A GEOSTATISTICAL APPROACH TO PREDICT 'OPERATING SPEED' ALONG HORIZONTAL CURVES ON TWO-LANE HIGHWAYS

MAZZELLA, ALESSANDRO;PINNA, FRANCESCO;PIRAS C.

2010-01-01

---

Abstract

Injuries are one of the main causes of death according to W.H.O. For this reason the attention of road safety researchers especially regards the study of the relationship between driver and road environment. Several research works show that operating speed is an excellent driver behavior parameter. This article describes a different approach to the classical definition of prediction models for operating speed on horizontal curves. In this paper, the fundamental theories, the applied operating procedures and the first results obtained with the application of Geostatistics are discussed. The mathematical models expressing operating speed in function of horizontal curves characteristics found in International scientific literature, have mainly been built on the basis of Classical Statistics. For this reason, it needs to be pointed out that the interpolative techniques found in Classical Statistics are based upon the use of canonical forms (linear or polynomial regressions) that completely ignore the correlation law between collected data. As such, the determined interpolation stems from the assumption that the data represent a random sample. The models described in this article have instead been created with the geostatistical interpolation technique (i.e. Kriging). This technique allows to obtain the "best" estimates possible because it considers the true correlation law between the measured data. The applied methods are then described along with the results obtained in the field of road safety by applying Geostatistics which, for several years, have been used, with positive results, in all scientific and engineering fields dealing with empirical data analysis and processing.