Transcript Document
Geography and road network vulnerability Erik Jenelius Div. of Transport and Location Analysis Royal Institute of Technology (KTH), Stockholm 1 Aims of the presentation • Study the vulnerability of different geographic regions in Sweden’s road network • Assess the regional equity of the road network in terms of vulnerability • Find properties of geography, network and traffic that explain regional differences, develop proxy variables 2 Road network vulnerability • Vulnerability is a susceptibility to incidents that can result in considerable reductions in road network serviceability (Berdica, 2002) • Typical scenarios: Extreme weather, landslides, major accidents, malevolent attacks • Vulnerability analysis should contain both probability/frequency and consequence • In the following, we focus on consequence (conditional vulnerability) 3 Consequence measure (1) • Considered incident: A single road link is completely cut off/blocked/closed a certain period • Some assumptions: 1. Changes only in route and departure time choices, not in trip generation, destination or mode choices 2. Users choose shortest route 3. Perfect information on the incident 4. Constant demand/hr • Consequence measure: Increased travel time/delayed arrival for car users 4 Consequence measure (2) • Two possibilities during closure: 1. No alt. routes: Users wait until link reopened 2 x t od open Todk xod = demand/hr, topen = closure duration 2 2. Alt. routes: Users take new shortest route, or if better, wait Todk • 2 xod t open 2 k k od xod od t open 2 if odk t open , if odk t open . Value of alt. routes increases with closure duration 5 Regional exposure (conditional vulnerability) • The average-case exposure of a region is the expected consequences for the region of a randomly located link closure • Two variants: 1. User exposure: The average increase in travel time per user starting in the region UE(r ) wk Todk k or d o x or d o t od open 2. Total exposure: The total increase in travel time for all users starting in the region (socio-economic consequence) TE(r ) wk Todk k or d o 6 Regional inequity • Large regional disparities in exposure indicate spatial inequity between users and regions • A measure of equity: Gini coefficient G G = 0: perfect equity G = 1: perfect inequity 7 Case study: Sweden • Two closure durations: 30 minutes and 48 hours • Average-case user and total exposure of every municipality (289) • Network, O-D demand and equilibrium link travel times from SAMPERS • No congestion effects - underestimation in dense areas • 77,769 nodes, 174,046 links, 8,764 centroids 8 The road network Population density Population density 0.3 - 1.1 1.1 - 2.3 2.3 - 10.5 10.5 - 26.4 26.4 - 4021.7 0 100 200 300 400 Kilometers 0 100 200 300 400 Kilometers 9 User exposure 30 mins: G = 0.35 48 hrs: G = 0.64 Long user exposure 3.9 - 15.6 15.6 - 33 33 - 91.5 91.5 - 193.6 193.6 - 1059.2 Short user exposure 1.76 - 5.91 5.91 - 8.05 8.05 - 11.68 11.68 - 20.57 20.57 - 52.37 0 100 200 300 400 Kilometers 0 100 200 300 400 Kilometers 10 Total exposure 30 mins: G = 0.43 48 hrs: G = 0.71 Long total exposure 0.06 - 0.31 0.31 - 0.71 0.71 - 1.19 1.19 - 4.59 4.59 - 28.09 Short total exposure 0.34 - 0.91 0.91 - 1.56 1.56 - 2.54 2.54 - 3.23 3.23 - 31.84 0 100 200 300 400 Kilometers 0 100 200 300 400 Kilometers 11 Proxy variables (1) • What affects the regional disparities in user and total exposure? • Long closure: Location of cut links • Short closure: 1. sparsity of the regional road network 2. average/total initial travel times of the users 12 Proxy variables (2) • Two measures of road network sparsity: 1. Geographic sparsity: 2. Network sparsity: GS r NS r Ar Lr lr r where Ar = surface area, Lr = length of road, lr = average link length, r = links-to-nodes ratio of region r 13 User exposure (30 mins) GPSUE(r ) GSr r Ar r Lr Short user exposure (10-6 h) 60 50 40 30 20 10 0 0 0.5 1 GPSUE (km 1.5 1/2 2 2.5 h) adj R2 = 0.87 14 Total exposure (30 mins) Ar Tr Lr GPSTE(r ) GSr Tr Short total exposure (10-3 h) 35 30 25 20 15 10 5 0 0 500 1000 1500 GPSTE (km 2000 1/2 2500 3000 h) adj R2 = 0.89 15 Conclusions • Considerable regional disparities in exposure and importance, larger for longer closures • Results are robust to change of partition • Interesting topics for further research: • How would congestion effects, more realistic closure probabilities etc affect the results? • Compare with other countries • Universality of proxy variables? 16 Thank you 17