Modernisation of passenger rail in Ireland: a very rough draft

Update: Jack Harman alerted me on Twitter that Irish Rail uses mostly Diesel Multiple Units outside of the Dublin-Cork line. This post has been edited.

Update #2: Here’s a spreadsheet with all the stops on all the lines. I didn’t write a timetable and I’m not planning to.

Update #3: @tggleeson on Twitter pointed out that I was sending no trains between Portarlington and Athlone. I fixed this by sending the Dublin-Westport trains there.

Update #4: @simonschre made an awesome map for the proposed IC network!

I’m dusting off my blog where I previously only wrote in French for a 6k-word English-written enormous post about what a good passenger rail network could look like in the Republic of Ireland. Be warned, this is going to be a long read.

The current set-up

There are 2733 km of track operated in Ireland, both in the Republic and in Northern Ireland. Of those, 53 km are electrified (in the Dublin area), 80 km are already upgraded to a good speed standard (in the Dublin-Cork line), and another 362 km are freight only [1]. All passenger traffic is operated by Irish Rail, the national public carrier.


What follows is a quick rundown of the Intercity services operated by Irish Rail. Dublin commuter rail is out of the scope of this post.

The best Intercity (IC) line Irish Rail operates is Dublin-Cork: trains depart every hour, and the line’s 266 km are covered in 2h15 by the fastest train. That’s an average of 118 km/h, which looks less impressive when you realise it’s a nonstop service. 80 km of track between Hazelhatch and Portlaoise are being upgraded in order to reduce travel times by 15 min, for an astonishingly small cost of 10 M€ (125k€ per km) [2]. The Dublin-Galway IC service is also decent, with hourly service, and a best journey time of 2h11 for 208 km of track (95 km/h).

Things start looking ugly when you consider the other lines. Dublin-Waterford is only operated 7 times a day, at irregular hours [3]. The fastest train is 2h03 for a distance of about 170 km (83 km/h). Luckily, there’s no motorway competition, so driving isn’t that much faster (1h52 according to Google). A major problem for improving travel times here is that Kilkenny, the one major town on the way, has a rail station in an awkward stub-end location, forcing any through-trains to reverse after getting there, which can easily add 5 minutes to any stop. And since there are no tail tracks, trains must approach very slowly because overshooting the platform would mean an instant crash. All in all, any stop in Kilkenny must delay a train by approximately 10 minutes, over and above the delays of a regular stop. But stopping in Kilkenny is necessary, since it’s such a major town.

The coastal Dublin-Rosslare line is an absolute disgrace. Trains average only 53 km/h, it’s not much of a surprise then that service is limited to four daily departures. Dublin-Sligo is barely better at 7 daily trains covering the 210 km in just under 3h (70 km/h). The fastest Dublin-Belfast service runs at 93 km/h but it’s apparently plagued by delays, and there isn’t even a train every hour. But improving this line could prove trickier than the others because it would depend on Northern Irish cooperation.

I expected the non-radial lines, which don’t serve Dublin at all, to be the worst, and Limerick-Waterford certainly fits the bill. There are only two trains every day, and they don’t even leave from Limerick proper, but from rural Limerick Junction. The average speed is a paltry 65 km/h.

On the other hand, Galway-Athenry-Ennis-Limerick has apparently seen some success lately. The Athenry-Ennis segment was reopened recently, at a cost of 110 M€ for about 60 km of track. One thousand daily passengers use it [4], or three times more than what the government projected. The line is single track and unelectrified and built on an old rail alignment, but the cost per km (a hair under 2 M€) is undeniably low, even if the cost per passenger, at 110k€, is high [5].

So, except for Dublin-Cork and arguably Dublin-Galway, IC trains in Ireland are too slow and infrequent to compete with driving. Major upgrades are necessary if we want a larger user base than just the people who can’t or won’t drive. The good news is that rail upgrades and construction seem to be on the cheap side in Ireland.

The vision

For rail to be faster than driving in Ireland while still being cheap enough to operate, a few general principles have to be followed. Before getting into the nitty-gritty details, I’ll first show an abstract view of what investments/choices need to be made:

  • All lines must be upgraded to a 160 km/h standard (mostly by straightening curves). This is the standard adopted throughout Europe for normal speed rail (as opposed to high speed rail), it’s well-suited for the 15 km stop spacing and for the rolling stock I’m imagining here (see below). Higher speeds might be warranted where it would allow trips to be completed in a bit less than an integer number of hours.
  • The target stop spacing is one every 15 km, on average. Stations may be built or abandoned accordingly. In many lines it’s actually hard to find big enough towns to stop that frequently, so in practice, the interstation will often be higher. But let’s keep 15 km as a convenient number. The goal is for all lines to run only one service pattern [6], instead of running both locals and expresses. Running two service patterns would greatly complicate operations and either double operating costs, or halve the frequency at local stations, to the point where rail would turn unattractive compared to roads.
  • Service should be hourly, and clockface [7]. Distances are generally small in Ireland, once-every-two-hours service wouldn’t compete with cars at all, hourly service is the bare minimum. 18 daily runs would also be desirable, this would allow departures starting as early as 6:00 and ending at 23:00. People would be able to rely on rail for all sorts of trips, not just for commuting but also for shopping, having dinner with family/friends, or an evening out at the pub.
  • The gaps which would not be covered by rail should be covered by rural buses. Routes should not be circuitous like the existing ones, though. Just like for the trains, speed is of the essence here, even if the reasons are slightly different. If a bus is fast enough that it can make a roundtrip from the station to the end of the line in less than one hour, then a single bus can run the line with hourly frequency. Buses should be scheduled to arrive at the rail station just before a train arrives/leaves. Details on how this should be implemented will be left for another section, with concrete examples.


  • All lines must be electrified, in order to use Electric Multiple Unit (EMU) rolling stock. Those are trains in which all cars are powered and help accelerate the train, instead of being pulled by a locomotive. They have much better acceleration/deceleration performance than diesel-powered multiple unit trains, let alone the loco-hauled diesel trains Irish rail runs on the Dublin-Cork line (the Dublin-area DART does run EMUs, and Diesel Multiple Units are used frequently elsewhere). Alon Levy has a handy tool to calculate the stop penalty of a train, i.e. the time it loses coming to a complete stop then accelerating again instead of maintaining full speed. Plugging the numbers for a FLIRT [8] cruising at 160 km/h, we get a stop penalty of 45 seconds, instead of the 120-180 seconds of loco-hauled trains. That’s a gain of 1.5-2 min for every single stop the train makes.
  • Single track or double track? Only DART, the Dublin-Belfast line and the Dublin-Cork line are double tracked (that’s roughly 400 km in total). All the rest (about 2000 km) is single track, with occasional double tracking at stations to allow trains to cross each other. There’s a reason most of the network is single track: the service levels are so abysmally low that double tracking is unnecessary. But even one train per hour per direction is a light enough service to run on single track: trains would only need to cross each other every half-hour, i.e. every 60 km (see calculation of the commercial speed below), i.e. once every 3-4 stops. Note that this would require excellent timekeeping, as any delay on one train would automatically delay the train coming in the other direction.
  • I’m on less firm ground here, but some old rail lines could be reactivated, and a new direct Navan-Dublin line built. At 30k inhabitants, Navan is the biggest Irish town unconnected to the rail network, and it’s only 45 km away from Dublin. The new rail alignment could go along the M3 motorway and join the Dublin-area network at the M3 Parkway stop [9]. The old rail line connecting Sligo to the rest of the West coast could be reactivated, via Claremorris and Athenry [10]. Less convincingly, the Waterford-Wexford link could be reopened, and the Athlone-Mullingar one as well (this last one would allow less circuitous connections between Mullingar and the West Coast, and between Athlone and the towns immediately to the East and North).

To summarise, running EMUs on straightened 160 km/h tracks would allow fast service even with multiple stops. FLIRTs have a stop penalty of 45 s, if the train stays at the stop for an extra 30 s (i.e. a total delay of 75 s per stop), and there is a stop every 15 km, then a commercial speed of 120 km/h [11] is what you get. This includes a schedule padding of a little over 7%. Padding the schedule is necessary because there will be unforeseen delays, and train drivers need a margin for recovery without going at unsafe speeds. 7% is standard in Switzerland.

Line modernisation costs

Let’s estimate the costs of the line upgrades suggested in the last section. Even though Ireland seems to build rail for very cheap, I’m going to assume that the cost of upgrading a double-track electrified line to a 160 km/h standard is the ~8 M€/km that Belgium is spending on its Brussels-Luxembourg modernisation programme. For a single-track line I’m going to assume the cost is ~6 M€/km (the cost must be somewhere between 50% and 100% of a double-track upgrade, 3/4 is an arbitrary pick).

Upgrading and double-tracking the entire 2371 km of the passenger rail network would then cost about 19 B€. If we’re more reasonable and only keep 500 km as double-track, then it would cost about 15 B€. Building an extra 160 km of new lines (Sligo-Athenry, Waterford-Wexford, Athlone-Mullingar) would cost an extra 1 B€. These would be major investments, given that the total Irish government spending is around 30 B€/year. Phasing the spending over 20 years would still cost 800 M€/year, or more than the current Irish Rail operating budget of 477 M€ (see p. 30 here). On the other hand, Ireland had apparently a capital spending budget of 5.3 B€ (including schools, hospitals and transport infrastructure) in 2018. Some of that money is paying for the widening of a motorway. Redirecting that spending and adding a few hundred million more on top isn’t impossible.

Intercity lines and rolling stock needs

Rolling stock needs are determined by the service frequency and how long the travel times are. If a train needs two hours to make the return trip, including turn-around time at the end stations, then only two trains are necessary to provide hourly service. But if the return trip lasts a bit more than 2h, say 2h10, then a third train would be needed, increasing rolling stock costs by 50%. This makes it very important for return trip lengths to be equal to an integer number of hours, i.e. for one-way trip lengths to be equal to a half-integer number of hours. Investments should be focused to make sure this always happens.

I’m proposing 11 different hourly rail services in all of Ireland (excluding the Dublin area local rail) which would provide decent complete coverage of the country. When drafting this plan I took into account the size of the cities and the shape of the network. The lines vary wildly in how many big cities they connect, so different train lengths are warranted, but to simplify operations I’m constraining the trains to be a multiple of 4 cars. This allows Irish Rail to only buy open gangway 4-car trainsets, which could be coupled into 8-car trainsets in the higher demand lines. All the rolling stock would be completely interchangeable, which simplifies maintenance and allows for a quick replacement of any broken down equipment. I’m assuming each car seats 60 passengers.


High definition version of the image here (credit to @simonschre)

1. Dublin (Heuston) – Cork: The local service could be done in ~2h15 at an average speed of 120 km/h, which would leave 15 minutes of turnaround time before the return trip. 5 trainsets would thus be needed. Since this will remain Ireland’s highest ridership line, it deserves an extra express service making the trip in less than 2h. It could stop just twice at Mallow and Kildare, two important stations allowing for many connections. An extra 5 trainsets would be necessary. All should be 8-car trains, of course.

2. Dublin (Connolly) – Galway (via Mullingar): Also an important line, which demands 8-car trains. It’s about 200 km long, so the trip could be done in less than 2h, with 4 trainsets.

3. Dublin (Heuston) – Limerick: Still an important line, 8-car trains. Again, the distance is close to 200 km (4 trainsets), investing enough here so that a one-way trip can be done in less than 2h is very important.

4. Dublin (Heuston) – Waterford: We’re connecting smaller towns now. But since Kilkenny is on the way, 8-car trains still make sense. The route is well under 200 km so a 2h trip time is doable even with the turnback at Kilkenny. 4 trainsets would be needed.

5. Dublin (Heuston) – Westport (via Tullamore): I wanted to send this service via Mullingar because combined with Dublin-Galway it would provide twice hourly trains between Athlone and Mullingar which deserve the frequency a lot more than the much smaller Kildare and Portarlington. But if this train doesn’t go though Kildare, the train stations in Offaly county (including the sizeable Tullamore) wouldn’t be served at all. This line is longer than 200 km so 5 trainsets would be needed. 4 cars each because the towns on the way are small.

6. Dublin (Connolly) – Sligo: It’s another one of those lines which is close to 200 km and where travel times should be kept under 2h so that 4x 4-car trains can provide the service. It’s actually about 215 km long, so that a <2h journey time would demand a more aggressive 180 km/h cruising speed. Stops which require 3-4 tight curves such as Leixlip should be either relocated or dropped.

7. East Coast Intercity: Dundalk – Dublin (Connolly) – Rosslare: This line is very constrained close to Dublin. For tens of km both North and South, it’s squeezed in by the sea and surrounding development. There’s not a lot of open space to straighten the curves and speed up the trains [12]. So even though the total line length is about 250 km, a realistic travel time would be close to 3h. 6x 4-car trains would be needed.

8. West Coast Intercity: Sligo – Galway – Limerick – Cork: This is our first line not serving Dublin. While it’s annoying that two turnbacks are needed at Galway and Limerick, it’s absolutely necessary to have a service connecting all of Ireland’s second-tier cities. Given the long distance and the two stub-end stations in the middle, a total travel time a little under 4h is to be expected. 8x 4-car trains could deliver that.

9. Limerick-Waterford-Wexford: The South Coast also deserves to be well-connected! This would be about 180 km, so 1h30 at 120 km/h. It might be doable with 3 trains, but in any case 4x 4-car trains would suffice. The Limerick Junction approaches would need to be rebuilt in order to allow East-West through service. A timed connection at Limerick Junction with the West Coast IC would allow Cork-Waterford-Wexford journeys to be competitive too. See Cork-Tralee for a discussion of what that entails.

10. Cork-Tralee: Tralee is still an important town by Irish standards, with about 20k inhabitants. It doesn’t get a direct train to Dublin because it would be silly to bring a train all the way from Dublin to Mallow and then divert it away from Cork. But care should be taken to time the connection of this train with the Cork-Dublin service, so that passengers who do wish to travel between Dublin and Tralee can do so with minimal wait at the transfer station. The Tralee -> Cork train should arrive at Mallow shortly before the Cork -> Dublin train, and the Cork -> Tralee train should leave Mallow shortly after the Dublin -> Cork train.

Cork-Mallow is 32 km, which is 12 min at 160 km/h. Let’s add 2 min for the slowdown on both sides and allow a comfortable padding: 14 min. A sample hourly clockface timetable would be the Dublin -> Cork train leaves Mallow at :31, arrives at Cork at :45, leaves at :00 and is back at Mallow at :14 and then goes on to Dublin. Meanwhile the Tralee -> Cork train leaves Mallow at :04, arrives at Cork at :18, leaves at :27, and is back at Mallow at :41. This way Dublin <-> Tralee passengers only have to wait 10 min in either direction.

The total Cork-Tralee travel time is about 1h10, so 3 4-car trainsets are needed.

11. Westport-Ballina: The branch to Ballina is still used for a few passenger services nowadays, it would be a shame to put it to waste! As with the Cork-Tralee service, we should ensure a timed connection with the Dublin-Westport service, at Manulla station. It’s very hard to get the travel time below 30 minutes, especially with the turnback at Manulla, so sadly two 4-car trains are needed to provide this service.

In total, 224 cars need to be in service at the same time in order to provide hourly frequency for all these lines (plus an extra express hourly Dublin-Cork train). Of course trains go through regular maintenance, so an extra number of cars are necessary in order to cover the shifts of those who are out of service. This Alon Levy post claims that for peak-less service, 86% of trains can be available at any time. So Irish Rail needs to buy roughly 260 cars.

Track sharing with DART

I didn’t want to talk about DART because it’s out of scope for a national rail network. It should run a lot more frequently than once an hour. But if Dublin is really ambitious with its local rail frequency, then the track sharing with the national trains might become a problem. Note that by « DART » here I mean all local rail in and around Dublin, within a radius of 30 km, even if it’s not officially called DART. I would include here services to e.g. Maynooth.

All lines using the approach to Heuston are fine. There are currently three tracks, but the right-of-way (ROW) is wide enough that a fourth track could be built. That would allow for a complete separation of local and long-distance traffic and there would be no capacity problems at all.

The Dublin (Connolly)-Mullingar line should also pose no problem. While the ROW is only two tracks wide, it’s possible to separate the long-distance line after Clonsilla. If long-distance trains stop at Broombridge and Coolmine to provide better coverage of the Dublin area, then local trains make only four extra stops: Ashtown, Navan Road Parkway, Castleknock and Clonsilla. The Navan Road Parkway stop is daft and should be eliminated, there is nothing around it except for a parking lot [13], and there is already a park-and-ride at M3 Parkway, this one is unnecessary. So, if the local trains make only three extra stops, and the stop penalty is 1min15s (it actually is lower since the top speed is also lower), then they will only accumulate 3min45s of delay w.r.t the fast service. Local frequency of up to one train every five minutes is thus fine, and I doubt Dublin will ever need more.

The East Coast line is more of a problem. It’s also two-track, and as already mentioned, the ROW is constrained and not easily widened. If 5 min local frequency is targeted, then the fast service can only skip 3-4 stops. This limit is relaxed to 7-8 stop skips if 10 min local frequency is the target.

As an aside, Cork could also have local service to Mallow, Cobh and, as long as we’re dreaming, Youghal. Cork has a beautiful through-running station and it would be a shame to put it to waste. An infill station on the North side of the city would be good to improve coverage. None of this will pose any track-sharing problems, as there is nothing between Cork and Mallow, so both local and long-distance trains would travel at the same speed.

Operating costs and ticket prices

Paying for all the tracks and the rolling stock isn’t enough, running the trains can also be costly. Operating costs can be decomposed into wages, rolling stock depreciation and maintenance, power, track maintenance and other infrastructure. As I’ll show in a minute, in this blog post’s proposal, track maintenance dominates operating costs.

Wages: Each train only needs one driver and one conductor to check passenger tickets. The conductor is necessary because on long-distance trains random spot checks don’t work [14]. From (again) Alon’s blog, rapid transit train operators drive between 450-867 hours of revenue service each year. Here we’re looking at long-distance rather than short-distance rail, so there might be some differences. On the one hand, operators might need to be ferried around far away from their homes if they’re working all over Ireland instead of a specific city, and this will decrease their efficiency. On the other hand, long-distance trips are longer, so this should increase efficiency. The extremely regular timetable and the total absence of peak service also increases efficiency. I’ll assume the positive and negative effects cancel out and that the same range of revenue-hours per driver per year can be expected for Irish Rail. Let’s take 700 hours/year and an average annual wage, including benefits, of 70k € [15], for a nice round 100 € per hour per operator. Since there are two operators per train, that’s 200 € per hour per train. Those two operators also need managers, but let’s say that’s covered by the train’s turn-around time at each terminus station. This isn’t technically revenue service, but I’ll be lumping the two together anyway because this makes all calculations easier.

Rolling stock: Depreciation+maintenance costs around 230k-260k $ per 25-metre car per year (according to the already linked post here). The cars we’re considering are about ~3/4 of that size, the middle of the range cost would then be ~180k € per car per year. With 18 daily departures, and average trips of around two hours, each car will spend around 20h each day on revenue service, except for when it needs to be at the maintenance yard, 86% of the time. The 180k € are thus amortised over 20h x 365 x 0.86 ~= 6200h. That gives an hourly operating cost of ~30 € per car.

Power: Since the trains are going to stop relatively frequently, power requirements will be dominated by the need to bring the train up to its cruising speed over and over again. A car weighing 44 tons with a cruising speed of 180 km/h = 50 m/s has 55 MJ of kinetic energy. With an average interstation of 15 km and an average speed of 120 km/h, the train will stop once every 7.5 min = 450 seconds. The average power requirement is then 122 kW. Over one hour, that’s 122 kWh of energy. Irish households seem to pay 0.20€/kWh, at that price the hourly electricity bill for each car is just under 25 €.

Of course, I did neglect many things here, not just friction but also losses in transmission and engine efficiency. But electric engines are very efficient (over 90% according to Wikipedia), the gains from regenerative braking roughly compensate the losses in transmission, one station every 15 km is probably too much [16], and industrial clients surely pay less for electricity than regular households.

Track maintenance: From Spinetta’s report (p.21) on French rail, SNCF and Deutsche Bahn both spend about 5 billion € each year on track rebuilding and maintenance (with different mixes on rebuilding vs maintenance). Since they both have ~50k km of single tracks, that works out to a neat ~100 k€/track-km maintenance+rebuilding budget. My proposed network would have just about 3000 km of track (remember that double track counts twice), so that’s an annual 300 M€ bill (sanity check: currently the railway infrastructure manager has annual costs of 224 M€, see p. 48 here). Some of that cost should be shouldered by DART, but for now let’s imagine that only the national trains are paying for it.

Remember, from a previous section, that at any one time, 224 train cars are circulating, and they do so for 20h a day. This means that there are ~1.64 million car-revenue hours each year. Divide our annual maintenance bill by that and we get ~185 € per hour per car. This is by far the dominant operating cost, for a 4-car train even the operator wages would amount to only 50 € per car.

Other infrastructure: I’m not sure how much of the cost of the support infrastructure (signaling, electric substations) is included in the track maintenance figure. But station costs are definitely not, so let’s try to estimate those. Most stations should be extremely simple affairs: two high platforms, a few shelters and that’s it. No need for a pedestrian bridge or underpass when only two trains per hour are expected. According to this article, a barebones 100 m long platform should cost 500 k – 1 M$, so let’s round it up to   2 M€ for a two-platform station. Assuming that the maintenance costs are the equivalent of rebuilding the whole station after 20 years, the annual bill is 100 k€. If the station is serviced by just one of the 11 lines I proposed, that’s 18 daily trains in each direction. Over one year, that’s ~13k trains, each of which should then contribute ~7.5 €. As calculated before, trains stop at a station every 7.5 min, or 8 times per hour, so the hourly cost per train is 60 €. Sure, I’m not counting fancier big town stations here, but those tend to see more than two trains per hour, they can offset some of the costs by renting out commercial space, and as mentioned before I’m probably overestimating the number of stations by setting it at one every 15 km.

Total costs: For a 4-car train, the hourly operating cost would then be 200 + 4 x (30 + 25 + 185) + 60 = 1220 €. For an 8-car train, it would be 2180 €. There are probably some factors I forgot about, but unless they come close in cost to track maintenance, they won’t change the conclusion here by a lot.

So how much would each passenger need to pay for the service to break even? First we need to estimate how many passengers would be in each train. Each car seats 60 people. What about the occupancy factor, how many seats will stay empty? An average occupancy of 25% may not seem like much, but it’s realistic. From my personal experience in Switzerland, this seems about right for trains there, when there are more people the website even warns that crowds are expected. And remember, we’re providing no extra peak service at all, and there are many late night trains which will be comparatively empty. 25% occupancy means 15 people per car, or 60 passengers in a short train and 120 in a long one. If we take the 8-car train as our baseline (that’s where most passengers will be!), each of them would need to pay ~18.20 €/h. That’s ~45 € one-way for the 2h30 Dublin-Cork. And wouldn’t you know, Irish Rail currently prices that very trip at 41 € [17].

So, great news: our system is fiscally viable with only modest 10% subsidies, provided the ridership to fill 25% of the trains is there. Is it realistic to expect that many riders? As always, let’s make a rough estimate and see how it compares internationally. We previously calculated that there were 1.64 million car-revenue hours per year, multiply that by 15 passengers per car and there are ~25 million passenger-hours per year. It seems reasonable to assume that average trip length would be one hour, so that’s 25 million passengers per year [18]. Currently, Irish Rail transports 45 million passengers every year, but 20 million of those are on DART, and I suspect a few million more are on local Dublin lines which aren’t currently called DART. In any case, 25 million annual passengers is if anything unambitious, it would mean a bit more than 5 rail trips per resident of the Republic. That compares to 21 trips/person in Belgium, a number which admittedly is mostly made up of commuters to Brussels and Antwerp. Switzerland is of course on another level with 50 annual trips/person, but that’s with a better system than what I’m proposing here.

The further good news is that since our costs are dominated by fixed infrastructure (the tracks), any extra service would be much cheaper to provide. If the trains are popular and it makes sense to go for twice as many trains, the average operating cost for an 8-car train would drop to 1410 €/h, a ~30% reduction. Dublin-Cork fares could then go down to 30-35 € without any subsidies.

Bus connections

I previously emphasised the need to time the connections between trains on different services so that transfering passengers don’t have to wait long. This generalises to a principle called a takt timetable: infrastructure investments should be focused in such a way that trains cross each other at stations, rather than in-between stations. If two trains are in the same station at the same time, passengers, can easily transfer from one to the other. And that time is also the perfect time for the local bus lines to get to the station, so that transfers between all possible services are possible with minimal wait.

With an hourly clockface schedule such as ours, trains on opposite directions will cross each other once every 30 min. Then the ideal location for a transfer station is one which is a multiple of 30 min away from a terminal station such as Dublin or Sligo. Other stations which are located at more awkward distances would get suboptimal service. Let’s consider both cases.

Carlow is about halfway between Dublin and Waterford, a journey which is supposed to take just under two hours. It’s thus the perfect transfer location, which is fortunate since it’s a 24k town with a higher-education centre. Bus lines should ideally be just under 1h long round-trip, so that a single bus can provide hourly service and meet up with the trains every time. If the average bus speed is 25 km/h (doable if it’s mostly in rural roads) and the round-trip 52 min, the bus line can be up to 10.5 km long. Multiples of 10.5 km are also fine (and then you need either N buses to provide hourly service or you accept having only one bus every N hours). Fiddling on Google Maps, I came up with a ~20 km line going to Tullow (a 4k town) via the Wexford Road Business Park, Tinryland, Kilmeany, Rathtoe and the Tullow Business Park. Another ~21 km line to Baltinglass (2k) via Castledermot (1.3k) also seems worthwhile. And I’m sure 3-4 other lines are also defensible. That would require 6 buses for service every two hours, or 12 buses for hourly services everywhere. Let’s go with 10 buses, with 8 on actual service and two spare for breakdowns and maintenance.

In towns which don’t benefit from two trains crossing each other at the station, bus service would be trickier. For trains which are close enough, say the first one comes at :00 and the second one at :15, the bus can idle 15 extra minutes at the station in order to catch passengers from both trains. Then the bus line roundtrip run time should be 45 min or 1h45. If the second train comes a bit later, say at :20, then the ideal roundtrip run time is 1h20 or 2h20, and once out of every two trips the bus should idle 20 min on the other end of the line. If the second train comes at :30 then a single bus with a 1h30 run time can alternate between serving one train and the other, each train would then get the perfect transfer once every 3h. None of this is ideal, but it beats having completely untimed transfers where passengers would need to wait 20-50 min for their train.

Now let’s estimate operating costs. The Brussels public transport operator recently bought 141 articulated hybrid buses for ~650 k€ each, including maintenance. If Ireland went for shorter buses, ~500 k€ each would be reasonable. With a bus lifetime of 12 years, that’s ~420 k€/year for 10 buses. If the buses run 20h per day like the trains, there are ~65k bus-hours per year. Unlike a train, the bus needs only one operator, the driver, so wages would be 6.5 M€/year. Add gas costs and you’re probably up to 8 M€/year total costs. If we go nuts and consider there will be 30 transfer stations like that, then it would all cost ~240 M€/year. I don’t expect any of this to be even close to profitable so almost all of that would need to come from subsidies. As a comparison, rail-only operating costs are about 455 M€/year (see footnote [18]) but as we’ve seen it can mostly be covered by fares.

Phasing the investments

It’s not possible to straighten the curves and electrify all the lines at once, an investment over 20 years would make a lot more sense. Today, the fastest lines are those which connect the biggest cities: Dublin-Cork, Dublin-Galway. At first Irish Rail needs to continue investing in those lines: start with electrifying and upgrading Dublin-Cork, and run the first EMUs there. The displaced diesel trains which are currently running on that line can be used to strengthen services to Galway, Limerick and Waterford. Then upgrading the short segment from Limerick Junction to Limerick proper would allow EMUs to serve Dublin-Limerick too. Then work on Dublin-Galway can start. At this point, there should be plenty of redundant diesel trains ready to be redeployed to lines which don’t go to Dublin at all. A steady trickle of EMUs should be coming as well and replacing the obsolete diesel trains as soon as the line is electrified (this is another reason why spacing the investments over 20 years is a good idea, there are plenty of still usable diesel trains in Ireland which shouldn’t be replaced all at once).

A similar phased approach should be used with the local buses. Send the first few buses ordered to serve ideal transfer stations on fast electrified lines. Then slowly expand the number of transfer stations served that way.


Building a modern, fast, Irish national rail network would demand heavy investments at first, but over the long run operations would require only small subsidies, even with modest ridership. With greater ridership, fares could even be lowered while keeping the network profitable. The connecting bus transfers can’t possibly turn a profit, but they will greatly improve the transit coverage in Ireland, and could well strengthen rail ridership. It will take a minimum of 20 years to make this network a reality, but improvements on the rail lines with the highest ridership should be seen after just a few years.

Compared to what exists today, this would be a truly revolutionary network. Nowadays, if you work in Dublin and have a business meeting in Tullow, you have no choice but to drive (1h30, with some traffic). After the national network is complete, it would be possible to get there by public transit in two hours (1h rail to Carlow + 1h bus to Tullow), multiple times a day. Most people would still drive, but it would no longer be a necessity. For people traveling to bigger towns which are served by a train station, the rail trip would always be faster than driving. Time to scale back investments on motorways.


[1] Unless noted otherwise, all information and pictures here are from Wikipedia.

[2] This figure comes from the reports before the upgrade started, in 2015. But even if it ends up costing twice as much, this is ridiculously low hanging fruit. As a comparison, Belgium is investing 1 billion euros for a time saving of 20 min in the Brussels-Luxemburg line. Though admittedly the passenger traffic is much higher in Belgium.

[3] This is from browsing the Irish Rail website.

[4] From the Irish Times.

[5] Transit journalist and blogger Alon Levy usually says that a reasonable cost for local rail transit is $20k per daily passenger. That figure is presumably higher for longer-distance rail, where passengers are willing to pay more.

[6] With one exception, the Dublin area, with warrants much more frequent commuter rail, running every 10 minutes, like DART.

[7] Clockface just means a train must be scheduled to arrive/leave at always the same amount of minutes past the hour, e.g. 6:34, 7:34, 8:34, etc. We can write :34 for short. It makes it very easy for passengers to remember when their train leaves. Unplanned, spontaneous trips become possible.

[8] An EMU trainset made by Stadler.

[9] Given the closeness to Dublin, this would warrant a DART-like frequency of one train every 20 minutes instead of the hourly service I’m proposing elsewhere. I’ll leave Navan aside in the rest of this post.

[10] Galway is on an awkward spur 25 km to the west of Athenry. Ideally some new tracks would be needed to link Galway to the mainline south of Athenry, to avoid backtracking all the way there again.

[11] That’s just the speed from the beginning to the end of your trip, including any stops.

[12] Though going through the golf clubs on the southern side of Dublin looks tempting…

[13] Another option would be to redevelop the site.

[14] A proper fine would need to be set way too high.

[15] That’s really the average of the driver and the conductor’s wages. So you can imagine the driver getting 80k € and the conductor 60k €.

[16] Seriously, staring at Google Maps, it’s sometimes hard to find worthwile places for the trains to stop at.

[17] I swear this is a happy coincidence, I didn’t contrive to get to about the same number.

[18] This allows for a quick sanity check on operating costs: 25 M * 18.20 € = 455 M€. This is less than the current Irish Rail operating costs (477 M€), but remember I’m excluding all Dublin local rail. The estimate then seems reasonable.

2 réflexions sur “Modernisation of passenger rail in Ireland: a very rough draft

  1. If Ireland is going to upgrade it’s rail infrastructure, re-adjust the geometry, electrify and buy hundreds of new trains, then the first question should be: « *Do we want to switch upgraded lines and newly built high speed lines to 1,435 mm (4 ft 8 1⁄2 in) standard gauge*? »

    The new high speed rail network in Iberia is not built in Iberian 1,668 mm (5 ft ​5 21⁄32 in) but in standard gauge. All existing Iberian gauge tracks that are renovated in Spain and Portugal have tie holes already in place for a later conversion to standard gauge.

    In the long run, it would make sense for Ireland to convert from 1,600 mm (5 ft 3 in) to standard gauge. Want to provide a new high speed service from Dublin to Cork, well Iarnród Éireann can shop for existing rolling stock off the shelf instead of requiring adaptation for their current gauge.


  2. It’s not expensive to adapt rolling stock to a different gauge. The main reason Iberian HSR uses standard gauge is to allow through-running to the French rail network. In Ireland, the through-running argument points to keeping the broad gauge in order to keep RoI-NI compatibility. Making new standard gauge ROWs would also be expensive, especially the city centre approaches. The Spanish can get away with it because their tunnels are really cheap and they have big cities to connect.


Votre commentaire

Entrez vos coordonnées ci-dessous ou cliquez sur une icône pour vous connecter:


Vous commentez à l’aide de votre compte Déconnexion /  Changer )

Photo Google

Vous commentez à l’aide de votre compte Google. Déconnexion /  Changer )

Image Twitter

Vous commentez à l’aide de votre compte Twitter. Déconnexion /  Changer )

Photo Facebook

Vous commentez à l’aide de votre compte Facebook. Déconnexion /  Changer )

Connexion à %s