How fast are our trains? You could ride each line to time it or look up a timetable. Then use published information (like Vicsig line guides) to establish the distance and work out the speed.
This was done last December by Declan Martin who made a nice network-wide comparison map (that got both newspaper and radio coverage at the time).
Not surprisingly outer suburban trains, with fewer inner area stops, were fastest with speeds of 55-60 km/hr to areas like Wyndham Vale, Melton and Sunbury. These are mostly but not always diesel V/Line services. Metro trains on lines with more stations were much slower at 35 to 40 km/h as we don't have a tradition of off-peak expressing. Longer lines like Pakenham, with some longer gaps between stations, did better at 50 km/h while shorter lines like Williamstown and Alamein were nearer to 30 km/h. For context buses on local streets average around 22km/h with Melbourne trams in mixed traffic often slower.
The table below lists lines in descending order of speed (km/h using figures from the abovementioned map).
Not just about in-train speed
End to end travel speed is only part of the story. Firstly no one lives and hardly anyone works at a station. So getting to it, possibly via another public transport service, needs to be factored in. And, unless you are very lucky, your journey will involve a wait. Count both these things and end-to-end speeds may halve or more. Reliability and consistency of travel time are also very important.
End to end travel speed is only part of the story. Firstly no one lives and hardly anyone works at a station. So getting to it, possibly via another public transport service, needs to be factored in. And, unless you are very lucky, your journey will involve a wait. Count both these things and end-to-end speeds may halve or more. Reliability and consistency of travel time are also very important.
I won't cover station access today but will discuss the waiting component of train journey speed. Including this gives a more representative view of how much time to allow for a trip. Especially if you've arrived at the station randomly, such as might be the case if coming off a bus (Melbourne still has many bus routes every 22 - 30 minutes that attempt to meet trains every 20 minutes) or returning from a job or show with a fixed finish time.
How much slower with average waits?
Unlike Perth and Sydney, which runs more consistent and usually higher frequencies, off-peak trains in Melbourne operate every 10, 15, 20, 30 or 40 minutes off-peak on weekdays depending on time and line. Such widely varying frequencies have a big average effect on travel time and an even larger influence if a train has just been missed.
How much slower with average waits?
Unlike Perth and Sydney, which runs more consistent and usually higher frequencies, off-peak trains in Melbourne operate every 10, 15, 20, 30 or 40 minutes off-peak on weekdays depending on time and line. Such widely varying frequencies have a big average effect on travel time and an even larger influence if a train has just been missed.
Just how big is frequency's influence? And could it even be more decisive than express running under some circumstances?
I took the abovementioned map and reworked the data so speeds included average waits. I assumed a random arrival at the station during an off-peak weekday period. That is passengers were using it as if it was a turn-up-and-go service. The average wait time in this case is half the interval between trains. Hence you would wait 10 minutes on average if you rocked up at a station with trains every 20 minutes. Adding the average wait time and recalculating the speed gives a fairer overall speed for passengers who waited an average time.
You can see the result of this below (click for better clarity):
I took the abovementioned map and reworked the data so speeds included average waits. I assumed a random arrival at the station during an off-peak weekday period. That is passengers were using it as if it was a turn-up-and-go service. The average wait time in this case is half the interval between trains. Hence you would wait 10 minutes on average if you rocked up at a station with trains every 20 minutes. Adding the average wait time and recalculating the speed gives a fairer overall speed for passengers who waited an average time.
You can see the result of this below (click for better clarity):
Which terminus stations had the most reduction in speed when the waiting experience was factored in? As you'd expect infrequency is the biggest factor. Also important was travel time; lines with low travel times have a proportionately higher waiting component so counting waiting meant a larger percentage reduction in speed.
Including average waits has also changed the relativities between how terminus stations rated for average speed. Frankston (on a line with many closely spaced stations) rated 10th out of 17 for speed if frequency wasn't considered. However counting frequency boosted its ranking to 7th place due to its excellent 10 minute frequency (or 5 min average wait).
The biggest relative fall was for Sunbury station due to the 20 minute average wait caused by its 40 minute frequency. Belgrave and Lilydale also dropped. After average waiting time travel from these stations was slower than for Frankston. This is remarkable as Belgrave and Lilydale have off-peak expressing whereas Frankston does not and started off at a much slower speed. The difference is entirely attributable to frequency with Belgrave and Lilydale being fast and infrequent versus Frankston station's slow but frequent service.
The above numbers are conservative as I've used actual time rather than perceived time (waiting is perceived to be longer than in-train travel time so some analysts give it a multiplier weighting). Intermediate stations will have different speeds but I'm considering only the termini here.
Factoring in maximum waits
Here's the above table expanded to include a maximum wait column. The numbers in it are the speeds you get if you just miss a train and need to wait for the next one. Re-sorting on maximum waits has changed the rankings and accentuated speed differences further.
Cutting variability the real way to speed
We've talked about comparisons between lines.
Now run you eye across some of the rows. Notice the extent to which speed drops off for some lines. Stations like Sunbury drop dramatically with average speeds almost halving if you happen to just miss a train. Whereas there's only a small change for Frankston. This is because Frankston has quadruple Sunbury's frequency with a quarter the wait.
We've talked about comparisons between lines.
Now run you eye across some of the rows. Notice the extent to which speed drops off for some lines. Stations like Sunbury drop dramatically with average speeds almost halving if you happen to just miss a train. Whereas there's only a small change for Frankston. This is because Frankston has quadruple Sunbury's frequency with a quarter the wait.
High frequency like Frankston also means less variable travel times. That's equally as important as speed for many passengers, especially those making shorter trips. Low variability means people can plan their day better and don't need fat contingencies for waiting that can easily double the time allowed for. Having to allow such high contingencies encourages too many to drive even for 'best case' scenario trips where both origin and destination are on the one train line.
I got a variability figure by comparing the travel time of worst case (train just missed) with best case (train just caught). For example if a train trip takes 60 minutes and the gap between trains is 30 minutes then travel time could be as high as 90 minutes if a train has just been missed. That's a 50% higher variation. In contrast the same length trip on a line with a 10 minute service would (at worst) take 70 minutes, or a 17% variability.
I got a variability figure by comparing the travel time of worst case (train just missed) with best case (train just caught). For example if a train trip takes 60 minutes and the gap between trains is 30 minutes then travel time could be as high as 90 minutes if a train has just been missed. That's a 50% higher variation. In contrast the same length trip on a line with a 10 minute service would (at worst) take 70 minutes, or a 17% variability.
Frequent service means low variability as the worst-case waiting time is never a high proportion of the journey time. The table below ranks termini by variability with Frankston performing the best due to its length and frequency. In contrast infrequent lines like Sunbury and short lines like Williamstown perform worst.
I didn't plot these but stations like Geelong, Ballarat and Bendigo have less variability than Wyndham Vale, Melton and Sunbury due to longer travel times. Boosting frequency may be desirable for capacity reasons but they don't need to be particularly high for acceptably low variability (at least for full length trips). Instead the emphasis should be on speed and reliability, though when that's done frequency needs boosting again to preserve the low variability previously obtained through slow travel times.
I didn't plot these but stations like Geelong, Ballarat and Bendigo have less variability than Wyndham Vale, Melton and Sunbury due to longer travel times. Boosting frequency may be desirable for capacity reasons but they don't need to be particularly high for acceptably low variability (at least for full length trips). Instead the emphasis should be on speed and reliability, though when that's done frequency needs boosting again to preserve the low variability previously obtained through slow travel times.
The average variability in travel time for most lines is around 50%. This means that travel can easily take 1.5 times as long as it would if a train was just at the platform when you arrived. These figures are for off-peak weekdays from terminus stations. But with a range from 15 to nearly 100%, the 6:1 variability in variability is higher than the 2:1 variation in travel time between our faster and slower lines. And, unless they enjoy higher frequency, the variability percentage tends to be higher at intermediate stations on a line as travel time is a smaller component of overall trip time.
I'm not denying the value of speed but it is often oversold compared to variability. And often you can actually get better overall speeds if you tackle variability first (which includes not only service design elements such as frequency but also reliable operations to minimise cancellations and delays).
Actual passenger experiences are better if using inner stations with more frequency (eg inward from Watergardens on the Sunbury line or Eltham on the Hurstbridge line). But they can be worse at night / weekends (with lower frequencies) and for the majority of passengers who use intermediate stations. Add this, plus indifferent bus connections and it is very possible for trips to take double the time they should, and maybe four times longer than driving.
Frequency's the fix
I'm not denying the value of speed but it is often oversold compared to variability. And often you can actually get better overall speeds if you tackle variability first (which includes not only service design elements such as frequency but also reliable operations to minimise cancellations and delays).
Actual passenger experiences are better if using inner stations with more frequency (eg inward from Watergardens on the Sunbury line or Eltham on the Hurstbridge line). But they can be worse at night / weekends (with lower frequencies) and for the majority of passengers who use intermediate stations. Add this, plus indifferent bus connections and it is very possible for trips to take double the time they should, and maybe four times longer than driving.
Frequency's the fix
What's the good news? The answer is that this is fixable. Often without the big builds and long shutdowns that have made taking trains such a gamble lately. Let's see what happens to speed and variability if we boost frequency, especially on lines to growth areas or through established areas undergoing residential densification and jobs growth.
For the sake of this exercise no line is worse than every 20 minutes with the majority getting a 10 minute service. Rather than being an unattainable dream, a network vision like this was 2013 official policy as published in the Network Development Plan - Metropolitan Rail. You can see how frequency uplift improves average speeds and lowers variability from terminus stations below:
For the sake of this exercise no line is worse than every 20 minutes with the majority getting a 10 minute service. Rather than being an unattainable dream, a network vision like this was 2013 official policy as published in the Network Development Plan - Metropolitan Rail. You can see how frequency uplift improves average speeds and lowers variability from terminus stations below:
Speeds for average waits are up by nearly 10% with typical variability cut from 50 to around 30%. In raw minutes terms this is a saving of 20 minutes for those who just miss a train on a 40 minute line and 10 minutes for those currently on 20 minute lines. Proportionate time savings for those making shorter trips or changing services are even better, especially if accompanied by bus reform, tram speed improvements and even higher frequencies for inner area stations (eg 5 min in from Dandenong or 10 min inwards from Ringwood, Watergardens and Greensborough).
The lessons are clear. If we want faster overall train travel we should be tackling frequency first. That speeds overall travel times and makes them less variable - a good aim in itself.
As well as higher frequency and fewer unique stopping patterns, revised greenfield timetables should incorporate quicker run times where there is known timetable padding for even greater time savings. Active management including headway running should be considered on very frequent lines that are operationally isolated without lower frequency branches. And all-day off-peak expressing is desirable on longer lines provided each metropolitan area stopping pattern can be run frequently.
Given that most of these gains are possible with existing infrastructure and rolling stock, the benefits are relatively high for the effort required. To say that better frequency cures all the network's issues is exaggerating. But not by much.
The lessons are clear. If we want faster overall train travel we should be tackling frequency first. That speeds overall travel times and makes them less variable - a good aim in itself.
As well as higher frequency and fewer unique stopping patterns, revised greenfield timetables should incorporate quicker run times where there is known timetable padding for even greater time savings. Active management including headway running should be considered on very frequent lines that are operationally isolated without lower frequency branches. And all-day off-peak expressing is desirable on longer lines provided each metropolitan area stopping pattern can be run frequently.
Given that most of these gains are possible with existing infrastructure and rolling stock, the benefits are relatively high for the effort required. To say that better frequency cures all the network's issues is exaggerating. But not by much.
3 comments:
Frequency is great for shorter trips, but the vast majority are doing longer trips.
I appreciate this is data is a bit older now, but I’d be curious to know how this has changed over time (changed timetables, increased services, line disruptions due to level crossing removal and other infrastructure projects), maybe even a timescale over 10 years and if there is a pattern in lines/areas gaining or losing avg speed p/hr.
Anon - in relation to services over the last decade (or three), it's been the politically marginal Frankston line that had the biggest increase, with Dandenong also getting big uplifts. Ringwood got big gains on weekends and Sandringham back in the '90s. The safe Labor seats in the west and north have got electrification extensions but their timetables are basically stuck in time with the likes of Craigieburn and Mernda having double the waits of the Frankston line despite high usage productivity. The other east beyond Ringwood got good gains on weekends but also remains stuck with 30 min gaps off-peak weekdays.
Geelong and Melton have done very well in relation to weekday off-peak service. However weekends still lag.
Line speeds slowed about 10 or 15 years ago with timetables slowed to reflect the patronage boom. Eg it used to be an hour from Frankston to Flinders St on an all stations train. Now its way slower (if you get a train - and it's very likely a replacement bus). However the frequency is much improved as a sort of trade-off. Dandenong (and Sunbury) should hopefully get big service and some speed gains with the Metro Tunnel - nothing public about the rest.
Post a Comment