Only the data knows for sure ! with apologies to Lamont Cranston . One would have to capture any ARIMA structure to be able to "SEE" other significant phenomenon like Pulses , Level Shifts, Seasonal Pulses , Local Time Trends , Change points in parameters over tine , Change points in model error variance over time , dependency between error variance and the level of the series etc...
If you post the data we can easily perform an analysis to suggest structure and violations thereof.
There might be hourly effects etc ... that once adjusted for might enable one to "SEE" better .