310 Investigating the Relationship Between Wildfire Smoke and Performance Metrics on Beef Cattle Through Bayesian Additive Regression Trees

Abstract

Abstract Identifying causal relationships on a large-scale system is highly complicated, and when not careful, statistical theories may be violated, and biological spurious claims generated. The causality between livestock health/performance and wildfire smoke has not been investigated. To critically evaluate the effects of wildfire, air quality data was obtained and validated with sensors (1.6 km from feedlot) before analyses. Herein, we examine the influence of environmental quality parameters effects on dry matter intake (DMI), body weight (BW), water intake (WI), and average daily gain (ADG) over a period of three years in Reno, Nevada. The datasets comprised represented controlled no-smoke exposure instances and days with air quality indexes (AQI) as high as 453 (hazardous). The dataset consisted of over 10,000 observations with daily mapping of WI, DMI, ADG, and BW. To evaluate the potential relationship between air quality and animal performance, variable tree length Bayesian additive regression trees (BART) were utilized to investigate the relationships between AQI, particulate matter (PM) 2.5 um and 10 um, NO2, SO2, ozone, and CO levels in the air were utilized as covariates for BART. All statistical analyses were performed on R Statistical Software (R Core Team 2022), with use of the bartMachine package. The correlation values (), 95 % confidence intervals (CI), root mean squared error (RMSE), and coefficient of determination (r2) are reported. A significant correlation was detected between the environmental parameters and BW, (= 0.863, CI = [0.849, 0.877], P > 0.001, r2 = 0.75, RMSE = 51.27) with PM2.5 and PM10 being the most included parameters in the regression trees. For WI, correlation was also significant ( = 0.447, CI = [0.402, 0.490], P < 0.001, r2 = 0.20, RMSE = 14.57), the most included variables on the regression trees were the 10-um particulate matter followed by the CO. For DMI (= 0.727, CI = [0.700, 0.752], r2 = 0.53, RMSE = 2.58) PM10 and AQI were the most important inclusion proportion parameters. Similarly, for ADG (= 0.354, CI = [0.305, 0.401], r2 = 13, RMSE = 1.60) with ozone and PM10 being the most included parameters. In conclusion, our results more carefully examine relationships between smoke parameters and livestock performance. Uncertainty of relationship estimates, and variable importance highlights how performance parameters are differently affected by different environmental parameters highlighting the need for experiments with controlled exposure through time.