Measures of controlling traffic operation and emissions are developing to be dynamic and detailed. We applied the dynamic speed data of social vehicles from traffic real-time platform and GPS data of buses from practical tests

then used LWLR (Locally Weighted Linear Regression) to forecast average speeds and operating mode distributions of some urban buses in Guangzhou. Finally a model for estimating emission factors (for liquefied natural gas buses) and electricity consumptions (for electric buses) was built. The results showed that

the average speeds of buses were positively correlated with those of social vehicles. The operating mode distributions under different average speeds had regular patterns. With the increase of the average speeds of buses

the frequency of Bin1 (idling mode) decreased

the frequencies of Bin11-12 increased firstly and then decreased

and the frequencies of Bin13-15 (denoted low speed and acceleration) were partly transformed to those of Bin16-17 (denoted high speed and acceleration). The proposed algorithm reflected the no-linear trend that the speeds of buses increased with that of social vehicles

and the mean absolute percentage error of speeds of buses was 1985%. The mean absolute percentage errors of NO

x

emission factors and electricity consumptions of buses were 20.27% and 26.52%.