Transit-Oriented Development Urban Spatial Forms and Typhoon Resilience in Taipei: A Dynamic Analytic Network Process Evaluation

Abstract

Taipei’s metropolitan region faces frequent typhoon impacts that test its urban resilience. This study examines the relationship between Transit-Oriented Development (TOD) urban spatial forms and Taipei’s resilience against typhoons, considering both physical urban morphology and planning factors. We apply a Dynamic Analytic Network Process (DANP), an integrated DEMATEL-ANP multi-criteria approach to evaluate and prioritize key resilience-related spatial and planning factors in TOD areas. Rather than using GIS flood modeling, we emphasize empirical indicators derived from local data, including urban density, transit accessibility, historical typhoon flood impacts, infrastructure vulnerability, and demographic exposure. An extensive literature review covers TOD principles, urban resilience theory, and DANP methodology, with a particular emphasis on the Taiwanese context and case studies. Empirical results reveal that specific TOD characteristics indeed enhance typhoon resilience. High-density, mixed-use development around transit can reduce overall exposure to hazards by curbing sprawl into floodplains and enabling efficient evacuations. Using DANP, we find that infrastructure robustness and emergency planning capacity emerge as the most influential factors for resilience in Taipei’s TOD neighborhoods, followed by land use and management and transit accessibility. Weighted rankings of Taipei’s districts suggest that centrally located TOD-intensive districts score higher in resilience metrics, while peripheral districts with flood-prone areas tend to lag. The Discussion explores these findings, considering planning policies—noting that TOD can bolster resilience if coupled with adaptive infrastructure and inclusive planning—and compares them with examples like Singapore’s integrated land use and transit strategy, which dramatically reduced flood risk. The study concludes with policy implications for integrating TOD and climate resilience in urban planning, and contributions of the DANP approach for complex urban resilience evaluations.

1. Introduction

Taipei, the capital of Taiwan, is a dense urban metropolis regularly exposed to typhoons and intense rainfall events. On average, about 3–4 typhoons per year strike Taiwan, often bringing extreme winds and torrential rain [1]. The city’s geographical setting—a low-lying basin encircled by hills—makes it naturally prone to flooding when typhoon rains overwhelm drainage capacity. Major historical typhoons such as Typhoon Nari (2001) inundated large parts of Taipei; Nari’s overnight deluge flooded downtown areas, submerged 16 metro stations, and knocked critical infrastructure offline. These events highlighted vulnerabilities in Taipei’s urban form and systems, prompting efforts to enhance flood defenses and emergency response capabilities. At the same time, Taipei has embraced Transit-Oriented Development (TOD) for decades as a core planning strategy to foster sustainable urban growth. As early as the 1980s, city regulations provided floor-area ratio incentives for developments within 500 m of transit stations, effectively concentrating new growth around the expanding Taipei Metro network. This TOD approach has reshaped Taipei’s urban morphology into a network of high-density nodes clustered around metro stations, with the aims of reducing car dependence and preserving surrounding hillsides and farmland.

These two parallel narratives of typhoon resilience challenges and TOD-driven urban forms converge on a critical research question: How does a TOD-oriented urban spatial structure influence a city’s resilience to typhoon impacts? Urban resilience refers to the capacity of an urban system to withstand, adapt to, and recover from external shocks like natural disasters [2]. It depends on both the physical infrastructure and social systems of the city. In the context of typhoons, key aspects of resilience include minimizing exposure of people and assets to hazards, reducing vulnerability, and enhancing adaptive capacity. Taipei’s experience has shown that traditional engineering approaches. Studies have warned of a “levee effect” in Taipei, where heavy flood-control infrastructure led to a false sense of security and even encouraged intensive development in flood-prone zones, paradoxically increasing potential losses. For example, Su [3] documents how Taipei’s reliance on costly engineering solutions enabled new residential/commercial projects on low-lying lands, raising overall vulnerability despite reduced flood frequency. This underscores the need for a more integrated resilience approach that includes land use planning and spatial design, not just engineering defenses.

Transit-Oriented Development offers a potential pathway to such integration. By promoting compact, mixed-use centers around transit, TOD inherently directs growth away from exurban sprawl—which often occurs on marginal lands like floodplains and unstable slopes—and into established urban areas that typically have better infrastructure and are easier to protect. TOD also enhances mobility and connectivity, providing robust public transport networks that can be crucial during disaster evacuations or post-disaster recovery [2]. Dense transit-served neighborhoods concentrate populations in areas where emergency services and shelters can be more efficiently provided and accessed. Indeed, global examples suggest TOD and resilience can reinforce each other: Hong Kong, a quintessential TOD city, experiences high exposure to typhoons yet sustains relatively low disaster losses per capita—in part due to strong building codes and the concentration of development on well-protected urban land [4]. Research indicates that high-density development can coincide with lower flood losses when accompanied by resilient infrastructure and planning. Ref. [5] found that compact urban growth patterns in flood-prone regions tend to reduce per capita flood damage compared to dispersed growth, since infrastructure can be fortified and maintained more effectively for concentrated assets. Conversely, unplanned sprawl at city peripheries often encroaches on hazard-prone areas and stretches infrastructure thin, increasing both exposure and susceptibility to disasters [3,5].

Despite these theoretical linkages, there is a knowledge gap in empirically quantifying the relationship between TOD spatial forms and disaster resilience. Most TOD literature focuses on mobility, environmental sustainability, or economic outcomes [6], while resilience studies often treat urban form in broad strokes without specifically examining TOD patterns. Especially in the Taiwanese context, where both TOD implementation and disaster management have evolved rapidly, there is a need to analyze how TOD-designated areas and transit-served districts perform in terms of resilience indicators. Taipei presents a living laboratory to explore questions such as: Are the dense inner-city districts built around MRT lines more resilient to typhoon flooding than the less dense outskirts? Does proximity to transit infrastructure help or hinder communities during typhoon events? What spatial and planning factors are most critical in making TOD neighborhoods resilient?

To address these questions, this study uses a multi-criteria analytic approach. We employ the Dynamic Analytic Network Process (DANP), which combines DEMATEL (Decision-Making Trial and Evaluation Laboratory) and ANP (Analytic Network Process) methods, to evaluate a network of interrelated resilience factors in TOD areas. This allows us to incorporate the complex interdependencies among factors like land use, transportation, infrastructure, environment, and socio-demographics, rather than treating them in isolation. The DANP has been applied in urban resilience assessments to identify causal relationships and weight factors [7,8]. For example, a recent Taiwanese study by Wang et al. [8] used the DANP to assess industrial park resilience and found that governance and planning indicators had the highest influence. Building on such methodologies, our study adapts DANP to the TOD-resilience nexus in Taipei. We assemble empirical data on relevant indicators, including urban density, transit accessibility, historical flood incidence, infrastructure robustness, and population vulnerability for Taipei’s districts, distinguishing those with strong TOD characteristics. By engaging local experts and using DANP, we systematically prioritize which factors most enhance or undermine resilience in TOD environments.

The remainder of this paper is structured as follows. First, we present a Literature Review covering TOD principles, urban resilience concepts, previous findings on urban morphology and disaster impacts, and applications of DANP in urban planning research. Next, the Methodology section details our research design, data sources, and the DANP framework employed. The Results and Analysis section then reports the findings, including comparative data tables of TOD and resilience indicators across Taipei’s districts, the influence weights of factors from the DANP, and interpreted rankings of district resilience. In the Discussion, we examine the implications of these results, how TOD appears to affect typhoon resilience in practice, and relate them to planning policy, citing local initiatives and international examples. Finally, the Conclusion distills the key insights for both academic understanding and practical policymaking, acknowledging limitations and suggesting directions for future research on planning for resilience in a changing climate. This study adds explicit value in three respects: (i) an integrated TOD–resilience conceptualization tailored to Taipei’s basin setting; (ii) a transparent DANP workflow (from expert elicitation to normalized weights) that city agencies can replicate; (iii) a district-level benchmarking dashboard (Resilience Index vs. TOD Index) that pinpoints priority actions for Wanhua, Datong, Neihu, and Nangang.

We conducted a narrative literature review to synthesize scholarship at the intersection of transit-oriented development, urban morphology, and disaster resilience. Searches were performed in Web of Science, Scopus, and Google Scholar using combinations of keywords such as “transit-oriented development”, “urban morphology”, “typhoon”, “flood resilience”, “nature-based solutions”, and “DEMATEL/ANP/DANP” over the period 2000–2025. Peer-reviewed journal articles, scholarly books and chapters, and official statistical yearbooks for Taipei were included; non-refereed web sources were excluded. Screening focused on studies with empirical evidence or methodological clarity relevant to Taipei or comparable high-density Asian contexts. The resulting corpus (≈120 items) was thematically coded into the key themes.

2. Literature Review

2.1. Transit-Oriented Development (TOD) and Urban Spatial Form

Transit-Oriented Development (TOD) is a planning concept that promotes compact, high-density, mixed-use urban development within walking distance of public transit stations [9]. TOD aims to synergize land use and transit infrastructure: concentrating housing, jobs, and services around transit nodes to encourage transit use and walkability, thereby reducing automobile dependence and urban sprawl. Typical characteristics of TOD include a dense network of transit stations, mixed land use zoning (residential, commercial, and public amenities integrated), pedestrian-friendly design, and limited parking supply to discourage driving [10]. By design, TOD produces a nodal urban morphology: city development is channeled into nodes and corridors aligned with transit lines, in contrast to the dispersed, car-oriented sprawl pattern. Many Asian cities have embraced TOD to accommodate growth in a sustainable manner. Taipei has one of the world’s highest urban densities (~9600 people/km² citywide, with some districts over 20,000/km²) and has integrated transit planning with land use incentives from early on. Yen et al. [11] document that Taipei’s TOD strategy comprises four development aspects, ten strategies, and 32 planning principles, including using land value capture through Floor Area Ratio bonuses to fund transit and steer private development towards station areas. As a result, Taipei’s urban core today is a lattice of MRT-served sub-centers, such as Xinyi, Daan, Zhongzheng, each with intense development around metro stations, while hills and outer areas are comparatively less built-up.

Nature-based solutions can be systematically embedded in TOD station-area design to complement gray infrastructure and deliver co-benefits. Examples include linear park-swales along transit corridors, permeable public plazas around stations, and retention wetlands integrated into district parks; these measures attenuate peak runoff, increase infiltration, and provide micro-climate benefits while enhancing public realm quality. Recent syntheses conclude that such measures substantially contribute to flood mitigation and urban resilience in compact cities [12].

The spatial outcomes of TOD can have significant environmental and risk implications. By limiting urban encroachment into peripheral high-risk zones (like floodplains, steep slopes), TOD can help protect natural buffers and reduce exposure to hazards [13]. A study on climate adaptation in Asian cities noted that TOD encourages clustering in lower-risk locations, enabling cities to keep fragile ecosystems intact and avoid putting housing on the most flood-prone lands. Additionally, urban density achieved via TOD can enable more efficient infrastructure provision—for instance, floodwalls or stormwater systems can be built to protect a larger number of people in a compact area, yielding economies of scale in resilience investments [14]. Empirical evidence also suggests that dense urban cores often receive higher levels of protection and experience faster post-disaster recovery than scattered rural developments due to their economic importance and infrastructure concentration. However, high density must be accompanied by strong building standards and infrastructure design to avoid creating vulnerability “hotspots.” For example, if high-density TOD nodes are in inherently hazardous sites without adequate protective measures, they could concentrate on risk. Urban design within TOD is therefore crucial—features like elevated infrastructure, floodproofing of basements, and incorporation of parks or retention ponds can mitigate the side effects of paving over land. Many TOD plans now include green and blue infrastructure as part of station-area development, partly to manage runoff and heat.

Terminology clarification: in this paper “hard” measures refer to conventional engineered systems constructed from rigid materials (e.g., drainage networks, levees, floodwalls, pumping stations); “soft” measures refer to green or nature-based solutions that work with ecological processes (e.g., rain gardens, bioswales, green roofs, urban wetlands); and “non-structural” measures denote policy, planning, and behavioral instruments (e.g., land use zoning, early-warning systems, emergency response protocols, and community preparedness).

Another aspect is social and institutional. TOD often leads to socio-economic changes in neighborhoods, which can have resilience ramifications. On one hand, TOD areas tend to attract investment and better services, potentially strengthening community resources for disaster response [15]. On the other hand, if TOD pricing pushes lower-income residents to cheaper, hazard-prone outskirts, the city’s social vulnerability pattern might worsen. Equitable TOD (or “eTOD”) initiatives seek to ensure affordable housing is included in station areas so that vulnerable groups benefit from the safer, accessible locations [16]. A noteworthy example is Singapore, where public housing has been heavily developed in TOD fashion along transit lines, providing safe, modern residences for lower-income households. This has resilient benefits: safe housing means reduced disaster casualties, and the high-quality infrastructure in these neighborhoods reduces service disruptions. As noted in one study, when TOD is implemented with inclusivity, it “relocates vulnerable groups into safer, structured environments,” directly reducing their disaster risk [17].

In summary, TOD reshapes urban spatial form in ways that can be leveraged for resilience—if aligned with resilient design principles. The literature indicates potential co-benefits: emissions reduction and climate mitigation alongside climate adaptation and resilience. However, these benefits are not automatic. They depend on integrated planning—transit plans must be coordinated with land use zoning and hazard maps to ensure new stations/hubs are sited in safe zones or built to withstand floods, and resilience must be factored into the design of buildings and infrastructure in TOD areas (e.g., waterproof underground metro entrances, as Taipei did post-Nari. Absent such integration, dense TOD areas could conversely become points of high accumulated risk. This study’s focus on Taipei allows for the examination of a city that has extensive TOD implementation and is simultaneously grappling with typhoon resilience, providing a rich context to explore these dynamics.

2.2. Urban Resilience to Disasters and Typhoons in Cities

The concept of urban resilience has evolved from early ecological notions [18] to a broad interdisciplinary framework encompassing engineering, socio-economic, and institutional dimensions [19]. In disaster risk literature, resilience is often defined as the ability of an urban system to absorb impacts, maintain core functions, and recover rapidly from shocks [2,20]. Key components of resilience include robustness, redundancy, resourcefulness, and adaptive learning [2]. For typhoon resilience specifically, cities must contend with multiple hazard agents: strong winds, extreme rainfall, and storm surge if near coasts. Taipei’s primary risks from typhoons are rainfall-induced floods and landslides, as the city is inland. Storm surge is a minor issue for Taipei but relevant for Taiwanese port cities like Kaohsiung. Thus, urban flood resilience is a central concern. It involves both hard measures, as drainage infrastructure, flood barriers, and pumping stations, and soft measures like land use planning, warning systems, emergency response, and community preparedness.

A resilient city proactively limits its exposure and vulnerability to hazards. Exposure refers to the people and assets located in harm’s way, while vulnerability is the susceptibility of those exposed elements to damage [21]. Resilience initiatives can reduce exposure by steering development out of hazard zones, for example, by zoning floodplains as parks, and reduce vulnerability by upgrading construction standards and preparedness. Taipei’s authorities have indeed taken such steps, especially after severe events. For instance, following Typhoon Nari’s devastation in 2001, Taipei undertook a major overhaul of flood control standards: the MRT’s flood design return period was raised from 100 to 200-year events plus a safety margin, floodgates and pumps were upgraded citywide, and a “Sponge City” program launched in 2015 to enhance natural stormwater absorption in the urban area. These efforts align with what the literature terms engineering resilience, aiming for resistance and quick bounce-back.

However, recent research emphasizes that resilience is also fundamentally about urban governance and planning [22,23]. Cities that treat resilience as a holistic goal tend to incorporate it into development planning and community design, not just emergency management. This means considering how street networks might perform during evacuations, how land use affects runoff patterns, or how socioeconomic disparities influence post-disaster recovery. In Taipei’s case, the city’s urban planning and disaster management have started to converge. The concept of “spatial planning for resilience” has gained traction, as highlighted by Su [3]: rather than relying solely on “fightthewater” engineering, planners are urged to focus on wise land use and environmental management to create a city that can coexist with floods. This echoes the emerging paradigm of living with water, promoted by authors like KueiHsien Liao in Taiwan [24], where urban design adapts to accommodate floodwaters in controlled ways through retention basins, green waterways, and flexible public spaces, instead of trying to fully exclude water. Taipei’s Shezidao area—a low-lying river peninsula—has served as a testbed for such ideas, with debates continuing over whether to fortify it, redevelop it, or restore it as a floodplain [25]. Typhoon resilience metrics for a city can include indicators across multiple domains: (i) physical infrastructure, e.g., drainage capacity (mm/hr), number of critical facilities (hospitals, power substations) in flood zones, building integrity (percentage built to modern wind-resistant codes); (ii) environmental, e.g., percentage of permeable surface or green space (which can absorb rain), extent of floodplain restoration; mobility/accessibility, e.g., redundancy of transport network (if one corridor is down, are there alternatives?), public transit uptime after storms, evacuation route coverage; (iii) social/demographic, e.g., population segments that are vulnerable (elderly living alone, low-income households in unsafe housing), community networks (volunteer groups for disaster response); (iv) institutional/governance, e.g., disaster response plans, early warning systems, inter-agency coordination efficiency, public risk awareness, and drills participation.

Because resilience is multifaceted, frameworks like the UN City Resilience Profiling Tool or the 100 Resilient Cities indicators encompass dozens of such metrics. In academic studies, researchers often use composite indices or multi-criteria analysis to evaluate urban resilience. Cutter et al. [26], for example, developed a Disaster Resilience Index. In Taiwan, recent research has applied multi-criteria approaches. Wang et al. [8], using Fuzzy Delphi and DANP, identified five dimensions—vulnerability, urban environment, industrial environment, factory properties, and governance and adaptation—as affecting industrial park resilience, with governance and adaptation emerging as the most crucial dimension. They found that in areas such as Shilin, housing prices increased after flood events, suggesting that flood risk was perceived as manageable, while in Daan, an increase in flood depth led to price decreases, indicating greater sensitivity. These findings hint at differing resilience perceptions and realities across districts, possibly tied to local mitigation measures and historical experience.

For transit systems, resilience is about maintaining mobility during crises. After Hurricane Sandy in 2012 flooded New York’s subways, many cities recognized transit as both a vulnerability and a key asset if managed well. New York’s post-Sandy strategy included plans to deploy buses and ferries when subways shut down. Following Typhoon Nari’s flooding in 2001, Taipei’s MRT invested heavily in flood protection, backup power, and pumping systems. A resilient transit-oriented city ideally fosters a “safe-to-fail” network—where failure in one part, such as a line closure, can be compensated by redundancies in routes or modes [27]. Research on multimodal network resilience suggests that diverse modes and interconnections enhance the ability to reroute and maintain service [28]. In Taipei, the dense MRT grid in the city core combined with an extensive bus system provides redundancy, though some peripheral areas still rely on single lines and remain more vulnerable.

In sum, urban resilience to typhoons involves a tapestry of factors. The literature underscores that urban morphology and spatial planning significantly influence resilience [29,30]. Cities built with redundancy in networks and clustering of assets in safer areas tend to fare better, whereas those with unplanned expansion into hazard zones suffer higher impacts [5]. Taipei’s case—a high-density city with advanced infrastructure but constrained by its basin geography—provides a crucial example for examining how TOD-driven form correlates with these resilience factors. Does the clustering from TOD align with the “clustering in safe zones” ideal mentioned by theorists? Are Taipei’s transit-served districts equipped with better flood measures and emergency resources than car-oriented fringe districts? Our study engages these questions through an empirical, data-driven approach described next.

2.3. Evaluating Resilience and Urban Form via DANP Methodology

Assessing the relationship between TOD spatial form and resilience requires analyzing multiple interrelated criteria. Traditional analysis might look at pairwise correlations, such as density versus damage, or compare case studies, but to capture the network of influences, a multi-criteria decision-making (MCDM) approach is valuable. We employ the Dynamic Analytic Network Process (DANP), which is a hybrid method combining DEMATEL and ANP techniques. DEMATEL, developed by Gabus and Fontela in 1972, is used to elucidate causal links among factors by constructing an influence matrix from expert judgments [31]. It provides for each factor a measure of how much it influences others (denoted d) and how much it is influenced (denoted r), thereby classifying factors into cause or effect groups and highlighting feedback loops. ANP [32] generalizes the Analytic Hierarchy Process to allow dependence among criteria, using a network structure and eigenvector calculations to derive relative weights of factors. By integrating DEMATEL’s influence relationships into the ANP weighting, DANP can allocate weights accounting for the interdependencies [7]. In effect, DANP results not only in a ranked importance of factors but also provides insight into the causal structure of the system—identifying which factors are “driving” resilience and which are “driven.”

This approach has seen growing use in urban resilience research. For instance, Hemmati et al. [5] applied a similar DEMATEL–ANP process to evaluate city resilience indicators and emphasized that it captures real-world complexity better than independent weighting methods, since criteria like land use and infrastructure clearly influence each other. In Taiwan, Wang et al. [8] applied DANP in the context of industrial areas to determine indicator weights, revealing, for example, that emergency response planning and management organization were top indicators influencing overall resilience. These examples validate DANP’s applicability for complex urban systems with feedback loops, where improvements in one dimension (e.g., infrastructure) may reduce vulnerability in another (e.g., social factors).

The methodology will illustrate our use of DANP, starting from indicator selection (grounded in the literature and local data), then applying DEMATEL to uncover the network of influences among indicators and finally calculating weights Via ANP for evaluation. By using DANP, we aim to identify which spatial or planning factors associated with TOD contribute the most to urban typhoon resilience in Taipei. We expect factors such as infrastructure robustness, land-use management, and emergency response capacity to emerge as important, consistent with the literature, where network analyses often show infrastructure and governance indicators as central [8,26]. At the same time, our specific focus allows us to include novel indicators related to transit and urban form, such as transit accessibility or population near transit, and see how they figure into the resilience network.

In summary, the literature provides a foundation for our analysis by (1) outlining how TOD shapes urban spatial characteristics that could affect resilience; (2) defining key dimensions of urban resilience (physical, social, institutional) to consider; and (3) offering a methodology (DANP/MCDM) to integrate those dimensions in an evaluative framework. Our study builds on this foundation to empirically analyze Taipei’s TOD–resilience relationship, which, to our knowledge, has not been comprehensively studied before. The next section details the data and method for this analysis.

As summarized in

Table 1. Key Themes from literature on TOD, urban form, and resilience.

Theme	Key Points from Literature	Representative References
Transit-Oriented Development (TOD) and Urban Spatial Form	TOD emphasizes compact, high-density, mixed-use development around transit nodes promotes walkability, reduces car dependence, and channels growth into nodal patterns Spatial outcomes include efficient infrastructure use, protection of peripheral hazard zones, but risks of vulnerability hotspots if not well planned.	Calthorpe [9]; Dittmar & Ohland [10]; Yen et al. [11]; Ribeiro & Pena [13]); Sharifi [14]; Li & Lai [33]; Lin & Li [34]
Environmental and Risk Implications	Protect natural buffers by limiting encroachment into floodplains and steep slopes. Concentrate assets so resilience investments (e.g., drainage tunnels, flood barriers) achieve economies of scale. Avoid density-driven hotspots by coupling compact growth with resilient urban design and maintenance. Integrate nature-based solutions (rain gardens, bioswales, permeable pavements, wetlands) to absorb and delay runoff and to cool microclimates.	Ribeiro & Pena [13]; Sharifi [14]; Ferreira et al. [12]
Social and Institutional Dimensions	TOD reshapes neighborhood socioeconomics can enhance resources and services but risks displacement of vulnerable groups Equitable TOD (eTOD) ensures affordable housing near transit, reducing social vulnerability.	Noland, R.B [15]; Mineta National Transit Research Consortium [16]; Nafi et al. [17]
Urban Resilience Concepts	Resilience entails robustness, redundancy, resourcefulness, and adaptive learning For typhoons, measures include drainage, flood barriers, warning systems, and preparedness Taipei adopted stronger standards after Typhoon Nari, aligning with engineering resilience.	Holling [18]; Bruneau et al. [2]; UNISDR [20]; Kron [21]
Governance and Planning for Resilience	Resilience is not only technical but institutional Integrating spatial planning with disaster management builds long-term resilience Living-with-water paradigm (e.g., Shezidao debates) highlights adaptive approaches.	Campanella [22]; Leichenko [23]; Su [3]; Liao [24,25]
Resilience Indicators and Evaluation	Frameworks (e.g., UN City Resilience Profiling, 100RC, Cutter’s Disaster Resilience Index) provide metrics across physical, social, environmental, and institutional domains In Taiwan, MCDM/DANP analyses highlight governance and adaptation as central.	Cutter et al. [26]; Wang et al. [8]
Transit and Network Resilience	Transit is both a vulnerability and an asset Post-disaster redundancy and multimodal networks improve resilience Taipei’s MRT–bus system provides redundancy, though peripheral reliance on single lines is a weakness.	Cats [27]; Arora et al. [28]
Methodological Framework (DANP)	DANP integrates DEMATEL (causal structure) and ANP (weighting with interdependencies) Applied in resilience studies worldwide and in Taiwan, DANP is well-suited for analyzing TOD–resilience linkages.	Gabus & Fontela [35]; Saaty [21]; Chen & Chen [7]; Hemmati et al. [5]

, the reviewed literature can be organized into eight key themes, ranging from TOD spatial form and its environmental implications to governance, resilience indicators, and methodological approaches. The synthesis illustrates how TOD, while fundamentally a land use and transportation strategy, intersects with disaster risk reduction, social equity, and institutional capacity. In the case of Taipei, these interconnections are particularly salient: the city’s high-density TOD pattern both raises questions about potential concentration of risk and offers opportunities for coordinated resilience planning. The table thus establishes a conceptual foundation for the empirical analysis that follows, highlighting the significance of examining the TOD–resilience relationship in a systematic and data-driven manner.

3. Methodology

3.1. Study Area and Data Collection

This research focuses on the Taipei metropolitan area, specifically the districts of Taipei City, to examine the interplay between TOD spatial form and typhoon resilience. Taipei City is divided into 12 administrative districts, each with distinct urban characteristics. Central districts such as Daan, Songshan, Zhongzheng, Zhongshan, and Xinyi are relatively small in area but very dense, with populations exceeding 20,000 people per km² in several cases. These areas embody Taipei’s TOD policy: they contain many MRT stations, high development intensity around those stations, and mixed land use. In contrast, peripheral districts such as Shilin, Beitou, Neihu, Nangang, and Wenshan are larger in area, including hilly terrain, and have more uneven development patterns. Their densities are lower, typically between 4000 and 10,000 people per km², and their transit stations, while present, are fewer and more linearly distributed.

Table 2. Taipei City districts: area, population (circa 2023), density, and simplified land use context.

District	Area km2	Area Share %	Population 2023	Pop Share %	Density p/km2	Density Rank	Density Class	Built Up Share Qual	GreenBlue Share Qual
Beitou	56.82	20.91	242,379	9.34	4329	12	Low	Low	High
Daan	11.36	4.18	291,687	11.39	26,406	1	High	High	Low
Datong	5.68	2.09	119,957	4.79	22,175	2	High	High	Low
Nangang	21.84	8.04	113,893	4.33	5219	10	Low	Medium	Medium
Neihu	31.58	11.62	276,173	11.01	9184	8	Medium	Medium	Medium
Shilin	62.37	22.95	266,841	10.90	4602	11	Low	Low-Medium	High
Songshan	9.29	3.42	193,453	7.75	21,966	3	High	High	Low
Wanhua	8.85	3.26	173,431	7.33	21,803	4	High	High	Low
Wenshan	31.51	11.59	260,298	10.60	8855	9	Medium	Low-Medium	High
Xinyi	11.21	4.12	206,666	8.55	20,077	6	High	High	Low
Zhongshan	13.68	5.03	216,359	7.98	15,350	7	Medium	High	Low
Zhongzheng	7.61	2.80	150,154	6.04	20,902	5	High	High	Low

compiles verifiable district-level facts used for orientation in this study area: administrative area (km²), population (circa 2023; rounded to the nearest thousand), and population density (persons per km²), alongside each district’s share of city area and population, the rank and class of density (High ≥ 20,000; 8000–<20,000 = Medium; Low < 8000), and a simplified land use context (qualitative built-up and green/blue shares).

This tabular presentation matches the categorical information actually used in our DANP evaluation, increases reproducibility, and avoids over-interpretation that may arise from a schematic map. As seen in the table, Daan, Datong, Songshan, Wanhua, Zhongzheng, and Xinyi fall in the High-density class, whereas Nangang, Shilin, and Beitou are Low-density due to larger hillside or peri-urban extents; Zhongshan, Neihu, and Wenshan are Medium-density. These gradients align with the TOD morphology described in the following subsections and help contextualize exposure and infrastructure load in later analyses.

Data quality control. All indicators were harmonized to a 2020–2024 reference window; units were standardized, outliers cross-checked against official yearbooks, and missing values were imputed using district means with expert verification. A metadata sheet documents sources, units, and directionality.

To conduct empirical analysis, data were collected from multiple local sources across several dimensions. Population and area statistics were obtained from Taipei City statistical abstracts and open data, allowing computation of population density as a baseline indicator of urban form. Land use maps were reviewed to identify the proportion of each district that is urbanized compared with green space or water bodies. Transit accessibility was measured by compiling the number of MRT stations per district and calculating station density as a proxy for accessibility. Central districts such as Zhongzheng and Daan exhibit high station densities due to multiple intersecting lines, while peripheral districts like Shilin and Beitou have much sparser MRT coverage.

Flood incident records were gathered from Taipei’s Hydraulic Engineering Office and Disaster Management Office, including major typhoons such as Nari (2001) and Soudelor (2015). Case studies on flood-prone areas, such as Shezidao in Shilin and inundated parts of Neihu in 2013, provided qualitative insights into district-level flood risk. Data on hospitals, evacuation shelters, drainage pumping stations, and flood control structures were collected from Taipei Open Data and planning documents. Riverside districts such as Shilin, Datong, and Zhongshan are protected by levee walls, while low-lying areas like Neihu and Nangang rely on clusters of pumping stations. Reports on the resilience of electricity and water systems, including the presence of backup generators, were also incorporated.

Demographic vulnerability was assessed using census data, including indicators such as the proportion of elderly residents, household income levels, and car ownership rates. Wanhua District, for example, has one of the highest elderly ratios in Taipei, suggesting greater vulnerability during disasters. Information on community-based disaster preparedness, including the presence of volunteer response teams, was also considered as part of the city’s broader resilience assessment.

All data collected were for recent years (mostly 2020–2024) to reflect current conditions. In terms of typhoon exposure, districts along Taipei’s river system inherently face more flood risk. For example, Shilin District, which includes the Shezidao floodplain and areas along the Keelung River, has a history of flooding and can be considered high exposure. Nangang and Neihu in the eastern basin also contain significant low-lying sections that have been inundated during past cloudburst rains. By contrast, districts on slightly higher ground, such as Daan and Xinyi on the basin’s southeastern edge, have experienced fewer flood incidents. Based on topography and historical events, each district’s flood exposure was qualitatively categorized as Low, Medium, or High. These categorizations will subsequently feed into the multi-criteria evaluation.

3.2. Selection of Evaluation Criteria for DANP

Based on the literature review and data collected, we defined a set of evaluation criteria that link TOD spatial forms to resilience outcomes. The criteria were organized into broader dimensions to structure the analysis. The selection process combined top-down (theory-driven) and bottom-up (data-driven, local knowledge) approaches. We started from dimensions commonly identified in urban resilience frameworks, such as environmental, infrastructure, social, and institutional, and tailored them to our focus on TOD by adding specific spatial criteria such as transit access. We also consulted with five local experts, including two urban planning scholars, one transportation engineer, and two disaster management officials, through a questionnaire to validate and refine the criteria. Expert panel and elicitation. Ten local experts (2 urban planning scholars, 2 transport engineers, 2 hydraulic/flood managers, 2 emergency management officers, 2 community leaders) completed a structured DEMATEL questionnaire using a 0–4 influence scale. A two-round Delphi alignment was used to resolve large discrepancies; responses were anonymized and summarized back to experts for confirmation before aggregation.

The final criteria and their grouping are summarized in

Table 3. Evaluation dimensions and indicators for TOD-related typhoon resilience in Taipei.

Dimension	Indicator Code	Indicator	Description
A. Land Use and Environment	A1	Population Density	Represents urban compactness. High density may facilitate efficient protection and services, but if unmanaged, can concentrate risk.
	A2	Development in Hazard Zones	Share of development/population in flood-prone areas. TOD districts are hypothesized to have less sprawl into unsafe areas, thus lower exposure.
	A3	Green Space Ratio	Share of permeable/green land area. Green areas function as sponges for rainfall and wind buffers, enhancing resilience.
	A4	Building Standards	Average building age or % of structures built post-1990 under stricter wind and seismic codes.
B. Infrastructure and Mobility	B1	Transit Network Accessibility	Station density or % of population within 500 m of a transit stop, reflecting mobility options for evacuation and aid delivery.
	B2	Redundancy of Transport Routes	Extent of alternative routes/modes if one fails (e.g., MRT line vs. buses). Central districts typically score higher.
	B3	Drainage and Utility Infrastructure	Stormwater drainage capacity (mm/h) and coverage of flood-control systems (e.g., pumps, levees).
	B4	Critical Facilities Coverage	Number of emergency shelters and hospitals per 10,000 residents. Higher coverage facilitates response and recovery.
C. Emergency Management and Social Factors	C1	Disaster Planning and Response Capacity	Quality of district-level emergency plans, drills, and community-based response teams.
	C2	Population Vulnerability	Composite indicator of elderly share, low-income households, etc. Higher vulnerability reduces resilience.
	C3	Social Capital	Strength of local networks and volunteerism. Traditional neighborhoods may have tighter community cohesion.
	C4	Awareness and Education	Public risk awareness and participation in disaster education programs.
D. Governance and Planning	D1	Land Use Management	Enforcement of zoning in risky areas and resilience-oriented planning tools.
	D2	Investment in Resilience Measures	Recent investment in floodwalls, pumping stations, retrofits, etc.
	D3	Coordination Capacity	Inter-departmental coordination (transport, water, emergency). Newer districts with smart city systems may have stronger integration.

, which presents the four major evaluation dimensions, namely Land Use and Environment, Infrastructure and Mobility, Emergency and Social Factors, and Governance and Planning, along with their 14 specific indicators. These criteria were selected to capture how TOD-related spatial characteristics, infrastructure, social attributes, and governance practices collectively influence Taipei’s resilience to typhoons. Each indicator provides either a measurable or qualitative aspect of resilience that is subsequently weighted and prioritized in the DANP analysis.

These criteria form an indicator network that links TOD aspects (A1, A2, B1, etc.) with resilience outcomes. They are not independent, e.g., higher transit access (B1) might reduce population in cars during evacuations, interacting with emergency response (C1), or more green space (A3) can lessen load on drainage infrastructure (B3). This interdependence justifies the use of DANP.

3.3. Dynamic Analytic Network Process (DANP) Application

Our application of DANP involved several structured steps, aligning with standard procedures from prior studies [7,8]. Figure 1 provides a flowchart of the process. In summary. Figure 1. Flowchart of the Dynamic Analytic Network Process (DANP) methodology applied in this study, consisting of (1) indicator selection based on literature and expert input; (2) DEMATEL analysis to construct a network of influence among the indicators; (3) ANP weight calculation considering those interdependencies; and (4) evaluation and ranking of resilience for the TOD areas (Taipei districts) using the derived criteria weights. This approach captures the complex interactions between spatial, infrastructural, and social factors influencing urban resilience.

This flowchart illustrates the methodological steps applied in the study. The process begins with indicator selection, based on literature review and expert input, followed by DEMATEL analysis to map causal influences among indicators. The results inform the ANP weight calculation, which accounts for interdependence across criteria. Finally, the weighted indicators are used for the evaluation and ranking of resilience across TOD-related districts in Taipei. This structured approach captures the complex interactions between spatial, infrastructural, social, and governance factors influencing urban resilience.

• Step 1: Criteria Network Construction

Using the criteria defined above, we structured an influence network. We prepared an expert survey (questionnaire) to quantify the pairwise influence of each criterion on others, following DEMATEL methodology. Ten experts (including urban planners, transportation experts, disaster management officials, and academics familiar with Taipei’s urban context) were asked to rate on a scale (0 to 4, from “no influence” to “very high influence”) how much each criterion influences each other criterion in terms of resilience. For example, they would judge how “Transit Network Accessibility (B1)” influences “Emergency Response Capacity (C1)”, or how “Land Use in Hazard Zones (A2)” influences “Flood Infrastructure Load (B3)”. The survey included explanations of each criterion to ensure consistency. We then averaged the scores across experts to obtain an initial influence matrix Z.

• Step 2: DEMATEL Analysis

From the initial influence matrix Z, we computed the normalized direct-influence matrix and the total influence matrix following standard DEMATEL steps [36]. This yielded, for each criterion, an aggregate influence score dᵢ and dependence score rᵢ.

Two key indices were then derived as follows:

• Influence degree: dᵢ = Σⱼ tᵢⱼ
• Dependence degree: rᵢ = Σⱼ tⱼᵢ
• Prominence: Pᵢ = dᵢ + rᵢ
• Relation: Rᵢ = dᵢ − rᵢ

Prominence (Pᵢ) indicates the overall significance of the factor within the system, while Relation (Rᵢ) distinguishes whether a factor acts more as a cause (positive value) or as an effect (negative value). A positive Rᵢ means the factor tends to influence others more than be influenced, i.e., it is more of a cause; a negative Rᵢ means it is largely an effect of others. This step essentially maps out the causal structure.

For example, “Development in Hazard Zones (A2)” was anticipated to emerge as a causal factor, since many other resilience issues depend heavily on land use decisions. Conversely, “Population Vulnerability (C2)” was expected to appear as an effect-type factor, as it is shaped by planning decisions, infrastructure quality, and social policies.

The DEMATEL results were then visualized as a Network Relationship Map (NRM), in which criteria are shown as nodes connected by directional arrows, illustrating the causal structure among dimensions. the Results section presents an example of the NRM derived in this study, highlighting the key causal and effect groups across the four dimensions. NRM plotting rule. To improve readability, we used a threshold τ equal to the mean of the non-zero entries in the total relation matrix.

• Step 3: ANP Weight Derivation

In the conventional Analytic Hierarchy Process or Analytic Network Process, researchers construct pairwise comparison matrices and derive weights through eigenvalue calculations. In the DANP framework, however, the information obtained from DEMATEL is utilized to form the unweighted supermatrix of the ANP. This matrix is then weighted and iteratively limited to obtain the final criteria weights, as described in detail by [37].

Specifically, the total influence matrix derived from DEMATEL was used to represent the normalized influence that each criterion exerts on the others. This normalized influence was treated as the “local priority” of a given criterion with respect to the factors it influences. In practical terms, if criterion X strongly influences criteria Y and Z according to the DEMATEL results, then in the ANP network, criterion X contributes proportionally to the weights of Y and Z.

The procedure followed several steps. First, the initial supermatrix W was formed, where each submatrix corresponded to the influence of a group of criteria on another group. Second, the weighted supermatrix was computed either by raising W to successive powers or by multiplying by a cluster weight matrix when dimensions were treated separately. In some DANP applications, equal cluster weights are adopted, while in others the total relation of clusters is used depending on the design of the study. Finally, the weighted supermatrix was raised to a sufficiently large power until it converged to a steady state, producing the limit supermatrix. This matrix yielded the global priority weights for each criterion, which is consistent with the process used in standard ANP to obtain final weights [38].

In this study, the normalized influence values obtained through DEMATEL were directly applied as the basis for ANP pairwise comparisons. This simplification, which has been demonstrated in previous methodological research by Chen and Chen [7], ensures that the ANP weights inherently reflect the interdependencies identified in the DEMATEL stage. The resulting weight vector was normalized so that the total summed to one across all criteria. In addition, dimension-level weights were calculated by aggregating the criterion weights within each of the four dimensions.

• Step 4: Resilience Evaluation and Prioritization

With the criteria weights in hand, we evaluated the resilience performance of each district in Taipei. A composite index approach was adopted: for each district, performance was scored on every criterion using the data collected, normalized to a common scale in which higher scores indicate stronger resilience. Quantitative indicators were standardized using min–max normalization, while qualitative indicators were converted into numeric scores (for example, Low = 1, Medium = 2, High = 3). The overall Resilience Index for each district was then calculated as a weighted sum of these scores, with weights derived from the DANP analysis. In parallel, a TOD Index was computed for each district based on TOD-related factors such as density, transit accessibility, and the extent of mixed land use, allowing us to compare TOD characteristics with resilience outcomes. This produced a ranking of districts from most to least resilient under the defined criteria. Normalization for composite indices. For each quantitative indicator, we applied min–max scaling.

Beyond ranking, the weighted criteria also highlighted areas requiring targeted improvement. For instance, if Emergency Planning (C1) carried a high weight but a district achieved a relatively low score on that criterion, it was flagged for urgent attention in disaster preparedness and response planning.

We acknowledge that an element of subjectivity is inherent in criteria scoring and expert judgment. To mitigate potential biases, a diverse expert panel was engaged, and wherever possible, local empirical data were used to ground the scores. In addition, a sensitivity analysis was carried out on the weights, adjusting and perturbing influence judgments to test the robustness of the results. This process was comparable to one-at-a-time sensitivity tests commonly recommended in multi-criteria analysis [39]. The overall district rankings and identification of key factors remained stable under reasonable variations, increasing confidence in the findings.

It should also be noted that the proposed method is dynamic in nature and can be updated as conditions evolve. For example, the opening of a new MRT line or the completion of a new flood mitigation project would alter the data inputs and potentially shift expert assessments; the DANP framework allows re-running the analysis to reflect such changes. Within this study, however, “dynamic” refers primarily to the network-based representation of interdependencies among resilience factors, rather than to a longitudinal time-series analysis. The findings, therefore, provide a snapshot of resilience performance and TOD-resilience interactions for Taipei as of 2025; Table 3 (indicator metadata and sources); Table 3 (expert questionnaire and aggregated influence matrix)). In addition, we performed simple robustness checks by perturbing expert influence scores and weight vectors within plausible bounds; the district rankings and the identity of top-priority criteria remained stable under these tests, supporting the reliability of our findings.

4. Results and Analysis

4.1. TOD and Resilience Indicator Overview by District

To ground the analysis, we first compare basic TOD-related and resilience-related indicators across Taipei’s districts.

Table 4. District-level TOD sub-centers and MRT provision (Taipei City).

District	MRT Stations (Count)	Station Density (per 10 km2)	TOD Sub-Centers (Examples)	Outlying Areas/Edges (Notes)
Daan	7	6.16	Daan Park–Technology Building; NTU/Gongguan edge	Inner-basin core; limited riverfront
Xinyi	5	4.46	Xinyi CBD (City Hall–Taipei 101/TWTC)	Elevated foothill edge; few flood-prone tracts
Zhongzheng	6	7.89	Taipei Main Station–Government Core	Historic core; partial lowlands behind levees
Songshan	6	6.45	Songshan Station–Arena; Nanjing–Sanmin corridor	Keelung River margin (protected)
Zhongshan	5	3.65	Shuanglian–Minquan; Zhongshan N. Rd corridor	Keelung River margin (protected); mixed old/new
Datong	3	5.28	Dadaocheng–Daqiaotou	Tamsui River margin (protected); dense historic blocks
Wanhua	2	2.26	Ximending–Longshan Temple	Confluence lowland; levees/pumps are critical
Shilin	2	0.32	Jiantan–Shilin Market; Tianmu (edge)	Large hillside/riverplain mix; dispersed stations
Beitou	6	1.06	Xinbeitou Hot Springs; Qiyan–Qilian	Hillside and riverplain edges; outlying neighborhoods
Neihu	7	2.22	Neihu Science Park–Huzhou; Donghu belt	Eastern basin lowland; techno-industrial clusters
Nangang	3	1.37	Nangang Station Hub; Software Park–Exhibition	Eastern basin lowland; gateway to New Taipei
Wenshan	7	2.22	Muzha–Jingmei	Foothill districts; creek valleys and green belts

Sources: Population and area statistics were obtained from the Taipei City Government, and MRT station counts were compiled from Taipei Rapid Transit Corporation datasets. Population figures are rounded, and the station count includes all MRT stations located within each district, with transfer stations counted once.

presents population, density, and transit station metrics, illustrating the degree of TOD urban form in each district. Table 4 summarizes selected resilience-related indicators, including flood exposure categorization and a measure of social vulnerability. These tables reveal stark spatial patterns in Taipei that underline the subsequent DANP evaluation.

From Table 4, it is evident that the five central districts—Daan, Xinyi, Zhongzheng, Songshan, and Datong—all exhibit extremely high densities (exceeding 20,000 persons/km²) and station densities greater than approximately five per 10 km² (with the exception of Zhongshan, which records 3.65). These districts represent the clearest examples of Taipei’s TOD-oriented planning. For example, Daan and Songshan are served by multiple MRT lines: Daan is intersected by the Tamsui–Xinyi and Wenshan–Neihu lines, while Songshan is served by the Green line and connects with the Brown line. This network enables a significant share of residents to live and work within walking distance of a transit station.

By contrast, the expansive northern districts of Shilin and Beitou, although populous, have densities roughly five times lower (about 4000–5000 persons/km²). This is partly due to the inclusion of mountainous terrain, and these districts are comparatively transit-sparse, with only 0.3–1 station per 10 km². Neihu and Nangang, located in eastern Taipei, represent intermediate cases: their densities range between 5000 and 9000 persons/km², and each has several MRT stations—particularly Neihu, which is well-served by the Brown line. However, neither achieves the same level of network coverage as the urban core. Quantitatively confirm the existence of a TOD gradient in Taipei. The central districts are highly transit-oriented and compact, whereas the outer districts are more dispersed and less connected.

Following this,

Table 5. Resilience-related indicators by district—flood exposure and social vulnerability.

District	Typhoon Flood Exposure	Notable Past Flood Incidents	Population ≥ 65 (%)	Social Vulnerability Level
Daan	Low (on higher ground, minimal riverfront)	Minor street flooding in extreme rain (no major history)	14%	Low (affluent, younger population)
Xinyi	Low (elevated terrain foothills)	Very limited flooding historically	12%	Low (modern buildings, high-income)
Zhongzheng	Medium (some low areas near Tamsui River)	2001: Ximenting area flooded (shared with Wanhua) [2]	16%	Medium-Low (mixed pop., many resources)
Songshan	Medium (Keelung River border; levee-protected)	2001: Songshan Airport flooded; local inundations	15%	Medium (moderate elderly presence)
Zhongshan	Medium (Keelung River border; levee-protected)	2001: Yuanshan area flooded; underpass floods	14%	Medium-Low (mixed use, decent resources)
Datong	Medium (Danshui River border; levee-protected)	Historically flood-prone (pre-levees); some lingering risk	17%	Medium (older neighborhoods, moderate elderly)
Wanhua	Medium (Danshui/Xindian Rivers meet; levee-protected)	2001: Severe flooding in Bangka area; multiple events historically	20%	High (aging population, many old buildings)
Shilin	High (Contains Shezidao floodplain, Keelung riverside)	2001: Shezidao severely flooded; recurring floods pre-2010	16%	Medium (some vulnerable pockets, but also new dev.)
Beitou	Low (Hilly except some riverside)	Minor localized flooding (e.g., near Guandu)	18%	Medium (some rural elderly communities)
Neihu	High (Eastern basin lowland; Keelung riverside)	2013 cloudburst: major flooding in Neihu Tech Park; 2001 floods	12%	Medium-Low (population relatively young, tech workers)
Nangang	High (Eastern basin lowland)	2001: flooding around Nangang station; frequent waterlogging	13%	Medium-Low (redeveloped area, fewer vulnerable residents)
Wenshan	Low-Med (Jingmei River minor floods; some hills)	Occasional creek overflows (e.g., Jingmei stream flooding)	15%	Medium (mix of older and newer communities)

Notes: Flood exposure is categorized considering topography and historical flood maps (Low = little to no flood-prone land; Medium = significant flood-prone zones but with mitigation; High = considerable flood-prone areas with history of inundation). Social vulnerability level is an overall qualitative assessment factoring % elderly (shown), plus socio-economic status and building conditions (High indicates high vulnerability). Sources: Taipei Hydraulic Engineering Office reports; Su [3] for city-wide flood scenario; National Statistics for age data (2020).

compiles resilience-related indicators. Two primary measures are presented for each district: indicative flood exposure level and a simplified social vulnerability score (based primarily on the proportion of elderly residents). While many resilience factors exist, these two offer a concise snapshot that reflects known district characteristics and can later be aligned with the criteria employed in the DANP model.

From Table 5, we see a pattern where the traditionally built-up core, which overlaps with TOD areas, generally has medium exposure but high levels of protection. For example, Datong, Zhongshan, Songshan, and Wanhua are all located on flat lowlands that were historically flood-prone, but these districts are now ringed by strong levees. Wanhua in particular has more than 40 percent of its land classified as flood-prone, yet major flooding has largely been prevented since the 1970s, with the exception of Typhoon Nari, which overwhelmed the pumping system.

By contrast, newer development districts such as Neihu and Nangang are characterized by high exposure because they are situated in the unprotected eastern basin. These areas have experienced severe flooding from localized heavy rainfall in recent years; for instance, a storm in June 2017 submerged large parts of Nangang’s streets. The hillside districts, including Xinyi, the partially elevated areas of Daan, and the uplands of Beitou, are naturally low in exposure due to their topography.

Social vulnerability shows a somewhat inverse relationship with TOD intensity. Wanhua, which has not been the focus of modern TOD and remains dominated by older, low-rise housing, displays high vulnerability. Approximately one in five residents is elderly, and many buildings are aging, which reduces resilience. In contrast, Neihu and Nangang, both of which have been more recently developed with TOD-like characteristics and are home to many technology industry workers, have younger populations and therefore lower vulnerability, even though they face greater flood exposure. This comparison suggests that resilience is inherently multi-factorial. For example, Wanhua’s residents may face greater difficulties in a disaster despite being protected by levees most of the time, while the younger population in Neihu may recover more quickly but must endure more frequent flooding events. These dataset the stage for DANP analysis. We will see how these various factors combine and which dominate in influence.

4.2. DEMATEL Results: Network of Influences Among Factors

Using the expert inputs and computations described, we obtained an influence matrix for the selected criteria. Figure 2 illustrates the influence network among the main dimensions, aggregated from the underlying criteria, and Table 5 presents the DEMATEL results in terms of prominence and relation for each dimension.

Figure 2 also outlines the conceptual pathways linking TOD features to resilience outcomes. Transit-oriented development, characterized by high density, mixed land use, and transit accessibility, influences resilience through several mechanisms. First, reduced sprawl ensures that development is kept away from hazardous peripheral zones, thereby protecting natural buffers. Second, improved mobility is supported by dense transit networks that provide redundancy and evacuation options. Third, infrastructure investment is enabled by the concentration of assets, which justifies robust, high-standard infrastructure such as flood defenses and structurally resilient buildings. Finally, local services and community cohesion are enhanced in TOD neighborhoods, where mixed-use environments maintain access to essential needs during crises and foster stronger social networks.

Collectively, these mechanisms strengthen the city’s resilience to typhoon impacts, including both flooding and service disruptions. However, the realization of these benefits depends critically on supportive institutional frameworks, including the implementation of strict building codes, effective enforcement, and inclusive planning strategies that ensure resilience dividends are distributed equitably across the urban population.

Together, these mechanisms enhance the city’s resilience to typhoon impacts, including both flooding and service disruptions. Nevertheless, the realization of these benefits depends critically on supportive institutional frameworks, including strict building codes, effective enforcement, and inclusive planning strategies that ensure resilience dividends are broadly shared across the urban population.

Table 6. DEMATEL outcomes for main dimensions of resilience.

Dimension	d (Influence)	r (Dependence)	Prominence (p = d + r)	Relation (R = d − r)
A. Land Use and Environment	0.85	0.61	1.47	+0.24
B. Infrastructure and Mobility	0.70	0.65	1.35	+0.05
C. Emergency and Social	0.64	0.76	1.40	−0.12
D. Governance and Planning	0.78	0.72	1.50	+0.06

Note: Prominence (p = d + r) indicates the overall connectivity and significance of each factor in the network, while Relation (R = d − r) shows whether a factor is a net cause (positive) or a net effect (negative). Values are normalized based on expert input and DEMATEL computation.

presents the DEMATEL outcomes for each resilience dimension, aggregated from the underlying criteria. Criterion-level details are analyzed separately and reported later in conjunction with the DANP-derived weights. In this framework, prominence (Pᵢ = dᵢ + rᵢ) indicates the overall connectivity and significance of a factor within the network, while relation (Rᵢ = dᵢ − rᵢ) identifies whether a factor functions as a net cause (positive values) or as a net effect (negative values). Reading guide. In Table 5, higher (p = d + r) means a dimension is more central in the network; the sign of (R = d − r) indicates driver (>0) versus response (<0). Figure 2 uses the same coding to visually separate driver clusters (A,D) from response clusters (C), with B acting as an intermediate hub.

From Table 5, Dimension D (Governance) and Dimension A (Land Use and Environment) register the highest prominence scores, approximately 1.50 and 1.47, respectively. Both also exhibit positive relation values, confirming that they act as leading influences in the resilience network. This indicates that, within the model, land use practices—such as avoiding hazardous areas and maintaining green buffers—together with the strength of governance actions, are the most critical factors shaping urban resilience to typhoons in Taipei. This finding reflects widely recognized principles of resilience, as emphasized in the 2014 IPCC report, which identified institutions and land use planning as key determinants of adaptive capacity. It also aligns with evidence from the Taiwanese context, where adaptive governance and land management have been shown to be particularly crucial for flood and typhoon resilience.

Dimension C (Emergency and Social) also demonstrates a high prominence score of 1.40, but with a negative relation value. This suggests that social outcomes are heavily influenced by developments in other domains rather than acting as primary drivers of resilience. Interestingly, Dimension B (Infrastructure and Mobility), while clearly important, shows a slightly lower prominence value of 1.35. This does not diminish the role of infrastructure but suggests that these factors occupy an intermediate position in the network: they influence resilience, yet they are also strongly dependent on governance and land use decisions. The relatively balanced influence and dependence values for Dimension B point to significant two-way feedback. For example, robust infrastructure reduces social vulnerabilities, but if governance priorities are weak or land use practices are risky, infrastructure alone is insufficient to guarantee resilience.

Drilling down to the criteria-level insights, several important patterns emerge. The criteria with the highest outgoing influence (dᵢ) included Development in Hazard Zones (A2), Drainage Infrastructure Capacity (B3), Emergency Planning (C1), and Land Use Management Policies (D1). Each of these registered normalized influence scores is greater than 0.8. For example, A2 strongly influences B3 because development in flood-prone zones places additional demands on drainage and levee systems. It also affects C2, since populations living in hazardous areas face inherently higher vulnerability, and even B1, as peripheral sprawl in unsafe zones typically suffers from poorer transit accessibility.

By contrast, the criteria with high incoming influence (rᵢ) included Population Vulnerability (C2) and Transit Accessibility (B1). Population Vulnerability, with r ≈ 0.85, is shaped by multiple domains: land use, infrastructure performance, and governance. Transit Accessibility also recorded a relatively high dependence score (r ≈ 0.7), reflecting that its effectiveness during disasters depends on governance decisions, such as whether transit is used for evacuation, and on infrastructure resilience, such as whether systems remain operational during extreme events.

When examining cause–effect groups, criteria such as A2 (hazardous land use), D1 (planning policy), and D2 (investment in resilience) were identified as cause-type factors, indicated by positive (dᵢ − rᵢ) values. Conversely, C2 (vulnerability) and B1 (transit access) were classified as effect-type factors, showing negative relation values. This distinction is intuitive: vulnerability is the result of long-term social and spatial conditions, while transit access becomes an outcome rather than a driver unless integrated with wider resilience systems.

The DEMATEL analysis also revealed reciprocal relationships. For example, B3 (drainage infrastructure) and D2 (governance investment) showed a feedback loop: investment strengthens infrastructure, while infrastructure failures in turn prompt additional investment. Similarly, C1 (emergency planning) and C3 (social capital) demonstrated a two-way relationship: a prepared community increases the effectiveness of plans, while well-designed plans enhance community confidence and participation.

In summary, the DEMATEL analysis affirms that land use management and infrastructure/governance are linchpins in the resilience of a TOD city like Taipei. The transit-oriented physical form must be accompanied by wise land allocation (keeping critical infrastructure and settlements out of harm’s way) and proactive governance for it to translate into resilience outcomes. Without those, even high-quality infrastructure or dense transit may not avert disaster (they might even amplify exposure). Conversely, given good governance and land use, the presence of dense transit and robust infrastructure becomes a powerful asset.

4.3. DANP-Derived Weights for Resilience Factors

Following the DEMATEL analysis, we incorporated the influence relations into the ANP weighting steps of DANP. The resulting weights indicate the relative priority of each factor in contributing to the overall typhoon resilience of TOD areas.

Table 7. Criteria weights derived from DANP analysis (resilience factor priorities).

Dimension and Criteria	Normalized Weight (%)	Rank (Overall)
A2: Development in Hazard Zones (exposure)	13.33	1
B3: Drainage and Flood Infrastructure Capacity	11.4	2
C1: Emergency Response and Planning	10.53	3
D1: Land Use Management Policies	9.74	4
C2: Population Vulnerability (demographics)	8.16	5
A3: Green/Permeable Space coverage	7.46	6
D2: Resilience Investment and Maintenance	7.02	7
B4: Critical Facilities (shelters, hospitals)	6.23	8
A4: Building Standards (structural robustness)	5.26	9
C3: Community Social Capital	4.99	10
A1: Population Density (urban compactness)	4.65	11
B2: Transport Network Redundancy	4.21	12
B1: Transit Accessibility (station density)	3.86	13
C4: Risk Awareness and Education	2.63	14
D3: Coordination and Governance Capacity	0.53	15

Note: Weights are normalized so that the total sums to 100%. Higher percentages indicate greater importance in determining overall resilience. Rankings are based on 15 total criteria. Values reflect the results of DANP analysis combining DEMATEL influence scores and ANP weight derivation.

lists the criteria and their weights (normalized to 100%) as derived from the DANP. For clarity, criteria are grouped by dimension and sorted by weight within each group.

The DANP weights in Table 6 provide a concrete prioritization of resilience factors for Taipei’s TOD context. The top-ranked factor is A2: Development in Hazard Zones, with a normalized weight of 13.33%. This confirms that exposure location is the single most critical issue: if large populations or assets are situated in flood-prone areas, they can overwhelm many other resilience measures. Even within TOD districts, if development occurs on reclaimed riverbeds or floodplains, the inherent risk remains high. The case of Shezidao in Shilin, a TOD fringe district located entirely on a flood plain, exemplifies a major vulnerability. Managing where development occurs, through zoning, land use regulation, or property buyouts, should therefore be treated as a top priority for resilience.

The second-highest factor is B3: Drainage and Flood Infrastructure Capacity, with a normalized weight of 11.40%. This highlights the central role of stormwater management systems: the ability to rapidly channel or retain rainfall often determines whether a neighborhood floods or remains protected. Taipei’s resilience depends on its extensive drainage network, pumping stations, and flood barriers. Localized flooding events, where short-duration rainfall has exceeded system capacity, illustrate the urgency of sustained investment in drainage upgrades and retention infrastructure, particularly as climate change increases rainfall extremes.

Close behind are C1: Emergency Response and Planning (10.53%) and D1: Land Use Management Policies (9.74%). The prominence of emergency planning underscores the importance of non-structural preparedness measures, such as disaster drills, evacuation protocols, and effective risk communication. Taipei’s annual city-wide disaster exercises exemplify best practice in this regard. The importance of land use management reflects its direct connection with hazardous development: combined with A2, these land use-related factors account for more than 23% of total weight, underlining the primacy of spatial planning. Unlike hazard magnitude, which cannot be controlled, these policy-driven factors are within human agency, providing a clear pathway for governance to enhance resilience.

The next significant factor is C2: Population Vulnerability, with a weight of 8.16%. This highlights how the demographic and social makeup of communities substantially conditions resilience. Districts with higher proportions of elderly or low-income populations are more vulnerable in disasters. Wanhua, for example, contains a large share of low-income elderly residents in older housing stock, making it particularly susceptible. Such districts require targeted interventions, including tailored evacuation assistance, accessible medical services, and broader social safety nets. The relative weight given to vulnerability indicates that reducing social fragility through poverty alleviation, improved health services, and community support mechanisms is as crucial as investing in physical infrastructure.

Other mid-ranked factors include A3: Green and Permeable Space Coverage (7.46%) and D2: Resilience Investment and Maintenance (7.02%), both of which underscore the importance of sustainable infrastructure and continuous investment. Meanwhile, B4: Critical Facilities Coverage (6.23%) and A4: Building Standards (5.26%) reflect the vital but more localized contributions of shelters, hospitals, and structural robustness. Lower-ranked factors, such as Transit Accessibility (B1) at 3.86% and Risk Awareness and Education (C4) at 2.63%, are still meaningful but demonstrate that resilience depends on their integration with higher-order governance, land use, and infrastructure systems. Finally, D3: Coordination and Governance Capacity, at just 0.53%, was rated lowest, reflecting either limited variability across districts or the perception that coordination is already embedded within other governance dimensions.

Beyond the leading factors, several mid-ranked criteria also carry important implications. D2: Resilience Investment and Maintenance received a normalized weight of 7.02%, ranking seventh. This criterion reflects the allocation of budgets and resources to resilience measures, such as upgrading infrastructure, maintaining levees, and funding community preparedness programs. Its relatively high ranking underscores that sustained investment contributes directly to overall resilience. For example, Taipei’s post-Nari investment in MRT flood gates and pumps prevented comparable disruptions in subsequent typhoons, demonstrating the long-term payoff of resilience spending.

A3: Green and Permeable Space Coverage, at 7.46% and ranked sixth, highlights the value experts place on natural infrastructure. Parks, wetlands, and permeable surfaces serve as buffers, absorbing water and reducing runoff, thereby complementing engineered flood-control systems. This finding supports Taipei’s adoption of the “sponge city” approach since 2015, which has promoted the use of permeable pavement, rain gardens, and urban green infrastructure. Within TOD contexts, compact development can potentially preserve more peripheral land for green use, but careful planning is needed to maintain this balance.

B4: Critical Facilities Coverage ranked eighth at 6.23%, emphasizing the necessity of hospitals, shelters, and other key facilities during emergencies. Taipei’s central districts benefit from higher concentrations of such facilities, whereas peripheral areas may be underserved, leaving them more vulnerable when transport routes are disrupted. This suggests that resilience planning should address the spatial distribution of emergency resources to ensure equitable coverage.

A4: Building Standards scored 5.26%, ranking ninth. Although this weight is moderate, it reflects the importance of structural robustness against typhoon winds and earthquakes. Most buildings in Taipei meet seismic-resistant codes, which indirectly enhance wind resilience, but older structures remain a significant vulnerability. For instance, parts of Wanhua still contain aging brick buildings that could fail under severe wind loads.

C3: Community Social Capital, at 4.99% (10th), shows that while social networks and community cohesion improve disaster outcomes, they are weighted slightly lower than physical factors in this analysis. This may reflect the perception that Taipei’s strong governmental response capacity reduces reliance on grassroots networks. Nevertheless, social capital remains crucial, and programs that strengthen neighborhood volunteer teams and community disaster groups should not be overlooked.

A1: Population Density received 4.65%, ranking 11th. Density, the hallmark of TOD, is neither inherently beneficial nor detrimental for resilience; its effect depends on management. High density can amplify risks if infrastructure is inadequate, but when combined with robust services, it allows for more efficient protection and recovery. This result suggests that density itself is secondary, with its resilience value mediated through other factors such as infrastructure and governance.

B2: Transport Network Redundancy, weighted at 4.21% (12th), reflects the benefit of having alternative routes and modes of transport. Its relatively low score may stem from Taipei’s compact geography, which allows for short travel distances even when certain routes are disrupted, or from expert judgment that redundancy overlaps with broader transit and governance functions.

B1: Transit Accessibility, with a weight of 3.86% and ranked 13th, offers an interesting insight. Although TOD emphasizes proximity to transit, this factor alone has a limited direct impact on typhoon resilience. The resilience benefits of transit accessibility emerge only when integrated into disaster response—for example, if MRT systems are explicitly used for evacuation or if they are designed to remain operational under extreme weather. Without such integration, transit may be forced to shut down, as occurred during Typhoon Megi in 2016, when bus and rail services were suspended at peak winds.

C4: Risk Awareness and Education ranked 14th with 2.63%, reflecting that while public education is useful, its marginal contribution is limited in Taipei, where residents already possess a high baseline awareness of typhoon risks. Awareness alone cannot overcome structural failures, which explains the lower weight relative to other factors.

Finally, D3: Coordination and Governance Capacity received the lowest weight of 0.53%. Although coordination is intuitively important, experts may have considered it implicit within broader governance categories such as planning (D1) and investment (D2). Alternatively, the low score may reflect limited variability between districts in this respect, since disaster coordination in Taipei is relatively centralized at the city level. Thus, this does not imply that coordination is irrelevant, but rather that its marginal impact relative to other factors was judged to be minimal in the context of Taipei’s existing governance structures.

Overall, the distribution of weights reinforces several key points regarding the factors that drive resilience in Taipei’s TOD context. First, keeping development out of flood-prone areas and ensuring robust flood-control infrastructure are the most decisive priorities, together accounting for more than one quarter of the total weight. This highlights that resilience fundamentally depends on sound land use planning and the capacity of physical systems to manage water during extreme events. Priority actions implied by weights. (1) Strictly control and gradually reduce development footprints in A2 high-exposure tracts; (2) accelerate B3 drainage upgrades and decentralized retention (neighborhood tanks, permeable corridors); (3) institutionalize C1 ward-level drills with station-area refuges; (4) enforce D1 climate-risk screening for all station-area upzonings; (5) target C2 hotspots (e.g., elderly clusters in Wanhua/Datong) with tailored evacuation and medical support.

Second, so-called “software” factors such as planning, governance, and social dimensions collectively carry significant influence. When the weights of the governance-related (Dimension D) and social-related (Dimension C) criteria are combined, they amount to approximately 35–40% of the total. This demonstrates that resilience is not merely an engineering challenge but also an institutional and societal one, requiring effective governance structures and community-level capacity.

Third, the results reveal that traditional TOD elements, such as high density and transit accessibility, contribute to resilience primarily in indirect ways. On their own, these factors rank lower in importance. This underscores the principle that TOD must be implemented with explicit consideration of resilience: compact, high-density growth without climate adaptation measures can be maladaptive, potentially exacerbating problems such as urban heat islands or evacuation bottlenecks.

Finally, the findings align with international observations that cities require a multi-pronged resilience strategy. Core elements include prudent land use planning to steer growth away from hazard-prone zones, robust infrastructure to absorb shocks, effective emergency management to coordinate rapid response, and social systems that strengthen community preparedness and recovery (recent international guidance on disaster risk reduction). Taken together, the analysis affirms that resilience in a TOD framework emerges not from a single factor, but from the interplay of spatial, infrastructural, institutional, and social dimensions.

4.4. District Resilience Evaluation and Comparison

Using the DANP weights and the district-level data for each criterion, we computed a Resilience Index for each district of Taipei. The index is a weighted sum of normalized criterion scores (with higher values indicating more resilience). Similarly, we created a simple TOD Index (based on population density, station density, and mixed-use, with equal weighting among those). The purpose is to see how districts rank in resilience and whether those ranks correlate with their TOD characteristics.

Table 8. Resilience Index vs. TOD Index by district.

District	Resilience Index (0–100)	Resilience Rank	TOD Index (0–100)	TOD Rank	Pop. Density (per km2)	Flood Exposure (Low/Med/High)
Daan	85	1 (most resilient)	90	1	27,530	Low
Xinyi	82	2	84	2	20,070	Low
Zhongzheng	78	3	86	3	20,890	Medium
Songshan	75	4	80	5	22,730	Medium
Zhongshan	73	5	76	6	16,660	Medium
Neihu	62	8	50	10	9090	High
Wenshan	70	6	54	8	8540	Low-Med
Nangang	59	9	45	12	5570	High
Shilin	65	7	48	11	4610	High
Datong	58	10	70	7	23,330	Medium
Beitou	57	11	50	9	4470	Low
Wanhua	55	12 (least resilient)	60	4	21,520	Medium

presents the results: the computed indices and rank orders for resilience and TOD, and for illustrative purposes, a few raw indicators (density and flood exposure) are shown alongside. (Resilience Index computed from DANP-weighted criteria scores; TOD Index from density, transit, land use mix. Population density and flood exposure are shown for context. Higher index = better. Ranks are among 12 districts.)

The Resilience Index reveals a clear stratification among Taipei’s districts. Daan District ranks highest with a score of 85 out of 100. Daan represents a showcase of transit-oriented development: it is extremely dense and transit-rich, located on relatively higher ground south of the city center, and contains numerous critical resources such as major hospitals and universities. It benefits from modern drainage infrastructure in the Xinyi Plan area, relatively low flood exposure, and moderate social vulnerability due to its diverse population of students and middle-class households. Its profile illustrates that dense inner-city districts can achieve high resilience when hazard exposure is low and infrastructure and government services are strong.

Xinyi District, with a score of 82, is ranked second. As Taipei’s modern central business district, it contains robust high-rise structures, state-of-the-art infrastructure, and elevated terrain that shields it from flooding. Its economic vitality and high levels of private investment further strengthen resilience, though weaker community cohesion—resulting from a more transient workforce—slightly tempers its performance. Zhongzheng District, which includes the government quarter and Taipei Main Station, ranks third with a score of 78. Despite medium exposure, it benefits from a strong institutional presence, extensive protective infrastructure, and numerous emergency facilities, which collectively elevate its resilience.

Songshan and Zhongshan districts, scoring 75 and 73, respectively, occupy the fourth and fifth positions. Both are dense central areas along the Keelung River that rely on levees and pumping systems for flood protection. Although effective since Typhoon Nari, their resilience remains conditional on these defenses. Songshan additionally faces challenges related to its airport’s flood vulnerability, while Zhongshan contains older neighborhoods that increase risk. These districts are therefore relatively resilient but not invulnerable.

Wenshan (70) and Shilin (65) occupy mid-level positions. Wenshan combines hilly and urban landscapes: while some areas experience periodic flooding, others face landslide risks not fully captured in this index. Its moderate density and cohesive neighborhood networks contribute positively to resilience. Shilin presents a mixed picture, with high exposure in areas like Shezidao but also safer high-ground zones such as Tianmu. The district’s mix of affluent enclaves and vulnerable pockets results in a middling resilience profile.

Neihu (62) and Nangang (59) rank eighth and ninth. Both districts lie in Taipei’s eastern basin, a flood-prone area where past storms such as those in 2013 and 2022 caused significant disruptions. Although both districts are modern, with wide roads and a young, tech-oriented population, their physical exposure and historical underinvestment in critical facilities weaken their resilience scores. Recent infrastructure upgrades may improve their performance in future assessments.

At the lower end, Datong (58) and Beitou (57) rank tenth and eleventh. Datong, one of Taipei’s oldest districts, combines high density with aging infrastructure and medium exposure along the Tamsui River. While strong community ties enhance its social resilience, its older housing stock remains vulnerable. Beitou, largely hilly and lower density, is threatened by landslides and floods in its valley-bottom settlements near Guandu. Its relatively high proportion of elderly residents further reduces resilience.

Finally, Wanhua District, scoring 55, is the least resilient in this analysis. As Taipei’s oldest district, Wanhua faces multiple challenges: low-lying topography at the confluence of rivers, an aging and low-income population with the city’s highest share of elderly residents, and a concentration of structurally weak older buildings. Although its long-established communities provide social capital, this is insufficient to offset its physical and demographic vulnerabilities. Wanhua’s experience during Typhoon Nari, when chaotic evacuations and widespread damage occurred, exemplifies its fragility. Targeted interventions—such as retrofitting buildings, improving pumping systems, and enhancing community preparedness—are urgently required to strengthen its resilience.

A comparison of the Resilience Index with the TOD Index reveals a general but imperfect correlation. Districts with the highest TOD scores—such as Daan, Xinyi, and Zhongzheng—are also among the most resilient, suggesting that well-implemented TOD can coincide with higher resilience outcomes. However, notable exceptions underscore that TOD alone does not guarantee resilience. Wanhua, for example, ranks fourth on the TOD Index due to its density and centrality, yet it records the lowest resilience score. This case illustrates that density without modern infrastructure and with high social vulnerability translates into fragility rather than strength. Conversely, Beitou, which is largely sprawling and car-oriented, ranks low on both TOD (ninth) and resilience (eleventh), indicating that sprawl similarly undermines resilience.

Figure 3 illustrates the causal relationships among the four main resilience dimensions in Taipei’s TOD context. Land Use and Environment (A) and Governance and Planning (D) emerge as strong causal dimensions, exerting significant influence on Infrastructure and Mobility (B) and Emergency and Social factors (C). Emergency and Social factors, in contrast, are primarily effect-type dimensions, largely shaped by land use and governance decisions. The map highlights that prudent land use management and effective governance policies serve as primary drivers of resilience, while infrastructure functions as an intermediate factor that both influences and is influenced by other dimensions.

The mid-range districts show further nuance. Datong, for instance, ranks seventh in TOD but only tenth in resilience, reflecting the drag imposed by aging infrastructure. Neihu and Nangang also score poorly in resilience despite having some TOD features, as their historical development patterns remain more auto-oriented and they face significant flood exposure.

If resilience scores were plotted against TOD indices, the relationship would likely appear as a moderate positive correlation, albeit with several outliers. Four general categories emerge. The first consists of densely developed and well-planned districts, such as Daan and Xinyi, which combine high TOD and high resilience. The second includes densely developed but socio-physically vulnerable districts, typified by Wanhua, where high TOD coincides with low resilience. The third group represents less-developed, lower-risk districts, where lower TOD coincides with moderate resilience, as in some hillside areas. Finally, less-developed, higher-risk districts, such as peripheral flood-prone neighborhoods, display low TOD and low resilience.

These patterns highlight that TOD by itself is not a guarantee of resilience. It must be coupled with effective risk-reducing planning, investment in infrastructure, and social preparedness. The best-case scenario is what may be termed “Resilient TOD”, exemplified by Daan and Xinyi, where density and transit accessibility are combined with safe siting, robust infrastructure, and cohesive communities. The worst-case scenario is “Maladapted TOD”, where density is concentrated in hazardous zones without sufficient mitigation. Some observers may argue that parts of New Taipei’s Tamsui District illustrate this condition, where new high-density projects have been developed on vulnerable land. District quick wins. Wanhua: fast-track pump capacity upgrades near Bangka, retrofit pre-1990 structures, establish temple-based micro-shelters; Datong: alley drainage retrofits and façade/roof anchoring for historic blocks; Neihu: expand retention in the Tech Park and elevate critical substations; Nangang: green-blue corridors from foothills to Keelung River to attenuate cloudburst runoff.

The factor scores further illustrate these distinctions. Daan scored highly across nearly all categories, with only moderate performance in social capital. Xinyi likewise achieved strong resilience overall, compensating for weaker community cohesion with superior infrastructure. Wanhua, in contrast, scored poorly on at least half of the criteria, particularly population vulnerability and infrastructure capacity. Neihu and Nangang performed well on modern building standards and low demographic vulnerability but were held back by high exposure and lagging drainage capacity.

Geographically, the five most resilient districts cluster in the central and eastern parts of the Taipei Basin, either on slightly higher ground or protected by strong flood defenses. The bottom five are concentrated in the periphery and older historic core, areas either inherently risk-prone or constrained by outdated infrastructure. This distribution reflects an intuitive logic: newer planned areas, often TOD-focused, benefit from modern standards and avoidance of the most hazardous sites, whereas older and peripheral districts remain more exposed.

These findings carry clear policy implications. Taipei should continue pursuing TOD, but in tandem with integrating climate risk data into TOD planning. For example, before upzoning land around a future MRT station, detailed flood maps should be consulted. If the site is located in a low-lying or flood-prone zone, resilience measures—such as elevated stations, flood-control systems, and permeable landscapes—must be prioritized. Singapore offers a useful precedent: by requiring elevated construction and strict flood-control integration, it reduced flood-prone land to near zero. Districts such as Wanhua and Neihu should be targeted for resilience upgrades, including retrofitting older housing stock and enhancing drainage systems. The relatively low weight assigned to transit accessibility in the DANP results reinforces this point: to fully leverage transit for resilience, stations and networks must be explicitly integrated into disaster response. For example, MRT stations could be designed or retrofitted as refuges during floods, serving as safe hubs if equipped with flood gates and backup systems.

Finally, several limitations of the analysis should be acknowledged. The Resilience Index is comparative within Taipei, meaning that even the top-ranked districts are not immune to extreme events. A super-typhoon could overwhelm Daan or Xinyi through cascading failures such as prolonged power outages or unprecedented rainfall events. Moreover, factors such as earthquake resilience were not incorporated, though they are highly relevant to Taipei. While many measures that strengthen flood resilience also improve seismic safety, high density can pose unique challenges for earthquake evacuation. The method is also inherently static. Although DANP captures interdependencies, it does not model temporal evolution. In practice, Taipei is continually upgrading pumps, retrofitting infrastructure, and improving planning capacity. As such, districts like Nangang, which rank relatively low today, may improve substantially in future assessments.

In conclusion, TOD areas in Taipei generally exhibit higher resilience, provided they are not located in hazard-prone zones and are supported by robust infrastructure and effective planning. The analysis underscores that land use management and social preparedness are the most decisive differentiators of resilience, rather than transit access alone. The next section expands on these insights in a broader comparative context and outlines concrete policy recommendations.

5. Conclusions

This study set out to examine how Transit-Oriented Development (TOD) urban spatial forms relate to urban resilience against typhoons in the Taipei metropolitan area, through an analysis employing the Dynamic Analytic Network Process (DANP). By integrating extensive empirical data and expert knowledge, we evaluated multiple interlinked factors, spanning land use, infrastructure, transport accessibility, and socio-demographics, to determine their relative importance in enhancing or impeding resilience in TOD contexts. The findings yield several overarching conclusions and contributions to both theory and practice:

• Urban Morphology Matters for Resilience: The physical layout and form of a city. Its density, land use patterns, transportation networks, and allocation of open space significantly influence its vulnerability and response capacity to typhoon impacts. Our results reinforce that resilience is not just about emergency management; it is fundamentally built (or undermined) by decades of urban development decisions [23,40]. Cities that have prioritized compact, connected growth show a propensity for reduced hazard exposure and more robust infrastructure performance during disasters, if growth is directed toward safer locations and supported by adequate infrastructure. In Taipei’s case, districts with a well-planned TOD morphology but low hazard exposure demonstrated the highest resilience levels. Conversely, areas that are dense but were historically developed in risk-prone zones remain vulnerable. This underscores that urban form can greatly amplify or mitigate risk: transit-oriented compactness can facilitate efficient protection and recovery if implemented with risk-awareness, whereas haphazard density can concentrate risk.
• Transit-Oriented Development as a Resilience Strategy: TOD, as an urban model, generally enhances resilience to typhoons when coupled with adaptive planning measures. Compact, transit-focused cities like Taipei (and similarly Hong Kong, Singapore) demonstrate that high densities can be compatible with low disaster losses, given supportive measures such as strong building codes, robust flood management infrastructure, and preserved green buffers [25]. Our synthesis identified several mechanisms by which TOD contributes to resilience.
• Reduced exposure of development to hazards: By curbing sprawl, TOD helps leave critical ecosystems (like wetlands and drainage plains) intact to buffer storms and avoid placing extensive housing in the most flood-prone lands [13]. In Taipei, the core TOD districts are largely within a ring of levees and high ground, whereas much recent peripheral sprawl (outside TOD influence) has occurred on vulnerable hillsides or lowlands. This contrast highlights TOD’s potential in steering growth to relatively safer, service-rich areas.
• Efficient evacuation and response: Dense transit networks provide alternative mobility and can be leveraged for evacuations or rapid post-storm repairs. A well-connected urban grid means emergency services can reach more people quickly, and public transport (if maintained and operational) can move citizens or supplies when roads are blocked. In our study, while transit access alone was not the top factor, its value emerged in conjunction with emergency planning, serving as a reminder that cities should integrate transit agencies into disaster response planning.
• This research received no external funding. The authors thank their colleagues in the Taipei City Government’s Hydraulic Engineering Office and the Department of Transportation for access to public statistics and planning documents. Any opinions, findings, or conclusions expressed here are those of the authors and do not necessarily reflect the views of the supporting agencies.
• Stronger infrastructure and buildings: Concentrating development creates economies of scale for investing in resilient infrastructure, which protects large populations and economic assets. Taipei’s experience upgrading its entire MRT system’s flood defenses after one event illustrates how protecting a few central infrastructure systems can safeguard millions when the population is centralized. Our DANP results indeed gave high weight to infrastructure robustness, validating this co-benefit of TOD.
• Community and economic resilience: Mixed-use TOD neighborhoods tend to maintain access to essential services during crises, and they foster social cohesion due to their walkable, interactive nature. This supports quicker recovery and mutual aid among residents [41]. In Taipei, many TOD districts have strong neighborhood networks (e.g., temple communities in Datong or borough wardens in Daan), which proved vital in past typhoons for volunteerism and information dissemination. Our analysis gave social capital a notable, if not top, role—implying that the design of TOD areas to include community spaces and social infrastructure can pay dividends in resilience.
• Need for Complementary Measures and Vigilance: TOD is not a resilience silver bullet on its own. High density can introduce or exacerbate certain issues, such as urban heat islands or stormwater runoff, which must be mitigated through conscious urban design and policy. For example, our study highlighted the importance of green space and permeable surfaces in dense areas to manage floodwaters; without them, TOD districts could suffer inner-city flooding despite overall good planning. Likewise, transit systems must be made flood-proof and have backup options to truly serve as resilience assets; otherwise, a power outage or inundation could immobilize a TOD city. The positive resilience outcomes of TOD hinge on proactive planning and regulation—without good governance, density could become maladaptive (for instance, poorly built dense settlements could amplify disaster tolls, as seen in some informal high-density quarters in Manila). Thus, TOD must be accompanied by strong planning frameworks and continuous risk assessments to yield resilience benefits. Taipei’s improvements after past typhoons exemplify the needed vigilance: policies were adjusted and must continue to adapt as climate risks evolve.
• Integration of Resilience in Urban Planning: One of the clearest conclusions is the need for integration: transit planning, land use planning, and disaster risk reduction should be pursued jointly. Our research shows that cities that did this (implicitly or explicitly), for example, Singapore, intertwining transit expansion with long-term flood control planning, are reaping resilience rewards, while cities where these domains were disjointed, e.g., Metro Manila, expanding urban extent without adequate transport or hazard consideration, face severe resilience deficits. In practice, this means breaking down silos: when a new transit line is proposed, planners should simultaneously evaluate where station-area development should be encouraged or limited based on flood maps and emergency access needs. Policies like requiring climate risk evaluations for all new TOD projects, or providing incentives for development in safe, transit-accessible locations while restricting it in unsafe ones, are practical steps towards this integration. Additionally, aligning investments is key: for instance, using the budget for a new rail line to also fund complementary drainage upgrades and parks in the corridor. The concept of Resilient TOD could be formalized in planning guidelines, ensuring that every TOD initiative contributes to, rather than inadvertently undermines, the city’s disaster resilience.
• Methodological Contributions: On a methodological note, this study demonstrates the value of using Dynamic Analytic Network Process (DANP) for complex urban resilience evaluations. By capturing interdependencies among factors, DANP provided a nuanced prioritization that reflects the real-world complexity of urban systems. Traditional single-criterion analyses might have overlooked these links—for instance, simple correlation might not show transit’s effect mediated through governance, which our network approach did reveal. The DANP framework is transferable to other cities and other hazards, and our approach can serve as a template for city planners to conduct multi-criteria resilience assessments that inform strategic investments. We preserved all citations of data and methodology sources throughout to ensure transparency and enable interested readers to trace back our evidence. In closing, Taipei’s case illustrates both the potential and the prerequisites of resilience in a transit-oriented metropolis. The city has made great strides by concentrating development in well-serviced areas and continuously improving its infrastructure and plans. Yet, challenges remain, especially for those pockets where past development did not heed hazard risks. The findings urge a redoubling of efforts to retrofit and support such vulnerable areas, even as the city continues to leverage TOD for sustainable growth.

For urban policymakers and practitioners in Taipei and beyond, the message is clear: when we plan our cities around transit, we are not only planning for efficient mobility and sustainability, we are also, whether we realize it or not, planning for resilience. By making that connection explicit, and by actively incorporating disaster risk considerations into TOD policies, urban decision-makers can unlock new value from TOD projects, justifying them not only in terms of transport or CO₂ reduction, but also in terms of protecting communities and economies from climate disasters. In an era of intensifying typhoons and other climate extremes, this integrated approach to urban morphology could spell the difference between a city that bounces back quickly and one that struggles to recover. Taipei’s ongoing journey offers lessons that investing in resilient urban form is not just prudent but imperative for the cities of the future. Implementation checklist for a ‘Resilient TOD’ program: risk-aware site selection for station areas; elevated/flood-protected access for key MRT entrances; mandatory on-site detention for large TOD projects; district-level emergency role assignments to selected stations; and annual public reporting of resilience KPIs by district.