By matching fatal traffic accidents across the contiguous United States over a six-year period from the Fatality Analysis Reporting System (FARS) with radar-based precipitation estimates from the newly-development National Mosaic and Multisensor Quantitative Precipitation Estimate (NMQ/Q2) reanalysis, we examine the frequency of accidents which occur during observed precipitation, compared to the climatological frequency of precipitation in that area. In the absence of a relationship, one would expect these frequencies to be approximately equal, simply by random chance. A significant departure from the expected frequency gives cause for suspicion of a causal relationship.
In all climate regions and 40 of 45 examined states, the frequency of crashes that occur during precipitation is significantly higher than would be expected in the absence of any relationship, from a 9% difference in the Northeast to a 51% difference in the Northwest. These results hold true when controlling for time of day, season, and other risk factors such as alcohol and drug involvement.