Linear Geostatistical Estimating Methods Compared Using Annual Average Daily Traffic Data from Low-Volume Roads in Minnesota

Abstract

Geostatistical methods are utilized to analyze and predict values associated with spatial or spatiotemporal phenomena. These techniques incorporate both the spatial and, in some cases, temporal coordinates of the data. The mining industry prefers these methods for ore reserve estimation, and they have also gained popularity in other specialized fields. Notably, the transportation industry is gradually adopting them to estimate annual average daily traffic (AADT). Although researchers use various predictive techniques to estimate AADT, many of them are experiential or based on simple subjective assessments, especially for low-volume roadways. The shortcomings of these assessment techniques include high margins of error, non-integrated spatial and temporal variability in the datasets, and inadequate consideration of total coverage. Geostatistical techniques are superior predictive tools in many fields because their algorithms account for these limitations. Moreover, geostatistical methods can predict data at both sampled and unsampled locations with minimal available data. Therefore, the crucial question arises as to which method best estimates AADT. This paper focuses on investigating linear univariate geostatistical methods, namely empirical Bayesian Kriging (EBK), ordinary Kriging (OK), simple Kriging (SK), and universal Kriging (UK), along with proportional valuations, to estimate AADT on low-volume roadways and provides a recommendation based on the findings. The study utilizes AADT data from 1,486 and 2,731 locations on low-volume roads throughout Minnesota in 2009 and 2016, respectively. The results obtained from this investigation hold significant systematic and practical implications for estimating AADT on low-volume roads and can be adapted for high-volume roadways as well.