spatiotemporal land use random forest model for estimating metropolitan NO2 exposure in Japan

Abstract

Adequate spatial and temporal estimates of NO concentrations are essential for proper prenatal exposure assessment. Here, we develop a spatiotemporal land use random forest (LURF) model of the monthly mean NO over four years in a metropolitan area of Japan. The overall objective is to obtain accurate NO estimates for use in prenatal exposure assessments. We use random forests to convey the non-linear relationship between NO concentrations and predictor variables, and compare the prediction accuracy with that of a linear regression. In addition, we include the distance decay effect of emission sources on NO concentrations for more efficient model construction. The prediction accuracy of the LURF model is evaluated through a leave-one-monitor-out cross validation. We obtain a high R value of 0.79, which is better than that of the conventional land use regression model using linear regression (R of 0.73). We also evaluate the LURF model via a temporal and overall cross validation and obtain R values of 0.84 and 0.92, respectively. We successfully integrate temporal and spatial components into our model, which exhibits higher accuracy than spatial models constructed individually for each month. Our findings illustrate the advantage of using a LURF to model the spatiotemporal variability of NO concentrations.

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

“…Land-cover terms are proxies for traffic emissions and related to NO 2 concentrations indirectly. A typical NO 2 model is based on a land-cover regression with such quantities as road length, population density, tree canopy coverage, impervious surface, elevation, distance to coast, 12 traffic flow, 13 traffic intensity, 14 land-cover type, 13,15,16 road types, 17 building density, 18 and urban density, 19 as predictor variables.…”

Low-cost sensor networks and land-use regression: Interpolating nitrogen dioxide concentration at high temporal and spatial resolution in Southern California

“…RF models have successfully been used in previous studies aiming to predict NO2 concentrations using land use (Araki et al, 2018;Chen et al, 2019;Hu et al, 2017;Zhan et al, 2018). This approach was chosen to minimise the risk of overfitting given that there are relatively few monitoring sites and also to capture non-linear relationships observed between air pollutant concentrations and predictor variables (Araki et al, 2018;Chen et al, 2019;Vizcaino and Lavalle, 2018). RF models are bagged decision tree models, where each tree consists of a random subset of predictor variables from the training dataset and where the final output is the average of multiple decision trees (Breiman, 2001;Grange et al, 2018;Vizcaino and Lavalle, 2018).…”

Machine Learning Approaches for Outdoor Air Quality Modelling: A Systematic Review

“…Araki et al [57] also build a spatiotemporal model based on land use to capture interactions and non-linear relationships between pollutants and land. The study consists of comparing two algorithms-Land Use Random Forest (LURF) vs. Land Use Regression (LUR)-to estimate the concentration levels of NO 2 .…”

“…Brokamp et al [31] perform a deeper analysis based on a prediction of the concertation of the chemical composition of PM 2.5 . As Beckerman et al [57], they compare the performance of a LURF versus LUR approach. The dataset is built from the measurements of 24 monitoring stations in the city of Cincinnati (USA) during the period 2001-2005.…”