Identifying risk factors for road traffic injuries is one of the main priorities of transportation agencies. More than 12,000 fatal work zone crashes were reported between 2000 and 2013. In spite of recent efforts on improving work zone safety, the frequency and severity of work zone crashes are still a big concern for transportation agencies. The effect of work zone on traffic safety is shown to be intensified by adverse weather conditions. Although many studies have been conducted on different work zone safety-related issues, there is a lack of studies that investigate the effect of adverse weather conditions on work zone crash severity. This paper utilized probit-decision tree; a relatively recent and promising combination of machine learning technique and conventional parametric model, to identify factors affecting work zone weather-related crashes severity using a unique dataset collected by the second Strategic Highway Research Program Roadway Information Database. The key strength of this technique lies in its capability of alleviating the shortcomings of both parametric and non-parametric models. The results were compared to a conventional probit model and the proposed probit-decision tree is considered a better technique that outperformed the conventional probit regression because of its high estimation accuracy, robustness and reliability, and its ability of estimating marginal effects of risk factors.