Spatial Prediction of AADT in Unmeasured Locations by Universal Kriging

Brent Selby, Kara M. Kockelman

Transportation Research Board 90th Annual MeetingTransportation Research Board · 2011 · 27 citations · 8 references

Concepts

Abstract

This work explores the application of kriging methods for prediction of average daily traffic counts across the Texas network. Both results based on both euclidean and roadway network distances (between new count sites and existing data-collection sites) are compared, allowing for strategic spatial interpolation of count values, while comparing for functional classification, lane numbers, speed limits, and other site attributes. Universal kriging is found to reduce errors (in practically and statistically significant ways) over non-spatial regression techniques, thought, at some sites, errors remain quite high, particularly in less dense areas and on small roads near major highways. Interestingly, the estimation of kriging parameters by network distances showed no enhanced performance over Euclidean distances, which require less data and are much more easily computed.

References

8