It’s probably quite difficult to tease out all of the inter-dependencies between different factors when you don’t have a lot of data points, but did you consider plotting the charts with time on the x-axis and the ratio of patronage to possible driving factor (eg trips per head of population) on the y-axis? A direct 1-1 relationship would then be a flat line and it would be easier to see the time periods where different relationships applied. This would still be a very rough proxy of a multilinear regression and driver analysis, but at least it would be easier to compare the (possible) influence of different factors at different times.

Just one minor question about the Sydney heavy rail data – do the 2000/2001 numbers have Olympic patronage out or in? Once again, great stuff.

HW

Big ask I know, but there is a vital need to publish proper tables from reliable, expensive ABS Tablebuilder type stats for all State Sburb names (within LGA bounds) for the real parking rates in such blocks, inclusive and exclusive of the Llandlord type of ‘State Housing Authority’.
VCAT is digging us into a hole here!
thanks
Martin

I’ve discovered I was using the wrong field hence the strange occupancy by hour figures. I’ve been using the ‘stops’ table, and “STARTHOUR” appears to apply for this leg, while “DEPHOUR” appears to be for the next trip from the destination location (not sure why). I needed to use “STARTHOUR” and I’ve now replaced the chart and it looks much more sensible.

To derive “trip purpose”, I use the destination purpose. unless the destination is “go home” – in which case I use the origin purpose instead. It gets around the 49% “go home” problem, but it does mean that the last non-home activity in a trip chain will be counted twice. I probably should have explained that.

Don’t disagree with your point about all car kms on the road between 8 and 9 am. There will be some school drop off trips that start before 8 am, but a lot more work trips that start before 8 am, and work trips are generally much longer too.

Couple of comments. First, you’re right to suspect a quirk in the VISTA data to explain why there’s an obvious peak in occupancy at school drop-off time but not at pick-up time. This appears to be a direct result of the VISTA design accounting for each leg of a trip separately.

In the afternoon, the trip from home to school will have its purpose coded as ‘pick up someone’ but with only one person in the car, while the return trip with passengers will most likely be coded as ‘go home’. I’ve noticed you don’t have ‘go home’ included as a trip purpose, so your data must be accounting for go-home trips some other way. The VISTA database provides data for trips linked into ‘home-based chains’, but this isn’t really what you want either.

The problem is that if school trips are being accounted in a way that excludes the go-home stage, it will greatly overestimate occupancy in the morning and underestimate it in the afternoon.

But one valid reason for the afternoon pick-up not being so prominent is that there’s more of a tendency by working parents to use after-school care rather than before-school care, which tends to spread the pick-up peak over later times.

One other thing to keep in mind about the effect of school drop-offs on peak hour traffic congestion, is that even if school runs account for 19% of vehicle-kilometres for trips *starting* between 8am and 9am, that doesn’t mean 19% of cars on the road between 8am and 9am are doing school runs. There’s an increasing tendency for journeys to work to commence before 8am, and to take longer. So in that busiest hour you’re also going to see a lot of cars on the road making trips that commenced before 8am. But relatively few school runs commence before 8am (unless they’re combined with work trips), so that 19% is in practice quite a bit less.