Further to my previous comments on the Science March, the graph below shows the (somewhat dubious) attendance estimates from Wikipedia for various cities (excluding vague counts like “thousands”), compared to the power-law predictor 0.47 D1.49 P0.78, where D is the fraction of the relevant state voting for Clinton last year (from Wikipedia), and P is the city population (also from Wikipedia).
The population P predicts 56% of the variance in turnout (not surprisingly), and D an additional 7%. Both factors were significant (p = 0.000000055 and p = 0.014 respectively). Prediction could probably be improved by using metro area population numbers for the cities, by using metro area election results (rather that state results), and by adding factors indicating the number of other marches in the relevant state (Colorado Springs, for example, was rather overshadowed by Denver) and the presence of universities (Ann Arbor, for example, is a university town). But the basic messages seem to be: Democrat voters do not like Donald Trump and Large cities attract large crowds. It would be interesting to compare the numbers here against other recent political marches which focused on different issues.