Abstract. An important topic for climate change investigation is the behavior of severe weather under future scenarios. Given the fine-scale nature of the phenomena, such changes can only be analyzed indirectly, for example, through large-scale indicators of environments conducive to severe weather. Climate models can account for changing physics over time, but if they cannot capture the relevant distributional properties of the current climate, then their use for inferring future regimes is limited. In this study, high-resolution climate models from the North American Regional Climate Change Assessment Program (NARCCAP) are evaluated for the present climate state using cutting-edge spatial verification techniques recently popularized in the meteorology literature. While climate models are not intended to predict variables on a day-by-day basis, like weather models, they should be expected to mimic distributional properties of these processes, which is how they are increasingly used and therefore this study assesses the degree to which the models are actually suitable for this purpose. Of particular value for social applications would be to better simulate extremes, rather than inferring means of variables, which may only change by small increments thereby making it difficult to interpret in terms of the impact on society. In this study, it is found that the relatively high-resolution NARCCAP climate model runs capture areas, spatial patterns, and placement of the most common severe-storm environments reasonably well, but all of them underpredict the spatial extent of these high-frequency zones. Some of the models generally perform better than others, but some models capture spatial patterns of the highest frequency severe-storm environment areas better than they do more moderate frequency regions.