Quantifying the Uncertainty in Annual Average Daily Traffic (AADT) Count Estimates

This paper describes how Average Annual Daily Traffic (AADT) values provide a key variable in many models and policy decisions; however, these are simply rough estimates of traffic counts along the vast majority of roadway sections. This research quantifies the level of uncertainty in AADT estimates and compares these across sampling strategies. Variations in AADT estimation errors are investigated across roadway and area types, for both Minnesota and Florida automatic traffic recorder (ATR) sites. Errors as a function of distance to the nearest sampling site are also studied, using predictions of network travel patterns in Austin, Texas. Overall errors at ATR sites are found to be highest (averaging 24.6%) when data come from misclassified sites on weekends. Spatial and temporal (inter-sampling year) extrapolations can further add to such error, in a sizable way. The analytical results of this investigation suggest a variety of recommendations for agencies seeking to reduce and appreciate errors in their AADT estimates. These include sampling in spring and summer months (on weekdays), exercising greater caution with counts on multi-lane and low-AADT roadways, pursuing appropriate site assignment to ATR groups, and recognizing the effects of distance to the sampling site. With adequate attention, (average) errors in AADT estimates can probably be reduced to the 10 percent level. Nevertheless, these still will have an impact on investment decisions, crash rate calculations, travel demand model validation, and other analyses.