Quality Control of Precipitation Data for Wet Pavement Accident Analysis

Abstract

Rainfall and wet pavement have significant impacts on road safety. Wet percent time was found to be an important variable in wet pavement accident analysis. However, the accuracy of wet percent time can be impaired by data with quality issues, which have negative impacts on the identification of high wet collision locations. Hence, data quality control is a necessary and essential part of any weather station network used for generating wet percent time. The absence of quality control can result in poor quality data that severely limits their usefulness. This study showcased the first application of a methodology for the multi-step quality control of precipitation data. The methodology includes data preprocessing, various quality checks, and treatment of missing data. Historical hourly precipitation data reported by rain gauges were obtained from five network data sources to demonstrate this methodology. These five network sources provided data from over 2,000 weather stations in California. The results of quality control were employed by the California Department of Transportation to update its wet percent time information. The methodology can be adopted by other states conducting similar studies or for applications such as climate change research.