FLOODS AND URBAN CONNECTIVITY: A TOOLKIT FOR PRIORITIZING RESILIENCE INVESTMENTS Demonstration note with case studies from Kinshasa and Kigali The challenge: floods disrupt urban prioritize cost-effective measures for urban resilience? This note connectivity and livelihoods describes an analytical approach that can help prioritize investments in urban transport resilience and public transport (figure 1), while Cities are intricately interconnected socioeconomic systems, also strengthening the economic case for such investments. with transport networks connecting people to their jobs, health and education facilities, and ensuring the smooth functioning of supply chains. When floods happen, they isolate people and firms An analytical solution, in a nutshell from these vital networks, causing cascading disruptions and losses. Such floods are not limited to rare and extreme events. Assessing urban network disruptions and prioritizing Especially in developing country cities, the lack of resilient investments in urban resilience infrastructure systems means that even regular rainfall events— We have developed a replicable novel methodology to: for example, during rainy seasons—can cause havoc. Attention is often biased towards direct asset losses from floods, 1. Document the impacts of flooding on the performance of rather than the wider economic costs of disrupted networks. collective transport systems1 in developing country cities This is due primarily to the complex dynamics of economic and 1 In many African cities, most “public” transport is in fact privately infrastructure networks. But public transport and road usage data owned and operated. In this paper, we refer to collective transport and are also often limited, especially when the predominant modes of public transport, using the latter as an overarching term that includes transport are informal and walking. So how can we identify and privately owned collective transport services. Figure 1. Kinshasa’s most critical transport nodes: candidates for priority resilience upgrades Betweenness based on travel paths ▲ NORTH 0 2.5 5 10 KM Note: Betweeness is an indicator of the criticality, i.e. importance, of a road segment to the network’s overall functionality. RESULTS IN RESILIENCE SERIES 2 | Demonstration note with case studies from Kinshasa and Kigali 2. Measure the consequences of these disruptions on employ- This methodology relies on an innovative, dual condition, ment and key service accessibility transit feed data collection campaign (under flooded and normal conditions) combined with global flood maps, commuter or 3. Estimate the economic costs linked to travel delays household surveys, and open-source information retrieved from 4. Guide local decision makers by identifying the most critical Open Street Map (OSM). The first applications of the methodology links in transport networks that would benefit from cli- focus on Kinshasa, Democratic Republic of Congo and Kigali, mate-proofing, increased maintenance, or resilience up- Rwanda. grades. Key insights: Illustrations from Kinshasa Figure 2. Trip consistency and rerouting in Kinshasa under and Kigali normal and flooded conditions flooded condition route flooded condition route Public transport performance under flooded conditions dry condition route dry condition route To compare public transport performance, we conducted travel surveys under normal (“dry”) and flooded (“wet”) conditions for Kigali and Kinshasa. These surveys include data on stops, schedules, travel times, and routing related to different public transport services, and they can often include supplementary information such as fares. We conducted the wet survey in the height of the rainy season in both cities, within 48 hours of a heavy rain episode. By comparing General Transit Feed Specification (GTFS)2 feeds under normal and flooded conditions, we can document the impact of rain and flooding on three dimensions of public transport performance: trip rerouting, headway a. Esprit de Vie, Rond Point Ngaba— b. Taxi Jaune, Gombe (difference in minutes between two departures along a given Université Pédagogique Nationale Centre Ville—Victoire route)/frequency, and travel speeds. de Kinshasa Figure 2 shows examples of itinerary changes between normal c. Esprit de Mort, flooded condition route and flooded conditions in Kinshasa, with similar behaviors Gombe Centre Ville— dry condition route identified in Kigali. While most itineraries remain identical DGC whatever the conditions, others can change slightly (panels a and d) or substantially (b and c) when heavy rain occurs, as drivers adapt trips to avoid the worst-affected road segments. These sudden changes in itinerary impact on travelers’ ability to reach their jobs and other destinations. Some transport services, such as those provided by Kinshasa’s formal bus company TRANSCO, were suspended under heavy rain conditions. Heavy rain episodes can also reduce travel speeds. In Kigali, d. Taxi Jaune, speeds were reduced by 1.8 kilometers per hour on average (an flooded condition route Gombe Centre Ville— 8 percent reduction); in Kinshasa, they were reduced by 1–4 dry condition route Velodrome kilometers per hour, depending on service type. Headways also increased in both cities, with the difference in minutes between two departures along a given route rising by approximately 28 percent on average in Kigali, and 30 percent (Esprit de Vie vehicles) and 75 percent (Taxi Jaune vehicles) in Kinshasa. Note: Esprit de Vie, Esprit de Mort, and Accessibility of employment and other services Taxi Jaune are modes of public transport in The combination of rerouting, cancelled trips, skipped bus stops, the city. slower travel speeds, and increased headways leads to travel delays for commuters trying to reach different parts of both cities. This in turn results in lower accessibility of employment 2 A set of text files that describe particular aspects of transit information: stops, routes, trips, and other schedule data. Floods and urban connectivity: a toolkit for prioritizing resilience investments | 3 Figure 3. Accessibility of jobs: Difference between normal and flooded conditions in Kigali Focus area Study boundary Official Kigali boundary Reduction in jobs accessible within 60 minutes (%) 10 km 2 km 1 Akamatamu 4 Kacyiru 7 Kimironko 10 Nyamirambo 12 Nyarugenge 14 Ruyenzi 2 Batsinda 5 Kicukiro 8 Kinyinya 11 Nyanza 13 Remera 15 Taba 3 Gisozi 6 Kigali International 9 Nyabugogo Airport Note: Blue cells indicate locations where the share of jobs accessible within 60 minutes decreases by up to 80 percent. Red cells indicate locations with increased job accessibility due to flood-induced rerouting. opportunities from public transport, or the share of jobs that can accounting for flood probabilities and aggregating for the 10, 50, be reached within 60 minutes. This is a commonly used metric to 100, 500 and 1000-year flood events, the average opportunity cost assess urban labor market physical integration. per flood day represents $1.2 million. In flooded conditions, average accessibility of jobs reachable in The average figures are large. They also hide important spatial 60 minutes falls from 34 to 25 percent in Kigali and from 20 to and socioeconomic variations that we documented. Figure 4, 15 percent in Kinshasa. This represents a 25 percent decrease for instance, shows travel delays and associated opportunity in both cities, in flood events typical of 5- and 10-year return costs broken down by income group in Kinshasa for five return periods, respectively. These decreases get larger for less frequent periods. On average, more extreme floods with a longer return and more intense flood events as more road links are flooded period induce longer travel delays and higher economic costs and travel is increasingly impeded. Figure 3 also shows sizeable compared with those with a shorter return period. It is clear from spatial heterogeneity in accessibility losses. figure 4 that commuters with medium income levels (monthly salary of $100–1,000) experience higher travel delays than those Economic costs and heterogenous socioeconomic on a low (under $100) or high (over $3,000) incomes. It is also clear impacts of travel delays that higher income groups incur the highest opportunity costs in terms of time, despite their comparatively small travel delays. With information about commuters’ incomes and their residential Note, however that different metrics such as opportunity costs and work locations, we can calculate travel delays from flood in relation to income would show higher economic impacts for disruptions and estimate a lower-bound economic cost of floods the poorer income groups. Estimated travel delays for the lower by calculating their opportunity costs at city scale. We find income groups are more extreme (over one hour), suggesting that in Kinshasa, travel delays alone represent $5.4 million in the existence of highly disadvantaged commuter groups who are opportunity costs per flood day for a typical 10-year event. When heavily impacted by flood disruptions on road networks. 4 | Demonstration note with case studies from Kinshasa and Kigali Figure 4. Histogram of travel delays and estimated economic loss in Kinshasa, by income level, under five flood scenarios 800 Home-to-work travel delay 600 (minutes) 400 200 0 Less than $100 $100–199 $200–299 $300–399 $400–499 $500–999 $1,000–1,499 $1,500–1,999 $2,000–2,999 $3,000 and above 0 Home-to-work travel cost 20,000 10-year flood 50-year flood 40,000 ($) 100-year flood 500-year flood 60,000 1000-year flood 80,000 Monthly income level Investment prioritization: starting from the most resilience of these points will yield the largest reductions in critical points economic disruption and flood loss. Our work does not stop at assessing the physical and economic Figure 5 shows the most critical nodes in Kigali’s public transport consequences of floods on transport systems and accessibility; system, which would create the largest travel delays if they become we also consider how to prioritize resilience upgrades to reduce impassable due to heavy rains or other exogenous factors. Six of disruptions most effectively. In Kinshasa and Kigali, we identify Kigali’s 13 most critical nodes are at risk of flooding. Figure 1 (page the most critical transport links that are at risk of flooding and 1) highlights Kinshasa’s most critical road segments. Upgrading the that create the most extensive travel delays for commuters. In most critical nodes should be considered a priority for resilience other words, we identify the road segments and intersections that investments. With additional information about resilience upgrade are responsible for the highest travel delays, and should therefore options, the system costs from flood-induced travel delays could be be prioritized for upgrade. Targeted investments to increase the compared to costs of climate-proofing road segments. Figure 5. Weighted betweenness centrality for Kigali’s public transport network Legend Study boundary Official Kigali boundary Nodes 50 most critical nodes Betweenness 1 km 5k 10k 15k 20k Betweenness 5k 10k 15k 20k 10 km 1 km Note: KN = Kigali Nyarugenge; KG = Kigali Gasabo; RD = Road; RN and NR = National Road; CBD = Central Business District Floods and urban connectivity: a toolkit for prioritizing resilience investments | 5 Figure 6. Overall project workflow from data collection to processing and analysis Transported feed data Road network data Flood maps Household and (GTFS) (OSM) (Fathom) commuter surveys Public transport mapping under normal and flooded Walk and drive Flood depths and extents Travel patterns and travel conditions networks for multiple return periods modes + income data Public transport mapping under normal and flooded conditions Flood impacts Accessibility under normal Critical road segments Economic costs due on transit patterns and flooded conditions identification to travel delays Overall, this information can help local decision makers channel 4. A commuter travel survey with origin-destination information their resources—for example, for public transport investment of commuters’ trips (and ideally, travelers’ socioeconomic and road maintenance—to where they will deliver the highest attributes)3, or a firm census or business registry dividends in ensuring smooth urban mobility under normal and flooded conditions. Challenges in replication and available support The Global Facility for Disaster Reduction and Recovery’s (GFDRR) Methodology and data Global Programs on Resilient Infrastructure, Disaster Risk Analytics, In many developing countries, and Africa in particular, collective and Resilient Health Systems can provide support in implementing and informal transport modes—such as matatus in Kenya, tap taps network analysis approaches to quantify the wider economic in Haiti, tro tro in Ghana, and dala dala in Tanzania—are the most impacts of disasters and identifying investment or policy options to used means of travel besides walking. Recent initiatives (such as mitigate their costs and harness the dividends or resilience. Digital Matatus) that survey these transport modes in urban areas The GFDRR team can also advise teams on the procurement in Africa, Latin America, and the Caribbean document how these process to contract firms for the travel surveys and the GTFS data systems operate and offer crucial information to mainstream collection. accessibility analyses. Yet, coverage is far from universal. And even less is known about how these systems are impacted by disasters. Finally, the GFDRR team is available to discuss with teams whether the approach proposed in this short note is the most appropriate To overcome such data shortfalls, we combine four key datasets for the task at hand and can suggest alternative methods if it is not. to assess the impact of floods on urban transport in Kinshasa and Kigali: Contacts and additional information 1. An innovative GTFS dataset collected for these analyses The interested reader is welcome to explore our journal paper on under normal and flooded conditions for collective modes of Kinshasa4 and our blog post summarizing its findings.5 transport For more information, or if you are interested in applying this 2. OSM transport network vector data methodology to your projects or analyses, please feel free to 3. A set of high-resolution global flood maps capturing the contact: extent and depth of pluvial and fluvial floods, which can be • Paolo Avner, Urban Economist, GFDRR: pavner@worldbank.org supplemented with coastal floods as needed. When local flood maps are available, they should be preferred as they are • Jun Rentschler, Senior Economist, Office of the Chief Economist typically of higher quality and resolution and can account for for Sustainable Development: jrentschler@worldbank.org local conditions, such as drainage capacity. 3 Note that a commuter survey with Origin/Destination pairs and socioeconomic information is not mandatory but will considerably enrich analyses by allowing for a more meaningful criticality analysis and the economic evaluation of travel delay costs. In the absence of such surveys, a dataset with information about the spatial distribution of jobs will be enough to measure the impacts of floods on accessibility and a more standard criticality analysis can be performed. 4 He, Y, Thies, S, Avner, P and Rentschler, J. 2021. “Flood impacts on urban transit and accessibility—a case study of Kinshasa.” Transportation Research Part D: Transport and Environment, 96 (102889). https://doi.org/10.1016/j.trd.2021.102889 5 Avner, P and Rentschler, J, He, Y and Thies, S. 2021. “When floods hit the road – impacts of rainy season flooding on transit and commuting in Kinshasa.” World Bank blog, February 8. https://blogs.worldbank.org/developmenttalk/when-floods-hit-road-impacts-rainy-season-flooding-transit-and-commuting- kinshasa.