Mobility in the Lisbon Metropolitan Area (AML) is continuously under debate in the public and also political space. With this analytical work, I intend to contribute to this discussion with data. I looked for the areas with the highest residential density in AML and where it takes too long to walk to a subway or train station.

These were some of the conclusions I came to:

• Apart from Bethlehem and Ajuda, the parishes of the municipality of Lisbon have most of their populations served by train or subway stations;
• The same is not true for other areas in the LMAincluding some areas with population densities similar to those of Lisbon's parishes;
• O The most flagrant case in this respect is the one I called "West Odivelas".where more than 200,000 people live, of which about 60% live more than 20 minutes away from a subway or train station;
• On the other hand, there are success cases such as the Sintra Linewhich aligns almost perfectly with the degree of urbanization along its route - probably because it already existed there as housing was built. It is a good example of "public transport driven development".

Grade: this article has several interactive maps that may take a while to load in some situations, depending on the quality of your internet connectivity.

---

Part 1 - Introduction

Why only train and subway?

It is not by chance that so many medium or large cities bet on trains, surface streetcars, or underground railways. The table below shows the maximum number of people passing per hour on a 3-meter wide lane, using figures taken from page 14 of Global Street Design Guide.

Track type (width: 3 meters)	People per hour
Private Cars	600 – 1 600
Private cars and frequent buses	1 000 – 2 800
Bi-Directional Bike Path	6 500 – 7 500
Buses in BUS lanes	4 000 – 8 000
People walking on the sidewalk	8 000 – 9 000
Light rail or BRT/Metrobus	10 000 – 25 000

Some of these numbers may not be intuitive - for example, a 3-meter wide sidewalk allows 8,000-9,000 people per hour to pass, which is much more than a lane used only by private cars (600-1600 people per hour). Our intuition tells us that the faster we walk, the more people can pass by per hour, but a car takes up much more space than a person on foot. In other words, people on foot walk slower, but they can walk much closer to each other.

It is clear that in certain circumstances the lower capacity of private car transport is justified, for example on the network of national roads and freeways. It makes no sense to think about replacing a highway with sidewalks, because we will never have 8 to 9 thousand people per hour interested in doing the Lisbon-Oporto A1 on foot. However, in an urban environment, the density of people is higher and the average speeds of a car are much lower due to intersections, crosswalks, traffic, etc. In this sense, a stroll through the center of a city often moves more people than a car lane.

On the other hand, according to this same table, the light rail is able to transport, in the same space, about 10 times more people per hour compared to a mobility made in private cars.

Examples in AML

Taking an example from AML itself, the Sintra Line's capacity is at least 16 thousand people per hoursince each train on this line has capacity for 1,600 passengers and at peak hours there are up to 10 trains per hour per direction on this line, according to CP website. To move the same number of people using private cars in an urban environment, you would need to use 10 to 27 lanes... each way!

Another example is the trains of the Lisbon Metro; each train has a capacity of about 1,070 peopleaccording to this Wikipedia article - assuming four non-motor cars and two motor cars per train - and they pass every 4 minutes and 35 seconds on the Blue Line, according to Lisbon Metro's own website. This gives an approximate capacity of 14,000 people per hour for each side.

It is possible that in both cases the actual capacity is higher, since the infrastructure may not be being used to its maximum capacity, for example due to lack of trains, limitations of the signaling system, lack of drivers, etc.

So, what about the buses?

As for buses, they may have the advantage of lower initial investment and greater flexibility if you want to change the route. However, given the absence of dedicated channels (BUS lanes and traffic lights that turn green for a bus whenever it approaches), especially outside the Lisbon municipality, they will have to mix with general traffic. In this case, they allow roughly doubling the number of people moved per hour compared to cars (to 1000 to 2800 people), which is rather less than multiplying it by 10 or more as is achieved with railway channels.

Despite these advantages of the rail mode within metropolitan areas, it is still a mode that requires high upfront investment, especially if it is underground, and whose route is difficult to change. It is therefore important to decide well where it should go so that the investment is justified.

Population density in the LMA

Let's look, then, at population (i.e. residential) density within the LMA. We can't have train stations on every corner, so it makes sense to have them in areas where there is high demand, or where it is anticipated that there will be such demand in the future.

Why density population? Work and study locations also generate travel, right? Quite right. The choice to use population density is just for pragmatism. As far as I know, there is no open data in Portugal about the routes taken by people in their daily commutes. I also don't know any data on job density, which would be important for this analysis.

However, in 2021 Censuswhose provisional data are already available, we have information on the number of residents at the block and neighborhood level. We can analyze the population density of each neighborhood from this data, and look for high density areas that are far away from metro and train stations.

To calculate the population density this process is followed:

• The provisional data from the 2021 Census indicate, for each neighborhood, how many people live there.
• For each of the 2399 neighborhoods in the LMA, the Census Bureau also tells us the map area they occupy, thus allowing us to calculate the area of each neighborhood in square meters.
• Using the resident population of each neighborhood and its area, we can calculate the population density of each neighborhood.

The interactive map above shows population density on a 3D map of the LMA. The higher the height of the buildings, the higher the population density of that neighborhood. You can do zoom with the mouse wheel, left click to move the map, or right click to rotate it. If you hover the mouse pointer over a neighborhood you will see detailed information about that neighborhood - this is Census data.

Knowing where more or fewer people live, we can move on to the next step: include information on the map about whether it is easy or difficult to use rail modes. This is the subject of Part 2.

---

Part 2 - Walking to Railway Stations

At Part 1We looked at a map with the population density of the LMA. Now it remains to find which areas the subway and train are not a practical option. For this we can calculate the walking time to the nearest train station.

For this analysis I looked at the following railway lines, in a total of 155 stations:

• the four lines of the Lisbon Metro;
• the four suburban CP lines within the AMLThe Sado Line (connecting Barreiro to Praias do Sado, passing through Setúbal), the Azambuja Line, the Sintra Line, and the Cascais Line;
• I also added the stations of CP's West Line that are within the AMLalthough they are not suburban services;
• a South Lineoperated by Fertagus, which crosses the 25 de Abril Bridge and connects the municipality of Lisbon to Setúbal, in a different route from the Sado line;
• o Metro Sul do Tejo.

Taking the map from Part 1, we can now add information about the proximity to a train station. The process is as follows:

• o Open Street Map tells us where each train station is located.
• for each of the 2399 Census districts, which are on the map in Part 1I calculated the centroid, that is, the center point of this block.
• finally, I used Open Route Service to calculate the time needed to walk between all the centroids and all the train stations. I saved, for each block, the nearest station and the time it takes to walk to it.

This last step is not unlike what the reader has probably already done using applications like Google Maps or Apple Maps, but it is automated. This allows you to calculate walking times between the 2399 AML neighborhoods and the 155 train stations I mentioned in a few minutes on a modern computer.

On the map above, the shades of green represent the walking time to the nearest train station. There are five shades of green to represent the following ranges:

• 0 to 5 minutes walk - the darkest green
• 5 to 10 minutes on foot
• 10 to 15 minutes on foot
• 15 to 20 minutes on foot
• more than 20 minutes on foot - the lighter shade

Running and doing zoomOn the map we can see several densely populated areas where most neighborhoods are more than 20 minutes walk from train stations.

In the following sections we will look at some of them in more detail, starting with western zone of Odivelaswhich is the subject of Part 3. A Part 4 analyzes the eastern zone of Odivelas together with the South of Loures.

---

Part 3 - West Odivelas

At Part 2 we looked at the areas of high population density and how far they are within walking distance of train stations. We can use this map to look in more detail at certain parts of the LMA. The first zone, Odivelas Oeste, under analysis is marked on the map below in blue. It includes the following parishes:

• municipality of Odivelas: Odivelas (parish); UF Pontinha and Famões; UF Ramada and Caneças (UF = Union of Parishes, created during the administrative reorganization of 2012).
• municipality of Amadora: Slope of the Sun; Water Mine.
• municipality of Sintra: Casal de Cambra.

This area has:

• Area: 36.97 km²
• Resident Population: 212,680 people, of which 125,547 (59%) live more than 20 minutes' walk from a railway station;
• Population density: 5,753 people per km².

A population density of 5,753 people per km² is quite high. For comparison, below is a table with similar information for the 24 parishes of the municipality of Lisbon, from the most dense to the least dense.

line	Parish	Population	Area (km²)	Population density (pax/km²)	Population 20+ min from station	% Population 20+ min station
1	Arroios	33 307	2,127	15 656	0	0%
2	Campo de Ourique	22 146	1,651	13 413	1 329	6%
3	Areeiro	21 167	1,719	12 316	0	0%
4	Penha de França	28 485	2,710	10 510	4 249	15%
5	São Domingos de Benfica	34 081	4,293	7 938	0	0%
6	New Avenues	23 261	2,994	7 769	0	0%
7	Saint Anthony	11 062	1,494	7 404	0	0%
8	Santa Clara	23 650	3,355	7 050	8 061	34%
9	Lumiar	46 338	6,574	7 049	3 132	7%
10	Saint Vincent	13 956	1,985	7 029	0	0%
11	Alvalade	33 313	5,341	6 238	64	0%
12	Campolide	14 794	2,774	5 334	0	0%
13	Marvila	35 482	7,121	4 983	424	1%
14	Help	14 313	2,876	4 977	10 008	70%
15	Beato	12 185	2,483	4 907	641	5%
16	Carnide	18 029	3,688	4 888	3 590	20%
17	Star	20 308	4,603	4 412	0	0%
18	Benfica	35 367	8,023	4 408	1 740	5%
19	Mercy	9 660	2,192	4 407	0	0%
20	Park of Nations	22 382	5,434	4 119	0	0%
21	Olivais	32 184	8,087	3 980	0	0%
22	Santa Maria Maior	10 052	3,011	3 338	0	0%
23	Alcantara	13 852	5,074	2 730	1 991	14%
24	Bethlehem	16 549	10,425	1 587	8 830	53%

The analyzed zone, of the Odivelas municipality, would be between lines 12 and 13 of this table in terms of population density - right in the middle of the table. However, apart from the parishes of Ajuda and Belém (which will also be analyzed in the future), it is easy to see that the Lisbon parishes always have the vast majority of the population less than 20 minutes away from a train station.

Parts of this area of Odivelas are relatively close to Lisbon Metro stations, namely the Yellow Line (to the east) and the Blue Line (to the south). But it is undeniable that most of the population in this area will not use the subway regularly, even if they work or study near a subway station, since the stations are too far from their residence. It is not by chance that 2009 Lisbon Metro expansion plan included expansions serving this area, with extension of the Yellow Line, now planned as surface metro (future Violet Line).

Next, in the Part 4We will look at the eastern part of Odivelas together with the southern part of Loures.

---

Part 4 - Odivelas East and Loures South

At Part 3 We looked at the western part of the Odivelas municipality. Let's now look at the eastern part of this municipality, adding also the southern part of the municipality of LouresThere is some continuity in population density between these two municipalities.

The area under review includes the following parishes:

• municipality of Odivelas: Póvoa de Santo Adrião and Olival Basto UF.
• municipality of Loures: UF Moscavide and Portela; UF Sacavém and Prior Velho; UF Santa Iria de Azóia, São João da Talha and Bobadela; UF Camarate, Unhos and Apelação.

This area has:

• Area: 44 86 km²
• Population: 149,988 people, of which 115,936 (77%) live more than a 20-minute walk from a train station;
• Population density: 3,343 people per km².

This zone is larger than the previous one, and less dense. It is possible to see large areas in its interior with low population density, for example where the Trancão River passes. It is also interesting to verify that the easternmost zone of this area, which is the Santa Iria de Azóia, São João da Talha and Bobadela UF, is even close to a railway line (the Azambuja Line, operated by CP)but the two closest stations, Bobadela and Santa Iria, do not adequately serve the areas where people live.

Since the railway line already exists, one could naively think of making a new station or halt between these two stations. However, no matter how realistic video games about cities are getting, in real life it's not just clicking a button to make a station in the middle of a consolidated urban area. More credible solutions for this UF would be a good bus service, with dedicated channels, and a good cycling network that doesn't exist at the moment (as you can see on the ciclovias.pt).

A curiosity: technically, the municipality of Loures has metro - The border between Lisbon and Loures runs along Rua 1º de Maio, in Moscavide, and two of the entrances to the Moscavide metro station on the Red Line are on the north side of this street, already in the municipality of Loures. However, this map shows that there is a large extension of Loures, with reasonable population density, with no train or subway nearby.

The areas of Portela and Sacavém were also contemplated in the 2009 Metro Expansion PlanIn this case with a new branch of the Red Line. Currently, this expansion is being considered through a surface metro system, the future LIOS Ocidental. Another extension, from the Yellow Line to Infantado, would also serve the municipality of Loures, but not the areas analyzed here. It would pass further north, through Santo António dos Cavaleiros and the city of Loures. This extension is also being studied at the moment, in the future Violet Line already mentioned. Unlike the Western LIOS, this Violet Line already has financing secured.

Now that we have finished looking at these two areas, we can talk about what the next steps would be. This is the subject of Conclusion of this text.

Part 5 - Conclusion

Next steps

There is a lot to go on after this analysis. First, there are other areas of the LMA whose analyses I have not yet published, but which are in progress, including:

• West Lisbon
• Oeiras
• Almada
• Montijo
• Seixal

The analysis can also be extended to other parts of the country, namely the Metropolitan Area of Porto or the Coimbra Region, which has recently been discussing the closure of its most central railway station.

It would also be important to analyze the performance of other parts of the AML mobility system, in particular the performance of buses. Unlike the rail systems reviewed here, buses are generally vulnerable to traffic. Even if there are bus lines serving certain areas, this does not mean that this service has the frequency, punctuality and speed necessary to be competitive with private car use.

Technical details

It would be possible to hand pick neighborhoods one by one to include in these zones, which would allow us to delineate denser areas or areas with a higher percentage of population away from stations. I still chose to use parishes because:

• is simpler to explain which area is under analysis;
• makes it easier for the reader to know which political bodies govern that area (the corresponding Parish Councils and City Councils);
• is less arbitrary.

It would also be possible to correct, for the parishes that include "neighborhoods" on the Tagus River, the respective area values. I chose not to do so for the following reasons:

• Some neighborhoods include large areas over the river but also some land areas, and it is not possible to separate one from the other in the Census data;
• the same reasoning would justify other corrections, such as removing the Lisbon airport area from the parish of Olivais. If we were to go that way, there would be a large number of somewhat arbitrary corrections to make.

The provisional data from the Census 2021 were obtained through INE website.

The walking times were calculated using the Open Route Servicewhich is based on Open Street Map. The Open Route Service allows you to request directions on their website, but we quickly hit the daily limit when trying the number of paths needed for this analysis. So I chose to run on my computer a copy of the Open Route Service that they provide on GitHub. This allows me to calculate as many walking paths as I want, limited only by the computing power of my computer.

Open Street Map was also used to determine the position of each train station. It would be possible to use alternatives such as Google Maps, and its directions API, but I preferred to go with open data and programs that anyone can consult and even correct. Also, the Google Maps API is paid.

Data processing was done using a personal computer and Python software. Most of the numerical calculations were done with GeoPandas and the maps were made using Kepler.glan open geographic data visualization software made by Uber.

---

This work was developed by Miguel Almeida - a data scientist with a great passion for maps that help to understand cities, mobility and sustainability - and was originally published in May 2022 here. It was duly edited by Mário Rui André for publication in Lisboa Para Pessoas.