Using Mobile Monitoring To Measure Traffic-related Pollutants In Fine-scale Urban Streets

In the last two decades, increasing vehicular pollutants(CO and NOx) are severely threatening travelers’ health status. Air quality in street microenvironment is closely related to each traveler. With the increasing demand for the traffic travel in urban area, air quality in street microenvironment has gained more attention than ever before.Diffusion model is suitable for studying the air quality in urban street canyon with symmetrical aspect ratio, and numerical simulation model based on hydrodynamics can deal with streets with asymmetrical aspect ratio,their predicting results are hardly to be validated. Mobile monitoring, for its high sampling resolution, sampling data’s reliability, as well as its ability of periodic monitoring, is gradually becoming a fresh method for studying the street microenvironment.We obtained 8 i nfluencing factors of street environment: wind speed, wind direction, temperature, humidity, speed, volume, diesel volume, aspect ratio. We use mobile monitoring platform to sample traffic-related pollutant concentration and its influencing factors to analyze: 1) Reliability of data sampled by mobile monitoring,2) We defined an ideal street environment, which assumed that meteorological and traffic factors were stable. Under the ideal state, correlations of moving speed and the measured concentration were analyzed, so were relationships between mobile sampling and fixed sampling,3) We defined a normal street environment, which assumed that traffic-related pollutant in street level was jointly influenced by meteorological, traffic and local geographical factors. Under normal status, both parametrical and non-parametrical predicting models of traffic-related pollutant concentration in street microenvironment were studied.Reliability of data sampled by m obile monitoring was conducted in two parts: stable calibration and mobile calibration. Mobile calibration of mobile monitoring analysis revealed that sampling measurements under different moving speeds can be used to simulate travelers’ exposure level under different transport modes.Analysis of street microenvironment concentration in ideal state demonstrated that the moving speed and its measured concentrations did not show obvious linear relationship, the optimal moving speed of a mobile monitoring platform to get traffic-related fixed sampling pollutants was 0-5km/h and 10~15km/h for NOx and CO, respectively.In normal state, we built four parametrical models of polynomial, power function, logarithmic, and hyperbolic regression models, power function(CO_R2=0.76,NOx_R2=0.44)demonstrated best performance. And three non-parametrical models were also built namely as neural networks, support vector machines and random forest, while random forest’s performance takes the lead.Under normal state, predicting model of power function and random forest were validated by new sampling data. Results showed that predicting performance of power function(CO_R2=0.68,NOx_R2=0.5) is less superior to random forest(CO_R2=0.83,NOx_R2=0.53). Analysis of predicting error indicated that traffic volume plays an important role in predicting the pollutant concentrations(CO, NOx) in urban microenvironments.Predicting traffic-related pollutant concentration in fine-scale streets improves urban governors’ regulate and control ability on street air quality, showing potential guidance for travelers’ green travel mode.