When a pavement is flooded, transportation agencies are faced with the decision to leave the road open or close it to traffic. On the one hand, agencies may want to keep the road open and prioritize connectivity if the road serves a critical economic purpose for a region. On the other hand, agencies may choose to close the road to prevent additional structural damage (which may incur significant repair costs) because of high water content in (or even saturation of) the subgrade. Nondestructive testing (NDT) tools such as the falling weight deflectometer (FWD) can be applied to assist decision makers by facilitating a more reliable evaluation of the structural condition of the pavement. Reliable decision making ought to incorporate inherent uncertainties in the process, including the structural state of the pavement after flooding, as well as the reliability of FWD testing and variability in costs. This paper presents a method that uses a Bayesian decision tree approach for highway emergency operations after flooding. Uncertainties in the structural state of the pavement after flooding, NDT, and costs are addressed with Monte Carlo simulations. A case study of a flooded roadway section in North Dakota demonstrated this approach: user delay costs caused by closure of the roadway were explicitly considered. The results of the decision tree analysis provide objective recommendations about opening or closing a road after flooding, as well as whether FWD testing of the flooded road should be conducted once the water recedes.