Liu, YueThis study was conducted to analyze and quantify the impact of weather factors on road accident severity, based on Maryland accident data during 2007-2010. In order to find a better model fitted related variables, three candidate models multinomial logit (MNL), ordered probit logit (OP), and neural networks were chosen to examine in SAS. The results showed that the Multilayer Perceptron Model in neural networks performed the best and is the accident severity model of choice. During the model construction, eight factors related to weather condition were considered. They were: air temperature, average wind speed, total precipitation in the past 24 hours, visibility, slight, moderate, heavy precipitation and relative humidity. Based on the comparison criteria, we concluded that MNL regression is more interpretive than OP and Neural Networks models. All factors except visibility and heavy precipitation had significant impact on accident severity when considering the data from the entire Maryland highway system. Using MNL, a data subset with accident records only in a section of US route 50 was examined. After excluding the impact factors other than weather, a narrow significant variable set was obtained.ThesisTransportation planningCivil engineeringaccident severityhighway accident factorsregressionweather impact