Conceptual cost estimation of highway bid unit prices using ordinary kriging

Abstract

State-of-the-art encompasses the application of deterministic and geostatistical algorithms to quantify the effect of spatial variation of unit prices across different locations. No studies have evaluated the efficacy of the different combinations of variogram and kriging models when used to characterize the spatial variability of major highway bid unit prices, as well as how well they predict future unit prices at unsampled locations. In this paper, ordinary kriging was combined with three commonly used semivariograms (spherical, exponential, Gaussian) models one at a time to predict six years of the top five common highway bid items obtained from the Wisconsin Department of Transportation (WisDOT) and evaluated the prediction uncertainty for each bid item. The results show that the best-performing combination of ordinary kriging and semivariogram varied depending on the type of bid unit-price data. This study advances the state of practice by comparing the efficacy of various combinations of the variogram and ordinary kriging models in predicting highway bid unit prices and provides insight into the models’ behavior in describing bid unit price spatial heterogeneity. Transportation agencies can employ this methodology to model other highway bid items and project types that have not been experimented in this study.