One flaw in your math is that 1900 is not what arterial streets carry. That is the “ideal saturation flow rate” which means if there were green lights all the time and no other interference, you’d probably measure 1900. From there you apply reduction factors. The biggest factor is the green-time factor, which may be .6 on arterials, and .35 or so for collectors. So .6*1900 = 1100. But in truth, a road like the Provo BRT corridor will be closer to 750 or 800 vphpl. Then you have the occupancy factor, which at peak times might be 1.3 or 1.4. So say 800*1.4 = 1100 people/hr/lane, if carried in cars.
Taking this into account, I'll set forth another set of scenarios:
But if we make that more realistic, and assume that the bus is half full during the peak hour (perhaps generous, but plausible), the numbers are much less flattering to BRT. Even at max buses/hour, it moves about as many as an actual travel lane (1215 vs. 1330/1064). However, during rush hour, 30 buses/hour compares favorably: 1215 vs. 1045/836. Hence, during rush hour, a BRT simply carries more persons than any traffic lane, even at 50% full.
However, dropping the number of buses per hour significantly undermines that advantage. As 5 minute headway (12 buses/hour), a half-full BRT only has a capacity of 486, less than half of that for traffic lane, even during rush hour. Each bus would need to be full (90 people) during rush hour, to equal the capacity of a traffic lane.
For automobile travel, passengers per vehicle is the real wild-card. As the chart shows, an HOV lane, even minimally loaded (2 persons) at its worse (1520) carries more persons than all the BRT systems except the full loaded BRT at 3 or 4 minute headways.
This suggests a system of on-arterial HOV lanes would be a more effective strategy than BRT. But that will be the topic for a later post.