Land-Use Regression Estimation of Cumulative Environmental Noise Exposure in Jefferson County, Kentucky

Abstract

BACKGROUND AND AIM: Noise exposure is generally reported as year-round and source-specific. Recent studies, predominately outside of the US, have reliably estimated environmental noise with land-use regression (LUR). We estimated cumulative environmental noise exposure in Jefferson County, Kentucky via LUR for different seasons and time windows. METHODS: Noise data were collected using a Class 1 sound level meter at 15 sites for 24 hours in January 2021 and April 2021. Noise data were averaged to represent combinations of season and time of day. LUR models were built based on prior literature and statistical thresholds using geographic predictors, such as Normalized Difference Vegetation Index (NDVI), length of streams, railways, and major roads within a buffer; distance to airports; hospitals and police stations; traffic volume and more. R2 values assessed model fit and the leave-one-out cross validation and root-mean squared errors (RMSE) assessed the predictive ability. RESULTS:Models resulted with R2 values ranging from 0.50 to 0.73 and RMSE values between 4.57 and 6.49 decibels. NDVI and distance to flightpaths were consistent predictors of noise across seasons and time windows. Compared to living within 1 km of flightpaths, living further away resulted in lower noise estimates (beta coefficient range: -0.99 to -4.86). A 0.1 increase in NDVI resulted in lower noise estimates (coefficients range: -2.01 to -5.97). Length of major roads (beta=1.5 95% CI -3.3, 6.4) and streams (beta=-4.1 95% CI -13.0, 4.9) were predictors of January noise. Traffic volume and railroad length were positive predictors of April noise but not January noise. Noise was approximately 10 decibels louder in April than in January. CONCLUSIONS:Multiple models may be needed to estimate environmental noise for varying seasons and times when considering total ambient noise exposures. Consideration of other noise sources (ex. NDVI, streams) in noise LUR models is warranted in future work. KEYWORDS: Noise, spatial statistics, modeling, exposure assessment, environmental epidemiology