I'm trying to model the relationship between hourly traffic count data and some weather variables such as temperature. Both show strong recurring patterns, i.e., traffic data on daily and weekly basis and temperature on daily basis. By going through some previous research, some fitted a univariate SARIMA model to the targeted time series before doing the SARIMAX; some others fitted a 'common' SARIMA model to both response and explanatory time series before SARIMAX. So I'm confused what are the correct steps to conduct a SARIMAX model using these two variables? 

Many thanks,

Sui