Annual average daily traffic (AADT) values play an important role in transportation design, operation, and planning. Each year, transportation agencies spend a significant amount of resources collecting this information. However, AADT values are mostly rough estimates based on the closest short-period traffic counts, factored up using adjustment factors derived from permanent continuous count stations. For example, in New York State, the unadjusted ADT obtained from the short period traffic counter is adjusted by the seasonal adjustment factors and the axle adjustment factor. Thus the accuracy of AADT relies heavily on the precision of adjustment factors. New York State calculates the seasonal adjustment factors using the average of three years’ continuous count data. The factors are then grouped into three categories based on road segment locations and functions: urban, suburban and recreational. (NYSDOT, 2010) Although convenient to use, such factor categorization leads to aggregate and arbitrary estimates. For example, in the transitional areas between suburban and recreational sites, it is often difficult to determine which group of factors should be used. Different land use types and demographic distribution in the surrounding neighborhood may also lead to different temporal fluctuation. For example, though both in “urban” areas, roads along commercial development tend to have peak volume during weekends of holiday seasons while those within residential development normally have lower volume in the same time period.

This research develops a method that will generate site-specific seasonal adjustment factors based on (1) site conditions such as the number of lanes, road functional classification and surrounding neighborhood information, and (2) the spatial dependence of traffic flows over road network. That is, in addition to the consideration of various site-specific variables, each road segment will also obtain its unique adjustment factor based on its spatial connections to the surrounding permanent continuous count stations. The theoretical foundation of such method is Tobler’s first law of geography: "Everything is related to everything else, but near things are more related than distant things." The proposed model is built on the Kriging method, which presumes spatial autocorrelation in unobserved factors as a function of distance. This study further advances the standard Kriging approach by utilizing network connectivity indicators instead of Euclidean distance. The indicators to be evaluated include network distance, network topology and equilibrium flow sensitivity. The validated model will be applied to the whole road network in the State of New York and yields a continuous adjustment factor map.

The proposed research is of significant importance to the region. With the proposed method and the resulted adjustment factor map, the New York State will get more reliable AADT estimates, which are crucial to the planning and management of transportation systems. The approach also provides a promising way to explore spatial relationships across a wide variety of network-based data sets, including, for example, pavement conditions, traffic speeds, percentages of trucks, land values, and trip generation rates. All of these are critical components of the transportation system. In short, the proposed research will provide means to utilize the geographic information, develop new directions to solve traditional problems, and lead to significant impact on the efficient management of transportation systems in New York State.