Modeling traffic noise intensity and comparative validation analysis of Arima and MLR models

Abstract

The traffic noise big data collected from studying traffic situations in Port-Harcourt Nigeria selected trunks A and C roads subsectioned as flexible pavements locations 1, 2, and 3 and flexible and rigid pavements locations 4 and 5 respectively has been analyzed by using the  multi-linear regression (MLR) technique. Traffic noise is an acoustic hazard affecting mostly people living closest to the roadway  pavement. The solution of such a high degree of discomfort on roadside dwellers deserves serious study. This work considered traffic  parameters like distance between dwellers and the roadway, traffic count, vehicular speed, traffic periods, etc. in modeling the traffic  noise intensity (TNI) of the selected road. The average peak traffic noise for location 1 obtained at various distances of 5m, 10m and 15m  from the centre of the roadway are 85.59dB, 84.93dB and 83.97dB respectively, for location 2 are 86.52dB, 85.34dB and 84.26dB  respectively, for location 3 are 84.38dB, 83.88dB and 83.32dB respectively, for location 4 are 85.16dB, 84.56dB and 83.55dB respectively,  for location 5 Trunk C Flexible Pavement are 55.46dB, 54.36dB and 53.99dB respectively and for Trunk C Rigid Pavement are 60.58dB,  59.58dB and 58.96dB respectively. The traffic noise values for location 1-4 had higher noise intensity and same range, it was categorized  as Trunk A flexible pavement and classified as heavy-trafficked routes while location 5 (Trunk C) had lower noise intensity and same range  which was classified as light-trafficked routes. MLR predicted the TNI with R2 (0.2015, 0.2110, 0.1894, 0.2203, 0.2275, 0.1983, 0.4398,  0.4398, 0.3907, 0.3952, 0.3427, 0.3355, 0.3149, 0.1505, 0.1526, 0.1441, 0.002, 0.0012, 0.001) values for the model along the selected routes.  From the result, the distance of noise measurement from the centre of the roadway of Trunk C flexible pavement with the most  significant p-value of 0.804145, the equivalent traffic volume and traffic speed had p-values of 0.014782 and 3.22E-50 respectively whereas  that of Trunk C rigid pavement with the most significant p-value of 0.872625, the equivalent traffic volume and traffic speed had  p-values of 0.265025 and 3.67E-61 respectively. The noise level increased more on rigid pavements than that of flexible pavements, which is attributed to more voids on rigid pavements and the higher frictional noise due to increased frictional force between the vehicle tires  and road surfaces with the grip being more in rigid pavements. At the end of the exercises, it was observed that ARIMA (R² greater 90%)  performed better than MLR even with the technical advantage of determining noise difference between interfering points using the auto- correlation factor (ACF) and the partial auto-correlation factor (PACF).