Spatiotemporal exposure modeling of ambient erythemal ultraviolet radiation

Abstract

Background: Ultraviolet B (UV-B) radiation plays a multifaceted role in human health, inducing DNA damage and representing the primary source of vitamin D for most humans; however, current U.S. UV exposure models are limited in spatial, temporal, and/or spectral resolution. Area-to-point (ATP) residual kriging is a geostatistical method that can be used to create a spatiotemporal exposure model by downscaling from an area- to point-level spatial resolution using fine-scale ancillary data. Methods: A stratified ATP residual kriging approach was used to predict average July noon-time erythemal UV (UVEry) (mW/m2) biennially from 1998 to 2012 by downscaling National Aeronautics and Space Administration (NASA) Total Ozone Mapping Spectrometer (TOMS) and Ozone Monitoring Instrument (OMI) gridded remote sensing images to a 1 km spatial resolution. Ancillary data were incorporated in random intercept linear mixed-effects regression models. Modeling was performed separately within nine U.S. regions to satisfy stationarity and account for locally varying associations between UVEry and predictors. Cross-validation was used to compare ATP residual kriging models and NASA grids to UV-B Monitoring and Research Program (UVMRP) measurements (gold standard). Results: Predictors included in the final regional models included surface albedo, aerosol optical depth (AOD), cloud cover, dew point, elevation, latitude, ozone, surface incoming shortwave flux, sulfur dioxide (SO2), year, and interactions between year and surface albedo, AOD, cloud cover, dew point, elevation, latitude, and SO2. ATP residual kriging models more accurately estimated UVEry at UVMRP monitoring stations on average compared to NASA grids across the contiguous U.S. (average mean absolute error [MAE] for ATP, NASA: 15.8, 20.3; average root mean square error [RMSE]: 21.3, 25.5). ATP residual kriging was associated with positive percent relative improvements in MAE (0.6-31.5%) and RMSE (3.6-29.4%) across all regions compared to NASA grids. Conclusions: ATP residual kriging incorporating fine-scale spatial predictors can provide more accurate, high-resolution UVEry estimates compared to using NASA grids and can be used in epidemiologic studies examining the health effects of ambient UV.