Abstract
In this work, we show the importance of considering a city’s shape, as much as its population density figures, in urban transport planning. We consider in particular cities that are “circular” (the most common shape) compared to those that are “rectangular”: for the latter case we show greater utility for a single line light rail/tram system. We introduce the new concepts of Infeasible Regions and Infeasibility Factors, and show how to calculate them numerically and (in some cases) analytically. A particular case study is presented for Galway City.
Similar content being viewed by others
Change history
13 July 2020
The equation at the beginning of Sect.��4.4, and also Eq.��(45) at the end of Appendix B, should both be replaced by IF��=��23/72.
Notes
So, the results here will not apply in “large” cities, where there are skyscrapers/tower blocks/large apartment blocks in city centres. They will also not apply in small cities in countries where there is a tradition of people living in apartment blocks in city centres (much of continental Europe, for example). But our results will apply in smaller cities in USA, UK, Ireland, for example, which generally do not have large apartment blocks in their centres.
For example: by foot, \(\bar{s}_{\mathrm{nt}} \approx 0.1\) km/min, by bicycle \(\bar{s}_{\mathrm{nt}} \approx 0.2\) km/minute. On the Paris metro, \(\bar{s}_{\mathrm{t}} \approx 0.5\) (measured by the author on line 4, between Jussieu and Mairie d’Ivry, November 2018), while the Dublin Luas has \(\bar{s}_{\mathrm{t}} \approx 0.28\)
We do not consider any details of city topography, road layout, physical geography (river, etc. ) here, and leave it to others to take these in to account. Because our layout models link square kilometre areas, they do not have fine grain detail, and so leave room for the actual line to be placed within 100 metres of the centre of each square in our grid, without affecting substantially the details of our calculations.
References
Alagar VS (1976) The distribution of the distance between random points. J Appl Prob 13(3):558–566. https://doi.org/10.2307/3212475
Aldous D, Barthelemy M (2019) Optimal geometry of transportation networks. Phys Rev E 99(5):052303. https://doi.org/10.1103/PhysRevE.99.052303
Barthelemy M (2016) The structure and dynamics of cities: urban data analysis and theoretical modeling. Cambridge University Press, Cambridge
Burago D, Burago Y, Ivanov S (2001) A course in metric geometry. American Mathematical Society, Providence
Census 2016 reports (2016a). https://www.cso.ie/en/census/census2016reports/. Accessed 10 Jan 2019
Census 2016 results (2016b). http://airomaps.nuim.ie/id/Census_2016/P2_PopDist/. Accessed 10 Jan 2019
Dan C (2011) EU population 2011 by 1 km grid. https://dancooksonresearch.carto.com/maps. Accessed 18 Oct 2018
Derrible S, Kennedy C (2009) Network analysis of world subway systems using updated graph theory. Trans Res Rec 2112(1):17–25. https://doi.org/10.3141/2112-03
García-Pelayo R (2005) Distribution of distance in the spheroid. J Phys A Math General 38(16):3475–3482. https://doi.org/10.1088/0305-4470/38/16/001
Klain DA, Rota G-C (1997) Introduction to geometric probability. Cambridge University Press, Cambridge
Louf R, Roth C, Barthelemy M (2014) Scaling in transportation networks. PLoS ONE 9(7):e102007. https://doi.org/10.1371/journal.pone.0102007
Mathai AM, Moschopoulos P, Pederzoli G (1999) Random points associated with rectangles. Rendiconti del Circolo Matematico di Palermo 48(1):163–190. https://doi.org/10.1007/BF02844387
Okabe A, Suzuki A (1997) Locational optimization problems solved through Voronoi diagrams. Eur J Oper Res 98(3):445–456. https://doi.org/10.1016/S0377-2217(97)80001-X
Okabe A, Boots B, Sugihara K (1992) Spatial tesselations: concepts and applications of Voronoi diagrams. Wiley, New York
Acknowledgments
Thanks to Ulf Strohmayer for pointing this out.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendixes
A Examples of (\(\alpha\),\(\beta\))–infeasible regions
To help the reader appreciate the role of the tram speed, and the frequency of service, on the infeasible regions, we show in Fig. 9 some further plots of infeasible regions for some realistic values of \(\alpha\) and \(\beta\) and departure point (8, 3). We remind the reader that \(\alpha = \bar{s}_{\mathrm{t}}/\bar{s}_{\mathrm{nt}}\) measures the (relative) tram speed while \(\beta = t_f/2\) measures the frequency. Note
-
Fig. 9a corresponds to the extreme case scenario where the tram does not move (its speed is zero) and/or the time interval between trams is infinite. In this (ridiculous!) situation, obviously the infeasibility factor is \(100\%\) (there is no reason to take such a tram!)
-
Plots (c) and (d) of Fig. 9 are very similar: (d) has higher speed trams, while (c) has lower speed trams but more frequent ones.
B Infeasibility factor for square city
Without loss of generality, we restrict ourselves to points in the upper right quadrant, i.e. \(0\le x_1 \le a\) and \(0\le y_1 \le a\). For points in this quadrant, the boundary of the infeasible region is parabolic (as described in Sect. 4.3). Depending on the point \((x_1,y_1)\) chosen, the parabolic boundary may intersect the city boundaries along the horizontal line \(y=a\) or along the vertical lines \(x=-a\) or \(x=a\). Points that are close to the tram line (i.e. small values of \(y_1\)) will give rise to “narrow” parabolas (i.e. with small latus rectum), while larger values of \(y_1\) will give “wider” parabolas. There are three possibilities:
-
(i)
Both intersection points are along \(y=a\) (“narrowest” parabolas);
-
(ii)
One intersection point is along \(y=a\) and one is along \(x=a\);
-
(iii)
One intersection point is along \(x=-a\) and one is along \(x=a\) (“widest” parabolas).
We examine these three cases separately.
- Case (i):
-
From Eq. (24), the parabolic boundary is \(y = (x - x_1)^2/4y_1\). The right arm of this parabola intersects the vertex (a, a) when \(a = (a - x_1)^2/4y_1\), i.e. when \(y_1 = (a-x_1)^2/4a\). For fixed \(x_1\), values of \(y_1\) less than \((a-x_1)^2/4a\) will give (narrow) parabolas that intersect only with \(y=a\). This region is thus defined by \(0\le x_1\le a\) and \(0\le y_1\le (a-x_1)^2/4a\). The area of the infeasible region (bounded by the parabola and the line \(y=a\)) is
$$\begin{aligned} A1(x_1, y_1) = \int _{y=0}^a\int _{x=x_1-2\sqrt{y_1y}}^{x_1+2\sqrt{y_1y}} dx dy = \int _0^a 4\sqrt{y_1y} dy = \frac{8\sqrt{a^3y_1}}{3}. \end{aligned}$$(38)The average value of this area is thus
$$\begin{aligned} {\overline{A1}} = \frac{\int _{x_1=0}^a\int _{y_1=0}^{(a-x_1)^2/4a} A1(x_1, y_1) dy_1 dx_1}{\int _{x_1=0}^a\int _{y_1=0}^{(a-x_1)^2/4a} dy_1 dx_1} = \frac{a^4/18}{a^2/12} = \frac{2a^2}{3}. \end{aligned}$$(39) - Case (ii):
-
In this region, the right arm of the parabola intersects the vertical line \(x=a\) while the left arm intersects the horizontal line \(y=a\). As \(y_1\) increases, the left arm eventually intersects the vertex \((-a,a)\) when \(a = (-a-x_1)^2/4y_1\), i.e. \(y_1 = (a+x_1)^2/4a\). This region is thus defined by \(0\le x_1\le a\) and \((a-x_1)^2/4a\le y_1\le (a+x_1)^2/4a\). The area of the infeasible region (bounded below by \(y = (x - x_1)^2/4y_1\), to the right by \(x=a\) and above by \(y=a\)) is
$$\begin{aligned} \begin{aligned} A2(x_1, y_1)&= \int _{y=0}^{(a-x_1)^2/4y_1}\int _{x=x_1-2\sqrt{y_1y}}^{x_1+2\sqrt{y_1y}} dx dy + \int _{y=(a-x_1)^2/4y_1}^{a}\int _{x=x_1-2\sqrt{y_1y}}^a dx dy\\&= a\left( a-\frac{(a-x_1)^2}{4y_1}\right) + \int _{y=0}^{(a-x_1)^2/4y_1} (x_1+2\sqrt{y_1y}) dy \\&\quad - \int _{y=0}^a (x_1-2\sqrt{y_1y}) dy\\&= (a-x_1)\left( a-\frac{(a-x_1)^2}{4y_1}\right) +2\sqrt{y_1} \left( \int _{y=0}^{(a-x_1)^2/4y_1} + \int _{y=0}^a \right) (\sqrt{y}) dy\\&= a(a-x_1+4\sqrt{ay_1}/3) - (a-x_1)^3/(12y_1). \end{aligned} \end{aligned}$$(40)We average to get
$$\begin{aligned} {\overline{A2}} = \frac{\int _{x_1=0}^a\int _{y_1=(a-x_1)^2/4a}^{(a+x_1)^2/4a} A2(x_1, y_1) dy_1 dx_1}{\int _{x_1=0}^a\int _{y_1=(a-x_1)^2/4a}^{(a+x_1)^2/4a} dy_1 dx_1} = \frac{a^4(1+6\ln {2})/9}{a^2/2} = \frac{2a^2(1 + 6\ln {2})}{9}. \end{aligned}$$(41)
- Case (iii):
-
Here, the right arm of the parabola intersects with the vertical line \(x=a\) and the left parabola arm intersects with \(x=-a\) (we have a “wider” parabola compared to the other cases). The region is defined by \(0\le x_1\le a\) and \((a+x_1)^2/4a\le y_1\le a\). The area of the infeasible region (bounded below by \(y = (x - x_1)^2/4y_1\), to the right by \(x=a\), to the left by \(x=-a\) and above by \(y=a\)) is
$$\begin{aligned} A3(x_1, y_1)= & {} \int _{x=-a}^a\int _{y=(x-x_1)^2/4y_1}^{a} dy dx = 2a^2 - \frac{1}{4y_1}\int _{x=-a}^a (x-x_1)^2 dx =2a^2 \nonumber \\&- \frac{a(a^2 + 3x_1^2)}{6y_1}. \end{aligned}$$(42)Averaging gives us
$$\begin{aligned} {\overline{A3}} = \frac{\int _{x_1=0}^a\int _{y_1=(a+x_1)^2/4a}^{a} A3(x_1, y_1) dy_1 dx_1}{\int _{x_1=0}^a\int _{y_1=(a+x_1)^2/4a}^{a} dy_1 dx_1} = \frac{a^4(2+6\ln {2})/9}{5a^2/12} = \frac{8a^2(1 + 3\ln {2})}{15}. \end{aligned}$$(43)
Taking the weighted average over the three cases gives
Since the city has area \(4a^2\), this gives a final (dimensionless) Infeasibility Factor of \({\overline{A}}/(4a^2)\), or
C Infeasibility factor examples
We present in Fig. 10 plots of IF(p) for various values of \(\alpha\) and \(\beta\). Our city here has an area of 300 square kilometres, and is a rectangle of size \(10 \times 30\). Figure 10a and f present the extreme case scenarios with fewest/most infeasible journeys, respectively. As we would expect, points at right angles to the tram line show highest Infeasibility Factors for example points (5, 0) or \((-5,0)\) in Fig. 10b–e (Fig. 10a is identical to Fig. 6, where a different coloring scheme is used for the contours).
In Fig. 11 we present the corresponding contour plots for a circular city of approximately equal area (radius 10 km).
D Case study of Galway City
We present here a few generic schematicFootnote 3 suggestions for the layout of a single line light rail/tram for Galway City. Each suggestion differs by taking in to account in greater detail the variations of population density, and building a successively longer line. So in Fig. 15 our (short) line just links the highest population densities (from Fig. 14), leading eventually to Fig. 19 which is the longest line, taking into account all the data (Figs. 16, 17, 18).
Figure 13 shows the overall population densities, drawn to scale, with some important points of interest indicated. Figure 12 shows the single tram line, connecting highest density areas, superimposed on a GOOGLE map of Galway City. Figure 20 presents data from the CSO (Central Statistics Office, see Census 2016 reports 2016a), via AIRO (All-Island Research Observatory, see Census 2016 results 2016b), showing the linear/rectangular nature of (the population distribution of) Galway City.
Rights and permissions
About this article
Cite this article
Mc Gettrick, M. The role of city geometry in determining the utility of a small urban light rail/tram system. Public Transp 12, 233–259 (2020). https://doi.org/10.1007/s12469-019-00226-9
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12469-019-00226-9