Close. The problem is not so much that the historical evidence disagrees with the models, but rather that it is consistent with both the models, and the null hy

Close. The problem is not so much that the historical evidence disagrees with the models, but rather that it is consistent with both the models, and the null hypothesis. A major reason for this cited by the article is the low quality of historical records makes it difficult to use historical evidence. A related problem is that the trends observed in the historical record could be noise. Similarly, the historical record does not provide enough evidence to conclude that are models are wrong. As an extreme example, suppose our entire historical record was "September 10, 1900: sunny skys in Florida. September 10, 2017: Category 5 hurricane in Florida". This is simply not enough information to draw any conclusion.