Statistical Analysis of Automated versus Manual Pavement Condition Surveys

Abstract

The Alabama Department of Transportation (ALDOT) has used a vendor to perform automated pavement condition surveys for the Alabama pavement network since 1997. In 2002, ALDOT established a quality assurance (QA) program to check the accuracy of the automated pavement condition data. The QA program revealed significant discrepancies between manual and automatically collected data. ALDOT uses a composite pavement condition index called pavement condition rating (PCR) in its pavement management system. The equation for PCR was developed in 1985 for use with manual pavement condition surveys; however, ALDOT continues to use it with data from automated condition surveys. Since the PCR equation was developed for manual surveys, the discrepancies between the manual and automated data led ALDOT to question the continuity between its manual and automated pavement condition survey programs. A regression analysis was completed to look for any systematic error or general trends in the error between automated and manual data. Also, Monte Carlo simulation was used to determine which distress parameters most influence the PCR and whether they require more accuracy. The regression analysis showed the following general trends: automated data overreport outside wheelpath rut depth, under-report alligator severity Level 1 cracking, and overreport alligator severity Level 3 cracking. Through Monte Carlo simulation, it was determined that all severity levels of transverse cracking, block cracking, and alligator cracking data require greater accuracy.