I would have guessed that it is just due to humans being very visual beings.

The paper briefly considers whether the effect is due to graphs being easier to process, but apparently not (p.32):

These results do not appear to be driven by systematic differences how respondents processed Text or Graph – a post-treatment check of recall of a primary data source (NASA) found few significant differences between the treatments. Moreover, we observe no significant difference in the length of time respondents spent considering each treatment. These results suggest that the greater effectiveness of Graph is not simply a function of ease or depth of processing (results available upon request).